Sound Insulation in

Buildings
Sound Insulation in
Buildings

Jens Holger Rindel

Cover photo by Iiris Turunen-Rindel. The photo on the cover shows a concrete building from the
1960s being demolished.

CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742

© 2018 by Taylor & Francis Group, LLC

CRC Press is an imprint of Taylor & Francis Group, an Informa business

No claim to original U.S. Government works

Printed on acid-free paper

International Standard Book Number-13: 978-1-4987-0041-2 (Hardback)

This book contains information obtained from authentic and highly regarded sources.
Reasonable efforts have been made to publish reliable data and information, but the author and
publisher cannot assume responsibility for the validity of all materials or the consequences of
their use. The authors and publishers have attempted to trace the copyright holders of all mate-
rial reproduced in this publication and apologize to copyright holders if permission to publish
in this form has not been obtained. If any copyright material has not been acknowledged please
write and let us know so we may rectify in any future reprint.

Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, repro-
duced, transmitted, or utilized in any form by any electronic, mechanical, or other means, now
known or hereafter invented, including photocopying, microfilming, and recording, or in any
information storage or retrieval system, without written permission from the publishers.

For permission to photocopy or use material electronically from this work, please access www.
copyright.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc.
(CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organi-
zation that provides licenses and registration for a variety of users. For organizations that have
been granted a photocopy license by the CCC, a separate system of payment has been arranged.

Trademark Notice: Product or corporate names may be trademarks or registered trademarks,

and are used only for identification and explanation without intent to infringe.

Library of Congress Cataloging-in-Publication Data

Names: Rindel, J. H. (Jens Holger), author.

Title: Sound insulation in buildings / by Jens Holger Rindel.
Description: Boca Raton : CRC Press, [2018] | Includes bibliographical
references and index.
Identifiers: LCCN 2017024121| ISBN 9781498700412 (hardback : alk. paper) |
ISBN 9781498700429 (ebook : alk. paper)
Subjects: LCSH: Soundproofing. | Architectural acoustics.
Classification: LCC TH1725 .R56 2018 | DDC 693.8/34--dc23
LC record available at https://lccn.loc.gov/2017024121

Visit the Taylor & Francis Web site at

http://www.taylorandfrancis.com

and the CRC Press Web site at

http://www.crcpress.com
Contents

Preface xv
About the Author xix
Introduction xxi

1 Basic concepts in acoustics 1

1.1 Introduction  1
1.2 The decibel scale  2
1.3 Frequency and related concepts  3
1.3.1 Harmonic vibrations  3
1.3.2 Speed of sound  4
1.3.3 Frequency bands  4
1.4 Loudness level and A-weighted sound pressure level  6
1.5 Propagation of sound in plates  8
1.5.1 Longitudinal waves  9
1.5.2 Bending and shear waves in plates   9
1.5.3 Critical frequency  12
1.6 Averaging  14
1.6.1 Arithmetic average  14
1.6.2 Geometric average  14
1.6.3 Harmonic average  14
1.6.4 Energetic average  14
1.7 Useful mathematical functions and formulae  15
1.7.1 Trigonometrical formulae  15
1.7.2 Euler’s formulae  15
1.7.3 The cardinal sine function  15
1.7.4 Logarithmic functions  15
1.7.5 Basic equations for complex numbers  16
References 17
vi Contents

2 Mechanical vibrations 19
2.1 A simple mechanical system  19
2.2 Mechanical impedance and mobility  21
2.3 Free vibrations  23
2.4 Forced vibrations  26
2.5 Resonance 30
2.5.1 Resonance frequency  30
2.5.2 Bandwidth of a resonant system  31
2.6 Vibration insulation  33
2.6.1 The power input  33
2.6.2 Force transmitted to the foundation  34
2.6.3 Vibration isolators with hysteretic damping  37
2.6.4 The Q-factor 37
2.6.5 Insertion loss  38
2.6.6 Insulation against structure-borne vibrations  40
2.6.7 Design of vibration isolators  42
2.7 Human response to vibrations  44
2.7.1 Examples of evaluation criteria  46
References 47

3 The sound field in front of a wall 49

3.1 Normal incidence of a plane wave  49
3.2 Oblique incidence of a plane wave  52
3.3 Random incidence in a diffuse sound field  54
3.4 The sound field at edges and corners in a room  59
3.5 The influence of bandwidth on reflection of noise  61
References 67

4 Introduction to room acoustics 69

4.1 Sound waves in rooms  69
4.1.1 Standing waves in a rectangular room  69
4.1.2 Modal reverberation times  72
4.1.3 Transfer function in a room  74
4.1.4 Modal density  74
4.1.5 The Schroeder limiting frequency  79
4.2 Statistical room acoustics  79
4.2.1 The diffuse sound field  79
4.2.2 Incident sound power on a surface  80
4.2.3 Equivalent absorption area  82
4.2.4 Energy balance in a room  82
 Contents vii

4.2.5 Reverberation time: Sabine’s formula  83

4.2.6 Stationary sound field in a room:
Reverberation distance  85
4.3 Geometrical room acoustics  89
4.3.1 Sound rays and a general reverberation formula  89
4.3.2 Sound absorption in air  91
4.3.3 Reflection density in a room  92
4.4 Calculation of reverberation time  93
4.4.1 Using Sabine’s formula  93
4.4.2 Reverberation time in non-diffuse rooms  94
4.5 Measurement of reverberation time  96
References 98

5 Introduction to sound insulation 101

5.1 Airborne sound insulation  101
5.1.1 The sound transmission loss  101
5.1.2 Sound insulation of a partition between
two rooms  102
5.1.3 Measurement of sound insulation  103
5.1.4 Multi-element partitions and apertures  104
5.2 Single leaf constructions  106
5.2.1 Sound transmission through a solid material  106
5.2.2 The mass law  109
5.2.3 Sound insulation at random incidence  111
5.2.4 The critical frequency  112
5.2.5 A simple model of sound insulation
of single constructions  113
5.2.6 Examples of single constructions  116
5.3 Double leaf constructions  121
5.3.1 Impedance model for sound transmission  121
5.3.2 The mass-air-mass resonance frequency  123
5.3.3 Sound insulation of double constructions  124
5.3.4 Examples of double constructions  125
5.4 Slits and leaks  128
5.5 Flanking transmission  131
5.6 Enclosures 135
5.7 Impact sound insulation  136
5.7.1 Impact sound pressure level  136
5.7.2 Historical note  138
5.7.3 Sum of airborne and impact sound insulation  138
5.8 Single-number rating of sound insulation  140
viii Contents

5.8.1 Weighted sound reduction index and

the sound transmission class  140
5.8.2 Weighted impact sound pressure level and
the impact insulation class  142
5.8.3 Alternative reference curve for evaluating
the impact sound insulation  144
5.8.4 Previous methods for rating sound insulation  145
5.8.5 Historical note  145
5.8.6 Spectrum adaptation terms  146
5.9 Requirements for sound insulation  150
References 151

6 Sound radiation from plates 155

6.1 Normal modes in a rectangular plate  155
6.2 Sound radiation from a propagating transverse wave  160
6.3 The radiation efficiency  163
6.4 Sound radiation calculated by Rayleigh’s method  164
6.5 Sound radiation from forced vibrations
in a rectangular plate  166
6.5.1 Approximate results  171
6.6 Sound radiation from supersonic bending waves  174
6.7 Sound radiation from resonant
vibrations in a rectangular plate  174
6.7.1 Approximate results  179
6.7.2 Directivity of resonant radiation  181
6.7.3 Combined forced and resonant radiation  183
6.8 The refraction effect  184
References 187

7 Statistical energy analysis, SEA 189

7.1 Energy balance in a resonant system  189
7.2 Coupling between resonant systems  191
7.3 Resonant transmission through a single wall  194
7.4 Forced and resonant transmission through a single wall  197
7.5 Acoustic excitation of resonant vibrations in a plate  198
7.5.1 Excitation from one side  198
7.5.2 Freely hung plate in a room  199
7.6 Resonant transmission through a slab with extended area  199
References 201
 Contents ix

8 Airborne sound transmission through single constructions 203

8.1 Sound insulation of homogeneous,
isotropic plates at random incidence  203
8.1.1 Total transmitted sound power  203
8.1.2 The mass law  204
8.1.3 Wall impedance  205
8.1.4 The loss factor  210
8.1.5 Forced transmission  211
8.1.6 Resonant transmission  214
8.1.7 Total transmission: a calculation model  215
8.2 Sound insulation of acoustically thick plates  218
8.3 Sound insulation at very low frequencies  221
8.4 Sound insulation of subdivided structures  224
8.5 Sound insulation at oblique incidence  226
8.5.1 Definition of external transmission loss  226
8.5.2 Transmission from outside to inside  227
8.5.3 Transmission from inside to outside  228
8.5.4 External transmission loss for
a single construction  228
8.6 Orthotropic plates  229
8.7 Sandwich plates  234
8.7.1 Speed of transverse waves  234
8.7.2 Resonant transmission  238
8.7.3 Forced transmission  240
8.7.4 Dilatational resonance  240
References 244

9 Airborne sound transmission through

double constructions 247
9.1 Transmission via the cavity  247
9.1.1 Normal incidence of sound  247
9.1.2 The resonance frequency  249
9.1.3 Standing waves in the cavity  250
9.1.4 Calculation for random incidence  250
9.1.5 Weakly attenuated cavity  252
9.2 Transmission via sound bridges  254
9.2.1 Point connections  255
9.2.2 Line connections  257
x Contents

9.2.3 Boundary connections  259

9.2.4 Resonant sound radiation  260
9.2.5 Elastic connections  262
9.2.6 Coupling factor  263
9.2.7 Summary on sound bridges  264
9.3 Sound insulation of drywalls  265
9.3.1 Influence of stud distance  265
9.3.2 Examples 268
9.4 Sound insulation of windows  269
9.4.1 Gas filling in windows  269
9.4.2 Examples 270
References 272

10 Impact sound insulation 275

10.1 The impact sound pressure level  275
10.2 Massive floors  276
10.2.1 Point excitation of a plate  276
10.2.2 Impact sound pressure level for massive floors  278
10.3 Resilient floor coverings  280
10.4 The relationship between R and L n 282
10.5 Floating floors  283
10.5.1 Attenuation of impact sound pressure level  285
10.5.2 Surface-supported floors  287
10.5.3 Dynamic stiffness of loaded mineral wool  288
10.5.4 Resonantly reacting floors  290
10.5.5 Locally reacting floors  292
10.5.6 Point-supported floors  295
10.5.7 The effect of sound bridges  296
10.5.8 Summary on floating floors  296
10.6 Suspended ceilings  297
10.7 Massive wood floor constructions  297
10.8 Sound insulation in practice  302
10.8.1 A live walker and the ISO tapping machine  302
10.8.2 Lightweight and heavyweight floor constructions  304
10.8.3 Typical complaints about impact sound  305
References 310

11 Flanking transmission 313

11.1 Flanking transmission of airborne sound  313
11.1.1 Transmission paths  313
 Contents xi

11.1.2 Junction attenuation  314

11.1.3 Flanking transmission loss  315
11.1.4 The total apparent transmission loss  317
11.1.5 The reciprocity method  317
11.2 Flanking transmission of impact sound  318
11.2.1 The impact sound pressure level
for one transmission path  318
11.2.2 Total impact sound pressure level  320
11.2.3 Combination of airborne and impact transmission  321
11.3 Junction attenuation  321
11.3.1 Empirical model  321
11.3.2 The structural transmission coefficient  321
11.3.3 Relation between junction attenuation
and structural transmission loss  324
11.3.4 Vibration reduction index for rigid connections  326
11.3.5 Estimated attenuation for elastic connections  326
11.4 Example 329
References 332

12 Measurement methods 335

12.1 Sound pressure level and low-frequency noise  335
12.1.1 Microphone 335
12.1.2 Frequency weighting  336
12.1.3 Averaging time  337
12.1.4 Correction for background noise  337
12.1.5 Measurement positions and spatial averaging  338
12.1.6 Correction for reverberation time  339
12.2 Airborne sound insulation between rooms  340
12.2.1 Generation of sound  340
12.2.2 Measurement positions and spatial averaging  341
12.2.3 Low-frequency procedure  345
12.3 Windows and facades  346
12.3.1 The global method  346
12.3.2 The element method  347
12.4 Impact sound insulation  348
12.5 Reverberation time  350
12.5.1 Measurement positions  350
12.5.2 The interrupted noise method  350
12.5.3 The integrated impulse response method  350
12.5.4 Lower limits for reverberation time  351
12.5.5 Spatial averaging  353
xii Contents

12.6 Vibrations 353
12.6.1 Calibration 353
12.6.2 Measurement positions  354
12.6.3 Frequency weighting  354
12.6.4 Integration time  356
12.6.5 Noise floor  356
12.6.6 Averaging of results  356
12.6.7 The vibration dose value  356
12.7 Loss factor  357
12.7.1 Structural reverberation time  357
12.7.2 Upper limits for loss factor  357
12.8 Radiation efficiency  358
12.9 Junction attenuation  359
References 360

13 Measurement uncertainty 363

13.1 Some statistical concepts  363
13.1.1 Uncertainty of a measurement result  363
13.1.2 The Gaussian distribution  364
13.1.3 Confidence levels  364
13.2 Variance of sound pressure in a room  365
13.2.1 Pure tone excitation  365
13.2.2 Band-limited noise excitation  367
13.2.3 Variance at low frequencies  368
13.3 Measurement uncertainty of sound
pressure level measurements  374
13.3.1 Probability density functions  374
13.3.2 Zones to be avoided  377
13.3.3 The influence of the measurement time  380
13.3.4 The integrated impulse response method  381
13.4 Spatial averaging  381
13.4.1 Correlation coefficient  381
13.4.2 Discrete spatial averaging  382
13.4.3 Continuous spatial averaging  384
13.5 Measurement uncertainty on
reverberation time measurements  385
13.5.1 General 385
13.5.2 The interrupted noise method  386
13.5.3 The integrated impulse response method  387
13.6 Measurement uncertainty on sound
insulation measurements  388
 Contents xiii

13.6.1 Combined measurement uncertainty  388

13.6.2 Example of airborne sound
insulation measurement  389
13.6.3 Measurement uncertainty on
single-number quantities  389
13.6.4 Application of uncertainties in relation
to requirements for sound insulation  390
References 392

14 Noise effects and subjective evaluation of sound insulation 395

14.1 Noise and health  395
14.1.1 Noise descriptors  395
14.1.2 Sleep disturbance  396
14.1.3 Dose-effect relationships  398
14.2 Noise from neighbours  400
14.2.1 Acoustical comfort  400
14.2.2 Sound insulation and social disruption  402
14.2.3 Influence of age and lifestyle  403
14.2.4 Willingness to pay for good sound insulation  404
14.3 Dose-response functions  405
14.3.1 Statistical methods  405
14.3.2 Traffic noise annoyance  407
14.3.3 Socio-acoustic surveys  408
14.3.4 Different traffic noise sources  411
14.3.5 Airborne sound insulation  413
14.3.6 Impact sound insulation  415
14.3.7 Vibrations  416
14.4 Low-Frequency noise from neighbors  417
14.5 Design criteria for satisfactory sound insulation  419
References 421

15 Experimental buildings with high sound insulation 425

15.1 ‘The Quiet House’ Lund, Sweden  425
15.1.1 General 425
15.1.2 Purpose of experiment  425
15.1.3 Building concept  426
15.1.4 Acoustical results  427
15.2 ‘OPEN HOUSE Urban Building System’, Denmark  428
15.2.1 General 428
15.2.2 Purpose of experiment  429
xiv Contents

15.2.3 Building concept  429

15.2.4 Acoustical results  430
15.3 ‘ECO-House’, Denmark  431
15.3.1 General 431
15.3.2 Purpose of experiment  432
15.3.3 Building concept  432
15.3.4 Acoustical results  433
15.4 ‘BRF Kajplatsen’, Stockholm, Sweden   434
15.4.1 General 434
15.4.2 Purpose of experiment  435
15.4.3 Building concept  436
15.4.4 Acoustical results  437
15.5 ‘Ylöjärvi’ apartments, Finland  440
15.5.1 General 440
15.5.2 Purpose of experiment  440
15.5.3 Building concept  440
15.5.4 Acoustical results  440
15.6 Nordic Wood project  443
15.6.1 General 443
15.6.2 Purpose of experiment  443
15.6.3 Acoustical results  443
References 444
Index 445
Preface

This book on sound insulation is based on my lecture notes developed over

a long span of time, beginning with a note on sound insulation from 1975.
After a few years of teaching basic acoustics, I wanted to establish a more
advanced course on architectural acoustics, meant for those students from
the building engineering faculty who wanted to study more in-depth about
acoustics and might consider doing their masters in acoustics. The course
included room acoustics as well as sound insulation. Hence, from 1977 to
1982, a series of lecture notes were written, covering both fields.
I am indebted to late Professor Fritz Ingerslev (1912–1994), who was my
teacher and mentor in the first years of my career. Quite early in my studies
at the Technical University of Denmark (DTU), I discovered his textbook
on building acoustics, and there is no doubt that this book was the rea-
son that I decided to turn my education in the direction of house building
and to specialize in acoustics. Professor Ingerslev suggested the topic of my
master’s thesis – and later also the topic of my PhD thesis – to be related to
the sound insulation of windows. I was very fortunate that the acoustical
laboratories at DTU had an incredibly good library with complete collec-
tions of all relevant acoustical journals and a rich collection of textbooks
and research reports. From around 1980, I was guided by another men-
tor, Jørgen Kristensen (1925–1992), who was head of the building acoustic
section at the Danish Building Research Institute (SBI), today a part of
Aalborg University, Copenhagen. He invited me to coauthor a book on
building acoustics, which was published in 1989. He also suggested that
I should take over his role as acoustical expert for the Danish Ministry
of Housing, which in turn led to my work for the Nordic Committee for
Building Regulations in 1993–1996.
My research has always been divided between building acoustics and
room acoustics. From around 1985, the development of room acoustic cal-
culation models was in focus and the software ODEON was launched in
1991. The following years were devoted to research projects, some of them
in buildings acoustics, but most of them related to room acoustics, musical

xv
xvi Preface

instruments and auralisation. In the 1990s, I also became involved with

standardisation and new measurement methods.
In 2008, I left the university. I started to work as a consultant for the
Norwegian company, Multiconsult and continued my work with the
Danish company, ODEON. After a while, I realised that some of my old
lecture notes might still be relevant in consulting and in 2014, I decided to
collect my former lecture notes on sound insulation and make a book of
it. Fortunately, Tony Moore, senior editor at CRC Press (imprint of Taylor
and Francis), was supportive of the idea. However, it was necessary to bring
everything up-to-date in accordance with the latest research results and
current international standards. Concerning references, I feel it important
to give credit to those who originally came up with ideas and contributed
to the field, and thus, the reader will find many references to old papers and
books. Science is not only about developing new knowledge but it is also
important to know the history behind the theories and methods and be able
to trace back to the origins, when needed.
The first five chapters of the book deal with fundamental matters,
mechanical vibrations, sound reflections and basic room acoustics. Chapter
5 describes sound insulation in accordance with the generally known clas-
sical theory. This first part of the book is more or less, what I would
teach in a general course on acoustics for building engineers. Chapters 6
through 15 are meant for the more advanced student, who wants a deeper
understanding of the physics behind sound transmission and measurement
methods. In recent years, there has been a growing focus on measure-
ment uncertainty in building acoustics, and especially at low frequencies.
Chapter 13 explains the fundamentals of measurement uncertainty, with
some emphasis on the spatial variance of the sound pressure in a room.
For some years, there has been a debate on the importance of the low fre-
quencies between 50 Hz and 100 Hz in relation to sound insulation. The
origin of this is the complaints from inhabitants in multi-unit houses con-
cerning insufficient sound insulation, especially against noise from traffic
and neighbours. Thus, Chapter 14 deals with annoyance and other sub-
jective aspects of sound insulation, explaining the statistical methods used
in socio-acoustic investigations, and presenting some of the findings. The
final chapter, Chapter 15, provides a background for discussions and ideas
on how to develop building technology, so that future houses can be built
with satisfactory sound insulation.
I am very grateful to Multiconsult for supporting me in my work with
this book by offering an office, where I could concentrate on writing,
surrounded by my books and papers. I have enjoyed the support of all
my colleagues in the acoustical section of the company. In particular, I
thank the head of the acoustics section, Ingunn Milford for her enthusi-
asm and support. Dr Arild Brekke has provided very important comments
 Preface xvii

and suggestions to improve Chapter 2 on vibrations. Special thanks go

to Dr Anders Løvstad, who has read the entire manuscript and provided
invaluable feedback and comments.

Oslo, August 2017

Jens Holger Rindel
About the Author

Jens Holger Rindel is currently a senior consultant at Multiconsult in

Norway and senior researcher at Odeon A/S in Denmark. He holds a
MSc in civil engineering (house building) and a PhD in acoustics, both
from the Technical University of Denmark. Until 2008, he was profes-
sor in acoustics at the Technical University of Denmark, Copenhagen,
Denmark. He is also past president of the Nordic Acoustical Association
and past chairman of the Technical Committee for Building and Room
Acoustics of the European Association of Acoustical Societies. He is
a fellow of the Institute of Acoustics and of the Acoustical Society of
America. He has been guest professor at universities in Sydney, Auckland
and Tokyo and worked as guest researcher at research institutions in
Braunschweig, Germany and Oslo, Norway.

xix
Introduction

The science of architectural acoustics is divided into room acoustics and

building acoustics. The origin of room acoustics as a scientific discipline is
closely connected to the pioneering work done by Wallace C. Sabine (1868–
1919) around 1896–1900. A theoretical model and a calculation method
for reverberation time were developed together with the first measurement
method. Carefully tuned organ pipes were used as sound sources and the
human ear as measuring instrument.
The beginning of building acoustics was quite different. Although there
was increasing awareness of the importance of sound insulation in the
beginning of the twentieth century, the theoretical models and calculation
methods available then were very simple and insufficient. The importance
of mass for the sound insulation of massive single constructions was known
since the work of Berger (1911). The need for completely separate wall con-
structions for high sound insulation was also a known principle, which
was used in special cases, such as theatres and concert halls. Sabine (1915)
described three cases where he had tried to design for good sound insula-
tion but, to his frustration, the results did not meet his expectations. With
a touch of bitterness, he says “It is always easier to explain why a method
does not work than to know in advance whether it will or not. It is espe-
cially easy to explain why it does not work when not under the immedi-
ate necessity of correcting it or of supplying a better.” This was in direct
response to the unsuccessful attempts at insulation for the New England
Conservatory of Music.
Airborne sound insulation was measured with organ pipes as an exten-
sion of the method used for reverberation time. For laboratory measure-
ments, a test wall was built between a large reverberant room and a smaller
receiving room with high acoustic damping. The operator used a stopwatch
to measure the decay time from when the source was stopped until the
time that the sound was inaudible. When measured in the receiving room,
the sound intensity was lower and thus, the audible decay time was much
shorter than in the source room, and the transmission coefficient of the test
wall could be calculated from the difference in decay times. The reverber-
ation method for sound insulation is described in detail by Paul E. Sabine

xxi
xxii Introduction

(1930), who succeeded Wallace C. Sabine as head of the acoustical test

laboratories at Riverbank (Kopec, 1997). In 1930, the organ pipes had
been replaced by loudspeakers and 17 test tones were used from 128 Hz to
4096 Hz.
Barkhausen (1926) developed an instrument for measuring the loudness
level in phon. The operator listened to the noise with one ear and adjusted
the level of a 1000 Hz test sound heard in the other ear, to get equal loud-
ness. This made it possible to do sound insulation measurements with a
stationary noise source, and several mechanical devices were put to use in
the following years. Dr H. Reier from the technical university in Stuttgart
used a very noisy rattle machine with a frequency range from 100 Hz to
2000 Hz. He was asked to do measurements in a very complicated case in
Copenhagen, where new broadcasting studios had been built on top of a
theatre using separated constructions. The measurement report by Reier
(1931) is reproduced in the book by Oelsner (1935). The measured result
was a very high airborne sound insulation of 83 phons. The unit was phon
since the sound insulation was measured as the difference between the
loudness levels on either side of the construction.
Vern O. Knudsen (1930) describes three measurement methods for air-
borne sound insulation, which he applied in the newly built acoustic test
facilities at the University of California. One of them is Sabine’s reverbera-
tion method with loudspeakers as sound sources. The second method is the
insertion loss, i.e., a measurement with and without the test panel. The third
method uses an electric attenuator to adjust the level of the loudspeaker in
the source room until the sound is inaudible. The reported results of the
three methods are in surprisingly good agreement (within about 1 dB) with
each other. A few years later, a new test method was developed by Meyer
and Keidel (1935) using a logarithmic attenuator and a voltage meter to
draw a curve of the sound pressure level as a function of frequency. For the
first time, it was possible to measure sound insulation without using the
human ear as a measuring instrument.
Another issue that would need attention was the impact sound from
floors, especially footfall noise. An early method of measuring impact sound
used the dropping of a steel ball from different heights above the floor; the
height of fall that made the sound just audible in the receiving room was
taken as a measure of the impact sound. Later, the tapping machine, as it
is known today, was developed and the impact sound level of the tapping
noise was measured as a loudness level in phon. The comprehensive book by
Schoch (1937) presents the state of the art in building acoustics at the time.
Although the measuring methods for airborne and impact sound insulation
had been developed, the physical and technical foundation for understand-
ing sound insulation problems was still very poor. It was restricted to the
mass law and the advantage of separated structures.
A scientific approach to the physics of sound insulation began in 1942,
when Lothar Cremer (1905–1990) explained the importance of bending
 Introduction xxiii

waves and introduced the concept of coincidence in sound transmission in

his pioneering article (Cremer, 1942). Albert London (1950) was the first to
present a theoretical model for sound transmission through double walls.
Today, there is no doubt that bending waves, the coincidence effect and
the critical frequency are key points in understanding sound insulation.
However, the knowledge of Cremer’s findings spread very slowly and did
not reach the majority of those working with building acoustics until sev-
eral decades later. Looking into the textbooks by Brüel (1946), Ingerslev
(1949) and Knudsen and Harris (1950), the coincidence effect is not even
mentioned. The book by Ingerslev was in use at the Technical University
of Denmark until 1971, and not until then, would the teaching in building
acoustics include models of sound transmission that were more advanced
than the mass law. The book by Knudsen and Harris was reprinted in 1978,
because it was still in use in education in many countries.
From around 1930, building technology changed dramatically. The
introduction of concrete slabs with thickness of only 80 mm to 100 mm
caused severe noise problems in houses. It became very clear that dimen-
sions determined solely from the point of view of structural engineering
were not sufficient from an acoustical point of view. Several countries
introduced minimum requirements for sound insulation between dwellings
in the 1940s or 1950s, and these first requirements were typically in terms
of specific acceptable constructions, e.g. a 250 mm full-brick wall with
plaster on both sides. The industrialised building technology, which started
in the 1960s, and the extensive use of reinforced concrete led to buildings
that were uniform and relatively cheap, but not particularly satisfactory
for the inhabitants. These houses became synonymous with noise prob-
lems and still today, concrete as a building material is often believed by
the layman to cause poor sound insulation. The introduction of building
codes with minimum requirements for sound insulation had some negative
effects; building apartment houses was a matter of money investment, and
thus, it was important to keep the building costs low. The result was that
any sound insulation better than the minimum requirement was considered
a waste of money. This led to houses with a sound insulation that was just
acceptable in those days, but is clearly insufficient today, about 50 years
later. Since it is always difficult and expensive to improve the sound insu-
lation in an existing house, these houses from the 1960s are often pulled
down and replaced by completely new buildings.
New building technologies, new powerful sound sources in homes (hi-fi
loudspeakers, home cinemas with subwoofers, etc.) and increased aware-
ness of noise in the society have all contributed to the development of bet-
ter sound insulation design. An important part of this development is the
acousticians’ understanding of the physical principles in sound insulation.
Another important part of the development is the knowledge about how
individuals react to noise from neighbours, and how the sound insulation
of a building is related to the degree of satisfaction that can be expected
xxiv Introduction

among the inhabitants. In recent years, this has led to raised minimum
requirements for sound insulation in many countries. The idea of a sound
classification system for houses, with specified sound classes better than the
minimum requirements, has aroused interest, and an international stand-
ard for sound classification of houses is under preparation (ISO/DIS 19488,
2017). Projects with experimental houses that have shown new directions
for building technology with better sound insulation are very important.
Much more can be done along this line in the future, and it is important to
spread the information to the public that good sound insulation in houses
is possible. A major problem may be to have a general acceptance among
architects and in the building industry that sound insulation is a design
parameter that must be taken seriously.

REFERENCES

H. Barkhausen (1926). Ein neuer Schallmesser für die Praxis (A new sound meas-
uring device for practical application, in German). Zeitschrift für technische
Physik 7, 599–601.
R. Berger (1911). Über die Schalldurchlässigkeit. (On the Sound Transmission, in
German). Dissertation, Munich, Germany.
P.V. Brüel (1946). Lydisolation og rumkustik (Sound insulation and room acous-
tics, in Danish). Jul. Gjellerups Forlag, Copenhagen.
L. Cremer (1942): Theorie der Schalldämmung dünner Wände bei schrägem
Einfall. (Theory of sound insulation of thin walls at oblique incidence, in
German). Akustische Zeitschrift 7, 81–104.
F. Ingerslev (1949). Akustik. Lærebog i bygningsakustik for ingeniører (Acoustics.
Textbook on building acoustics for engineers, in Danish). Teknisk Forlag,
A-S Dansk Ingeniørforenings Forlag, Copenhagen.
ISO/DIS 19488 (2017). Acoustics—Acoustic classification of dwellings. (Under
development.) International Organization for Standardization, Geneva,
Switzerland.
V.O. Knudsen (1930). Measurement and calculation of sound-insulation. Journal
of the Acoustical Society of America 2, 129–140.
V.O. Knudsen and C.M. Harris (1950). Acoustical Designing in Architecture. John
Wiley & Sons, Inc., New York. Reprint 1978 by the Acoustical Society of
America, New York.
J.W. Kopec (1997). The Sabines at Riverbank. Their Role in the Science of
Architectural Acoustics. Acoustical Society of America, New York.
A. London (1950). Transmission of reverberant sound through double walls.
Journal of the Acoustical Society of America 22, 270–279.
E. Meyer and L. Keidel (1935). Röhrenvoltmeter mit logarithmischer Anzeige
und seine Anwendungen in der Akustik, (Vacuum-tube voltmeter with log-
arithmic display and its applications in acoustics, in German). Elektrische
Nachrichtentechnik (ENT) 12(2), 37–46.
 Introduction xxv

W. Oelsner (1935). Det teknisk-videnskabelige Grundlag for Rumakustik og

Lydisolation (The technical-scientific foundation of room acoustics and
sound insulation, in Danish). Hjorths Tryk, Copenhagen.
P.E. Sabine (1930). The transmission of sound by walls. Journal of the Acoustical
Society of America 1, 181–201.
W.C. Sabine (1915). The insulation of sound. The Brickbuilder 24(2). Paper No. 10
in Collected Papers on Acoustics. Harvard University Press, 1922. Published
1964 by Dover Publications Inc., New York.
A. Schoch (1937). Die physikalischen und technischen Grundlagen der
Schalldämmung im Bauwesen (The physical and technical fundamentals of
sound insulation in buildings, in German). S. Hirzel Verlag, Leipzig.
Chapter 1

Basic concepts in acoustics

This chapter summarizes the most important basic concepts used in building
acoustics. The definitions and symbols are in accordance with International
Organization for Standardization (ISO) technical report (TR) 25417.

1.1 INTRODUCTION

The sound insulation of walls and floors in a building is often of major con-
cern for the people using the building. It is of special interest in residential
buildings, but also in, e.g. offices, hospitals and school buildings. Sound
insulation in buildings has been under research and standardization for a
long time – many contributions were made in the 1950s. However, since that
time, the behaviour and habits of people, as well as their expectations, have
changed. Noise sources have become louder and cover a broader frequency
spectrum, and it is possible that the demands of the tenants have increased.
At the same time, there are a gradual development and change in build-
ing elements and methods. Thus, knowledge of physical and other aspects
of sound insulation in buildings is essential for an acoustician. Moreover,
basic knowledge of sound insulation is also important in other branches of
acoustics, e.g. in the development of vehicles, ships and aeroplanes.
Sound insulation is divided into airborne sound insulation, where the
noise source is in the air inside or outside the building, and structure-borne
sound insulation, where the noise source is located at a building structure.
The most important case of structure-borne sound insulation is the impact
noise sound insulation, which includes noise generated by people walk-
ing. Chapter 5 gives a basic overview of the principles of sound insulation.
Chapters 8 and 9 deal more thoroughly with airborne sound insulation,
and Chapter 10 addresses the impact noise.
Airborne noise sources in a building are typically people talking, sound
from stereo equipment, televisions and musical instruments. The develop-
ment of hi-fi stereo equipment means that these sources are much louder
and cover more of the low frequencies than was the case when the building
regulations for sound insulations were developed. It is thus not given that

1
2  Sound Insulation in Buildings

a regulation originally intended to avoid noise discomfort still does so. It is

therefore most often recommended to use a broader frequency spectrum (50
Hz to 5000 Hz) than is used in the original measurement description (100
Hz to 3150 Hz), and sometimes it is also important to perform a subjective
evaluation of the noise situation in more experimental buildings. Examples
of external airborne noise sources are traffic and industrial noise, where the
windows of a building are of major interest.

1.2 THE DECIBEL SCALE

Sound is an audible vibration, either in the air or in liquids or solid materi-

als. The human ear is sensible to the pressure variations in a sound wave,
and since the audible dynamic range is immense, the logarithmic decibel
scale is used. The sound pressure level Lp is

p2
Lp = 10 ⋅ lg 2
(dB) (1.1)
pref

where pref = 20 μPa is the reference sound pressure. The sound pressure is
the root-mean-square (RMS) value of the time-varying sound pressure. For
harmonic vibrations, p 2 = 1/2 |p|2 , where |p| is the amplitude of the sound
pressure. In rough terms, the range of audible sound is 0 dB to 120 dB.
A sound source is characterized by its emitted sound power, which can
also be expressed in decibels. If P is the sound power in watts, the sound
power level is

P
LW = 10 ⋅ lg (dB) (1.2)
Pref

Table 1.1  Definition of levels in decibels

Quantity Symbol Reference Level (dB)
Sound power P 1 pW  Quantity 
10lg 
Sound intensity I 1 pW/m2  Reference 
Sound pressure (in air) p 20 μPa  RMS quantity 
2
10lg 
Force F 1 μN  Reference 
Acceleration a 1 μm/s2
 RMS quantity 
Velocity v 1 nm/s = 20lg 
 Reference 
Displacement x 1 pm
Note the use of prefixes: μ = 10−6, n = 10−9, p = 10−12.
 Basic concepts in acoustics  3

where Pref = 10 −12 W is the reference sound power. The relation between
sound power level and sound pressure level is addressed in Chapter 4.
The decibel scale is also used for many other acoustic parameters,
e.g. vibratory acceleration, vibratory velocity and sound intensity (see
Table 1.1). The reference values are defined in International Organization
for Standardization (ISO) 1683.

1.3 FREQUENCY AND RELATED CONCEPTS

1.3.1 Harmonic vibrations
If the velocity of particle motions is small compared to the speed of sound
propagation in the medium, the vibrations can be considered harmonic
vibrations. Concepts such as sound pressure, velocity, acceleration, and
force can be described mathematically by either real or complex numbers as

y = y cos(ω t + ϕ ) (1.3)

or in the complex notation as

y = y ej(ωt+ϕ ) = y ( cos(ω t + ϕ ) + jsin(ω t + ϕ )) (1.4)

In the latter case, y is a complex number that can be considered in two dif-
ferent ways: either having a size (the modulus |y|) and an angle (ωt + φ) in
the complex plane or having a real part (the cosine term) and an imaginary
part (the sine term). Whatever notation is used, the physical vibration that
can be observed is just the cosine part. However, the complex notation has
some mathematical advantages.
The imaginary unit j = −1 may also be written as j = 1ej(π /2) , i.e. as the
unit vector turned perpendicular to the real axis.
In the aforementioned expressions, t is the time, φ is a phase angle and ω
is the angular frequency:

2π
ω = 2π f = (rad/s) (1.5)
T1

where f is the frequency (Hz) and T1 is the period duration (s).

A wave propagating in the direction of the x-axis can be described math-
ematically by

y = y ej(ω t −kx) = y ejω t e− jkx (1.6)

4  Sound Insulation in Buildings

The factor ejωt is called the time factor, and it is often omitted in theo-
retical derivations if it does not directly affect the result or calculations.
Sometimes, the time factor e−iωt is used instead, which is equally fine; but it
is very important to be consistent and use the same time factor throughout
any derivations or calculations.
The place factor e− jkx describes the propagation in the positive direction
of the x-axis, and it contains the angular wave number k, which is

2π ω
k= = (m −1) (1.7)
λ c

where λ is the wavelength (m) and c is the speed of sound (m/s).

1.3.2 Speed of sound
The speed of sound is the distance travelled by sound per unit time. It is
found as the distance travelled during one period (one wavelength λ) multi-
plied by the number of periods per second (the frequency f):

c = λf (m/s) (1.8)

In air and other gases or fluids, the speed of sound is a function of the bulk
modulus K s (Pa) and the density ρ (kg/m3) of the media it travels in

Ks
c= (m/s) (1.9)
ρ

In indoor air with a temperature of ~ 20 °C, the density is ρ ≈ 1.205 kg/m3

and the speed of sound is c ≈ 344 m/s.

1.3.3 Frequency bands
Acoustic measurements are usually made in frequency bands. The
bandwidth Δf is the difference between the upper and lower limiting
frequencies:

∆f = f2 − f1 (Hz) (1.10)

The centre frequency is the geometrical mean of the limiting frequencies:

fcenter = f1f2 (Hz) (1.11)

 Basic concepts in acoustics  5

The relative bandwidth is the ratio between the bandwidth and the centre
frequency:

∆f f2 f
= − 1 (1.12)
fcenter f1 f2

In building acoustics, the most often used frequency bands are one-third
octave bands:

f2
= 101/10 ≅ 3 2 ∆f ≅ 0.23fcenter (1.13)
f1

Some times octave bands are used in stead:

f2
= 103/10 ≅ 2 ∆f ≅ 0.71fcenter (1.14)
f1

In music, the octave is defined as a frequency interval where the upper fre-
quency is exactly two times the lower frequency. (The name comes from the
Latin Octavus, meaning ‘eight’.) Playing an ascending scale on the white keys
of a keyboard, the octave is step number 8. Using both black and white keys,
the octave is divided into 12 semitones. Thus, the one-third octave is the interval
of four semitones, which is called a major third in music. (In German literature
“Terz” is often used for one-third octave.) Using just intonation, the major third
has the frequency interval = 5/4 = 1.25. In the modern equal temperament, the
semitones all have equal interval = 21/12, and the major third is 21/3 ≈ 1.259921.
For comparison, the exact interval of the one-third octave is 101/10 ≈ 1.258925.
When applied to filters for acoustic measurements, ISO 266 defines the
exact and nominal center frequencies to be periodic within a decade, i.e.
the one-third octave bands are, strictly speaking, one-tenth decade bands
(see Table 1.2). If n denotes the band number, the exact center frequency is
calculated from the following equation:

fcenter = 10n /10 (1.15)

The standard range for the one-third octave center frequencies is 1

­ 00 Hz
to 3150 Hz, although an extended frequency range of 50 Hz to 5000 Hz
is often preferred. Sometimes, even lower frequencies are used for impact
sound, and for vibrations, the standard range is 1 Hz to 80 Hz.
6  Sound Insulation in Buildings

Table 1.2  Exact and nominal centre frequencies in hertz of the one-third octave
bands often used in building acoustics
Band no. Exact Nominal Band no. Exact Nominal
14 25.12 25 26 398.11 400
15 31.62 31.5 27 501.19 500
16 39.81 40 28 630.96 630
17 50.12 50 29 794.33 800
18 63.10 63 30 1000.00 1000
19 79.43 80 31 1258.93 1250
20 100.00 100 32 1584.89 1600
21 125.89 125 33 1995.26 2000
22 158.49 160 34 2511.89 2500
23 199.53 200 35 3162.28 3150
24 251.19 250 36 3981.07 4000
25 316.23 315 37 5011.87 5000

1.4 LOUDNESS LEVEL AND A-WEIGHTED

SOUND PRESSURE LEVEL

The sensation of sound is very frequency dependent. The subjective loud-

ness of a low-frequency sound (e.g. below 200 Hz) is much lower than a
sound with the same physical sound pressure level in the frequency range of
1 kHz to 5 kHz. This is displayed in Figure 1.1 as equal-loudness contours,
i.e. curves showing combinations of sound pressure level and frequency
that give the same subjective loudness. The loudness level of a sound is
defined as the sound pressure level of a 1000 Hz tone that has the same
loudness as the sound. The unit of the loudness level is phon.
In building acoustics, it is usual that sound is measured in one-third
octave bands because transmission properties are very frequency depen-
dent. However, it is practical to express the sound level as a single-number
value, and the commonly used value is the A-weighted sound pressure level.
A-weighting applies a frequency-dependent attenuation, which is in the
A-weighting filter in a sound level meter (IEC 61672-1). The A-weighted
sound pressure level can also be calculated from measured sound pressure
levels in one-third octave bands Lp,i using the following equation:

∑ 10
0,1(Lp ,i + Lcorr,i )
LA = 10 lg (dB ) (1.16)
i

where L corr,i is the correction at frequency band i (Table 1.3). Actually, the
A-weighting filter is a rough approximation of the 40 phon curve.
 Basic concepts in acoustics  7

130

110
100 phon

90
80 phon
Sound pressure level, dB

70
60 phon

50
40 phon

30
20 phon

10

Hearing threshold
–10
10 100 1,000 10,000
Frequency, Hz

Figure 1.1 Equal loudness contours and hearing threshold. (Adapted from ISO 226,
Acoustics – Normal Equal-Loudness-Level Contours, International Organization
for Standardization, Geneva, Switzerland, 2003.)

Table 1.3  A-weighting corrections in one-third octave bands

Frequency (Hz) A-weighting (dB) Frequency (Hz) A-weighting (dB)
10 −70.4 500 −3.2
12.5 −63.4 630 −1.9
16 −56.7 800 −0.8
20 −50.5 1000 0
25 −44.7 1250 0.6
31.5 −39.4 1600 1
40 −34.6 2000 1.2
50 −30.2 2500 1.3
63 −26.2 3150 1.2
80 −22.5 4000 1
100 −19.1 5000 0.5
125 −16.1 6300 −0.1
160 −13.4 8000 −1.1
200 −10.9 10000 −2.5
250 −8.6 12500 −4.3
315 −6.6 16000 −6.6
400 −4.8 20000 −9.3
Source: 
IEC 61672-1, Electro acoustics – Sound level meters – part 1: specifications.
International Electrotechnical Commission, Geneva, Switzerland, 2002.
8  Sound Insulation in Buildings

1.5 PROPAGATION OF SOUND IN PLATES

In order to understand the physics behind sound insulation, it is necessary

to have some basic knowledge about structure-borne sound, especially the
wave phenomena in plates. A single wall is basically a plate.
Solid materials differ from fluids (such as air and water) in that they can
resist shear forces. This implies that apart from the longitudinal wave (that
is present in fluid), there will also be shear waves as well as combinations
of these two basic wave types (Figure 1.2). Combined waves are typical for
plates, rods and beams. The bending wave is the most important combined
wave type.
The two basic wave types are generally uncoupled in extended materials
and therefore independent of each other. At a boundary, there will however
be coupling and transformation from one wave type to the other. This is the
reason for the combined waves in plates, rods and beams.
The basic wave types are both controlled by the wave equation; in the
case of a plane wave propagating in the x-direction, it will be the ordinary
wave equation, which can be written as follows:

∂2 u 1 ∂2 u (1.17)
=
∂ x2 c 2 ∂t 2

where u is the particle velocity. In the longitudinal case, this velocity will be
in the x-direction (particle movement in the same direction as the propaga-
tion), whereas in the transversal case, it will be in the y and z directions
(more or less perpendicular to the direction of propagation). Note, how-
ever, that in the case of general non-plane wave motion, the differential
equations will be a little bit more complicated, especially in the transverse
case. As the waves are described by the ordinary wave equation, they will

(a) λI CI (b) λs CS

(c) λk CK (d) λB CB

Figure 1.2 Important types of waves. (a) Longitudinal; (b) quasilongitudinal; (c) shear; (d)
bending. (Reproduced from Kristensen, J., Rindel, J.H., Bygningsakustik – Teori og
praksis (Building acoustics – Theory and practice, in Danish), SBI-Anvisning 166,
Danish Building Research Institute, Hørsholm, Denmark, 1989. With permission.)
 Basic concepts in acoustics  9

behave in the same way as the acoustic wave in air; only the magnitude of
the wave speed c will differ. Also, note that c will be constant with respect
to the frequency. However, a combined wave such as the bending wave can-
not be described by the ordinary wave equation, and the wave speed will be
frequency dependent. This is called dispersion.

1.5.1 Longitudinal waves
The bulk modulus of a solid is the ratio between a small pressure increase
and the resulting decrease in volume or increase in density. It is slightly
different under adiabatic or isothermal conditions, which depends on the
thermal conductivity of the material. In solid structures, the bulk modulus
and thus the speed of longitudinal waves depend on the actual shape of the
structure. The simplest case is that of a bar whose sides are free to move
as the compression travels through the bar. Hence, the bulk modulus of
the bar equals the Young’s modulus E, and the speed of the compression
wave is

E
cL,bar = (m/s) (1.18)
ρm

where ρm is the density of the solid material. In a plate, the Poisson

ratio μ tends to increase the bulk modulus, and the speed of a longitudinal
wave travelling along the plate is

E
cL,plate = (m/s) (1.19)
ρ m (1 − µ 2 )

In an extended solid medium, i.e. without nearby surfaces, the speed is

E(1 − µ)
cL,extended = (m/s) (1.20)
ρ m (1 + µ)(1 − 2µ)

Most solid materials have Poisson ratios between 0.1 and 0.5, the latter
being the theoretical upper limit. Some examples are given in Table 1.4.

1.5.2 Bending and shear waves in plates

In plates and beams, there exists wave motion in the form of bending –
that is, the same type of motion as when you bend a stick. The motion
is perpendicular to both the direction of propagation and the surface
of the structure (Figures 1.2 and 1.3). The bending wave is very impor-
tant in acoustics in general as it can radiate sound very efficiently to the
10  Sound Insulation in Buildings

h λB CB

Figure 1.3 A bending wave in a plate with thickness h.

surrounding air. The reason for this is partly due to the transverse out-
of-plane motion and the low impedance associated with this motion, and
partly due to the frequency dependency of the wave speed, as will be dis-
cussed later.
A plate with thickness h can be characterized by the mass per unit area,

m = ρ m h (kg/m 2 ) (1.21)

the bending stiffness per unit width

Eh3
B= (Nm) (1.22)
12(1 − µ 2 )

and the shear modulus

E
G= (N/m 2 ) (1.23)
2(1 + µ)

The pure bending wave, typical for thin plates and long wavelengths, is
controlled by the bending wave equation. In the case of a plane wave propa-
gating in the x-direction, the bending wave equation is

∂4 v ∂2 v
B 4
+m 2 =0 (1.24)
∂x ∂t

where v is the vibration velocity perpendicular to the plate. It should be

noted that compared to the ordinary wave equation, the bending wave
equation is of the fourth order in the spatial derivative and of the second
 Basic concepts in acoustics  11

order in the time derivative. This implies that the wave speed is frequency
dependent and given by

B Eh2
cB = ω 4 = 2π f 4 (m/s) (1.25)
m 12ρ m (1 − µ 2 )

The bending wave speed is thus proportional to the square root of the fre-
quency. This is in good agreement with reality at low frequencies (and thin
plates). However, it is unrealistic to have a wave speed that approaches
infinity in the limit of high frequencies. The bending wave equation and
wave speed are low-frequency approximations.
Another transversal wave is the shear wave, in which the sections of a
plate are shifted in parallel, perpendicular to the direction of propagation
(Figure 1.2). The speed of a shear wave is

G E
cS = = (m/s) (1.26)
ρm 2ρ m (1 + µ)

In contrast to bending waves, shear waves depend neither on the frequency

nor on the thickness of the plate. If the thickness of the plate exceeds
approximately one-sixth of the wavelength, the resistance of the bending
wave exceeds the resistance of the shear wave. Thus, with increasing fre-
quency, there is a gradual transition from bending to shear, which can be
approximated by the following expression for the effective transverse wave
speed (Rindel 1994):

−1/3
 1 1
cB,eff ≅  3 + 3 (1.27)
 cB cS 

The frequency dependency of the effective transverse wave speed compared

to the bending wave speed is shown in Figure 1.4. When the speed of propa-
gation depends on the frequency, it is called dispersion, and the curve in
Figure 1.4 is a dispersion curve.
The effect of dispersion of bending waves can be observed in winter-
time when skating on the ice on a lake. An impact on the ice some dis-
tance away can be heard as a characteristic chirp sound, starting at high
frequencies and quickly descending to low frequencies. The explanation
of this phenomenon is that the vibrations propagate as bending waves
along the plate of ice, and the high frequencies travel faster than the low
frequencies.
12  Sound Insulation in Buildings

1000
Phase speed, m/s

c = 344 m/s

fc fs
100
10 100 1,000 10,000
Frequency, Hz
cB cS cB,eff

Figure 1.4 The effective speed of transverse waves in a plate showing the transition from
bending to shear. In this example, the speed of shear waves is three times
c, the critical frequency is 80 Hz, and the crossover frequency is 720 Hz.

1.5.3 Critical frequency
The acoustic wave speed in air is independent of the frequency (c ≈ 344 m/s
at 20 °C), whereas the bending wave speed and the effective transverse wave
speed are frequency dependent. This has the consequence that there exists a
frequency where the two wave types have the same speed and wavelength,
and therefore couple easily. A good coupling means that the plate will easily
radiate sound at and above this frequency. Actually, it is tempting to adopt
the Mach number M known from fluid dynamics and aircraft technology.
This is the ratio of speed of propagation to the speed of sound in air. Thus,
the Mach number of a structural wave is M = cB,eff /c.
The critical frequency fc is defined as the frequency at which cB = c = 344 m/s
(Mach 1). Thus, from Equation 1.25, it follows that

c2 m c 2 3ρ m (1 − µ 2 ) Kc
fc = = = (1.28)
2π B πh E h

where Kc is material constant. It should be noted that for a given material,

the critical frequency is inversely proportional to the plate thickness, h.
The crossover frequency from bending waves to shear waves, fs is defined
as the frequency at which cB = c S :

c s2 m 1 3E(1 − µ) (1.29)
fs = =
2π B 2π h ρ m (1 + µ)

 Basic concepts in acoustics  13

Table 1.4  Examples of typical parameters for common building materials

Loss Shear
Young’s Poisson factor, wave
Material Density, modulus, ratio, internal speed,
ρm E η Kc cs
(103 kg/m3) (109 N/m2) μ (10−3) (Hz∙m) (m/s)
Concrete 2.30 26 0.20 10 19 2150
Clinker concrete 1.70 4.6 0.10 15 39 1100
Light clinker concrete 1.30 3.8 0.10 15 38 1150
Porous concrete 0.60 0.9 0.10 15 53 820
Brick 1.90 24 0.10 10 18 2400a
Light brick 1.60 14 0.10 10 22 2000a
Gypsum board 0.84 3 0.30 15 33 1150
Glass 2.50 60 0.24 1 13 3000
Steel 7.80 210 0.33 0.1 12 3200
Aluminium 2.70 72 0.33 0.1 12 3150
Lead 11.30 17 0.45 100 48 720
Wooden chipboard 0.70 4.6 0.20 20 25 1650
Wood, pine, parallel 0.50 9.8 0.10 10 15 2980
to fibres
Wood, pine, 0.50 0.2 0.10 10 103 420
perpendicular to fibres
Laminated wood, 0.60 6.2 0.10 28 20 2150
parallel to fibres
Laminated wood, 0.60 2.3 0.10 28 33 1300
perpendicular to fibres
Source: 
Cremer, L., M. Heckl, Structure-Borne Sound, Structural Vibrations and Sound
Radiation at Audio Frequencies, 2nd edition, Springer, Berlin and Heidelberg, 1988;
Fasold, W., E. Sonntag, Bauakustik. Bauphysikalische Entwurfslehre, (Building acous-
tics. Designing in building physics, in German). Band 4, 3rd edition, VEB Verlag,
Berlin, 1978; Bodlund, K., Luftlydisolering. En sam­manstälning av tillämplig teori.
(Airborne sound insulation. A collection of applicable theory, in Swedish). Rapport
R60:1980, Statens råd för Byggnadsforskning, Stockholm, 1980; Rindel, J.H., Appl.
Acoust., 41, 97–111, 1994.
a Brick and light brick walls are not homogeneous, and thus, the actual shear wave speed is
2–3 times lower than that calculated from the equations.

Thus, the ratio between fs and fc is

f s cS2 E (1.30)
= 2 = 2
fc c 2c ρ m (1 + µ)

i.e. only dependent on the material properties. Examples for some building
materials are given in Table 1.4.
14  Sound Insulation in Buildings

1.6 AVERAGING

Four different kinds of averaging are applied in acoustics – arithmetic, geo-

metric, harmonic and energetic.

1.6.1 Arithmetic average
1
a=
2
( x + y) (1.31a)

n

a=
1
n ∑x i (1.31b)
i =1

Example: The average reverberation time in a room is the arithmetic aver-

age of the reverberation times in n positions.

1.6.2 Geometric average

a= xy (1.32)

Example: The centre frequency of an octave band is the geometric average

of the upper and lower limiting frequency.

1.6.3 Harmonic average

2xy
a= (1.33)
x+y

The harmonic average is, in fact, the arithmetic average of the reciprocal
values, 1/a = (1/x + 1/y)/2.
Example: The characteristic distance from a reflecting surface is the
­harmonic average of the distance to the source and the distance to the
receiver.

1.6.4 Energetic average

1 n

La = 10 lg 
 n
∑10
i =1
0.1⋅Li
 (dB)

(1.34)

Example: L a is the sound pressure level averaged over n positions in a room,

and Li is the sound pressure level at position i.
 Basic concepts in acoustics  15

1.7 USEFUL MATHEMATICAL FUNCTIONS AND FORMULAE

1.7.1 Trigonometrical formulae

2 cos ( x ) sin ( x ) = sin ( 2x ) (1.35)

cos2 ( x ) − sin2 ( x ) = cos ( 2x ) (1.36)

2 cos2 ( x ) = 1 + cos ( 2x ) (1.37)

2 sin2 ( x ) = 1 − cos ( 2x ) (1.38)

1.7.2 Euler’s formulae

ejx + e− jx
cos ( x) = (1.39)
2

ejx − e− jx
sin ( x) = (1.40)
2j

ejx = cos ( x) + j sin ( x) (1.41)

1.7.3 The cardinal sine function

sin ( x)
sinc ( x) = (1.42)
x

Note that sinc (0) = 1. This function is also called the 0th order spherical
Bessel function of the first kind, j0(x). The function is shown graphically in
Figure 1.5.

1.7.4 Logarithmic functions
The common logarithm, base 10:

y = 10x ⇔ x = log10 ( y ) = lg ( y ) (1.43)

lg ( xy ) = lg ( x ) + lg ( y ) (1.44)
16  Sound Insulation in Buildings

0.8

0.6

0.4
sinc (x)

0.2

–0.2

–0.4
0 5 10 15 20
x

Figure 1.5 The sinc function.

( )
lg xn = n lg ( x )

(1.45)

The natural logarithm, base e:

y = ex ⇔ x = ln ( y ) (1.46)

Conversion between common and natural logarithms:

lg ( x)
ln ( x) = (1.47)
lg (e)

ln ( x )
lg ( x ) = (1.48)
ln (10)

1.7.5 Basic equations for complex numbers

Complex number : z = Re { z} + jIm { z} = z ejθ (1.49)

Real part : Re { z } = z cos θ (1.50)

Imaginary part : Im { z } = z sin θ (1.51)

 Basic concepts in acoustics  17

Modulus : z = Re { z } + Im { z } 2
2
(1.52)

 Im { z } 
Argument : θ = arctan   (1.53)
 Re { z } 

Complex conjugate : z* = Re { z} − jIm { z} = z e− jθ (1.54)

Imaginary unit : j = ejπ /2 , j2 = ejπ = −1 (1.55)

Multiplication : z1z2 = z1 z2 ej(θ1+θ2 ) (1.56)

z1 z1 j(θ1−θ2 )
Division : = e (1.57)
z2 z2

REFERENCES

K. Bodlund (1980). Luftlydisolering. En sammanstälning av tillämplig teori,

(Airborne sound insulation. A collection of applicable theory, in Swedish).
Rapport R60:1980, Statens råd för Byggnadsforskning, Stockholm, Sweden.
L. Cremer and M. Heckl (1988). Structure-Borne Sound, Structural Vibrations and
Sound Radiation at Audio Frequencies. Translated and revised by E.E. Ungar,
2nd edition. Springer-Verlag, Berlin.
W. Fasold and E. Sonntag (1978). Bauakustik. Bauphysikalische Entwurfslehre
(Building acoustics. Designing in building physics, in German), Band 4, 3rd
edition. VEB Verlag, Berlin.
IEC 61672-1 (2002). Electro acoustics – Sound level meters – Part 1: specifications.
International Electrotechnical Commission, Geneva, Switzerland.
ISO 1683 (2008). Acoustics – Preferred reference values for acoustical and vibratory
levels. International Organization for Standardization, Geneva, Switzerland.
ISO 226 (2003). Acoustics – Normal equal-loudness-level contours. International
Organization for Standardization, Geneva, Switzerland.
ISO 266 (2003). Acoustics – Preferred frequencies. International Organization for
Standardization, Geneva, Switzerland.
ISO/TR 25417 (2007). Acoustics – Definition of basic quantities and terms. Technical
report. International Organization for Standardization, Geneva, Switzerland.
J. Kristensen and J.H. Rindel (1989), Bygningsakustik – Teori og praksis (Building
acoustics – Theory and practice, in Danish). SBI-Anvisning 166. Danish
Building Research Institute, Hørsholm, Denmark.
J.H. Rindel (1994). Dispersion and absorption of structure-borne sound in acousti-
cally thick plates. Applied Acoustics 41, 97–111.
Chapter 2

Mechanical vibrations

2.1 A SIMPLE MECHANICAL SYSTEM

In this chapter, we shall look at one-dimensional vibrations in a simple

mechanical system consisting of a mass, a spring and a damping element.
Since the vibrations can be fully described in one dimension, this is also
called a system with one degree of freedom.
First, the use of complex notation is shown for harmonic vibrations as
a basis for later applications in acoustic vibrations and sound fields. Next,
the vibrations in a resonant mechanical system are dealt with, as a basis
for numerous applications in resonant acoustic systems. Finally, the theory
for vibration isolation is explained as this has fundamental importance for
noise control of machines and equipment in practice. The description fol-
lows the terminology and concepts in ISO 2041.
Figure 2.1 shows a mechanical system consisting of a mass m placed on
an elastic layer on top of a rigid, unmovable surface. The elastic layer has
the stiffness k symbolized as a spring and the resistance r symbolized as
a dashpot. Some materials have slightly different stiffness for static load
and for a dynamic excitation, and thus, we will distinguish between those
cases. The acceleration due to gravity means that the system has a static
excitation, which leads to a compression of the elastic layer. The static dis-
placement is by definition:

mg
xs = (2.1)
ks

where g = 9.81 ms−2 is the acceleration due to gravity and ks is the static
stiffness (N/m).
In the case of a dynamic excitation of the system by the force F, we have
the equation of motion from Newton’s second law:

d2 x dx
m 2
+r + kd x = F (2.2)
dt dt

19
20  Sound Insulation in Buildings

(a) (b) (c) F

xs
m m·g m
x
k r ks r kd r

Figure 2.1 (a) Simple mechanical system. (b) Static deflection due to mass load and ­gravity.
(c) Dynamic excitation by an external force.

where x is the displacement from the rest position, t is the time and kd is the
dynamic stiffness (N/m).
As long as the external force is active, the system is said to display forced
vibrations. However, when the force stops, the system can still exhibit
vibrations for a shorter or longer time after the force is stopped. These are
resonant or free vibrations.
Let us assume that the external force is harmonic, i.e. it varies with time
as a sinusoidal function:

F = F0 cos(ω t + ϕ ) (2.3)

where F 0 is the amplitude of the force, ω is the angular frequency and φ

is the phase angle. Using the complex notation, the harmonic force can be
written as

F = F0 cos(ω t + ϕ ) + jF0 sin(ω t + ϕ ) = F0 e j(ω t +ϕ ) = Fˆ e jω t (2.4)

The force is expressed here as a complex phasor. F̂ is the complex amplitude,

and the magnitude of the phasor is the amplitude of the force, F = F0. The
advantage of the complex notation appears when the force is expressed as a
phasor, and we want to do multiplications, divisions, or other mathemati-
cal operations with harmonic signals. However, only the real part has a
physical reality, i.e. we can neglect the imaginary part when the force is
seen in relation to physical conditions.
As a consequence of the complex notation, the harmonic varying sig-
nal can be displayed in the complex plane, where the real and imaginary
parts are marked on two orthogonal axes. At a certain time, the signal
is displayed as a vector, starting from origin, and with a length equal to
the amplitude (the modulus). The direction of the vector is determined by
the argument, which is the angle relative to the real axis. The force in
Equation 2.4 can be displayed as shown in Figure 2.2.
The harmonically varying signal is represented in the complex plane as a
rotating vector, the phasor F̂. In the scientific literature, it is common not
to include the time factor ejωt, as it is understood.
 Mechanical vibrations  21

Imaginary

|F |

‹
F
ωt

φ
Real

Figure 2.2 A harmonically varying force displayed as a phasor in the complex plane.

2.2 MECHANICAL IMPEDANCE AND MOBILITY

Again, we consider the simple mechanical system in Figure 2.1. The system
is excited by the harmonic force F:

F = F ej(ωt+ϕ ) (2.5)

The vibrations exhibited by the system can be characterized by the deflec-
tion x, the velocity v, or the acceleration a. As long as the force is acting, the
system exhibits forced vibrations, implying that the vibrations have the same
angular frequency ω as the acting force. Thus, the velocity can be written as

v = vejωt (2.6)

where we have chosen to set the phase angle of the velocity to zero.
Mechanical impedance Zm of a system is defined as the complex ratio
between a harmonic force and the resulting velocity:

F F jϕ
Zm = = e (2.7)
v v

The unit is N s/m. The phase angle between force and velocity is φ.
The force and the velocity can be determined either in the same point
or in two different points. Therefore, we distinguish between the driving
point impedance and the transfer impedance. The latter is defined by

F1
Zm,12 = (2.8)
v2
22  Sound Insulation in Buildings

where F 1 is a harmonic force that excites the system in one point, and v 2 is
the resulting velocity in another point of the system, both taken as complex
quantities.
The mobility Ym of a mechanical system is the inverse of mechanical
impedance, i.e.

v v − jϕ
Ym = = e (2.9)
F F

The mobility is sometimes called mechanical admittance.
For a mass m, we assume a harmonic vibration (Equation 2.6) and get
the following relation from Newton’s second law:

dv
F=m = jω mv (2.10)
dt

Considering a spring with the dynamic stiffness kd , we get

F = kd x = kd ∫ v dt = kjωv
d
(2.11)

For a dashpot with resistance r, we get by definition

F = rv (2.12)

Table 2.1 shows the point impedance and mobility for the three basic
mechanical elements and for an extended system, namely, a large homo-
geneous plate.
An important system for applications in building acoustics is a large and
thin, homogeneous plate. This may be excited in a point by a harmonic
force in the normal direction perpendicular to the surface of the plate.
Bending waves are generated propagating radially in the plate outgoing
from the excitation point. The driving point impedance of the plate is found
to be surprisingly simple (Cremer and Heckl, 1967):

F0 4c 2m ′′
Z0 = = 8 m ′′B = (2.13)
v0 π fc

where m″ is the mass per unit area and B is the bending stiffness per unit
width. The critical frequency fc (Equation 1.28) is inserted in the last term.
It is noted that the point impedance is independent of frequency and real,
i.e. force and velocity are in phase. It can be shown that the impedance
(Equation 2.13) is also a good approximation for the average velocity due
to point excitation of a finite plate.
 Mechanical vibrations  23

Table 2.1  Examples of mechanical point impedance Zm and mobility Ym

Mechanical Mobility
System impedance Zm Ym
Mass: m (kg) jω m 1
jω m
Spring: kd (N/m) kd jω
jω kd

Dashpot: r (Ns/m) r 1
r

Plate 8 m′′B 1
Mass per unit area: 8 m′′B
m″ (kg/m2)
Bending stiffness:
B (Nm)

2.3 FREE VIBRATIONS

Free vibrations of the mechanical system are described through the solution
of the equation of movement (Equation 2.2) when the force is F = 0:

d2 x dx
m +r + kd x = 0 (2.14)
dt 2 dt

The solution of this homogeneous differential equation is well known from

mathematical physics. Three categories of solutions appear:

r < mkd : The system is weakly damped and can exhibit free harmonic
vibrations.
r = mkd : The system is critically damped.
r > mkd : The system is overdamped and a deflection slowly returns to the
steady-state position.

Only the first case is relevant for acoustical applications and is discussed
in the following.
The equation of movement (Equation 2.14) can be rewritten in the case
of a weakly damped system as

d2 x dx (2.15)
2
+ 2δ + ω 02 x = 0
dt dt
24  Sound Insulation in Buildings

where δ is the damping coefficient:

r
δ = (2.16)
2m

and ω 0 is the angular frequency of the equivalent undamped system (i.e.

for r = 0):

kd
ω0 = (2.17)
m

A solution to Equation 2.15 is

x = x0e−δ t cos(ω r t − θ ) (2.18)

where x0 and the phase angle θ depend on the initial conditions and the
angular frequency of the damped vibration is

ω r = ω 02 − δ 2 (2.19)

The time function of the vibration is shown in Figure 2.3. The phase angle
is set to zero. The amplitude of the vibration is x 0 e−δt, and thus, it decreases
exponentially with time. It can be noted from Equation 2.19 that the angu-
lar frequency of the system, ωr, is lower than the angular frequency of the
undamped system ω 0.

1
Deflection x / x0

–1
Time t

Figure 2.3 Time function of the free vibration of a weakly damped system with one
degree of freedom.
 Mechanical vibrations  25

The system is said to exhibit a natural mode of vibration, and the associ-
ated frequency is called the natural frequency fr = ωr/2π.
The undamped natural frequency f 0 = ω 0/2π is often referred to as the
resonance frequency. However, this is only correct if a velocity resonance
is understood (see more about this in Section 2.5). The undamped natural
frequency is

ω0 1 kd
f0 = = (Hz) (2.20)
2π 2π m

The period duration of the damped natural vibration is the reciprocal of

the natural frequency:

1 2π 2π
T1 = = = (s) (2.21)
fr ω r ω 02 − δ 2

From Equation 2.18, it can be seen that the ratio between deflections x
at time t and t + T1 is e δT 1. The natural logarithm of this ratio is called the
logarithmic decrement Λ:

x(t)
Λ = ln = δ T1 (2.22)
x(t + T1)

The damped vibrations can also be characterized by the time constant τ

defined as the reciprocal of the damping coefficient:

1
τ= (s) (2.23)
δ

The energy of the vibration is related to the velocity v, which is easily found
when Equation 2.18 is written in the complex notation:

x = x0e−δ t ejω rt e− jθ (2.24)

dx (2.25)
v= = (jω r − δ )x0e−δ t ejω rt e− jθ
dt

v = ω r2 + δ 2 x0e−δ t = ω 0 x0e−δ t (2.26)

The energy as a function of time is

1 2 1 (2.27)
E= m v = mω 02 x02e−2δ t = E0e−2δ t
2 2
26  Sound Insulation in Buildings

where E 0 is the initial energy at t = 0.

The loss factor η is the most used descriptor of the damping of vibrating sys-
tems in building acoustics. It is defined as the relative energy loss per radian:

−dE / dt 2δ r r
η= = = = (2.28)
ω 0E ω0 mkd ω 0m

The reverberation time T is primarily used in room acoustics, but it may
also be used in relation to the decay of free vibrations. By definition, T is
the time for the energy to decrease to 10 −6 of the initial energy, which cor-
responds to a decay of 60 dB. For t = T, we get from Equation 2.27

E
= e−2δ T = 10−6 = e−6 ln(10) (2.29)
E0

and the reverberation time is

6 ln(10) 6.9 2.2

T= ≈ ≈ (2.30)
2δ δ f0η

The latter relation is very useful for experimental determination of the

loss factor at the resonance frequency – this can be done by measuring the
reverberation time of the system.

2.4 FORCED VIBRATIONS

We now consider the situation where the mechanical system is excited by a

harmonic force F = F̂ejωt , which means that there are forced vibrations with
the same angular frequency as the force, i.e. the deflection is x = x̂ ejωt . By
insertion in the equation of movement (Equation 2.2), and since

dx d2 x
= jω x and = −ω 2 x (2.31)
dt dt 2
we get

Fˆ Fˆ
xˆ = = (2.32)
−ω m + jω r + kd jω r + j(ω m − kd / ω)
2

The velocity of the vibration is

dx F
v= = jω xˆejωt = (2.33)
dt r + j(ω m − kd / ω)

 Mechanical vibrations  27

From the definition of mechanical impedance Zm (Equation 2.7), we get

F
Zm = = r + j(ω m − kd / ω) (2.34)
v
It may be noted that this formula is analogous to Ohm’s law for a simple
electrical circuit.
The modulus and argument of the mechanical impedance are

Zm = r 2 + (ω m − kd / ω )2 (2.35)

 ω m − kd / ω 
θ = arctan   (2.36)
 r

where Zm = |Zm|ejθ.
Applying the mechanical impedance of the system, we get the following
formulas for the force, velocity and deflection in complex notation:

F = F ejωt (2.37)

F F j(ωt −θ )
v= = e (2.38)
Zm Zm

F F
x= = ej(ωt −θ −π /2) (2.39)
jω Zm ω Zm

Here we have chosen the force as a reference for the phase angle. Figure 2.4
shows these three parameters as phasors in the complex plane. In the physi-
cal world, the parameters appear as projections on the real axis with har-
monic variation (the exponential functions are replaced by cosine functions).
The displacement x can be found from Equation 2.39 by inserting
Equation 2.34 and using Equations 2.17 and 2.28:

F F / kd
x= = (2.40)
m 2  ω2 

( 2 
jω  r + j ω − ω 0  jη ⋅
 ω 
)ω
+ 1 − 2 
ω0  ω0 

This leads us to the following frequency dependency of the displacement

amplitude relative to the amplitude of the force:
x 1 / kd
= (2.41)
F 2 2
 ω   ω2 
 η ⋅ ω  +  1 − ω 2 
0 0

28  Sound Insulation in Buildings

Imaginary

|F|

θ |F|
v=
ωt |Zm|

Real
π
2

|F |
x=
ω|Zm|

Figure 2.4 Display of force, velocity and displacement as vectors in the complex plane.
The vectors rotate with the angular frequency ω, but the relative positions remain
constant.

The phase angle between the force and deflection is

π   ω 0 ω  −1 
θ + = arctan  η ⋅  − (2.42)
 
2   ω ω0  

Figure 2.5 shows the amplitude and phase of the displacement as functions
of frequency for different values of the loss factor. (Note that the relative
angular frequency and the relative frequency are identical, i.e. ω/ω 0 = f/f 0.)
At low frequencies (ω ≪ ω 0), the deflection amplitude is almost indepen-
dent of the frequency and the phase angle is close to zero, which means that
the deflection follows the force without delay. The mechanical impedance is
dominated by the stiffness term |Zm| ≈ kd /ω.
At the resonance frequency (ω = ω 0), the deflection amplitude increases
dramatically for small loss factors, and more moderately for large loss fac-
tors. The phase angle is π/2 and the deflection is one-fourth period behind
the force, while the velocity is in phase with the force. The mechanical
impedance is dominated by the damping term |Zm| ≈ r.
At high frequencies (ω ≫ ω 0), the deflection amplitude decreases rapidly
with frequency and the phase angle is close to π, which means that deflec-
tion and the force are almost in opposite phase. The velocity is behind
the force and the acceleration is in phase with the force. The mechanical
impedance is dominated by the mass term |Zm| ≈ ωm.
 Mechanical vibrations  29

10

0,1
(x/F) kd

1
0,2
0,5
1

0.1
0.1 1 10
f /f0

3
Phase angle, θ + π/2 (rad)

2
0,1
0,2
0,5
1
1

0
0.1 1 10
f /f 0

Figure 2.5 Amplitude of deflection (upper graph) and phase angle between deflection
and force (lower graph) for different values of the loss factor.

The described behaviour of the system is typical for a resonant system.

When the system is excited with a frequency close to the natural frequency
of the system, the amplitude of the deflection is significantly larger than at
other frequencies.
30  Sound Insulation in Buildings

2.5 RESONANCE

Resonance in a system is defined as a state in which any change in excita-

tion frequency results in a decreased response. Thus, the resonance fre-
quency depends on what kind of response is observed – deflection, velocity
or acceleration.

2.5.1 Resonance frequency
If we consider the deflection of the system, it follows from Equation 2.40
that the maximum deflection occurs when the denominator

2 2
 ω   ω2 
η ⋅ +
 ω   1 −
0 ω 02 

is minimum, which happens for the angular resonance frequency:

1
ω res, x = ω 0 1 − η 2 = ω 02 − 2δ 2 (2.43)
2

The maximum deflection at resonance is

F F
x max = = (2.44)
1 r ω −δ2
2
k d η 1 − η2 0

However, if we instead chose to consider the velocity of the system, inser-

tion of Equations 2.17 and 2.28 in Equation 2.33 yields

F F / mkd
v= = (2.45)
ω0mη + jm / ω (ω 2 − ω02 ) η + j(ω / ω0 − ω0 / ω)

From this, it is obvious that the angular resonance frequency for velocity
resonance is

ω res, v = ω 0 (2.46)

and the corresponding maximum velocity is

F F
v max = = (2.47)
η mkd r

 Mechanical vibrations  31

Table 2.2  N
 atural frequencies and resonance frequencies of a simple
mechanical system with one degree of freedom
Natural frequency
Undamped Damped
Angular kd ω r = ω 02 − δ 2
frequency ω0 =
m
(rad/s)
1 kd η2
Frequency (Hz) f0 = fr = f0 1 −
2π m 4

Resonance frequency
Deflection Velocity Acceleration
Angular ω 02 − 2δ 2 ω0 ω 02
frequency
(rad/s) ω − 2δ 2
2
0

η2 f0
Frequency (Hz) f0 1 − f0
2 η2
1−
2

Finally, if we chose to consider the acceleration of the system, we find in

a similar way that the maximum acceleration occurs at the angular reso-
nance frequency for acceleration:

ω0 ω0
ω res, a = = (2.48)
1 ω 0 − 2δ 2
2
1 − η2
2

An overview of the various resonance frequencies and natural frequencies

is found in Table 2.2.
It is particularly interesting to note that the resonance frequency for
velocity is identical to the natural frequency of the undamped system,
ωres,v = ω 0. This is a very convenient coincidence.

2.5.2 Bandwidth of a resonant system

The mechanical energy of a vibrating system is (½)mv 2 and thus propor-
tional to the velocity squared. The ratio between this and the force squared
is from Equation 2.45:

2
v 1 / (mkd ) 1 / (mkd )
2 = = (2.49)
F η 2 + (ω / ω 0 − ω 0 / ω )2 η 2 + (f / f0 − f0 / f )2

32  Sound Insulation in Buildings

In this section, the frequency f in Hz is used instead of the angular fre-

quency. Equation 2.49 is illustrated as a function of the relative frequency
in Figure 2.6. The maximum is at the velocity resonance, i.e. at the fre-
quency fres,v = f 0.
Resonance covers a certain frequency range and in order to characterize
this, we introduce the half-power bandwidth Br defined as the frequency
interval within which the energy is at least half the maximum energy at reso-
nance. If f1 and f 2 are the lower and upper limiting frequencies, respectively,
it can be seen from Equation 2.49 that they can be found as solutions to

η 2 = (f / f0 − f0 / f )2

1 1 2 2 2
⇒ f = ± η f0 ± η f 0 +f 0
2 4

Since we accept only positive solutions, we get

1
f1 = fm − η f0
2
(2.50)
1
f2 = f m + η f0
2
where the mean frequency fm is

1 1
fm =
2
( f1 + f2 ) = f0 1 + η 2
4
(2.51)

1
Relative |v|2

Br = η f 0
0.5

0
f1 f0 f2
Frequency

Figure 2.6 The frequency course of energy in a resonant system.

 Mechanical vibrations  33

and the half-power bandwidth is

Br = f2 − f1 = η f0 (2.52)

However, instead of the arithmetic average frequency, a frequency band is

usually described by the centre frequency fc , which is calculated as the geo-
metrical average of the limiting frequencies:

1 2 2
fc = f1f2 = fm2 − η f0 = f0 (2.53)
4

Hence, the centre frequency is identical to the resonance frequency for

velocity, which again is the same as the natural frequency of the undamped
system.
By means of a frequency generator, we can find the resonance frequency
and the bandwidth. Thus, it is possible to apply Equation 2.52 for the mea-
surement of the loss factor in a weakly damped system.

2.6 VIBRATION INSULATION

An important practical application of the mechanical vibration model is

the vibration isolation of machines and other equipment from the building
construction as a means of noise and vibration control.

2.6.1 The power input

When the force F excites a mechanical system, the power input P is defined
as the time-average of the power multiplied by the velocity at the point of
excitation, measured in the same direction as the force:

T1
1
P = v⋅F =
T1 ∫ v ⋅ Fdt (2.54)
0

where T1 = 1/f = 2π/ω here is the period duration of the forced vibration. The
velocity and the force are assumed to be harmonic functions and, thus, the
effective RMS (root-mean-square) values and the amplitudes are related as

2 1 2 1 2
veff = v and Feff2 = F
2 2 .

Inserting the real part of v and F from Equations 2.37 and 2.38 yields
34  Sound Insulation in Buildings

T1

P=
1
T1 ∫ v ⋅ cos(ωt − θ) ⋅ F ⋅ cos(ωt)dt
0

T1

=v⋅F
1
T1 ∫ 12 [cos(2ωt − θ) + cosθ ] dt
0
(2.55)

1
= v ⋅ F cos θ
2
Note the importance of the phase angle θ.
The same derivation can be made by using complex notation:
1
P=
2
v ⋅ F ⋅ Re ejθ { }
1
=
2
{
⋅ Re v e− jω t ejθ ⋅ F ⋅ ejω t } (2.56)

1
= ⋅ Re {v* ⋅F }
2
where v* is the complex conjugate of v (see Equation 1.54).The power
input can be related to the velocity at the excitation point by means of the
mechanical point impedance Zm (Equation 2.7):

1 v  1 2 F 
P= 2
⋅ Re  v * ⋅F  = v ⋅ Re   = veff ⋅ Re {Zm } (2.57)
2 v  2 v 
In a similar way, the power input can be related to the excitation force and
the mobility Ym (Equation 2.9):

1 2 v 
P= F ⋅ Re   = Feff2 ⋅ Re {Ym } (2.58)
2 F 
It may be noted that the following relation exists:

 1  Re {Zm }
Re {Ym } = Re  = 2 (2.59)
 Zm  Zm

2.6.2 Force transmitted to the foundation

Figure 2.7 shows a machine that excites the mechanical vibration system
with a harmonic force F:

F = F ejωt = Zm v = ( r + j(ω m − kd ω)) v (2.60)

 Mechanical vibrations  35

m v

k r
F΄

Figure 2.7 Principle of a machinery supported on vibration isolators.

where ω is the angular frequency of the excitation force, Zm is the mechan-

ical impedance (Equation 2.34) and v is the velocity of the mechanical
vibration system. The vibrating mass m may be that of the machine, but the
mass is often increased by attaching the machine rigidly to a heavy plate. A
heavy mass is an advantage for setting up good vibration isolation.
The force Fʹ that is transferred to the supporting building structure is
determined by the excitation of the spring and the dashpot:

dx
F′ = r + kd x = ( r − j kd ω ) v (2.61)
dt

From this, we define the transmissibility H of the system as the ratio

between the transferred force and the excitation force:

F′ r − jkd / ω
H= = (2.62)
F r + j(ω m − kd ω)

Introducing the natural frequency ω 0 (Equation 2.17) and the loss factor η
(Equation 2.28), yields

η − jω0 / ω
H= (2.63)
η + j(ω / ω0 − ω0 ω)

Figure 2.8 displays the numerical value of the transmissibility, i.e. the ratio
between the amplitude of the forces expressed in dB, as a function of the
relative frequency for various values of the loss factor. Note that ω/ω 0 = f/f 0.
The transmissibility in dB is calculated from

η 2 + (ω 0 / ω )
2
F′ 2
20 lg H = 10 lg = 10 lg (2.64)
F2 η 2 + (ω / ω 0 − ω 0 ω )2

At low frequencies, i.e. for ω ≪ ω 0, the vibration isolator is stiff and the force
is transmitted to the floor without attenuation. If the excitation frequency
36  Sound Insulation in Buildings

ω is near ω 0, the transmitted force is stronger than the excitation force. The
amplification is due to resonance, and if the loss factor is small, the ampli-
fication can grow very high. Thus, the frequency range near the resonance
frequency should always be avoided.
Only when the frequency ω > 2ω 0, does the attenuation start. It is vital
that vibration isolators are designed to give a resonance frequency well
below the lowest excitation frequency, and a factor of three may be a practi-
cal minimum (ω ≈ 3ω 0) to obtain an efficient isolation.
If the loss factor is small and η ≪ ω 0/ω ≪ 1, the approximate transmis-
sibility is

( )
−2
≅ (ω 0 / ω )
4
H 2 ≅ 1 − (ω / ω 0 )2 (2.65)

This means a slope of –12 dB per octave. However, at very high frequen-
cies, ω 0/ω ≪ η, the slope changes to –6 dB per octave, which follows from
the approximation:

40

20
Transmissibility, 20 log |H| (dB)

–20

–40

–60

–80
0.1 1 10 100
Relative frequency, f /f0

η = 0,03 η = 0,1 η = 0,3 η=1

Figure 2.8 The numerical value of the transmissibility with viscous damping, shown as a
function of the relative frequency for four different values of the loss factor.
 Mechanical vibrations  37

2
η2  ηω 0 
H 2≅ ≅ (2.66)
η + (ω / ω 0 )
2 2
 ω 

In practice, attenuation at high frequencies is limited due to wave transmis-

sion through the isolator, and high-frequency resonances in the isolator
may occur.

2.6.3 Vibration isolators with hysteretic damping

The viscous damping (dashpot type) assumed in the previous section is
usual in cars and other places, but not in buildings. Common vibration
isolators made of rubber have the losses inherent in the spring, and the
dynamic stiffness is complex: kd (1 + jη). This type of damping is called hys-
teretic damping, and the losses are proportional to the deflection, not the
velocity. Inserting the complex stiffness and r = 0 in Equation 2.62 yields
the transmissibility of a system with hysteretic damping:

−jkd (1 + jη) / ω η−j

H= = (2.67)

(
j(ω m − kd (1 + jη) ω) η + j (ω ω0 )2 − 1
)
This leads to the transmissibility in dB:

η2 + 1
20 lg H = 10 lg (2.68)
( )
2
η 2 + (ω / ω 0 ) − 1
2

This is displayed in Figure 2.9. The main difference between this and vis-
cous damping is that the damping above the resonance frequency is practi-
cally independent of the loss factor. The slope is –12 dB per octave.

2.6.4 The Q-factor
When a resonant system is excited with a frequency near the resonance
frequency f 0, the system will cause an amplification instead of an attenua-
tion. The Q-factor is defined for a lightly damped system as a measure of
the sharpness of the resonance peak as

f0 1
Q= = (2.69)
Br η

Hence, the Q-factor is just the reciprocal of the loss factor.

For vibration isolators, the resonance peak in the transmissibility is
found from Equation 2.64 or 2.68 by setting ω = ω 0:
38  Sound Insulation in Buildings

40

20
Transmissibility, 20 log |H| (dB)

–20

–40

–60

–80
0.1 1 10 100
Relative frequency, f /f0

η = 0,03 η = 0,1 η = 0,3 η =1

Figure 2.9 The numerical value of the transmissibility with hysteretic damping, shown as
a function of the relative frequency for four different values of the loss factor.

 1
20 lg H = 10 lg  1 + 2  = 10 lg 1 + Q2 ≈ 20 lg Q
 η 
( ) (2.70)

2.6.5 Insertion loss
The efficiency of vibration isolators is not solely determined by the trans-
missibility of the vibrating system. The supporting floor may also influence
the result. Hence, a better measure for the efficiency of vibration isolators
is the insertion loss, defined by

P
∆L = 10lg (dB) (2.71)
P′
where P and P′ refer to the power transferred to the floor without or with
the vibration isolator, respectively (Figure 2.10).
The point impedance of the floor is called Zm and the velocity generated
in the floor is v or vʹ without or with the vibration isolator, respectively.
From Equation 2.57, we get
 Mechanical vibrations  39

(a) (b)
F

m v
F kd r
F΄
m
vu v΄

Figure 2.10 Vibrating machinery on a supporting floor (a) without vibration isolators,

(b) with vibration isolators.

2
2
veff ⋅ Re {Zm } v
∆L = 10lg = 10lg 2 (dB) (2.72)
v′eff ⋅ Re {Zm }
2
v′

The insertion loss is simply the difference between the velocity levels mea-
sured on the floor without and with the vibration isolator.
The force F acts against the point impedance, which is Zm + jωm without
the vibration isolator, where m is the total mass of the machine. With refer-
ence to Figure 2.10, we have

F F′
= Zm + jω m , = Zm (2.73)
v v′

Insertion in Equation 2.72 yields

2 2
F ⋅ Zm jω m
∆L = 10lg 2 2
= −20lg H − 20lg 1 + (dB) (2.74)
F′ ⋅ Zm + jω m Zm

where H is the transmissibility. If the point impedance of the floor is very

large, |Zm| ≫ ωm, the last term in Equation 2.74 is negligible. Thus, the
insertion loss is, with a good approximation, described by the graphs
shown previously in Figure 2.9. If the floor is a homogeneous plate, e.g. a
concrete slab, the point impedance can be taken from Table 2.1, and we get


(ω m)  (dB)
2

∆L = −20lg H − 10lg 1 + (2.75)

 64m′′B 
 

where m″ is the mass per unit area (kg/m 2) and B is the bending stiff-
ness (Nm) of the floor. Obviously, the efficiency of the vibration isolators
depends on whether the mass and stiffness of the floor are sufficient to pro-
vide a reasonable withhold against the isolators. In order to insulate against
40  Sound Insulation in Buildings

(a) (b) (c)

Figure 2.11 Vibration isolation of a machinery: (a) Insufficient resistance from the floor,
(b) increased mass of floor, (c) increased stiffness of floor. Drawing by Bernt
Forsblad. (Reproduced with permission from Arbetarskyddsnämnden, Buller
bekämpning. Principper och tillämpning. (Noise abatement. Principles and
applications, in Swedish), Arbetarskyddsfonden, Sweden, 1977.)

large forces, it may be necessary to take special steps in order to increase

the mass or the stiffness of the supporting floor as illustrated in Figure 2.11.

2.6.6 Insulation against structure-borne vibrations

Vibration insulation is not only applied in cases of vibrating equipment and
machines. It is equally important in cases of sensitive equipment that must
be protected against vibrations from the environment. Examples include
delicate equipment in laboratories or hospitals, and rooms that need very
low background noise levels like concert halls and recording studios.
In principle, the vibrating system in such situations is not different from
the one discussed in the previous sections, except that the excitation of the
system is caused by vibrations in the supporting floor (Figure 2.12). The
velocity ve causes a vertical force that reacts on the spring and the dashpot
of the system, and this, in turn, causes the mass of the equipment m to
vibrate with the velocity v. Assuming harmonic vibrations, i.e. d 2 x/dt 2 = jωv
and x = v/jω, application of Newton’s second law (Equation 2.2) leads to

m v

kd r
ve

Figure 2.12 Principle of a mass that is isolated from vibrations in the floor.

 Mechanical vibrations  41

ve − v
r ( ve − v ) + kd = jω mv (2.76)
jω

leading to

v r − jkd / ω
= =H (2.77)
ve r + j (ω m − kd / ω )

Comparing this with Equation 2.62, we realize that the transmissibility H
of the system describes the vibration insulation in this case too. The only dif-
ference being that H is now a ratio of velocities, instead of a ratio of forces.
An example is the large anechoic chamber at the Technical University of
Denmark, DTU. For such a facility, it is crucial that the insulation against
noise and vibrations from the environment is as good as possible. The
building is a box-in-box system, where the inner box is made from 40 cm
thick concrete. The total mass is 1200 tons supported by 24 vibration iso-
lators made of rubber. Six vibration isolators are placed near each corner
(Figure 2.13).
To check the efficiency, vertical vibrations were measured simultane-
ously below and above the vibration isolators. The accelerometers were
mounted on small steel plates, which can be seen in Figure 2.13. The source
of vibrations was pile driving for the foundation of a neighbour building.
The measured difference between the velocity levels is shown in Figure 2.14
as a function of the frequency. The resonance of the system appears as an
amplification around 8 Hz; but at frequencies above 20 Hz, the attenuation
is better than 10 dB.

Figure 2.13 Vibration isolators of rubber under the large anechoic chamber at DTU.
(Photo by J. H. Rindel.)
42  Sound Insulation in Buildings

10

0
Transmissibility (dB)

–10

–20

–30
1 10 100
Frequency (Hz)

Figure 2.14 Measured attenuation of vibrations in the vertical direction obtained by the

vibration isolators under the large anechoic chamber at DTU.

2.6.7 Design of vibration isolators

For the practical design of vibration isolators, the first step is to identify the
lowest excitation frequency f that should be attenuated. If the source is an
engine with rotating parts, e.g. an electric motor, and the number of rota-
tions per minute is n (min−1), the lowest excitation frequency is

n
f = (Hz) (2.78)
60

The resonance frequency of the system can then be decided. If there are no
detailed requirements, we may apply the simple criterion:

f
f0 ≅ (Hz) (2.79)
3

From Figure 2.9 it can be seen that at frequency 3f 0, the attenuation is

around 15 dB to 18 dB.
We shall now recapitulate Equation 2.1 for the static load and Equation
2.17 for the dynamic vibration. Combining these equations leads to

1 kd g
f0 = (Hz) (2.80)
2π ks xs

where
g is the gravity acceleration (9.81 m/s2)
kd is the dynamic stiffness (N/m)
 Mechanical vibrations  43

ks is the static stiffness (N/m)

xs is the static deflection (m)
The ratio of dynamic stiffness and static stiffness depends on the actual
construction and material of the vibration isolator. A typical value of kd /
ks ≈ 1.3 is applicable for many rubber-based isolators. Insertion of this and
the value of g yield the simple relationship between resonance frequency
and static deflection:

1
f0 ≅ (Hz) (2.81)
3xs

Applying the design criterion (Equation 2.79) leads to the very simple
design guide:

1 3
xs ≅ ≅ (m) (2.82)
3f02 f 2

Figure 2.15 is a graphical representation of Equation 2.82. With this graph,

it is easy to find the necessary static deflection from the known frequency of

100.0
Static deflection (mm)

10.0

1.0

0.1
1 10 100
Frequency of excitation (Hz)
Stiff 0 dB 10 dB 20 dB 30 dB

Figure 2.15 Diagram for the design of vibration isolators. The necessary static deflection
is determined from the excitation frequency and the attenuation in dB. The
shaded area is the resonance region and must be avoided. Example: excita-
tion at 24 Hz, 20 dB attenuation requires 6 mm static deflection.
44  Sound Insulation in Buildings

Figure 2.16 Water pump and electric motor mounted on a concrete block, which is
supported by four vibration isolators (not visible on the photo). (Photo by
J. H. Rindel.)

excitation and the wanted amount of attenuation. The resonance region as

clearly marked and should always be avoided. The example shown in Figure
2.15 is for a water pump driven by an electric motor with 1450 turns per min-
ute; that gives f ≈ 24 Hz, and 20 dB attenuation would then require xs = 6 mm.
A photo of this water pump is seen in Figure 2.16. In this case, vibra-
tions cannot be tolerated in the building, so the vibration isolation had to
be extremely good. Several practical measures were taken to obtain this.
First, the vibrating equipment was mounted on a big, heavy concrete block;
this means that the velocity created by the equipment is reduced to a mini-
mum, even before the vibration isolators. Figure 2.16 also shows that all
tubes and connections to and from the equipment are made flexible – this
is vital to avoid any rigid connections between the vibrating equipment and
building.

2.7 HUMAN RESPONSE TO VIBRATIONS

The human body can sense vibrations in the frequency range from 1 Hz
to 80 Hz. The sensitivity depends on the direction of vibration relative
to the body. Hence, in a standing position, vertical vibrations are sensed
 Mechanical vibrations  45

particularly well for accelerations in the range from 4 Hz to 8 Hz, while

in a lying position, the body is most sensitive in the frequency range from
1 Hz to 2 Hz. Thus, it is common practice to apply a combined sensitivity
curve (Figure 2.17).
The base curves represent curves of equal human response with respect
to annoyance or complaints. The base curves can be considered the approx-
imate lower threshold for vibrations, i.e. no adverse sensations or com-
plaints are expected below these curves.
Vibrations are measured either with an accelerometer or with a geo-
phone. While the former responds to acceleration and is commonly used in
acoustics, the latter converts velocity into voltage and is commonly used in
seismology and geo-techniques.
It is convenient to apply a frequency weighting filter for measurements
(Figure 2.18). These filters approximately mirror the base curves for human
sensitivity. This is analogous to using the A-weighting filter for measuring
sound pressure levels.

1 120
Acceleration (RMS), m/s2

0.1 100 La, dB re 10–6 m/s2

0.01 80

0.001 60
1 2 4 8 16 32 63
Frequency (Hz)
Vertical Horizontal Combined
10 140
Velocity (RMS), mm/s

Lv, dB re 10–9m/s

1 120

0.1 100

0.01 80
1 2 4 8 16 32 63
Frequency (Hz)
Vertical Horizontal Combined

Figure 2.17 Base curves for building vibrations: (top): acceleration; (bottom): velocity.
(Adapted from ISO 2631-2, Evaluation of human exposure to whole-body vibra-
tion – Part 2: Continuous and shock-induced vibration in buildings (1 to 80 Hz),
International Organization for Standardization, Geneva, Switzerland, 1989.)
46  Sound Insulation in Buildings

–10
Weighting (dB)

–20

–30

–40
0.1 1 10 100 1000
Frequency (Hz)
Acceleration, Wm, dB Velocity, Wmv, dB

Figure 2.18 Frequency weighting applied to vibration measurements, acceleration and

velocity. (Adapted from ISO 2631-2, Evaluation of human exposure to whole-
body vibration – Part 2: Continuous and shock-induced vibration in buildings (1 to
80 Hz). International Organization for Standardization, Geneva, Switzerland,
2003; NS 8176, Vibration and shock – Measurement of vibration in buildings from
land based transport and guidance to evaluation of effects on human beings. (In
Norwegian, 2nd edition, Standard Norge, Lysaker, Norway, 2005.)

An interesting feature of the frequency weighting filters is that there is

a simple relationship between the weighted acceleration and the weighted
velocity:

v w ≅ k ⋅ aw (m/s) (2.83)

where k = 0.028 s is a constant. Similarly, for the frequency weighted vibra-

tion levels, the relationship is

Lv , w ≅ La, w + 29 dB (2.84)

Without the weighting filters, the relationship is frequency dependent, a = jωv.

2.7.1 Examples of evaluation criteria

The threshold for whole-body sensation of vibrations are, according to
the curves in Figure 2.17, approximately a weighted acceleration level of
71 dB re 1 μm/s2 (RMS acceleration 3.6 mm/s2). Using the velocity level, the
threshold is 100 dB re 1 nm/s (RMS velocity 0.1 mm/s). This may be used as
an evaluation criterion for critical areas, e.g. hospital operation theatres or
precision laboratories. Other examples are shown in Table 2.3.
 Mechanical vibrations  47

Table 2.3  Examples of evaluation criteria in terms of weighted acceleration level

or weighted velocity level
La,w dB re 1 μm/s2 Lv,w dB re 1 nm/s
Place Continuous Transient Continuous Transient
Critical working areas 71 71 100 100
Residential, day 74–83 107–110 103–112 136–139
Residential, night 74 74–97 103 103–126
Office 83 107–113 112 136–142
Workshop 89 107–113 118 136–142
Source: ISO 2631-2, Evaluation of human exposure to whole-body vibration – Part 2:
Continuous and shock-induced vibration in buildings (1 to 80 Hz). International
Organization for Standardization, Geneva, Switzerland, 1989.

REFERENCES

Arbetarskyddsnämnden (1977). Buller bekämpning. Principper och tillämpning (Noise

abatement. Principles and applications, in Swedish). Arbetarskyddsfonden,
Stockholm, Sweden.
L. Cremer, and M. Heckl (1967). Körperschall. (Structure-borne sound, translated by
E.E. Ungar, 1973). Springer, Berlin, Germany.
ISO 2041 (2009). Mechanical vibration, shock and condition monitoring – Vocabulary.
International Organization for Standardization, Geneva, Switzerland.
ISO 2631-2 (1989). Evaluation of human exposure to whole-body vibration –
Part 2: Continuous and shock-induced vibration in buildings (1 to 80 Hz).
International Organization for Standardization, Geneva, Switzerland.
ISO 2631-2 (2003). Mechanical vibration and shock – Evaluation of human expo-
sure to whole-body vibration – Part 2: Vibration in buildings (1 to 80 Hz).
International Organization for Standardization, Geneva, Switzerland.
NS 8176 (2005). Vibration and shock – Measurement of vibration in buildings from
land based transport and guidance to evaluation of effects on human beings.
(In Norwegian), 2nd edition. Standard Norge, Lysaker, Norway.
NT ACOU 082 (1991). Buildings – Vibration and shock, evaluation of annoyance.
Nordtest, Espoo, Finland. ISSN 0283-7145.
Chapter 3

The sound field in front of a wall

3.1 NORMAL INCIDENCE OF A PLANE WAVE

We will consider the sound field in front of a wall, which is assumed to be

perfectly plane, smooth and infinitely large. However, the results will also
be valid approximations for walls encountered in practice as long as (1)
their dimensions are much larger than the wavelength and (2) any irregu-
larities in the surface are much smaller than the wavelength of the sound.
First, the case of normal incidence is considered, i.e. the wall is perpendicu-
lar to the direction of propagation of the incoming sound wave, which we
assume to be a single harmonic sound with angular frequency ω. We place the
x-axis in the direction of propagation, so that x = 0 at the surface (Figure 3.1).
The sound pressure pi and the particle velocity ui in the incoming sound
wave can be written as

pi = pi ej(ωt −kx) (3.1a)

pi j(ωt −kx)
ui = e (3.1b)
ρc

For a harmonic sound wave, the particle velocity component along a given
axis direction x can be obtained through partial differentiation of the sound
pressure:

∂Φ 1 ∂p
ux = − =− ⋅ (3.2)
∂x jωρ ∂x

with Φ = Φ ejωt being the velocity potential, and the sound pressure being

∂Φ
p=ρ = jωρ ⋅ Φ (3.3)
∂t

49
50  Sound Insulation in Buildings

ur

pr
x
pi
v
ui

x=0

Figure 3.1 Reflection of a plane wave at normal incidence.

Normally, the reflected sound wave will have lower amplitude and a dif-
ferent phase compared to the incoming sound wave. Both can be consid-
ered by introducing a complex reflection factor, R. Consequently, the sound
pressure and particle velocity of the reflected wave can be written as

pr = R ⋅ pi ej(ωt+kx) (3.4a)

pi j(ωt+kx)
ur = −R ⋅ e (3.4b)
ρc

Notice that the change in direction of propagation implies a shift in the

sign of kx. Also, the sign of the particle velocity has changed compared to
that of the incoming sound wave in Equation 3.1b. The sound pressure and
particle velocity in the combined sound field in front of the wall are found
by adding the expressions in Equations 3.1 and 3.4:

( )
p = pi ⋅ e− jkx + R ⋅ ejkx ejωt (3.5a)

pi
u=
ρc
( )
⋅ e− jkx − R ⋅ ejkx ejωt (3.5b)

The particle velocity at the wall surface is perpendicular to the wall and
equal to the normal component of the wall velocity: v. Setting x = 0 in
Equation 3.5b yields

pi
v = u(x=0) = ⋅ (1 − R) ejωt (3.6)
ρc
 The sound field in front of a wall  51

If the wall is sufficiently hard and heavy, v = 0 resulting in R = 1, and the
wall is said to be totally reflecting. In front of such a wall, we can find p
and u from Equation 3.5 as

p = 2 pi ⋅ cos (kx) ejωt (3.7a)

pi
u = −2j ⋅ sin (kx) ejωt (3.7b)
ρc

Notice that none of these expressions contain a term of the form ejkx, i.e.
the combined wave is not propagating; we have a ‘standing wave’ in front
of the wall.
The energy density in the sound field is distributed between a potential
energy density and a kinetic energy density given by

w = wpot + wkin (3.8)

where

2
1 p 2 1 p (3.9a)
wpot = ⋅ 2 = ⋅ 2
2 ρc 4 ρc

1 1
⋅ ρ u 2 = ⋅ ρ u (3.9b)
2
wkin =
2 4

These energy densities are proportional to the amplitudes of the pressure

squared and the velocity squared, respectively. For the incident sound field,
which is a plane propagating wave, we have

2 2 2
1 pi 1 pi 1 pi (3.10)
wi = ⋅ 2 + ⋅ ρ = ⋅ 2
4 ρc 4 (ρ c) 2
2 ρc

In the standing wave in front of the wall, we have from Equation 3.7:

2 2 2
p = 4 pi ⋅ cos2 (kx) = 2 pi ⋅ (1 + cos(2kx)) (3.11a)

2 2
2 pi 2
pi
u =4 ⋅ sin (kx) = 2 ⋅ (1 − cos(2kx)) (3.11b)

( ρ c )2 ( ρ c )2

From this it can be seen that the potential energy has a maximum and the
kinetic energy has a minimum (zero) when 2kx = n2π, i.e. for x = nλ/2 where
52  Sound Insulation in Buildings

n = 0, 1, 2, 3, etc. The kinetic energy is maximum and the potential energy
is minimum (zero) when 2kx = π + n2π, i.e. for x = λ/4 + nλ/2. These are the
characteristic nodal points and maxima in a standing wave in front of a
wall.
The total energy density is found by inserting Equation 3.11 in Equations
3.9 and 3.8:

2 2 2
1 pi 1 pi pi
w= ⋅
4 ρc2
(
2 1 + cos ( 2kx ) + ⋅ ρ )
4 ( ρ c )2
2 1 − cos ( 2kx ) =
ρc2
(
(3.12) )

It appears that the total energy density is independent of the distance from
the wall. By comparison with Equation 3.10, it can also be seen that the
energy density is simply doubled compared to that of the incoming plane
wave alone.

3.2 OBLIQUE INCIDENCE OF A PLANE WAVE

We now assume that the angle between the direction of propagation of

the plane sound wave and the normal of the wall equals θ. We also assume
the direction of propagation to be in the xy plane as shown in Figure 3.2.
This corresponds to a transformation of the coordinates in which x in
Equation 3.1 is substituted by x cos θ + y sin θ:

pi = pi e− jk(x cos θ +y sin θ )ejωt (3.13a)

pi
uix = cos θ ⋅ e− jk(x cos θ +y sin θ )ejωt (3.13b)
ρc

urx
ury
pr ur

θ
x
θ

pi
ui
uiy v
uix

Figure 3.2  Reflection of a plane wave at oblique incidence.

 The sound field in front of a wall  53

pi
uiy = sin θ ⋅ e− jk(x cos θ +y sin θ )ejωt (3.13c)
ρc

in which the components of the particle velocity in the two axis directions
have been found using Equation 3.2.
Likewise for the reflected wave, and remembering the change in sign for
the x term, we get

pr = R ⋅ pi e− jk(− x cos θ +y sin θ )ejωt (3.14a)

pi
urx = −R ⋅ cos θ ⋅ e− jk(− x cos θ +y sin θ )ejωt (3.14b)
ρc

pi
ury = R ⋅ sin θ ⋅ e− jk(− x cos θ +y sin θ )ejωt (3.14c)
ρc

Through addition of Equations 3.13 and 3.14, we obtain for the total sound
field in front of a hard wall with reflection factor R = 1:

p = 2 ⋅ pi cos(kx cos θ )e− jky sin θ ejωt (3.15a)

pi
ux = −2j ⋅ cos θ ⋅ sin(kx cos θ )e− jky sin θ ejωt (3.15b)
ρc

pi
uy = 2 ⋅ sin θ ⋅ cos(kx cos θ )e− jky sin θ ejωt (3.15c)
ρc

By comparing with Equation 3.7, we see that the sound field corresponds to
a standing wave with the distance λ/(2 cos θ) between the nodal points and
between maxima; but at the same time, this field propagates parallel with
the wall with the speed given by

c (3.16)
cy =
sin θ

Notice that the speed of the phase is c y > c, i.e. supersonic. The ‘trace’ of the
sound field seems to move with the phase speed along the y-axis.
By squaring the amplitudes in Equation 3.15, we obtain

2
p = 2 ⋅ pi
2
(1 + cos (2kx cos θ )) (3.17a)
54  Sound Insulation in Buildings

2
pi
2
ux = 2 ⋅ (
cos2 θ ⋅ 1 − cos ( 2kx cos θ ) ) (3.17b)

( ρ c )2
2
pi
2
uy = 2 ⋅ (
sin2 θ ⋅ 1 + cos ( 2kx cos θ ) ) (3.17c)

( ρc ) 2

Thus, we find for the resulting particle velocity:

2
pi
2 2
u = ux + uy = 2 ⋅
2
(
⋅ 1 − cos (2θ ) cos (2kx cos θ ) )
( ρc )
2
(3.18)

Here we have made use of the mathematical relation (Equation 1.36).
While the potential energy equals zero in certain nodal points (see
Equation 3.17a), this is not the case for the kinetic energy when θ ≠ 0. The
reason is that only the normal component of the particle velocity, ux is zero
at the wall surface. The tangential component uy reaches a maximum value
at the wall and is proportional to p:

2
2 p
uy = sin2 θ

( ρ c )2 (3.19)

The total energy density can be found by entering Equations 3.17a and 3.18
into Equations 3.8 and 3.9 and making use of the mathematical relation
(Equation 1.38):

2
pi
w=
ρc2
(
⋅ 1 + sin2 θ ⋅ cos ( 2kx cos θ )
(3.20) )

It is worth mentioning that the total energy density is not constant in all
parts of the sound field in the case of oblique incidence. The maxima occur,
e.g. at the wall surface.

3.3 RANDOM INCIDENCE IN A
DIFFUSE SOUND FIELD

In this section, we shall assume a hard wall only with reflection factor
R = 1. If the intensity of the incoming wave is the same from all angles of
incidence (or the probability of all angles of incidence is the same) and if
all incoming sound waves are uncorrelated, then it is possible to calculate
the potential and kinetic energy densities in the diffuse field by integrating
 The sound field in front of a wall  55

the expressions valid for a single angle of incidence over the solid angle 2π.
From Equation 3.17a, we get for the pressure squared:
2π π 2
1
∫ ∫ 2 ⋅ p (1 + cos (2kx cos θ )) ⋅ sin θ d θ
2 2
p = dφ i
2π
0 0

sin(2kx)
= 2 ⋅ pi (1 +
2

2kx
2
(
) = 2 ⋅ pi 1 + sinc ( 2kx ) ) (3.21)

Here and in the following, the sinc function (Equation 1.42) is applied for
simplicity. Notice that in large distances from the wall and very close to the
wall, we have, respectively:
2 2
p ∞ = 2 ⋅ pi for x → ∞ (3.22a)

2 2 2
p x = 0 = 4 ⋅ pi = 2 ⋅ p ∞ for x → 0 (3.22b)

Likewise, from Equation 3.18, we get for the squared particle velocity:

 sin ( 2kx ) 
pi
2
 − cos ( 2kx ) 
2 sin(2kx) 2 kx
u = 2⋅ 2 1 − + 
ρc  2kx (kx)2 
 

sinc ( 2kx ) − cos ( 2kx ) 

2
pi 
= 2⋅
ρc2  1 − sinc ( 2kx ) + (kx)2  (3.23)

In a large distance from the wall, Equation 3.23 yields
2
2 pi
u ∞ = 2⋅ for x → ∞ (3.24a)
(ρ c)2

Close to the wall, Equation 3.23 cannot be used, but we find the limit by
setting x = 0 in Equation 3.18 and repeating the integration:
2 π 2
pi
∫ (1 − cos (2θ)) ⋅ sinθ dθ
2
u x=0 = 2 ⋅
ρc2
0

π 2

∫ 2 ⋅ sin θ d θ
2 3
= u∞
0

= 4
u∞
2
for x → 0 (3.24b)
3
56  Sound Insulation in Buildings

The total energy density is found by inserting Equations 3.21 and 3.23 in
Equation 3.8:

sinc ( 2kx ) − cos ( 2kx ) 

2
pi 
w=
ρc2  1 + 2(kx)2  (3.25)

Far from the wall, Equation 3.25 yields, as expected, recalling Equation 3.12:

2
pi
w∞ → = 2w i for x → ∞ (3.26a)
ρc2

Very close to the wall, we find by applying Equations 3.22b and 3.24b:

2
pi  2 5
wx =0 →  1 +  = w∞ for x → 0 (3.26b)
ρc2 3 3

Hence, we see that in a diffuse sound field, the total energy density at the
walls (with R = 1) is 10 log (5/3) = 2.2 dB higher than the asymptotic value
far from the walls. Likewise, from Equations 3.22 and 3.24, we see that the
sound pressure is 3 dB higher and the particle velocity is 1.2 dB higher than
the respective asymptotic values far from the walls. In Table 3.1, the results
for the sound field at the surface of a hard wall relative to the asymptotic
values far from the wall are summarized.
The main results so far are illustrated in Figure 3.3. For normal, oblique
and random sound incidence on a hard wall, the patterns of potential,
kinetic and total energy densities are shown – all relative to the asymp-
totic values far from the wall (two times the corresponding values in the
incoming wave). In the case of random incidence, notice how the ampli-
tudes of the fluctuations decrease quickly with the distance from the wall.
Already at the distance of λ/4, the sound pressure has reached its asymp-
totic value, and only minor deviations from this value occur further away
from the wall.

Table 3.1  L
 evels of sound pressure, particle velocity and total energy density at the
surface of a hard wall, relative to the asymptotic values far from the wall
Sound pressure (dB) Particle velocity Total energy density
Normal incidence 3 → −∞ dB 0 dB
Oblique incidence 3 10 lg (2 sin2θ) 10 lg (1 + sin2θ)
Random incidence 3 1.2 dB 2.2 dB
 The sound field in front of a wall  57

(a) 2 (b) 2 (c) 2

–
ρ2 υ2 E
1 θ = 0˚
1 1

0 0 0
2 kx 2 kx 2 kx

(d) 2 (e) 2 (f ) 2
–
ρ2 υ2 E

1 1 1
θ = 30˚

0 0 0
2 kx 2 kx 2 kx

(g) (h) (i) 2

2 2
–
ρ2 υ2 E
1 1 1
θ = 60˚

0 0 0
2 kx 2 kx 2 kx

(j) 2 (k) (l) 2

2
–
ρ r2 υr2 Er
1 1 1

0 0 0
2 kx 2 kx 2 kx

Figure 3.3 Sound field energy components in front of a totally reflecting wall. Four rows
from above: plane incident sound waves with angles of incidence 0°, 30°, 60°
and diffuse field random incidence. In each case, the three columns show
potential, kinetic and total energy densities relative to the values far from the
wall. (Reproduced from Waterhouse, R.V., J. Acoust. Soc. Am., 27, 247–258,
1955, Figure 1. With the permission of the Acoustical Society of America.)

For physical reasons, one should expect the total energy density to be
equally distributed between potential and kinetic energy in the total sound
field, i.e.

∞ ∞

∫
0
∫
wpot d x = wkin d x (3.27)
0
58  Sound Insulation in Buildings

As we see immediately from Equations 3.9, 3.22a and 3.24a, the asymp-
totic values are equal:
2
pi
wpot, ∞ = wkin, ∞ = 2 ⋅ for x → ∞ (3.28)
ρc2

Consequently, we only need to calculate any addition to this value due to

the increase in potential energy density close to the wall:

∞ ∞ ∞
wpot − wpot, ∞  p2  1 π λ
∫ wpot, ∞ ∫ p 
∫
d x =  2 − 1 d x = sinc ( 2kx ) d x =
2
⋅ = (3.29)
k 2 8
0 0 ∞ 0

As shown by Waterhouse (1955), the same result is obtained for the kinetic
energy density and, thus, for the total energy density, too. It is remarkable
that acoustic energy is slightly increased near the walls in a diffuse sound
field. The additional energy corresponds to a sound field with constant
energy density also occupying the volume extending λ/8 beyond the physi-
cal walls. In a large room (dimensions  wavelength) , the total energy is
approximately

 λ
Etot = w ⋅  V + S  (3.30)
 8

where w is the energy density averaged over the room volume avoiding
regions very close to the walls, V is the room volume and S the total area of
the inner wall surfaces.
This implies that to find the total energy in a room based on a measure-
ment of the average energy density in the central part of the room, the
measurement result should be corrected with the amount:

 λS 
CW = 10 lg  1 +  (dB) (3.31)
 8V 

This correction is sometimes called the ‘Waterhouse correction’ because
it was first derived by Waterhouse (1955) in a pioneering paper that also
treats the sound field at the edge between two walls and in wall corners.
While negligible at high frequencies, the correction amounts to a few dB
below 500 Hz, depending on the volume. Let us summarize the assump-
tions on which the Waterhouse correction is based:

• Totally reflecting walls. This may be approximately right at low fre-

quencies in a room with heavy walls, but questionable with light-
weight drywalls.
• Random incidence on the surfaces, i.e. a diffuse sound field. This
assumption is not fulfilled at very low frequencies in small rooms
 The sound field in front of a wall  59

where axial modes dominate the sound field. Axial modes correspond
to the case of normal incidence, where there is no correction.
• Only the energy condition of a large surface is considered. The energy
conditions near edges and corners are not included.

3.4 THE SOUND FIELD AT EDGES

AND CORNERS IN A ROOM

In the previous sections, we have derived expressions for the behaviour of

the sound field in front of a totally reflecting wall. In the same way, expres-
sions can be derived for the sound field in front of room edges and corners
as performed by Waterhouse (1955). In the following, only a qualitative
description and presentation of the main results will be given and only in
the case of diffuse sound field incidence.
As can be seen from Figure 3.4, at a certain point of observation, the sound
field can be regarded as the combination of the undisturbed, incident sound
field at this point and at one or more images of this point. A single wall only pro-
vides one image point and hereby, the sound field consists of two ­contributions.
A right-angled edge results in four contributions of which one is a second-order
image, and a right-angled corner results in eight contributions totally.
The properties of the resulting sound field depend on the correla-
tion between the individual contributions to the sound field. If the point
of ­observation is sufficiently far away from the reflecting surfaces (in a
diffuse field at ­
­ least λ/4), we can regard the different contributions as
­mutually   uncorrelated, and the resulting sound field can, therefore, be
regarded as the energy sum of the c­ ontributions. With the pressure of the
incoming sound wave denoted as p ­ i and the resulting sound field far from
the walls (the far field) denoted as p∞, we have

2 at a wall
p∞
2

2 = 4 at an edge (3.32)
pi 
8 in a corner

(a) (b) (c)

Figure 3.4 Illustration of (a) a point of observation (black spot) and its images in a wall;
(b) a right-angled edge; and (c) a right-angled corner.
 60  Sound Insulation in Buildings

However, if the point of observation is moved close to the walls so that

the distance between the point of observation and its images becomes very
small compared to the wavelength, all the contributions to the sound field
become correlated and in phase. The resulting sound field can, therefore, be
described by the pressure sum of the contributions, and we get
2 at a wall
p x=0 
= 4 at an edge (3.33)
pi 
8 in a corner

In a diffuse sound field (random incidence), the squared sound pressure
in front of a totally reflecting wall relative to the asymptotic value, from
Equation 3.21, is
2
p
2 = 1 + sinc ( 2kx ) (3.34)
p∞

Similarly, near a rectangular edge lying along the x-axis, the result is
(Waterhouse 1955)
2
p
2 = 1 + sinc ( 2ky ) + sinc ( 2kz ) + sinc ( 2kd x ) (3.35)
p∞

where dx = √(y 2+z 2). Near a rectangular corner with the x-, y- and z-axis
along the three edges forming the corner, the result becomes
2
p
2 = 1 + sinc ( 2kx ) + sinc ( 2ky ) + sinc ( 2kz ) + sinc ( 2kd x ) + sinc ( 2kd y ) + sinc ( 2kd z ) + sin
(3.36)
p∞

(2kz ) + sinc (2kdx ) + sinc (2kdy ) + sinc (2kdz ) + sinc (2kd )

where dy = √(x 2+z 2), dz = √(x 2+y 2) and d = √(x 2+y 2+z 2).
The relative sound pressure level (Figure 3.5) is
2
p
∆L = 10lg 2
(dB) (3.37)
p∞

as function of the distance d, which in this case means either the distance
perpendicular from a wall, or the distance from a 2D edge along the axis of
symmetry (y = z in Equation 3.35), or the distance from a 3D corner along
the axis of symmetry (x = y = z in Equation 3.36). It is obvious that the fluc-
tuations of ΔL near the reflecting surfaces are exaggerated when going from
 The sound field in front of a wall  61

12

9
Wall Edge Corner
Sound pressure level re far field (dB)

–3

–6

–9

–12
0 5 10 15 20 25
2kd

Figure 3.5 Interference pattern near a wall, an edge and a corner.

Table 3.2  Some characteristics of the sound pressure level near walls, edges and
corners assuming hard reflective surfaces and a diffuse sound field
Relative sound
Relative sound pressure level ΔL
pressure level Distance d to Distance d to at first minimum
ΔL at d = 0 (dB) first ΔL = 0 dB first minimum (dB)
Wall 3 0.25 λ 0.35 λ −1
Edge (2D) 6 0.32 λ 0.45 λ −3
Corner (3D) 9 0.53 λ 0.78 λ −11
Note: ΔL is sound pressure level relative to the asymptotic values in the far field.

a single wall to an edge and further to a corner. In the latter case, we have
variations from +9 dB to −11 dB and the distance from the corner must be
d > λ in order to reach small fluctuations within ca. ±1 dB. More details from
the curves are listed in Table 3.2.

3.5 THE INFLUENCE OF BANDWIDTH

ON REFLECTION OF NOISE

In the previous section, we have treated the sound field in front of a wall
for a single frequency. If the sound field contains white or pink noise in a
certain frequency band Δf = f 2 − f1, we can carry out a simple energy summa-
tion of the mutually uncorrelated frequency components.
62  Sound Insulation in Buildings

For noise incident from one direction with angle of incidence θ relative
to the normal of the totally reflecting wall, Equation 3.17a gives us the
squared sound pressure amplitude. Using k = 2πf/c and the distance from
the wall x = d, we get

2 f2
p 1   4π 
2 ⋅ pi
2 =
f2 − f1 ∫  1 + cos 
f1
c
fd cos θ   d f


  4π   4π 
sin  f2d cos θ  − sin  f1d cos θ  
1   c   c 
=  f2 − f1 + 
f2 − f1  4π
f2d cos θ 
 c 

  4π (f2 + f1)   4π (f2 − f1) 

 2 ⋅ cos  d cos θ  ⋅ sin  d cos θ  
 2c   2c 
= 1 + 
 4 π 
(f 2 − f1)d cos θ
 c 

= 1 + cos ( 2X ) ⋅ sinc (Y ) (3.38)

where
2π
X=
c
fm d cos θ (f m = 12 (f2 + f1)) (3.39a)

2π
Y= ∆fd cos θ (∆f = f2 − f1) (3.39b)
c
This result was first derived and validated by Rindel (1978).
It appears that the interference pattern in front of the wall depends on
two parameters – X, which is controlled by the arithmetic mean frequency
of the frequency band, and Y, which is controlled by the bandwidth. In
addition, the distance from the wall and the angle of incidence are variables
in both parameters. The difference between the centre frequency, which is
the geometric mean, and the arithmetic mean frequency of the frequency
band is small for octave bands (fm /fc = 1.06) and negligible for one-third
octave bands (fm /fc = 1.0066).
Unless the bandwidth is zero, Equation 3.38 has the asymptotic value
of unity. This value corresponds to a simple energy summation of inci-
dent and reflected noise without any interference. Consequently, this is a
natural normalization factor for the interference function in Equation 3.38.
Figure 3.6 shows four examples of what the interference pattern looks like
for different bandwidths of noise. The abscissa is X = (2π/c)fmd cos θ. As
can be seen, a standing wave occurs near the wall – like in the case of
a pure tone of frequency fm; but the larger the bandwidth, the faster the
 The sound field in front of a wall  63

(a) 2

|p|2
1
2•|pi|2

0
0.1 1 10 100
x
(b) 2

0
0.1 1 10 100
x
(c) 2

0
0.1 1 10 100
x
(d) 2

0
0.1 1 10 100
x

Figure 3.6 Examples of interference pattern corresponding to Equation 3.38 for four dif-
ferent values of the noise bandwidth. Valid for totally reflecting walls and a
single angle of incidence. (a) pure tone; (b) one-third octave bandwidth; (c)
octave bandwidth; and (d) wide band. The dashed lines are the sinc (Y) function.

interference pattern converges towards the asymptotic value. The band-

width determines an envelope function that limits the fluctuations. For
wide band noise, Δf = 2fm , noticeable interference occurs only very close to
the wall, see Figure 3.6d.
The interference in the sound pressure level close to the wall can be
expressed as a correction relative to the sound pressure level when incident
and reflected sound are added on a simple energy basis:

 p2 
∆Lint = 10lg   = 10lg (1 + cos(2X) ⋅ sinc(Y )) (dB) (3.40)
2⋅ p 2 
 i 
64  Sound Insulation in Buildings

(a) 3

–3
ΔLint (dB)

–6

–9

–12

–15
0.1 1 10 100
X

(b) 3

–3
ΔLint (dB)

–6

–9

–12

–15
0.1 1 10 100
X

Figure 3.7 Interference correction in front of a large reflecting wall as a function of the

parameter X. (a) one-third octave bands and (b) octave bands.

In Figure 3.7, tha interference correction is shown for one-third octave

bands and octave bands as a function of frequency, distance and angle of
incidence; the abscissa is X = (2π/c)fmd cos θ. Significant negative interfer-
ence is noted around X = 1.6, which is at the distance d = λ/(4 cos θ).
When measuring the sound pressure level in front of a large reflecting
surface, e.g. a building façade, there are two possibilities for the choice of
microphone position; either very close to the surface where the interference
causes ΔLint = 3 dB, or in a sufficiently long distance from the surface where
the interference is negligible, ΔLint ≈ 0 dB. The former requires the distance
to be d < 0.075λ/cos θ in order to ensure ΔLint > 2 dB. The most critical here
are the high frequencies and normal incidence of sound. Hence, the dis-
tance should be d < 8 mm at 3150 Hz. This may just be obtained with a ½
inch microphone mounted with axis parallel to the wall, whereas a micro-
phone with 1 inch diaphragm is too big. Alternatively, the microphone may
be mounted into a hole in the wall with the diaphragm flush with the wall.
 The sound field in front of a wall  65

The other option is to measure in distance d > 1.7λ/cos θ in order to ensure

|Lint| < 1 dB. In this case, the most critical are low frequencies and oblique
incidence of sound (see examples in Table 3.3). Hence, at 100 Hz one-third
octave, the minimum distance should be around 6 m, and even more if the
angle of incidence exceeds 15°. This means that in practice it is not possible
to avoid measurement errors due to interference.
When measuring façade sound insulation in accordance with ISO
16283-3, there are two possibilities for the sound source. Either it can be
a loudspeaker with an angle of incidence 45°, or the source can be traffic
noise emitted from a line, preferably covering a range of angles within ±60°.
For such a line source, the time-averaged sound field can be assumed to con-
tain a constant sound intensity from all directions of incidence within the
angular range, under which the road is seen from the measurement position.
The resulting interference pattern is calculated by numerical integration of
Equation 3.40 and the result is shown in Figure 3.8. It is to be noted that

Table 3.3  M
 inimum distance (in m) from a reflecting wall in order to have |ΔLint|
< 1 dB at 100 Hz one-third octave band or 125 Hz octave band for
oblique incidence at various angles or for a line source covering the
range from −60° to +60°

Angle of incidence (°) One-third octave100 Hz Octave125 Hz

0 5.8 2.7
15 6.0 2.8
30 6.7 3.1
45 8.2 3.8
60 11.6 5.4
75 22.3 10.5
Line (±60°) 2.2 1.2

0
ΔLint (dB)

–3

–6

–9
0.1 1 10 100
X

Figure 3.8 Interference correction for one-third octave bands, averaging angles of inci-
dence with in the range from −60° to +60°. The parameter is X = (2π/c)fmd.
66  Sound Insulation in Buildings

the interference pattern is less pronounced compared to the case of a fixed

angle of incidence.
The standard ISO 16283-3 prescribes measurements to be made either
flush with the façade or at a distance of 2.0 m ± 0.2 m. The theoretical mea-
surement errors in the lower one-third octave bands are shown in Table 3.4
for the distance 2 m. Below 400 Hz, the interference causes errors around
±2 dB for the loudspeaker measurements, whereas the error is around ±1 dB
for the traffic noise measurements.
Measurements in front of building façades put great demands on the
choice of microphone position; but if the bandwidth can be extended, e.g.
to octave instead of one-third octave, the interference problems are sub-
stantially reduced. By measurements of A-weighted traffic noise levels, the
bandwidth is so large that interference is no longer a problem. However, in

Table 3.4  T
 heoretical measurement errors due to interference for one-third
octave bands with the microphone 2 m in front of the façade
Frequency (Hz) 45° ΔLint (dB) Traffic ΔLint (dB)
100 1.7 −1.3
125 2.7 1.2
160 −2.3 0.8
200 −2.2 −0.9
250 2.0 0.8
315 −2.0 −1.0
400 −0.6 0.1
500 0.1 0.1
Note: Either loudspeaker measurements using angle of incidence 45° or measurements with
traffic noise covering the range from −60° to +60°.

3
ΔLint (dB)

–3
0.1 1 10 100
X

Figure 3.9 Interference correction for one-third octave bands and random incidence
(3D diffuse field). The parameter is X = (2π/c)fmd.
 The sound field in front of a wall  67

case the noise contains substantial pure tones, the problems occur again.
Sometimes, the problem can be solved by carrying out a spatial average in
an area in front of the façade, as is the normal procedure in closed rooms.
For random incidence or 3D diffuse sound field, the effect of the band-
width can be found through numerical integration of Equation 3.21 over
the frequency range of relevance. For the one-third octave band, the result
is shown in Figure 3.9. The first minimum occurs at a distance d = 0.35 λ
from the wall, i.e. at a little longer distance than in the case of normal inci-
dence (0.25 λ).

REFERENCES

ISO 16283-3 (2016). Acoustics – Field measurements of sound insulation in build-

ings and of building elements – Part 3: Façade sound insulation. International
Organization for Standardization, Geneva, Switzerland.
J.H. Rindel (1978). Lydfeltet foran en stor, reflekterende plan (The sound field in
front of a large reflecting surface, in Danish). Proceedings of NAM-78, 16–18
August 1978, Odense, Denmark, pp. 101–104.
R.V. Waterhouse (1955). Interference patterns in reverberant sound fields. Journal of
the Acoustical Society of America 27, 247–258.
Chapter 4

Introduction to room acoustics

This chapter summarizes the basic concepts in room acoustics as a basis

for sound insulation theory and building acoustic measurement methods.

4.1 SOUND WAVES IN ROOMS

4.1.1 Standing waves in a rectangular room

A rectangular room has the dimensions lx, ly and lz. The wave equation can
then be written as (e.g. Kuttruff 1973, p. 51)

∂2 p ∂2 p ∂2 p
+ + + k2 p = 0 (4.1)
∂ x2 ∂ y 2 ∂ z 2

where p is the sound pressure and k = ω/c is the angular wave number, ω is
the angular frequency and c is the speed of sound in air. The equation can
be solved by separation of the variables, and it is assumed that the solution
can be written in the form:

p = X(x) ⋅ Y (y) ⋅ Z(z) ⋅ ejωt

Insertion of this expression in Equation 4.1 and division by p gives

1 ∂2 X 1 ∂2 Y 1 ∂2 Z
+ + + k2 = 0
X ∂ x2 Y ∂ y 2 Z ∂ z 2

This can be separated in various directions, and for the x-direction, it yields

1 ∂2 X
+ kx2 = 0
X ∂ x2

69
70  Sound Insulation in Buildings

Similar equations hold for the y and z directions. The angular wave num-
ber k is divided into three along the directions as

k2 = kx2 + ky2 + kz2 (4.2)

The general solution to the aforementioned one-dimensional equation is

X(x) = Cx cos ( kx x + ϕ x )

in which the constants C x and φx are determined from the boundary

conditions.
The room surfaces are now assumed to be rigid, i.e. the normal compo-
nent of the particle velocity is zero at the boundaries

1 ∂p
ux = − =0 for x = 0 and x = lx
jωρ ∂x

This means that φx = 0 and

π
kx = ⋅ nx , where nx = 0,1, 2,3, … (4.3)
lx

Two similar boundary conditions hold for the y and z directions. With
these conditions, the solution to Equation 4.1 is

 x  y  z
p = p0 ⋅ cos  π nx  ⋅ cos  π ny  ⋅ cos  π nz  (4.4)
 lx   ly   lz 

The time factor ejωt is omitted. The amplitude of the sound pressure does not
move with time, so the waves that are solutions to Equation 4.4 are called
standing waves. They are also called the modes of the room, and each of
them is related to a certain natural frequency (or eigenfrequency) given by

ω n ck c
fn = = = kx2 + ky2 + kz2
2π 2π 2π
2 2
(4.5)
2
c  nx   ny   nz 
fn = + +
2  lx   ly   lz 

Modes can be divided into three groups:

• Axial modes are one-dimensional, i.e. the amplitude of the sound

pressure varies in one direction, only – one of nx, ny, nz is > 0.
 Introduction to room acoustics  71

• Tangential modes are two-dimensional, i.e. the amplitude of the

sound pressure varies in two directions – two of nx, ny, nz are > 0.
• Oblique modes are three-dimensional, i.e. the amplitude of the sound
pressure varies in three directions – all three of nx, ny, nz are > 0.

Some examples are shown in Figure 4.1. It is observed that the set of numbers
(nx, ny, nz) indicate the number of nodes (places with p = 0) along each coordi-
nate axis. Although it is usual to display the room modes as the distribution
of p2 , it may actually be more relevant to look at the distribution of the sound
pressure level in dB; see the last example, mode (2,2,0), in Figure 4.1.

(a) p2 p2

(2,0,0) (2,1,0)

p2 p2

(1,2,0) (3,1,0)

(b) SPL, dB
p2

(2,2,0) (2,2,0)

Figure 4.1 Examples of room modes. (a) The vertical axis is the sound pressure squared,
and the horizontal axes are along the length and width of the room. Mode (2,0,0)
is an axial mode and the others are tangential modes and (b) Mode (2,2,0) is
shown both as the sound pressure squared and as the sound pressure level in dB.
72  Sound Insulation in Buildings

4.1.2 Modal reverberation times

A rectangular room with dimensions lx, ly, lz is considered. Using Euler’s
formula (Equation 1.39) for the solution (Equation 4.4) to the wave equa-
tion, the sound pressure in a room mode can be written as

  n 

p=
p0
8 ∑expjπ ± nl x

x
x±
ny
ly
y ± z z 
lz 

(4.6)

As before, the time factor ejωt is omitted, and the summation is taken over
all eight possible combinations of the + and − signs. Equation 4.6 shows
that the total sound pressure can be interpreted as the interference between
eight plane waves travelling in different directions. The wave number of the
propagation along the x-axis is kx = πnx /lx. The direction of propagation
expressed as an angle relative to the x-axis is

nx
lx nx ⋅ c
cos ϕ x = =± (4.7)
2 2 2 2 ⋅ lx ⋅ fn
 nx   ny   nz 
 l  +  l  +  l 
x  y z

It is analogous for the y and z directions.

Two examples of tangential modes (A and B) in a rectangular room are
visualized in Figure 4.2. The nodal lines (actually vertical planes) represent
the standing wave pattern, whereas the corresponding wave fronts repre-
sent the plane waves travelling in four different directions. The normal
distance between parallel wave fronts is the wave length λ = c/f. Figures 4.2c
of modes A and B show the four possible directions of propagation that
are perpendicular to the wave fronts. The direction of propagation and the
directional cosines in Equation 4.7 are displayed in Figure 4.3 in the case
of an oblique mode.
The sound absorption coefficients of the six surfaces of the room are
denoted by αx1, αx2 , αy1, αy2 , αz1 and αz2 . The reverberation time Tn of a room
mode with modal number (nx, ny, nz) can be derived – see Rindel (2016) for
the details – and the result is

−1
 nx
( ny
) (
 l 2 ⋅ ln (1 − α x1 )(1 − α x2 ) + l 2 ⋅ ln (1 − α y1 )(1 − α y 2 ) )
55.3 ⋅ fn  x y
 (4.8)
Tn =
−c 2  n 
 lz
(
 + 2z ⋅ ln (1 − α z1 )(1 − α z 2 ) ) 

 
 Introduction to room acoustics  73

(a)

(b)

(c)

A B

Figure 4.2 Two examples of tangential modes. A: (1,2,0) mode, left side and B: (3,1,0)
mode, right side. (a) nodal lines; (b) wave fronts in two positions with a short
time delay; and (c) directions of propagation, perpendicular to the wave
fronts.

kz

nz/lz
φz
φy
ky
φx
nx/lx

ny/ly
kx

Figure 4.3 Direction of propagation and directional cosines of an oblique mode.

74  Sound Insulation in Buildings

If we assume all absorption coefficients to be equal to the mean absorption

coefficient αm , we get the modal reverberation time:

−1
55.3 ⋅ fn   nx ny nz  
Tn = 2 ⋅  2 + 2 + 2  ⋅ ln (1 − α m )  (4.9)
−c 2   lx ly lz  

This shows that the axial modes have longer reverberation times than tan-
gential modes, and oblique modes have shorter reverberation times.

4.1.3 Transfer function in a room

The transfer function is the frequency response from a source position to
a receiver position in a room. When a sound source submits a tone at a
certain frequency, a large number of room modes will be excited and the
sound pressure at a point in the room can be described as the addition
of the contributions from all modes with phase (e.g. Kuttruf 1973, eq.
(III.30)):

p(ω) = ∑ω 2
An
− ω − j2δnω n
2
n
(4.10)
n

Here An is the amplitude of mode n, ω is the angular frequency emitted by

the source, ωn is the angular frequency of room mode n and δn is the damp-
ing constant.
An example of a calculated transfer function is shown in Figure 4.4. It
fluctuates very much with frequency, and the maxima can be identified as the
natural frequencies of the room. The example in Figure 4.4 has the same room
dimensions as was used for the calculations in Table 4.1. The calculations
are made as explained in Rindel (2015). While the first five modes are clearly
separated, there are several examples of coinciding or overlapping modes that
cannot be separated: modes (1,0,1) and (2,1,0) around 54 Hz; modes (1,1,1)
and (0,2,0) around 62 Hz; and the three modes (2,0,1), (1,2,0) and (3,0,0)
around 67 Hz.

4.1.4 Modal density
A closer inspection of Equation 4.5 shows that the natural frequen-
cies of a rectangular room may be interpreted in a geometrical way. A
three-­d imensional frequency space is shown in Figure 4.5. The natu-
ral frequencies of the one-dimensional modes are marked on each of
the axes, representing the axial modes of the length, the width and the
height, respectively. Interestingly, the points in the grid represent the
oblique modes, the distance to each point from the origin is the natural
 Introduction to room acoustics  75

60

50

40
SPL, dB

30

20

10

0
20 40 60 80 100 200
Frequency, Hz

Figure 4.4 Transfer function in a rectangular room with dimensions 7.7 m, 5.5 m and
3.5 m. At low frequencies, it is possible to identify the modes by their modal
numbers (compare with Table 4.1).

Table 4.1  Calculated natural frequencies at low

frequencies using Equation 4.5 in a
rectangular room with dimensions 7.7 m,
5.5 m and 3.5 m
nx ny nz fn (Hz)
1 0 0 22.3
0 1 0 31.2
1 1 0 38.4
2 0 0 44.6
0 0 1 49.0
1 0 1 53.9
2 1 0 54.4
0 1 1 58.1
1 1 1 62.3
0 2 0 62.4
2 0 1 66.3
1 2 0 66.3
3 0 0 66.9
2 1 1 73.3
3 1 0 73.8
2 2 0 76.7
0 2 1 79.4
76  Sound Insulation in Buildings

fz

c/2lz fy

c/2lx

c/2ly

fx

Figure 4.5 Frequency grid in which each grid point represents a room mode.

frequency of that mode. Hence, the number of oblique modes below a

certain frequency f is equal to the number of grid points inside the sphere
with radius f.
The volume is one-eighth of the sphere with radius f, i.e. (4πf 3/3)/8 =
πf 3/6. Each mode occupies a volume c 3/(8lxlylz) = c3/(8V). Hence, the number
of modes below f is approximately

3
π f 3 8V 4 π V  f 
N= ⋅ 3 = ⋅  (4.11)
6 c 3  c

The axial modes are marked along each of the three axes and the tan-
gential modes are found in each of the three planes between two axes. If
these modes are also taken into account, the number of modes with natural
frequencies below the frequency f, the mode count, is (Pierce 1989, p. 293)

3 2
4π V f πS  f  Lf 1
N=  c  + 4  c  + 8 c + 8 (4.12)
3

where V is the volume of the room, S is the total area of the surfaces and L
= 4 (lx + ly + lz) is the total length of all edges. At high frequencies, the oblique
modes dominate and the first term in Equation 4.12 is a good approxima-
tion for any room, not only for rectangular rooms.
The modal density is the average number of modes per hertz. From
Equation 4.12, we get
 Introduction to room acoustics  77

d N 4π V 2 π S L
n(f ) = = 3 f + 2f+ (4.13)
df c 2c 8c

In Figure 4.6, this is compared to the actual modal density in a room. At

very low frequencies, the estimate can give fractional values between 0 and
1, whereas the actual number of modes within a frequency band must be
a whole number. In the example room, there is one mode in the 20 Hz and
the 31.5 Hz bands, but none in the 25 Hz band.
For high frequencies (e.g. above 200 Hz), it may sometimes be sufficient
to use only the first term in Equation 4.12 for the modal density:

4π V 2
n(f ) ≅ f (4.14)
c3

A more detailed analysis of the modal grid points in the frequency domain
(Rodríguez Molares 2010, Table 7.1) leads to the modal densities of axial,
tangential and oblique modes in a three-dimensional room:

L
nax (f ) ≅ (4.15)
2c

πS L
ntan (f ) ≅ f− (4.16)
c2 2c
Number of modes per 1/3 octave band

100

n(ax)
10 n(tan)
n(ob)
n(total)
1
n(count)

0.1
20 25 31.5 40 50 63 80 100 125 160 200
1/3 octave center frequency, Hz

Figure 4.6 Modal density as a function of frequency in a rectangular room with dimen-

sions 7.7 m, 5.5 m and 3.5 m. Dots indicate the actual number of modes in
the third octave bands from 20 Hz to 200 Hz (compare with Table 4.1). The
curves show the number of modes estimated for axial, tangential, oblique and
total modes by Equations 4.13 and 4.15 through 4.17.
78  Sound Insulation in Buildings

4π V 2 π S L
nob (f ) ≅ 3
f − 2f+ (4.17)
c 2c 8c

Adding the modal densities of the three groups of modes yields again
Equation 4.13.
Each normal mode is in itself a resonant system that can be associated
with a bandwidth around the natural frequency. From Equation 4.6, it fol-
lows that the squared sound pressure of mode n is

2
An
pn2 = (4.18)
(ω )
2
2
− ω n2 + 4ω 2δ n2

The bandwidth Br is defined as the frequency interval in which the acous-

tic energy is at least half the maximum energy at the natural frequency,
the  3 dB bandwidth (Figure 4.7). It follows from Equation 4.18 that the
maximum energy is at ω = ωn and the half energy is at ω1,2 = ωn ± δn. The
bandwidth in angular frequency is Δω = ω2−ω1 ≈ 2δn and, thus, the band-
width Br in Hz can be related to the reverberation time, T:

∆ω δ 2.2
Br = ≅ = (4.19)
2π π T

The relation between damping constant and reverberation time follows

from the definition of T as the time for the energy to decay by 60 dB after
the source is interrupted. A resonant system with damping constant δ has
the decay 10 lg (exp(−2δT)) = −60 dB, and so δT = 3/lg (e) = 2.2π.

Measurement band
Room response, Lp, dB

3 dB

Br

Frequency

Figure 4.7 Frequency response of three modes within a frequency band of measure-

ment. The modal overlap index is M = 0.7. (Adapted from Bies, D.A. and
Hansen, C.H., Engineering Noise Control, Theory and Practice, Unwin Hyman
Ltd., London, 1988, Figure 7.3.)
 Introduction to room acoustics  79

In a room, the natural frequencies of the normal modes are quite spread
at low frequencies, but at higher frequencies, they are very close together.
The number of modes within the bandwidth of the modes is called the
modal overlap index, M:

8.8 ⋅ π ⋅ V 2
M = Br ⋅ n(f ) ≈ f (4.20)
T ⋅ c3

The modal overlap index in a room increases rapidly with increasing fre-
quency, which is clearly seen in the approximation where Equations 4.14
and 4.19 are inserted. Figure 4.7 shows an example of low modal overlap.
There are three modes within the measurement band, which is 4.2 times
the modal bandwidth. Thus, M = 4.2/3 ≈ 0.7.

4.1.5 The Schroeder limiting frequency

Due to the high number of modes in a room at high frequencies and in
order to avoid too complicated methods, it is necessary to apply statisti-
cal considerations, e.g. to assume the same energy in all modes within a
certain frequency band. As a minimum requirement, it has been suggested
by Schroeder (1962) that the modal overlap should be M = 3, although this
cannot be considered a very sharp borderline. Using Equation 4.20, this
leads to a lower limiting frequency for statistical methods, the so-called
‘Schroeder’s limiting frequency’, fg:

T
f g = 2000 (4.21)
V

where T is the reverberation time (s) and V is the volume (m3).

As an example, a room with volume 50 m3 and reverberation time 0.5 s
has a limiting frequency of fg = 200 Hz. Thus, below this frequency, a
weak coupling between the room modes can be assumed, or – in other
words – the sound field below 200 Hz may be dominated by single room
modes.

4.2 STATISTICAL ROOM ACOUSTICS

4.2.1 The diffuse sound field

In this section, the acoustical behaviour of a room is treated from a statis-
tical point of view, based on energy balance considerations. It is assumed
that the modal density is high enough, so that the influence of single modes
in the room can be neglected. It is also assumed that the reflection density
80  Sound Insulation in Buildings

is high enough so that the phase relations between individual reflections

can be neglected. This means that the reflections in the room are assumed
to be uncorrelated and their contribution can be added on an energy basis.
The diffuse sound field is defined as a sound field in which

• The energy density is the same everywhere.

• All directions of sound propagation occur with the same probability.

It is obvious that the direct sound field near a sound source is not included
in the diffuse sound field. Neither are the special interference phenomena
that are known to give increased energy density near the room boundaries
and corners. The diffuse sound field is an ideal sound field that does not
exist in any room. However, in many cases, the diffuse sound field can be a
good and very practical approximation to the real sound field.

4.2.2 Incident sound power on a surface

In a plane propagating sound wave, the relation between RMS sound pres-
sure p1 and sound intensity I1 is

p12 = I1 ⋅ ρ c

In a diffuse sound field, the RMS sound pressure pdiff is the result of sound
waves propagating in all directions, and all having the sound intensity I1.
By integration over a sphere with the solid angle ψ = 4π, the RMS sound
pressure in the diffuse sound field is

2
pdiff =
∫
ψ = 4π
I1 ⋅ ρ c dψ = 4π ⋅ I 1 ⋅ ρ c
(4.22)

In the case of a plane wave with the angle of incidence θ relative to the
normal of the surface, the incident sound power per unit area on the
surface is

2
pdiff
Iθ = I1 cos θ = cos θ (4.23)
4πρ c

where pdiff is the RMS sound pressure in the diffuse sound field. This is just
the sound intensity in the plane propagating wave multiplied by the cosine,
which is the projection of a unit area as seen from the angle of incidence
(Figure 4.8).
The total incident sound power per unit area is found by integration over all
angles of incidence covering a half sphere in front of the surface (Figure 4.8b).
 Introduction to room acoustics  81

(a) p1 (b)

pdiff

θ Iθ Iinc

Figure 4.8 (a) Plane wave at oblique incidence on a surface and (b) diffuse incidence on
a surface.

The integration covers the solid angle ψ = 2π and Figure 4.9 shows the defini-
tion of the angles.

2π π / 2
2
1 pdiff
Iinc =
∫
ψ = 2π
Iθ dψ =
4π ∫∫ 0 0
ρc
cos θ sin θ d θ d φ

1
1 p2 2
1 pdiff 1
=
4π
⋅ 2π ⋅ diff
ρc ∫
0
sin θ d(sin θ ) = ⋅
2 ρc 2
⋅ (4.24)

2
p diff
Iinc =
4ρc

It is noted that this is four times less than in the case of a plane wave of
normal incidence.

sinθ dφ

dθ
θ

φ y

Figure 4.9 Definition of angles of incidence in a diffuse sound field.

82  Sound Insulation in Buildings

4.2.3 Equivalent absorption area

The absorption coefficient α is defined as the ratio of the non-reflected
sound energy to the incident sound energy on a surface. It can take values
between 0 and 1; α = 1 means that all incident sound energy is absorbed in
the surface. An example of a surface with absorption coefficient, α = 1 is an
open window. Sabine in his original paper on reverberation from 1898 sug-
gested the use of open-window unites for the absorption area (Paper one in
collected papers, Sabine, 1922).
The product of area and absorption coefficient of a surface material is the
equivalent absorption area of that surface, i.e. the area of open windows
giving the same amount of sound absorption as the actual surface. The
equivalent absorption area of a room is

A= ∑Sα
i
i i = S1α 1 + S2α 2 + … = Sα m

(4.25)

where S is the total surface area of the room and αm is the mean absorption
coefficient. The unit of A is m 2 . In general, the equivalent absorption area
may also include sound absorption due to the air and due to persons or
other objects in the room.

4.2.4 Energy balance in a room

The total acoustic energy in a room is the sum of potential energy and
kinetic energy, or twice the potential energy, since the time average of the
two parts must be equal. The total energy E is the energy density multiplied
by the room volume V:

p2
E = ( wpot + wkin ) V = 2wpotV = V (4.26)
ρc2

Here and in the following, p denotes the RMS sound pressure in the diffuse
sound field (called pdiff in Section 4.2.2). The energy absorbed in the room
is the incident sound power per unit area (Equation 4.24) multiplied by the
total surface area and the mean absorption coefficient, i.e. the equivalent
absorption area (Equation 4.25),

p2
Pa,abs = Iinc Sα m = Iinc A = A (4.27)
4ρ c

If Pa is the sound power of a source in the room, the energy balance equa-
tion of the room is

dE
Pa − Pa,abs =
dt
 Introduction to room acoustics  83

p2 V d 2
Pa − A= (p ) (4.28)
4ρ c ρc2 d t

With a constant sound source, a steady state situation is reached after some
time, and the right side of the equation is zero. Hence, the absorbed power
equals the power emitted from the source, and the steady state sound pres-
sure in the room is

4Pa
ps2 = ρc (4.29)
A
This equation shows that the sound power of a source can be determined by
measuring the sound pressure generated by the source in a room, provided
that the equivalent absorption area of the room is known. It also shows
how the absorption area in a room has a direct influence on the sound pres-
sure in the room. In some cases, it is more convenient to express Equation
4.29 in terms of the sound pressure level Lp and the sound power level LW,

 4A 
Lp ≅ LW + 10 lg  0  ( dB ) (4.30)
 A 

where A 0 = 1 m 2 is a reference area. The approximation comes from neglect-
ing the term with the constants and reference values:

ρ cPref 1.204 × 343 × 10−12

10 lg = 10 lg = 0.14 dB ≅ 0 dB

2
A0 pref 1 × (20 × 10−6 )2

4.2.5 Reverberation time: Sabine’s formula

If the sound source is turned off after the sound pressure has reached the sta-
tionary value, the first term in the energy balance equation (Equation 4.28)
is zero, and the RMS sound pressure is now a function of time:

A 2
4ρ c
V d 2
p (t) + 2
ρc d t
p (t) = 0 ( ) (4.31)

The solution to this equation can be written as
cA
− (t – t 0 )
p2 (t) = ps2e 4V (4.32)

where ps2 is the mean square sound pressure in the steady state and t 0 = 1 s is
the time when the source is turned off. It can be seen that the mean square
sound pressure, and hence, the sound energy, follow an exponential decay
84  Sound Insulation in Buildings

function. On a logarithmic scale, the decay is linear, and this is called the
decay curve (Figure 4.10).
If instead the source is turned on at time t = 0, the sound build up in the
room follows a similar exponential curve, also shown in Figure 4.10.

 −
cA
t
p2 (t) = ps2  1 − e 4V  (4.33)
 

The reverberation time T is defined as the time it takes for the sound energy
in the room to decay to one millionth of the initial value, i.e. a 60 dB decay
of the sound pressure level. Hence, for t = T

(a)
1
Relative sound pressure squared

0.8

0.6

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3
Time, s
(b)
0
Relative sound pressure level, dB

–10

–20

–30

–40

–50

–60
0 0.5 1 1.5 2 2.5 3
Time, s

Figure 4.10 Build up and decay of sound in a room. Here, the source is turned on at t = 0
and turned off at t0 = 1 s. The reverberation time is 1.8 s. (a) Linear scale
(sound pressure squared), and (b) logarithmic scale (dB).
 Introduction to room acoustics  85

cA
− T
p2 (t) = ps210−6 = ps2 e 4V

Hence, the reverberation time is

4V 55.3V
T = 6 ⋅ ln(10) ⋅ = (4.34)
cA cA

This is Sabine’s formula named after Wallace C. Sabine (1868–1919). He

introduced the reverberation time concept in a pioneering paper published
first time 1898 in Proceedings of the American Institute of Architects.
(Paper one in collected papers, Sabine, 1922.). He was the first to demon-
strate that T is inversely proportional to the equivalent absorption area A.
Note: Sabine’s formula is often written as T = 0.16 V/A. However, this
implies that V must be in m3 and A in m 2 .
There is no doubt that Sabine’s formula is the most well-known formula
in room acoustics, and it is widely used to calculate reverberation time.
However, the preconditions for the formula to be valid are often neglected:

• The sound field should be a three-dimensional, diffuse sound field.

• The absorption should be evenly distributed on all surfaces.

None of these conditions are fulfilled in ordinary rooms like classrooms and
offices that typically have a sound absorbing ceiling and highly reflecting
walls.
In Figure 4.10 we see the rapid build-up of sound energy after the source
is turned on. The point of −3 dB (half of the asymptotic final energy) is
reached after 5 % of the reverberation time, −1 dB after 11 % and −0.1 dB
after 27 %. This means that using a stationary sound source, the sound
field can be considered stationary after one-third of the reverberation time,
and measurements can start.

4.2.6 Stationary sound field in a room:

Reverberation distance
A reverberation room is a special room with long reverberation time and a
good diffusion. In such a room, the diffuse sound field is a good approxi-
mation, and the results for stationary conditions (Equation 4.29) and for
sound decay (Equation 4.34) can be applied to measure the sound power
of a sound source:

ps2 55.3V
Pa = ⋅ (4.35)
4 ρc cT

The reverberation time and the average sound pressure level in the rever-
beration room are measured, and the sound power level is calculated from
86  Sound Insulation in Buildings

2
pref ⋅ 55.3 ⋅ V
LW = Lp + 10lg
Pref ⋅ 4 ρc 2 ⋅ T
(4.36)
V T
= Lp + 10lg − 10lg − 14dB
V0 t0

where V0 = 1 m3 and t 0 = 1 s.

In most ordinary rooms, the diffuse sound field is not a good approxima-
tion. Each of the following conditions may indicate that the sound field is
not diffuse:

• An uneven distribution of sound absorption on the surfaces, e.g. only

one surface is highly absorbing
• A lack of diffusing or sound scattering elements (e.g. furniture) in the
room
• The ratio of longest to shortest room dimension is higher than three
• The volume is very large, say more than 5000 m3

A rather simple modification to the stationary sound field is to separate the

direct sound. The sound power radiated by an omnidirectional source is
the sound intensity at the distance r in a spherical sound field multiplied by
the surface area of a sphere with radius r

Pa = I r ⋅ 4π r 2 (4.37)

Thus, the sound pressure squared of direct sound in the distance r from the
source is

Pa
2
pdir = ρc (4.38)
4π r 2

The stationary sound is described by Equation 4.29:

4Pa
ps2 = ρc
A

2
The reverberation distance rrev is defined as the distance where pdir = ps2
when an omnidirectional point source is placed in a room. It is a descrip-
tor of the amount of absorption in a room since the reverberation distance
depends only on the equivalent absorption area:

A (4.39)
rrev = = 0.14 A
16π
 Introduction to room acoustics  87

At a distance closer to the source than the reverberation distance, the direct
sound field dominates, which is called the direct field. At longer distances,
the reverberant sound field dominates, and in this so-called far field, the
stationary, diffuse sound field may be a usable approximation.
An expression for the combined direct and diffuse sound field can
derived by the simple addition of the squared sound pressures of the two
sound fields. However, since direct sound is treated separately, it should be
extracted from the energy balance equation, which was used to describe
the diffuse sound field. To do this, the sound power of the source should
be reduced by a factor of (1 −αm), which is the fraction of the sound power
emitted to the room after the first reflection. Hence, the squared sound
pressure in the total sound field is

 r2 
2
ptotal 2
= pdir + ps2 (1 − α m ) = ps2  rev2 + 1 − α m  (4.40)
 r 

 1 4 
2
ptotal = Pa ⋅ ρ c  + (1 − α m ) (4.41)
 4π r 2
A 

The absorption area A divided by (1 − αm) is sometimes called the room
constant.
Normal sound sources like a speaking person, a loudspeaker, or a musical
instrument radiate sound with different intensities in different directions.
The directivity factor Q is the ratio of the intensity in a certain direction to
the average intensity,

4π r 2
Q= I⋅ (4.42)
Pa

Hence, the squared sound pressure of the direct sound is

Q ⋅ Pa
2
pdir = ρc (4.43)
4π r 2

This leads to a general formula for the sound pressure level as a function of
the distance from a sound source in room.

4 A0  r2 
Lp ≅ LW + 10 lg + 10 lg  Q rev2 + 1 − α m  ( dB ) (4.44)

A  r 

where A 0 = 1 m 2 . This formula is displayed in Figure 4.11 for the case of
an omnidirectional sound source (Q = 1). In a reverberant room with little
sound absorption (e.g., αm < 0.1), the sound pressure level in the far field
will be approximately as predicted by the diffuse field theory, i.e. the last
88  Sound Insulation in Buildings

15

10

1
5
Relative sound pressure level, dB

0 5

10
–5
20

–10 50

100
–15

∞
–20
0.1 0.2 0.5 1 2 5 10 20
Distance from sound source r, m

Figure 4.11 Relative sound pressure level as a function of distance in a room with an

approximately diffuse sound field. The source has a directivity factor of one.
The parameter on the curves is the room constant A/(1 −αm) in m2 .

term will be close to zero. In the case of a highly directive sound source
like a trumpet (Q  1) , the direct field can be extended to distances much
longer than the reverberation distance. In the latter situation, the last
term in Equation 4.44 raises the sound pressure level above the diffuse
field value.
In large rooms with medium or high sound absorption (e.g., αm > 0.2), the
sound pressure level will continue to decrease as a function of the distance
because the diffuse field theory is not valid in such a room. Instead, the slope
of the spatial decay curve may be taken as a measure of the degree of acoustic
attenuation in a room. Hence, in large industrial halls, the attenuation in dB
per doubling of the distance may be a better descriptor than the reverberation
time.
A reverberation room can be used to determine the sound power of a
sound source by measuring the average sound pressure level in the room.
If measurements are made in positions that avoid the direct sound, the last
term in Equation 4.44 becomes more correctly (Vorländer 1995):
 Introduction to room acoustics  89

 −A  A
10 lg (1 − α m ) → 10 lg  e S  ≅ −4.34 ⋅ ( dB )
  S

where A is the equivalent absorption area, S is the total surface area and tak-
ing the air attenuation into account (see Section 4.3.2). For correct results
at low frequencies, it is also necessary to apply the Waterhouse correction
(Equation 3.31)), and, thus, the sound power level is determined from

A A  cS 
LW = Lp + 10 lg + 4.34. + 10 lg  1 +  (dB) (4.45)
4 A0 S  8Vf

This is how the sound power of sound sources can be measured according
to ISO 3741. In addition, the possible variation of temperature and static
pressure may also be taken into account for very high precision.

4.3 GEOMETRICAL ROOM ACOUSTICS

4.3.1 Sound rays and a general reverberation formula

In geometrical acoustics, rays are used to describe sound propagation.
The concept of rays implies that the wavelength and the phase of the
sound are neglected, and only the direction of sound energy propagation
is treated.

Figure 4.12 A plane wave travelling as a ray from wall to wall in a room.
90  Sound Insulation in Buildings

Sound decay shall now be studied by following a plane wave travelling as

a ray from wall to wall (Figure 4.12). The energy of the wave is gradually
decreased due to absorption at the surfaces, all of which are assumed to
have the mean absorption coefficient αm.
The ray representing a plane wave may start in any direction and it is
assumed that the decay of energy in the ray is representative of the decay of
energy in the room. The room may have any shape.
By each reflection, the energy is reduced by a factor (1 −α m). The ini-
tial sound pressure is p 0 and after n reflections, the squared sound pres-
sure is

p2 (t) = p02 ⋅ (1 − α m )n = p02 ⋅ en⋅ln(1−α m ) (4.46)

The distance the ray travels from one reflection to the next is li and the total
distance travelled by the ray up to the time t is

∑l = c ⋅t = n ⋅l
i m
(4.47)
i

where lm is the mean free path. Hence, the squared sound pressure is

c
⋅ln(1− α m )⋅t
p2 (t) = p02 ⋅ e lm (4.48)

When the squared sound pressure has dropped to 10 −6 of the initial value,
the time t is by definition the reverberation time T:
c
⋅ln(1−α m )⋅T c
10−6 = e lm ⇒ −6 ⋅ ln(10) = ⋅ ln(1 − α m ) ⋅ T
lm

This leads to an interesting pair of general reverberation formulae:

13.8 ⋅ lm 13.8 ⋅ lm
T= ≈ (4.49)
− c ⋅ ln(1 − α m ) c ⋅αm

The last approximation is valid if αm < 0.3, i.e. only in rather reverberant
rooms. The approximation comes from

 1  α 2 α 3
− ln(1 − α m ) = ln   = αm + m + m +
 1− αm  2 3

With the assumption that all directions of sound propagation appear with
the same probability, it can be shown from Kosten (1960) that the mean
free path in a three-dimensional room is
 Introduction to room acoustics  91

4V
lm =
S
( three-dimensional ) (4.50)

where V is the volume and S is the total surface area.

Similarly, the mean free path in a two-dimensional room can be derived.
This could be the narrow air space in a double wall, or structure-borne
sound in a plate. The height or thickness must be small compared to the
wavelength. In this case, the mean free path is

π Sx
lm =
U
( two-dimensional ) (4.51)

where Sx is the area and U is the perimeter. The one-dimensional case is

just the sound travelling back and forth between two parallel surfaces with
the distance l = lm.
Insertion of Equation 4.50 in the last part of Equation 4.35 gives Sabine’s
formula (Equation 4.34), whereas insertion in the first part of Equation 4.49
leads to the so-called Eyring’s formula for reverberation time in a room:

55.3 ⋅ V
T= (4.52)
− c ⋅ S ⋅ ln (1 − α m )

In a reverberant room (αm < 0.3), it gives the same result as Sabine’s for-
mula, but in highly absorbing rooms, Eyring’s formula is theoretically
more correct. In practice, the absorption coefficients are not the same
for all surfaces and the mean absorption coefficient is calculated as in
Equation 4.20:

αm =
1
S
⋅ ∑Sα i i (4.53)
i

In the extreme case of an anechoic room (αm = 1), Eyring’s formula gives

correctly a reverberation time of zero, whereas Sabine’s formula is obvi-
ously wrong, giving the value T = 55.3 V/(c S). However, in normal rooms
with a mixture of different absorption coefficients, it is recommended to
use Sabine’s formula.

4.3.2 Sound absorption in air

A sound wave travelling through air is attenuated by a factor m, which
depends on the temperature and the relative humidity (RH) of the air. The
unit of the air attenuation factor is m−1. If this attenuation is included in
Equation 4.48, the squared sound pressure in the decay is
92  Sound Insulation in Buildings

c ct
⋅ln(1−α m )⋅t ⋅( ln(1−α m )− m⋅lm )
2 2
p (t) = p ⋅ e lm
⋅e − mct 2
= p ⋅e lm (4.54)
0 0

The general reverberation formula then becomes

13.8 ⋅ lm 13.8 ⋅ lm
T= ≈ (4.55)
(
c − ln (1 − α m ) + m ⋅ lm )
c (α m + m ⋅ lm )

In the three-dimensional case with Equation 4.50, we then have

55.3 ⋅ V 55.3 ⋅ V (4.56)

T= ≈
c(−S ⋅ ln(1 − α m ) + 4mV ) c(S ⋅ α m + 4mV )

These two expressions are Eyring’s and Sabine’s formula, respectively,

with the air absorption included. By comparison with Equation 4.34, the
equivalent absorption area including air attenuation is

A= ∑ S α + 4mV
i i (4.57)
i

Some typical values of m are found in Table 4.3.

4.3.3 Reflection density in a room

The image source principle can easily be applied to higher order reflections
in a rectangular room. An infinite number of image rooms make a grid,
and each cell in the grid is an image room containing an image source. The
principle is shown for the two-dimensional case in Figure 4.13.

ct

Figure 4.13 Rectangular room with a sound source and image sources, shown in two
dimensions. Image sources located inside the circle with radius ct will con-
tribute reflections up to time t.
 Introduction to room acoustics  93

If an impulse sound is emitted, the number of reflections that will arrive

within the time t can be calculated as the volume of a sphere with radius ct
divided by the room volume V:

4
π (ct)3
N (t) = 3 (4.58)
V

The reflection density is then the number of reflections within a small time
interval dt, and by differentiation:

dN c3
= 4π t 2 (4.59)
dt V

The reflection density increases with the time squared, so the higher order
reflections are normally so dense in arrival time that it is impossible to
distinguish separate reflections. If Equation 4.59 is compared to Equation
4.14, it is striking to observe the analogy between reflection density in the
time domain and modal density in the frequency domain.

4.4 CALCULATION OF REVERBERATION TIME

4.4.1 Using Sabine’s formula

Sabine’s formula (Equation 4.34) is the most well-known and simple
method for calculation of reverberation time in a room:

55.3V 0.16V (4.60)

T= ≅
cA A

with volume V in m3 and A in m 2 . The equivalent absorption area is cal-

culated as in Equation 4.57, but in addition to absorption from surfaces
and air, the absorption from persons or other items in the room should be
included, if relevant:

A= ∑ S α + ∑ n A + 4mV
i
i i
j
j j (4.61)

Here n j is the number of items, each contributing with an absorption

area A j. Examples of absorption coefficients of common materials are
given in Table 4.2. The air attenuation for different values of RH at
20 °C can be taken from Table 4.3. It is noted that high RH means low
air attenuation.
94  Sound Insulation in Buildings

Table 4.2  Typical values of the absorption coefficient α for some common materials
Frequency (Hz)
Material 125 250 500 1000 2000 4000
Brick, bare concrete 0.01 0.02 0.02 0.02 0.03 0.04
Parquet floor on studs 0.16 0.14 0.11 0.08 0.08 0.07
Needle-punch carpet 0.03 0.04 0.06 0.10 0.20 0.35
Window glass 0.35 0.25 0.18 0.12 0.07 0.04
Curtain draped to half its 0.10 0.25 0.55 0.65 0.70 0.70
area, 100 mm air space

Table 4.3  Examples of air attenuation coefficient m (m−1) at a temperature of 20 °C

Frequency (kHz)
Relative humidity (%) 1 2 4 8
40 0.0011 0.0026 0.0072 0.0237
50 0.0010 0.0024 0.0061 0.0192
60 0.0009 0.0023 0.0056 0.0162
70 0.0009 0.0021 0.0053 0.0143
80 0.0008 0.0020 0.0051 0.0133
Source: Harris, C.M., J. Acoust. Soc. Am., 40, 148, 1966.

4.4.2 Reverberation time in non-diffuse rooms

In a room with the sound absorption unequally distributed on the surfaces,
the assumption of a diffuse sound field is not fulfilled and, thus, Sabine’s
formula will not be reliable. The measured reverberation time may be either
shorter or longer than predicted by Sabine’s formula.
A shorter reverberation time will appear in a room in which the first
reflections are directed towards the most absorbing surface. In an audito-
rium, the most absorbing surface is typically the floor with the audience.
In a rectangular room without sound scattering surfaces or elements,
there is a possibility of prolonged decay in certain directions. In order to
give an idea of the problem, it is possible to calculate the different rever-
beration times associated with one-dimensional decays in each of the three
main directions using the general reverberation formula (Equation 4.49).

13.8 ⋅ lm l
T≈ ≈ 0.04 ⋅ m ( lm in m ) (4.62)
c ⋅αm αm

As an example, the room in Figure 4.14 is considered. The ceiling has a

high absorption coefficient (α = 0.8), but all other surfaces are acoustically
hard (α = 0.1).
 Introduction to room acoustics  95

αfloor = 0.1 αceiling = 0.8

αwalls = 0.1

5m

20 m
10 m

Figure 4.14 A rectangular room with indicated absorption coefficients.

Volume V = 5 × 10 × 20 = 1000 m3
Surface area S = 700 m 2
Equivalent absorption area A = 200 × 0.8 + 500 × 0.1 = 210 m 2
Mean absorption coefficient αm = A/S = 210/700 = 0.30
Mean absorption coefficient (height) αm = (0.8 + 0.1)/2 = 0.45
Mean free path (three-dim.) lm = 4V/S = 4 × 1000/700 = 5.7 m
Mean free path (two-dim.) lm = πSx /U = π × 200/60 = 10.5 m

The results are shown in Table 4.4. A two-dimensional reverberation in

the horizontal plane between the walls has also been calculated (4.2 s). The
one-dimensional decays are the extreme cases with the longest reverbera-
tion time being 20 times the shortest one, 8.0 s and 0.4 s, respectively.
The real decay that is measured in the room will be a mixture of these dif-
ferent decays, and the reverberation time will be considerably longer than pre-
dicted from Sabine’s formula. Eyring’s formula is even worse. The measured
decay curve will be bent, and thus, the measuring result depends on which
part of the decay curve is considered for the evaluation of reverberation time.
In a room with long reverberation time due to non-diffuse conditions and at
least one sound-absorbing surface, introducing some sound scattering elements

Table 4.4  Calculation of the one-dimensional reverberation times of the

rectangular room in Figure 4.14
Direction lm (m) αm T (s)
Three-dimensional (Sabine) 5.7 0.30 0.8
Three-dimensional (Eyring) 5.7 0.30 0.6
Two-dimensional (horizontal) 10.5 0.10 4.2
One-dimensional (length) 20 0.10 8.0
One-dimensional (width) 10 0.10 4.0
One-dimensional (height) 5 0.45 0.4
96  Sound Insulation in Buildings

in the room can have a significant effect. It could be furniture or machines on

the floor or some diffusers on the walls. This will make the sound field more
diffuse, and the reverberation time will be reduced, i.e. it will come closer to
the Sabine value. In other words: The sound absorption available in the room
becomes more efficient when scattering elements are introduced to the room.
Note: In the one-dimensional case, it is strictly not correct to use the
arithmetic average of the absorption coefficients for calculating the rever-
beration time of the axial mode. By inspection of Equation 4.46, it can be
seen that the mean absorption coefficient should be calculated from

(1 − α m ) = (1 − α1 )(1 − α 2 ) (4.63)

Hence, if one of the surfaces is is totally absorbing, αm = 1 and, therefore,

the reverberation time is zero.

4.5 MEASUREMENT OF REVERBERATION TIME

The reverberation time in a room can be measured with a noise signal

or with an impulse. Actually, the modern measurement technique is to
measure the impulse response using a sine-sweep or an maximum length
sequence signal, see also (see Chapter 12). Both methods (interrupted noise
and integrated squared impulse response) are described in the measurement
standard ISO 3382-2. The traditional method uses pink noise or white
noise emitted by a loudspeaker and a microphone to measure the sound
pressure level as a function of time after the source is switched off. This
gives a decay curve. A typical example is shown in Figure 4.15.
From the microphone, the signal is led to a frequency filter, which is either an
octave filter or a one-third octave filter. If the sound in the room is adequately
diffuse and a sufficiently large number of modes are excited, the decay curve is

SPL, dB
Noise on Noise off

Average stationary level

Decay

Background
noise level

t=0 Time, s

Figure 4.15 Typical decay curve measured with noise interrupted at time t = 0. SPL,
sound pressure level.
 Introduction to room acoustics  97

close to a straight line between the excitation level and the background level.
The dynamic range is seldom more than around 50 dB and the whole range
of the measured decay curve is not used. The lower part of the decay curve is
influenced by the background noise and the upper part may be influenced by
the direct sound, which gives a steeper start of the curve. Hence, the part of the
decay curve used for evaluation begins 5 dB below the average stationary level
and ends normally 25 dB or 35 dB below the same level. The evaluation range
is, thus, either 20 dB or 30 dB and the slope is determined by fitting a straight
line or automatically by calculating the slope of a linear regression line. From
the slope of the decay curve in dB per second, the reverberation time is calcu-
lated, being the time for a 60 dB drop following the straight line. According to
ISO 3382-2, the preferred evaluation range is 20 dB and the result is denoted as
T20. Similarly, T30 is used if the 30 dB evaluation range has been applied. If the
sound field in the room is sufficiently diffuse, the decay curves are reasonably
straight and T20 and T30 are approximately the same.
Sometimes, the decay curves are not nice and straight and it is difficult
to measure a certain reverberation time; T20 and T30 yield different results.
One reason can be that it is a measurement at low frequencies in a small
room and maybe only a few modes are excited within the frequency band
of the measurement. In this case, interference between the modes can cause
very irregular decay curves.
The reverberation time is measured in a number of source and receiver
positions, often two source positions and three microphone positions, i.e.
six combinations are used. If the interrupted noise method is used, several
decays should be measured and averaged in each position because the noise
signal is a stochastic signal and, thus, the measured decay curves are always a
little different. However, if the integrated squared impulse response method
is used, a single measurement in each position is sufficient; the decay curve
obtained with this method corresponds to averaging an infinite number of
decays with the interrupted noise method (Schroeder, 1965). Figure 4.16
shows an example of a measured squared impulse response (the curve with
fluctuations) and the corresponding integrated decay curve.
Annex B of ISO 3382-2 contains two parameters that are useful for
describing the quality of reverberation time measurements. One is the
degree of non linearity of the part of the decay curve within the evaluation
range. Since the evaluation of the decay curve is made by a least-squares fit
of a straight line, the correlation coefficient r is known. The non linearity
parameter ξ is defined as the permillage deviation from perfect linearity:

(
ξ = 1000 1 − r 2 ) (‰) (4.64)

Values higher than 10 ‰ indicate a decay curve being far from a straight
line, and the value of the reverberation time derived from the curve may be
suspicious.
98  Sound Insulation in Buildings

–10

–20
SPL, dB
–30

–40

–50

0 0.5 1 1.5
Time, s

Figure 4.16 Typical decay curve measured using the integrated squared impulse response
method. The lower curve is the squared impulse response and the upper
curve is the decay curve.

The other quality parameter is the degree of curvature C defined as

T 
C = 100  30 − 1 (%) (4.65)
 T20 

Values higher than 10 % indicate a bent decay curve; the decay may be a
combination of room modes with very different reverberation times.

REFERENCES

D.A. Bies and C.H. Hansen (1988). Engineering Noise Control, Theory and Practice,
Unwin Hyman Ltd., London.
C.M. Harris (1966). Absorption of sound in air versus humidity and temperature.
Journal of the Acoustical Society of America 40, 148–159.
ISO 3382-2 (2008). Acoustics – Measurement of room acoustic parameters – Part
2: Reverberation time in ordinary rooms, International Organization for
Standardization, Geneva, Switzerland.
ISO 3741 (2010). Acoustics – Determination of sound power levels and sound
energy levels of noise sources using sound pressure – Precision methods for
reverberation test rooms, International Organization for Standardization,
Geneva, Switzerland.
C.W. Kosten (1960). The mean free path in room acoustics. Acustica 10, 245–250.
H. Kuttruff (1973). Room Acoustics, Applied Science Publishers, London.
A.R. Molares (2010). A Monte Carlo approach to the analysis of uncertainty in
acoustics. PhD Dissertation, Escola Técnica Superior de Enxeñeiros de
Telecomunicación Universidade de Vigo, Spain.
A.D. Pierce (1989). Acoustics. An Introduction to Its Physical Principles and
Applications. 2nd Edition, The Acoustical Society of America, New York.
 Introduction to room acoustics  99

M.R. Schroeder (1962). Frequency correlation functions of frequency responses in

rooms. Journal of the Acoustical Society of America 34, 819–1823.
M.R. Schroeder (1965). New method of measuring reverberation time, Journal of
the Acoustical Society of America 37, 409–412.
J.H. Rindel (2015). Modal energy analysis of nearly rectangular rooms at low fre-
quencies. Acta Acustica United with Acustica 101, 1211–1221.
J.H. Rindel (2016). A note on modal reverberation times in rectangular rooms. Acta
Acustica United with Acustica, 102, 600–603.
W.C. Sabine (1922). Collected Papers on Acoustics. Harvard University Press.
Published 1964 by Dover Publications Inc., New York.
M. Vorländer (1995). Revised relation between the sound power and the average
sound pressure level in rooms and consequences for acoustic measurements.
Acustica 81, 332–342.
Chapter 5

Introduction to sound insulation

This chapter describes the basic principles of sound insulation with simpli-
fied assumptions. The methods for measuring airborne and impact sound
insulation are briefly described. Also presented are the methods of objective
evaluation of the measurement results. As a supplement to the fundamental
theoretical models, the principles are illustrated through examples of mea-
surement results on typical constructions.

5.1 AIRBORNE SOUND INSULATION

5.1.1 The sound transmission loss

A sound wave incident on a wall or any other surface separating two adja-
cent rooms is partly reflected back to the source room, partly dissipated as
heat within the material of the wall, propagates partly to other connecting
structures and is partly transmitted into the receiving room. Note that it is
important to distinguish between sound insulation and sound absorption.
Sound insulation has to do with the reduction of sound energy when sound
is transmitted through a barrier or a wall, whereas sound absorption has to
do with the sound energy that is lost (to heat) in some material. A material
that is a good sound absorber is not necessarily a good material for sound
insulation and vice versa. In fact, a good sound absorber is typically light
and porous, whereas a good sound insulator is heavy and airtight.
If the power incident on the wall is P1 and the power transmitted into the
receiving room is P2 , the sound transmission coefficient τ is defined as the
ratio of transmitted to incident sound power:

P2
τ= (5.1)
P1

However, the sound transmission coefficients are typically very small num-
bers, and it is more convenient to use the sound reduction index R with the
unit decibel (dB). It is defined as:

101
102  Sound Insulation in Buildings

P1 1
R = 10 lg = 10 lg = −10 lg τ (dB) (5.2)
P2 τ

Another name for the same term is sound transmission loss, which will be
used throughout this book.

5.1.2 Sound insulation of a partition between

two rooms
Consider the sound insulation between two rooms. With the assumption
of diffuse sound fields in both rooms, it is possible to derive a simple rela-
tion between the transmission loss and the sound pressure levels in the
two rooms. The rooms are called the source room and the receiving room,
respectively. In the source room is a sound source that generates the average
sound pressure p1. The sound power (see Equation 4.24) incident on the
wall is (Figure 5.1),

p12S
P1 = Iinc S = (5.3)
4ρ c

The area of the wall is S. In the receiving room, the sound pressure
p 2 is  generated from the sound power P 2 radiated into the room (see
Equation 4.29),

4P2
p22 = ρc (5.4)
A2

A 2 denotes the absorption area in the receiving room. Insertion in Equation

5.2 gives

Room 1 Room 2

P1 P2

A2

L1 L2

Figure 5.1 Airborne sound transmission from source room (1) to receiving room (2).
 Introduction to sound insulation  103

p12S p2 S S
R = 10 lg = 10 lg 21 = L1 − L2 + 10 lg (dB) (5.5)
4 ρ cP2 p2 A2 A2

L1 and L 2 are the spatially averaged sound pressure levels in the source and
receiving rooms, respectively. This important result is the basis for trans-
mission loss measurements.

5.1.3 Measurement of sound insulation

Sound insulation is measured in one-third octave bands covering the
­frequency range from 100 Hz to 3150 Hz. In recent years, the international
standards for measurement of sound insulation have been revised, and it
is recommended to extend the frequency range down to 50 Hz and up to
5000 Hz. One reason for this is that the low frequencies 50 Hz to 100 Hz
are very important for the subjective evaluation of the sound insulation
properties of lightweight constructions. Lightweight constructions have
been more commonly used in new building technology, whereas heavy con-
structions have traditionally been used for sound insulation.
The sound pressure levels are measured as the average from a number of
microphone positions or as the average from microphones slowly moving
on a circular path (if mechanically operated) or on other paths (if manually
operated). The results are usually averaged over two different source posi-
tions. More details can be found in Chapter 12.
In addition to the average sound pressure levels in the source and receiv-
ing rooms, it is also necessary to measure the reverberation time in the
receiving room to calculate the absorption area. Sabine’s equation is used
for this (see Equation 4.34):

55.3V2
A2 = (5.6)
cT2

where V2 and T2 are the volume and reverberation time of the receiving room,
respectively, and c is the speed of sound. Only under special laboratory con-
ditions can the transmission loss of a wall be measured without influence
from other transmission paths. In a normal building, sound will not only be
transmitted through the separating construction but the flanking construc-
tions will also influence the result, as discussed later in Section 5.5.
For measurements of sound insulation in buildings, the apparent sound
reduction index is

S
R′ = L1 − L2 + 10 lg (dB) (5.7)
A2
The apostrophe after the symbol indicates that flanking transmission can
be assumed to influence the result. Thus, R′ is typically a few dB less than
104  Sound Insulation in Buildings

R for the same partition wall measured in a laboratory. For the character-
ization of sound insulation in buildings, the R′ has a problem – the area
S of the partition must be known, and this is not always the case. For
example, the source and receiving rooms may be located with an offset
(staggered rooms) so that the area S does not exist or is very small? In the
latter case, Equation 5.7 would yield a very low result, not corresponding
to the ­perceived sound insulation.
To overcome this problem in field measurements, the area S can be set
to a constant value of 10 m 2 , and the result using Equation 5.7 is called the
normalized level difference Dn. An alternative measure that overcomes the
problem with the partition area is the standardized level difference DnT :

T2
DnT = L1 − L2 + 10 lg (dB), T0 = 0.5 s (5.8)
T0

This is the level difference between the source and receiving rooms if the
reverberation time of the receiving room is 0.5 s at all frequencies. This rever-
beration time is typical for a furnished living room in a dwelling. According
to ISO 16283-1, the standardized level difference DnT provides a straightfor-
ward link to the subjective impression of airborne sound insulation.
the relation between DnT and R′ is approximately: From Equations 5.6
through 5.8,

V2
DnT = R′ + 10 lg (dB), d0 = 3 m (5.9)
S ⋅ d0

where d 0 is a reference distance, V2 is the volume of the receiving room and

S is the area of the partition (if any). In the case of a rectangular receiving
room where one of the walls is the partition wall, the relation depends on the
depth of the receiving room perpendicular to the partition wall, d2 = V2/S.

If d2 < 3 m; DnT < R′

If d2 ≈ 3 m; DnT ≈ R′
If d2 > 3 m; DnT > R′

As a consequence, DnT depends on the measurement direction if the size

of the source room and the receiving room is different. For testing to show
compliance with a minimum airborne sound insulation requirement, the
smaller room should be used as the receiving room to get the lowest result.

5.1.4 Multi-element partitions and apertures

A partition is often divided into elements with different sound insulation
properties, e.g. a wall with a door. Each element is described by the area
Si and the transmission coefficient τi. If the sound intensity incident on
 Introduction to sound insulation  105

the surfaces of the source room is denoted by Iinc, the total incident sound
power on the partition is
n

P1 = ∑SI i inc = SIinc

i =1

The total area of the partition is called S. The total sound power transmit-
ted through the partition is
n

P2 = ∑τ S I i i inc

i =1

Thus, the transmission coefficient of the partition is

n

τ res =
P2 1
=
P1 S ∑ τ S (5.10)
i =1
i i

The same result can also be written in terms of the transmission losses Ri
of each element
1 n

Rres = −10 lg τ res = −10 lg 
S
∑ S 10
i =1
i
−0,1Ri
 (5.11)


In the simple case of only two elements, the graph in Figure 5.2 may be used.

0 S2/(S1+S2), %
20 1

2
15
5
Rres–R2 (dB)

10 10

20

5
40

80
0
0 5 10 15 20 25 30
R1–R2 (dB)

Figure 5.2 Estimating the transmission loss of a multi-element partition.

106  Sound Insulation in Buildings

Sap/S
60 10–6

50 10–5

40 10–4
Rres (dB)

30 10–3

20 10–2

10 10–1

0
0 10 20 30 40 50 60 70 80
R1 (dB)

Figure 5.3 Estimating the transmission loss of a construction with an aperture.

An aperture in a wall is a special example of an element with different

transmission properties. As an approximation, it can be assumed that the
transmission coefficient of the aperture is 1. If the area of the aperture S ap is
also very small compared to the total area, the resulting transmission loss
of the wall with aperture will be:

1
( 
) 
Rres = −10 lg  S110−0,1R1 + Sap  ≅ −10 lg  10−0,1R1 +
S  
Sap 
S 
(5.12)

Figure 5.3 can illustrate the result. The relative area of the aperture defines
an upper limit of the sound insulation that can be achieved.

5.2 SINGLE LEAF CONSTRUCTIONS

5.2.1 Sound transmission through a solid material

Consider a solid material shaped as a large plate with thickness h. The
material is characterized by the density ρm and the speed of longitudinal
waves cL (Equation 1.20). The surface of the material defines two transi-
tion planes, where the sound waves change from one medium to another.
It is assumed that the medium on either side is air with density ρ and the
 Introduction to sound insulation  107

speed of sound c (also longitudinal waves). The symbols and notation are
explained in Figure 5.4.
The sound pressure is equal on either side of the two transition planes:

pi + pr = p1 + p4

pt = p2 + p3 (5.13)

Also, the particle velocity is equal on either side of the two transition planes:

ui − ur = u1 − u4

ut = u2 − u3 (5.14)

The characteristic impedance in the surrounding medium (air) is denoted

by Z 0 and in solid material it denoted by Zm. Thus, the ratio of sound pres-
sure to particle velocity in each of the plane propagating waves is:

pi pr pt
= = = Z0 = ρ c
ui ur ut
p1 p2 p3 p4
= = = = Zm = ρ mcL (5.15)
u1 u2 u3 u4

Using Equation 5.15 in Equation 5.14 leads to:

Z0
pi − pr =
Zm
( p1 − p4 )
Z0
pt =
Zm
( p2 − p3 ) (5.16)

pi p1 p2 pt

pr p4 p3

Figure 5.4 Thick wall with incident, reflected and transmitted sound waves. p1, p2 , p3 and
p4 denote the sound pressures of the longitudinal waves in the solid material,
incident and reflected at the boundaries. pi, pr and pt denote the incident,
reflected and transmitted sound pressures in the air.
108  Sound Insulation in Buildings

Assuming propagation from one side of the material to the other without
losses means that there is only a phase difference between the pressures at
the two intersections:

p2 = p1e− jkm h
p4 = p3e− jkm h (5.17)

km = ω/cL is the angular wave number for longitudinal sound propagation in

the solid material.
From Equations 5.13, 5.16 and 5.17, the ratio between the sound pres-
sures pi and pt and, thus, the transmission loss at normal incidence can be
expressed by:
2
p   Z0 Zm 
2

R0 = 10 lg i 2 1
= 10 lg  cos (km h) +  + sin2 (km h) (5.18)
pt 
4  Zm Z0 
 

The resulting transmission loss is shown in Figure 5.5, where some dips
can be observed at high frequencies. They occur at frequencies where the
thickness is equal to half a wavelength in the solid material, or a multiple of
half wavelengths. However, the dips are very narrow and they are mainly
of theoretical interest.

100

80

60
R0 (dB)

40

20

0
10 100 1,000 10,000
Frequency (Hz)

Figure 5.5 Transmission loss at normal incidence of sound on a 600 mm thick concrete wall.
 Introduction to sound insulation  109

Two special cases can be studied. First, the case of a thin wall: Zm ≫ Z 0
and kmh ≪ 1

  Z 2    ωρ h  2 
R0 ≅ 10 lg  1 +  m  sin2 (km h) ≅ 10 lg  1 +  m   (5.19)
  2Z0     2ρ c  

The other special case is a very thick wall: Zm ≫ Z 0 and kmh ≫ 1

2
 Z  ρ c 
R0 ≅ 10 lg  m  ≅ 20 lg  m L  (5.20)
 2Z0   2ρ c 

The cross-over frequency from Equation 5.19 to Equation 5.20 is the fre-
quency f h at which kmh = 1:

cL
fh = (5.21)
2π h
This is the frequency at which the thickness is approximately one-sixth of
the longitudinal wavelength λL in the material:

cL λ
h= = L
2π f 2π

The result for the thin wall (Equation 5.19) is the so-called mass law, which
will be derived in a different way in the next section. The result for a very
thick wall (Equation 5.20) means that there is an upper limit on the sound
insulation that can be achieved by a single-leaf construction, and this limit
depends on the density of the material. For wood, it is 68 dB, for concrete
80 dB and for steel 94 dB. These numbers should be reduced by 5 dB in the
case of random incidence, instead of normal incidence – see Section 5.2.3.

5.2.2 The mass law

Consider a thin wall with the mass per unit area m (see Figure 5.6). The
application of Newton’s second law (force = mass × acceleration) gives:

dvn
∆ p = pi + pr − pt = m = jω mvn (5.22)
dt

where vn is the velocity of the wall vibrations (in the direction normal to the
wall). Wall impedance Zw is introduced:

∆p
Zw = = jω m (5.23)
vn
110  Sound Insulation in Buildings

vn

pi
vt = vn /cos θ

θ
pt
pr

Figure 5.6 Thin wall with sound pressures and particle velocities.

The wall impedance will be more complicated if the bending stiffness of the
wall is also taken into account (see Section 5.2.4).
The particle velocities in the sound waves are called u with the same
indices as the corresponding sound pressures. Due to the continuity
requirement, the normal component of the velocity on both sides of the
wall is:

vn = ut cos θ = (ui − ur ) cos θ (5.24)

which leads to:

vn
pt = pi − pr = Z0 (5.25)
cos θ

The sound transmission loss R θ at a certain angle of incidence θ is:

2 2
pi Zw cos θ
Rθ = 10 lg = 10 lg 1 + (dB) (5.26)
pt 2Z0

In the special case of normal sound incidence (θ = 0), the insertion of

Equation 5.23 gives the important mass law of sound insulation:

2
ωm  π fm 
R0 = 10lg 1 + j ≅ 20lg   (dB) (5.27)
2ρc  ρc 

Since m = ρmh, this result is the same as that derived earlier in a different
way (Equation 5.19). The mass law for sound insulation was first suggested
in a dissertation by Berger (1911).
 Introduction to sound insulation  111

5.2.3 Sound insulation at random incidence

The transmission coefficient at the angle of incidence θ is, from Equation 5.26,

1
τ (θ ) = 2 (5.28)
 ωm
1+  cos2 θ
 2ρ c 

Random incidence means that the sound field on the source side of the
partition is approximately a diffuse sound field. In a diffuse sound field, the
incident sound power P1 on a surface is found by integration over the solid
angle ψ = 2π, assuming the same sound intensity I1 in all directions. Since,
in each direction, the transmitted sound power is equal to the incident
sound power multiplied by the transmission coefficient, the ratio between
transmitted and incident power is:
π /2

P ∫ψ = 2π
τ (θ )I1S dψ ∫ τ (θ) cos θ sin θ dθ
τ= 2 = = 0

∫
π /2
P1 I1S dψ
ψ = 2π
∫ cos θ sin θ dθ
0

π /2 1

=2
∫0
∫
τ (θ ) cos θ sin θ dθ = τ (θ )d(cos2 θ )
0

( )
2
d(cos2 θ ) 2ρ c 
∫ ln 1 + (ω m 2ρ c )
2
= = 
1 + (ω m 2ρ c ) cos θ  ω m 
2 2
0

R = −10 lg τ = R0 − 10 lg (0, 23R0 ) (dB) (5.29)

where R0 is the mass law for normal incidence, Equation 5.27. This is the
theoretical result for random incidence, and for typical values (R0 between
30 dB and 60 dB), it means that R is 8 dB to 11 dB lower than R0. However,
this is not true in real life, and it can be shown that the aforementioned
result is related to partitions of infinite size. Taking the finite size into
account, the result is approximately:

R ≅ R0 − 5 dB (5.30)

The transmission loss at random incidence is approximately the normal

incidence transmission loss less 5 dB. This is in good agreement with mea-
suring results on real walls. The sound transmission through finite-size
walls will be described in Chapter 8.
112  Sound Insulation in Buildings

5.2.4 The critical frequency

The bending stiffness of a beam with height h is the product of Young’s mod-
ulus E and the moment of inertia, h3/12. In a plate, E must be replaced by
E/(1−μ2), and the bending stiffness per unit length of a plate with thickness h is:

Eh3
B= (5.31)
12(1 − µ 2 )
where E is Young’s modulus of the material and μ is Poisson’s ratio (μ ≅ 0.3
for most rigid materials).
The speed of propagation of bending waves in a plate with bending stiff-
ness per unit width B and mass per unit area m is (see Equation 1.25):

B f
cB = ω 4 =c (5.32)
m fc

fc is introduced as the critical frequency and is defined as the frequency at

which the speed of the bending waves equals the speed of sound in air, cB = c.
The critical frequency is:

c2 m
fc = (5.33)
2π B
A sound wave with the angle of incidence θ propagates across the wall
with the phase speed c/sin θ, i.e. the phase speed is in general higher than c
(Figure 5.7). If the bending wave speed happens to be equal to the phase
speed of the incident sound wave, this is called coincidence:

c pr pt c
θ θ

c /sin θ

c
pi θ

Figure 5.7 Thin wall with bending wave and indication of speed of propagation along the wall.
 Introduction to sound insulation  113

cB = c / sin θ
The coincidence leads to a significant dip in the sound transmission loss.
The coincidence dip will be at the coincidence frequency, which is higher
than or equal to the critical frequency:

fco = fc sin2 θ
(5.34)
The coincidence effect was first described by Lothar Cremer (1905–1990)
in a pioneering paper (Cremer 1942). Taking the bending stiffness into
account, the wall impedance (Equation 5.23) is replaced by:

  2 
f
Zw = jω m 1 −   sin4 θ  (5.35)
  fc  
 
Insertion in the general form, Equation 5.26, leads to the sound transmis-
sion loss at a certain angle of incidence:

Rθ = R0 + 20 lg cos θ + 20 lg 1 − ( f fc ) sin4 θ (dB) (5.36)

2

The frequency course of the transmission loss at different angles of inci-

dence is displayed in Figure 5.8. At normal incidence (θ = 0), the transmis-
sion loss increases monotonically with frequency, but at all other angles
of incidence, there is a coincidence dip at the coincidence frequency
(Equation 5.34). At grazing incidence, i.e. when θ approaches 90°, the
coincidence dip is at the critical frequency, but at 45°, the coincidence
dip is one octave higher at f = 2 fc. Another interesting observation from
Equation 5.36 is that the transmission loss is extremely low near grazing
incidence. If we go back to the original formula, Equation 5.26, we see
that R θ = 0 dB at 90°. This is a theoretical result that obviously disagrees
with reality. In Chapter 8, we shall derive a more realistic solution by tak-
ing the finite area into account.
The three-dimensional contour plot in Figure 5.9 shows the transmis-
sion loss of a single panel as a function of frequency along one axis and
as a function of the angle of incidence along the other axis. At normal
incidence, there is a monotone increase with frequency and no coincidence
dip. At other angles of incidence, there is a dip at the coincidence frequency
given by Equation 5.34.

5.2.5 A simple model of sound insulation

of single constructions
The simple model of sound insulation is based on the mass law as given
in Equation 5.27, but with the influence of the dip around the critical
114  Sound Insulation in Buildings

70

60

50
Transmission loss, Rθ (dB)

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

θ = 0° θ = 30° θ = 45° θ = 80°

Figure 5.8 Transmission loss of a single panel as a function of frequency for different

angles of incidence. The panel has m = 25 kg/m2 and fc = 1000 Hz.

frequency. Only bending waves are assumed. This model does not include
the influence of a finite area, the normal modes at low frequencies, nor
the shear waves at high frequencies. These topics will be dealt with in
Chapter 8.
The following results are valid for sound insulation between rooms
with approximately diffuse sound fields. In the frequency range below the
critical frequency, f < fc , the sound insulation is mass-controlled and the
approximate transmission loss is:

R ≅ R0 + 20 lg 1 − ( f fc ) − 5 dB (5.37)
2

 Introduction to sound insulation  115

100
90
80
70
60
Rθ (dB)

50
40
30
20
10
0
0
An 3
gle 0
of 0
in 1000
cid 60
en 1000
ce 90 Hz)
,θ 100 ncy (
Freque

Figure 5.9 Contour plot of the transmission loss of a single panel as a function of angle
of incidence and frequency. The step between contours is 10 dB. The panel
has m = 25 kg/m2 and fc = 1000 Hz.

In the frequency range above the critical frequency, f  ≥  fc , the sound insula-
tion is stiffness-controlled and the resonant transmission due to the struc-
tural modes dominates the transmission. The approximate transmission
loss in this region is:

2η f
R ≅ R0 + 10 lg (dB) (5.38)
π fc

where η is the loss factor (see Equation 2.28). This result will be derived
and explained later in Chapter 7.
A sketch of the transmission loss as a function of frequency is shown
in Figure 5.10. Well below the critical frequency, the course increases by
6 dB per octave, but it changes towards the dip at the critical frequency as
described by Equation 5.37. The maximum transmission loss below fc will
be at f = fc / 3 ≅ 0.58 ⋅ fc . Insertion of this frequency and Equation 1.28
in Equation 5.37 leads to:

 c2 ρ 3 

ρ E 
( )
Rmax, f < fc = 10 lg  2 m 1 − µ 2  − 8.5 dB (5.39)

This is a theoretical maximum, and for some building materials (Table 1.4),
this can be calculated: around 33 dB for gypsum board; 35 dB for glass,
116  Sound Insulation in Buildings

R (dB)

fc
Frequency (Hz)

Figure 5.10 Simplified sound insulation of a single-leaf construction, fc is the critical fre-

quency and the horizontal dotted line at A indicates the maximum below
fc (Equation 5.39). The dotted line at B indicates the upper limit due to lon-
gitudinal waves (Equation 5.20).

brick and porous concrete; 37 dB for concrete; 44 dB for steel and 59 dB for
lead. In reality, the maximum value may be a few dB lower because reso-
nant transmission contributes to the total sound transmission, especially
when the physical dimensions are relatively small, like windows, or when
the plate is supported by laths, like gypsum boards. Hence, a maximum
around 30 dB to 32 dB is typical for single layers of gypsum board and glass.

5.2.6 Examples of single constructions

Some examples of measured transmission loss of single constructions will
illustrate the physical principles explained so far.
Figure 5.11 shows the effect of increasing the thickness h of a glass pane.
This means a shift of the frequency course towards lower frequencies such
that the critical frequency goes down, fc ∼ 1/h (see Equation 1.28). Below fc,
the mass law predicts an increase of R with 6 dB per doubling of h. The weak
point of a single glass pane is the dip around the critical frequency, and for a
thin glass (4 mm), this dip is at a sufficiently high frequency, so it is not a seri-
ous problem. But increasing the thickness means that the frequency region
with weak sound insulation moves down into more important f­ requencies –
in general, solid glass should not be more than 6 mm to 8 mm thick.
If a heavier glass is needed, it should be a laminated glass, which is made
from two or more layers of glass that is glued together with a special soft
material. An example of the transmission loss of a laminated glass is shown
in Figure 5.12. The dip around the critical frequency is almost removed due
to the losses inherent in the lamination.
The previous examples indicate that it may be difficult to take full
advantage of the mass law. However, if we could increase the mass without
 Introduction to sound insulation  117

40

30
Transmission loss, R (dB)

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
6.5 mm glass 4 mm glass

Figure 5.11 Influence of thickness on sound insulation of a single construction. (Adapted

from Lewis, P.T., Building Acoustics, Oriel Press, London, 1971, Figure 7.18.)

40

30
Transmission loss, R (dB)

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
Laminated glass, 2 × 3.2 mm glass with 1.1 mm interlayer 6.5 mm glass

Figure 5.12 Transmission loss of a laminated and a solid glass pane with same mass
­per unit area. (Adapted from Marsh, J.A., Building Acoustics, Oriel Press,
London, 1971, Figure 5.3.)
 118  Sound Insulation in Buildings

increasing the bending stiffness, this might be interesting. The example in

Figure 5.13 is a door leaf with cylindrical cavities, which can be filled with
sand. This doubles the mass without changing the bending stiffness, and
the improvement in sound insulation is much better than the theoretically
expected 6 dB. In addition to the weight, the sand probably also increases
the internal loss factor.
The opposite solution is also possible – reducing the bending stiffness
without significantly changing the mass. In Figure 5.14, we see the effect of
carving slits halfway through a plywood panel. The reduced bending stiff-
ness means lower speed of propagation of bending waves, a higher critical
frequency and less radiation of sound from resonant modes.

39

30 22

50

40
Transmission loss, R (dB)

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
With sand, 33 kg/m2 Without sand, 17 kg/m2

Figure 5.13 Transmission loss of a door leaf, 39 mm hardboard with 22 mm cylindrical

cavities per 30 mm. (Adapted from Gösele, K., Schalldämmung von Türen,
Berichte aus der Bauforschung, Heft 63, 1969, Figure 10.)
 Introduction to sound insulation  119

40

1 2
30

Transmission loss, R (dB)

20

15 15
10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
Without carving With carving

Figure 5.14 Transmission loss of a 15 mm plywood plate with or without slits. (Adapted

from Fasold, W. and Sonntag, E., Bauakustik. Bauphysikalisches Entwurfslehre,
Band 4, VEB Verlag für Bauwesen, Berlin, 1978, Figure 252.)

A simple and widely used method to increase the mass without too much
trouble with the critical frequency is simply to layer two or more sheets on
top of each other. With two layers instead of one, the bending stiffness is
doubled and so is the mass, i.e. the ratio is constant and the critical fre-
quency is the same as for the single sheet. Figure 5.15 shows the transmis-
sion loss of one or two layers of 13 mm gypsum board, with the critical

40

1 2 30
Transmission loss, R (dB)

20

13 26
10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
13 mm gypsum board Two layers of 13 mm gypsum board

Figure 5.15 Transmission loss of one or two layers of 13 mm gypsum board. (Adapted

from Fasold, W. and Sonntag, E., Bauakustik. Bauphysikalisches Entwurfslehre,
Band 4, VEB Verlag für Bauwesen, Berlin, 1978, Figure 253.)
120  Sound Insulation in Buildings

frequency unchanged at 2500 Hz. Three or even four layers are sometimes

used when very high sound insulation is required. It is important that the
layers are not glued together, but free to slide against each other. In praxis,
the plates may be screwed together and they still behave as layered plates.
Good sound insulation requires that the partition wall is airtight. In the case
of porous concrete and for some types of brick walls, this means that the sur-
face must be treated with plaster, preferably on both sides. The effect of plaster
on the sound insulation of a porous concrete wall is shown in Figure 5.16.

70

60

50
Transmission loss, R (dB)

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
250 mm porous concrete with plaster 250 mm porous concrete without plaster

Figure 5.16 Transmission loss of 250 mm porous concrete with or without plaster.

(Adapted from Kristensen, J., Lydisolation: Teori, måling, vurdering og
bestemmelser. SBI-notat 24 (In Danish), Danish Building Research Institute,
Aalborg University Copenhagen, 1973, Figure 14.)
 Introduction to sound insulation  121

60

50

40
Transmission loss, R (dB)

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
Empty cavity With mineral wool

Figure 5.17 Transmission loss of a 10 mm gypsum board with or without sound absorb-

ing treatment on both sides. (Adapted from Fasold, W. and Sonntag, E.,
Bauakustik. Bauphysikalisches Entwurfslehre, Band 4, VEB Verlag für Bauwesen,
Berlin, 1978.)

Sometimes, the partition wall should also provide sound absorption in

the rooms. The treatment of the surface with a sound absorbing material
may influence the sound transmission. An example of the sound insulation
of a 10 mm gypsum board with or without sound absorbing treatment on
both sides is seen in Figure 5.17. The sound absorbing treatment is 30 mm
mineral wool covered with a perforated plate.

5.3 DOUBLE LEAF CONSTRUCTIONS

5.3.1 Impedance model for sound transmission

Consider a double construction with two plates separated by distance d
(Figure 5.18). The wall impedance of the two plates is denoted by Z1 and
122  Sound Insulation in Buildings

m1 d m2

pi p1 p2 pt

pr p4 p3

v1 v2

Figure 5.18 A double construction showing sound pressures and wall velocities. p1, p2 ,
p3 and p4 denote the sound pressures in the cavity, incident and reflected at
the boundaries. pi, pr and pt denote the incident, reflected and transmitted
sound pressures. v1 and v 2 are the velocities of the two walls when excited
by sound.

Z 2 , respectively. As for the single construction in Equation 5.12, the move-

ment of each wall is:

pi + pr − (p1 − p4 ) = Z1v1

p2 + p3 − pt = Z2v2 (5.40)

The velocity of each wall equals the particle velocity on either side:

1
v1 = ui − ur = (pi − pr )
Z0
1
v1 = u1 − u4 = (p1 − p4 )
Z0
1
v2 = u2 − u3 = (p2 − p3)
Z0
1
v2 = ut = pt (5.41)
Z0
Assuming propagation from one side of the cavity to the other without
losses means that there is only a phase difference between the pressures at
the two intersections:

p2 = p1e− jkd

p4 = p3e− jkd (5.42)

 Introduction to sound insulation  123

From Equations 5.40 through 5.42, the ratio between the sound pressures pi
and pt can be derived and, thus, the transmission loss can be expressed by:
2
p
R0 = 10 lg i
pt
2
 Z + Z2   Z1 + Z2 Z1Z2 
= 10 lg 1 + 1  cos(kd) + j 1 + +  sin(kd) (5.43)
 2Z0   2Z0 2Z02 

If only the mass of each wall is taken into account, the wall impedances are:

Z1 = jω m1
Z2 = jω m2 (5.44)

Neglecting the smaller parts and inserting Z 0 = ρc together with Equation
5.44 yields:

 ω (m + m ) 2
  ω (m1 + m2 ) ω 2m1m2  
2

R0 ≅ 10 lg  1 2
sin(kd) +  cos(kd) − sin(kd)  (5.45)
 2ρ c   2ρ c 2(ρ c)2
 


This result will be discussed and simplified in the following section.

5.3.2 The mass-air-mass resonance frequency

The transmission loss is minimum when the last term in Equation 5.45 is
zero, for example:

m1 + m2 ρ c
tan(kd) = (5.46)
m1m2 ω
For a cavity that is narrow compared to the wavelength (kd << 1), we get:

ωd
tan(kd) ≅ kd =
c

The solution to Equation 5.46 is the mass-air-mass resonance frequency

f 0 = ω 0 /2π:

c ρ 1 1 
f0 ≅ + (5.47)
2π d  m1 m2 

If the depth d of the cavity is comparable to the wavelength, there are many
solutions to Equation 5.46, and they are approximately kd = nπ, where n is
124  Sound Insulation in Buildings

a natural number. The dips in the sound insulation occur at frequencies at

which the cavity depth equals one or more half wavelengths: d = n λ/2.
However, more important than these dips is the shift from low- to high-
frequency behaviour of the air cavity. The cross-over frequency is the fre-
quency fd at which kd = 1:

c
fd = (5.48)
2π d

This is quite similar to the result found for the sound transmission through
a solid material (Equation 5.21). Only, in this case, the transmission is
through air. The spring-like behaviour of the air cavity changes from that
of a simple spring below the cross-over frequency to that of a transmission
channel at higher frequencies.

5.3.3 Sound insulation of double constructions

The result (Equation 5.45) can be simplified in different ways depending on
the frequency range. In the frequency range below the resonance frequency,
f < f 0,

 ω (m1 + m2 ) 
R0 ≈ 20 lg   = R(1+ 2) (5.49)
 2ρ c

This means that the construction behaves as a single construction with the
mass per unit area (m1 + m2). In the frequency range above the resonance
frequency and below the cross-over frequency, f 0 < f < fd ,

 ω 3m1m2d 
R0 ≈ 20 lg  ≈ R1 + R2 + 20 lg(2kd) (5.50)
 2ρ 2c 3 

In this range, a much better sound insulation can be obtained, and the
transmission loss depends on the product of the three parameters, m1, m2
and d. At frequencies above fd where the cavity is wide compared to the
wavelength, sin (kd) is replaced by its maximum value 1, and for f ≥ fd:

 ω 2m1m2 
R0 ≈ 20 lg  ≈ R1 + R2 + 6 dB (5.51)
 2(ρ c)2 
In this high-frequency range, d is no longer an important parameter.
A sketch of the transmission loss as a function of frequency is shown in
Figure 5.19. We recall that the transmission loss of a single leaf will exhibit
a dip at the critical frequency, and if the two leafs have different critical fre-
quencies, there will be two corresponding dips in the transmission loss of
 Introduction to sound insulation  125

R (dB)

f0 fd fc1 fc2
Frequency

Figure 5.19 Sound insulation of a double leaf construction, f0 is the resonance frequency

and fd is the cross-over frequency of the cavity. The critical frequencies of
the two plates are fc1 and fc2 , respectively. Note that fc1 = fc2 if it is a symmet-
ric construction.

the double construction. To include this in the calculations, Equations 5.37

and 5.38 should be used to represent R1 and R 2 in Equations 5.50 and 5.51.
A more accurate approximation of Equation 5.45 valid at frequencies
below fd is:

  2
 
2
 ω 
R0 = R(1+ 2) + 10 lg sin (kd) +  cos(kd) −   ⋅ sinc(kd) 
2
   ω0   
 
2
  f 2
≅ R(1+ 2) + 10 lg  1 −    (5.52)
  f0  

5.3.4 Examples of double constructions

Some examples of measured transmission loss of double constructions will
illustrate the physical principles described here.
Figure 5.20 shows the effect of increasing the distance d between the
glasses in a double window. With 12 mm cavity, the resonance frequency
is 250 Hz, which is very unfortunate in relation to traffic noise. Increasing
the distance to 85 mm yields a much better sound insulation at all frequen-
cies, as the resonance frequency is lower and the resonance dip not so pro-
nounced. The critical frequency of a 4 mm glass is 3 kHz, and this is noted
in the measured frequency course.
In a double window, it can be a good idea to use glass of different
thicknesses because the dips at the critical frequencies, then, are less
 126  Sound Insulation in Buildings

60

50

40
Transmission loss, R (dB)

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
2 × 4 mm glass with 85 mm distance 2 × 4 mm glass with 12 mm distance

Figure 5.20 Sound insulation of a double window with two 4 mm glass panes, and the
influence of the cavity depth. (Adapted from Ingemansson, S., Ljudisolerande
fönsterkonstruktioner (In Swedish), Stockholm, Sweden, 1968, Figure 8.)

pronounced compared to the dip in a symmetric construction. The exam-

ple in Figure 5.21 is a double window with a 100 mm cavity and glass
thicknesses of 4 mm and 6.5 mm. The advantage of the asymmetric solu-
tion is obvious.
In a drywall, it is important that the cavity is acoustically damped, e.g.
with mineral wool. Without this, the sound in the cavity can resonate
and significantly reduce the sound insulation. The example in Figure 5.22
shows the result for a wall with two layers of 13 mm gypsum board on steel
profiles per 600 mm. The cavity depth is 70 mm and this is either empty or
 Introduction to sound insulation  127

60

50

40
Transmission loss, R (dB)

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
6.5 mm and 4 mm glass with 100 mm distance
2 × 4 mm glass with 100 mm distance
2 × 6.5 mm glass with 100 mm distance

Figure 5.21 Influence of symmetry on the sound insulation of a double construction.

(Adapted from Lewis, P.T., Building Acoustics, Oriel Press, London, 1971,
Figure 7.19.)

filled with 50 mm mineral wool. The effect of the mineral wool is signifi-
cant in the entire frequency range.
Asymmetric double constructions have a special application in relation
to heavy- or medium-weight single walls. Such walls have a relatively low
critical frequency and, thus, the sound radiation from resonant modes in the
wall is very important for sound transmission. If the structural vibrations
are transferred to a thin plate with high critical frequency, the sound radia-
tion is reduced and, thus, the sound insulation is significantly improved, as
shown in the example in Figure 5.23. In fact, it is possible to improve the
sound insulation even more by applying a similar light construction to the
other side of the wall.
 128  Sound Insulation in Buildings

60

50

40
Transmission loss, R (dB)

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
50 mm mineral wool in cavity Empty cavity

Figure 5.22 The influence of absorption material in the cavity on the sound insulation of
a double construction.

5.4 SLITS AND LEAKS

In Figure 5.3 and Equation 5.12, it was shown that an aperture in a wall
leads to an upper limit of the sound insulation. As a first approximation, this
limit is independent of the frequency, which in practice means that the high
frequency sound insulation is limited while the lower frequencies can remain
more or less unaffected by a small slit or a leakage. In the following, we shall
see a few examples of how slits and leaks influence the sound insulation.
The effect of the width of a slit is seen in Figure 5.24. While transmis-
sion loss of the wall without slits increase with frequency, a 1 mm wide
slit means an upper limit around 40 dB. In addition, we see very efficient
transmission around 2000 Hz and 4000 Hz. This can be explained by the
depth of the slit that equals one or two half wavelength at these frequencies.
 Introduction to sound insulation  129

70

60

50
Transmission loss, R (dB)

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

With treatment Without treatment

Figure 5.23 Effect of a thin panel construction to reduce sound radiation from a single
wall, shown here as a 68 mm lightweight concrete wall. The thin panel is
3 mm of hardboard on 25 mm wooden laths with 25 mm mineral wool
in the cavity. (Adapted from Brandt, O., Akustisk planering. (In Swedish),
V. Pettersons Bokindustri AB, Stockholm, Sweden, 1958, Figure 3.24b.)

In Figure 5.25, the 3 mm slit from the previous figure is filled with felt,
and we see that such a filling can be efficient; the transmission loss with felt
in the slit is nearly the same as that without any slit in Figure 5.24.
Doors can be particularly difficult in terms of sound insulation, because
it sometimes becomes necessary to avoid any sealing lists at the floor, e.g.
in hospitals. In such cases, it is possible to apply special moving sealing lists
that close when the door is closed. However, it is also possible to leave the
130  Sound Insulation in Buildings

60

50

40
Transmission loss, R (dB)

30

20

10

0
100 200 500 1000 2000 5000 10000
Frequency (Hz)
No slit b = 1 mm b = 3 mm b = 10 mm

Figure 5.24 Transmission loss of a 1 m2 wall with a 1 m long slit. The depth is 75 mm and
the width of the slit varies from 0 mm to 10 mm. (Adapted from Gösele, K.,
Schalldämmung von Türen, Berichte aus der Bauforschung, Heft 63, 1969,
Figure 17.)

slit under the door open, if there is an efficient sound absorbing lining in
the door, like a resonator. In the measured result shown in Figure 5.26, the
underside of the door has an opening into a 100 mm deep cavity filled with
mineral wool.
Sound insulation of doors is also particularly difficult in practice due to
the risk of leaks and because the sealing between the frame and the sur-
rounding wall is important. Good workmanship is important here.
Openable windows are also an issue, where leaks can easily lead to poor
sound insulation. It is important that a soft sealing list is applied on all
sides of the sash, and if it is possible to have two sets of continuous sealing
that is even better that a single sealing (Figure 5.27).
 Introduction to sound insulation  131

60

50

40
Transmission loss, R (dB)

30

20

10

0
100 200 500 1000 2000 5000 10000
Frequency (Hz)
3 mm slit with felt 3 mm slit

Figure 5.25 Effect of filling a slit with felt. Transmission loss of a 1 m2 wall with a 1 m
long and 75 mm deep slit. The width of the slit is 3 mm. (Adapted from
Gösele, K., Schalldämmung von Türen, Berichte aus der Bauforschung, Heft
63, 1969, Figure 24.)

5.5 FLANKING TRANSMISSION

The transmission of sound from a source room to a receiver room can

be via flanking constructions like the floor, the ceiling or the façade.
When all relevant transmission paths are considered, the sound insulation
is described by the apparent sound transmission loss (apparent sound
­reduction index):

P1 S
R′ = 10 lg = L1 − L2 + 10 lg (dB) (5.53)
P2 + P3 A2
132  Sound Insulation in Buildings

60

50

40
Transmission loss, R (dB)

30

20

10

0
100 200 500 1000 2000 5000 10000
Frequency (Hz)
2 mm slit with absorber 2 mm slit without absorber

Figure 5.26 Effect of a Helmholtz resonator to reduce sound transmission through

a slit. (Adapted from Gösele, K., Schalldämmung von Türen, Berichte aus der
Bauforschung, Heft 63, 1969, Figure 25.)

where P1 is the incident sound power on the partition, P2 is the sound

power transmitted through the partition wall to the receiver room and P3
is the sound power radiated to the receiver room from the flanking surfaces
and other flanking paths:

P3 = ∑P
i
F ,i (5.54)

Each single flanking transmission path i can be characterized by the flank-

ing transmission loss, R F,i:

P1
RF ,i = 10 lg (dB) (5.55)
PF ,i
 Introduction to sound insulation  133

60

50

40
Transmission loss, R (dB)

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

Two sets of sealing One set of sealing No sealing

Figure 5.27 Effect of sealing in an openable double window with two panes of 4 mm glass
and 85 mm cavity depth. (Adapted from Ingemansson, S., Ljudisolerande
fönsterkonstruktioner (In Swedish), Stockholm, Sweden, 1968, Figure 6.)

It is convenient to keep the incident sound power P1 on the partition wall

as a reference for all the flanking transmission losses. In this way, it is very
simple to add all the contributions. The apparent transmission loss is cal-
culated from:

 
R′ = −10 lg  10−0,1R +

∑ 10
i
−0,1RF ,i
 (dB) (5.56)


In the typical case of horizontal transmission through a wall, there will be

12 flanking paths, namely, three possible paths for each of the four sur-
rounding flanking constructions (Figure 5.28).
Flanking transmission can be controlled in different ways. Three basic
principles include the following:
134  Sound Insulation in Buildings

Figure 5.28 Direct transmission and three flanking transmission paths via the floor.
(Reproduced from Kristensen, J. and Rindel, J.H., Bygningsakustik – Teorti og
praksis. SBI-Anvisning 166 (In Danish), Danish Building Research Institute,
Hørsholm, Denmark, 1989. With permission.)

• Use of heavyweight constructions. The heavy mass of the flanking

construction means that the excitation of structural vibrations is rela-
tively weak. If the transmission loss of the flanking construction itself
is at least as good as that of the partition wall, the flanking transmis-
sion should not be a problem.
• Interruption of the flanking construction, either by a slit or by elastic
interlayers. This calls for careful design of the details in the junction.
Some examples of good and bad solutions are shown in Figures 5.29
and 5.30.

Bad Good Bad Good

Bad Good Bad Good

Figure 5.29 Examples of junctions between external and internal walls in h­ orizontal

view. Solutions are shown in pairs, those with a flanking transmission
issue to the left.
 Introduction to sound insulation  135

Bad Good Bad Good

Bad Good Bad Good

Figure 5.30 Examples of junctions between roof and internal wall (upper part) and
between floor and internal wall (lower part) in vertical view. Solutions are
shown in pairs, those with a flanking transmission issue to the left.

• Covering the flanking surface with a board having a high critical fre-
quency, i.e. preferably above 2000 Hz, the higher the better. This can
be a secondary surface with one or two layers of plasterboard. The
reduction of flanking transmission is efficient on both surfaces of the
flanking structure, in the source room as well as in the receiving room.

5.6 ENCLOSURES

A noise source is supposed to radiate sound power Pa. Consider that the
noise source is totally covered by an enclosure with surface area S, absorp-
tion coefficient α on the inside and the enclosure is made from a plate
with transmission loss R or transmission coefficient τ. The average sound
pressure in the enclosure pencl can be estimated, if a diffuse sound field is
assumed:

4Pa

2
pencl = ρ c (5.57)
αS
The sound power incident on the inner surface of the enclosure is (still with
the assumption of a diffuse sound field):

2
pencl S
Pinc = (5.58)
4ρ c
136  Sound Insulation in Buildings

The sound power transmitted through the enclosure is then:

τ
Pout = τ Pinc = Pa (5.59)
α
The insertion loss of the enclosure is the difference in radiated sound power
level without and with the enclosure:

Pa α
∆L = 10 lg = 10 lg = R + 10 lg α (dB) (5.60)
Pout τ
This result cannot be considered to be very accurate. The assumption of
a diffuse sound field inside the enclosure is especially doubtful. However,
the result is not bad as a rough estimate for the design of an enclosure. It is
clearly seen from Equation 5.60 that both transmission loss and absorption
coefficient are important for an efficient reduction of noise by an enclosure.

5.7 IMPACT SOUND INSULATION

5.7.1 Impact sound pressure level

The noise generated from footsteps on floors is characterized by the impact
noise level. It is measured using a standardized tapping machine according
to ISO 16283-2 in buildings or ISO 10140-3 in the laboratory. The main
data for the tapping machine are the following:

• The noise is generated by steel hammers with a fall height of 40 mm.

• Each steel hammer has a mass of 500 g.
• The number of taps per second is 10.

The tapping machine is placed on the floor in the source room in a number
of positions. The calibrated sound pressure level L 2 is measured in the room
below – or any other room in the building (see Figure 5.31). The reverbera-
tion time in the receiving room must also be measured in order to calculate
the absorption area A 2 . The impact sound pressure level is the sound pres-
sure level in dB re 20 μPa that would be measured if the absorption area is
A 0 = 10 m 2:

A2
Ln = L2 + 10 lg (dB) A0 = 10 m 2 (5.61)
A0

The frequency range is the same as for airborne sound insulation, i.e. the
16 one-third octave bands from 100 Hz to 3150 Hz. However, it is recom-
mended that the frequency range be extended down to 50 Hz, especially in
the case of lightweight floors and heavyweight constructions with a floating
floor.
 Introduction to sound insulation  137

Room 1

P2

Room 2

A2

L2

Figure 5.31 Principle of measuring the impact sound pressure level from a floor to a
receiving room (Room 2).

The normalized impact sound pressure level L n is used for measure-

ments in the laboratory to characterize specimens of floor constructions.
However, for field measurements in buildings, the standardized impact
sound pressure level Ln′ T is sometimes preferred.

T2
Ln,′ T = L2 − 10 lg (dB) T0 = 0.5 s (5.62)
T0
Just like DnT for airborne sound insulation, Ln′ T is corrected for the
reverberation time in the receiving room if different from 0.5 s. According
to ISO 16283-2, Ln′ T provides a straightforward link to the subjective
impression of impact sound insulation.
The relationship between the standardized and the normalized impact
sound pressure level is approximately:

V2
Ln′ T = Ln′ − 10 lg (dB), V0 = 30 m3 (5.63)
V0
where V2 is the volume of the receiving room. For typical volumes in a
dwelling around 50 m3 to 80 m3, the difference will be −2 dB to −4 dB.
138  Sound Insulation in Buildings

5.7.2 Historical note
The tapping machine originated from the German standard DIN 4110
(1938), but the impact sound pressure level was measured in Phon as a
broad-band level applying a frequency weighting filter B. The practice to
measure impact sound in one-third octave bands was developed in Denmark
(Ingerslev et al. 1947) – they used the frequency range from 100  Hz
to 2000 Hz. An alternative method was developed in America, using a tap-
ping machine with different specifications and measuring in seven octave
bands from 63 Hz to 4 kHz (Lindahl and Sabine 1940).

5.7.3 Sum of airborne and impact sound insulation

Figure 5.32 shows the measured impact sound pressure level of a 230 mm
clinker concrete floor, either bare or with the traditional Danish flooring,
a 22 mm wooden floor on joists on resilient pads. The flooring makes a
big difference on the impact sound, and the improvement of the impact
sound pressure level is denoted by ΔL. It increases from about 200 Hz with

90

80
Impact sound pressure level, Ln (dB)

70

ΔL
60

50

40

30

20

10
50 100 200 500 1000 2000 5000
Frequency (Hz)
Bare slab With wooden floor on laths on soft pads

Figure 5.32 Impact sound pressure level of a 200 mm clinker concrete slab with 30 mm
concrete finish, measured without and with a 22 mm wooden floor on
laths (38 mm × 57 mm) on 10 mm fiberboard pads. (Data from Lydteknisk
Institut, Building Acoustic Laboratory Measurements. Bygningsakustiske labora-
toriemålinger, in Danish), Lyngby, Denmark, LL 906/80.
 Introduction to sound insulation  139

9 dB per octave; a slight amplification is seen at 125 Hz, which is due to a

mass-spring resonance of the floor. Figure 5.33 shows the airborne sound
insulation of the same construction, and it is seen that the improvement of
the transmission loss due to the flooring ΔR is almost negligible.
For massive homogeneous floors, there is a relationship between the air-
borne and impact sound insulation. In Chapter 10, it will be shown that
for homogeneous floors with a hard surface, the relationship is as simple as:

R + Ln = 38 + 30 lg f (5.64)

With a floating floor or a soft flooring, the relationship changes above the
resonance frequency f 0 of the flooring:

R + Ln = 38 + 30 lg f0 − ∆L + ∆R (f > f0 ) (5.65)

80

70
ΔR
60
Transmission loss, R (dB)

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
Bare slab With wooden floor on laths on soft pads

Figure 5.33 Transmission loss of a 200  mm clinker concrete slab with 30 mm

­concrete finish, measured without and with a 22 mm wooden floor on
laths (38 mm × 57 mm) on 10 mm fiberboard pads. (Data from Lydteknisk
Institut, Building Acoustic Laboratory Measurements. Bygningsakustiske
­laboratoriemålinger, in Danish), Lyngby, Denmark, LL 884/80.
140  Sound Insulation in Buildings

160

150

140

130
Sum R + Ln (dB)

ΔL – ΔR
120

110

100

90

80
50 100 200 500 1000 2000 5000
Frequency (Hz)
Bare slab With wooden floor on laths on soft pads Theory

Figure 5.34 Example of sum curves obtained by adding R and L n, using the measured
results from Figures 5.32 and 5.33.

Using the example of the measured sound insulation in Figures 5.32

and 5.33, the sum curve with and without the wooden floor is shown in
Figure 5.34. The agreement with the theoretical curves is quite satisfactory.
The wooden floor on joists on resilient pads is efficient in attenuating the
impact sound at frequencies above 200 Hz.

5.8 SINGLE-NUMBER RATING OF
SOUND INSULATION

5.8.1 Weighted sound reduction index and

the sound transmission class
The single-number rating of sound insulation is practical for several purposes:

• To characterize in a single number the measured result of a building

construction
• For quick comparison of the sound insulation obtained with different
constructions
• To specify requirements for sound insulation

The weighted sound reduction index Rw is based on a standardized refer-

ence curve that is defined in one-third octaves in the frequency range from
 Introduction to sound insulation  141

100 Hz to 3150 Hz. The method is defined in the international standard

ISO 717-1. The reference curve has three straight lines with a slope of 9 dB
per octave from 100 Hz to 400 Hz, 3 dB per octave from 400 Hz to 1250 Hz
and 0 dB per octave from 1250 Hz to 3150 Hz.
The measured transmission loss is compared to the reference curve, and
the difference is calculated in each of the 16 one-third octave bands. The
sum of unfavourable deviations (deficiencies) is calculated. An unfavour-
able deviation is the deviation between the reference curve and the mea-
sured curve when the measured sound insulation is lower than the value of
the reference curve. Deviations where the measured value is better than the
reference curve do not count.
The reference curve is shifted up or down in steps of 1 dB, and the correct
position of the reference curve is found when the sum of unfavourable devi-
ations is as large as possible, but do not exceed 32 dB. This limit of 32 dB
is chosen because it means an average deviation of 2 dB in the 16 frequency
bands covered by the curve. No credit is given for (favourable) deviations
above the reference curve. The value of the reference curve at 500 Hz is
taken as the single-number value of the measuring result. The application
of the method on an example is shown in Figure 5.35.

80

70

60
Transmission loss, R (dB)

Rw = 30 dB
50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 5.35 Determination of the weighted sound reduction index. Full line is the mea-
sured transmission loss, in this example, a double window measured in the
laboratory. The dotted curve is the reference curve, which has been shifted
to the position 30 dB at 500 Hz. The sum of the unfavourable deviations is
7.0 + 8.0 + 8.0 + 4.0 + 2.0 + 3.0 = 32.0 dB. The result is Rw = 30 dB.
142  Sound Insulation in Buildings

The same procedure is applied to field measurement results of airborne

sound insulation, e.g. as the weighted apparent sound reduction index Rw′
and the weighted standardized level difference DnT,w.
The American standard ASTM E413-16 defines a very similar single-­number
rating called sound transmission class (STC). The method and the reference
curve are the same as in ISO 717-1, but there are the following differences:

• The frequency range is 125 Hz to 4000 Hz, i.e. there are still 16 one-
third octave bands but they are shifted one band up; the 100 Hz value
of the reference curve is not used and the reference curve is extended
horizontally up to 4000 Hz.
• The measured transmission loss results are rounded to whole dB val-
ues before comparison with the reference curve.
• An additional criterion is that the maximum unfavourable deviation
is 8 dB; in the case of a dip in the frequency curve of sound insulation,
this may be decisive for the single-number value.
• The result is written STC xx (without dB), where xx is the single-
number value in dB.

The Rw and STC values are usually very close, within 1 or 2 points, but
larger differences may occur in the case of a resonance dip.

5.8.2 Weighted impact sound pressure level and

the impact insulation class
The weighted impact sound pressure level L n,w is based on a standard-
ized reference curve that is defined in one-third octaves in the frequency
range 100 Hz to 3150 Hz. The method is defined in the international stan-
dard ISO 717-2. The reference curve has three straight lines with a slope of
0 dB per octave from 100  H z to 315 Hz, −3 dB per octave from 315 Hz to
1000 Hz and −9 dB per octave from 1000 Hz to 3150 Hz.
The measured impact sound pressure level is compared to the reference
curve, and the sum of unfavourable deviations (deficiencies) is calculated.
An unfavourable deviation is the deviation between the reference curve and
the measured curve if the measured impact sound pressure level is higher
than the value of the reference curve.
The reference curve is shifted up or down in steps of 1 dB, and the cor-
rect position of the reference curve is found when the sum of unfavourable
deviations is as large as possible, but do not exceed 32 dB. The value of
the reference curve at 500 Hz is taken as the single-number value of the
measuring result. The application of the method on an example is shown
in Figure 5.36.
The same procedure is applied to laboratory and field measurement
results of impact sound insulation, e.g. as the weighted standardized impact
sound pressure level Ln′ T ,w .
 Introduction to sound insulation  143

80

L'n,w = 48 dB
70

60
Impact sound pressure level, L'n (dB)

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 5.36 Determination of the weighted impact sound pressure level. The full line is
the measured impact sound pressure level, in this example, a field measure-
ment of a concrete slab with a floating floor. The dotted curve is the refer-
ence curve, which has been shifted to the position 48 dB at 500 Hz. The sum
of the unfavourable deviations is 6.3 + 3.7 + 1.7 + 4.4 + 3.0 + 5.0 + 2.3 = 26.4 dB.
The result is L′n,w = 48 dB.

The American standard ASTM E989-06 defines a very similar single-­

number rating called impact insulation class (IIC). The method, the frequency
range and the reference curve are the same as in ISO 717-2, but there is a
significant difference in the presentation of the result: The IIC value is equal
to 110 dB − Ln,w. This implies that a high impact sound insulation gives a high
IIC value, but a low value of Ln,w. The IIC is characterized by the following:

• The frequency range is 100  H z to 3150 Hz, i.e. the same 16 one-third

octave bands as used for L n,w.
• The measured impact sound pressure levels are rounded to whole dB
values before comparison with the reference curve.
• An additional criterion is that the maximum unfavourable deviation
is 8 dB; in the case of a peak in the frequency curve of impact sound
pressure level, this may be decisive for the single-number value.
• The result is written IIC xx (without dB), where xx is the s­ ingle-number
value in dB.
144  Sound Insulation in Buildings

The importance of frequencies below 100 Hz, especially in connection with

lightweight floor constructions and floating floors on heavyweight con-
structions, has made it relevant to extend the frequency range for measure-
ments of impact sound insulation down to 50 Hz, or even further down to
20 Hz. More about this is explained in Chapter 12.

5.8.3 Alternative reference curve for evaluating

the impact sound insulation
There has been some controversy about the rating curves, especially the one
for impact sound. An alternative rating curve was suggested by Bodlund
(1985) based on a thorough analysis and comparison with results from a
socio-acoustic investigation. The so-called Bodlund curve is a straight line
with a slope of 1 dB per one-third octave from 50 Hz to 1000 Hz. Although
this covers only 14 frequency bands, the evaluation method is the same
as for the ISO curve, i.e. the sum of unfavourable deviations should be as
large as possible without exceeding 32 dB. An example of the application
is shown in Figure 5.37, using the same measured data as in Figure 5.36.

80

L'B = 60 dB
70

60
Impact sound pressure level, L'n (dB)

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 5.37 The Bodlund curve and its application. The full line is the measured impact
sound pressure level, in this example, a field measurement of a concrete
slab with a floating floor. The dotted curve is the reference curve, which has
been shifted to the position 60 dB at 500 Hz. The sum of the unfavourable
deviations is 3.3 + 10.9 + 11.7 + 3.3 = 29.2 dB. The result is L′B = 60 dB.
 Introduction to sound insulation  145

This alternative evaluation method for impact sound is often applied or

referred to in research. The reason is that the results of this evaluation
method have proved to be highly correlated with subjective response to
footfall noise from neighbours in multi-unit houses. If Psatisfied is the per-
centage of persons rating the impact sound insulation quite or nearly quite
satisfactory, the corresponding Bodlund-weighted impact sound pressure
level, according to Hveem et al. (1996, 34–35), is:

LB′ = 74, 7 − 0, 247 ⋅ Psatisfied (dB) (5.66)

It has been found that the result, using the Bodlund curve, is highly cor-
related with the ISO method when extended with the spectrum adaptation
term going down to 50 Hz (the spectrum adaptation term is explained in
Section 5.8.6). As an example, we find that more than 80 % should be
satisfied with the impact sound if L′B < 55 dB, or approximately L′n,w  + C I,50–
2500 < 49 dB, because there is an average difference of 6.4 dB between the
two ratings (Hveem et al. 1996, 35).

5.8.4 Previous methods for rating sound insulation

The idea of using a shifted reference curve for evaluating airborne and
impact sound insulation comes from Germany (DIN 4109, 1959) and was
adapted in ISO/R 717 in 1968. It was called the airborne sound insula-
tion index Ia and was similar to Rw with the exception, that maximum
unfavourable deviation in any one-third octave band was limited to 8 dB.
Hence, Ia ≅ Rw. In the 1982 edition of ISO 717, the symbol was changed to
Rw and the 8 dB-rule was removed.
Similarly, for the rating of impact sound insulation, the impact sound
index Ii was defined in ISO/R 717. The rating curve was the same as shown
in Figure 5.31, but the impact sound pressure levels should either be mea-
sured in 1/1 octave bands or measured in one-third octave bands and then
corrected +5 dB before comparison with the reference curve. Hence, Ii ≅
L n,w + 5 dB. In the 1982 edition of ISO 717, the 5 dB correction and the 8
dB-rule was removed and the symbol was changed to L n,w.

5.8.5 Historical note
It is interesting to note that ISO/R 717, which is specifically meant for mea-
surements in dwellings, only mentions the normalized measures for sound
insulation R′ and L n, not the standardized measures that adjust the result to
0.5 s reverberation time. Since some countries were using the standardized
measures, it is not surprising that there was some controversy about the
proposal among the ISO member counties; 20 countries approved the rec-
ommendation, but six countries opposed the approval (Belgium, Denmark,
France, Italy, Norway and USA).
146  Sound Insulation in Buildings

Before the first ISO standard on evaluation of sound insulation, many

countries used a fixed reference curve to define minimum requirements in
the building code. In Britain, the aggregate adverse deviation (AAD) was
used as single-number rating of airborne sound insulation. This measure is
the sum of all deviations below a fixed two-segment reference curve and
ignores bands where measured performance is above the reference curve.
An approximate conversion is Rw′ ≅ STC ≅ −0.113 AAD + 56.7 dB (Bradley
1982). Hence, AAD = 15 corresponds to Rw′ ≅ 55 dB and AAD = 60 corre-
sponds to Rw′ ≅ 50 dB.

5.8.6 Spectrum adaptation terms

Historically, two different methods were developed for single-number rat-
ing of airborne sound insulation. One was the comparison with a weighting
curve as described earlier, and the other method was the French ‘Indice
d’affaiblissement acoustique R vis-à-vis d’un bruit rose’, R rose. This is cal-
culated as the difference between two A-weighted sound pressure levels: (1)
a sound with pink spectrum from 100 Hz to 5000 Hz, and (2) the sound
with a pink spectrum reduced in each one-third octave band by the sound
reduction index. For the evaluation of windows and facades, a generalized
traffic noise spectrum was applied instead of the pink noise spectrum, and
the term was ‘Indice d’affaiblissement acoustique R vis-à-vis d’un bruit de
traffic routier’, R route.
The revision of ISO 717 published in 1996 was the result of a compro-
mise, merging the German and the French methods into one. The reference
curve and Rw was kept, but new spectrum adaptation terms C and C tr were
defined in such a way that

Rrose ≅ Rw + C and Rroute ≅ Rw + Ctr

Single-number values that include a spectrum adaptation term characterize

the sound insulation as the A-weighted level difference for noise with either
a pink noise spectrum or a road traffic noise spectrum. The latter has more
energy at the low frequencies and is primarily intended for the evaluation
of windows and façades. The frequency range used in this method is more
flexible, and in addition to the frequencies used for Rw, (100 Hz to 3150 Hz),
the range can be extended to include higher frequencies (4000 Hz and
5000 Hz) and lower frequencies (50 Hz, 63 Hz and 80 Hz).
The spectrum adaptation terms are calculated by the following equation:
n

C j = −10 lg ∑ 10
i =1
(Lij − Xi )/10
− Xw (dB) (5.67)

where
 Introduction to sound insulation  147

Cj is the spectrum adaptation term, e.g. C, C tr, C 50–5000, C tr,50–3150.

j is index for the sound spectrum, either 1 or 2.
i is index for the one-third octave band in the relevant range, up to n.
Lij is the level of the spectrum j in the frequency band i (Table 5.1).
Xi is the sound insulation at frequency band i, e.g. R or DnT, given to 0.1 dB.
Xw is the single-number quantity calculated with the shifted reference
curve, e.g. Rw.

For simplicity, DnT,w,50 is sometimes used instead of the more correct

DnT,w + C 50–3150 and, similarly, Rw,50 is used as abbreviation for Rw + C 50–3150.

Table 5.1  S pectra for calculation of the spectrum adaptation terms according to
ISO 717-1
Spectrum 1 Spectrum 1 Spectrum 2
From 50 Hz or From 50 Hz or
100 Hz up to 100 Hz up to Any frequency
3150 Hz 5000 Hz range
Frequency
(Hz) Lj,i (dB) Lj,i (dB) Lj,i (dB)
50 −40 −41 −25
63 −36 −37 −23
80 −33 −34 −21
100 −29 −30 −20
125 −26 −27 −20
160 −23 −24 −18
200 −21 −22 −16
250 −19 −20 −15
315 −17 −18 −14
400 −15 −16 −13
500 −13 −14 −12
630 −12 −13 −11
800 −11 −12 −9
1000 −10 −11 −8
1250 −9 −10 −9
1600 −9 −10 −10
2000 −9 −10 −11
2500 −9 −10 −13
3150 −9 −10 −15
4000 −10 −16
5000   −10 −18
Note: Spectrum 1 is A-weighted pink noise and spectrum 2 is A-weighted traffic noise.

148  Sound Insulation in Buildings

Note that there is no need to calculate the weighted value with the shifted
reference curve if only the extended frequency range is relevant. For
­example, Rw,50 is calculated directly from:
n

Rw,50 = −10 lg ∑10

i =1
(Lij − Ri )/10
(dB) (5.68)

The principle of applying an assumed noise spectrum and characterizing

the sound insulation by the A-weighted level difference is particularly use-
ful for windows and façade elements, and many different traffic noise spec-
tra are available for this purpose. The collection of spectra in Nordtest
ACOU 061 contains seven spectra representing various kinds of road traffic
noise, railway noise and aircraft noise. Nordtest Spectrum 1 is for mixed
urban road traffic at 50 km/h and about 10 % heavy vehicles, and this is the
traffic noise spectrum adapted in ISO 717-1.
The ASTM E1332 describes a similar method, using a spectrum repre-
senting a mixture of road, rail and air traffic. The one-third octave bands
from 80 Hz to 4000 Hz inclusive are used, and the calculated A-weighted
level difference is called the outdoor-indoor transmission class (OITC). This
spectrum has a substantial amount of energy at low frequencies. However,
Davy (2004) brought important new information about the measurements
behind the OITC spectrum, and he concluded that there is some doubt
about the validity of the spectrum, especially concerning the high levels at
low frequencies.
Figure 5.38 shows the traffic noise spectra from ISO 717-1 and ASTM
E1332 together with a spectrum for city street traffic noise derived from
newer measurements and proposed by Buratti et al. (2014). It is clearly
seen that the newer spectrum for city street traffic has less energy at low
frequencies than the older Nordtest/ISO spectrum; whether this is due to a
general change in cars as noise sources, or to differences in the measurement
conditions, is not known. However, there are certainly differences from
one country to another in the typical composition of road traffic (e.g. the
number of very light vehicles, scooters, frequent use of horns etc.). Further
discussion of the applicability of the Nordtest/ISO traffic noise spectrum is
found in the work of Mesihovic et al. (2016).
Spectrum adaptation terms are also used for impact sound insulation.
This allows the frequency range to be more flexible, and in addition to the
frequencies used for L n,w (100 Hz to 3150 Hz), the range can be extended
to  include lower frequencies (50 Hz, 63 Hz and 80 Hz), or even lower
if desired.
The spectrum adaptation term for impact sound is calculated by the fol-
lowing equation:
n

CI = 10 lg ∑10
i =1
Li /10
− 15 − Xw (dB) (5.69)
 Introduction to sound insulation  149

A-weighted SPL in 1/3 octave bands (dB) 0

–10

–20

–30
50 100 200 500 1000 2000 5000
Frequency (Hz)

ISO 717-1 ASTM E1332 Buratti et al. (2014)

Figure 5.38 Traffic noise spectra in third octave bands relative to the A-weighted level.
The ISO 717-1 spectrum used for the spectrum adaptation term Ctr, the
ASTM E1332 spectrum used for OITC, and the city street traffic spectrum
from Buratti et al. (2014).

where
C I is the spectrum adaptation term; C I, 50–2500 in the case of extended
frequency range.
i is index for the one-third octave band in the relevant range, up to n.
Li is the level of the measured impact sound pressure level in the fre-
quency band i, given to 0.1 dB.
Xw is the single-number quantity calculated with the shifted reference
curve, e.g. L n,w or Ln′ T ,w .

The number 15 in Equation 5.69 has been so determined that the value of
C I is approximately 0 dB for solid floors with effective floor coverings.
For simplicity, L n,w,50 is sometimes used instead of the more correct
L n,w + C I, 50–2500 and, similarly, for Ln′ T ,w,50 . Note that there is no need to
calculate the weighted value with the shifted reference curve if only the
extended frequency range is relevant. For example, L n,w,50 is calculated
directly from:

Ln,w,50 = 10 lg ∑10
i =1
Li /10
− 15 (dB)

(5.70)
150  Sound Insulation in Buildings

5.9 REQUIREMENTS FOR SOUND INSULATION

Requirements for sound insulation are usually given in the national build-
ing codes and may differ from one country to another. Some countries
have detailed and complicated requirements for different types of buildings
and rooms, while other countries have few and simple requirements. Some
examples of requirements between dwellings are shown in Table 5.2. The
highest level of sound insulation requirements is found in the Nordic coun-
tries and in Austria, while Portugal has significantly lower requirements.
The latter is characteristic for countries in southern Europe. It appears that
sound insulation requirements depend on the climate of the country, and
a warm climate means that people spend most time with open windows or
outdoors, while the opposite is true in a cold climate.
In 2003, the UK changed the compliance limits for England and Wales
to use the traffic noise spectrum adaptation term C tr for the evaluation of
party walls and floors; Australia did the same a few years later. The reason-
ing behind this rather surprising adoption was that low-frequency sound
from music, Hi-Fi equipment, etc. was recognized as an increasing problem
between neighbours in multi-unit buildings, but there was at the same time
a strong resistance against extending the measurements to include frequen-
cies below 100 Hz. The traffic noise spectrum would certainly put more
weight on the lower frequencies (e.g. 100 Hz to 500 Hz), but instead new
problems have appeared with complaints about insufficient sound insula-
tion at the mid and high frequencies.

Table 5.2  Examples of sound insulation requirements between dwellings in

multi-unit buildings (as of 2016)
Country Airborne Impact
Some Nordic countries (DK, ES, FI, IS, Rw′  ≥ 55 dB ′  ≤ 53 dB
Ln,w
NO)
Sweden DnT,w,50 ≥ 52 dB Ln′T ,w,50  ≤ 56 dB
Austria DnT,w ≥ 55 dB Ln′T ,w  ≤ 48 dB
Germany (walls/floors) Rw′  ≥ 53/54 dB ′  ≤ 53 dB
Ln,w
Italy Rw′  ≥ 50 dB ′  ≤ 63 dB
Ln,w
France DnT,w + C ≥ 53 dB Ln′T ,w  ≤ 58 dB
Portugal DnT,w ≥ 50 dB Ln′T ,w  ≤ 60 dB
Spain DnT,w + C ≥ 50 dB Ln′T ,w  ≤ 65 dB
Scotland DnT,w ≥ 56 dB Ln′T ,w  ≤ 56 dB
England and Wales DnT,w + Ctr ≥ 45 dB Ln′T ,w  ≤ 62 dB
Australia DnT,w + Ctr ≥ 45 dB Ln′T ,w  + CI ≤ 62 dB
New Zealand STC 55 (FSTC 50) IIC 55 (FIIC 50)
USA (2006 International Building Code) STC 50 (FSTC 45) IIC 50 (FIIC 45)
Canada STC 50 (ASTC 47) N/A
 Introduction to sound insulation  151

In the USA, most of the states have adapted the International Building
Code (IBC), which requires STC 50 and IIC 50. However, this is the require-
ment for laboratory tests of the building components, and in the case of
a field test, 5 dB lower values are allowed (field sound transmission class
(FSTC) and field impact insulation class (FIIC)). It should be noted that the
American measurement standard ASTM 336 requires that ‘the flanking
paths be shielded’ during measurement of transmission loss in the field.
Therefore, FSTC 45 is a rather weak requirement for airborne sound insu-
lation. Canada has no requirement for impact sound and the requirement
for airborne sound insulation is similar to that in the USA, except that field
measurements should be made in terms of the apparent sound transmission
class (ASTC), which includes the flanking paths in the building.
The progress of lightweight building systems has led to increased aware-
ness of the importance of low frequencies for evaluation of impact sound,
and as a consequence, a revised building code SS 25267 was introduced in
Sweden in 2015. This requires, as a minimum, the sound insulation to be
evaluated from 50 Hz, and impact sound to be measured down to 20 Hz in
case of sound classes above the minimum requirement.

REFERENCES

ASTM E90-09 (2009). Standard test method for laboratory measurement of air-
borne sound transmission loss of building partitions and elements, American
Society for Testing and Materials, West Conshohocken, PA.
ASTM E336-16 (2016). Standard test method for measurement of airborne sound
attenuation between rooms in buildings, American Society for Testing and
Materials, West Conshohocken, PA.
ASTM E413-16 (2016). Classification for rating sound insulation, American Society
for Testing and Materials, West Conshohocken, PA.
ASTM E492-09 (2016). Standard test method for laboratory measurement of impact
sound transmission through floor-ceiling assemblies using the tapping machine.
American Society for Testing and Materials, West Conshohocken, PA.
ASTM E989-06 (2012). Standard classification for determination of impact
insulation Class (IIC), American Society for Testing and Materials, West
Conshohocken, PA.
ASTM E1007-16 (2016). Standard test method for field measurement of tapping
machine impact sound transmission through floor-ceiling assemblies and asso-
ciated support structures, American Society for Testing and Materials, West
Conshohocken, PA.
ASTM E1332-16 (2016). Standard classification for rating outdoor-indoor sound atten-
uation, American Society for Testing and Materials, West Conshohocken, PA.
R. Berger (1911). Über die schalldurchlässigkeit. (On the Sound Transmission, in
German). Dissertation, Munich, Germany.
K. Bodlund (1985). Alternative reference curves for evaluation of the impact sound
insulation between dwellings. Journal of Sound and Vibration 102, 381–402.
152  Sound Insulation in Buildings

J.S. Bradley (1982). Subjective Rating of the Sound Insulation of Party Walls.
Building Research Note 196. National Research Council Canada, Division of
Building Research.
O. Brandt (1958). Akustisk planering. (Acoustical planning, in Swedish). V. Pettersons
Bokindustri AB, Stockholm, Sweden.
C. Buratti, E. Belloni, E. Moretti (2014). Façade noise abatement prediction: New
spectrum adaptation terms measured in field in different road and railway traf-
fic conditions. Applied Acoustics 76, 238–248.
L. Cremer (1942): Theorie der Schalldämmung dünner Wände bei schrägem Einfall.
(Theory of sound insulation of thin walls at oblique incidence, in German).
Akustische Zeitschrift 7, 81–104.
J. Davy (2004). Insulating buildings against transportation noise. Proceedings of
ACOUSTICS 2004, 3–5 November 2004, Gold Coast, Australia, 447–453.
DIN 4109 (1959). Schallschutz im Hochbau, Entwurf. (In German). Berlin 1959.
(Final edition 1962).
DIN 4110 (1938). Technische Bestimmung für die Zulassung neuer Bauweisen. (In
German), 2nd edition. Berlin.
W. Fasold, E. Sonntag (1978). Bauakustik. Bauphysikalisches Entwurfslehre, Band 4,
3rd edition. VEB Verlag für Bauwesen. Berlin.
K. Gösele (1969). Schalldämmung von Türen (Sound insulation of doors, in German).
Berichte aus der Bauforschung, Heft 63.
S. Hveem, A. Homb, K. Hagberg, and J.H. Rindel (1996). Low-frequency footfall
noise in multi-storey timber frame buildings. NKB Work and Committee
Report 1996:12 E. Nordic Committee on Building Regulations, Acoustics
Group, Helsinki.
S. Ingemansson (1968). Ljudisolerande fönsterkonstruktioner. (Sound insulating
window constructions, in Swedish). Report from the Building Research, 3/68,
Stockholm, Sweden.
F. Ingerslev, A. Kjerbye Nielsen, and A. Falck Larsen (1947). The measuring of
impact sound transmission through floors. Journal of the Acoustical Society of
America 19, 981–987.
ISO 10140-2 (2010). Acoustics – Laboratory measurement of sound insulation
of building elements – Part 2: Measurement of airborne sound insulation,
International Organization for Standardization, Geneva, Switzerland.
ISO 10140-3 (2010). Acoustics – Laboratory measurement of sound insulation
of building elements – Part 3: Measurement of impact sound insulation,
International Organization for Standardization, Geneva, Switzerland.
ISO 16283-1 (2014). Acoustics – Field measurements of sound insulation in build-
ings and of building elements – Part 1: Airborne sound insulation, International
Organization for Standardization, Geneva, Switzerland.
ISO 16283-2 (2015). Acoustics – Field measurements of sound insulation in build-
ings and of building elements – Part 2: Impact sound insulation, International
Organization for Standardization, Geneva, Switzerland.
ISO 16283-3 (2016). Acoustics – Field measurements of sound insulation in build-
ings and of building elements – Part 3: Façade sound insulation, International
Organization for Standardization, Geneva, Switzerland.
ISO 717-1 (1996). Acoustics – Rating of sound insulation in buildings and of build-
ing elements – Part 1: Airborne sound insulation, International Organization
for Standardization, Geneva, Switzerland.
 Introduction to sound insulation  153

ISO 717-2 (1996). Acoustics – Rating of sound insulation in buildings and of build-
ing elements – Part 2: Impact sound insulation, International Organization for
Standardization, Geneva, Switzerland.
ISO/R717 (1968). Rating of sound insulation for dwellings, International
Organization for Standardization, Geneva, Switzerland.
J. Kristensen (1973). Lydisolation: Teori, måling, vurdering og bestemmelser, (Sound
insulation: Theory, measurement, evaluation and requirements, in Danish). SBI-
notat 24 (In Danish). Danish Building Research Institute, Aalborg University
Copenhagen.
J. Kristensen, J.H. Rindel (1989). Bygningsakustik – Teori og praksis (Building
acoustics – Theory and practice, in Danish) SBI-Anvisning 166. Danish
­
Building Research Institute, Hørsholm, Denmark.
P.T. Lewis (1971). Real windows, Chapter 7, in Building Acoustics (Eds. T. Smith,
P.E. O’Sullival, B. Oakes, R.B. Conn). British Acoustical Society, Special Volume
No. 2, Oriel Press, London.
R. Lindahl and H.J. Sabine (1940). Measurement of impact sound transmission
through floors. Journal of the Acoustical Society of America 11, 401–405.
J.A. Marsh (1971). The airborne sound insulation of glass, Chapter 5, in Building
Acoustics (Eds. T. Smith, P.E. O’Sullival, B. Oakes, R.B. Conn). British
Acoustical Society, Special Volume No. 2, Oriel Press, London.
NT ACOU 061 (1987). Windows: Traffic Noise Reduction Indices, Nordtest
Method. Helsinki, Finland.
SS 25267 (2015). Acoustics – Sound Classification of Spaces in Buildings, Swedish
Standards Institute, Stockholm.
Lydteknisk Institut. Building Acoustic Laboratory Measurements (Bygningsakustiske
laboratoriemålinger, in Danish). Lyngby, Denmark.
M. Mesihovic, J.H. Rindel, and I. Milford (2016). The need for updated traffic noise
spectra, used for calculation of sound insulation of windows and facades.
Proceedings of InterNoise 2016, Hamburg, Germany, 3890–3897.
Chapter 6

Sound radiation from plates

The sound radiation from plates is described in this chapter. Both forced and
resonant vibrations are considered, and the application of Rayleigh’s method
of radiation calculation is demonstrated in the case of radiation from forced
bending waves. The influence of the finite area of a plate is also shown. Forced
vibrations are the part of the vibrations that are directly due to the surround-
ing sound field exciting the plate. In contrast, resonant vibrations are the free
vibrations caused by reflections of the forced vibrations from the boundaries.

6.1  NORMAL MODES IN A RECTANGULAR PLATE

To study radiation of the reverberant part of the vibration field in a plate,

we need some basic information about this field. The modal density for the
plate is especially needed, as this is an important parameter for the statisti-
cal energy analysis (SEA). The reverberant vibration field will be analysed
with the aid of SEA in Chapter 7.
The normal modes are the free vibration patterns that can exist in a
finite system without excitation or after an excitation has stopped. They
are also responsible for building up the field when an excitation starts. Each
normal mode is associated with a natural frequency (resonance frequency)
and – for a given excitation – a modal amplitude. The normal modes are
orthogonal to each other, meaning among other things that the energy is
determined from the sum of the squared amplitudes of the normal modes.
For a plate, the normal modes shall fulfil the bending wave equation
(Equation 1.24) – if we can assume to be in the low-frequency region, i.e.
well below the cross-over frequency fs (Equation 1.29) – and simultaneously
fulfil the boundary conditions. In the present case, we are only interested
in the general trends (valid for all boundary conditions), so it is practical to
use the simplest of all situations – the simply supported rectangular plate.
The boundary condition of a simple support is that the displacement and
the moment reaction at the boundary is zero – the plate is, however, free
to rotate at the boundary. The moment reaction is related to the second
derivative of the displacement (or vibration velocity) field. The modes we

155
156  Sound Insulation in Buildings

are looking for are, therefore, to have zero value at the boundary, and the
second-order derivative should also be zero at this location. A function
fulfilling these conditions is sin (kxx) (convince yourself). It is also easy
to show that this function fulfils the bending wave equation (once again,
convince yourself). These two facts prove that the normal modes we are
looking for are sine functions. The wave number k is, however, still not
fixed. If there is a simple support both at x = 0 and x = lx, then kx must equal

mπ
kx, m = (6.1)
lx

where m is an arbitrary integer. (OBS: In other connections m is the mass per

unit area.) However, we are talking about a plate. Thus, the y-direction must
also be included. This direction will have the same type of field, sin (kyy) and

nπ
ky, n = (6.2)
ly

where ly is the plate length in the y-direction and n is an arbitrary integer.

A normal mode can then be written as

 mπ   nπ 
φm, n (x, y) = sin  x sin  y  (6.3)
 lx   ly 

Figure 6.1 shows an example of a normal mode in a plate. Some other

examples are shown in Figure 6.2. Normal modes can be used to find solu-
tions for the (forced) bending wave equation in the form
∞

w(x, y) = ∑ Aφ
m , n =1
n m, n (x, y) (6.4)

However, that is not what we shall do here. Instead, the starting point will
be the Pythagoras relationship between the free bending wave numbers and
the mode numbers m and n,
2 2
 mπ   nπ 
kb2 = km2 , x + kn2, y =  + (6.5)
 lx   ly 

If the bending wavelength is used instead of the wave numbers, the relation-
ship that shall be fulfilled is
2 2 2
 2  m  n 
 λ  =  l  +  l  (6.6)
B x  y
 Sound radiation from plates  157

λ/2
λy/2 ly

lx

λx/2

Figure 6.1 Example of a normal mode in a simply supported rectangular plate. In this

case, the mode numbers are (m, n) = (5, 3). The corresponding half wave-
length and direction of a propagating bending wave are shown.

where, as before, m and n are natural numbers (Figure 6.1). The numbers
represent the number of half wavelengths along the x- and y-axis, respec-
tively. Some examples are shown in Figure 6.2. With simply supported
edges, the lowest natural frequency is the one with (m, n) = (1, 1).
In general, it is a little complicated to calculate the natural frequen-
cies of the normal modes, because of the transition from bending to shear

(a) (b)

(c) (d)

Figure 6.2 Examples of normal modes in a simply supported rectangular plate, illus-

trated as maximum deflection. The modal numbers are (a) (1,1), (b) (2,1), (c)
(2,2), (d) (3,2). The arrows indicate the nodal lines.
158  Sound Insulation in Buildings

waves (see Equation 1.27). At low frequencies, however, it may normally

be acceptable to assume bending waves, i.e. the frequency is well below
the cross-over frequency for shear waves (Equation 1.29). Then, the wave-
length in the panel is λB = cB / f = c / f ⋅ fc and, thus, the natural frequency
of the mode with numbers (m, n) is:

c2  m 2  n 2
fmn =    +    (6.7)
4 fc   lx   ly  

Here c is the speed of sound in air and fc is the critical frequency

(Equation 1.28). Considering the distribution of the natural frequencies, it
can be found that, on average, the number of natural frequencies below a
certain frequency, i.e. for wavelengths longer than λB , is:

1 1 1
N = π ⋅ lx ⋅ ly − ( lx + ly ) + (6.8)
λB
2
λB 4
Assuming bending waves and introducing the area S = lx∙ly and the perim-
eter of the panel U = 2(lx + ly), the modal density yields:

∆N π S U fc π S (6.9)
= 2 fc − ≅ 2 fc
∆f c 4c f c

Similarly, by assuming shear waves, λB = c s /f and the modal density

yields:

∆N 2π S U 2π S (6.10)
= 2 f− ≅ 2 f
∆f cs 2c s cs

We notice that the modal density in a plate is constant in the bending wave
region, but increases with frequency in the shear wave region.
The modal density of a homogeneous, isotropic plate depends not only on
the length and width but also on the thickness. If the plate is thin (thickness
less than approximately one-sixth of the wavelength of bending waves),
bending waves can be assumed; but if the plate is thick compared to the
wavelength, shear waves will occur instead. If the bandwidth is one-third
octave, Δf = 0.23 f and, thus, the approximate number of modes within the
band is:

 0.23π
 2
Sfc f (f < f s / 2) (bending)
 c
∆N1/3 ≅ (6.11)
 0.46π Sf 2 (f > f s / 2) (shear)

 c s2

 Sound radiation from plates  159

Here fc is the critical frequency, fs = fc (cs /c)2 is the cross-over frequency for
transition to shear waves and cs is the speed of shear waves. The two for-
mulas in Equation 6.11 are shown graphically in Figure 6.3. The curves
cross at the frequency fs /2, so this is the cross-over frequency from bending
to shear when the modal density is considered. In other words, the cross-
over for modal density is one octave lower than the cross-over for speed of
propagation.
As an example, the first modes have been calculated for a 10 m 2 brick
wall, 250 mm thick (Table 6.1). The actual number of modes in each one-
third octave band is displayed in Figure 6.4 and compared with the sta-
tistical estimate (Equation 6.11). The shear wave cross-over frequency is
fs = 4 kHz, so for modal density, the bending wave assumption holds up
to 2 kHz. It is noted that up to 250 Hz, many one-third octave bands are
without any modes and the first modes from f 11 to f 22 are widely separated.
Actually, this interval is two octaves (f 22 = 4 f11) and there may be only
two or three modes within that range. Thus, the statistical estimates in
Equation 6.11 should only be used for frequencies above f 22 = 4 f11, that is
366 Hz in this example.

100.0

Bending
Number of modes per one-third octave band

Shear

10.0

1.0

1/2 fs

0.1
50 100 200 500 1,000 2,000 5,000 10,000
Frequency (Hz)

Figure 6.3 Number of normal modes per one-third octave bands in a 250 mm thick
brick wall, calculated for bending waves and shear waves with Equation 6.11.
fc = 72 Hz, fs ≈ 4000 Hz, f11 = 91.4 Hz. Below 315 Hz, there is statistically less
than one mode per band. Note that the ordinate is logarithmic.
160  Sound Insulation in Buildings

Table 6.1  F
 irst 14 modes calculated for a 250 mm
thick brick wall, 4 m × 2.5 m
m n fmn (Hz)
1 1 91.4
2 1 168
1 2 289
3 1 297
2 2 366
4 1 477
3 2 494
1 3 617
4 2 674
2 3 694
5 1 708
3 3 823
5 2 905
6 1 990
Number of modes per one-third octave band

7 Actual number
Statistical
6

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 6.4 Number of normal modes per one-third octave bands in a 250 mm thick brick
wall, 4 m × 2.5 m. The first normal mode is f11 = 91.4 Hz. The statistical modal density
calculated from Equation 6.11 is also shown. Bending waves only.

6.2 SOUND RADIATION FROM A

PROPAGATING TRANSVERSE WAVE

Consider a large plate located in the xy-plane with sound radiation into
the side where z > 0. A bending wave is assumed to travel in the plate as a
 Sound radiation from plates  161

plane wave in the x-direction. Hence, the velocity of the vibration can be
written as:

v = v e− jkBx (6.12)

The time factor ejωt is omitted here and in the following. The angular wave
number of the bending wave is k B = ω/cB . In the air above the plate, there
is no variation in the y-direction, so the sound pressure can be written in
the form:

p = p e− j(kx x+kz z) (6.13)

where kx and kz are the wave numbers that represent the sound propaga-
tion in the x- and z-directions, respectively. When this general solution is
inserted in the Helmholtz equation:

∇2 p + k2 p = 0 (6.14)

it is found that the wave numbers are related as:

k2 = kx2 + kz2 (6.15)

The z-component of the particle velocity in the air field is:

1 ∂p p kz − j(kx x+kz z)
uz = − = e (6.16)
jωρ ∂z ωρ

At the surface of the plate (z = 0), this must equal the normal velocity of the
plate (Equation 6.12) and thus:

p kz
v= and kx = kB (6.17)
ωρ

From Equation 6.15, it follows that

kz = k2 − kB2 (6.18)

Insertion in Equation 6.13 yields the following result for the sound pressure:

ρc v − j k2 −kB2 z
p= e− jkBxe (6.19)
1 − (kB / k)2

This result is discussed here for three cases.

162  Sound Insulation in Buildings

The first case is k B < k (meaning that cB > c and f > fc). The solution rep-
resents a sound field propagation in both the x- and z-directions. The angle
of propagation (Figure 6.5c) is determined by the following:

c k fc
sin θ = = B = (6.20)
cB k f

In this case, the sound pressure can be written in terms of the angle of
radiation, θ:

ρc v − j(x sin θ +z cos θ )

p= e (6.21)
cos θ

The second case is k B > k (meaning that cB < c and f < fc). The sound pres-
sure (Equation 6.19) yields:
jρ c v − j kB2 − k2
p= e− jkB xe z (6.22)
2
(kB / k) − 1

This represents a sound field that only propagates in the x-direction, but
not in the z-direction; the amplitude of the sound pressure decreases expo-
nentially with the distance from the plate. There is an acoustic near field,
but no sound radiation (Figure 6.5a).

(a) (b) (c)

–
θ

+ CB CB
CB C
CB
C
–
–

f < fc f = fc f > fc

Figure 6.5 A propagating bending wave in a plate cannot radiate sound if the speed
is below c (case a). If the speed equals c, the sound is radiated along the
surface of the plate (case b). A plane wave is radiated in the direction θ if
the speed exceeds c (case c). (Reproduced from Kristensen, J., Rindel J.H.,
Bygningsakustik – Teori og praksis (Building acoustics – Theory and practice,
in Danish). SBI Anvisning 166. Danish Building Research Institute, Hørsholm,
Denmark, 1989. With permission.)
 Sound radiation from plates  163

(a) (b)

Figure 6.6 Simulated particle movements in the air due to a bending wave travelling from
left to right. (a) cB < c. (b) cB > c. (Reproduced from Heckl, M. and Heckl,
M., Darstellung von abstrahlvorgängen im film. Fortschritte der Akustik – DAGA
1976, VDI-Verlag, Düsseldorf, 165–168, 1976. With permission.)

The third case is the transition from non-radiation to radiation, which

is at k B = k, i.e. when the bending wave number equals the wave number in
air, or the speed of the propagating bending wave equals the speed of sound
(Figure 6.5b). This is called coincidence. Also note that when coincidence
occurs, the sound radiation is extremely high since p → ∞ as the denomina-
tor of Equation 6.22 becomes zero.
A simulation of the sound field above a propagating bending wave with-
out and with sound radiation is shown in Figure 6.6.

6.3 THE RADIATION EFFICIENCY

The result in the previous section is for structures with infinite extent.
However, real structures are finite and this has an important effect on the
sound radiation. The rest of this chapter is about finite structures. In order
to study the radiation of these structures, radiation efficiency is introduced.
The concept of radiation efficiency was first introduced by Gösele (1953).
Radiation efficiency is a measure of the radiated sound power from a source
relative to the sound power radiated from an equivalent piston source that
generates a plane wave. The sound power from a very large piston with a
harmonic vibration v = |v| ejωt and the area S is:
2
Ppiston = 1
2 ρ cS v = ρ cS v 2 (6.23)

Here v denotes the RMS (root mean squared) value of the velocity, and the
brackets mean the average taken over the surface of the vibrating source.
If the actual radiated sound power is P from a surface with area S and
velocity v, the radiation efficiency σ is defined as:
P
σ = (6.24)
ρ cS v 2
164  Sound Insulation in Buildings

The radiation efficiency has no dimension and it may take values between
0 and around 4. Values higher than unity are possible because bending waves
in a plate may radiate more efficiently than the piston, as we shall see later.
The radiation efficiency will be used to describe the radiated sound power
from various kinds of vibrations. When sound is transmitted through a
plate, the radiated sound power can be assigned to two different kinds of
vibrations in the same plate: forced and resonant vibrations. As these two
vibration fields have different physical origins and different wave numbers
associated with them, the radiation efficiency and velocity magnitude will
also be very different. The resonant vibration is associated with free bend-
ing wave numbers and the forced vibration is associated with the wave
number in the sound field in the air. The total transmission through the
wall is found by adding together the transmitted sound powers:

Ptot = Pfor + Pres = ρ cS v for

2
σ for + ρ cS v res
2
σ res (6.25)

6.4 SOUND RADIATION CALCULATED

BY RAYLEIGH’S METHOD

In this section, a general method for calculating the sound radiation from a
vibrating surface is introduced – the so-called Rayleigh’s method to be used
in the following sections. (Rayleigh, 1896, §278, Equation 3). The method
is described in Lord Rayleigh’s “The Theory of Sound” first p­ ublished 1877,
and the steps are the follows:

dS΄
β0 r
dS ro α0
Iy X
2
Iy θr
2
Ix Ix
2 2
Z

Figure 6.7 A sound-radiating surface surrounded by a large, rigid baffle.

 Sound radiation from plates  165

• Knowing the velocity pattern of the vibrating panel, the velocity

potential (and thus the sound pressure) is found in an arbitrary point
by integration over the surface of the panel.
• The squared sound pressure is integrated over the surface of an infi-
nite half-sphere.
• The result is the radiated sound power.

The surface is assumed to be rectangular with the dimensions lx and

ly and the position in a coordinate system is shown in Figure 6.7. A small
area dS can be considered as a point source with the volume velocity v dS.
Since the radiation into a half sphere is considered, the sound pressure in
the distance r is

vdS − jkr
p = jωρ e (6.26)
2π r

Here and in the following, the time factor e− jω t is omitted. The total sound
pressure at the point of observation can be found by integration over the
surface of the plate:

∫ ver
− jkr
jωρ
p= dS (6.27)
2π S

The distance from the centre of the plate to the point of observation is r 0
and it is assumed that this is large compared to the dimensions of the plate.
In Equation 6.27, r appears in two places – in the divisor, the approxima-
tion r ≈ r 0 is sufficient, whereas in the exponent, a more accurate approxi-
mation is needed:

r ≈ r0 − (x cos α 0 + y cos β0 ) (6.28)

Here x and y are the coordinates of the source point dS, and the angles α 0
and β 0 are as defined in Figure 6.7. With these approximations, the equa-
tion becomes:

p=
jωρ − jkr0
2π r0
e ∫ ve
S
jk( x cos α0 + y cos β0 )
dS (6.29)

With this equation, the sound pressure can be calculated in any direction
(α 0, β 0) if the velocity v of the plate is known as a function of the position
(x, y).
2
The radiated sound intensity at the point of observation is I = 12 p ρ c,
and the total sound power radiated into the half sphere S′ = 2π r 0 is:
2

P= ∫ S'
IdS′ =
r02
2ρc ∫ 2π
2
p dψ (6.30)
166  Sound Insulation in Buildings

6.5 SOUND RADIATION FROM FORCED

VIBRATIONS IN A RECTANGULAR PLATE

We are now ready to study sound radiation and the radiation efficiency of a
finite plate. The theoretical derivation and experimental verification of the
results were done by Rindel (1975). To start with, consider the case of forced
vibrations (the resonant vibration case will be discussed in Section 6.6).
It is assumed that the vibrations of the plate are generated by a plane
sound wave incident from the side z < 0. The angles of incidence relative to
the axis of the coordinate system are α, β and θ, respectively (Figure 6.8).
The sound pressure of the incident sound can be described as:

pi = pi e− jk(x cos α +y cos β +z cos θ ) (6.31)

Accordingly, the forced vibrations of the plate can be written as:

v = v e− jk(x cos α +y cos β ) (6.32)

At this point of our discussion, it is not important whether there is a phase

shift between sound pressure and velocity. For the sake of simplicity, it is
assumed that the velocity amplitude is constant within the area of the plate
and zero on the rigid baffle surrounding the plate. Then, the sound pressure of
the radiated sound field is found by inserting Equation 6.32 in Equation 6.29:

1l 1l
2 y 2 x

p=
jωρ
2π r0
v e− jkr0 ∫ ∫e jkx(cos α0 −cos α ) jky (cos β0 −cos β )
e dx dy
− 12 ly − 12 lx

j 1 klx (cos α0 −cos α ) − j 1 kl (cos α0 −cos α )

jωρ e2 −e 2 x
= v e− jkr0 ⋅
2π r0 jk(cos α0 − cos α)
j 1 kly (cos β0 −cos β ) − j 1 kl (cos β0 −cos β )
e2 −e 2 y
⋅
jk(cos β0 − cos β )

Even if this result looks complicated, we note that the variables are sepa-
rated. With the use of Euler’s formula (Equation 1.40), the result can be
rewritten in the simple form:

jωρ sin X sin Y

p= v Se− jkr0 ⋅ ⋅
2π r0 X Y

jωρ
= v Se− jkr0 ⋅ sinc(X) ⋅ sinc(Y ) (6.33)
2π r0
 Sound radiation from plates  167

where S is the area of the plate and

X = 12 klx (cos α 0 − cos α )

Y = 12 kly (cos β0 − cos β )

(6.34)

The function sinc (X) = sin X/X is symmetric (same value for positive and
negative X), it has a maximum of 1 for X = 0 and → 0 for X → ∞ (Figure 1.5).
This means that the sound radiation is maximum in the direction α 0 = α and
β 0 = β, i.e. in the same direction as the incident sound, as could be expected.
The result also shows that smaller amounts of sound are radiated in other
directions; especially for small values of klx and kly, i.e. if the plate dimen-
sions are small compared to the wavelength. If the ratio of dimension to
wavelength is very small, the radiation approaches an omnidirectional pat-
tern in a half sphere.

The characteristic dimension a is introduced as a general measure of the

size of a plate:

2S lxly
a= = (6.35)
U lx + ly

where S is the area and U is the perimeter. 2a is in fact the harmonic aver-
age of lx and ly (see Equation 1.33), and a must be between 0.5 and 1
times the shorter dimension of the plate. In the special case of a square
plate, a = 1 / 2 S . If the sides are very different in length, the characteristic
dimension does not exceed the shorter dimension.

α
θ
β

pi

Figure 6.8 Definition of the angles of incidence of the plane sound wave generating the
forced vibrations in the plate.
168  Sound Insulation in Buildings

(a)
1
0.9
0.8 ka = 1/2
Radiation function, sinc2 (X)

0.7 ka = 1
0.6 ka = 2
θ = 0° 0.5
ka = 4
0.4
ka = 8
0.3
ka = 16
0.2
ka = 32
0.1
0
–90 –60 –30 0 30 60 90
Angle of radiation, (θ0)
(b)
1
0.9
0.8 ka = 1/2
Radiation function, sinc2 (X)

0.7 ka = 1
0.6 ka = 2
θ = 45° 0.5
ka = 4
0.4
ka = 8
0.3
ka = 16
0.2
ka = 32
0.1
0
–90 –60 –30 0 30 60 90
(c) Angle of radiation, (θ0)
1
0.9
ka = 1/2
Radiation function, sinc2 (X)

0.8
0.7 ka = 1
0.6 ka = 2
θ = 90° 0.5 ka = 4
0.4
ka = 8
0.3
ka = 16
0.2
ka = 32
0.1
0
–90 –60 –30 0 30 60 90
Angle of radiation, (θ0)

Figure 6.9 Radiation function sinc2(X) showing the sound radiation pattern from forced
bending waves. Three examples of the angle of incidence, (a) 0°, (b) 45°,
(c) 90°, are shown.
 Sound radiation from plates  169

The radiation pattern is shown in Figure 6.9 for the two-dimensional case,
i.e. β0 = β = π/2. It is seen that the radiation has little directivity for low ka, but
gets more directive for high values of ka. In the case of grazing incidence,
θ = 90°, the radiation covers a relatively wide angle even for high values of ka.
The total radiated sound power into the half sphere z > 0 is found by
inserting Equation 6.33 in Equation 6.30:

P=
1
2ρc ∫ ω2ρ2 2 2
2 π 4π
2
v S (sincX ⋅ sincY ) dψ
2

=
ρc 2 2 2
8π 2
⋅k S v ∫ 2π
(sincX ⋅ sincY )
2
dψ

(6.36)

This integral cannot be solved analytically; so, instead, numerical approxi-

mations can be made. The radiation efficiency is denoted by σ θ when the
radiation is due to a forced bending wave generated by a sound field with
the angle of incidence θ:

σθ =
1
2 ρcS v
P
2
=
k2 S
4π 2 ∫ 2π
(sincX ⋅ sincY )
2
dψ (6.37)

The result of a numerical integration is shown in Table 6.2 and Figure 6.10

as a function of the angle of incidence θ in the case of a square plate.

Table 6.2  Calculated radiation efficiency for forced vibrations in a square plate
ka 0° 15° 30° 45° 60° 75° 90° Random
0.5 −8.3 −8.3 −8.3 −8.4 −8.5 −8.6 −8.6 −8.50
0.75 −5.0 −5.1 −5.2 −5.4 −5.6 −5.7 −5.7 −5.53
1 −3.0 −3.0 −3.3 −3.6 −3.9 −4.1 −4.2 −3.79
1.5 −0.6 −0.8 −1.1 −1.7 −2.2 −2.6 −2.7 −2.01
2 0.3 0.2 −0.2 −0.8 −1.3 −1.8 −2.0 −1.15
3 −0.1 0.2 0.5 0.4 −0.2 −0.7 −0.9 −0.14
4 −0.3 0.0 0.7 1.0 0.6 0.1 −0.2 0.41
6 0.0 0.2 0.6 1.6 1.7 1.1 0.7 1.16
8 0.1 0.2 0.5 1.6 2.4 1.8 1.4 1.61
12 0.0 0.2 0.6 1.4 3.1 2.9 2.3 2.18
16 0.0 0.2 0.6 1.5 3.2 3.7 3.0 2.57
24 0.0 0.2 0.6 1.5 2.9 4.7 3.9 3.04
32 0.0 0.2 0.6 1.5 3.0 5.4 4.6 3.36
48 0.0 0.2 0.6 1.5 3.0 6.0 5.5 3.77
64 0.0 0.2 0.6 1.5 3.0 6.1 6.2 4.04
Source: Sato, H., J. Acoust. Soc. Jpn., 29, 509–516, 1973.
170  Sound Insulation in Buildings

10

5
ka ∞
64
10 lg σθ (dB)

32
16
8
4
0
2
1

–5
0° 15° 30° 45° 60° 75° 90°
Angle of incidence (θ)

Figure 6.10 Radiation efficiency for forced vibrations in a square plate as a function of

the angle of incidence for different values of ka. (Adapted from Sato, H., J.
Acoust. Soc. Jpn., 29, 509–516, 1973.)

It is particularly interesting to note the behaviour when the angle of

incidence approaches grazing incidence; the radiated power is limited,
instead of growing to infinity as it would theoretically for an infinite
plate.
Table 6.2 also gives the result of an additional integration over all angles
of incidence in a diffuse sound field, and this is the radiation efficiency for
random incidence.
The limiting case of an infinitely large plate relative to the wavelength is:

1
σθ → for ka → ∞ (6.38)
cos θ
 Sound radiation from plates  171

10

Random
10 lg σθ (dB)

0 90°
60°
45°
0°

–5

–10
0.5 1 2 5 10 20 50 100
ka

Figure 6.11 Radiation efficiency for a square plate with forced bending waves for differ-
ent angles of incidence and for random incidence.

6.5.1 Approximate results
The following approximation to the result in Equation 6.37 was derived by
combining three asymptotic results for high frequencies, for low frequen-
cies and for grazing incidence θ = 90° (Rindel 1993).

−1/ 4
 4

π   π  
2 2
  π  
σ for,θ ≅  cos θ − 0.6  + π  0.6  + 
2
 (6.39)
 ka   ka   2 ( ka )2  
 

This approximation is shown in Figure 6.11 as a function of ka for some

values of θ.
The radiation efficiency for random incidence is found by another
numerical integration over all angles of incidence. The result is also shown
172  Sound Insulation in Buildings

in Figure 6.11. A numerical approximation for a square plate and valid for
ka > 0.5 is:

1
σ for, d ≅
2
(0.20 + ln(2ka)) (6.40)
(excitation by a diffuse sound field, so the index d). Usually, Equation 6.40
is sufficiently accurate for an estimate. However, a more exact formula (still
for ka > 0.5) was derived by Sewell (1970):

σ for, d =
1
2( (
0.160 + ln(2ka) + 16π (ka)2 )
−1
)
+ C(Λ) (6.41)

Here C is a function of the edge ratio Λ = lx /ly, where lx ≥ ly. In the original
paper by Sewell (1970), C was given in a table for values of Λ between
1 and 10. The function can be approximated by:

C(Λ) = −0.0022 ⋅ Λ2 + 0.0821 ⋅ Λ − 0.08 (6.42)

0
10 lg σ (dB)

Λ=1

–5
Λ=2

Λ = 10

–10
0.1 1 10 100
ka

Figure 6.12 R
 adiation efficiency for a plate with forced bending waves excited by a dif-
fuse sound field as a function of ka for three different edge ratios.
 Sound radiation from plates  173

The radiation efficiency for diffuse sound field excitation is displayed in

Figure 6.12.
In the low-frequency region, ka < 0.5, the radiation is weak and not
directive. Thus, we can apply the low-frequency asymptotic result for a
square piston in a large rigid baffle (Stenzel 1952):

2
σ for ≅
π
(ka )2 (6.43)

This is valid for ka < 0.5 and for any angle of incidence including random
incidence.

10

5
ka ∞
64
10 lg σ (dB)

32
16
8
4
0
2
1

–5
1 2 5 10
(cB/c)2 ≈ f/fc

Figure 6.13 Radiation efficiency for a square plate as a function of the phase speed of
bending waves relative to the speed of sound in air.
174  Sound Insulation in Buildings

6.6 SOUND RADIATION FROM

SUPERSONIC BENDING WAVES

The solution to the sound radiation problem as derived earlier is not

restricted to the case of forced vibrations. There is a direct relation between
the phase speed of the propagating bending wave and the angle of incidence
(see Equation 6.20). This means that the radiation efficiency as displayed
in Figure 6.10 can be considered as a function of the relative speed of bend-
ing waves, instead of as a function of the angle of incidence. As long as
the bending wave is supersonic (cB > c), it does not matter whether it is a
forced or a free bending wave, the physical principle of sound radiation is
the same. So, the radiation efficiency can be displayed in a new way using
(cB /c)2 ≈ f/fc as the independent parameter (Figure 6.13).
The usual assumption that sound radiation is maximum at the critical
frequency is true with modifications. We can observe from Figure 6.13 that
the radiation is maximum at a frequency a little higher than the critical fre-
quency for a finite plate with high ka. However, for ka < 4, the maximum is
shifted up about one octave or more and the peak has almost disappeared.
The radiation is weak for very low ka. We also observe that for ka > 2, the
radiation efficiency has the asymptotic value of unity (0 dB) for f ≫ fc.
Thin plates like a gypsum board wall or a glass pane in a window have
high ka values at the critical frequency (fc ≈ 2.5 kHz to 4 kHz) and thus,
a clear peak of radiation at the critical frequency. But thick plates like a
masonry wall or a concrete slab have low ka values at the critical frequency
(fc ≈ 100 Hz) and thus, no peak of radiation. In Section 6.7, we shall elabo-
rate more on these findings.

6.7 SOUND RADIATION FROM RESONANT

VIBRATIONS IN A RECTANGULAR PLATE

We shall now consider the sound radiation from resonant vibrations of a

plate. Sound radiation is mainly dependent on two conditions: the ratio
between propagation speeds in the plate and in the air (the Mach number),
and the size of the plate compared to the wavelength in air. The first con-
dition determines whether the radiation is at all efficient or not, while the
second condition determines an upper limit for the radiation efficiency.
The excitation of resonant vibrations can be either mechanical impact,
structural vibrations (e.g. flanking transmission) or airborne excitation
from an incident sound wave. As we have seen in the beginning of this
chapter, the normal modes have wavelengths λB = cB/f that are determined
solely by the dimensions of the plate and the modal numbers (m, n) (see
Equation 6.6). Thus, for a given frequency, the wavelength of the plate
vibrations can be either smaller than or larger than the wavelength in air.
In other words, it is important to determine whether the frequency is below
 Sound radiation from plates  175

(a) (b) (c)

+ + –
+
– – +
(1,1) (1,2) (2,2)

Figure 6.14 Radiation from low-numbered modes, (a) (1,1), (b) (1,2) and (c) (2,2) in
a plate. The shaded areas are those radiating sound. Areas with + and − are
in opposite phase.

or above the critical frequency of the plate. If f < fc , we have cB < c, λB < λ,
and vice versa for f > fc.
At the first normal mode (1,1) of a simply supported plate (Figure 6.14a),
all parts of the plate move in phase and the sound radiation is similar to
that from the diaphragm of a loudspeaker. At the second and third normal
mode (1,2) or (2,1), the movements of the plate are divided by one nodal
line (Figure 6.14b). Below the critical frequency, the pressure variations
are cancelled across the modal line, but sound is radiated from a zone near
the edges. If we consider the next normal mode (2,2), the sound radiation
is further limited to four zones near the corners (Figure 6.14c). The radia-
tion efficiency of these first modes is displayed in Figure 6.15 as functions
of the relative speed cB /c. All modes have radiation efficiency σ ≅ 1 above

10

0
10 lg σm,n (dB)

–10

(1,1)
(1,2)
–20
(2,2)

–30
0.1 1
cB/c

Figure 6.15 Radiation efficiency for low-numbered modes, (1,1) (1,2) and (2,2). (Principle
adapted from Wallace, C.E., J. Acoust. Soc. Am., 51, 946–952, 1972, Figure 2.)
176  Sound Insulation in Buildings

the critical frequency (cB /c > 1); but at lower frequencies, the radiation is
less efficient. The sound radiation from normal modes in plates has been
analysed by Wallace (1972), but here it is sufficient to look at approximate
results for a square panel.

27 Sf 2 4Sf 2
σ res (1,1) = ⋅ ≅ 2 mode(1,1) (6.44a)
π 3 c2 c

25 S 2 f 4 3.4S 2 f 4
σ res (1, 2) = ⋅ ≅ mode(1,2) (6.44b)
3π c 4 c4

8π S3 f 6 1.7 S3 f 6
σ res (2, 2) ≅ ⋅ ≅ mode(2,2) (6.44c)
15 c 6 c6

Modes with higher mode numbers radiate from the edges or from the cor-
ners when the frequency is below the critical frequency, but from the entire
area above the critical frequency (Figures 6.16 and 6.17). There is cancel-
lation of sound pressures across the nodal lines of the plate as long as the
wavelength in the plate is shorter than the wavelength in the air. Therefore,
some mode patterns give radiation along two edges and other mode pat-
terns only allow the four corners to radiate sound. Above the critical fre-
quency, the entire surface contributes to the radiation and σ → 1 at high
frequencies.
The following result for the radiation efficiency was derived by Maidanik
(1962) for the case of resonant vibrations in a plate generated by random
incidence sound at frequencies below the critical frequency.

c2 Uc
σ res = 2
⋅ g1(M) + ⋅ g 2 (M) (f11 < f < fc ) (6.45)
Sfc Sfc

where g 1 and g 2 are help functions:

 8(1 − 2M 2 )
 (M < 2 / 2)
g1(M) =  π 4 M 1 − M 2 (6.46)

 0 (M ≥ 2 / 2)

1  (1 − M )ln ((1 + M) (1 − M)) + 2M 

2
g 2 (M) =  (6.47)
4π 2  (1 − M 2 )3/ 2 
 Sound radiation from plates  177

λ/2 λ/2 λ/2

Figure 6.16 Radiation from a high-numbered mode (3,6) scaled to different frequencies.

The shaded areas are those radiating sound.

10

C: Surface
0
10 lg σm,n (dB)

–10

B: Edges

–20
A: Corners

–30
0.1 1
cB/c

Figure 6.17 Radiation efficiency for high-numbered modes. A: Corner radiation from,

e.g. mode (11, 11) or (12, 12). B: Edge radiation from, e.g. mode (1, 11) or
(1, 12). C: Surface radiation from all modes when cB/c > 1. (Principle adapted
from Wallace, C.E., J. Acoust. Soc. Am., 51, 946–952, 1972, Figure 3.)

Here the Mach number M is used as a frequency parameter.

cB,eff f
M= ≈ (6.48)
c fc

The parameter M is the Mach number, as it gives the ratio between the
speed of the travelling bending wave and the speed of sound in air. The
approximation using the critical frequency is only valid for thin plates,
whereas a more general result is obtained by using the ratio between the
propagation speed in the plate and in the air, cB,eff /c.
178  Sound Insulation in Buildings

The derivation of Equation 6.45 is based on the following assumptions:

• The modal density is high.

• All modes have the same energy.
• Contribution from ‘even-odd’ and ‘even-even’ modes is negligible
compared to the contribution from ‘odd-odd’ modes.

The first assumptions means in reality that the result is not valid below
f 22 = 4 f11 (see Section 6.1).
Above the critical frequency, we have radiation from supersonic vibra-
tions and for ka ≫ 1, the radiation efficiency is:

( )
−1/2
≅ (1 − fc f )
−1/2
σ res = 1 − M −2 (f > fc ) (6.49)

At frequencies near the critical frequency fc , the radiation efficiency is

determined by a maximum related to the dimensions of the plate compared
to the wavelength, as discussed in Section 6.6. A simple empirical approxi-
mation for the maximum is:

π fU
σ res ≅ (f ≈ fc ) (6.50)
16c
Combining the aforementioned equations into one:

σ res = (σ 1−4 + σ 2−4 + σ 3−4 )−1/ 4 (6.51)

10

4000
10 lg σ (dB)

2000
–10
1000
500

–20 250
125

–30
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 6.18 Radiation efficiency for resonant vibrations in plates with dimension

4.0 m  ×  2 .5 m and critical frequencies from 125 Hz to 4000 Hz.
 Sound radiation from plates  179

where σ1, σ 2 and σ 3 are from Equations 6.45, 6.49 and 6.50, respectively. In
this way, the transition between each equation is smoothed.
Some calculation examples are shown in Figure 6.18. The previous obser-
vations concerning a peak near the critical frequency for thin plates and a
very flat radiation course for thick plates are clearly seen. For frequencies
above the critical frequency, the results shown in Figure 6.18 are in fact the
same as those in Figure 6.13.

6.7.1 Approximate results
Assuming bending waves, i.e. plates that are relatively thin and suffi-
ciently large so that the first mode of vibration f 11 ≪ fc , we get the following
approximations for the frequency range below the critical frequency (also
shown in Figure 6.19).

Uc f  3c 1 
σ res ≅  < f < fc  (6.52)
π 2 Sfc fc U 2 

c2  3c 
σ res ≅  f11 < f <  (6.53)
Sfc2 U

4Sf 2
σ res ≅ (f < f11) (6.54)
c2

10
fc

fb
10 lg σ (dB)

–10

–20

f11 f22 3c/U fc /2

–30
10 20 50 100 200 500 1,000 2,000 5,000 10,000
Frequency (Hz)

Figure 6.19 Radiation efficiency from resonant vibrations in a large thin plate (f11 < fc).
The maximum radiation from Equation 6.50 is shown as a dashed line.
180  Sound Insulation in Buildings

10
fc

0
10 lg σ (dB)

–10 fa f11 fb f22

–20
10 20 50 100 200 500 1,000 2,000 5,000 10,000
Frequency (Hz)

Figure 6.20 Radiation efficiency from resonant vibrations in a small thick plate (f11 > fc).
The maximum radiation from Equation 6.50 is shown as a dotted line.

The area is S and U is the perimeter. The range between ½ fc and fc is

not covered by any formula, but a simple interpolation may be sufficient.
The frequency calculated from 3c/U represents the approximate transition
from mainly corner radiation at low frequencies to mainly edge radiation at
somewhat higher frequencies. The frequencies that separated the range for
Equations 6.49 through 6.54 are marked in Figure 6.19. In addition, the f 22
frequency is shown, which is the lower frequency limit for application of
statistical modal density (see Section 6.1).
As an example, a 6.5 mm glass pane with dimensions 1.4 m  ×  1.4 m is
considered (Table 6.3). The first normal mode is at 15 Hz and the statistical
modal density may be applied above 60 Hz.

Table 6.3  E
 xamples of key frequencies for a 6.5 mm
glass pane and a 250 mm brick wall
Glass pane Brick wall
h, mm 6.5 250
lx , m 1.4 4
ly, m 1.4 2.5
S, m2 2 10
f11, Hz 15 91.5
f21, Hz 37 168
f12, Hz 37 289
f22, Hz 60 366
fa, Hz 89 43
3c/U, Hz 183 79
fc, Hz 2000 72
fb, Hz 2305 204
 Sound radiation from plates  181

In contrast, Figure 6.20 shows the radiation efficiency in the case of a

thick plate. The critical frequency is below the first normal mode of the
plate, fc < f11. This implies that there are no resonant modes available
around and below the critical frequency, so the regions of edge radiation
and corner radiation have disappeared from the graph in this case. Another
difference between Figures 6.19 and 6.20 is the frequency of maximum
radiation. Whereas the maximum is close to the critical frequency of the
thin plate, this is not the case for the thick plate. The maximum radiation
is shifted to a somewhat higher frequency, f b, which can be found as the
frequency of the crossing of the two expressions, Equations 6.49 and 6.50:

5c
fb = fc + (6.55)
U
At very low frequencies, the crossing of Equations 6.50 and 6.54 is at the
frequency fa:

πU
fa = 3
64c
( f11fc ) (6.56)
2

As an example of a thick plate, a 250 mm brick wall with dimensions

2.5 m × 4.0 m is considered (Table 6.3). Here 5c/U = 5 × 344/13 = 132 Hz,
which means that although the critical frequency is around fc = 72 Hz, the
frequency of maximum radiation is actually shifted more than one octave
up, to 204 Hz. The course of the curve in Figure 6.20 is quite flat. Actually,
the radiation efficiency is close to unity (0 dB) at all frequencies above
50 Hz.

6.7.2 Directivity of resonant radiation

The radiation pattern of resonant vibrations in a plate can be found for
each of the normal modes by integration of the velocity distribution over
the plate according to Rayleigh’s method. The problem is reduced from
three to two dimensions by assuming that the plane of radiation lies in the
xz-plane. So, only modes related to the lx dimension of the plate are con-
sidered. Taking  only the most important contributions into account (the
‘odd-1’ mode), the approximate result is:

k2 S
3 (
sincX1 + sincX2 ) (6.57)
2
σ res,θ =
2π

where

1 1
X1 = klx ( sin θ r + c cB ) and X2 = klx ( sin θ r − c cB ) (6.58)
2 2
 182  Sound Insulation in Buildings

(a) f = 0.6 fc (b) f = 0.8 fc

60 60

50 50

Radiation efficiency,(σ)
Radiation efficiency,(σ)

40 40

30 30

20 20

10 10

0 0
–90 –75 –60 –45 –30 –15 0 15 30 45 60 75 90 –90 –75 –60 –45 –30 –15 0 15 30 45 60 75 90
Angle of radiation,(θr) Angle of radiation,(θr)
ka = 2π ka = 4π ka = 8π ka = 16π ka = 2π ka = 4π ka = 8π ka = 16π
(c) f = fc (d) f = 4/3 fc
60 60

50 50
Radiation efficiency,(σ)

Radiation efficiency,(σ)

40 40

30 30

20 20

10 10

0 0
–90 –75 –60 –45 –30 –15 0 15 30 45 60 75 90 –90 –75 –60 –45 –30 –15 0 15 30 45 60 75 90
Angle of radiation,(θr) Angle of radiation,(θr)
ka = 2π ka = 4π ka = 8π ka = 16π ka = 2π ka = 4π ka = 8π ka = 16π

(e) f = 2fc (f) f = 4fc

60 60

50 50
Radiation efficiency,(σ)

Radiation efficiency,(σ)

40 40

30 30

20 20

10 10

0 0
–90 –75 –60 –45 –30 –15 0 15 30 45 60 75 90 –90 –75 –60 –45 –30 –15 0 15 30 45 60 75 90
Angle of radiation,(θr) Angle of radiation,(θr)
ka = 2π ka = 4π ka = 8π ka = 16π ka = 2π ka = 4π ka = 8π ka = 16π

Figure 6.21 Examples of the radiation efficiency as a function of the angle of radiation

calculated by Equation 6.57. (a) f = 0.6 fc, (b) f = 0.8 fc, (c) f = fc, (d) f = 1.33
fc, (e) f = 2 fc, (f) f = 4 fc.
 Sound radiation from plates  183

It is seen from the result (Equation 6.57) that the radiation has two domi-
nating, symmetric lopes as long as cB > c (f > fc). The main direction of
radiation is:

θ r = ± arcsin(c cB ) = ± arcsin ( )
fc f (6.59)

Some examples of calculated radiation patterns are shown in Figure 6.21. At

frequencies below the critical frequency, the radiation is weak and without
any specific directionality, but at higher frequencies, the radiation is highly
directional and symmetric around the normal of the plate. For increasing
plate dimensions compared to the wavelength (ka increasing), the directivity
of the radiation becomes sharper. At and just below the critical frequency, the
radiation is mainly in directions close to 90°, but covering a relatively broad
angle, especially for low ka values. Above the critical frequency, we  have
radiation from supersonic bending waves, and the radiation is highly direc-
tive in accordance with Equation 6.59, especially for high ka values.

6.7.3 Combined forced and resonant radiation

A summary of the radiation properties of forced and resonant vibrations in
a plate is given in Table 6.4.
A sound wave incident on a plate generates a forced bending wave, and
when the vibrations are reflected from the boundaries, the resonant modes
of vibration are excited, mainly those with a natural frequency close to
the frequency of the incident sound. At high frequencies, where the modal
density in the plate is high, many modes will be excited simultaneously and
contribute to resonant radiation, but at low frequencies, the modal density
in the plate is low and the resonant radiation can only take place at the
natural frequencies of the plate.

Table 6.4  Comparison of forced and resonant vibrations

Forced vibrations Resonant vibrations
Excitation Incident sound wave Any excitation
Wave type Propagating wave Standing wave
Phase speed cB Depends on angle of Depends on frequency
incidence cB = c / sin θ ≥ c cB = c f / f c
Direction of radiation One direction No direction for f < fc
θ r ≅ arcsin ( c / cB ) Two directions for f ≥ fc
(
θ r ≅ ± arcsin fc / f)
Strength of radiation σ < 1 for ka < 10 σ ≪ 1 for f < fc
σ > 1 for ka > 10 σ ≅ 1 for f > fc
184  Sound Insulation in Buildings

(a) (b)
f < fc f = fc

(c) (d)
f = 2fc f = 4fc

Figure 6.22 Sound radiation from a panel with excitation by a plane wave at normal
incidence. (a) Below the critical frequency. (b) At the critical frequency. (c)
One octave above the critical frequency. (d) Two octaves above the critical
frequency. Dotted line arrows indicate resonant radiation.

Examples of combined forced and resonant sound radiation are shown in

Figure 6.22. Below the critical frequency, the resonant vibrations can only
radiate from corners or edges, and the contribution is relatively weak, but
above the critical frequency, the resonant radiation is strong and may be
more important than the forced transmission.

6.8 THE REFRACTION EFFECT

The sound transmission through a surface of finite area includes diffraction

effects when the wavelength is not very small compared to the character-
istic dimension of the surface. Diffraction effects include the scattering of
radiated sound in many directions and the refraction of transmitted sound
for incident sound near grazing incidence. Hence, the angle of radiation
may be different from the angle of incidence (Figure 6.23).
We return to Figure 6.10, which shows that near the grazing incidence,
i.e. θ between 75° and 90°, the radiation from a forced bending wave is lim-
ited and the maximum radiation depends on ka. The asymptotic value of
the radiation efficiency is 1/cos θ (see Equation 6.37). This means that the
actual value of the radiation efficiency can be converted into an equivalent
angle of radiation:

 1 
θ r = arccos   (6.60)
σθ 
 Sound radiation from plates  185

a θr

a θ

Figure 6.23 Angle of incidence θ and main direction of sound radiation θr for forced
vibrations in a finite plate with characteristic dimension a.

10

ka ∞

32

5
10 lg σθ (dB)

θr

–5
0° 15° 30° 45° 60° 75° 90°
Angle of incidence (θ)

Figure 6.24 Equivalent angle of radiation from a square panel is 70° when the angle of
incidence is 85° and ka = 32.
186  Sound Insulation in Buildings

Figure 6.24 shows the principle of this conversion for the angle of incidence
θ = 85° and ka = 32. The equivalent angle of radiation is found to be around
θr = 70°.
The simulations shown in Figure 6.25 are four examples of radiation
from resonant vibrations in a plate that is excited by an incident plane
wave. The normal modes radiate sound equally to both sides of a plate, and
the radiation is symmetric around the normal to the plate. In Figure 6.25a,
the mode with modal number 1 is excited, and the radiation is nearly omni-
directional. With higher modal numbers, the radiation is more directive,
but with a clear refraction effect (Figure 6.25b and c). In these simulations,
the resonant radiation is strong and the radiation from the forced bending
wave is not visible.

(a) (b)

(c) (d)

Figure 6.25 Computer simulation of sound radiation from a plate at the coincidence

frequency. The excitation is a plane wave incident at angle 60°. (a) ka = π, (b)
ka = 2π, (c) ka = 4π, (d) ka = 14π. (Reproduced from Heckl, M., J. Sound Vib.,
77, 165–189, 1981. With permission from Elsevier.)
 Sound radiation from plates  187

REFERENCES

K. Gösele (1953). Schallabstrahlung von Platten, die zu Biegeschwingungen angeregt

sind (Sound radiation from plates that are excited to bending waves, in
German). Acustica 3, 243–248.
M. Heckl (1981). The tenth Sir Richard Fairey memorial lecture: Sound transmission
in buildings. Journal of Sound and Vibration 77, 165–189.
M. Heckl, M. Heckl (1976). Darstellung von abstrahlvorgängen im film. Fortschritte
der Akustik – DAGA 1976, VDI-Verlag, Düsseldorf, 165–168.
J. Kristensen, J.H. Rindel (1989). Bygningsakustik – Teori og praksis (Building acous-
tics – Theory and practice, in Danish). SBI Anvisning 166. Danish Building
Research Institute, Hørsholm, Denmark.
G. Maidanik (1962). Response of ribbed panels to reverberant acoustic fields.
Journal of the Acoustical Society of America 34, 809–826. Erratum (1975).
Journal of the Acoustical Society of America 57, 1552.
J.W.S. Rayleigh (1896). The Theory of Sound. Vol. 2. Reprint of 2nd edition by
Dover Publications, New York, 1945.
J.H. Rindel (1975). Transmission of traffic noise through windows. Influence of inci-
dent angle on sound insulation in theory and experiment. Report No. 9, The
Acoustics Laboratory, Technical University of Denmark, Lyngby.
J.H. Rindel (1993). Modelling the angle-dependent pressure reflection factor. Applied
Acoustics 38, 223–234.
H. Sato (1973). On the mechanism of outdoor noise transmission through walls
and windows. Journal of the Acoustical Society of Japan 29, 509–516. (In
Japanese).
E.C. Sewell (1970). Transmission of reverberant sound through a single-leaf parti-
tion surrounded by an infinite rigid baffle. Journal of Sound and Vibration 12,
21–32.
H. Stenzel (1952). Die akustische Strahlung der rechteckiger kolbenmembran.
Acustica 2, 263–281.
C.E. Wallace (1972). Radiation resistance of a rectangular panel. Journal of the
Acoustical Society of America 51, 946–952.
Chapter 7

Statistical energy analysis, SEA

Sound fields in rooms and mechanical vibrations in plates are examples of

systems with a high number of modes, the so-called resonant systems. In
case of combinations of two or more such resonant systems, it can often be
advantageous to take a statistical approach in order to analyse the propaga-
tion and distribution of the acoustic energy in the coupled resonant systems.
The method of statistical energy analysis (SEA) was originally developed
by Lyon and Maidanik (1962) and applied for sound insulation by Crocker
and Price (1969) and others. The method has been particularly successful
for the investigation of sound propagation through complicated solid struc-
tures like ships and buildings.
Even if SEA is a powerful tool, there are some general underlying assump-
tions that put restrictions and drawbacks on the methodology. Some of the
underlying assumptions are as follows:

• Only resonant fields are of importance.

• The modal density should be high and smooth.
• There should be a more or less diffuse field.
• The coupling between different subsystems should be weak.

Some of these assumptions can be violated if one is careful, but it is impor-

tant to be aware of these underlying facts. One consequence of the underly-
ing assumptions is that structures that are periodic in nature – as in many
lightweight and wooden building structures – are not well suited for SEA
(these structures violate the first three points). However, for heavy and
homogeneous traditional building structures, the underlying assumptions
are well met.

7.1 ENERGY BALANCE IN A RESONANT SYSTEM

The total mechanical energy in a vibration system is the sum of the kinetic
and the potential energies. In a plate with normal modes, the kinetic and
the potential energies are on average equal. An explanation of this is that

189
190  Sound Insulation in Buildings

at resonance the kinetic and the potential energies are per definition equal,
and the average response magnitude is mainly determined by the response
at the resonances. Since it is common to describe mechanical vibrations by
the velocity of the surface, i.e. the kinetic part of the energy, we calculate
the total energy E as twice the kinetic energy, as in:

E = mS v 2 (7.1)

where m is the mass per unit area of the plate, S is the area of the plate and
v 2 is the mean square of the velocity averaged over the area of the plate.
For a sound field in a room, the energy is commonly described by the
sound pressure, i.e. the potential part of the energy. The total energy E in
the room can be expressed as twice the potential energy, as follows:

p 2
E= V (7.2)
ρ c2

where V is the volume of the room, p is the mean square of the sound
2

pressure averaged over the volume of the room, ρ is the density of air and c
is the speed of sound in air.
The loss factor η of a resonant system is defined as the dissipated energy
in a time period of vibration divided by 2π and the total mechanical energy:
Pd T0 P
η= = d (7.3)
2π E ω E

where Pd is the dissipated power (energy per time unit), T0 is the time period
of the vibration, E is the current total energy in the system and ω = 2πf = 2π/
T0 is the angular frequency.
In a room, the energy losses are usually described by the equivalent
absorption area A or by the reverberation time T. If the volume is V and the
sound field in the room can be assumed to be approximately diffuse, the
energy density in the stationary sound field is E / V = 4Pin / cA, where Pin is
the sound power emitted in the room. Energy balance at stationary condi-
tions means that Pin = Pd and, thus, the loss factor of a room is:

Pd cA cA
η= ⋅ = (7.4)
ω 4Pd V 4ω V
The reverberation time is given by Sabine’s equation T = 55V /cA, and the
loss factor of a room can also be expressed as:

55 2.2
η= ≅ (7.5)
4ω T fT
 Statistical energy analysis, SEA  191

The latter expression can be shown to be valid for the vibrations in a

solid structure as well, and it is a well-established measuring method for
the loss factor. The reverberation time is derived from the slope of the mea-
sured decay curve after the source is stopped. The frequency f is taken as
the centre frequency of the frequency band used during the measurement.

7.2 COUPLING BETWEEN RESONANT SYSTEMS

Two resonant systems are said to be coupled if energy can be transferred from
the normal modes in one system to the normal modes in other system. This
requires two conditions to be fulfilled, a physical one and an acoustical one:

• The resonant systems must have a physical connection in order that

the energy can be exchanged, e.g. a junction between two plates or a
common surface between a plate and a room.
• The resonant systems must have normal modes with natural frequen-
cies sufficiently close to each other in order that the energy can be
exchanged. This may be fulfilled if each system has at least one natu-
ral frequency within the frequency band under consideration, e.g. a
one-third octave band.

The coupling loss factor is introduced to describe the transfer of energy

from one system to another. The definition is an analogue to that of the loss
factor; the coupling loss factor η12 from system 1 to system 2 is:

P12′
η12 = (7.6)
ω E1

where P12′ is the power (energy per time unit) transferred from system 1 to
system 2, E1 is the current total energy in system 1 and ω is the angular
frequency.
Example. The sound power radiated from resonant modes in a plate
(system 1) into a room (system 2) can be derived. The radiated sound power
from one side of the plate with velocity of vibration v 1, area S1 and radia-
tion efficiency σ 1,res is (Equation 6.24):

P12′ = v12 S1ρ cσ 1 (7.7)

Insertion of this and the total energy (Equation 7.1) into the definition
yields:

ρc
η12 = σ 1,res (7.8)
ω m1
192  Sound Insulation in Buildings

This shows that the energy transfer to the room is strongest at low frequen-
cies and for low mass per unit area if the radiation efficiency is constant.
Two coupled resonant systems are considered (Figure 7.1). Each system
may have a source that supplies the system with mechanical or acous-
tic energy, P1i and P2i, respectively. The dissipated energy is P1d and P2d ,
respectively.
The net energy transfer per unit time from system 1 to system 2 can be
written as:

P12 = P12′ − P21′ = ωη12E1 − ωη21E2 (7.9)

However, the consideration of the energy balance between the two systems
should be made on the basis of the normal modes that carry the energy.
It is the energy per mode in each system that is important for the energy
balance, as it is the modes that are the degree of freedom of the system. In
system 1, the average energy per mode Em1 is:

E1 E ∆f
Em1 = = 1⋅ (7.10)
∆N1 ∆f ∆N1

where Δf is the frequency band under consideration and ΔN1 is the number
of modes in system 1 within this frequency band. The ratio ΔN1/Δf is called
the modal density of system 1, i.e. the number of modes per Hz around the
centre frequency f.
The energy balance equation for the two systems can now be expressed
in terms of the average energy per mode:

P12 = ω∆N1η12 Em1 − ω∆N2η21Em 2 (7.11)

Room Plate

P1, in P2, in

P΄12

P΄21
E1 E2

P1, d P2, d

Figure 7.1 Block diagram to show the energy transfer between two coupled resonant
systems.
 Statistical energy analysis, SEA  193

The idea is that each normal mode in a resonant system can be considered a
degree of freedom, and it is a general physical principle that energy transfer
between coupled systems will always be in the direction from the system
with higher energy per degree of freedom to the system with lower energy
per degree of freedom. This means that in the case of energy balance,
Em1 = Em2 , the net power flow P12 is zero and thus Equation 7.11 yields,

∆N1
η21 = η12 (7.12)
∆N2

This is the very important reciprocity relation for coupled resonant sys-
tems. It says that it is possible to derive the coupling loss factor in one
direction if the coupling loss factor is known in the opposite direction and
if the modal densities of the two systems are also known. As an example,
it is well known how a plate radiates sound into a room, and with the
reciprocity relation, it will then be possible to estimate how much energy is
transferred into a plate from the sound field in a room.
The energy balance equation can be expressed by only one coupling loss
factor using Equation 7.12:

P12 = ω∆N1η12 ( Em1 − Em 2 ) (7.13)

This is sometimes called the fundamental equation of SEA. It clearly shows

that the difference in average modal energy (Em1 − Em2) determines the direc-
tion and amount of energy flow between the two systems. Equation 7.13
also shows a certain analogy to thermodynamics; the average energy per
mode has the same role in acoustics as the temperature has in thermody-
namics, and the heat is transferred from a body with a higher temperature
to a body with a lower temperature.
It is emphasised that two statistical assumptions have been made:

• The energy in each system is assumed to be equally distributed on the

number of normal modes ΔNi available within the frequency band Δf
with the centre frequency f.
• All pairs of normal modes in the two systems have the same cou-
pling loss factor within the frequency band Δf with the centre fre-
quency f.

These assumptions are the reason for calling the method statistical. None
of the assumptions are strictly correct, but on average they may be the best
available estimate. Later, it will be shown that the results from the SEA
method may also be used without the statistical part in case of very low
modal density in one of the systems. The modal densities of some typical
systems are collected in Table 7.1.
194  Sound Insulation in Buildings

Table 7.1  Statistical modal density in different one-, two- and three-dimensional
systems
System Modal density, ΔNi/Δf
Room, 3D modes only 4π
Volume V Vf 2
c3
More accurate, see Equation 4.13
Narrow cavity, 2D modes only 2π
Area S = lx ly Sf
c2
Tube, both ends either closed or open, 1D modes only 2
Length lx lx
c
Thin plate, bending waves π
Area S, critical frequency fc Sfc
c2
More accurate, see Equation 6.9
Thick plate, shear waves 2π
Area S, speed of shear waves cs Sf
cs2
More accurate, see Equation 6.10
Beam, longitudinal waves 2
Length lx, speed of sound cm lx
cm

7.3 RESONANT TRANSMISSION THROUGH
A SINGLE WALL

A partition wall between two rooms can be considered a resonant sys-

tem that is coupled to the two resonant systems representing the rooms.
A sketch is shown in Figure 7.2 together with a SEA flow diagram. System 1
is the source room, system 2 is the wall and system 3 is the receiving room.
The forced sound transmission through the wall cannot be derived by SEA
since the normal modes of the wall are not involved in this transmission.

Room Wall Room

P1,in

P΄12 P΄23

P΄21 P΄32
E1 E2 E3

P1,d P2,d P3,d

Figure 7.2 Block diagram to show the energy transfer between three coupled resonant
systems.
 Statistical energy analysis, SEA  195

So, only the resonant transmission is considered in the following. Later, it

will be discussed how to combine the two different kinds of transmission.
With the usual assumption of diffuse sound fields in the two rooms, the
transmission loss of the wall can be written in terms of the total energies in
the rooms using Equation 7.2 in combination with Equation 5.5:

p12 S2 EV S
R = 10 lg 2 = 10 lg 1 3 2 (7.14)
p3 A3 E3V1A3

where S 2 is the area of the wall and A 3 is the absorption area of the receiv-
ing room.
The SEA method implies to set up the energy balance equations for each
of the systems, i.e. three in this case:

P1,in + P21′ = P1, d + P12′ (7.15)

P12′ + P32′ = P2, d + P21′ + P23′ (7.16)

P23′ = P3, d + P32′ (7.17)

System 1 is excited by a stationary sound source, and the equation for that
system will only be used if it is relevant to analyse the results as a function
of the power of the source. In the present situation, this is not the case; it is
sufficient to find the ratio between the energy of systems 1 and 3, which is
independent of the power of the source. So, it will be sufficient to continue
with the equations from systems 2 and 3, i.e. the systems without a source.
The loss factors (Equation 7.3) and the coupling loss factors (Equation 7.6)
are inserted in Equations 7.16 and 7.17:

η12E1 + η32E3 = η2E2 + η21E2 + η23E2 = η2,tot E2 , (7.18)

η23E2 = η3E3 + η32E3 = η3,tot E3 . (7.19)

Here, the total loss factors have been introduced in systems 2 and 3. The
loss factor of the wall is combined from internal losses in the solid material
and boundary losses (both contained in η2) and radiation losses to either
side (these are the coupling loss factors from the wall to each of the rooms,
η21 and η23). The total loss factor is the one that can be measured experi-
mentally since it is not possible to avoid the radiation losses. The total loss
factor of the receiving room combines the absorption of sound from the
wall (η32) and the absorption from all other surfaces in the room (η3,d).
Again, the total loss factor is the one that can be measured experimentally.
196  Sound Insulation in Buildings

It is possible to eliminate E 2 and derive the ratio E1/E 3:

E1 η2,totη3,tot η32
= − (7.20)
E3 η12η23 η12

The last term can be neglected because it must be close to unity and the
total ratio is expected to be >>1. In the first term, it is an advantage to
change the coupling loss factor η12 with η21, which is done by applying the
reciprocity relation (Equation 7.12):

E1 ∆N1 η2,totη3,tot (7.21)

≅ ⋅ .
E3 ∆N2 η21η23

From Equations 7.4 and 7.8 and Table 7.1, we have

cA3 ρcσ 2,res ∆N1 4π

η3,tot = and η21 = η23 = and ≅ 3 V1f 2
4ωV3 ω m2 ∆f c

leading to

2
E1 4π ∆f cA3  ω m2 
≅ 3 V1f 2 ⋅ ⋅ η2,tot ⋅ ⋅  (7.22)
E3 c ∆N2 4ωV3  ρcσ 2,res 

and the transmission loss for resonant transmission (Equation 7.14):

2
E1V3S2  ω m2   c 2σ 2,res
2
∆N2 
Rr = 10 lg = 10 lg   − 10 lg  ⋅  (7.23)
E3V1A3  2ρc   2η2,tot S2 f ∆f 

In this result, the first term is recognised as the mass law at normal inci-
dence R 0. In this form, Equation 7.23 is a generally valid expression for
thin or thick walls, having either high or low modal density. The modal
density can be inserted according to the type of wall (Table 7.1) (bending
waves or shear waves). If a thin wall is assumed (only bending waves and
high modal density), the resonant transmission loss is approximately:

 πσ 2 f 
Rr = R0 − 10 lg  2,res c ,2  (7.24)
 2η2,tot f 

It should be noted that the radiation efficiency is squared, i.e. it appears

twice. If we trace back in the derivation, it is seen that the radiation effi-
ciency comes from the two coupling loss factors η21 and η23. So, the first one
 Statistical energy analysis, SEA  197

represents the excitation of the plate from the source room (reciprocity),
while the second one represents the radiation into the receiving room.

7.4 FORCED AND RESONANT TRANSMISSION

THROUGH A SINGLE WALL

The forced part of the transmission cannot be handled with SEA. However,
it will be more thoroughly studied in Chapter 8. The forced transmission
can be added as an independent transmission path connecting source room
and receiving room (Figure 7.3). This is equivalent of adding a direct field
to a diffuse field in room acoustics.
The total transmission through the wall is found by adding the sound
powers of forced and resonant transmission:

Ptot = Pfor + Pres = ρ cS2 v 2,for

2
σ 2,for + ρ cS2 v 2,res
2
σ 2,res (7.25)

Here, the velocity of forced vibrations is v 2,for and the radiation efficiency
of the forced vibrations is σ2, for (Equation 6.40). This equation is a good
basis for a sound transmission model; forced and resonant transmission is
combined, and for both parts, the transmission process is divided into exci-
tation (generating the vibrations in the panel) and radiation.
The transmission loss including both forced and resonant transmission
is then:

R = 10 lg
Pin
Ptot
= 10 lg
Pin
Pfor + Pres
( )
= −10 lg 10−0.1Rf + 10−0.1Rr (7.26)

A simple approximation for the forced transmission loss Rf was derived in

Chapter 5 (see Equation 5.37).

Room Wall Room

P13,for
P1,in

P12,res P23,res
E1 E2 E3

P1,d P2,d P3,d

Figure 7.3 Block diagram to show the combination of forced and resonant transmission
through a wall.
198  Sound Insulation in Buildings

7.5 ACOUSTIC EXCITATION OF RESONANT

VIBRATIONS IN A PLATE

7.5.1 Excitation from one side

The acoustic excitation of a plate by sound incident from one side can be
derived from the previous example with one room on either side of the plate
or it can be simplified to one room and one plate only. If the room is system
1, the plate is system 2, and if the room is excited by a stationary source, we
only need to consider the energy balance equation for system 2:

P12′ = P2, d + P21′ (7.27)

In terms of loss factor and coupling loss factors, this leads to:

η12E1 = η2E2 + η21E2 = η2,tot E2 (7.28)

Here, it should be noted that in this particular case, the loss factor η2
includes the radiation loss from the back of the plate, since η21 represents
the radiation loss from the exposed side of the plate back to the source
room. Compare Equation 7.28 with the three-system model in the previous
section (Equation 7.18). In any case, it is more practical to introduce the
total loss factor of the plate. Then, the energy ratio is:

E2 ∆N2 η21 (7.29)

≅ ⋅
E1 ∆N1 η2,tot

If the plate is excited from one side as a wall that is part of the surfaces of a
room, the coupling loss factor is as in Equation 7.8 except that the number-
ing is different here. Together with the modal density for room 1, this leads
to the result:

2
v 2,res c2 σ 2,res ∆N2
≅ ⋅ ⋅ (7.30)
(ω m2 ) 2f η2,tot S2 ∆f
2
p12

This is a general result for the acoustic excitation of the resonant modes
in a plate from one side, as a function of the modal density of the plate.
Assuming a thin plate with bending waves, this simplifies to:

2
v 2,res π fc ,2 σ 2,res (7.31)
≅ 2 ⋅
p12 (ω m2 ) 2f η2,tot
 Statistical energy analysis, SEA  199

7.5.2 Freely hung plate in a room

If instead the plate is freely hung in the room and is excited from both sides,
the coupling loss factor must include the radiation to both sides, i.e. a factor
of 2 must be added and instead of Equation 7.8, we get:
2ρ cσ 2,res
η21 = (7.32)
ω m2

This leads to a factor of 2 in the results corresponding to Equations 7.30

and 7.31. However, this does not mean that the energy of the resonant
modes in the plate is twice as high as in the case of excitation from one side.
The radiation efficiency is not the same for a freely hung plate; it will be
lower than for the plate in a wall because the free edges of the plate allow
some degree of acoustic cancellation around the edges. At low frequencies,
it will be considerably lower than for a plate surrounded by a baffle, but
how much lower shall not be discussed here.
Assuming a freely hung thin plate with bending waves, we get instead of
Equation 7.31:

2
v 2,res π fc ,2σ 2,res (7.33)
≅ 2 ⋅
p12 (ω m2 ) f η2,tot

7.6 RESONANT TRANSMISSION THROUGH

A SLAB WITH EXTENDED AREA

A more general example of sound transmission is shown in Figure 7.4. The

source room (1) has the volume V1 and the floor area S1, while the receiving

Room (1)

V1

Slab (2) S2 S1

S3

V3

Room (3)

Figure 7.4 Case of a large slab separating a source room and a receiving room. Other
rooms are not shown. All walls are assumed to be lightweight and the vibra-
tion energy equally distributed in the slab.
200  Sound Insulation in Buildings

room (2) has the volume V3 and the floor area S 3. The sound is transmitted
through a large slab with area S 2 that is greater than both S1 and S 3. We
also assume that the walls are lightweight and, thus, the junction attenu-
ation can be neglected. So, we have three coupled resonant systems as in
Figure 7.3, but now the area of the slab is different from the excitation area
S1 and the radiation area S 3.
The steady-state energies in the three systems are:

p12
E1 = V1 (7.34)
ρc2

E2 = m2S2 v 22 (7.35)

p 32
E3 = V3 (7.36)
ρc 2

The coupling loss factor from plate to receiving room yields:

ρ cS3σ
η23 = (7.37)
ω m2S2

Resonant transmission, only, is assumed and σ is the radiation efficiency.

The modal densities in source room and slab (with bending waves) are:

∆N1 4π
≅ 3 V1f 2 (7.38)
∆f c

∆N2 π
≅ 2 S2 fc (7.39)
∆f c

where fc is the critical frequency of the slab. The reciprocity relation

(Equation 7.12) leads to the coupling loss factor from source room to plate:

∆N2 π c 3 S2 fc ρcS1σ ρc 2 fc S1σ

η12 = η21 = ⋅ = ⋅ (7.40)
∆N1 4π c V1f ω m2 S2 4V1f 2 ω m2
2 2

In the same way as in Equation 7.22, we find the ratio of energy in source
and receiver rooms:

2
E1 2η2,tot f V1A3S2  ω m2  (7.41)
≅ ⋅ ⋅
E3 πσ 2 fc V3S1S3  2ρ c 
 Statistical energy analysis, SEA  201

The transmission loss for resonant transmission yields:

2
EV S  ω m2   πσ 2 fc   S2 
Rr = 10 lg 1 3 1 = 10 lg  − 10 lg  + 10 lg  S3 
E3V1A3  2ρ c   2η2,tot f

 πσ 2 fc   S2 
Rr = R0 − 10 lg   + 10 lg  S3  (7.42)
 2η2,tot f

Comparing with Equation 7.24, we see that the extended area of the slab
gives rise to an increased transmission loss. If the area of the slab is 10 times
the area of the ceiling radiating into the receiving room, the transmission
loss may be up to 10 dB higher than in the normal case where S 2 = S 3.

REFERENCES

M.J. Crocker, A.J. Price (1969). Sound transmission using statistical energy analysis.
Journal of Sound and Vibration 9, 469–486.
R.H. Lyon, G. Maidanik (1962). Power flow between linearly coupled oscillators.
Journal of Acoustical Society of America 34, 623–639.
Chapter 8

Airborne sound transmission

through single constructions

The sound insulation of single constructions is rather complicated. One

reason for this is that two transmission mechanisms, the forced and the
resonant, are in play at the same time, as mentioned in Section 7.4. The
resonant vibrations using statistical energy analysis (SEA) were discussed
in Chapter 7. Thus, the forced vibrations remain and will be discussed in
Section 8.1. Another reason for sound insulation being rather complicated
is the large variation in building constructions covering a range from thin
plates of steel, wood or glass to thick structures of brick or concrete. To this
picture, orthotropic and sandwich panels with complicated structural waves
are added. The aim of this chapter is to describe the physical principles of
sound transmission and derive a general formula for the transmission loss
that covers the range from thin to thick panels and includes effects due to
the finite area of the panels. The influence of the angle of incidence is also
analysed with special address to the enigma of grazing incidence; according
to classical theory, the transmission loss approaches 0 dB (total transmis-
sion), but the incident sound power also approaches zero. Obviously, the
classical theory is insufficient in the case of grazing incidence of sound.

8.1 SOUND INSULATION OF HOMOGENEOUS,

ISOTROPIC PLATES AT RANDOM INCIDENCE

The physical behaviour of a single homogeneous isotropic plate is now to

be studied. The vibrations generated in a plate by excitation of an incident
sound field can, as already discussed, be divided into forced and resonant
vibrations. In the present chapter, the forced transmission is analysed in
some detail and combined with the results from SEA in order to establish
more general models for the total transmission.

8.1.1 Total transmitted sound power

In Figure 8.1, the principle of sound transmission through a single panel
is shown. The excitation of the panel gives rise to two kinds of vibrations:
forced and resonant. So, at the same time, the panel exhibits forced and

203
204  Sound Insulation in Buildings

(a) (b)

Pres

vres
vfor

Pfor

Figure 8.1 Principle of sound transmission through a panel. Excitation and radiation are
different for forced transmission (a) and resonant transmission (b).

resonant waves with different velocities of vibration and different radiation

efficiencies. When S is the area of the panel, the total transmitted power
can be written as:

Ptot = Pfor + Pres = ρ cS v for

2
σ for + ρ cS v res
2
σ res (8.1)

For each transmission part, the physical process can be divided into excita-
tion and radiation. The excitation gives rise to the vibrations of the panel,
and the velocity is controlled by the impedance (mass and stiffness) of the
panel. The radiation is described by the radiation efficiency, which is differ-
ent for forced and resonant radiation (see Chapter 6).

8.1.2 The mass law

The mass law expresses the airborne sound insulation of a plate in depen-
dency of its mass per unit area. A very large uniform/homogeneous plate
of mass m per unit area is examined as the resonant normal modes are dis-
regarded. The plate is excited by a plane sound wave at normal incidence,
so all parts of the plate will move in phase and the bending stiffness can be
disregarded. Under these idealised conditions, the transmission loss is R 0,
where only m and the frequency f are variable. Using 1 kg/m 2 and 100 Hz
as reference values, we get from Equation 5.27:

  ωm 2
R0 = 10 lg  1 +    (dB) (8.2)
  2ρ c  

 m   f 
R0 ≈ 20 lg  + 20 lg  − 2 dB (8.3)
 1kg/m 2   100 Hz 
 Airborne sound transmission through single constructions  205

This simple relation between transmission loss, mass per unit area and
­frequency is called the mass law, and is the basis of all the following
­considerations on airborne sound insulation. The mass law is characterised
by a 6 dB increase every time m or f is doubled.

8.1.3 Wall impedance
The resistance of a wall or a plate against a sound wave is called the wall
impedance Zw or the transmission impedance. Zw is defined as the ratio
between pressure difference ∆p across the plate and the velocity v of the
vibrations of the plate (Equation 5.23):

∆p
Zw = (8.4)
v

When ∆p and v are out of phase, the wall impedance is a complex number,
which is normally the case. The wall impedance is determined from the
bending wave equation (see Equation 1.24). Assuming a harmonic time
dependency (with angular frequency ω = 2π f) of the pressure exciting the
plate, the bending wave equation is:

2
 ∂2 ∂2 
B  2 + 2  v − mω 2 v = jω∆p (8.5)
 ∂x ∂y 

where v is the vibration velocity, B is the bending stiffness, m is the mass

per unit area and Δp is the applied force per unit area. Wave motion is in
both x- and y-directions. Assuming an applied pressure from the surround-
ing air in the form:

p = p̂e− jkr sin θ (8.6)

where k = ω/c and θ is the angle of incidence and r is an arbitrary distance
parameter, the bending wave equation will have the form:

2
 ∂2 ∂2 
B  2 + 2  vˆ e− jkr sin θ − mω 2 vˆ e− jkr sin θ = jω∆pˆ e− jkr sin θ (8.7)
 ∂x ∂y 

and the wall impedance can be written as:

Zw =
∆pˆ
vˆ
=
1
jω ( )
B (k sin θ ) − mω 2 = ZB + Zm
4
(8.8)
206  Sound Insulation in Buildings

where the mass impedance is:

Zm = jω m (8.9)

and the bending wave impedance is:

4
1 ω  c4
ZB = B  sin θ  = −jω m ⋅ B4 ⋅ sin4 θ (8.10)
jω  c  c

This would be sufficient for thin panels. However, in order to derive results
applicable also to thick panels, it is necessary to consider shear waves in
addition to bending waves. The shear wave impedance can be found from
a wave equation for shear waves in a way similar to that for bending waves
and this leads to:
2
1 ω  c
ZS = Gh  sin θ  = −jω m ⋅ S ⋅ sin2 θ (8.11)
jω c  c

where G is the shear modulus, h is the thickness of the panel and c s is the
speed of shear waves in the panel.
Like a simple mechanical system, the wall impedance contains three
elements representing mass, stiffness and losses. The wall impedance
can be described by an electrical analogy, as shown in Figure 8.2. The
mass is represented by an inductor and the losses by a resistance r ≈ ηωm
(Equation 2.28). The inductor is L = m, and the capacitors are:

c4 c2
C1 = , C2 = 2
ω B
4
ω Gh

Bending Shear

C1 C2

Figure 8.2 Electrical analogy of the wall impedance of a thick panel.

 Airborne sound transmission through single constructions  207

Bending stiffness and shear stiffness are considered to work in parallel, so

the weaker one is always dominating. The wall impedance yields:

1 Z Z
(C1 + C2 ) + r = Zm + B S + r
−1
Zw = jωL +
jω ZB + ZS

 (c /c)2 (c /c)4 
= jω m 1 − S 2 B 4
⋅ sin4 θ  + ηω m (8.12)
 (cS /c) + (cB /c) 
  c 2  c 4  
−1

= jω m 1 − sin θ ⋅   +     + ηω m
4

  cS   cB   


In case of a material with c S ≫ c like glass or steel, the wall impedance can
be approximated by the more simple formula 5.35.
The imaginary part of the wall impedance (Equation 8.12) relative to the
mass impedance is displayed as a function of the frequency in Figure 8.3a
for grazing incidence and in Figure 8.3b for the angle of incidence 45°.
The wall impedance is zero when coincidence occurs. We notice that
coincidence at grazing incidence is not exactly at the critical frequency,
but slightly above fc for c S /c = 2. The minimum of the impedance is shifted
to very high frequencies for c S /c→ 1. If c S /c < 1, which can be the case in
sandwich panels, there is no possibility of coincidence, simply because the
speed of the transverse waves is subsonic. At the angle of incidence 45°, the
coincidence dip is shifted to higher frequencies by one octave for c S /c ≫ 1.
The wall impedance (Equation 8.12) at grazing incidence (θ = 90°) can
be used to derive the effective speed of free transverse vibrations, as shown
by Kurtze and Watters (1959). A plane sound wave with variable speed c
is assumed to propagate along the surface of the panel and excite a forced
transverse motion. The ratio of the pressure to the transverse velocity is
the wall impedance, which is a function of the speed of the exciting wave.
The speed of free transverse waves in the panel cB,eff is found by the mini-
mum energy condition, i.e. the impedance must be minimum that means
Im{Zw} = 0. This leads to an equation in c of the fourth degree:

cS2c 4 + cB4 c 2 − cB4 cS2 = 0 (8.13)

The speed cB,eff equals the speed c of the exciting wave in this equation, and
according to Rindel (1994), the solution is:

4 −1/3
cB2 1 1 c   1 1 (8.14)
cB,eff = − + 1+ 4 S  ≅  3 + 3
cS 2 2  cB   cB cS 
208  Sound Insulation in Buildings

(a) 2

θ = 90°

1.5
|Zw / Zm|2

0.5

0
0.1 0.2 0.5 1 2 5 10
f/fc

10 2 1 0.5

(b) 2

θ = 45°

1.5
|Zw / Zm|2

0.5

0
0.1 0.2 0.5 1 2 5 10
f /fc

10 2 1 0.5

Figure 8.3 Wall impedance relative to the mass impedance as function of frequency for
different values of cS /c. (a) At the angle of incidence 90° and (b) at the angle
of ­incidence 45°. Solid line: cS /c = 10, dashed line: cS /c = 2, dotted line: cS /c = 1,
dashed–dotted line: cS /c = 0.5.

where the approximation is accurate within ±1 % (see also the dispersion
curve in Figure 1.4).
At low frequencies, the impedance of the plate is dominated by the reso-
nance at the first normal mode. If the size of the panel is not extremely
small, the cross-over frequency from bending waves to shear waves is
fs ≫ f11 and bending waves can be assumed at the first normal mode. Thus,
 Airborne sound transmission through single constructions  209

the natural frequency of the first normal mode f 11 and the critical frequency
fc are closely related, as seen from Equation 6.7 with (m, n) = (1, 1):

c2  1 2  1 2
f11 =    +    (8.15)
4 fc   lx   ly  

where lx and ly are the plate dimensions. Therefore, a high fc means a low f11
and vice versa. Some examples are given in Table 8.1. It is noted that the thick
plates (concrete and brick) have a much lower modal density than thin plates
(porous concrete, glass or gypsum), even if the latter have smaller dimensions.
At frequencies below and around the natural frequency of the first struc-
tural mode, f11, the stiffness comes not only from bending the plate but
also from the connections to the surroundings at the boundaries. As an
approximation, we assume the system to behave as a simple mechanical
system with the natural frequency f 11.
 k 
Zw = jω m − d  + r
 ω
(8.16)
  2 
f
= jω m 1 −  11   + ηω m
  f  
 
Combining with the above result (Equation 8.12) and using Equation 5.32 to
replace cB with fc, we get the wall impedance at the angle of sound ­incidence θ:

  2    2  2  
−1
f  c f 
Zw = jω m ⋅ 1 −    ⋅ 1 − sin θ ⋅   +     + ηω m (8.17)
11 4 c
  f     c S   f   
     

In analogy with Equation 6.38, we can take into account the finite size of
the panel by replacing the term sin 2 θ with the radiation efficiency:
−2
sin2 θ = 1 − cos2 θ → 1 − σ for,θ (8.18)

Table 8.1  C
 alculated critical frequency, first natural frequency and number of
modes for examples of thin and thick plates
ΔN in One-Third
Octave
Plate
Dimensions (m) fc (Hz) f11 (Hz) 100 Hz 500 Hz
4 mm glass 1.2 × 1.3 3250 12 3 15
13 mm gypsum board 2.5 × 1.2 2615 10 5 24
75 mm porous concrete 2.5 × 4.0 707 9 4 22
180 mm concrete 2.5 × 4.0 106 62 0.7 3
250 mm brick 2.5 × 4.0 72 91 0.4 2
210  Sound Insulation in Buildings

where σfor,θ is the radiation efficiency of forced waves (Equation 6.39). This
can be used for relatively high frequencies where σfor,θ > 1, but for low
­frequencies (small ka) coincidence is not possible, and sin 2 θ → 0.
For a diffuse sound field with random incidence, a good approximation
valid for ka > 3.84 ≈ 4 is obtained by the corresponding value for grazing
incidence, θ = 90° in Equation 6.39:

−2 3.84
sin2 θ → 1 − σ for,90 ≅ 1− (8.19)
ka

For ka < ca. 4, the plate is acoustically small and coincidence is not
possible. The reason is that the dimensions of the plate are too small
compared to the wavelength of the incident sound. So, instead of sound
propagation along the plate, diffraction effects mean that the transmit-
ted sound is refracted (see Figure 6.25; also Figure 6.10 for a graphical
presentation of the radiation efficiency as function of ka and the angle
of incidence).

8.1.4 The loss factor

There are three different contributions to the energy losses of vibrations in
a plate: internal losses, boundary losses and radiation losses.
The internal loss is part of the energy that is transferred to heat and this
is quite small for most materials and approximately frequency independent
(see examples in Table 1.4). The loss factor due to boundary losses can be
calculated as shown by (Craik 1981):

U λBα Ucα
= 2 ( f ⋅ fc )
−1/ 2
ηborder = (8.20)
π 2S π S

where α is the average structural absorption coefficient of vibrations along

the boundary, which has a total length of U.
Typical values of the structural absorption coefficient are within 0 <
α ≤ 0.3, and it depends both on the type of connection to the surrounding
structures and on the thickness of the plate relative to the thickness (or
mass) of the connecting structures. In Figure 8.4, theoretical values of the
structural absorption coefficient for junctions where all plates are made of
the same material but with different thickness are shown.
The radiation loss factor was dealt with in Chapter 7 and it is closely
related to the resonant radiation efficiency (Equation 7.8). Since the reso-
nant modes of the plate radiate to both sides, the radiation loss factor is
twice the coupling loss factor from plate to room:

2ρ c
ηrad = ⋅ σ res (8.21)
ωm
 Airborne sound transmission through single constructions  211

1.000

cross

Absorption coefficient, α tee C

0.100

0.010

0.001
0.1 1 10
Ratio of plate thickness h2/h1

Figure 8.4 Absorption at the boundary where all plates are connected rigidly and made
of the same material but with different thickness. The absorption relates
to the plate with thickness h1, and h2 is the thickness of the crossing plate.
(Adapted from Craik, R.J.M., Appl. Acoust., 14, 347–359, 1981.)

Below the critical frequency, the radiation loss factor is small, but it can be
quite significant around the critical frequency. The total loss factor is:

Ucα 2ρ c
2 (
f ⋅ fc )
−1/ 2
ηtot = ηint + ηborder + ηrad = ηint + + ⋅ σ res (8.22)
π S ωm
When measuring the loss factor of a plate, it is the total loss factor that is
measured, as it is not possible to eliminate the radiation losses.
Examples of measured loss factors are shown in Figure 8.5. The wall
elements were 2.54 m high and 1.80 m wide and were connected in a
T-junction to a 180 mm clinker concrete slab. The loss factor generally
increases towards the lower frequencies. One of the walls was measured
with an elastic connection to the slab, and in this case the loss factor was
clearly higher than with the rigid connection, which may be due to viscous
losses in the elastic connection.

8.1.5 Forced transmission
Let the sound pressure of the incident, reflected and transmitted sound be
denoted as pi, pr and pt, respectively. Then, ∆p = pi + pr − pt ≅ 2pi and the
squared sound pressure in the source room in some distance from the wall
is p12 ≅ p i2 + p r2 ≅ 2p i2 because we assume nearly total reflection from the
panel, i.e. pi ≈ pr. The relation between velocity of the forced vibrations of
the plate vfor and the sound pressure p1 in the source room yields:

2
2 p12
v for ≅ 2 (8.23)
Zw
212  Sound Insulation in Buildings

0.1
Loss factor, η

0.01

0.001
50 100 200 500 1000 2000 5000
Frequency, Hz
Rigid connection Elastic connection

Figure 8.5 Measured loss factor as a function of frequency for two walls, which were
the lower part of T-junction to a slab of 180 mm clinker concrete (1700 kg/
m3). Full line: 150 mm porous concrete wall (660 kg/m3) with rigid ­connection
and d­otted line: 100 mm porous concrete wall (660 kg/m3) with elastic
­connection. (Adapted from Rindel, J.H., Appl. Acoust., 41, 97–111, 1994.)

Forced transmission means the sound transmission due to transverse waves,

which are directly created by the sound pressure of incident sound wave.
The radiated (transmitted) sound power is:

Pfor = ρ cS v for
2
σ for (8.24)

where σfor is the radiation efficiency for forced bending waves (Equation
6.40). From Equation 5.5 and the above Equations 8.23 and 8.24, we get
the transmission loss for forced transmission:

2
p12 S 2
Zw ⋅ v for
Rf = 10 lg = 10 lg
4 ρ cPfor 8ρ c ⋅ Pfor
(8.25)
Z
= 20 lg w − 10 lg ( 2σ for ) (dB)
2ρ c
 Airborne sound transmission through single constructions  213

Considering a diffuse sound field in the source room, the wall impedance
(Equation 8.17) is inserted with θ = 90° because this angle of incidence cor-
responds to the coincidence dip at random incidence. The forced transmis-
sion for random incidence yields:

  −1 2 −1 2  
 f11 2 f 2   c  2 cf2c  2f  2  
2 2

Rf = R0f += 10 
R0lg+ 10 
1lg− 1 − 11
⋅ 1 − ⋅ 1 − + + c 2 
 + η + η 
2
   f  f    cS   cS f   f   tot  tot 
             
 

−10lg−(10lg
2σ for()2(dB)
σ for ) (dB) (8.26)

The loss factor ηtot determines how sharp the dip is, either at coincidence or
at the first structural mode.
For thin plates (f11 ≪ fc) and at frequencies well below the critical fre-
quency, the forced transmission (Equation 8.26) can be approximated by
the transmission loss:

 1 
Rf ≅ R0 − 10lg (2σ for ) ≅ R0 − 5 (dB)  f11 < f < fc  (8.27)
 2 

Here, the last approximation is valid within ±2 dB for surfaces above 10 m 2
and frequencies above 100 Hz. However, for small areas and low frequen-
cies, the area has a significant influence on the transmission loss (Figure 8.6).
The reason for this area effect of sound transmission is the same as applied
to a loudspeaker; the diameter of the loudspeaker membrane determines
how well low frequencies can be radiated. The low-­frequency limit is deter-
mined by the ratio between wavelength and diameter.
At frequencies above the critical frequency, the bending stiffness starts
to dominate the wall impedance and the transmission loss increases rapidly
with frequency:

Rf ≅ R0 + 40lg ( f fc ) − 5 (dB) ( f > fc ) (8.28)

However, in this frequency range, the forced transmission has no practical

importance because the resonant transmission dominates over the forced
transmission above the critical frequency, as shown in the following.
At very low frequencies, (f < f11) the transmission loss is controlled
by the stiffness of the plate, and the transmission loss increases rapidly
towards low frequencies by 12 dB per octave. The approximate result from
Equation 8.26 is:

Rf ≅ R0 − 40 lg ( f f11 ) − 10lg (2σ for ) (dB) (f < f11) (8.29)

We shall return to this subject in Section 8.3.

214  Sound Insulation in Buildings

50

40

30 R0
S = 1 m2
R, dB

S = 4 m2

20 S = 12 m2
S = 50 m2
S ∞

10

0
50 100 200 500 1000
Frequency, Hz

Figure 8.6 Influence of area on the transmission loss of a square panel with m = 25 kg/m2 .
(Adapted from Sato, H., J. Acoust. Soc. Jpn. 29, 509–516, 1973.)

8.1.6 Resonant transmission
As explained earlier, a plate that is excited by airborne sound will exhibit
forced vibrations and resonant vibrations, if the plate has any normal
modes in the frequency band of excitation. The resonant transmission
cannot be described in the same way as the forced transmission by
the wall impedance method, but instead the SEA can be applied (see
Chapter 7).
The transmission loss for resonant transmission alone Rr has been found
in Equation 7.23:

 c 2σ res
2
∆N 
Rr = R0 − 10lg  ⋅  (dB) (8.30)
 2ηtot Sf ∆f 

Here, the frequency dependency of the radiation efficiency σres is very

important (see Figure 6.18). For many constructions, especially thin plates,
the resonant transmission has only minor importance at frequencies below
 Airborne sound transmission through single constructions  215

the critical frequency, but it is the dominating transmission above the criti-
cal frequency where σres ≈ 1.
In the frequency range where bending waves can be assumed, the modal
density can be approximated by Equation 6.9 and the resonant transmis-
sion loss is:

πf σ 2   1 
Rr ≅ R0 − 10 lg  c res  (dB)  4 f11 < f < f s  (8.31)
 2ηtot f   2 

The cross-over frequency from bending to shear waves is fs (Equation 1.29).

At frequencies above fs /2, the wall is too thick for bending waves and the
modal density for shear waves (Equation 6.10) must be used instead,
­leading to:

 π c 2σ res
2
  1 
R ≅ R0 − 10 lg  2 
(dB)  f > f s  (8.32)
 ηtotcS   2 

A slight change in the slope with frequency can be noted at fs /2. While R0
has a slope of 6 dB per octave, the resonant transmission increases with
9 dB per octave below fs /2 and continues with 6 dB per octave above fs /2.
As pointed out in Chapter 6, the statistical approximations for modal
density in a plate should not be applied at frequencies below 4 f 11.

8.1.7 Total transmission: a calculation model

The transmission loss of a single construction can be described by adding
the forced and the resonant transmissions, as in Equation 7.26:

( )
R = −10 lg 10−0.1Rf + 10−0.1Rr (8.33)

Using this with Equation 8.26 for forced transmission and Equation 8.30
for resonant transmission, this makes a general calculation model for
the random incidence transmission loss of homogeneous and isotropic
constructions.
In addition to the area of plate, angle of incidence, speed of sound and
frequency, the main parameters are

R0: mass law (Equation 8.2)

f11: first natural frequency (Equation 8.15)
fc: critical frequency (Equation 1.28)
c S : speed of shear waves (Equation 1.26)

∆N
: modal density in the plate (Equation 6.9) bending or (Equation 6.10)
∆f shear
216  Sound Insulation in Buildings

ηtot: total loss factor (Equation 8.22)

σfor: radiation efficiency of forced vibrations at random incidence
(Equation 6.40)
σres: radiation efficiency of resonant vibrations at random incidence
(Equation 6.51)

We shall discuss this result for thin and thick plates, which behaves quite
differently from an acoustical point of view. An acoustically thin plate has
f11 ≪ fc , whereas an acoustically thick plate can be defined as one having
fc ≤ f 22 = 4 f11.
The critical frequency plays an important role for the transmission loss
of thin plates, and there is usually a local dip in the transmission loss curve
at or slightly above the critical frequency. Below the critical frequency,
the radiation from resonant vibration is weak and the forced transmis-
sion will normally be the dominating part of the transmission. However,
above the critical frequency, the resonant vibrations radiate efficiently and
dominate over the forced transmission. This behaviour is shown graphi-
cally in Figure  8.7. Between f 22 and ½ fc , the sound transmission is con-
trolled by the mass per unit area following the mass law, i.e. with a slope
of 6 dB per octave. It is noted that the increase of R by 6 dB per doubling
of the mass (the mass law) is only valid within a limited frequency range.
Increasing the thickness of the plate causes a lower critical frequency and a
higher modal resonance frequency, narrowing the mass law region.
Figures 8.8 and 8.9 show two examples of application of the calculation
model for an acoustically thin panel, namely, a 10 mm glass pane, but with
different dimensions. In the calculations, the internal loss factor is 0.01 and

f11 f22 fc 1/2 fs f

Figure 8.7 Principle of sound transmission loss for an acoustically thin single construc-
tion as a function of frequency. The critical frequency is high, fc > f 22 . Increasing
the thickness of the plate causes a shift from the solid curve to the dotted
curve and the key frequencies move up and down as indicated by the arrows,
narrowing the mass law region.
 Airborne sound transmission through single constructions  217

80

70
fc
60
Transmission loss, R, dB

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency, Hz

10 mm glass 1.6 × 1.3 Calculated Equation 8.27

Figure 8.8 Measured and calculated transmission loss of a 10 mm glass pane (25 kg/m2).
Dimensions 1.6 m × 1.3 m. The solid line is measured, the dashed line is calcu-
lation with the complete model (Equation 8.33) and the dashed-dotted line is
calculation with a simplified model (Equation 8.27). (Adapted from Cops, A.,
Myncke, H., The Sound Insulation of Glass and Windows – Theory and Practice. IV
Jornadas G.A.L.A., Cordoba, 1971.)

absorption coefficient at the edges is 0.2. The coincidence dip at the critical
frequency (ca. 1300 Hz) is clearly seen in the large glass pane (Figure 8.8).
Here, the first structural modes are below f 22 = 90 Hz and thus of no impor-
tance. Below the critical frequency, the transmission loss is controlled by
forced transmission, and the agreement with the simplified equation 8.27
is good.
However, several changes can be observed in the case of a small pane
(Figure 8.9). Here, f11 = 90 Hz and f 22 = 360 Hz and the low frequencies are
strongly influenced by the structural modes. The coincidence dip is shallow
and shifted to a somewhat higher frequency and the loss factor is higher
due to increased influence of boundary losses. The simplified model does
not work in this case, but the agreement with the complete model (Equation
8.33) is very good. The area effect shown in Figure 8.6 is not so clear in
reality; it is to some extent overruled by dips at the lower structural modes.
An example of a wall with a somewhat lower critical frequency is a
100 mm thick porous concrete wall. The measured transmission loss is
 218  Sound Insulation in Buildings

80

70

f11 fc fb
60

50
Transmission loss, R, dB

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency, Hz

10 mm glass 0.8 × 0.65 Calculated Equation 8.27

Figure 8.9 Measured and calculated transmission loss of a 10 mm glass pane (25 kg/m2).
Dimensions 0.8 m × 0.65 m. The solid line is measured, the dashed line is calcu-
lation with the complete model (Equation 8.33) and the dashed-dotted line is
calculation with a ­simplified model (Equation 8.27). (Adapted from Cops, A.,
Myncke, H., The Sound Insulation of Glass and Windows – Theory and Practice. IV
Jornadas G.A.L.A., Cordoba, 1971.)

compared with the calculated one, using Equation 8.33 in Figure 8.10. In
the calculations, the internal loss factor is 0.001 and absorption coeffi-
cient at the edges is 0.1. The critical frequency is calculated to 374 Hz, but
the coincidence dip is located at a somewhat higher frequency, f b = 508 Hz
(Equation 6.55). The first structural modes are below f 22 = 67 Hz and thus
of no importance.

8.2 SOUND INSULATION OF ACOUSTICALLY

THICK PLATES

We define an acoustically thick plate as one, with a critical frequency fc

lower than the f 22 structural mode. In this case, the critical frequency has
no practical importance and we have a general picture of the transmission
loss, which is quite different from the previous one (Figure 8.11). Since the
critical frequency is very low, the resonant radiation is efficient (σres ≈ 1) and
 Airborne sound transmission through single constructions  219

80

70

60
Transmission loss, R, dB

50

40

30

20

10
fc fb 1/2 fs

0
50 100 200 500 1000 2000 5000
Frequency, Hz
Aerated concrete Calculated R0, dB

Figure 8.10 Measured and calculated transmission loss of a 100 mm thick wall of porous
concrete. Dimensions are 3.7 m × 2.69 m. Mass per unit area is 130 kg/m2. (Data
from Lydteknisk Institut, Bygningsakustiske laboratoriemålinger, Building Acoustic
Laboratory Measurements, in Danish, Lyngby, Denmark, LL 871/68.)

the resonant transmission will dominate in most of the frequency  range.

The transmission loss is controlled by the bending stiffness of the plate,
gradually taken over by the shear stiffness above the crossover frequency ½
fs. The transmission loss is well approximated by the simplified equations

f11 f22 1/2 fs fh f

Figure 8.11 Principle of sound transmission loss for an acoustically thick single construc-
tion as a function of frequency. The critical frequency is low, fc ≤ f 22 .
220  Sound Insulation in Buildings

8.31 and 8.32 above the frequency f 22 = 4 f11. The slope of the transmis-
sion loss curve changes from 9 dB per octave below ½ fs to 6 dB per octave
above ½ fs (Figure 8.11). At very high frequencies, the longitudinal waves
cause an upper limit of the transmission loss, although this region if very
seldom seen in practice. The cross-over frequency is f h as previously defined
(Equation 5.21).
At the lower frequencies, the sound transmission of thick plates is domi-
nated by single structural modes. In the example of a 180 mm concrete
wall, the first structural mode is at 106 Hz, and already below 400 Hz the
modal density is so low that the equations derived from SEA cannot be used
directly. A statistical approach is not valid when the modal density is low
i.e. below f 22 .
However, we may still use the general Equation 8.33 but with great cau-
tion. With low modal density, the SEA can be replaced by a modal energy
analysis. This implies that the first structural modes are calculated one
by one, and then the actual mode count ΔN in each frequency band is
observed. If one-third octave bands are used, Δf = 0.23 f and the actual
number of modes ΔN are inserted in Equation 8.30. In some frequency
bands ΔN = 0, which means that no resonant transmission is present in that
particular frequency band; only forced transmission remains. Other bands
may contain just one mode, ΔN = 1, e.g. the one-third octave band where
f11 is located. Since the first structural modes are much spread (there are
normally only two modes between f 11 and f 22 that are two octaves apart),
the transmission loss may exhibit great variations and pronounced dips
between f11 and f 22 .
The example in Figure 8.12 is a 185 mm hollow concrete slab with criti-
cal frequency around 88 Hz and the first natural frequencies are within
the important frequency range (f11 ≈ 67 Hz and f 22 ≈ 268 Hz). In the calcu-
lations, the internal loss factor is 0.01 and absorption coefficient at the
edges is 0.05. Above 250 Hz, the statistical modal density can be used and,
thus, the transmission loss can be estimated from Equations 8.31 and 8.32.
However, below 250 Hz there are very few resonant modes, so the one-
third octave bands 63 Hz, 160 Hz and 200 Hz each contain just one mode
(ΔN = 1) while the one-third octave bands 80 Hz, 100 Hz and 125 Hz have
none (ΔN = 0). Insertion in the general Equation 8.33 and using the sta-
tistical modal density above 200 Hz leads to the result shown in Figure
8.12. The dip at 160 Hz might look like a coincidence dip, but it is not. It
is actually the opposite, namely, the consequence of relatively high sound
insulation in the lower frequency range from 80 Hz to 125 Hz. In this fre-
quency range, the slab has no modes and, thus, the resonant transmission
is missing and the sound insulation at these frequencies is relatively high. At
frequencies above 160 Hz, the slope is close to 6 dB per octave and the tran-
sition to shear waves is around 840 Hz (calculated fs = 1680 Hz). The critical
frequency is also marked in the figure, but it is typical for an acoustically
thick construction that there is no dip at the critical frequency.
 Airborne sound transmission through single constructions  221

80

70

60

50
Transmission loss, R, dB

40

30

20

10 f11 fc f21 f12 1/2fs

0
50 100 200 500 1000 2000 5000
Frequency, Hz
Hollow concrete slab Calculated R0, dB

Figure 8.12 Measured and calculated transmission loss for a 185 mm hollow concrete
slab (density 1700 kg/m3). Dimensions are 3.37 m × 2.99 m. Mass per unit
area is 315 kg/m2 . (Data from Lydteknisk Institut, Bygningsakustiske laborato-
riemålinger, Building Acoustic Laboratory Measurements, in Danish, Lyngby,
Denmark, LL 709/74.)

Another example is a 240 mm clinker concrete wall with transmission

loss as shown in Figure 8.13. In the calculations, the internal loss factor is
0.01 and absorption coefficient at the edges is 0.3. Again, it is remarkable
that there is no dip at the critical frequency (91 Hz), but some small dips at
the low structural modes (f11 = 69 Hz, f 21 = 140 Hz, f12 = 204 Hz). The speed
of shear waves is assumed to be around 1100 m/s, and this means that the
curve at frequencies above ca. 500 Hz is controlled by shear waves, which
yields the slope of 6 dB per octave.

8.3 SOUND INSULATION AT VERY

LOW FREQUENCIES

At very low frequencies, e.g. under 100 Hz, the sound insulation is a com-
plicated interaction between the structure and the connected rooms.
Particularly, the dimensions of the receiving room have been shown to be
important (Gibbs and Maluski 2004). A pronounced dip in the transmis-
sion loss curve can be expected in the one-third octave band that contains
222  Sound Insulation in Buildings

80

70

60

50
Transmission loss, R, dB

40

30

20
f11 fc f21 f12 1/2fs

10

0
50 100 200 500 1000 2000 5000
Frequency, Hz
240 mm clinker concrete Calculated R0, dB

Figure 8.13 Measured and calculated transmission loss for a 240 mm clinker concrete
wall (density 1800 kg/m3). Dimensions are 3.70 m × 2.69 m. Mass per unit
area is 430 kg/m2 . (Data from Lydteknisk Institut, Bygningsakustiske laborato-
riemålinger, Building Acoustic Laboratory Measurements, in Danish, Lyngby,
Denmark, LL 1156/77.)

the frequency of the first axial room mode perpendicular to the partition
wall (or floor). If the room dimension perpendicular to the surface is called
L x, the dip will be at the frequency c/(2L x). Gibbs and Maluski (2004) have
found this dip to be on average

• 5 dB for lightweight constructions

• 10 dB for heavyweight walls/floor greater than 10 m 2
• 20 dB for heavyweight walls/floor less than 10 m 2

Heavyweight constructions are normally acoustically thick plates (fc < f 22)
and at frequencies below the critical frequency, the transmission loss is con-
trolled by forced transmission. Thus, the transmission loss below f 11 can be
approximated by the simple Equation 8.29. This means that theoretically
the transmission loss below f 11 increases towards lower frequencies with
12 dB per octave. The plate may behave like a stiff membrane, and the stiff-
ness comes from the fixations at the edges of the plate.
 Airborne sound transmission through single constructions  223

80

70

60
Transmission loss, R, dB

50

40

30

20

10
f11

0
20 50 100 200 500 1000
Frequency, Hz
200 mm masonry walls Calculated Min. Max.

Figure 8.14 Calculated transmission loss and average of a large number of measured

sound level difference between equal rooms of various length. Heavyweight
walls (200 mm masonry walls, 385 kg/m2). The dashed line is calculation with
the complete model (Equation 8.33). (Adapted from Gibbs, B.M., Maluski,
S., Build. Acoust., 11, 61–78, 2004.)

The results from the survey conducted by Gibbs and Maluski (2004)
included ~100 room pairs and showed a clear difference in the low-fre-
quency level difference for lightweight and heavyweight constructions. In
the frequency range from 20 Hz to 50 Hz, the lightweight walls had average
sound insulation around 20 dB to 25 dB decreasing towards lower frequen-
cies, whereas the heavyweight walls had average sound insulation around
40 dB to 60 dB increasing towards lower frequencies (Figures 8.14 and 8.15).
The explanation behind this finding must be that the frequency of the
first structural mode in lightweight walls is very low (f11 < 20 Hz), whereas
for heavyweight walls, this frequency is relatively high (f11 > 50 Hz). It is
quite remarkable how different the heavyweight walls and the lightweight
walls behave at the low frequencies. The measured data are averaged over
many different room dimensions and, thus, the above-mentioned dip due
to room depth has been averaged out. For the heavyweight walls, the struc-
tural modes have a strong influence and cause great variations in the fre-
quency range from 50 Hz to 125 Hz, but below 50 Hz the spread of the
measured data is surprisingly small.
 224  Sound Insulation in Buildings

80

70

60
Transmission loss, R, dB

50

40

30

20

10

0
20 50 100 200 500 1000
Frequency, Hz
Lightweight walls Calculated Min. Max.

Figure 8.15 Calculated transmission loss and average of a large number of measured

sound level difference between equal rooms of various length. Lightweight
walls (approximated by 58 kg/m2). The dashed line is calculation with the
complete model (Equation 8.33). (Adapted from Gibbs, B.M., Maluski, S.,
Build. Acoust., 11, 61–78, 2004.)

8.4 SOUND INSULATION OF SUBDIVIDED STRUCTURES

Some structures like drywalls and windows are often subdivided into
smaller areas, and this may influence the sound insulation in various
ways:

• The frequency of the first structural mode f 11 is increased when the

plate dimensions get smaller (Equation 8.15).
• The modal density is decreased (Equation 6.9).
• The boundary losses are increased (Equation 8.20).
• The resonant radiation efficiency below and around the critical fre-
quency is increased (Equations 6.45, 6.50 and 6.52, 6.53).

This means that the resonant transmission is changed, and most signifi-
cantly above the critical frequency, where the transmission loss is controlled
 Airborne sound transmission through single constructions  225

by the modal density. The dimensions, circumference and area of the

­subdivided sections should be used in all the above-mentioned equations.
However, the forced transmission is only influenced to a minor extent,
since the excitation of the forced vibrations is correlated for the whole area,
and the radiation is always from the entire area. So, in the calculations,
σfor in Equation 8.26 and S in Equation 8.30 refer to the total area of the
construction, not the area of each subdivision.
As a calculated example we consider an 8 mm glass window with dimen-
sions 2 m × 2 m. If subdivided into a 2 × 2 grid or a 4 × 4 grid by a perfectly
rigid structure, the transmission loss is changed as seen in Figure 8.16. The
first structural mode is increased from far below 50 Hz (9 Hz) for the large
pane to around 146 Hz for the small panes measuring 0.5 m × 0.5 m. At fre-
quencies above the critical frequency, the transmission loss is increased due
to the decreased modal density in the smaller panes (see Equation 8.30).
This high-frequency effect of subdivided s­urfaces has been found experi-
mentally (e.g. Figure 7.10 in Lewis 1971) but was previously explained by
an increase of the ­boundary loss factor.

80

70

60
Transmission loss, R, dB

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency, Hz

8 mm glass, 2 m × 2 m Subdivided 2 × 2 panes Subdivided 4 × 4 panes

Figure 8.16 Calculated transmission loss of 8 mm glass with dimensions 2 m × 2 m, either

as a large pane or subdivided into 4 or 16 square panes.
226  Sound Insulation in Buildings

8.5 SOUND INSULATION AT OBLIQUE INCIDENCE

8.5.1 Definition of external transmission loss

In the case of windows and façades in a building, there is no source room,
and the incident sound power is not well defined. For instance, the incident
power of a plane wave at grazing incidence is normally assumed to be zero,
but nevertheless some sound will be transmitted. A solution to this prob-
lem is to introduce the so-called external transmission loss R E . The usual
definition of transmission loss is based on the transmitted sound power in
relation to the incident sound power. However, for this purpose, we define
the external transmission loss directly by Equation 5.5:

 p12 S  S
RE = 10 lg  2 ⋅  = L1 − L2 + 10 lg (dB) (8.34)
 p2 A  A

Figure 8.17 shows the parameters in Equation 8.34. The angle of incidence
does not go into the definition of external transmission loss.
Note that <p12> (and thus L1) is not really a spatial average since this is
the sound pressure in an open space. Instead, L1 is the sound pressure level
close to the wall including the energy reflection but excluding the interfer-
ence fluctuations.
It is also noted that RE is in reality the same as the apparent sound reduc-
tion index R’S as defined in ISO 16283-3 when the source is traffic noise [road,
rail or air traffic (see Equation 12.10)]. The only difference being that in the
ISO standard L1 is measured as the sound pressure level at the façade −3 dB.
The sound pressure level at a reflecting surface is known to be ~3 dB higher
than the average sound pressure level a few meters in front of the surface
­(see Figures 3.7 and 3.8).
An advantage of the definition (Equation 8.34) is that information about
the direction of the incident sound is not necessary. It may have any angle of
incidence, or it may be a more complicated sound field, like sound from line

S L2
L1
θ

Figure 8.17 External transmission loss of a window with area S defined from the sound
pressure levels L1 and L 2 and the absorption area A in the receiving room.
The angle of incidence is θ.
 Airborne sound transmission through single constructions  227

(e.g. a road). In the special case of a diffuse sound field on the source side,
the external transmission loss equals the normal transmission loss, R E = R.

8.5.2 Transmission from outside to inside

First, a point source with spherical radiation located outside in the distance r
from the window is considered. It is assumed that r is much greater than the
dimensions of the sound-transmitting surface. The squared sound pressure
outside the window p12 is taken as twice the value in a free field because of the
reflection from the façade. In a real situation, this is approximately the sound
pressure that can be measured in a distance of 2 m in front of the façade.

2Pa
p12 = ρ c (8.35)
4π r 2
Thus, the external transmission loss can be expressed in terms of the sound
power of the outside point source and the sound pressure in the receiving
room:

 P ρc S
a
RE = 10 lg  ⋅  (dB) (8.36)
 2π r 2 p22 A 

The reciprocity principle states that in linear sound fields, the sound pres-
sure in one point due to an omnidirectional point source in another point
remains the same if the source and receiver are switched. Figure 8.18 shows
the reciprocity principle applied to the case of sound transmission through
a window. The room with absorption area A can be either the source or
the receiving room. The relation between the sound power of the source Pa
and the sound pressure p 2 remains the same when the source and receiver
switch positions.

(a)
A
S p2

Pa r

(b)
A
S Pa

p2 r

Figure 8.18 Reciprocity principle applied to the sound transmission through a window in

a façade. (a) From outside to inside and (b) from inside to outside.
228  Sound Insulation in Buildings

8.5.3 Transmission from inside to outside

With transmission from inside to outside, the source is in the room with the
absorption area A. The source still has the sound power Pa and assuming a
diffuse sound field, it produces the average sound pressure in the source room:

4Pa
p12 = ρ c (8.37)
A

So, Pa in Equation 8.36 can be replaced by p1 and the external transmission

loss can be written as:

 P ρc S  p12 S 
a
RE = 10 lg  ⋅  = 10 lg  2 ⋅  (dB) (8.38)
 2π r p2 A 
2 2  p2 8π r 2 

This gives the following formula for calculation of the sound pressure level
L 2 in the outside receiver position when the indoor sound pressure level is
L1:

S
L2 = L1 − RE + 10 lg (dB) (8.39)
8π r 2

8.5.4 External transmission loss for

a single construction
For the forced transmission, it is necessary to go back to the wall imped-
ance as a function of the angle of incidence (Equation 8.17) and insert in
Equation 8.25. The radiation efficiency of the forced waves depends on the
angle of incidence σfor, θ, and it can be found in Equation 6.39. The forced
external transmission loss yields:

  2
 22   −1 2 2 −1 2
  
  f11  2  f11   c 
2
  cfc  fc   2 
22

   θ −⋅ sin 
  tot + ηtot 
 c θ ⋅ +c f  +  f + η
4 4 2
RE, f =R
= RR0E+, f 10 lg0 + 1  f 1 − ⋅f 1 −sin⋅  1
10−lg
  
         S  S    
 

σ for,θ )( 2(dB)
−10lg (−210lg σ for,θ ) (dB) (8.40)

The resonant transmission was derived by SEA in Equation 7.22. The result
contains twice the radiation efficiency for the resonant modes, one repre-
senting the excitation of the vibrations in plate and one representing the
sound radiation to the receiving room. In the case of the external transmis-
sion loss, these two radiation efficiencies must be treated separately because
the sound field is very different on either side. With transmission from the
outside to the inside, the first radiation efficiency is a function of the angle
 Airborne sound transmission through single constructions  229

of incidence σres,θ (Equation 6.57) and the other one represents the radiation
into a room σres,d (Equation 6.51). Thus, the resonant external transmis-
sion loss at the angle of incidence θ can be calculated from the following
formula:

 c 2σ res,θ σ res,d ∆N 
RE,r = R0 − 10lg  ⋅  (dB) (8.41)
 2ηtot Sf ∆f 

Finally, the combined, total transmission loss can be calculated using

Equation 8.33.
Examples of measured and calculated external transmission losses are
shown in Figure 8.19. The measurements were made in an anechoic room
on a window with 3 mm glass in scale model 1:4. This corresponds to
12 mm glass in full scale (Rindel 1975). Dimensions of the window cor-
respond to 1.68 m × 1.20 m = 2.0 m 2 in full scale. The radiation efficiency
σres,θ has strong fluctuations with frequency: so, for these calculations, the
radiation efficiency was averaged over ten frequency steps within each one-
third octave. It is noted that the calculation model works quite well, even in
the extreme case of grazing incidence (see Figure 8.19d).

8.6 ORTHOTROPIC PLATES

Many building materials are orthotropic, i.e. they have different character-
istics in different directions as opposed to isotropic plates. In many wooden
plate materials, the bending stiffness is larger in the direction parallel with
the veins than perpendicular to the veins. Another example of an orthotropic
plate is a corrugated metal plate where a trapeze or waving profile involves a
very large difference on the bending stiffness in the two orthogonal directions.
As shown by Heckl (1960), it is possible to insert a geometric average
value Bxy of the two bending stiffnesses Bx and By as a good approximation
in many of the results already known from the isotropic plates:

Bxy = Bx By (8.42)

For sound insulation, the important difference between isotropic and

orthotropic plates is that the latter has two different critical frequencies
corresponding to sound propagation in the direction of the largest or the
smallest bending stiffness, respectively. If the two directions are called x
and y, we have as in Equation 1.28:

c2 m K
fcx = = cx (8.43)
2π Bx h
230  Sound Insulation in Buildings

(a)
60

θ = 0°

50
External transmission loss, RE, dB

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency, Hz
Measured Calculated
(b)
80

θ = 45°
70

60
External transmission loss, RE, dB

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency, Hz
Measured Calculated

Figure 8.19 Measured and calculated external transmission loss of a scale model window
made from 12 mm thick glass. (a) Angle of incidence 0°, (b) angle of incidence 45°
(Continued)
 Airborne sound transmission through single constructions  231

(c)
80

θ = 75°
70
External transmission loss, RE, dB

60

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency, Hz
Measured Calculated
(d)
80

θ = 90°
70
External transmission loss, RE, dB

60

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency, Hz
Measured Calculated

Figure 8.19 (c) angle of incidence 75° and (d) angle of incidence 90°. Dotted line:
­measured; full line: calculated.
232  Sound Insulation in Buildings

c2 m Kcy
fcy = = (8.44)
2π By h
where fcx is the critical frequency in the direction with the largest bending
stiffness, i.e. fcx < fcy, h is the plate thickness and Kcx and Kcy are material
constants. As a result of Equation 8.42, the critical frequency can in many
relations be replaced by:

fc → fcx fcy (8.45)

However, this should be done with caution. For the natural frequencies of
the structural modes, Equation 6.7 changes to:

c 2  m2 n2 
fmn = + (8.46)
4  fcxlx2 fcyly2 

The modal density is changed from Equation 6.9 to:

∆N π ⋅ S
≅ 2 fcx fcy (8.47)
∆f c

The two different critical frequencies involve the possibility of two dips in
the sound insulation and a relatively low sound insulation in the frequency
range between the dips. According to Heckl (1960), the radiation efficiency
in the frequency range fcx < f < fcy is determined by

2
1 fcx  4f 
σ res ≅  ln f  (8.48)
π2 fcy cx

Below the lower critical frequency (f < fcx), the transmission loss can be cal-
culated from the forced transmission (Equation 8.26) modified by introduc-
ing two critical frequencies instead of only one and simplified by neglecting
shear waves and the frequencies below f 11:

  f 
2
 f 
2 
R ≅ Rf = R0 + 10 lg  1 −   1 −   + ηtot 2
 − 10lg ( 2σ for ) (8.49)
  fcx   fcy  

Above the lower critical frequency fcx, the resonant transmission will nor-
mally be dominant, and the transmission loss is determined by inserting
Equations 8.47 and 8.48 in Equation 8.30:

 1 f fcx  4f  
4

R ≅ Rr = R0 − 10 lg  3
2π η
⋅ cx
f fcy  f   ( fcx < f < fcy ) (8.50)
ln
 tot cx 
 Airborne sound transmission through single constructions  233

In some cases, the forced and the resonant transmission are both important
in this frequency range, and Equation 8.33 is used to find the total trans-
mission loss.
In the upper frequency range above the higher critical frequency, Equation
8.30 can be applied with the modal density (Equation 8.47) and simplified
with the approximation σres ≈ 1:

π fcx fcy
R ≅ Rr = R0 − 10 lg
2ηtot f
( f > fcy ) (8.51)

For some orthotropic plates, the interval between fcx and fcy is so small that
only one coincidence dip appears but the dip may cover a larger frequency
range than with isotropic plates. Other orthotropic plates may have a very
large interval between the critical frequencies; some corrugated plates can
have critical frequencies from below 200 Hz to above 20 kHz.
Figure 8.20 shows an example of measured and calculated trans-
mission loss of a corrugated steel plate. With reference to the sketch in
Figure 8.20, the dimensions of the corrugation are a = 120 mm, b = 120 mm,
c = 45 mm, d = 30 mm and h = 0.7 mm. The plate has a mass of 6.3 kg/m 2
and assumed internal loss factor of 0.001. The critical frequencies used for
80

70 c
h

60 a d b d
Transmission loss, R, dB

50 fcx

40

30

20

10

0
100 200 500 1000 2000 5000 10000
Frequency, Hz
Corrugated steel Calculated R0, dB

Figure 8.20 Example of measured and calculated transmission loss of a 0.7 mm ­corrugated

steel plate with a profile as shown in the sketch. The critical frequencies are
fcx = 750 Hz and fcy = 18,500 Hz. (Data from Cederfeldt, L. Sound insulation of
corrugated plates – a summary of laboratory measurements. Report no. 55,
Division of Building Technology, Technical University of Lund, Sweden.)
234  Sound Insulation in Buildings

the calculations are 750 Hz and 18,500 Hz, respectively. Below the lower
­critical frequency, the transmission is forced (Equation 8.48) and above that
frequency, the transmission is resonant (Equation 8.49). It is ­typical for a
corrugated plate that the transmission loss is far below the mass law.

8.7 SANDWICH PLATES

A sandwich construction is a plate composed of two parallel, thin laminas

or face plates united with a lightweight core material. Such constructions
can combine a high static stiffness with low weight, especially useful in air-
planes and ships. The principle is also known from building materials, e.g.
the so-called zero-energy elements where the face plates are plywood plates
and the core is mineral wool placed edgeways with the fibre at right angles
to the plates. Other core materials could be plastic foam or honeycomb
core, which resembles honeycombs in a beehive but is made by metal foil,
paper or plastic. A very common building material that slightly behaves as
a sandwich plate is the gypsum (or plaster) board, covered on each side of a
sheet of paper. The paper is very thin and has ignorable bending stiffness,
but the in-plane tension stiffness is higher than for the gypsum core, forcing
extra shear motion on the gypsum core.
Figure 8.21 shows a sandwich plate and some of the wave types that may
appear in the sound transmission through sandwich plates. The dilatation
wave (Figure 8.21e) will be considered later, but is disregarded at this point,
i.e. the core material is assumed to act as an incompressible material that
simply keeps the two face plates at a constant distance from each other. The
face plates are assumed to be identical.

8.7.1 Speed of transverse waves

The mass per unit area of the sandwich plate is:

m = mc + 2mf = ρ c hc + 2ρ f hf (8.52)

(b) cB

(c) cS
(a) hf

(d) cBf
hc h
cBf
(e)
hf

Figure 8.21 (a) Section of sandwich panel, (b) bending wave, (c) shear wave, (d) bending
waves in the face plates and (e) dilatation resonance in elastic core.
 Airborne sound transmission through single constructions  235

Here, index c refers to the core material and index f refers to the face plates.
According to Ver and Holmer (1971, p. 313), the bending stiffness of the
sandwich plate is:
1
Ef hf ( hf + hc ) (8.53)
2
B=
2
The bending wave is shown in Figure 8.21b. This has the speed of propaga-
tion cB determined by Equation 1.25:

B
cB = 2π f 4 (8.54)
m

One peculiarity of sandwich plates is that the bending stiffness is very high
even if it is a lightweight construction and, thus, the speed of the bending
wave is extremely high. Shear wave motion plays an important role in sand-
wich plates. In the shear waves, there is no compression or dilatation of the
face plates, and the movements in the core material are perpendicular to
the direction of propagation (Figure 8.21c). The propagation speed of the
shear wave is:

Gh
cS = (8.55)
m
where G is the shear modulus, h is the total thickness of the plate and m is
the total mass per unit area. The core material is enclosed such that it can-
not extend to the sides. The shear modulus is determined by the modulus of
elasticity Ec of the core material and the Poisson’s ratio μ:
Ec
G= (8.56)
2 (1 + µ )
At high frequencies (short wavelengths), the shear wave is influenced by the
bending stiffness of the face plates. So, if we divide the sandwich plate into
two parts along the dotted line in Figure 8.21d, the shear wave actually
becomes a bending wave in each of the face plates. The bending stiffness of
the face plate is (Equation 1.22):

Ef hf3
Bf = (8.57)
12(1 − µ 2 )

The propagation speed cBf is determined by Equation 1.25, but with half
the mass of the core material added to the mass of the face plate:

Bf 2Bf
cBf = 2π f 4 = 2π f 4 (8.58)
mf + m c /2 m
236  Sound Insulation in Buildings

The described waves (Figure 8.21b through d) will not appear separately
as pure bending or pure shear waves, but there is a gradual transition as a
function of the frequency. An electrical model of the wall impedance of a
sandwich panel is seen in Figure 8.22. The bending and shear stiffnesses
are in parallel as for a thick panel, but in addition the bending stiffness of
the face plates adds to the shear stiffness.
The inductor is L = m, and the capacitors are:

c4 c2 c4
C1 = , C2 = , C3 =
ω 4B ω 2Gh ω 4 Bf

The wall impedance is then calculated from:
−1
1  1 
Zw = jωL + C1 +  +r
jω  1 / C2 + 2 / C3 
(8.59)
 c4 4
cBf + c 2cS2 
= jω m 1 − B4 ⋅ 4 4  + ηω m
2 2
 c cB + cBf + c cS 

The possibility of compression or dilatation of the core material is not

taken into account at this point, but the core is assumed to keep a constant
distance between the face plates. The resulting free transverse wave propa-
gates with the effective bending wave speed cB,eff, which can be found as the
speed c by setting Im{Zw} = 0. Using that cBf ≪ cB , this leads to an equation
in c of the sixth degree (Kurtze and Watters 1959):

cS2 6
c + c 4 − cS2c 2 − cBf
4
= 0 (8.60)
cB4

Bending Shear
C3
C1 C2
C3

Figure 8.22 Electrical analogy of the wall impedance of a sandwich panel.

 Airborne sound transmission through single constructions  237

The solution of this equation is not simple, but a good approximation

can be derived (Rindel 1988). The effective bending wave speed has ­a n
S-shaped function of frequency, as shown in Figure 8.23. It is ­obvious
that cB ≫ cBf and at the crossing point marked 1 in the figure cBf ≪ c S and
thus:

cS = cB ⇒ cB,eff = c s
1
2
( )
5 − 1 ≅ cS ⋅ 0.7862 ≅ cS ⋅ 2−1/3

At the crossing point marked 2 in Figure 8.23, cB ≫ cs and thus:

cS = cBf ⇒ cB,eff = cS
1
2
( )
5 + 1 ≅ cS ⋅ 1.2720 ≅ cS ⋅ 21/3

Based on these observations, the approximate solution to Equation 8.60

becomes:

−1/3
 1 1 
cB,eff ≅  3 + 3 3  (8.61)
c
 B c S + cBf 

CB
1000
Speed of propagation, m/s

1 fc

c = 344 m/s

Cs

CBf

100
log (frequency)

Figure 8.23 Dispersion curve for the effective bending wave speed in a sandwich
plate. Here, the shear wave speed is assumed to be less than the speed of
sound in the air. The critical frequency is where the effective bending wave
speed equals the speed of sound in the air. The crossing points 1 and 2 are
explained in the text.
238  Sound Insulation in Buildings

The influence of cB,eff is discussed on the basis of the dispersion curve

(propagation speed of transverse waves as function of frequency, see
Figure 8.23). It is noted that cB,eff increases with the frequency in a rela-
tively complicated way. At low frequencies, cB,eff → cB and the asymptotic
slope is cB ~ √f. In a mid-frequency range, the vibrations have the char-
acter of shear waves and cB,eff only increases slowly with the frequency.
At high frequencies, the slope of the curve increases again as cB,eff → cBf
because the bending stiffness in the face plates adds to the total stiffness
and becomes determining.
The critical frequency fc can be found by the definition, i.e. where cB,eff = c
(Figure 8.23). fc will be somewhere in the interval between the low critical
frequency corresponding to the plane bending and the high critical fre-
quency for the face plates alone. The crucial thing for the location of fc can
be seen to be the ratio of the speed of shear waves to the speed of sound
in air, i.e. c S /c. The best construction in terms of sound insulation will be
achieved when c S < c as this gives a high critical frequency and thus the low-
est possible contribution from the resonant transmission. The forced trans-
mission will dominate and the transmission loss will follow Equation 8.27.
However, a core material with very low shear modulus G will be needed
and in practice G will often be so large that c S ≈ c or c S > c. Increasing the
shear wave speed to exceed c has the consequence that the critical frequency
changes drastically from very high to a low frequency. So, instead of forced
transmission, the resonant transmission can be expected to dominate the
transmission loss.

8.7.2 Resonant transmission
The dispersion function for the transverse waves also has consequences
for the natural frequencies of the structural modes. Assuming simply sup-
ported edges, the natural frequencies fmn are determined as in Equation 6.6,
but now using the wavelength λB,eff = cB,eff /f:

2 2 2 2

 2   2fmn   m   n  (8.62)
 λ  =  c  =  l  +  l 
B,eff B,eff x y

However, the solution for the natural frequencies is not simple because of
the dispersion function of cB,eff.
Figure 8.24 shows a geometrical representation of Equation 8.62, each
dot represents a structural mode and the frequency is represented by the
radius 2/λB,eff. The statistical number of modes below a certain frequency f
is the number of dots within the quarter circle with the radius correspond-
ing to that frequency:
 Airborne sound transmission through single constructions  239

n/ly

n:
4
2/λB
3

m: 1 2 3 4 5 6 m/lx

Figure 8.24 Geometrical representation of Equation 8.62, each dot representing a

structural mode.

1/4π ⋅ ( 2 / λB,eff )
2
f2
N= = π l xl y (8.63)
(1 / l x ) ⋅ (1 / l y ) cB2 ,eff

With the area S = lx · ly, the modal density is found by differentiation with
frequency:

∆N f  f dc 
= 2π S 2 1 − ⋅ B,eff  (8.64)
∆f cB,eff  cB,eff df 

For pure bending waves, we have cB ~ √f, while the speed is constant for
shear waves.
The resonant transmission depends on the modal density and by inser-
tion in Equation 8.30, we get:

 c 2σ res
2
∆N 
Rr = R0 − 10 lg  ⋅ 
 2η Sf ∆f 
(8.65)
 2  2
 πσ res c  
= R0 − 10 lg ⋅  − ∆Rc (dB) (f > fc )
 2ηtot  cB,eff  
 
where the correction term ΔRc is determined from the slope of the disper-
sion curve:

  f dc 
∆Rc = 10 lg 2 1 − ⋅ B,eff  (dB) (8.66)
  cB,eff d f 
240  Sound Insulation in Buildings

This term takes values between 0 dB and 3 dB, the former referring to pure
bending waves and the latter referring to pure shear waves.
The sound radiation from the resonant modes depends on the speed of
the transverse wave relative to the speed of sound in air (see Equation 6.49).
This leads to the radiation efficiency above the critical frequency:

−1/2
  2
c  
σ res ≅ 1 −   ( f > fc ) (8.67)
  cB,eff  
 

This is close to unity, σres ≈ 1 for cB,eff ≫ c. Insertion in Equation 8.65 gives
the following result for the transmission loss above the critical frequency:

2ηtot  c 
2 
Rr ≅ R0 + 10 lg + 10 lg  B,eff  − 1 − ∆Rc (dB) ( f > fc ) (8.68)
π  c  

8.7.3 Forced transmission
The transmission loss for forced transmission through a sandwich panel
can be calculated as for a homogeneous panel (Equation 8.25). The coinci-
dence dip around the critical frequency can be quite pronounced if c S ≫ c.
However, some sandwich constructions have c S < c, and this means that the
coincidence dip is more shallow or disappears completely.
Inserting the wall impedance Equation 8.59 in Equation 8.25 yields the
transmission loss for forced transmission:

  22 22 4 2 22
2
 
c⋅ B1 − ccBBf + cc2BfcS2+ c cS 2 + η 2 
4
  f11 
44
= + 
 f11 − ⋅
R R 10 lg 1
Rf = R0f + 100lg  1 −     ⋅  1− 4 ⋅ 4 4 4 4 24 2  2+ η 
cB ++ccBfcS + c cS2 tot 
tot
   f   f   c cBc+ cBf 

 

σ for ()2(dB)
−10lg−( 210lg σ for ) (dB) (8.69)

An example of measured and calculated transmission loss for a sandwich

panel with c S /c > 1 is shown in Figure 8.25. With a stiff core material and
a low critical frequency (fc = 170 Hz), the sound transmission is resonant in
most of the frequency range, and in general the sound insulation is low. The
calculated forced and resonant transmission losses are shown separately.

8.7.4 Dilatational resonance
So far, the core material has been assumed to be incompressible. However,
for porous core materials like foam and mineral wool, the core is com-
pressible and dilatational resonances may occur (Figure 8.21e). The first
 Airborne sound transmission through single constructions  241

80

f11 fc fdil
70

60

50
Transmission loss, R, dB

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency, Hz
Measured R0 Rf Rr

Figure 8.25 Example of measured and calculated transmission loss of a sandwich panel

made from two sheets of 6.4 mm plywood and a 76 mm core of honeycomb
plastic. Panel dimensions 3.0 m × 2.4 m, G = 95 MN/m2 , cS /c = 2.18, fc = 170 Hz
and f11 = 50 Hz. The dilatational resonance is above 5000 Hz. (Measured data
from Jones, R.E., Control Eng., 16, 90–105, 1981.)

dilatational resonance is a mass-spring-mass resonance in which the faces

act as masses connected by a spring. Higher-order dilatational resonances
are also possible, in which a standing wave in the core connects the masses
of the face plates. The first dilatational resonance frequency can be calcu-
lated by Ver and Holmer (1971, p. 314):

1 4Ect
fdil = (8.70)
2π hc ( 2mf + mc / 3)

where Ect is the compression modulus of the core material given by:

Ec
Ect = (8.71)
3 (1 − 2µ )
242  Sound Insulation in Buildings

Ec is Young’s modulus of the core material and μ is the Poisson ratio. It

is seen that μ has a major influence on the dilatation resonance frequency,
which increases towards infinity if μ approaches 0.5. Rubber is an example
of a material with such a high Poisson ratio and rubber makes a very hard
and stiff spring if the material cannot expand sidewise. Normally, fdil can-
not be made sufficiently low to achieve a positive effect of compressible core
material. Therefore, this is in contrast to double constructions, which can
often have a low resonance frequency and very high sound insulation above
the resonance frequency.
Since the dilatational resonance behaves very much like the mass-spring-
mass resonance of a double wall, the transmission loss for forced trans-
mission (Equation 8.69) can be extended with an additional term like
(Equation 5.52):

 2 2 
 f  
Rf ,dil 
= Rf + 10 lg  1 −  2 
 + ηtot (dB) (8.72)
   fdil   
 

In summary, sandwich panels can be divided into two types. One type
has a core material that provides a shear wave speed less than the speed
of sound in air. Then, the critical frequency is high, and the transmission
loss is controlled by the forced transmission and the result can be close
to the mass law. However, the dilatational resonance may spoil the sound
insulation. The other type has a core with a high shear modulus and the
shear wave speed exceeds the speed of sound in air. The ­critical frequency
is low and the transmission loss is controlled by ­resonant transmission.
The mass law makes an upper limit for the obtainable sound insulation of
sandwich plates and even small alterations in the core material may lead
to major deteriorations of the sound insulation.
Another example of the transmission loss for a sandwich plate with a
relatively stiff but compressible core of mineral wool is given in Figure
8.26. Here, c S /c < 1 meaning that the free transverse vibrations propa-
gate at subsonic speed and, thus, the resonant transmission is negligible.
In such constructions, one or more dilatation resonances may occur as
the masses of the face plates oscillate with the core material as a spring
(Figure 8.21e). The fundamental dilatational resonance is clearly seen in
the example in Figure 8.26, butthere is also a dip at a higher-order dilata-
tional resonance. Although not included in the current calculation model,
this second dip can be associated with the thickness of the core hc being
equal to half a wavelength of longitudinal waves in the core material, the
 Airborne sound transmission through single constructions  243

80

70

fdil
60

50
Transmission loss, R, dB

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency, Hz
Measured Calculated R0, dB

Figure 8.26 Example of measured and calculated transmission loss of a sandwich panel

made from two sheets of 0.6 mm steel and a 50 mm core of mineral wool
with fibres perpendicular to the surface. Panel dimensions 3.7 m × 2.69 m,
G = 1.67 MN/m2 , cS /c = 0.22, fc = 20,500 Hz and f11 = 34 Hz. The dilatational
resonance is around 950 Hz.

speed of longitudinal waves being calculated from Equation 1.20 for an

extended solid medium. This example has a very high critical frequency,
and the forced transmission is dominating over the resonant transmis-
sion. The sound insulation is better than in the previous example, but
it is obvious that it is difficult to obtain good sound insulation with a
sandwich panel.
Examples of data for materials applicable as core material in sandwich
panels are collected in Table 8.2. Attempts to optimise the sound insulation
of sandwich panels should include a core material that can give a low speed
of shear waves and a high critical frequency, like this example, but with
a Young’s modulus high enough to avoid problems with the dilatational
resonance.
244  Sound Insulation in Buildings

Table 8.2  Examples of typical data for core materials in sandwich panel
Young
Density, Modulus, Shear Modulus,
Material ρm (kg/m3) E (106 N/m2) G (106 N/m2)
Urethane foam 33–72 7–19 2.1–3.7
Polystyrene foam 33–42 9–26 8.5
Honeycomb, aluminium 45–123 482–368 –
Honeycomb, plastic 64–149 248–1360 –
Plastic foam 30 6.2 2.1
Glass wool, 120 0.34–0.48 –
fibres ≠ surface
Glass wool, fibres ┴ 120 <6.5 9
surface
Glass wool, fibres ┴ 80 <1.9 3.6
surface
Glass wool, fibres ┴ 45 <1.2 1.3
surface
Source: Bodlund, K. Luftlydisolering. En sammanstälning av tillämplig teori, (Airborne sound
insulation. A collection of applicable theory, in Swedish). Rapport R60:1980, Statens
råd för Byggnadsforskning, Stockholm, 1980.

REFERENCES

K. Bodlund (1980). Luftlydisolering. En sammanstälning av tillämplig, (Airborne

sound insulation. A collection of applicable theory, in Swedish). Rapport
R60:1980, Statens råd för Byggnadsforskning, Stockholm.
L. Cederfeldt (1974). Sound insulation of corrugated plates – A summary of labora-
tory measurements. Report No. 55, Division of Building Technology, Technical
University of Lund, Sweden.
R.J.M. Craik (1981). Damping of building structures. Applied Acoustics 14,
347–359.
A. Cops and H. Myncke (1971). The Sound Insulation of Glass and Windows –
Theory and Practice. IV Jornadas G.A.L.A., Cordoba, September 1975.
B.M. Gibbs and S. Maluski (2004). Airborne sound level difference between dwell-
ings at low frequencies. Building Acoustics 11, 61–78.
M. Heckl (1960). Untersuchungen an orthotropen Platten. Acustica 10, 109–115.
ISO 16283-3 (2016). Acoustics – Field measurements of sound insulation in build-
ings and of building elements – Part 3: Façade sound insulation. International
Organization for Standardization, Geneva, Switzerland.
R.E. Jones (1981). Field sound insulation of load-bearing sandwich panels for hous-
ing. Noise Control Engineering 16, 90–105.
G. Kurtze and B.G. Watters (1959). New wall design for high transmission loss or
high damping. Journal of Acoustical Society of America 31, 739–748.
P.T. Lewis (1971). Real windows. Chapter seven in Building Acoustics (Eds. T. Smith,
P.E. O’Sullival, B. Oakes, R.B. Conn). British Acoustical Society Special, vol. 2,
Oriel Press, London.
 Airborne sound transmission through single constructions  245

Lydteknisk Institut. Bygningsakustiske laboratoriemålinger, Building Acoustic

Laboratory Measurements, in Danish, Lyngby, Denmark.
J.H. Rindel (1975). Transmission of traffic noise through windows. Influence of inci-
dent angle on sound insulation in theory and experiment. Report No. 9, The
Acoustics Laboratory, Technical University of Denmark, Lyngby.
J.H. Rindel (1988). Sound transmission through sandwich panels. Proceedings of
NAM-88, Tampere, Finland, 55–58.
J.H. Rindel (1994). Dispersion and absorption of structure-borne sound in acousti-
cally thick plates. Applied Acoustics 41, 97–111.
H. Sato (1973). On the mechanism of outdoor noise transmission through walls
and windows. Journal of the Acoustical Society of Japan 29, 509–516. (In
Japanese).
I.L. Ver and C.I. Holmer (1971). Interaction of sound waves with solid structures.
Chapter 11 in Noise and Vibration Control (Ed. L.L. Beranek). McGraw-Hill,
New York.
Chapter 9

Airborne sound transmission

through double constructions

The theoretical basis of sound insulation of double constructions is

described in this chapter. However, the theoretical model is combined with
semi-empirical models, which allows relatively simple mathematical formu-
lae. This may be a basis for prediction of the sound insulation of a number
of building components, e.g. double-glazed windows and double wall con-
structions with or without mechanical connections. The difficulties in cal-
culating the sound insulation of double constructions are especially related
to mechanical connections including boundary connections.

9.1 TRANSMISSION VIA THE CAVITY

A double construction can consist of two parallel plates separated by a cav-

ity d, which contains an elastic material with the density ρd , and the speed
of sound cd. The cavity will often contain air (with or without mineral
wool), but special window constructions may use other gases in the cavity.
As shown in Figure 9.1, the sound transmission will be partly through the
cavity and partly via possible mechanical connections, the so-called sound
bridges. The latter will be dealt with in the following paragraphs; here only
transmission via the cavity is dealt with. The two plates are characterized
by their mass per unit area m1 and m2 , their critical frequencies fc1 and fc2
and the transmission losses R1 and R 2 for each of the two plates.

9.1.1 Normal incidence of sound

If the boundary conditions of a plate are neglected, a plane sound wave
at normal incidence will not create bending waves in the plate, but the
plate will move with all parts in phase. From the equations of movement
(Newton’s second law) and from the conditions of continuity, the result in
Equation 5.43 was derived. Assuming the gas in the cavity to be different
from the atmospheric air outside, and wall impedances to be determined
from the mass per unit area, the following expression for the transmission
loss can be obtained:

247
248  Sound Insulation in Buildings

m1 m2
fc1 fc2
d
R1 R2
ρd cd
P1 P2I
P2b

Figure 9.1  Sound transmission through a double construction can be divided into one con-
tribution via the cavity and another contribution via connections (sound bridges).
(Reproduced from Kristensen, J., Rindel, J.H., Bygningsakustik – Teori og praksis.
(Building acoustics – Theory and practice, in Danish). SBI-Anvisning 166. Danish
Building Research Institute, Hørsholm, Denmark, 1989. With permission.)

(
R = 10 lg  cos(kd) − n ( K1 + K2 ) sin(kd) 
2

(9.1)
1 
2

+  (n + 1 / n)sin(kd) + ( K1 + K2 ) cos(kd) − 2nK1K2 sin(kd)  + η02 
2  
where

k = 2πf/cd is the angular wave number

d is the depth of the cavity
η 0 is the loss factor, which is only significant at the resonance

ρc
n= is the ratio of impedance of the air outside the wall and the air
ρ d cd
in the cavity
ω m1
K1 = is the normalized wall impedance for plate 1
2ρ c
ω m2
K2 = is the normalized wall impedance for plate 2
2ρ c

Normally, there is no need to include all terms in Equation 9.1. For most
double constructions, only two terms (and the loss factor) are of importance:

R ≅ 10 lg (((K + K ) cos(kd) − 2nK K sin(kd)) + η )(9.2)

1 2 1 2
2 2
0

It is seen from this that R will have a minimum if the two important terms
become numerically equal, which gives the following equation:
 Airborne sound transmission through double constructions  249

ω d m1 + m2 ρ d cd
tan(kd) = tan = ⋅ (9.3)
cd m1m2 ω

9.1.2 The resonance frequency

At frequencies where the wavelength is large compared to the depth of the
cavity, i.e. kd << 1, the tangent function equals the argument, and from
Equation 9.3 we find the resonance frequency:

1 ρ d cd2  1 1  (9.4)
f0 = +
2π d  m1 m2 


A mass-spring-mass resonance is caused by the spring effect of the gas in the

cavity. At the resonance frequency, a change occurs in the character of the
sound transmission compared to the single wall, as it is at frequencies above
the resonance frequency that double constructions are also acoustically double.
Figure 9.2 shows how to determine the resonance frequency for a
­symmetric double construction with air in the cavity.

1000

f0 fd

100
Panel distance d (mm)

m1 = m2(kg/m2)
5

10
10
20
50
10
0
20
0

1
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 9.2 Resonance frequency f0 and the knee frequency fd as functions of the cavity
depth for a symmetrical double construction with air in the cavity. The knee
frequency involves only the cavity depth, whereas the resonance frequency also
depends on the mass per unit area of the two plates. The indicated example
is for d = 200 mm and m1 = m2 = 5 kg/m2 , which yields f0 = 85 Hz and fd = 275 Hz.
250  Sound Insulation in Buildings

9.1.3 Standing waves in the cavity

At higher frequencies where the wavelength is not large compared to the
depth of the cavity, the spring effect of the air is no longer active. This hap-
pens at a knee frequency fd corresponding to kd = 1 or d ≈ λ/6:

cd
fd = (9.5)
2π d

Above this frequency, there are natural frequencies that are solutions to
Equation 9.3 as tan(kd) ≈ 0. These frequencies represent standing waves in
the cavity orthogonally on the two plates, and in principle this gives a num-
ber of dips in the frequency curve for the transmission loss. In practise,
however, these dips cannot be seen as the sound incidence not only occurs
orthogonally but also at diffuse sound incidence. For f > fd , the maximum
transmission loss is achieved by the insertion of cos (kd) ≈ 0 and sin (kd) ≈ 1
when calculating the transmission loss. These are good approximations
when the cavity is attenuated with absorbing material.

9.1.4 Calculation for random incidence

The previously given expressions for the transmission loss of double con-
structions Equations 9.1 and 9.2 give useful results by inserting the trans-
mission losses R1 and R 2 of the two plates; or even better the measured
transmission losses at random incidence. In this case, the equivalent wall
impedances for the two plates can be taken as:

K1 = 10R1 / 20 and K2 = 10R2 / 20 (9.6)

For estimated calculations, it can be fair to simplify Equation 9.2 depending

on the different frequency ranges. This can be done by means of Equations
9.5 and 9.6 and using that sin(kd) ≈ kd for f < fd and sin(kd) ≈ 1 for f > fd.

 R = 20 lg ( K1 + K2 ) (f < f0 )
 (1+ 2)
  f 
R ≅  R1 + R2 + 20 lg  2n  ( f0 < f ≤ fd ) (9.7)
  fd 
 R + R + 20 lg 2n
 1 2 ( ) (f > fd )

The corresponding simplified frequency course is seen in Figure 9.3. In the

case of thin plates and at frequencies below the critical frequency (f < fc), a
simplified mass law may be used for an estimate of R1, R 2 and R1+2 . For the
latter, the simplified estimate comes from Equations 8.3 and 8.27:
 Airborne sound transmission through double constructions  251

80

70
fd it y
lim vit
p er ca
Up d
60 pe
dam
y y
rtl vit
Pa ca
Transmission loss, R (dB)

50 p ed
m
y da
kl
f0 ea
40 W

t
imi
er l
30 Low

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 9.3 Simplified frequency course of the transmission loss of double constructions

with more or less sound absorption in the cavity, but without rigid connec-
tions. The three main sections correspond to Equation 9.7. The upper limit
for the sound insulation that can be achieved in practice is indicated. The
lower limit is also indicated. The two dashed-dotted lines show different
examples of a partly attenuated cavity.

 m + m2   f 
R(1+ 2) ≅ 20 lg  1 + 20 lg  − 7 dB (9.8)
 1 kg m −2   100 Hz 

However, if R1 and R 2 are known from measurements or a more reliable

prediction than the simple mass law, the application of the equivalent wall
impedances (Equation 9.6) should be preferred, instead of Equation 9.8.
It is implied that there are some absorbing material in the cavity and,
thus, standing waves are efficiently attenuated. As sound bridges are not
included at this stage, this is the maximum achievable sound insulation for
an ideal double construction.
It should be emphasized that normally n = 1, even if the cavity contains
mineral wool, as long as this does not contribute to the spring stiffness of
the cavity. The only case where n ≠ 1 is when the cavity is filled with a spe-
cial gas as sometimes used in windows for improved thermal insulation. If
the cavity contains atmospheric air, n = 1, and we get:
 252  Sound Insulation in Buildings

  f 
 R1 + R2 + 20 lg   + 6 dB ( f0 < f < fd )
R≅  fd  (9.9)

 R1 + R2 + 6 dB ( f ≥ fd )

9.1.5 Weakly attenuated cavity

Double constructions, in which the cavity is not efficiently attenuated with
sound absorbing material, will give a sound insulation between the upper and
lower bounds, as shown in Figure 9.3. With weak attenuation, a higher sound
pressure is built up in the cavity leading to an increased sound ­transmission. In
window constructions, a sound absorbing treatment of the window frame –
a so-called frame-absorber along the perimeter of the cavity – is sometimes
used. The transmission loss for such a construction can be estimated by:

 d ⋅U 
R ≅ R1 + R2 + 10 lg  α  (9.10)
 S 
80

70

it
100 mm im Empty
60
erl
p
50 mm Up
Transmission loss, R (dB)

50

40

t
imi
30
wer l
Lo

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 9.4 Measured transmission losses of a double wall with 6.4 mm and 3.2 mm
hardboard plates and a cavity of 160 mm without sound bridges. Curve 1: No
absorption in the cavity, curves 2 and 3: attenuated with 50 mm glass wool
and 100 mm glass wool, respectively. Upper and lower limits are indicated.
(Adapted from Sharp, B.H., Noise Control Eng., 11, 53–63, 1978.)
 Airborne sound transmission through double constructions  253

where
α is the sound absorption coefficient of the boundary in the cavity
d is the depth of the cavity
U is the perimeter of the boundary
S is the area of the plates

In a cavity wall without frame-absorber, the absorption coefficient α will

be very small. However, if the depth of the cavity is less than approximately
100 mm, some absorption will occur because of the friction in the viscous
boundary layer along the internal surface of the plates. From empirically
derived values (Brekke, 1980), the following equivalent absorption coef-
ficient of an empty cavity can be used:

 0.5 for d ≤ 20 mm

α ≅  10 mm (9.11)
 for d > 20 mm
 d

80

70
it
r lim
pe
60 Up
Empty
Transmission loss, R (dB)

50

150 mm
40
50 mm
t
imi
er l
30 Low

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 9.5 Measured transmission losses for a similar wall as in Figure 9.4, but with
frame-absorber of glass wool (48 kg/m3). The thickness of the frame-absorber
is either 150 mm or 50 mm. (Adapted from Sharp, B.H., Noise Control Eng.,
11, 53–63, 1978.)
254  Sound Insulation in Buildings

The use of Equation 9.10 is restricted to the range between lower and upper
limit as indicated in Figure 9.3. In Figure 9.4, some examples of measured
transmission losses can be seen with and without sound absorbing material
in the cavity. Corresponding results with a frame-absorber can be seen in
Figure 9.5.

9.2 TRANSMISSION VIA SOUND BRIDGES

A sound bridge means a mechanical connection by which vibration energy

is transferred between two separated plates in a double construction. As
illustrated in Figure 9.1, partly a sound power P2b is radiated due to the
transmission via sound bridges and partly a sound power P2l is radiated due
to the transmission via the cavity. The sum of the radiated sound powers
goes into the transmission loss of the construction.

 P1 
R = 10 lg  = −10 lg 10− R /10 + 10− Rb /10  (9.12)
 P2 + P2b 

The transmission loss Rl refers to the corresponding construction without

sound bridges and Rb is the transmission loss when only the transmission
via sound bridges is considered. The latter can be expressed by the trans-
mission loss R(1+2) (Equation 9.7) for a single construction with the plates 1
and 2 close together plus a correction term ΔR m that depends on the type
of the sound bridge.

Rb = R(1+ 2) + ∆ Rm(9.13)

In the following, it is shown how ΔR m can be determined for different types

of sound bridges. For the time being, they are presumed completely stiff and,
thus, the velocities are locally the same on each side of the sound bridge,
i.e. v 1,0 = v 2,0 (Figure 9.6). It is further presumed that the mass of the sound
bridges is insignificant compared to the total mass of the construction.
As shown by Sharp (1978), the relationship between the average velocity
in plate 1 and the local velocity at the sound bridge is determined by the
impedance of the two plates (point or line impedances) Z1 and Z 2:

2
v12 Z1 + Z2
2
= (9.14)
v 2,0 Z1

This expresses that there is a reduced movement at a sound bridge because

of the mutual effect of the plates.
 Airborne sound transmission through double constructions  255

V1,0

V1

V2 V2,0

Figure 9.6 Transmission of bending waves from plate 1 to plate 2 through a sound bridge.
(Reproduced from Kristensen, J., Rindel, J.H., Bygningsakustik – Teori og praksis.
(Building acoustics – Theory and practice, in Danish). SBI-Anvisning 166. Danish
Building Research Institute, Hørsholm, Denmark, 1989. With permission.)

If the average velocity in plate 1 is compared to that in the equivalent

single construction (where plates 1 and 2 are assumed close together), a fair
approximation should be:

2
v12+ 2  m1  
= (9.15)
v12  m1 + m2 

As the radiation efficiency for the equivalent single construction is pre-

sumed to be σ1+2 ≈ 1, the following expression for ΔR m is obtained:

v12+ 2 ρ cS
∆ Rm = 10 lg
P2b
(9.16)
 v 2 ρ cS   m1 Z + Z2 
= 10 lg  2,0 + 20 lg  ⋅ 1
 P 2b

  (m + m
1 2) Z1 

Further development of this expression depends on the type of the sound

bridge (see the following).

9.2.1 Point connections
The point excitation of a plate results in both a near-field radiation and a
resonant radiation, which is caused by the natural vibrations in plate 2. For
simplicity, the resonant radiation is disregarded here, but it will be dealt
with later. The sound power radiated from a number of equal excitation
points Np is (Cremer and Heckl, 1967, p. 471):

8c 2
P2b ≅ N p P2n = ρ cv2,0
2
N p (9.17)
π 3 fc22
256  Sound Insulation in Buildings

The point impedance for a plate is (Equation 2.13)

F 4c 2m
Zo = = 8 mB = (9.18)
vo π fc

Applying this for plates 1 and 2 (and inserted in Equation 9.16) yields:

 Sπ 3 fc22   m1 ( m1 /fc1 + m2 /fc 2 ) 

∆Rm, p ≅ 10 lg  2
+ 20 lg   (9.19)
 N p 8c   ( m1 + m2 ) ( m1 /fc1 ) 

Here, the last term can be combined with fc2 from the first term to yield an
auxiliary quantity fcp, which is a combination of the critical frequencies of
the two plates applying particularly for point connections:

m1fc 2 + m2 fc1
fcp = (9.20)
m1 + m2

and thus:

 Sπ 3 fcp2 
∆Rm, p ≅ 10 lg  2 
(9.21)
 N p 8c 

The principal course of the transmission loss as a function of frequency is

seen in Figure 9.7 for double constructions with sound bridges.
A high value of ΔR m can be obtained with a high critical frequency and
with the lowest possible number of connections per unit area. If the sound
bridges are placed in a quadratic pattern with a centre-to-centre distance
e = √S/Np, Equation 9.21 can be further simplified:

∆ Rm,p ≅ 20 lg ( e ⋅ fcp ) − 45 dB  (9.22)

From Equation 9.20, it is seen that for asymmetric double constructions,

the critical frequency of the lighter of the two plates is decisive for the size
of fcp. The direction of the transmission is of no importance. The simple for-
mula Equation 9.22 is illustrated in Figure 9.8. The advantage of choosing
a plate with a high critical frequency is obvious, if it is otherwise acceptable
to use plates with a low bending stiffness.
It is characteristic that ΔR m is not dependent on the frequency and the
transmission loss will, thus, go parallel with the transmission loss for the
corresponding single construction, i.e. typically with a slope of 6 dB/octave.
R(1+2) is the lower bound of the transmission loss; thus:

∆ Rm, p ≥ 0 dB(9.23)
 Airborne sound transmission through double constructions  257

80

70
fd it
lim
p er
60 Up
Transmission loss, R (dB)

50
ΔRm
f0
40

mit
er li
30 Low

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 9.7 Transmission loss for a double construction with rigid sound bridges.

In the above expression for point connections, it is an important assump-

tion that the frequency is below the critical frequency for the radiating
plate. Otherwise, the resonant radiation will result in ΔR m,p ≈ 0 dB.

9.2.2 Line connections
The sound transmission via line connections can be discussed analogously
to point connections. Instead of the point impedance, the line impedance is
used for each of the plates where the expression for a beam with excitation
in a point can be found (Cremer and Heckl, 1967, p. 273):

f (9.24)
Z = 2(1 + j)mcL 
fc

where Ll is the length of the line connection. As shown in the work of

Cremer and Heckl (1967, p. 481), the radiated sound power is:

2c
P2b ≅ P2n = ρ cv 2,0
2
L (9.25)
π fc 2
258  Sound Insulation in Buildings

30

Hz
000
=10
Correction for sound bridges ∆R (dB ) f c,p
Hz
00
20 50

Hz
00
20

Hz
00
10
10
z
0H
50

z
0H
20
0
0.1 0.2 0.5 1 2
Distance between connections e (m)

Figure 9.8 Influence of rigid point connections on the transmission loss. The connections
are assumed to have a square pattern with the centre distance e between the
points. The equivalent critical frequency fcp can normally be approximated by
the critical frequency of the thinner of the plates in the double construction.

disregarding the contribution from the resonant radiation. By inserting in

Equation 9.16 and by means of the auxiliary quantity

2
 m1 fc 2 + m2 fc1 
fc  =   (9.26)
 m1 + m2 

the following result applies for the sound transmission through a stiff line
connection:

 S π fc  
∆Rm, ≅ 10 lg  (9.27)
 Lt 2c 

The equivalent critical frequency fcl can normally be approximated by the

critical frequency of the thinner of the plates in the double construction. If
the sound bridges are located as parallel laths with a centre-to-centre dis-
tance b = S/Ll, the simple formula is:

∆Rm, ≅ 10 lg (bfc ) − 23 dB (9.28)

When comparing this with the result for point connections (Equation 9.22),
a slightly lower dependency of the critical frequency (3 dB/octave instead
 Airborne sound transmission through double constructions  259

of 6 dB/octave) can be observed. Thus, it is more difficult to achieve high

values of ΔR m with line connections than it is with point connections, as is
evident from Figures 9.8 and 9.9.
9.2.3 Boundary connections
A special variant of linear sound bridges are connections along the bound-
ary of a double construction. Compared to the line connections, the line
impedance for the boundaries is smaller by a factor of 4 (Cremer and Heckl,
1967, p. 273).

1 f
Zr = (1 + j)mcLr  (9.29)
2 fc

where Lr is the length of the boundary connection. From a simplified con-

sideration, the radiated sound power becomes one-half of that from a line
excitation at the middle of the plate. As in the previously described cases,
the resonant sound radiation is disregarded so far. The sound power radi-
ated from the boundary excitation is:

c
P2b ≅ P2n = ρ cv 2,0
2
Lr (9.30)
π fc 2

30
Correction for sound bridges ∆R (dB)

20
z
0H
1000
f c,l = z
0H
500
z
0H
200
10 z
0H
100
Hz
500
Hz
200

0
0.1 0.2 0.5 1 2 5
Distance between connections b (m)

Figure 9.9 Influence of rigid line connections on the transmission loss. The connections
are assumed to have the centre distance b between the lines. The equivalent
critical frequency fcl can normally be approximated by the critical frequency
of the thinner of the plates in the double construction.
260  Sound Insulation in Buildings

For a stiff boundary connection:

 S π fc  
∆Rm, r ≅ 10 lg  (9.31)
 Lr c 

where fcl is the previously introduced auxiliary quantity (Equation 9.26).

By comparison with Equation 9.27 we see that the result for ΔR m,r differs
by 3 dB from an ordinary line connection of the same length. The ΔRm,r
correction is displayed in Figure 9.10 for the case of boundary connection
all around the circumference.

9.2.4 Resonant sound radiation

The power that is provided to plate 2 partly causes a near-field radiation
from the affected area with the power P2n , and partly causes the build-up
of normal modes in the plate, and thus to a resonant sound radiation with
the power P2r. As the velocity in the excitation point is v 2,0, and the veloc-
ity of the normal mode vibrations is v 2 , an energy balance equation can be
set up for plate 2. The input power is determined from the impedance by

30
Correction for sound bridges ∆R (dB)

z
0000 H
f c,l =1
20 z
0H
500
z
0H
200
z
0H
100
Hz
10 500
Hz
200

0
0.1 0.2 0.5 1 2 5
Area to circumference ratio S/U (m)

Figure 9.10 Influence of rigid boundary connections on the transmission loss. The

abscissa is the ratio between the plate area S and the total length of the
boundary connection U. The equivalent critical frequency fcl can normally
be approximated by the critical frequency of the thinner of the plates in the
double construction.
 Airborne sound transmission through double constructions  261

Equation 2.57 and the dissipated power is expressed through the loss factor
by Equation 7.3:

2
v 2,0 Re {Z2 } = 2π fm2S v 22 η2 (9.32)

where η2 is the total loss factor for plate 2. The resonant radiated sound
power is:
2
v 2,0 Re {Z2 } ρ cσ 2
2
P2r = v 2,0 ρ cSσ 2 = (9.33)
2π fm2η2
The near-field radiation depends on the type of excitation; please refer to
the overview given in Table 9.1. For a point connection, the near-field radi-
ation (cf. Equation 9.17) is:

8c 2
P2n = ρ cv 2,0
2
(9.34)
π 3 fc22

The totally radiated sound power for a single point connection is:
P2b = P2n + P2r = P2nκ p(9.35)
where κp is inserted as a factor for the resonant radiation. For point connec-
tions, this becomes:

πσ 2 fc 2
κ p = 1+ (9.36)
4η2 f

Table 9.1  S ome quantities related to the excitation of a plate by point, line or
boundary connections
Point Line Boundary

Type of excitation

4c 2 m f f
Impedance Z0 2(1 + j)mcL 1
2 (1 + j)mcLr
π fc fc fc

π fc 1− j fc 1− j fc
Admittance Y0
4c 2 m 4mcL f mcLr f

Near-field 8c 2 2c c
ρ cv02 ⋅ Np ρ cv02 ⋅ L ρ cv02 ⋅ Lr
radiation P2n π 3 fc2 π fc π fc

Resonant radiation πσ fc σ fc σ fc
1+ 1+ 1+
factor κ 4η f 2η f 4η f
262  Sound Insulation in Buildings

Similarly, for line and boundary connections:

σ2 fc 2
κ = 1+  (9.37)
2η2 f

σ2 fc 2
κr = 1+ (9.38)
4η2 f

The significance of the resonant sound radiation will be that ΔR m is reduced

when the frequency approaches fc2 concurrently with an increasing value
of the radiation efficiency. For f > fcr the radiation efficiency is σ 2 ≈ 1 and
κ >> 1, which means that ΔR m becomes infinitesimal.

9.2.5 Elastic connections
Two plates connected with a sound bridge as shown in Figure 9.6 are con-
sidered. The mechanical connection can be of point or line shape, and it is
presumed to be elastic. The sound bridge is characterized by the mechani-
cal impedance Zb:

F k
Zb = ≅ d (9.39)
v1,0 − v2,0 jω

Here, F is the force whereby the sound bridge affects the plates, v 1,0 and v 2,0
are the velocities of the two plates at the point of the sound bridge. Zb will
primarily depend on the dynamic stiffness kd of the sound bridge, whereas
the internal losses can often be disregarded. However, the latter does not
apply to sound bridges of viscous elastic materials.
A bending wave propagating in plate 1 with a velocity amplitude v 1 will
get a slightly reduced velocity v 1,0 opposite the sound bridge:

v1,0 = v1 − Y1F (9.40)

where Y1 is the mechanical admittance of the plate, i.e. the reciprocal of the
impedance Z1.
In plate 2, the velocity opposite the sound bridge will be determined by
the admittance Y2 of the plate.

v2,0 = Y2F (9.41)

A new bending wave will consequently be generated in plate 2.

As shown by Ver (1971), the power P12 transferred from plate 1 to plate 2
can be determined:
2
Zb Y2 
P12 = v12 (9.42)
1 + 2 Re {Zb }(Y1 + Y2 ) + Zb (Y1 + Y2 )2
2
 Airborne sound transmission through double constructions  263

For a completely stiff sound bridge, the result is:

Y2
P12 = v12 for Zb → ∞ (9.43)
(Y1 + Y2 )2
The effect of an elastic sound bridge will, thus, be described by a coupling
factor γ, which is the ratio between the transferred power via an elastic
sound bridge and a totally stiff sound bridge, respectively.

γ =
(Y1 + Y2 )2
Zb
2

2 (9.44)
1 + 2 Re {Zb }(Y1 + Y2 ) + Zb (Y1 + Y2 )
2

−1
  2π f  
2

γ ≅ 1 +   (9.45)
  kd (Y1 + Y2 )  

In the latter formula, the internal losses of the sound bridge are disregarded,
as Equation 9.39 has been inserted.

9.2.6 Coupling factor
In the following, it is shown how the coupling factor γ can be determined
for three types of elastic sound bridges, point connections, line connections
and boundary connections. Indices p, l and r are used, respectively.
The calculations are based on Equation 9.45, and the dynamic stiffness
of the sound bridge is considered, but the internal losses are not. By insert-
ing the admittances as given in Table 9.1, it is found that the coupling fac-
tors can be written as:

(
γ p = 1 + ( f fkp ) )
2 −1
(9.46)

(
γ  = 1 + ( f fk ) )
3 −1
(9.47)

( 3 −1
γ r = 1 + ( f fkr ) ) (9.48)

Here, some knee frequencies are inserted, below which the sound bridges
behave as stiff connections:

fcp
fkp = kdp (9.49)
8c 2mm
264  Sound Insulation in Buildings

2/3
1 fc  
fk =  kd′  (9.50)
2 2π cmm 

2/3
1 2 fc  
fkr =  kdr
′ (9.51)
2 π cmm 

The dynamic stiffness of the point connection is kdp and k′dl and k′dr are
the dynamic stiffness’s per unit length for line connection and boundary
­connection, respectively. In these expressions, some auxiliary terms are
used, fcp (Equation 9.20) and fcl (Equation 9.26) and an equivalent mass per
unit area mm , which is determined by the mass per unit area of each of the
two plates:

m1m2
mm = (9.52)
m1 + m2

9.2.7 Summary on sound bridges

The coupling factors and factors for resonant sound radiation that has been
introduced above should be used with the expressions for radiated sound
power from the near field for rigid sound bridges. Contributions from dif-
ferent types of sound bridges can be combined in a single expression that
includes point connections, line connections and boundary connections:

 8c 2 2c 
∆Rm = −10 lg  3 2 N pγ pκ p + ( Lγ κ  + 12 Lrγ rκ r )  (9.53)
 Sπ fcp Sπ fc 

When this quantity has been determined, the transmission loss is found
from Equations 9.13 and 9.12. Examples of measured transmission loss
for light double walls with different types of sound bridges are shown in
Figure 9.11.
Heavy double constructions are especially sensitive to sound bridges,
as the critical frequency typically is rather low, which leads to a signifi-
cant resonant sound radiation. Figures 9.12 and 9.13 show examples of the
importance of sound bridges due to boundary coupling and point connec-
tions for double walls of clinker concrete.
Finally, Figure 9.14 shows an example of the use of secondary wall, i.e.
a double construction made by a massive wall and a light plate, which
reduces the sound radiation because of a high critical frequency. It is seen
that there is a good improvement of sound insulation even with rigid con-
nections, but the result is much better without sound bridges.
 Airborne sound transmission through double constructions  265

80

70

60
Transmission loss, R (dB)

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

Curve 1 Curve 2 Curve 3 Curve 4

Figure 9.11 Measured transmission loss for double walls of 13 mm gypsum board with
106 mm cavity depth. Wall 1: No attenuation in the cavity, no sound bridges;
wall 2: 50 mm of mineral wool in the cavity, no sound bridges; wall 3: rigid line
connections of wooden laths, b = 400 mm; wall 4: elastic line connections of
steel laths, b = 400 mm. (Adapted from Northwood, T.D., Transmission loss
of plasterboard walls. Build. Res. Note No. 66. National Research Council
of Canada, Ottawa, ON, 1970.)

9.3 SOUND INSULATION OF DRYWALLS

9.3.1 Influence of stud distance

In drywalls, e.g. made with gypsum boards, the stud distance can have
a major influence on the low-frequency sound insulation (Rindel and
Hoffmeyer, 1991). The studs have the function of stiffening the wall and,
thus, the plate cladding exhibits normal modes in the subareas between the
studs. Examples of measurement results from a laboratory test are shown
in Figure 9.15. The test opening was 2700 mm high and 3700 mm wide.
Thus, the standard stud distance of 600 mm could not be used throughout.
In the case where the first and the last sections were 350 mm, the measured
transmission loss was remarkably low in the range from 125 Hz to 160 Hz.
If instead the first stud distance is either 100 mm or 700 mm, the dip at low
frequencies is avoided and the overall result is much better. A difference of
18 dB is found at 160 Hz.
 266  Sound Insulation in Buildings

80

70 Leca, 150 mm
Plaster, 10 mm

60 Mineral wool, 50 mm
Leca, 100 mm
Transmission loss, R (dB)

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 9.12 Measured transmission loss for a single wall and a double wall with a rigid
boundary connection. Solid line: Single 150 mm clinker concrete wall with plas-
ter; dashed line: double wall, 150 mm and 100 mm clinker concrete on com-
mon concrete foundation. (Adapted from Vigran, T.E., Lydisolasjon i bygninger.
(Sound insulation in buildings, in Norwegian). Tapir, Trondheim, Norway, 1979.)

The results can be explained by the normal modes of the gypsum board
in the sections between the studs. The boundary condition also plays a
role here, and it may be something between simply supported and clamped
depending on the construction details. The natural frequency of the fun-
damental mode (m, n) = (1, 1) can be estimated from Equation 6.7 with an
extra boundary condition factor γ:

c2  1 2  1 2
f11 = γ    +    (9.53)
4 fc   lx   ly  

Here, c is the speed of sound in air, fc is the critical frequency and the
dimensions of the plate section are lx and ly. The factor γ is unity for sup-
ported plates but for a clamped boundary, it is a complicated function of
dimensions and mode number, as shown by Timmel (1991). Typical values
 Airborne sound transmission through double constructions  267

100

90

80

70
Transmission loss, R (dB)

60

50

40 Leca, 150 mm
Plaster, 10 mm
30 Mineral wool, 50 mm

Binder
20 Leca, 100 mm
Resilient filling

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 9.13 Measured transmission loss for a double wall of 150 mm and 100 mm clinker
concrete. Solid line: No boundary connections, Rw = 74 dB, dashed line:
With point connections, steel wire ties, e = 500 mm, Rw = 56 dB. (Adapted
from Vigran, T.E., Lydisolasjon i bygninger. (Sound insulation in buildings, in
Norwegian). Tapir, Trondheim, Norway, 1979.)

are in the range from 1.3 to 2, with the larger values for the fundamental
mode (1, 1).
For a gypsum board with critical frequency 2500 Hz and height 2700 mm,
the natural frequency f11 of the fundamental mode is calculated for some
examples of stud distances in Table 9.2. For the case of partly clamped
boundaries, the f 11 is very low at cc 600 mm and very high at cc 100 mm,
but at cc 350 mm, the frequency is exactly in the range from 125 Hz to 160
Hz where the severe dip in Figure 9.15 (solid line) is located. It is concluded
that stud distances around 300 mm to 400 mm should be avoided in this
type of wall.
268  Sound Insulation in Buildings

80

70

60
Transmission loss, R (dB)

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 9.14 Measured transmission loss for single and double walls. Solid line: 150 mm
clinker concrete wall with 10 mm plaster, Rw = 48 dB. Dashed line: Clinker
wall and a light secondary wall of 13 mm gypsum board with rigid wooden
line connections and 50 mm cavity with mineral wool, Rw = 54 dB. Dotted
line: The same, but secondary wall on steel lath only fixed in top and bottom,
75 mm cavity with mineral wool, Rw = 61 dB. (Adapted from Homb, A. et al.,
Lydisolerende konstruksjoner. (Sound insulation properties of windows, in
Norwegian). NBI-anvisning 28. Oslo, Norway, 1983, p. 61c.)

Table 9.2  Calculated natural frequency of fundamental mode (1, 1) for

sections of 13 mm gypsum board with height 2700 mm and different
stud distances
Stud distance lx (mm) 100 350 600
Simply supported, γ = 1 1185 Hz 98 Hz 34 Hz
Partly clamped, γ = 1.41 1671 Hz 138 Hz 49 Hz
Clamped, γ = 2 2370 Hz 196 Hz 69 Hz
Note: Three different boundary conditions are assumed.

9.3.2 Examples
Dry walls can be designed within a wide range of sound insulation depend-
ing on the number of gypsum board layers, the cavity depth and whether the
studs are common to both sides or separated. Table 9.3 gives some examples
 Airborne sound transmission through double constructions  269

80

70

60
Transmission loss, R (dB)

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 9.15 Measured transmission loss of a drywall with different stud distances at

the border region. Solid line: Stud distances 350 mm/5 × 600 mm/350 mm;
dashed line: stud distances 100 mm/6 × 600 mm; dotted line: Stud distances
700 mm/5 × 600 mm. (Adapted from Rindel, J.H., Hoffmeyer, D., Proceedings
of Inter-Noise '91, Sydney, Australia, pp. 279–282, 1991.)

and the measured Rw values spread from 38 dB to 70 dB. It is important to

note that the field performance in buildings is typically 3 dB to 10 dB lower
than when the same wall is measured in a laboratory. The low-frequency
performance is generally rather weak for lightweight double walls and, thus,
the spectrum adaptation term C 50–5000 for these walls is typically −4 dB to
−6 dB. The spectrum adaptation term also applies to the apparent sound
reduction index R′w.

9.4 SOUND INSULATION OF WINDOWS

9.4.1 Gas filling in windows
Windows are often made as units with two or three layers of glass –
especially in countries with a cold climate – for the reason of thermal
insulation. Sometimes, a special gas is used in the sealed cavity of the
270  Sound Insulation in Buildings

Table 9.3  E
 xpected transmission loss of drywalls with 13 mm gypsum boards, in
the laboratory (Rw and C 50–5000) and in the field (R′w)
Mineral
Wall Number Cavity wool
thickness of plate depth thickness Rw C50–5000 R′w
Type (mm) layers (mm) (mm) (dB) (dB) (dB)
Common 100 1 + 1 75 0 38–41 −1 ~35
studs 100 1 + 1 75 45 42–44 −3 ~40
125 2 + 2 75 45 48–50 −4 ~46
150 2 + 2 100 45 51–53 −4 ~47
175 2 + 2 125 90 56–58 −6 ~52
Separate 150 2 + 2 100 100 56–58 −5 ~51
studs 200 2 + 2 150 100 61–63 −6 ~54
275 1 + 1 250 250 60 −6 ~55
425 1 + 1 400 400 70 −4 ~60
Source: Byggforsk 524.325, Lydisolasjonsegenskaper til lette innervegger. (Sound insulation
properties of lightweight internal walls, in Norwegian). Byggforskserien, SINTEF
Building and Infrastructure, Oslo, Norway, 2000.

window unit as this can improve the thermal insulation. Examples of

measured sound insulation with different gasses in the cavity are shown
in Figure 9.16. The windows have one 8 mm and one 4 mm glass with a
distance of 12 mm. The gases are sulphur hexafluoride (SF6), argon (Ar)
and krypton (Kr). A window with normal air was also measured for com-
parison. The results show little difference between the transmission loss
using argon, krypton, or normal air. However, using SF6, the transmission
loss changes ­significantly, which must be due to a high density (6.17 kg/
m3) and a low speed of sound (134 m/s). The mass-spring-mass resonance
frequency is lowered and the dip is more pronounced, followed by a very
steep increase above the resonance frequency. The other gases are also
heavier than air and have a lower speed of sound, but the differences from
normal air are smaller.

9.4.2 Examples
In Table 9.4, some examples of measured sound insulation of windows with
two and three layers of glass are given. The constructions are explained by
glass thickness in mm and distance between panes in mm. Laminated glass
is denoted with a slash (/), so, for example, 4/1.14/4 means two panes of
4 mm glass with an interlayer of 1.14 mm.
For evaluation of the sound insulation against road traffic noise, it is
necessary to include the spectrum adaptation term C tr. The data used for
the Table 9.4 did not include frequencies below 100 Hz, but if available, the
 Airborne sound transmission through double constructions  271

80

70

60
Transmission loss, R (dB)

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

SF6 Argon Krypton Air

Figure 9.16 Measured transmission loss of sealed 8 – 12 – 4 window units with different

gasses in the cavity. (Adapted from Rasmussen, B., Optimering af vinduers
lydisolation. Karm/rammekonstruktioner. (Optimising the sound insulation of
windows. Frame/sash constructions, in Danish). Danish Acoustical Institute,
Denmark, 1990, p. A35.)

extended frequency range down to 50 Hz should be used for evaluation of

the performance of the windows in practice.
Some rough rules of thumb can be extracted from the examples in
Table 9.4:

• Asymmetry is an advantage. Better sound insulation is obtained in

two-layer units if the glass panes have different thickness.
• Better sound insulation is obtained if laminated glass is used, at least
in one of the panes.
• Three-layer units do not give better sound insulation than two layer
units with the same total thickness. Although asymmetry in the two
cavities is an advantage, the middle glass pane does not contribute
noticeably to the sound insulation.
272  Sound Insulation in Buildings

Table 9.4  E
 xamples of two- and three-layer windows, total thickness, expected
weighted sound reduction index (Rw) and practical sound insulation
against road traffic noise (Rw + C tr)
Total
thickness Rw + Ctr
Type (mm) Construction (mm) Rw (dB) (dB)
Two-layer 20 4 – 12 – 4 34 30
units 22 6 – 12 – 4 36 33
24 8 – 12 – 4 37 34
26 10 – 12 – 4 38 34
Two-layer 25 4 – 12 – 4/1.14/4 37 33
units with 35 6 – 20 – 4/0.76/4 41 36
laminated
glass 37 8 – 20 – 4/0.76/4 43 39
41 6/0.38/6 – 20 – 4/0.38/4 43 40
Three-layer 24 4–6–4–6–4 33 28
units 30 4–9–4–9–4 34 29
32 6 – 12 – 4 – 6 – 4 39 33
28 8–6–4–6–4 38 34
48 8 – 20 – 4 – 12 – 4 40 35
Three-layer 46 5 – 16 – 4 – 12 – 4/1.14/4 38 35
units with 51 8 – 18 – 4 – 12 – 4/0.2/4 43 39
laminated
glass 54 6/0.2/6 – 20 – 5 – 9 – 4/0.2/4 41 40

Single glass 67 5 – 42 – (4 – 12 – 4) 38 31
and 89 5 – 63 – (4 – 9 – 8) 40 35
two-layer
unit in 111 6 – 85 – (4 – 12 – 4) 43 40
separate 111 5 – 80 – (4 – 15 – 3/0.76/3) 45 38
sashes
Source: Byggforsk 533.109. Lydisolerende egenskaper for vinduer. (Sound insulation proper-
ties of windows, in Norwegian). Byggforskserien, SINTEF Building and Infrastructure,
Oslo, Norway, 2013.

REFERENCES

A. Brekke (1980). Beregningsmetoder for reduksjonstall til enkelt-, dobbelt- og

­trippelkonstruksjoner. (Calculation methods for reduction index of single-,
double- and tripple constructions, in Norwegian). Report STF 44 A 80021,
ELAB, NTH, Trondheim, Norway.
Byggforsk 524.325 (2000). Lydisolasjonsegenskaper til lette innervegger. (Sound insu-
lation properties of lightweight internal walls, in Norwegian). Byggforskserien,
SINTEF Building and Infrastructure, Oslo, Norway.
 Airborne sound transmission through double constructions  273

Byggforsk 533.109 (2013). Lydisolerende egenskaper for vinduer. (Sound insulation

properties of windows, in Norwegian). Byggforskserien, SINTEF Building and
Infrastructure, Oslo, Norway.
L. Cremer and H. Heckl (1967). Körpershall. (Structure-Borne Sound, Translated by
E.E. Ungar 1973). Springer, Berlin, Germany.
A. Homb, S. Hveem, S. Strøm (1983). Lydisolerende konstruksjoner. (Sound insulat-
ing constructions, in Norwegian). NBI-anvisning 28. Oslo, Norway.
J. Kristensen and J.H. Rindel (1989). Bygningsakustik – Teorti og praksis. (Building
acoustics – Theory and practice, in Danish). SBI Anvisning 166. Danish Building
Research Institute, Aalborg University Copenhagen, Hørsholm, Denmark.
T.D. Northwood (1970). Transmission loss of plasterboard walls. Build. Res. Note
No. 66. National Research Council of Canada, Ottawa, ON.
B. Rasmussen (1990). Optimering af vinduers lydisolation. Karm/rammekonstruk-
tioner. (Optimising the sound insulation of windows. Frame/sash construc-
tions, in Danish). Danish Acoustical Institute, Lyngby, Denmark.
J.H. Rindel and D. Hoffmeyer (1991). Influence of stud distance on sound insulation
of gypsum board walls. Proceedings of Inter-Noise '91, 2–4 December 1991,
Sydney, Australia, pp. 279–282.
B.H. Sharp (1978). Prediction methods for the sound transmission of building ele-
ments. Noise Control Engineering 11, 53–63.
R. Timmel (1991). Der Abstrahlgrad rechteckiger, dünner, homogener Platten in der
unendlich grossen Schallwand. (The radiation efficiency of rectangular, thin,
homogeneous plates in the infinitely large acoustic wall, in German). Acustica
73, 1–11.
I.L. Ver (1971). Impact noise insulation of composite floors. Journal of the Acoustical
Society of America 50, 1043–1050.
T.E. Vigran (1979). Lydisolasjon i bygninger. (Sound insulation in buildings, in
Norwegian). Tapir, Trondheim, Norway.
Chapter 10

Impact sound insulation

The impact sound pressure levels that can be obtained with homogeneous
massive plates can be described theoretically by the sound radiation from
point excitation of plates. However, supplementary attenuation of the
impact sound pressure level is normally necessary. This can be done by
soft resilient floor coverings or by elastically layered floating floors, which
can be considered double constructions. A suspended ceiling can also be
relevant in some cases.
The impact sound is the most important cause of complaints from the
dwellers in multi-storey houses, and the low frequencies below 100 Hz call
for special attention in the acoustic design.

10.1 THE IMPACT SOUND PRESSURE LEVEL

The impact sound pressure level L n is a measurement for the sound trans-
mission into a room when the floor in another room is mechanically excited
by a standardized tapping machine:

A2
Ln = L2 + 10 lg (dB) (10.1)
A0

where L 2 is the mean sound pressure level in dB re 20 μPa in the receiving

room. A 2 is the equivalent sound absorption area of the receiving room,
and A 0 = 10 m 2 is a standardized reference area. For theoretical calcula-
tions, Equation 10.1 can be rewritten by assuming the sound field in the
receiving room to be diffuse:

p 22 A2 4 ρcP2
Ln = 10 lg = 10 lg (dB) (10.2)
p02 A0 p02 A0

where p 22 is the mean sound pressure square in the receiving room, and P2 is the
sound power radiated into the receiving room. The formula in Equation 10.1 is

275
276  Sound Insulation in Buildings

the basis of measuring the impact sound pressure level in the laboratory as well
as in buildings. However, the symbol L′n is used in buildings in order to indicate
that the result may be influenced by flanking transmission.

10.2  MASSIVE FLOORS

10.2.1  Point excitation of a plate

The tapping machine is standardized in Annex E of ISO 10140–5 and used
for measuring impact sound pressure levels. The tapping machine has five
identical steel hammers, the mass of each is mh = 500 g. The hammers sup-
ply a power corresponding to a free fall of height h = 40 mm and they deliver
ten impacts per second, i.e. the tapping frequency is ftap = 10 Hz.
Under the assumption of a very heavy and stiff floor like a concrete slab,
the duration of the force impulse is very small and the mean-square-force
in a frequency band Δf ≫ 10 Hz can be shown to (e.g. Beranek and Ver,
1992, pp. 328–329) be:

F 2 = 4 ftap mh2 gh∆ f ≅ 3.9∆ f (N 2 ) (10.3)

The approximation comes by inserting the data from the tapping machine
and the acceleration of gravity, g = 9.81 m/s2. For less stiff structures like
wooden floors, the relation becomes more complicated and (Equation 10.3)
does not hold. For one-third octave bands, the relative bandwidth is 0.23
and the mean-square-force of the tapping machine on a heavy and stiff floor
is:

F 2 ≅ 0.91f (N 2 ) (10.4)

where f is the centre frequency in Hz.

The relationship between the force F in a point and the velocity v 0 in
the same point is called the point impedance, Z 0 . For an infinite, homo-
geneous plate that is excited to bending waves, i.e. the thickness less than
approximately one-sixth of the bending wave length, the point impedance
is (Equation 2.13):

F 4c 2m
Z0 = ≅ 8 mB = (10.5)
v0 π fc

The point impedance is in general a complex quantity, but in this case it

takes a real value, i.e. the force and the velocity are in phase. For a finite
plate, the point impedance varies from one point to another. However, it
can be shown that the mean value of the point impedance is equal to that
of the infinite plate. In a given point, the deviation from the mean value
 Impact sound insulation  277

depends on the position of the point relative to the modal pattern of the
plate.
A very important relationship exists between the modal density ΔN/Δf
of a finite plate and the point impedance of the equivalent infinite plate. If
m is the mass per unit area of the plate and S is the area, the real part of
the point admittance (the reciprocal of the impedance) (see Cremer et al.,
1988) is:

 1  1 ∆N
Re   = ⋅ (10.6)
Z
 0 4mS ∆f
The modal density of a homogeneous, isotropic plate depends on the thick-
ness. If the plate is thin compared to the wavelength, bending waves can
be assumed, but if the plate is thick, compared to the “wavelength” shear
waves will occur instead. The approximate modal density is:

 π
 Sfc (f < f s / 2) (bending)
∆N  c 2 (10.7)
≅
∆f  2π Sf (f > f s / 2) (shear)
 c s2
         
where fc is the critical frequency, fs = fc (cs /c) is the crossover frequency for
2

shear waves, and c s is the speed of shear waves (see Equation 1.30). Both
the general results in terms of the modal density as well as the bending wave
approximation will be stated in the following.
If φ is the phase angle between force and velocity in the excitation point,
the power input to the plate by the point force is (Equation 2.55):
 1 
Pi = 1
2 F ⋅ v0 ⋅ cos ϕ = v 02 Re {Z0 } = F 2 Re  
 Z0 

F 2 ∆N F 2π fc (10.8)
= ⋅ ≅ 2
4mS ∆f 4c m

In steady state, the power Pd that is lost by internal losses, boundary losses
and radiation losses is expressed by the total loss factor η (using Equations
7.1 and 7.3):

Pd = 2π fmS v r2 η (10.9)

At steady state, we have energy balance, Pi = Pd. Thus, the relationship

between the average velocity and the input force can be determined:

F 2 ∆f f
= 8π fm 2S 2η ≅ 8Sm 2c 2η (10.10)
2
v r ∆N fc

278  Sound Insulation in Buildings

The sound power P2 radiated into the receiving room depends on the radia-
tion efficiency σres and the average velocity (see Equation 6.24). From this
and Equation 10.10, the radiated sound power can be found as a function
of the input force F:

F 2 ρ cσ res ∆N F 2 ρσ res fc
P2 = ⋅ ≅ (10.11)
8π Sm 2η f ∆f 8m 2cη f

The sound power radiated from the near field around the excitation point
may also be included using the point impedance (Equation 10.5) together
with Equation 9.17, which leads to:

F 2 ρ F 2 ρ cσ res ∆N
P2 = + ⋅ (10.12)
2π m 2c 8π Sm 2η f ∆f

The first term is due to the near field radiation and the second term is the
resonant contribution from the natural modes in the plate.

10.2.2 Impact sound pressure level

for massive floors
For massive floors, the impact of sound pressure level can be determined
from Equation 10.2 by inserting the radiated power (Equation 10.12):

 4 ρ 2 F 2  1 c 2σ res ∆N  
Ln = 10 lg  2 2 
+ ⋅  (10.13)
 p0 A0 m  2π 8π fSη ∆f  

Insert the numerical constants and the force of the standardized tapping
machine per one-third octave, and the impact sound pressure level per one-
third octave is:

4 c 2σ res ∆N 
Ln ≅ 82 − 20 lg m + 10 lg  f + ⋅ (10.14)
π π Sη ∆f 
where m shall have the unit kg/m 2 . The last term represents the near field
radiation plus the resonant radiation from the natural modes. When the
loss factor η is small, the contribution of the near field is often insignificant
compared to the resonant radiation. For massive floors with a low critical
frequency, the radiation efficiency is σres ≈ 1 because f > fc. Assuming bend-
ing waves, an approximation for the impact sound pressure level is:

 η
Ln ≅ 82 − 10 lg  m 2  ( fc < f < fs / 2) (10.15)
 fc 

 Impact sound insulation  279

Above the frequency fs /2, the vibrations in the plate propagate as shear
waves and the approximation changes to:

 η  2f 
Ln ≅ 82 − 10 lg  m 2  + 10 lg   (f > fs / 2) (10.16)
 fc   fs 

If the frequency dependency of the loss factor is disregarded, these approxi-
mations show that the impact sound pressure level for a massive floor is
independent of the frequency in the bending wave range, but increases with
frequency at higher frequencies where the shear waves take over.
The influence of the plate thickness is seen from the fact that m is pro-
portional to the thickness and fc and fs are inversely proportional to the
thickness. The impact sound pressure level L n decreases 9 dB per doubling
of the plate thickness in the bending wave range, but decreases 12 dB per
doubling of the plate thickness in the shear wave range. For a rough esti-
mate, the relationship is shown for concrete and lightweight concrete in
Figure 10.1. However, it is not realistic to use so thick plates that the impact
sound pressure level is acceptable for dwellings without further attenuation
of the impact sound.

100
Impact sound pressure level per 1/3 octave (dB)

90

Por
80 ous
con
cret
e
70
Con
cret
e
60

50

40

30

20
50 100 200 500 1000
Plate thickness h (mm)

Figure 10.1 Approximate impact sound pressure level in one-third octave band for
massive floor of concrete (2300 kg/m3) or of porous concrete (600 kg/m3)
assuming bending waves, a radiation factor σ = 1, and a loss factor η = 0.05.
280  Sound Insulation in Buildings

10.3  RESILIENT FLOOR COVERINGS

With a carpet or some other thin flooring, e.g. linoleum with soft underlay,
a reduction of the hammer force of the floor is obtained by decreasing the
speed gradually from maximum at the surface to zero when a certain pres-
sure is reached in the flooring, after which the hammer returns to the sur-
face level. The course corresponds to half a cycle for a mass-spring system
where the spring constant is determined by the modulus of elasticity of the
flooring, its thickness ht and the actuated area Sh (Figure 10.2).
If the mass of the hammer is called mh, the resonance frequency of the
system can be determined by:

1 Sh Et (10.17)
f0 =
2π mh ht

For the standardized tapping machine, mh = 0.500 kg and Sh = 700 mm 2 .
The diagram in Figure 10.3 shows a relationship between the resonance
frequency f 0 and the elastic properties of the flooring. For many elastic
materials, the dynamic modulus of elasticity is about double that deter-
mined by a static load.
As shown by Ver (1971), if the resonance frequency f 0 is known, and
f > f 0 the attenuation of the impact sound pressure level can be determined
approximately for f > f 0:
f
∆L = Ln,without − Ln,with ≅ 40 lg (10.18)
f0
For f ≤ f 0, the attenuation is ΔL ≅ 0 dB. The idealized frequency course
of the impact sound attenuation is illustrated in Figure 10.4. For a good
design, f 0 should be as low as possible and preferably under 90 Hz. In
Figure 10.5, some examples of measured impact sound attenuation of thin
flooring can be seen.

mh Sh Et

ht

Figure 10.2 Resilient floor covering with thickness ht and Young’s modulus Et. The ham-
mer of the tapping machine has the mass mh = 500 g and the area Sh = 700 mm2 .
(Reproduced from Kristensen, J., Rindel, J. H., Bygningsakustik – Teori og
praksis, SBI-Anvisning 166 (In Danish), Danish Building Research Institute,
Aalborg University Copenhagen, 1989. With permission.)
 Impact sound insulation  281

1.E+11

1.E+10
Et/ht (N/m3)

1.E+09

1.E+08
100 200 500 1000
Resonance frequency f0 (Hz)

Figure 10.3 Diagram for estimation of the resonance frequency f0 for a soft floor cover-
ing with thickness ht and Young’s modulus Et.

40

30

20
ΔL, dB

10

–10
0.5 1 2 5 10
Relative frequency f/f0

Figure 10.4 Idealized frequency course of the impact sound attenuation of a soft floor
covering.
282  Sound Insulation in Buildings

70

60

50
Improvement, ΔLn (dB)

40

30

20

10

–10
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 10.5 Examples of impact sound attenuation for floor coverings. Solid line: Thin
linoleum. Dashed line: Vinyl on a layer of felt. Dashed–dotted line: Thick
and soft carpet.

10.4  THE RELATIONSHIP BETWEEN R AND L n

As regards massive floors, there exists a relationship between the airborne

sound insulation and the impact sound pressure level. This relationship is
surprisingly simple when the sound transmission is dominated by resonant
transmission, i.e. for frequencies above the critical frequency (f > fc). If soft
flooring is assumed and only resonant radiation is considered (the near field
radiation is neglected), we have from Equation 10.14:

 c 2σ res ∆N 
Ln ≅ 82 − 20 lg m + 10 lg  ⋅ − ∆L (10.19)
 π Sη ∆f 

Here ΔL is the impact sound attenuation of the flooring. The transmission

loss R is irrespective of the flooring given by (see Equation 7.23):

2
 ω m  c 2σ res
2
∆N 
R ≅ Rr = 10 lg   − 10 lg  ⋅
 2ρ c   2f S η ∆f 

 c 2σ res
2
∆N 
( )
= 10 lg m 2 f 3 − 10 lg  ⋅
 π S η ∆f 
− 44 (10.20)

 Impact sound insulation  283

Adding R and L n yields:

R + Ln = 38 + 30 lg f − 10 lg σ res − ∆L (10.21)

The radiation efficiency is approximately σres ≅ 1 since we have assumed

f > fc. Thus, for floors with a hard surface, the relationship is as simple as:

R + Ln = 38 + 30 lg f (10.22)

With a soft flooring or a floating floor that is locally reacting, the relation-
ship changes above the resonance frequency f 0 of the flooring:

f
R + Ln = 38 + 30 lg f0 − 10 lg
f0
(f > f0 ) (10.23)

The relationship is illustrated in Figure 10.6 when applied to typical

­ easurement results for heavy slabs in Danish domestic houses. The
m
lowest structural mode f11 is clearly seen at a frequency around 80 Hz.
However, the impact sound pressure mirror the course at frequencies
below f11, and the sum shown in Figure 10.6b is close to a straight line
below f 0 with the slope 9 dB per octave. With soft flooring or a floating
floor, the curve has a knee at f 0 and continues towards higher frequencies
with the negative slope −3 dB per octave. In the example, f 0 is around
250 Hz.
This simple relationship can be utilized for control of measuring results
as will be demonstrated in Chapter 11. If the measured airborne and impact
sound insulation added together fails to follow the predicted curve, this is
an indication of flanking transmission or leakage, i.e. some transmission
paths that only affect the airborne sound transmission and not the impact
sound transmission. It is emphasized that below the critical frequency of
the floor, the relationship between R and L n cannot be expected to be
equally simple.

10.5  FLOATING FLOORS

A special type of double construction is a floating floor where the elasticity

and construction of the inevitable mechanical connections are of impor-
tance for the sound insulating characteristics. As illustrated in Figure 10.7,
some airborne flanking transmission paths are not affected by a floating
floor, whereas the impact sound is effectively attenuated by a floating floor.
Thus, floating floors tend to improve the impact sound insulation more
than the airborne sound insulation.
284  Sound Insulation in Buildings

(a) 80
f0
70

60

50
R΄and L΄n (dB)

40

30

20

10
f11
0
20 50 100 200 500 1000 2000 5000
Frequency (Hz)
R΄ L΄n
(b) 140

e
av
130 oct
per
B
f0 9d
120
Sum R΄+ L΄n (dB )

110
–3 d
B pe
r oct
ave
100

90

80
20 50 100 200 500 1000 2000 5000
Frequency (Hz)
Heavy Theory Resilient floor

Figure 10.6 (a) Measured airborne and impact sound insulation of typical heavy floor
constructions. Average of 10 measurements. The frequency of the lowest
structural mode f11 and the resonance frequency of the flooring f 0 are indi-
cated. (Below 50 Hz the results are not corrected for reverberation time.)
(b) The sum R′ + L′n for a massive floor. Thick line: The sum of the measured
results in a. Thin straight line: Idealized frequency course for a hard floor.
Dashed line: Floor with impact sound attenuation by flooring. (Data from
DELTA, Målerapport: Boligers lydisolation ved lave frekvenser. (In Danish). PDØ
870028, DELTA, Aarhus, Denmark, 1999.)
 Impact sound insulation  285

Figure 10.7 Transmission of airborne and impact sound through floors with floating
floors. Due to flanking transmission, the attenuation of impact sound is bet-
ter than the attenuation of airborne sound. (Reproduced from Kristensen,
J., Rindel, J. H., Bygningsakustik – Teori og praksis, SBI-Anvisning 166 (In
Danish), Danish Building Research Institute, Hørsholm, Denmark, 1989.
With permission.)

In the following, the attenuation ΔL of the impact sound pressure level

is described. The airborne sound insulation for a construction with floating
floor may be evaluated by calculation of all possible transmission paths, as
described in Chapter 11 on flanking transmission.

10.5.1  Attenuation of impact sound pressure level

The excitation with the standardized tapping machine, as illustrated in
Figure 10.8, gives a well-defined pulse. However, the effect that is trans-
ferred to the construction will not only depend on the boundary impedance
of the construction but also on the impedance of the hammer. Instead of
Equation 10.8, the power input to the plate is more generally:

 1 
Pi = F 2 Re   (10.24)
Z
 0,1 + Zh 

where Z 0,1 is the point impedance of the floating floor slab. The impedance
Zh of the hammer is determined by its mass mh = 500 g:

Zh = jω mh (10.25)

The transferred power is reduced at high frequencies, where Zh > Z0,1 ,

i.e. for frequencies:

2m1 S1 ∆f
f > fZ = ⋅ (10.26)
π mh ∆N1
286  Sound Insulation in Buildings

Zh (mh)

F
Z1 (m1 , fc1)
d Zd (em, cm)

Z2 (m2 , fC2)

Figure 10.8 Floating floor with excitation from a falling mass. Each element is character-
ized by a mechanical impedance. (Reproduced from Kristensen, J., Rindel, J.
H., Bygningsakustik – Teori og praksis, SBI-Anvisning 166 (In Danish), Danish
Building Research Institute, Hørsholm, Denmark, 1989. With permission.)

Here, the point admittance has been inserted from Equation 10.6. For fre-
quencies higher than f Z , an extra contribution ΔL Z for the attenuation of
the impact sound pressure is obtained because of the impedance of the
hammer:

  f 2
∆LZ = 10 lg  1 +    (10.27)
  fZ  

For f < f Z , the attenuation is ΔLZ ≅ 0 dB. However, for lightweight floors this
attenuation has some importance. For example the frequency f Z is ≅ 700 Hz
for a 22 mm chip board and ≅ 200 Hz for a 12 mm chip board. Above f Z ,
the attenuation increases by 6 dB per octave in the case of a thin slab per-
forming bending waves, changing to 12 dB per octave in the case of a thick
slab performing shear waves. For measurements in the laboratory of impact
sound attenuation of floating floors, the measurement standard ISO 10140-1
(Annex H.4.1) prescribes an evenly distributed load of 20 kg/m 2 to 25 kg/m2 ,
which will diminish the problem of the hammer impedance.
The attenuation of the impact sound pressure level of a floating floor can
be expressed by the ratio of the velocities in the floor in the two situations
shown in Figure 10.9a and b:
v2 a
∆L = 20 lg (10.28)
v2b

By means of the mechanical impedances Z1 and Z 2 for the two plates, the
result is:
 Impact sound insulation  287

(a) (b) F v1.a

v1b v2.b v2b
F
v2a v2.b

Figure 10.9 Point force excitation of a slab (a) and a floating floor on the same slab
(b). (Reproduced from Kristensen, J., Rindel, J. H., Bygningsakustik – Teori og
praksis, SBI-Anvisning 166 (In Danish), Danish Building Research Institute,
Hørsholm, Denmark, 1989. With permission.)

F
v2 a = (10.29)
Z2

F
v1b = (10.30)
Z1

The force F′, which is transferred through the elastic layer from plate 1 to
plate 2, can be expressed as follows:

F ' = Zd ( v1b − v2b ) ≅ Zd v1b (10.31)

where Zd is the impedance of the elastic layer. It is presumed that v 1b > v 2b,
which means that the supporting slab must be heavy compared to the float-
ing floor. The velocity in the floor due to the force F′ is:

F' Z FZd
v2b = ≅ v1b d = (10.32)
Z2 Z2 Z1Z2

By inserting the above expression in Equation 10.28, the attenuation of the

impact sound pressure level is:
Z1
∆L ≅ 20 lg (10.33)
Zd
For further analysis, it is necessary to have more information about the
impedance Zd of the elastic layer, which is often made of mineral wool.
Furthermore, it is necessary to differ between floors that are more or less
locally reacting, as this is of importance for how the sound propagates
through the elastic layer. Depending on this, the impedances in Equation
10.33 can be seen either as point impedances or as surface impedances.

10.5.2  Surface-supported floors

The construction shown in Figure 10.9b can be regarded as a double con-
struction, which is characterized by a resonance frequency f 0:
288  Sound Insulation in Buildings

1  1 1 
f0 = k ''d  + (10.34)
2π  m1 m2 

where k″d is the dynamic stiffness per unit area of the elastic layer. This
equation is the same as Equation 9.4 for a double wall, except that here the
dynamic stiffness is not primarily due to the air in the cavity. In many floor
constructions, m1 ≪ m2 and, thus, the latter can be removed from Equation
10.34.
The dynamic stiffness per unit area is a combined contribution from the
structure of the elastic material and from the air contained in the pores of
the elastic material:

ρ mcm2 ρ c 2
k ''d = k ''m + k ''a = + (10.35)
d qd

Here, d is the thickness of the elastic layer and q is the porosity of the mate-
rial, i.e. the ratio between the pore volume and the total volume. ρm and cm
are the density of the material and the speed of longitudinal sound waves,
respectively.
An elastic material is characterized by the dynamic modulus of elasticity
ρmcm2 , which is the same as the stiffness per unit area per thickness unit.
Typical values for a selection of elastic materials are stated in Table 10.1.
Except from light mineral wool, the contribution from the air is often of
minor importance. The dynamic stiffness can be determined experimen-
tally after ISO 9052-1 by measuring the resonance frequency with a load
of known mass.

10.5.3  Dynamic stiffness of loaded mineral wool

Figure 10.10 shows the dynamic stiffness of some mineral wool products,
and how it increases with the static load. This behaviour is not very surpris-
ing since it is a general principle that the speed of propagation of a compres-
sion wave in an elastic material increases with increasing static pressure in
Table 10.1  Examples of dynamic modulus of elasticity for
materials that can be used for floating floors
Material Unit: MN/m2
Air 0.14
Mineral wool (50 kg/m3 to 160 kg/m3) 0.02–0.4
Coco mat 0.25–0.5
Polystyrene foam (10 kg/m3 to 20 kg/m3) 0.3–3
Wood fibre concrete 6–17
Cork 10–30
 Impact sound insulation  289

the material. If the results are corrected for the contribution from the stiff-
ness of the air, the following empirical relationship is found:

k ''m ≅ constant ⋅ m1 (10.36)

where m1 is the static load. Because it is mainly the speed of sound cm that
can change with different loads, Equation 10.35 leads to the relationship:

cm ≅ constant ⋅ 4 m1 (10.37)

Table 10.2 gives the data for the same types of mineral wool, as in
Figure 10.10. In a special experimental setup, the sound speeds cm are mea-
sured and the calculated dynamic modulus of elasticity correspond to the
curves in Figure 10.10.
The speed of sound in mineral wool is approximately one-tenth of the
speed of sound in air, which is of significance for the frequency depen-
dency of the impact sound attenuation. The sound propagation in an elas-
tic material, such as mineral wool, will be accompanied by losses. As
shown by Cremer and Heckl (1967, p. 169), this can be formulated by a
complex sound speed whose imaginary quantity is determined by the loss
factor η:

1
1
Dynamic elasticity module (MN/m2)

0.5 2

0.2

0.1
0.1 0.2 0.5 1 2 5 10
Static load (kPa)

Figure 10.10 Dynamic elasticity as a function of the static load for mineral wool. The
curves 1, 2 and 3 are regression lines for the same three types of min-
eral wool described in Table 10.2. (Adapted from Gudmundsson, S., Sound
insulation improvements of floating floors. A study of parameters. Report
TVBA-3017, LTH, Lund, Sweden, 1984.)
290  Sound Insulation in Buildings

Table 10.2  T hree examples of mineral wool for floating floors (same as in
Figure 10.10)
Dynamic elasticity
Type (number) Density, ρm (kg/m3) Speed of sound, cm (m/s) (MN/m2)
Rockwool (1) 160 49 0.38
Rockwool (2) 100 35 0.12
Fibre glass (3) 50 22 0.02
Source: G
 udmundsson, S., Sound insulation improvements of floating floors. A study of
parameters. Report TVBA-3017, LTH, Lund, Sweden, 1984, p. 159.
Note: The speed of sound is measured with a static load of 1 kPa.

 η
cm → cm  1 + j  (10.38)
 2
A number of measuring results of the loss factor have been referred by
Gudmundsson (1984, p. 162), but there is much uncertainty of this param-
eter. For Rockwool loaded with 1 kPa, the loss factor has been measured to
η = 0.23–0.28. For heavier loads, loss factors have been measured (η = 0.10–
0.15 for loads of 4 kPa to 6 kPa). In order not to complicate the description
further, it is presumed in the following that the loss factor is so small that
it will not contribute to the impact sound attenuation.

10.5.4  Resonantly reacting floors

Resonantly reacting floors means floating floors where the whole floor
plate contributes efficiently to the transmission of vibration energy through
the elastic layer. The impedances in Equation 10.33 should in this case
regarded as surface impedances. The impedance of the floor is:

Z1 = jω m1 (10.39)

where m1 is the mass of the floor per unit area. For the elastic layer, the
impedance is:

ρ mcm2
Zd = (10.40)
ω 
sin  d
 cm 

where the denominator goes towards 0 at frequencies where standing waves
in the elastic layer may occur. It will be practical to introduce a knee fre-
quency fd:

cm
fd = (10.41)
2π d
 Impact sound insulation  291

At low frequencies, the spring effect is not affected by standing waves and
sin(ω d/cm) ≅ ω d/cm. At higher frequencies f > fd a useful simplification is to
set sin(ω d/cm) ≅ 1, leading to:

 ρ mcm2 k ''m
 ω d = ω ( f ≤ fd )
Zd ≅  (10.42)

 ρ mcm ( f > fd )
At frequencies below the resonance frequency f 0, the construction does not
react as a double construction and the attenuation of the impact sound level
is insignificant. For floating floors, the expression (Equation 10.34) for the
resonance frequency (Equation 10.34) can often be simplified, as m1 < m2
and k″m > k″a.

1 ρ mcm2
f0 ≅
2π m1d (10.43)

The attenuation of the impact sound pressure level in the frequency

range  f 0  < f < fd is found by inserting Equations 10.39 and 10.42 in
Equation 10.33.

ω 2m1d f
∆L = 20 lg
ρ mcm2
= 40 lg
f0
( f0 < f ≤ fd ) (10.44)

This strong frequency dependency of 12 dB per octave continues, however,

only to the frequency fd , which is only a little higher than f 0 because of the
low speed of sound in the elastic layer. For f > fd the result is:

ω m1 f f
∆L = 20 lg
ρ mcm
= 40 lg d + 20 lg
f0 f0
(f > fd ) (10.45)

Thus, the frequency dependency is reduced to 6 dB per octave in the

­frequency range above fd.
In Figure 10.11, there are examples of measured impact sound attenua-
tion for resonantly reacting floating floors of 35 mm concrete with differ-
ent underlying layers, which give different resonance frequencies and knee
frequencies. The knee frequencies for the three materials are indicated in
the graph. The curves follow closely the two formulas (Equations 10.44
and 10.45).
Other examples of measured impact sound insulation for resonantly
reacting floating floors are shown in Figure 10.12. The only difference
between the three results is the different type of mineral wool under the
floating floor.
292  Sound Insulation in Buildings

50

40
2
Improvement, ΔLn (dB)

1
30 3

fd2
20

10
fd3
fd1
0
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 10.11 Impact sound attenuation of floating floors with three different elastic
materials. All curves are averaged results from three floors, measured
3 months and 1 year after inauguration. All floors are built from 150 mm
concrete with 35 mm floating concrete floor. Curve 1: 15 mm mineral wool;
curve 2: 25 mm mineral wool; curve 3: 15 mm polystyrene foam plate.
(Adapted from Kuhl, W., Acustica, 51, 103–115, 1982.)

10.5.5  Locally reacting floors

A locally reacting floor means a floor where the force from the hammers of
the tapping machine is mainly transmitted to a limited part of the construc-
tion close to the excitation point of the force. The difference of the force
transmission in resonantly reacting and locally reacting floors is illustrated
in Figure 10.13. In this case, the impedances in Equation 10.33 should be
regarded as point impedances.
The point impedance of the floor is (as in Equations 10.5 and 10.6):

∆f 4c 2m1
Z0,1 = 4m1S ≅ (10.46)
∆N1 π fc1

The impedance for a spring element consisting of the active section of the
elastic layer can be written as:

km
Z0, d = (10.47)
ω

where km is the dynamic stiffness of the active section. As shown by Kuhl

(1982), it is reasonable to assume that the active section in the floor plank
is a cylinder of a diameter, which equals a half bending wave length (see
Figure 10.13b).
 Impact sound insulation  293

70

60
6 dB per octave
50
Improvement, ΔLn (dB)

40

30

20

10

–10
50 100 200 500 1000 2000 5000
Frequency (Hz)
MW type 1 MW type 2 MW type 3

Figure 10.12 Measured impact sound attenuation of resonantly reacting floors made

of 50 mm dense concrete laid on different types of mineral wool, 50 mm
thick. The curves refer to the material numbers 1, 2, and 3, respectively,
in Table 10.2. The structural slab is 160 mm concrete. (Adapted from
Gudmundsson, S., Sound insulation improvements of floating floors. A
study of parameters. Report TVBA-3017, LTH, Lund, Sweden, 1984.)

If the corresponding active area is called S a , the result is:

π 2
km = km′′ Sa = km′′ λB1 (10.48)
16

which by means of Equation 10.35 can be rewritten as:

ρ mcm2 π c 2
km = ⋅ (10.49)
d 16ffc1

The attenuation of impact sound pressure level is found by inserting in

Equation 10.33:

 128m1d 2  f
∆L = 20 lg  f  ≅ 40 lg ( f > f0 ) (10.50)
 πρ mcm2  f0

where f 0 is the resonance frequency (Equation 10.43). The result for locally
reacting floors corresponds exactly to the previously found result for
294  Sound Insulation in Buildings

(a) (b) ½ λB

F F

Figure 10.13 Floating floors with a point force excitation. (a) Resonantly reacting floor.
(b) Locally reacting floor. (Reproduced from Kristensen, J., Rindel, J. H.,
Bygningsakustik – Teori og praksis, SBI-Anvisning 166 (In Danish), Danish
Building Research Institute, Hørsholm, Denmark, 1989. With permission.)

resonantly reacting floors below the knee frequency, but there is no knee
frequency for locally reacting floors. The slope of ΔL is 12 dB per octave in
the entire frequency range, which is in agreement with the model above for
the transmission through the elastic layer. The section of the active cylinder
is reduced by increasing frequency and, thus, the stiffness is inversely pro-
portional to the frequency (see Equation 10.49).
Examples of measured attenuation of impact sound pressure level for
locally reacting floors are seen in Figure 10.14. The types of applied mineral

70

60

50
12 dB per octave
Improvement, ΔLn (dB)

40

30

20

10

–10
50 100 200 500 1000 2000 5000
Frequency (Hz)
MW type 1 MW type 2 MW type 3

Figure 10.14 Measured impact sound attenuation of locally reacting floors made from
22 mm chipboard laid on different types of mineral wool, 50 mm thick. The
curves refer to the material numbers 1, 2 and 3, respectively, in Table 10.2.
The structural slab is 160 mm concrete. (Adapted from Gudmundsson, S.,
Sound insulation improvements of floating floors. A study of parameters.
Report TVBA-3017, LTH, Lund, Sweden, 1984. Fig. III.2.3.)
 Impact sound insulation  295

wool are the same as used in Table 10.2 and Figure 10.10. The curves
increase by 12 dB per octave in agreement with Equation 10.50.

10.5.6  Point-supported floors

Sometimes, floating floors are constructed with the flooring resting on
evenly distributed elastic patches (Figure 10.15). The dynamic stiffness of
each patch is called kdp and the number is Np. As the air contribution to
the stiffness is disregarded and the structural slab is assumed to be much
heavier than the top floor, the resonance frequency for the floor is:

1 kdp ⋅ N p
f0 = (10.51)
2π m1S

The attenuation of the impact sound by point supported floors that are
presumed to be resonantly reacting has been analysed by Ver (1971). With
indices 1 and 2 referring to the flooring plate and the structural slab,
respectively, his result is:

Z mη Z f3
∆L = 10lg  0,1 + 1 1 + 2πη1 0,1 ⋅ 2  (10.52)
 Z0,2 m2η2 kdp f0 

Above a certain frequency, the last term dominates and the impact sound
attenuation increases by 9 dB per octave, if the loss factor of the floor-
ing plate η1 and the stiffness of the patches are frequency independent.
Inserting the resonance frequency (Equation 10.51) and the point imped-
ance (Equation 10.46) yields:

 32π 3 m 2η ∆f  f
∆L ≅ 10lg  2 ⋅ 1 1 ⋅ ⋅ f 3  ≅ 30lg (10.53)
k
 dp N p ∆N 1  f0

Figure 10.15 Point supported floating floor with a point force excitation. The force is
transferred to the slab through a number of elastic patches. (Reproduced
from Kristensen, J., Rindel, J. H., Bygningsakustik – Teori og praksis, SBI-
Anvisning 166 (In Danish), Danish Building Research Institute, Hørsholm,
Denmark, 1989. With permission.)
296  Sound Insulation in Buildings

50

40

9 dB per octave
Improvement, ΔLn (dB)

30

20

10

–10
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 10.16 Measured impact sound attenuation of a point supported floor made from
22 mm wood on laths on 10 mm fibreboard pads. The structural slab is 200 mm
clinker concrete slab with 30 mm finish. Measured data from Figure 5.32.

Especially, the importance of a large loss factor and a low modal density
for the flooring plate should be noted. Thus, the flooring should preferably
be thick and with high damping.
An example of a flooring that performs in accordance with this model is
the traditional wooden floor on laths on resilient pads (Figure 10.16)

10.5.7  The effect of sound bridges

The sound transmission via sound bridges may be of decisive importance for
the impact sound pressure of floating floors. In Figure 10.17, examples of mea-
surements in laboratories are seen for a floating concrete floor where mechani-
cal contacts to the floor were made with different number of gypsum cylinders.
One small sound bridge is enough to ruin the result, which means that good
workmanship is extremely important for a successful result of a floating floor.

10.5.8  Summary on floating floors

The attenuated characteristics of floating floors are so complicated in prac-
tice that it is not possible with the available knowledge to draw up a gen-
erally acceptable basis of calculation. The survey given in this section is
mainly based on analyses by Kuhl (1982) and Gudmundson (1984).
 Impact sound insulation  297

40
n=0

30
Improvement, ΔLn (dB)

20 n=1

n = 10
10

–10
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 10.17 Harmful effect of rigid connections in a floating floor. The number of

bridges is n, and they were made as gypsum cylinders with a diameter of
30 mm. (Adapted from Gösele, K., Schallbrücken bei schwimmenden Estrichen
und anderen schwimmend verlegten Belägen. Berichte aus der Bauforschung,
Heft 35. Wilhelm Ernst und Sohn, Berlin, 1964.)

10.6  SUSPENDED CEILINGS

Compared to floating floors, the suspended ceilings have the advantage that
the resonance frequency can be significantly lower (preferably below 50 Hz)
because the gap between the slab and the suspended ceiling can be made
much wider. Still, the use of suspended ceilings for attenuation of impact
sound is not common practice.
The resonance frequency of a suspended ceiling is calculated as for an
asymmetric double construction with one heavy mass and one lighter mass:

1 ρc2
f0 ≅ (10.54)
2π m3d

Here, m3 is the mass per unit area of the suspended ceiling, d is the suspen-
sion, and ρc 2 is the dynamic elasticity of the air in the cavity.
Both perforated and closed gypsum board plates can provide attenuation
of impact sound. It is possible, at least to some extent, to combine a room
acoustic treatment of the room below with improved impact sound insula-
tion (Seidel and Hengst, 2017).

10.7  MASSIVE WOOD FLOOR CONSTRUCTIONS

Massive wood is an interesting building material with several good properties,

especially in relation to ecology. However, due to low density, there are obvi-
ously some challenges if good sound insulation shall be obtained. This was
298  Sound Insulation in Buildings

Table 10.3  O
 verview of measurement results on floor constructions made from
massive wood
Construction Description Rw (dB) Rw,50 (dB) Ln,w (dB) Ln,w,50 (dB)
185 mm wood 44 43 82 75

115 mm wood 56 52 59 60
30 mm mineral
wool
185 mm wood

185 mm wood 62 59 50 57
150 mm
suspended
ceiling
2 × 13 mm
gypsum board

115 mm wood 63 59 41 53
30 mm mineral
wool
185 mm wood
150 mm
suspended
ceiling
2 × 13 mm
gypsum board

115 mm wood 60 57 50 53
30 mm mineral
wool
50 mm concrete
tiles (200 × 400)
mm
185 mm wood

(Continued)
 Impact sound insulation  299

Table 10.3  (Continued) Overview of measurement results on floor constructions

made from massive wood
Construction Description Rw (dB) Rw,50 (dB) Ln,w (dB) Ln,w,50 (dB)
115 mm wood 62 60 47 48
30 mm mineral
wool
100 mm sand
185 mm wood

115 mm wood 63 61 33 48
30 mm mineral
wool
100 mm sand
185 mm wood
150 mm
suspended
ceiling
2 × 13 mm
gypsum board

Source: Eriksen, L., Ejlersen, C. V. J., Lydisolering – etageadskillelser af massive træelementer.

(In Danish). MSc Thesis, Byg-DTU and Ørsted-DTU, Technical University of Denmark,
Lyngby, Denmark, 2003.

studied through a series of laboratory measurements in a project by Eriksen and

Ejlersen (2003). A few examples of the results are listed in Table 10.3. The fol-
lowing abbreviations are used: Rw,50 = Rw + C50–3150 and Ln,w,50 = Ln,w + C I,50–2500.
The aim of the project was to search for a solution that could fulfil the
Danish requirements on airborne and impact sound insulation for floors
between dwellings, i.e. minimum R′w ≥ 55 dB and L′n,w ≤ 53 dB, but prefer-
ably sound class B: R′w + C 50–3150 ≥ 58 dB and L′n,w + C I,50–2500 ≤ 48 dB. The
wood material was glued with the same technique as used for laminated
wooden beams. The basic slab was a wooden plate 185 mm thick, and as
an upper floor, another wooden plate 115 mm thick was used.
The results show clearly that the 185 mm slab alone is far from meeting
the acoustic requirements (the first row in Table 10.3 and Figure 10.18).
The critical frequency of the slab is around 110 Hz, and there are very
few structural modes below 160 Hz. Above 160 Hz, the transmission is
due to resonant shear waves and the impact sound pressure level increases
with 3 dB per octave in agreement with Equation 10.16 until 1250 Hz. The
attenuation of impact sound at the high frequencies can be due to the rela-
tively soft surface of the wooden slab.
300  Sound Insulation in Buildings

90

80

70

60
R and Ln (dB)

50

40

30

20

10
50 100 200 500 1000 2000 5000
Frequency (Hz)
R Ln

Figure 10.18 Measured transmission loss and impact sound pressure level of a floor
of 185 mm massive wood. (Adapted from Eriksen, L., Ejlersen, C. V. J.,
Lydisolering – Etageadskillelser af massive træelementer. (In Danish). MSc
Thesis, Byg-DTU and Ørsted-DTU, Technical University of Denmark,
Lyngby, Denmark, 2003.)

The combination of a floating floor of 115 mm massive wood on a resil-

ient layer of 30 mm mineral wool gives promising improvements, but the
impact sound insulation is still insufficient (see second row in Table 10.3
and Figure 10.19). In order to achieve sufficient impact sound insulation at
low frequencies, it is necessary to increase the total mass and this was tried
with 50 mm concrete tiles and with 100 mm sand. Figure 10.20 shows the
results of the latter. Both solutions are efficient but sand yields results that
are a few dB better than results with concrete tiles, probably due to a higher
loss factor. Actually, the results with sand meet the goal, at least in the
laboratory. When applied in a real building, the sound insulation is always
a few dB lower than in the laboratory.
A suspended ceiling was also tried, both alone and in combination with
a floating floor. The impact sound attenuation of a 150 mm suspended
ceiling is better than that obtained by a floating floor compare (second and
third row in Table 10.3). Actually, the suspended ceiling meets the mini-
mum requirements for dwellings with a good margin. The combination of
a floating floor and a suspended ceiling is even more efficient as the attenu-
ations are additive (Figure 10.21).
 Impact sound insulation  301

80
Floating floor
70

60

50
R and Ln (dB)

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
Ri Ln

Figure 10.19 Measured transmission loss and impact sound pressure level of a floor of
185  mm massive wood with a floating floor of 115 mm massive wood on
30 mm mineral wool. (Adapted from Eriksen, L., Ejlersen, C. V. J., Lydisolering –
Etageadskillelser af massive træelementer. (In Danish). MSc Thesis, Byg-DTU
and Ørsted-DTU, Technical University of Denmark, Lyngby, Denmark, 2003.)

80
Floating floor + sand
70

60

50
R and Ln (dB)

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
R Ln

Figure 10.20 Measured transmission loss and impact sound pressure level of a floor of
185 mm massive wood with a topping of 100 mm sand and a floating floor
of 115 mm massive wood on 30 mm mineral wool. (Adapted from Eriksen,
L., Ejlersen, C. V. J., Lydisolering – Etageadskillelser af massive træelementer.
(In Danish). MSc Thesis, Byg-DTU and Ørsted-DTU, Technical University
of Denmark, Lyngby, Denmark, 2003.)
302  Sound Insulation in Buildings

70

60

50
Improvement, ΔLn (dB)

40

30

20

10

–10
50 100 200 500 1000 2000 5000
Frequency (Hz)
Floating floor Suspended ceiling Floating floor + suspended ceiling Sum

Figure 10.21 Impact sound attenuation of a floating floor of 115 mm massive wood on

30 mm mineral wool or a 15-mm suspended ceiling with 2 × 13 mm gypsum
board, or both combined. The supporting floor is 185 mm massive wood.
(Data from Eriksen, L., Ejlersen, C. V. J., Lydisolering – Etageadskillelser af
massive træelementer. (In Danish). MSc Thesis, Byg-DTU and Ørsted-DTU,
Technical University of Denmark, Lyngby, Denmark, 2003.)

10.8  SOUND INSULATION IN PRACTICE

10.8.1  A live walker and the ISO tapping machine

The tapping machine was developed to produce impact sound suited
for measurements. Hence, the tapping frequency (10 Hz) gives a suffi-
ciently dense line spectrum and the force of the impacts is high enough
to allow measurements without too much interference due to back-
ground noise. However, the resemblance with real footfall is not obvious.
Measurements with the tapping machine are compared with examples
of measured footfall noise from a live walker in Figure 10.22. In this
example, the construction is a concrete slab with a floating concrete floor.
At low frequencies (below 100 Hz), walking without shoes creates sound
pressure levels surprisingly close to the sound pressure levels from the tap-
ping machine, whereas walking with shoes gives about 10 dB lower levels.
The difference with or without shoes can be explained by the difference
between a hard and a soft impact; the latter makes an impulse stretched
in time, which means more energy at low frequencies. At higher frequen-
cies (above 100 Hz), the sound pressure levels from the tapping machine
are 20 dB to 30 dB higher than those from the live walker. Figure 10.22
 Impact sound insulation  303

80

70
Impact sound pressure level, L΄n (dB)

60

50

40

30

20

10

0
20 50 100 200 500 1000 2000 5000
Frequency (Hz)
Tapping machine Walking without shoes
Walking with shoes HT (ISO 226:2003)

Figure 10.22 Impact sound measured with the tapping machine compared with the
average sound pressure level measured with a live walker with or without
shoes. The floor is 250 mm concrete slab with 50 mm light concrete top-
ping, 17 mm resilient layer (polyurethane) and 65 mm floating concrete
floor. The hearing threshold for pure tones (HT) is shown for compari-
son. (Adapted from Wolf, M., Burkhart, C., Vermeidungsstrategien und
Ansätze einer Vermeidung des Estrichdrönens. Proceedings of DAGA 2016,
Aachen, Germany, 2016. Figure 3.)

also shows the hearing threshold (HT), although this is for pure tones
under special listening conditions. Nevertheless, this indicates that for
the actual example, the audible footfall noise is dominated by frequencies
below 100 Hz.
The relation between the spectrum from a live walker and from the tap-
ping machine depends to a great deal on the floor construction. Figure 10.23
shows the difference in spectrum for three different floors; a bare concrete
(difference increases monotonic with frequency), a floating top surface (the
difference is nearly constant around 20 dB above 100 Hz) and a floor with
carpet (maximum difference around 50 Hz to 200 Hz). A positive differ-
ence means that the tapping machine is louder than the live walker. These
results demonstrate that the tapping machine can be used effectively as a
testing device up to at least 500 Hz for all types of surface except carpet
(Warnock, 1998). With a carpeted floor, the difference curve is fairly close
to the others below 200 Hz. Since the impact sound from a carpeted floor
304  Sound Insulation in Buildings

60

50
Difference in impact sound pressure level (dB)

40

30

20

10

–10

–20
20 50 100 200 500 1000 2000 5000
Frequency (Hz)
Concrete surface Floating top surface Carpet

Figure 10.23 Difference between impact sound pressure levels measured with the
tapping machine and with a live walker with shoes, shown for three dif-
ferent floorings: Bare concrete, floating floor and carpet. (Adapted from
Warnock, A. C. C., Floor research at NRG Canada. Conference in Building
Acoustics “Acoustic Performance of Medium-Rise Timber Buildings”, Dublin,
Ireland, 1998. Figure 3.)

is dominated by low frequencies, it is concluded that a single number rating

that includes the low frequencies may correlate well with a rating of the
sound from a live walker (Warnock, 1998).

10.8.2 LIGHTWEIGHT AND HEAVYWEIGHT

FLOOR CONSTRUCTIONS

Complaints about impact noise in dwellings are very often related to the
thudding sound created by footfall. The low-frequency problems of impact
sound have been clearly related to lightweight, wood-frame building con-
structions (Blazier and DuPree, 1994). They found that the peak in the foot-
fall spectrum occurred at the fundamental natural frequency of the floor
construction, typically between 15 Hz and 30 Hz. The characteristic
­difference between impact sound spectra for lightweight and heavyweight
 Impact sound insulation  305

floor constructions is illustrated in Figure 10.24. The difference is obvious

below 100 Hz.

10.8.3 Typical complaints about impact sound

In multifamily houses, the footfall noise from neighbours is one of the most
common causes of complaints. From a technical and acoustical point of
view, it is a difficult problem to deal with, one of the reasons being that the
measurement and evaluation methods are still insufficient.
In fact, impact noise can be divided into two different types that cover
different frequency ranges and may vary independently:

• Low-frequency rumbling, typically below 100 Hz. This is related to

the total weight of the floor-ceiling construction.
• High-frequency heel clicks, typically above 400 Hz. This is related to
the surface of the floor.

It has been suggested to use two independent evaluation methods to cover

these two frequency ranges (LoVerde and Dong, 2017). For the low-fre-
quency range, they suggest using the impact sound pressure level in the
octave band including the third octaves 50 Hz, 63 Hz and 80 Hz:

(
L50 −80 = 10 lg 100.1⋅L50 + 100.1⋅L63 + 100.1⋅L80 ) (10.55)

For the evaluation of this low-frequency impact sound, the following guide
(LoVerde and Dong, 2017) is suggested:

• L 50–80 > 70 dB; the floor is ‘thuddy’ or ‘boomy’; substantial amount of

complaints
• L 50–80 < 60 dB; broadly acceptable
• L 50–80 < 50 dB to 55 dB; thudding inaudible; no complaints

In a German investigation on the problems of low-frequency impact sound

(Rittig, 2013), a clear difference was found between results from cases
of complaints and those from cases of no complaints (Figure 10.25). The
­question is how to characterize the difference in the best way? In Table 10.4,
some objective parameters derived from the two curves in Figure 10.25
are listed. There is no doubt that the frequencies 50 Hz to 80 Hz are very
­important, which means that we cannot rely on the L′n,w value. In other con-
texts (noise from Heating, Ventilation and Air Conditioning (HVAC) sys-
tems), it is known that rumble causes increased annoyance and this is related
to ­spectral imbalance. In other words, if the energy at mid-­frequencies
is much lower than the energy at low frequencies, this may cause higher
annoyance than noise with a more balanced spectrum (Beranek, 1989).
306  Sound Insulation in Buildings

80

70
Impact sound pressure level, L΄n (dB)

60

50

40

30

20

10

0
20 50 100 200 500 1000 2000 5000
Frequency (Hz)
Heavy Light

Figure 10.24 Measured impact sound pressure levels in Danish multifamily houses.

Average of 10 heavyweight and 8 lightweight floor constructions.
Normalization with reverberation time of receiving room is not applied
below 50 Hz. (Data from DELTA, Målerapport: Boligers lydisolation ved lave
frekvenser. (In Danish). PDØ 870028, DELTA, Aarhus, Denmark, 1999.)

The importance of spectrum balance has been found by many researchers,

and there seems to be a general agreement about the slope of a potentially
annoying sound being around 6 dB to 7 dB per octave or more (Waye, 2011).
In the cases of complaints about impact noise, the slope of the spectrum is
7 dB per octave, whereas the other curve with no complaints has a slope of
5 dB per octave. In this connection, we recall that the spectrum from a live
walker has more low-frequency energy than that measured with the tapping
machine so the spectral imbalance experienced with footfall noise is even
more pronounced.
However, it is a myth that low-frequency problems with impact sound
are only related to lightweight constructions. Floor constructions that
often give rise to complaints about impact noise can be divided into three
groups:

• Lightweight timber constructions. The problem is rumbling and the

important frequency range may be 20 Hz to 200 Hz. More weight is
necessary to solve this problem.
 Impact sound insulation  307

80

70
Impact sound pressure level, L΄n (dB)

60

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
Complaints Acceptable

Figure 10.25 Mean values of the impact sound pressure levels measured on site, from
10 cases of quality check and 12 cases of complaints. (Adapted from Rittig,
C., Übertragung tieffrequenter trittschallgeräusche auf massivdecken
mit schwimmendem estrich. Bachelor thesis, Building Physics, Technical
University, Stuttgart, Germany, 2013.)

• Heavyweight concrete constructions with tiles or similar hard sur-

face. The problem is high-frequency clicks due to hard-heeled shoes
or similar (chairs drags, dog toenails). The important frequency range
is 400 Hz to 3150 Hz (or 5000 Hz). The solution is a soft flooring or
a floating floor.
• Heavyweight construction with a heavy floating floor. The problem
that can occur is rumbling and the important frequency range may be
50 Hz to 100 Hz. If the floating floor has a resonance frequency above
50 Hz and the resilient layer has a high Q-factor, the impact sound is
amplified at frequencies around the resonance frequency. The solution

Table 10.4  S ome characteristics of the average impact sound pressure levels in
cases of complaints and in cases of no complaints
L′n,w (dB) L′n,w,50 (dB) L50–80 (dB) Slope (dB/octave)
Complaints 42 50 62 −7
No complaints 37 41 53 −5
Note: Same data as in Figure 10.25.
308  Sound Insulation in Buildings

can be to bring the resonance frequency below 50 Hz by changing the

resilient layer. Alternatively, a resilient material with a low Q-factor
may help.

Examples of measured results of constructions from each of the three groups

are shown in Figures 10.26 through 10.28. The ISO weighting curve and
the unfavourable deviations from the curve are also shown for comparison.
While the weighting curve works reasonably well in the example with a
hard surface, Figure 10.27, this is not the case in the other examples where
low-frequency rumbling is the problem. The parameter with extended fre-
quency range L’n,w,50 works fine in the two examples with low-frequency
problems, but fails in the example of a hard floor surface (Table 10.5).
The resonance frequency of a floating floor is calculated with Equation
10.43. The level of the resonance peak (negative ΔL) depends on the
Q-factor of the resilient layer, see the definition (2.69):

(
∆L0 = −10lg 1 + Q2 ≅ −20lg Q )
(10.56)

80

70
Impact sound pressure level, L΄n (dB)

60

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 10.26 Impact sound pressure level of a timber floor. Construction from top to
bottom: soft carpet, 22 mm chipboard, 48 mm × 198 mm wooden beams
with 200 mm mineral wool in the cavity, 48 mm × 48 mm lath with resil-
ient mounting of 2 × 13 mm gypsum board. (Adapted from Hagberg, K.,
Simmons, C., Bygg & teknik 1/97, 17–18, 1997.)
 Impact sound insulation  309

80

Impact sound pressure level, L΄n (dB) 70

60

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 10.27 Impact sound pressure level of a heavyweight floor with hard surface.
Construction from top to bottom: Clinker tiles on 250-mm concrete.
(Adapted from Hagberg, K., Simmons, C., Bygg & teknik 1/97, 17–18, 1997.)

80

70
Impact sound pressure level, L΄n (dB)

60

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 10.28 Impact sound pressure level of a heavyweight floor with floating floor.
Construction from top to bottom: 14 mm parquet, 3 mm ethafoam,
30 mm gypsum, 0.15 mm plastic foil, 15 mm mineral wool, 230 mm con-
crete. (Data from Multiconsult, Unpublished measurement report. Project
Number 103407, 2002.)
310  Sound Insulation in Buildings

Table 10.5  S ome characteristics of the impact sound pressure levels in cases of
complaints
Construction L′n,w (dB) L′n,w,50 (dB) L50–80 (dB) Slope (dB/octave)
Lightweight timber 44 57 72 −9
Concrete with hard surface 62 52 45 +5
Concrete with floating floor 42 56 71 −7
Note: Same data as in Figures 10.26 through 10.28.

Since ΔL is the positive attenuation of the impact sound, the amplification

at the resonance is a negative dB value. If the resilient layer is some kind of
mineral wool, as described by Schirmer and Schirmer (2016), the Q-factor
can be estimated from the flow resistance:

2π m1f0
Q= (10.57)
rsd

where m1 is the mass per unit area of the floating floor, rs is the airflow
resistivity per unit length and d is the thickness of the resilient layer. Typical
values of Q for floating floors are between 2.5 and 10.
Unfortunately, it is difficult to get information about the flow resistance
from the manufacturers of resilient materials for floors; normally, only data for
the dynamic stiffness are available. The relevant measurement standards are
ISO 9052-1 for the dynamic stiffness and ISO 9053 for the airflow resistance.

REFERENCES

L.L. Beranek (1989). Balanced noise-criterion (NCB) curves. Journal of the Acoustical
Society of America 86(2), 650–664.
L.L. Beranek and I.L. Vér (1992). Noise and Vibration Control Engineering. John
Wiley & Sons, New York.
W.E. Blazier and R.B. DuPree (1994). Investigation of low-frequency footfall noise
in wood-frame, multifamily building construction. Journal of the Acoustical
Society of America 96 (3), 1521–1532.
L. Cremer and M. Heckl (1967). Körperschall. (Structure-borne sound, translated by
E.E. Ungar 1973). Springer, Berlin, Germany.
DELTA (1999). Målerapport: Boligers lydisolation ved lave frekvenser. (Measurement
report: Sound insulation of dwellings at low frequencies, in Danish). PDØ
870028, DELTA, Aarhus, Denmark.
L. Eriksen and C.V.J. Ejlersen (2003). Lydisolering—Etageadskillelser af massive
træelementer. (Sound insulation – floor constructions of massive wood slabs,
in Danish). MSc Thesis, Byg-DTU and Ørsted-DTU, Technical University of
Denmark, Lyngby, Denmark.
S. Gudmundson (1984). Sound insulation improvements of floating floors. A study
of parameters. Report TVBA-3017, LTH, Lund, Sweden.
 Impact sound insulation  311

K. Gösele (1964). Schallbrücken bei schwimmenden Estrichen und anderen schwim-

mend verlegten Belägen. (Sound bridges in floating floors and other floating
layers, in German). Berichte aus der Bauforschung, Heft 35. Wilhelm Ernst und
Sohn, Berlin, Germany.
K. Hagberg and C. Simmons (1997). Nordisk ljudklassningsstandard – skärpta ljud-
krav. (Nordic standard for sound classification – increased acoustical require-
ments, in Swedish). Bygg & teknik 1/97, 17–18.
ISO 9052-1 (1989). Acoustics – Determination of dynamic stiffness – Part 1:
Materials used under floating floors in dwellings. International Organization
for Standardization, Geneva, Switzerland.
ISO 9053 (1991). Acoustics – Materials for acoustical applications – Determination
of airflow resistance. International Organization for Standardization, Geneva,
Switzerland.
ISO 10140-5 (2010). Acoustics – Laboratory Measurement of Sound Insulation of
Building Elements – Part 5: Requirements for test facilities and equipment.
International Organization for Standardization, Geneva, Switzerland.
ISO 10140-1 (2010). Acoustics – Laboratory measurement of sound insulation of
building elements – Part 1: Application rules for specific products. International
Organization for Standardization, Geneva, Switzerland.
J. Kristensen and J.H. Rindel (1989). Bygningsakustik – Teori og praksis. SBI-
Anvisning 166 (Building acoustics – Theory and practice, in SBI-Anvisning
166. Danish). Danish Building Research Institute, Aalborg University
Copenhagen.
W. Kuhl (1982). Impedanz von platten endlicher fläche, Körperschallpegeldifferenz
punktförmig angeregter paralleler platten und trittschallminderung.
(Impedance of plates with finite size, structure-borne sound level difference
of parallel plates with point excitation and attenuation of impact sound, in
German). Acustica 51, 103–115.
J.J. LoVerde and D.W. Dong (2017). A dual-rating method for evaluating impact
noise isolation of floor-ceiling assemblies. Journal of the Acoustical Society of
America 141, 428–440.
Multiconsult (2002). Unpublished measurement report. Project Number 103407.
C. Rittig (2013). Übertragung tieffrequenter trittschallgeräusche auf massivdecken
mit schwimmendem estrich. (Transfer of low-frequency impact sound of solid
slabs with a floating floor, in German). Bachelor’s thesis, Building Physics,
Technical University, Stuttgart, Germany.
H. Schirmer and W. Schirmer (2016). Schwimmender estrich und unterhangdecke
mit schwach gedämpfter eigenfrequenz fast gleicher grösse als ursache für
erhöhe tieffrequente trittschallübertragung. (Floating floor and suspended ceil-
ing with weakly attenuated eigenfrequency although the same as reason for
increase low-frequency transmission of impact sound, in German). Proceedings
of DAGA 2016, 4–17 March 2016, Aachen, Germany.
J. Seidel and K. Hengst (2017). Trittschallminderung und einfügungsdämmung
von raumakustik-decken. (Impact sound attenuation and insertion loss of
room acoustical ceilings, in German). Proceedings of DAGA 2017, 6–9
March 2017, Kiel, Germany.
I.L. Ver (1971). Impact noise isolation of composite floors. Journal of the Acoustical
Society of America 50, 1043–1050.
312  Sound Insulation in Buildings

A.C.C. Warnock (1998). Floor research at NRG Canada. Conference in Building

Acoustics “Acoustic Performance of Medium-Rise Timber Buildings”, Dublin,
Ireland.
K.P. Waye (2011). Noise and Health – Effects of Low Frequency Noise and
Vibrations: Environmental and Occupational Perspectives. Encyclopedia on
Environmental Health (J.O. Nriagu, Editor-in-Chief). Elsevier, Amsterdam.
M. Wolf and C. Burkhart (2016). Vermeidungsstrategien und ansätze einer verme-
idung des estrichdrönens. (Strategies for control and attempts to attenuate the
thump of floors, in German). Proceedings of DAGA 2016, 14–17 March 2016,
Aachen, Germany.
Chapter 11

Flanking transmission

The flanking constructions and the junctions between the constructions are
of high importance for the sound insulation between rooms. The number of
transmission paths to be considered is often quite high, so there is need for
a computer program to handle this in a reasonable way. Computation pro-
grams for sound insulation normally have the drawback that their results
are the maximum values that can be achieved under conditions with perfect
workmanship and acoustically correct building details, which can seldom
be achieved in practice. Yet, computation programs are useful tools at the
design stage for sound insulation.

11.1 FLANKING TRANSMISSION
OF AIRBORNE SOUND

11.1.1 Transmission paths
In buildings with traditional heavy constructions of concrete or masonry,
about 50 % of the sound transmission between rooms with a common
dividing partition is flanking transmission. Of course, the flanking trans-
mission is 100 % between rooms without a common dividing partition.
Figure 11.1 shows the principles of flanking transmission on four differ-
ent types of building. It is important whether the constructions are heavy-
weight or lightweight because the sound radiation of free bending waves
is different. Heavyweight walls and floors have a low critical frequency
and, thus, the radiation efficiency for free bending waves is high (approxi-
mately unity). Figure 11.1a shows a building type with heavyweight walls
and floors like the precast concrete houses typical from the 1960s and the
1970s. Flanking is both vertical and horizontal, and this may have contrib-
uted to the bad reputation of sound insulation in houses from that time.
Type B with lightweight walls and heavyweight floors is not unusual today,
and flanking is mainly in the horizontal direction. Type C with heavy-
weight, load-bearing walls and lightweight floors is very common in older
buildings with masonry walls and traditional timber floors. These houses

313
314  Sound Insulation in Buildings

(a) (b)

(c) (d)

Figure 11.1 Direction of flanking transmission in four building types. (a) Type A – heavy-
weight walls and floors; (b) type B – light-weight walls and heavy-weight floors;
(c) type C – heavy-weight walls and light-weight floors; and (d) type D – light-
weight walls and floors.

are characterized by flanking in the vertical direction. Finally, type D has a

load-bearing system of columns and beams, both wall and floor construc-
tions are lightweight, and flanking is not a big issue.
Figure 11.2 shows examples of transmission paths for airborne sound
from one room to the adjacent room. For each set of flanking surfaces,
there are three flanking transmission paths, i.e. with four flanking surfaces,
there are 12 transmission paths in addition to the direct transmission.

11.1.2 Junction attenuation
One flanking transmission path involves two surfaces: one in the source
room and another in the receiving room. The pair of plates is connected
through a junction (Figure 11.3). The junction attenuation is defined as the

Figure 11.2 Airborne sound transmission horizontally. The transmission paths are shown
via one flanking surface (the floor) and directly through the wall. (Reproduced
from Kristensen, J. and Rindel, J. H., Bygningsakustik – Teori og praksis (Building
acoustics – Theory and practice, in Danish). SBI-Anvisning 166, Danish Building
Research Institute, Hørsholm, Denmark, 1989. With permission.)
 Flanking transmission  315

P1,i P2,ij

Dv,ij vj
vi

Si σi Sj σj

P2,i

Figure 11.3 Simplified outline for flanking transmission from plate i to plate j at airborne
sound exciting. Dv,ij is the junction attenuation.

difference in velocity level on the plate i in the source room and that on the
plate j radiating into the receiving room:

v i2 1
Dv ,ij = 10 lg = 10 lg (11.1)
v 2j dij

The direction is i → j, which is important; the junction attenuation may

be different in the opposite direction. The determination of the junction
attenuation will be described later in this chapter. Since the plate i is in the
source room, both forced and resonant excitation will take place.

11.1.3 Flanking transmission loss

Each transmission path is described by a flanking transmission loss R ij

corresponding to the transmission via the surface i of the source room and
the surface j of the receiving room. The flanking transmission loss from i
to j is:

P1,0
Rij = 10 lg (11.2)
P2,ij

where P1,0 is the incident sound power on the partition wall area S 0 and P2,ij
is the power radiated into the receiving room from the transmission path
under consideration. The reason for choosing the same reference area S 0
for all flanking transmission losses is that they can easily be added together
without further area corrections. This would not be the case if each flank-
ing transmission loss had the incident sound power P1,i on the flanking
surface Si as a reference.
316  Sound Insulation in Buildings

Under the assumption of diffuse sound field in the source room, the
flanking transmission loss can be written as (see Equation 8.25):

p12 S0
Rij = 10 lg (11.3)
4 ρ cP2,ij

where p1 is the mean square sound pressure in the source room and S 0 is
2

the area of the common surface between the rooms. In cases where the rooms
have no common separation area, S0 is set to a fixed reference value of 10 m2 .
With a diffuse sound field in the source room, the mean velocity of forced
vibrations in the excited plate i can be determined from Equations 8.23 and
5.35:

2 p12
v i2,for = (11.4)
(ω mi )2  (1 − ( f fc , i )
2 2
) 
+ ηi2 


The resonant vibrations are similarly determined from Equation 7.29:

p12 c 2σ i ,res ∆Ni

v i2,res = ⋅ ⋅ (11.5)
(ω mi )
2
2ηi Si f ∆f

The total vibrations in plate i are combined from forced and resonant excita-
tion. It can be assumed that both are transferred to the receiving plate j with the
same junction attenuation and, thus, the resulting velocity in plate j becomes:

(
v 2j = dij ⋅ v i2,for + v i2,res ) (11.6)

The forced vibrations are restricted to the excited plate i. In plate j, only
resonant vibrations are possible although the energy of the vibrations is fed
from forced and resonant vibrations in plate i. By means of Equations 11.4
through 11.6, the radiated sound power from plate j is:

P2,ij = v 2j ρcS jσ j ,res

 
ρ c  2 c 2
σ ∆N 
i
= dij ⋅ p12 ⋅ ⋅ S jσ j ,res ⋅  + i ,res
⋅  (11.7)
( )
2
(ω mi )2
 1 − ( f fc ,i ) + ηi2 2ηi Si f ∆f 
2
 

The flanking transmission loss R ij can be determined by insertion of this

result in Equation 11.3:
 Flanking transmission  317

 
S jσ j ,res  2 c 2σ i ,res ∆Ni 
Rij = R0,i + Dv ,ij − 10lg − 10lg  + ⋅  (11.8)
( )
2
S0  1 − ( f fc ,i ) + ηi2 2ηi Si f ∆f 
2
 

where R0,i = 20lg (ω mi 2 ρc ) is the mass law at normal incidence for plate
i. From Equation 11.8, it is seen that a high flanking transmission loss can
be achieved when the plate being excited has a high transmission loss, i.e.
with heavy flanking constructions. Large junction attenuation or small
radiation efficiency for the radiating surface j may also contribute to a
high flanking transmission loss.
The transmission loss R of the direct transmission path can be written in
a form similar to Equation 11.8, using index 1 for this surface:

 
 2σ 1,for c 2σ 1,res
2
∆N1 
R = R0,1 − 10 lg  + ⋅  (11.9)
( )
 1 − ( f fc ,1 )2 + η12 2η1S1f ∆f 
2

 

For this direct transmission, the radiation into the receiving room is dif-
ferent for the forced and resonant vibrations, represented by the different
radiation efficiencies σ 1,for and σ1,res. Here, the area is called S1, but it is
normally the same as the reference area S0.

11.1.4 The total apparent transmission loss

The apparent sound transmission loss R′ can be determined by summing
up the radiated sound power from all possible transmission paths including
the sound power from the direct path P2,0:

R′ = 10 lg
P1,0
P2,0 + Σ P2,ij
(
= −10 lg 10− R / 10 + Σ 10− Rij / 10

) (11.10)

where each flanking transmission path is calculated according to Equation

11.8 and the direct transmission from Equation 11.9.

11.1.5 The reciprocity method

A simplification of the basis of calculation is possible when it is assumed
that sound transmission is independent of the transmission direction, as
shown by Gerretsen (1979). This will only be valid for heavy walls with
318  Sound Insulation in Buildings

sufficiently low critical frequency so only resonant transmission is consid-

ered. Under these assumptions, the flanking transmission loss may be writ-
ten in terms of the transmission loss Ri of the flanking wall:

S0σ i ,res
Rij = Ri + Dv ,ij + 10 lg (11.11)
S jσ j ,res

and if it is assumed that this is the same in both directions:

Rij = Rji = 1
2 ( Rij + Rji ) (11.12)

Thus, the radiation efficiencies will disappear from the equation, and

S0
Rij = 12 Ri + 12 Rj + Dv ,ij + 10 lg (11.13)
Si S j

where

Dv ,ij = 1
2 ( Dij + Dji ) (11.14)

is the junction attenuation as a mean value for the two directions i → j

and j → i. This reciprocity method has been the basis for the standard ISO
12354-1. However, as explained here, the applications are rather limited
since the method is only valid for heavy constructions where the forced
transmission can be neglected.

11.2 FLANKING TRANSMISSION OF IMPACT SOUND

Figure 11.4 shows examples of transmission paths for impact sound pres-
sure horizontally and vertically. Horizontally, there will normally only be
two significant transmission paths, whereas vertically, there will be one
transmission path for each of the four flanking walls, in addition to the
direct transmission through the horizontal division.

11.2.1 The impact sound pressure level

for one transmission path
The impact sound pressure level with sound transmission from i to j only is:

p 2,2ij A2 4 ρcP2,ij
Ln,ij = 10 lg = 10 lg (11.15)
2
p A0
0 p02 A0
 Flanking transmission  319

(a) (b)

Figure 11.4 Impact sound transmission paths. (a) Horizontal and (b) vertical. (Reproduced
from Kristensen, J. and Rindel, J. H., Bygningsakustik—Teori og praksis (Building
acoustics—Theory and practice, in Danish). SBI-Anvisning 166, Danish Building
Research Institute, Hørsholm, Denmark, 1989. With permission.)

where A 2 is the equivalent absorption area in the receiving room and

A 0 = 10 m 2 is a reference area (see also Figure 5.31). The transmission is
illustrated schematically in Figure 11.5. With the excitation force F, the
resonant modes in plate i are activated (see Equation 10.10) and the mean
velocity square can be found from:

F 2 Re {Yi } F 2 ∆Ni
v i2 = = ⋅ (11.16)
2π fmi Siηi 8π fmi Si ηi ∆f
2 2

where mi, Si and ηi are the mass per unit area, area and loss factor of plate
i, respectively. Yi is the mechanical point admittance of the plate, i.e. the
inverse of the point impedance Yi = 1/Zi. The modal density in plate i is ΔNi /
Δf, and this is a function of the frequency (different for bending waves and
shear waves).
The transmission of the vibration energy from plate i to plate j is described
with the junction attenuation Dv,ij (Equation 11.1). The radiated sound power
P2,ij is a function of the average velocity in plate j and the radiation efficiency:

F P2,ij

Dv,ij
vi vj

Zi Si mi ηi Sj σj

Figure 11.5 Simplified outline for flanking transmission from plate i to plate j by e­ xciting
with a tapping machine. Dv,ij is the junction attenuation and Zi is the mechani-
cal point impedance.
320  Sound Insulation in Buildings

P2,ij = v 2j ρcS jσ j ,res

v 2j F 2 ρc ∆Ni (11.17)

= ⋅ ⋅ ⋅ S jσ j ,res
v i2 8π fmi Si ηi ∆f
2 2

The impact sound pressure level for transmission path i → j can now be
determined from:

4 ρcP2,ij  ρ 2c 2 v 2j F 2 ∆Ni 
Ln,ij = 10 lg = ⋅ ⋅ σ 
10 lg  ⋅ S j j ,res  (11.18)
p02 A0  p02 A0 v i2 2π fmi2 Si2ηi ∆f 

By insertion of the force generated by the hammers of the tapping machine

in one-third octave frequency bands F 2 = 0.91f (N2), (Equation 10.4) and
other constants and including the possibility of soft flooring with the atten-
uation ΔL cover, the result is:

 c2 ∆Ni 
Ln,ij = 82 − 20 lg mi − Dv ,ij + 10 lg  2 ⋅ ⋅ S jσ j ,res  − ∆Lcover (11.19)
π
 i i
S η ∆f 

This general result can be simplified if it is assumed that the vibra-

tions in plate i are bending waves, and the appropriate modal density is
inserted (see Table 7.1):

 η   S jσ j ,res 
Ln,ij = 82 − 10 lg  mi2 i  − Dv ,ij + 10 lg   − ∆Lcover (11.20)
 fc ,i   Si 

For more complicated floor constructions, it may be a problem to deter-

mine the point impedance. A basis of calculation that considers the effect of
the lowest natural frequencies has been described by Breeuweer and Tukker
(1976). For a number of floor constructions, including ribbed decks and
plate girder systems, some measured admittances are reported in the work
of Gerretsen (1986).

11.2.2 Total impact sound pressure level

When the impact sound pressure levels for each transmission path are
determined, the total impact sound pressure level is calculated by addition
of the sound powers radiated into the receiving room:

Ln′ = 10 lg (∑10 Ln,ij /10

) (11.21)
 Flanking transmission  321

11.2.3 Combination of airborne and

impact transmission
The junction attenuation for the impact sound is assumed to be the same
as for the airborne flanking transmission. A double calculation can thus
be avoided by combining Equation 11.8 with Equation 11.19. Here, it is
simplified by neglecting the forced transmission in Equation 11.8, i.e. for
frequencies above fc,i we get:

 2fS0 
Rij + Ln,ij = 82 + R0,i − 20 lg mi + 10 lg  − ∆Lcover (11.22)
 π Siσ i ,res 

As mi is also inherent in the mass law R0,i it cancels out and we get:

 Siσ i ,res 
Rij + Ln,ij = 38 + 30 lg f − 10 lg  − ∆Lcover (11.23)
 S0 

This shows that the simple relation between the airborne sound insulation
and the impact sound pressure level for the total transmission (Equation
10.21) also applies to the individual transmission paths. The only differ-
ence is that the area correction, which is due to the flanking transmission
loss always refers to the area of the partition surface S 0. In case of vertical
transmission, S 0 = Si is the area of the floor in the source room.

11.3 JUNCTION ATTENUATION

11.3.1 Empirical model
One of the major problems when calculating sound insulation in buildings
is to provide data for the junction attenuation between the building parts.
For homogeneous massive plates, there is an extensive experimental basis,
as the data for cross junctions shown in Figure 11.6.
The junction attenuation is approximately 12 dB when the plates
have the same mass per unit area. Flanking transmission through a
structure intersected by a heavier structure is attenuated more than
12 dB and vice versa.

11.3.2 The structural transmission coefficient

Theoretical information on sound transmission between building parts
is available, but normally in another version, expressed as a structural
transmission coefficient γij. This is defined as the ratio between the
transmitted energy into plate j and the incident energy to the junction
from plate i:
322  Sound Insulation in Buildings

Pij (11.24)
γ ij =
Pinc

In Figures 11.7 and 11.8, the theoretical transmission coefficients are ren-
dered for different types of junctions. These are X-junctions or junctions
between three or two plates. Applying the numbering of Figure 11.9, the
corner joint only consists of plate 1 and plate 2, whereas the T-junction
misses either plate 3 or plate 4. The abscissa in Figures 11.7 and 11.8 is the
ratio:

m2 fc1 (11.25)
ψ =
m1fc 2

where m is the mass per unit area and fc is the critical frequency. The
index refers to the plates with numbers 1 and 2, respectively. In X- and
T-junctions, it is presumed that plate 3 equals plate 1, and plate 4 equals
plate 2.

60
st.d
d
50 a b
Junction attenuation Dv,ij, dB

c
40 transmission a b
transmission a c

30

20

10

0
0.1 0.2 0.4 0.6 0.8 1 2 4 6 8 10 20 40 60 80
Ratio of mass per unit area mcd / mab

Figure 11.6 Experimental junction attenuation for massive homogeneous junctions

shown as a function of the mass ratio (plates c–d/plates a–b). Solid line is
a regression line drawn through the measured points. Dashed lines are the
standard deviations. (Adapted from Gerretsen, E., Appl. Acoust., 12, 413–433,
1979, Figure 5. With permission from Elsevier.)
 Flanking transmission  323

0.01
0.02
Structural transmission loss, 10 lg (1/τij), dB
37 37 34 40
0.05
0.1
0.2
27 27 24 30
0.5
1
2
5
17 17 14 20 10
20
50
100

7 7 4 10 fc2
0 fc1
0 0
0
0.01 0.02 0.05 0.1 0.2 0.5 1 2 5 10 20 50 100
0.01 0.02 0.05 0.1 0.2 0.5 1 2 5 10 20 50 100
0.005 0.01 0.025 0.05 0.1 0.25 0.5 1 2.5 5 10 25 50
0.02 0.04 0.1 0.2 0.4 1 2 4 10 20 40 100 200
m f
ψ = 2 c1
m1 fc2

Figure 11.7 Theoretically determined transmission coefficients from plate 1 to plate 2

(γ12) for four different types of junctions. (Adapted from Craik, R. J. M., Appl.
Acoust., 14, 347–359, 1981, Figure 2. With permission from Elsevier.)
Structural transmission loss, 10 lg (1/τij), dB

0.01 0.02 0.05 0.1 0.2 0.5 1

2
40 40
5
10
20
30 30
50
100

20 20

fc2
10 10 fc1

0 0
0.01 0.02 0.05 0.1 0.2 0.5 1 2 5 10 20 50 100
0.02 0.04 0.1 0.2 0.4 1 2 4 10 20 40 100 200
m f
ψ = 2 c1
m1 fc2

Figure 11.8 Theoretically determined transmission coefficients from plate 1 to plate 3

(γ13) for X- and T-junctions. (Adapted from Craik, R. J. M., Appl. Acoust., 14,
347–359, 1981, Figure 3. With permission from Elsevier.)
324  Sound Insulation in Buildings

4 m4 = m2

m1 m3 = m1 m1 m3 = m1
1 3 1 3

2 m2 2 m2

Figure 11.9 Numbering the plates in a junction. Left: Cross-section, right: T-junction.

11.3.3 Relation between junction attenuation

and structural transmission loss
The energy in a wave travels with a speed that is called the group speed. In
an acoustically thick plate, the phase speed of transverse waves (Mindlin
waves) can be expressed by the speed of shear waves cs and the cross-over
frequency for shear waves fs. From Equation 8.14 and using Equation 1.30,
we get:

2
f 1 1  2f 
cB,eff = c s − + 1+  s   (11.26)
fs 2 2  f 

However, this is the so-called phase speed. In media with dispersion, the
group speed cg differs from the phase speed (Pedersen, 1995) and we get:

2
cB3 ,eff  2f  
cg = 1+  s  (11.27)
c s2  f 

or approximately:

 2⋅c = 2⋅c f / f (f < f s / 4)

 B c
cg ≅   (11.28)
 c s (f > f s / 4)

where fs is the cross-over frequency for shear waves (Equation 1.29).

The group speed is used for the calculation of the incident power of
vibrational energy on the connecting junction. Under the assumption that
 Flanking transmission  325

the vibrational energy in plate i travels equally in all directions (Rindel,

1994), the incident sound power on the connecting edge with length lij is:

1 1
Pinc =  ij mi v i2 c g ,i ≅  ij mi v i2 2cB,i (11.29)
π π

where cg,i is the group speed of the vibrations in plate i, and the approxima-
tion is valid for bending waves.
In the receiving plate j, the power transmitted from plate i must equal the
absorbed power under steady-state conditions Pij = Pj,abs and the steady-state
energy in the plate is Ej = mj Sj v 2j . Thus, we have the energy balance for
plate j:

Pij = Pj,abs = ω Ejη j = 2π fm j S j v 2j η j (11.30)

where Sj is the area and ηj is the total loss factor for plate j. The structural
transmission coefficient yields:

Pij m j v 2j S jη j m j v 2j 2.2S j
γ ij = = 2π f ⋅
2
2
= 2π ⋅
2
2
⋅ (11.31)
Pinc mi v i  ij c g ,i mi v i  ij c g ,iTj

where cg,i is the group speed of transversal waves in plate i and ℓij is the
length of the edge that adjoins the two plates. In the last expression, the
structural reverberation time Tj in plate j is inserted from Equation 2.30.
This formula can be used as the basis for measurements of the structural
transmission coefficient.
The structural transmission coefficient can be expressed in dB and is
called the structural transmission loss R s,ij:

1 m lij
Rs,ij = 10 lg = Dv ,ij + 10 lg i + 10 lg (dB) (11.32)
γ ij mj aj

where aj is the equivalent absorption length in plate j:

S jη j Sj
a j = 2π 2 f ⋅ ≅ 4.4 ⋅ π 2 (11.33)
c g ,i c g ,i ⋅ Tj

Insertion of the group speed (Equation 11.28) in Equation 11.33 yields the
structural transmission loss:
326  Sound Insulation in Buildings

  
 Dv ,ij + 10 lg mi + 10 lg  lij Tj f fc,i  + 12 dB ( f < fs,i 4)
 mj  Sj 
Rs,ij ≅ (11.34)
 mi  c s,i lij 
 Dv ,ij + 10 lg
mj
+ 10 lg  Tj  + 9 dB ( f > fs,i 4)
  cS j 

Thus, if we assume that both junction attenuation and structural rever-

beration time are frequency-independent, the structural transmission loss
increases slowly by 1.5 dB per octave for bending waves (at frequencies
below fs,i /4). At higher frequencies, the structural transmission loss becomes
independent of frequency.

11.3.4 Vibration reduction index

for rigid connections
The vibration reduction index Kij is used to characterize the junction atten-
uation in the calculation standards ISO 12354-1 and ISO 12354-2. It is
defined as the direction-averaged structural transmission loss:

1 1 1  lij lij 
Kij = (Rs,ij + Rs, ji ) = (Dv ,ij + Dv , ji ) + ⋅ 10 lg  ⋅ 
2 2 2  ai a j 

lij2
= Dv ,ij + 5lg (dB) (11.35)
ai ⋅ a j

Here, ai and aj are the equivalent absorption lengths of plates i and j, respec-
tively, and lij is the length of the connection.
As examples, we consider a rigid junction of homogeneous plates,
either as a cross-junction or a T-junction. See Figure 11.9 for definition
of K13 and K12 . Examples of the vibration reduction index suggested in
ISO 12354-1 are shown in Figures 11.10 and 11.11. These are assumed
frequency-independent.

11.3.5 Estimated attenuation for

elastic connections
The structural transmission loss was measured in the laboratory for a large
number of T-junctions by (Pedersen, 1995). Measurements were made with
two walls of either 100 mm clinker concrete (ρ = 1030 kg/m3) or 100 mm
porous concrete (ρ = 600 kg/m3) connected to a slab of 180 mm clinker con-
crete (ρ = 1700 kg/m3). For a rigid connection between two clinker concrete
walls and the slab, the transmission loss was almost frequency-independent
and in reasonable agreement with the prediction model for rigid connec-
tions in Figure 11.11.
 Flanking transmission  327

50

40
Vibration reduction index, Kij, dB

30

20

10

–10
0.1 0.2 0.5 1 2 5 10 20 50
Mass ratio m2 / m1
K13 K12

Figure 11.10 Vibration reduction index for a rigid cross-junction. (Adapted from ISO
12354-1, Building acoustics—Estimation of acoustic performance of buildings
from the performance of elements—Part 1: Airborne sound insulation between
rooms, International Organization for Standardization, Geneva, Switzerland,
2017. Figure E1.)

50

40
Vibration reduction index, Kij, dB

30

20

10

–10
0.1 0.2 0.5 1 2 5 10 20 50
Mass ratio m2 / m1

K13 K12

Figure 11.11 Vibration reduction index for a rigid T-junction. (Adapted from ISO 12354-
1, Building acoustics—Estimation of acoustic performance of buildings from the
performance of elements—Part 1: Airborne sound insulation between rooms,
International Organization for Standardization, Geneva, Switzerland, 2017.
Figure E3.)
328  Sound Insulation in Buildings

With two elastic interlayers, e.g. one below and the other one above the
slab, the transmission loss becomes strongly frequency dependent, increas-
ing with frequency above a certain frequency, which according to Pedersen
(1995) can be estimated from:

−3/ 2
 ρ1ρ 2 h1 
−6 
f1 = 2.5 ⋅ 10  d  (Hz) (11.36)
 G w

where

ρ1 is the density of plate 1 (kg/m3).

ρ2 is the density of plate 2 (kg/m3).
G is the shear modulus of the elastic interlayer (Pa).
d is the thickness of the elastic interlayer (m).
h1 is the thickness of plate 1 (m).
w is the width of the elastic interlayer (m).

h1

ρ1

G d

ρ2 w

ρ1

h1

Figure 11.12 Junction with two identical thin elastic interlayers. (Adapted from Pedersen,
D. B., Appl. Acoust., 46, 285–305, 1995. With permission from Elsevier.)

The parameters in the above equation are shown in Figure 11.12.

At frequencies below f1, the transmission loss is similar to that for a rigid
connection and frequency-independent. An empirical formula for the struc-
tural transmission loss is:
 Flanking transmission  329

Rs,ij = Rs,ij ,rigid + n ⋅ ∆Rs . (11.37)

where R s,ij,rigid is the structural transmission loss for the equivalent rigid
junction, n is the number of resilient layers in the junction (normally 1 or 2)
(Pedersen, 1995) and ΔR s is:

 0 dB (f < f1)

∆Rs ≅  f  (11.38)
 10 lg   (dB) (f > f1)
  f1 

With two elastic interlayers, the transmission loss increases by 6 dB per
octave above the frequency f1. In the case of only one interlayer, the improve-
ment of the transmission loss will be only about half the value for two inter-
layers. An improvement by 9 dB is obtained with either of these measures:

• halving the shear modulus G

• doubling of the density of the plates (ρ1 and ρ2)
• doubling of the thickness of the interlayers d
• doubling of the ratio between thickness of plate 1 and the width of
the interlayers h1/w

An example of a measurement result is shown in Figure 11.13 and com-

pared with the simple prediction method using Equations 11.37 and 11.38.
It is a T-junction with two 100 mm porous concrete walls and one 180 mm
clinker concrete slab, having two polyurethane interlayers, one below and
one above the slab. The measurement was from wall 1 to wall 3, referring
to the numbering in Figure 11.9. Speed of shear waves in the porous con-
crete wall was 750 m/s.
Annex E.3.4 in ISO 12354-1 has a similar formula for the vibration
reduction index Kij with flexible interlayers. It is suggested that the attenu-
ation depends on the load on the resilient layer. For one flexible interlayer,
the increase with frequency is 3 dB or 6 dB per octave if the load is >750 kN/
m 2 or <80 kN/m 2 , respectively. ISO 12354-2 deals with flanking transmis-
sion of impact sound in a similar way.

11.4 EXAMPLE

If the measured sound insulation is less than expected, it can be very dif-
ficult to point at the reason. However, the measures needed for an improve-
ment depend very much on which transmission path is most important. An
330  Sound Insulation in Buildings

80

70
Structural transmission loss Rs ,ij, dB

60

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency, Hz
Measured Corrected for Mindlin waves Simple estimate

Figure 11.13 Frequency dependency of the structural transmission loss with two elastic
interlayers. Measurement results without and with correction for group
speed of Mindlin waves (Equation 11.28). (Adapted from Pedersen, D. B.,
Appl. Acoust., 46, 285–305, 1995, Figure 4. With permission from Elsevier.)

obvious but rather complicated and time-consuming method is the measure

the sound intensity radiated from each surface into the receiving room.
In this case, the problem was the airborne sound insulation between
dwellings in the vertical direction. The cross-section of floor and inter-
nal wall construction is seen in Figure 11.14 and the measurement results
for both airborne and impact sound insulation are seen in Figure 11.15.
The single-number values are Rw′  = 49 dB and Ln,w ′  = 59 dB. The minimum
requirement for airborne sound insulation was Rw′  = 55 dB, so the result was
6 dB below the requirement and something had to be done about it.
If the floor construction was the cause of the problem, it might be a
solution to mount a suspended ceiling or to change the floor to a floating
floor. If flanking transmission was the cause, these measures would not
help; instead, something should be done with the walls.
In this case, it was helpful to look at the sum R′ + Ln′ (Figure 11.16). In
Section 10.4, we learned that the sum should follow approximately the the-
oretical straight lines also shown in Figure 11.16, and if that failed, this was
an indication of flanking transmission or leakage, i.e. some transmission
 Flanking transmission  331

Cast under wall

Cast
Wooden f loor on joists

75 mm mineral wool

220 mm clinker concrete

slab, 1700 kg/m3

Double wall, 100 mm light

clinker concrete, 1400 kg/m3
100 mm mineral wool

Figure 11.14 Cross-section of the two-story building.

80
f0

70

60

50
R' and L'n, dB

40

30

20
fc

10

0
50 100 200 500 1000 2000 5000
Frequency, Hz
R΄ before R΄ after L΄n

Figure 11.15 Measured impact and airborne sound insulation in vertical direction.

Indicated frequencies are assumed resonance frequency f 0 of the flooring
and critical frequency fc of the internal wall.

paths that only affect the airborne sound transmission and not the impact
sound transmission. In the current case, we see a clear deviation of the
sum curve at frequencies above 200 Hz and especially around 315 Hz. This
332  Sound Insulation in Buildings

140

f0
130

120
Sum R' + L'n, dB

110

100

90
fc

80
50 100 200 500 1000 2000 5000
Frequency, Hz

R' + L'n before R' + L'nafter Theory Resilient floor

Figure 11.16 Sum of measured R′ + L′ n before and after mounting of gypsum board pan-
els on the internal walls. Theoretical curves from Figure 10.6b are shown
for comparison.

observation points to the walls, which have a critical frequency around

380 Hz. (From Table 1.1 we have the coincidence number for light clinker
concrete Kc = 38.) Therefore, the sum-curve gives a strong indication that
flanking transmission via the 100 mm light clinker concrete walls is the
main cause of the problem.
A plasterboard panel was mounted in front of the light clinker concrete
walls and the sound insulation was measured again. The improved airborne
sound insulation is seen in Figure 11.15, and the result was Rw′  = 55 dB, just
as required. The new sum-curve also shown in Figure 11.16 indicates that
there is still some flanking transmission, but for practical reasons, it was
not possible to mount the plasterboard panels on all internal wall surfaces.

REFERENCES

R. Breuwer and J.C. Tucker. 1976. Resilient mounting systems in buildings. Applied
Acoustics 9, 77–101.
R.J.M. Craik. 1981. Damping of building structures. Applied Acoustics 14, 347–359.
E. Gerretsen. 1979. Calculation of the sound transmission between dwellings by
partitions and flanking structures. Applied Acoustics 12, 413–433.
 Flanking transmission  333

E. Gerretsen. 1986. Calculation of airborne and impact sound insulation between

dwellings. Applied Acoustics 19, 245–264.
ISO 12354-1. 2017. Building acoustics—Estimation of acoustic performance of
buildings from the performance of elements—Part 1: Airborne sound insula-
tion between rooms. International Organization for Standardization, Geneva,
Switzerland.
ISO 12354-2. 2017. Building acoustics—Estimation of acoustic performance of
buildings from the performance of elements—Part 2: Impact sound insula-
tion between rooms. International Organization for Standardization, Geneva,
Switzerland.
J. Kristensen and J.H. Rindel. (1989). Bygningsakustik—Teori og praksis (Building
acoustics—Theory and practice, in Danish). SBI-Anvisning 166. Danish
Building Research Institute, Hørsholm, Denmark.
D.B. Pedersen. 1995. Estimation of vibration attenuation through junctions of build-
ing structures. Applied Acoustics 46, 285–305.
J.H. Rindel. 1994. Dispersion and absorption of structure-borne sound in acousti-
cally thick plates. Applied Acoustics 41, 97–111.
Chapter 12

Measurement methods

Measurements related to building acoustics include noise measurements,

airborne and impact sound insulation, reverberation time, vibrations, loss
factor, radiation efficiency and junction attenuation. Measurements can
be made with various degrees of precision, and some measurement stan-
dards distinguish between three grades: survey, engineering and precision.
The engineering grade measurements are most relevant for field measure-
ments in buildings, but in some cases it may also be useful to take quick
measurements at the survey grade as an indication of the acoustical qual-
ity. Precision grade measurements are intended for laboratory tests and
research, where the measurement uncertainty must be as low as possible,
and time is of minor importance. The focus in this chapter is on field mea-
surements in buildings at the engineering grade. Details on the methods are
given in the relevant measurement standards and the most important ones
are listed in Table 12.1.

12.1 SOUND PRESSURE LEVEL AND

LOW-FREQUENCY NOISE

This section describes the principles of sound pressure level measurements

in rooms when the sound source is equipment for building services or some
outdoor noise source.

12.1.1 Microphone
The sound pressure level in a room is measured with a calibrated sound
level meter or a calibrated measurement microphone connected to an anal-
yser. The microphone must be selected to have a flat frequency response
in the frequency range of the measurement, and the electrical noise floor
must be sufficiently low to allow correct measurements. A microphone
with a big diaphragm, e.g. a 1 inch microphone. has a high sensitivity and

335
336  Sound Insulation in Buildings

Table 12.1  I nternational Organization for Standardization (ISO) and American

Society for Testing and Materials (ASTM) standards for
measurements of sound insulation and reverberation time
Building elements Performance
measured in a measurements in the
laboratory field
Airborne sound insulation, ISO 10140-2 ISO 16283-1
walls and floors ASTM E90 ASTM E336
Airborne sound insulation, ISO 10140-2 ISO 16283-3
windows and façades ASTM E90 ASTM E966
Impact sound insulation ISO 10140-3 ISO 16283-2
ASTM E492 ASTM E1007
Reverberation time ISO 3382-2 ISO 3382-2
ASTM E2235 ASTM E2235

an upper frequency limit around 8 kHz, which is sufficient for most cases
in building acoustics. However, a smaller ½ inch microphone is often pre-
ferred although the noise floor is higher, because it is less expensive and
easier to handle.
The frequency response should be flat for diffuse field incidence when
measuring in a room, i.e. a pressure microphone should be used, not a free
field microphone.

12.1.2  Frequency weighting

Measurements are normally made in one-third octave bands covering the
frequency range from 50 Hz to 5000 Hz. In case of measuring noise from
service equipment in buildings, octave bands are used as a minimum in the
frequency range from 125 Hz to 2000 Hz for survey measurements (ISO
10052) or for engineering grade measurements in an extended frequency
range from either 31.5 Hz or 63 Hz to 8000 Hz (ISO 16032). As single-
number values, the A-weighted and C-weighted sound pressure levels can
be calculated using Equation 1.16.
Low-frequency noise can be particularly annoying for the dwellers in
buildings, although the noise source may be far away from the building.
Sometimes, it is difficult to identify the origin at low-frequency noise.
One evaluation method that works satisfactorily was developed in the
Netherlands (Vercammen, 1992) and adapted in Denmark (Jakobsen,
2001; Poulsen and Mortensen, 2002). The one-third octave bands are mea-
sured from 10 Hz to 160 Hz and adjusted with the A-weighting filter correc-
tions (Table 1.3) and then the low-frequency noise level LpA,LF is calculated
for this frequency range using Equation 1.16.
 Measurement methods  337

12.1.3  Averaging time

The energy averaged sound pressure level is determined by:

 1 Tm

L = 10 lg 
 Tm ⋅ p02 ∫
0
p2 dt  dB (12.1)


where
p is the sound pressure as a function of time (Pa).
p 0 = 20 μPa is the reference sound pressure.
Tm is the integration time (s).

Measuring the sound pressure level of service equipment, the averaging

time must be long enough to cover typical time variations, but not less
than 30 s. For sound insulation measurements, the recommended averaging
times depend on the frequency (see Section 12.2.2).

12.1.4  Correction for background noise

For the measurement of service equipment noise, the background noise
level should be at least 4 dB, and preferably more than 10 dB, below the
measured sound pressure level in each frequency band.
If the background noise level is between 4 dB and 10 dB below the mea-
sured sound pressure level, the result is adjusted by:

( )
L = 10 lg 10Lsb /10 − 10Lb /10 dB (12.2)

where

L is the adjusted sound pressure level.

L sb is the measured sound pressure level (signal and background noise
combined).
Lb is the sound pressure level of the background noise.

When the difference is 4 dB, the correction is 2.2 dB. If the difference is less
than 4 dB, the correction 2.2 dB is applied and it must be clearly marked
that the result is influenced by background noise. For sound insulation
measurements, the background noise level should be at least 6 dB below the
measured sound pressure level (see Section 12.2.1).
338  Sound Insulation in Buildings

12.1.5  Measurement positions and spatial averaging

The recommended choice of measurement positions for the sound pressure
level in a room is very different for frequencies above or below the Schroeder
limiting frequency (see Equation 4.21). In Table 12.2, it is shown which vol-
ume corresponds to a certain limiting frequency, fg, assuming the reverbera-
tion time to be around 0.5 s as in a typical furnished room in a dwelling, or
around 0.8 s as in an unfurnished or sparsely furnished room. It is seen that
the limiting frequency is around 200 Hz for a 50 m3 furnished room.
Above the limiting frequency, statistical methods apply and the positions
represent samples from a more or less diffuse sound field. The ideal would
be to use a large number of microphone positions to find the grand aver-
age over the entire volume. However, this is not possible in practice, where
a rather limited number of positions must be selected. Alternatively, the
squared sound pressure can be averaged using a moving microphone fol-
lowing some path through the middle part or the room.
The single measurements can be considered uncorrelated if the distance
from other positions is at least half a wavelength, e.g. 0.86 m at 200 Hz
(Table 12.2). Near reflecting boundaries, the sound pressure increases
due to interference. This can actually be considered as the microphone
being close to the reflected image of the microphone and, thus, the posi-
tion should be at least a quarter wavelength from reflecting surfaces, e.g.
0.43 m at 200 Hz. Positions should be distributed randomly over the centre
part of the room, keeping a minimum distance from the room boundaries.
In accordance with statistical principles, the measurement uncertainty is
reduced if the number of uncorrelated positions is increased.
Below the Schroeder limiting frequency, the sound field in the room is deter-
ministic, dominated by a few room modes, as described in Section 4.1. This
implies that preferred positions are near the corners, where the sound pressure

Table 12.2  R
 elation between the Schroeder limiting frequency and the room
volume when the reverberation time is either 0.5 s or 0.8 s
V (m3)
fg (Hz) T = 0.5 s T = 0.8 s λ/2 (m)
100 200 320 1.72
125 125 200 1.38
160 80 125 1.08
200 50 80 0.86
250 32 50 0.69
315 20 32 0.55
400 13 20 0.43
500 8 13 0.34
Note: The minimum distance between uncorrelated microphone positions in a diffuse sound


field (λ/2) is also given.

 Measurement methods  339

level is well defined, whereas the centre region should be avoided because
extremely low sound pressure levels occur along the nodal lines. Since statisti-
cal methods do not apply at these low frequencies, it is impossible to avoid
correlation between the positions, and in contrast to the high-frequency range,
increasing the number of positions does not affect the measurement uncertainty.
Measuring the sound pressure level from building service equipment in
accordance with ISO 16032, three microphone positions are used; one in
a corner and two distributed in the reverberant field. The corner position
should be about 0.5 m from the nearest surfaces. Initial measurements are
made in all corners using the C-weighting filter, and the corner with the
highest C-weighted sound pressure level is chosen as position 1. Positions
2 and 3 shall be at least 1.5 m from position 1 and from any sound source.
The distance from room surfaces shall be at least 0.75 m. The result of mea-
sured sound pressure levels in the three microphone positions is calculated
as the energetic average using Equation 1.34.

12.1.6  Correction for reverberation time

It is known that the sound pressure level in a depends on the absorption
area, i.e. on the reverberation time and the volume. Therefore, the mea-
sured sound pressure level should be either standardized or normalized.
The standardized sound pressure level L nT is adjusted to a reverberation
time of T0 = 0.5 s:

T
LnT = L − 10 lg
T0
(dB ) (12.3)

It is generally assumed that the reverberation time in furnished dwellings is

around 0.5 s, and this has been confirmed by Takala and Kylliäinen (2013).
The normalized sound pressure level L n is adjusted to an equivalent
sound absorption area of A 0 = 10 m 2:

A0 c ⋅ A0 ⋅ T
Ln = L − 10 lg
A
= L − 10 lg
55.3 ⋅ V
(dB ) (12.4)

where

L is the measured sound pressure level (dB).

T is the reverberation time (s).
V is the volume (m3).
c = 344 m/s is the speed of sound in air.

Since the reverberation time is frequency dependent (more or less), the

corrections should be made in each frequency band (octave or one-third
octave) before the A-weighted or C-weighted levels are calculated.
340  Sound Insulation in Buildings

12.2 AIRBORNE SOUND INSULATION

BETWEEN ROOMS

12.2.1 Generation of sound
A loudspeaker is used to generate the sound, which must be at a sufficiently
high level in order to get a sound pressure level in the receiving room with-
out the influence of background noise. The directivity of the loudspeaker
should be close to omnidirectional, and the frequency response should
cover the relevant measurement range, which is at least from 100 Hz to
3150 Hz, but preferably from 50 Hz to 5000 Hz in one-third octave bands.
The traditional measurement technique uses stochastic noise (white or
pink), which can either be emitted as a broad-band signal or as a band-
limited signal with a bandwidth of at least one-third octave. The latter is, of
course, more time consuming, but has the advantage of allowing a higher
sound pressure level to be generated without overloading the loudspeaker.
Airborne sound insulation can also be measured using sound intensity, see
ISO 15186 (Part 1 to 3) and Machimbarrena and Jacobsen (1999).
More advanced measurement techniques use deterministic sound sig-
nals, either MLS (maximum-length-sequence) signal or swept-sine signals
in combination with either correlation signal processing or deconvolution
(see ISO 18233). The resulting impulse responses are squared and inte-
grated in order to get the stationary sound pressure level. Compared to the
traditional method, both methods have the advantage of suppressing the
influence of background noise. The swept-sine method has a number of
advantages over the MLS method:

• Nearly immune against harmonic distortion

• Higher excitation levels may be used, due to a lower crest factor
• More robust against time variance

The crest factor of a sound signal is defined as the ratio between the peak
value and the root-mean-square (RMS) value of the signal. A harmonic signal
that is composed of N sine functions has the crest factor √(2N), and a single
sine function has the crest factor √2 ≈ 1.414. White noise and pink noise have
a much higher crest factor, which means a higher risk that the loudspeaker
can be overloaded.
While distortion from the loudspeaker is not really a problem with the tra-
ditional measurement technique, it is fatal to the MLS technique. With the
swept-sine technique, some (but not all) of the harmonic distortion is separated
from the impulse response and does not influence the result. Both methods
require time-invariant systems, i.e. nothing must move in the rooms during the
measurements; even air movements may cause errors in the measurements and,
of course, moving microphones are out of question. While time invariance
is mandatory with MLS measurements, the swept-sine is much more robust
 Measurement methods  341

Table 12.3  I mproved SNR as function of sweep duration using the swept-sine
technique
Duration of sweep (s) 60 336 762
Duration per one-third octave (s) 2.9 16 32
Improved SNR (dB) 14 21 24

Note: The reference for SNR is here set to fast (0.125 s).



to this. Some years ago, when the computer capacity was more restricted, the
MLS technique had the advantage of requiring less memory, but since this is no
longer an issue, the many advantages of the swept-sine technique means that
the MLS technique is obsolete for building acoustic applications.
With the swept-sine technique, the advantage of improving the signal-
to-noise ratio (SNR) is connected to the length of the sweep signal. Only a
single sweep is used in each measurement position, which is in contrast to
the MLS technique, which uses the averaging of several repeated sequences.
If we compare measurements with the averaging time Fast (0.125 s), the
improved SNR with different sweep durations are shown in Table 12.3.
As an example, the sound level difference between two rooms has been
measured with different techniques and the results are shown in Figure 12.1.
The background noise in one-third octaves was around 43 dB at 125 Hz
to 200 Hz, decreasing towards lower and higher frequencies, being around
30 dB at 50 Hz to 80 Hz and around 25 dB at 2500 Hz to 5000 Hz. The influ-
ence of the background noise is worst using the broadband excitation and
parallel analysis. Some improvement was obtained using band-limited noise
excitation because the loudspeaker could be set to a level about 10 dB higher.
The best result was obtained with the swept-sine; the excitation signal could
be increased another 3 dB, but most importantly, the influence of the back-
ground noise in the receiving room was efficiently suppressed. The differ-
ence between the methods is clearly seen at medium and high frequencies.
The use of swept sine for measuring airborne sound insulation has been
further studied by Satoh et al. (2011). Using very long sweeps (10 min per
octave) and a technique for subtraction of background noise, they could
measure sound pressure levels 30 dB below the background noise with less
than 1 dB error. Løvstad et al. (2014) describe an example of the practical
application of the swept-sine technique for a rather extreme case where the
sound insulation was 95 dB at 500 Hz.

12.2.2 Measurement positions and spatial averaging

The requirements for measurement of airborne sound insulation in buildings
in ISO 16283-1 imply that two source positions shall be used in the source
room, the distance from room boundaries being at least 0.5 m and at least
1.0 m from the partition wall separating the source and receiving rooms.
 342  Sound Insulation in Buildings

80

70

60
Level difference, D (dB)

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)

Figure 12.1 Level difference between two rooms measured with different techniques.
Solid line: Swept-sine excitation. Dashed line: Serial excitation with noise
in one-third octave bands. Dotted line: Parallel excitation with pink noise.
(Adapted from Norsonic, Swept sine measurement technique using the Nor121
analyser. Application Note, Ed. 3, 10.06, Norsonic, Lierskogen, Norway,
2006. Figure 3.)

If measuring in vertical direction, and if the source is in the upper room,

the source position shall be at least 1.0 m above the floor. However, it is
advisable that measurements in the vertical direction are made with the
lower room as the source room in order to avoid the measurement to be
influenced by structure borne transmission from the loudspeaker stand.
If the rooms are of different volume, the larger room should normally be
chosen as the source room.
Instead of measuring with one loudspeaker in two different positions,
it is possible to use two loudspeakers simultaneously, and then calculate
the energy-averaged sound pressure levels in the source a nd receiving
rooms. If necessary, the receiving room levels are corrected for back-
ground noise and, finally, the reverberation time correction is applied
to get either the standardized level difference D nT or the apparent sound
reduction index R′.
 Measurement methods  343

If a single loudspeaker is used, the measurements must be repeated with

the source moved to a different position. In this case, the standardized level
difference DnT or the apparent sound reduction index R′ must be measured
and calculated for each source position separately and, then, the result is
averaged using one of these formulas:

 1 m 
DnT = −10 lg 
 m ∑
j =1
10− DnT , j /10  ( dB ) (12.5)


 m 
R′ = −10 lg 
1
∑ 10− Rj /10  (dB) (12.6)
′
m 
 j =1 
where
m is the number of loudspeaker positions.
DnT,j is the standardized level difference for loudspeaker position j.
Rj′ is the apparent sound reduction index for loudspeaker position j.
The microphones shall be calibrated as for normal sound level measure-
ments before the measurements, although strictly speaking, it is sufficient
that the same microphone is used in the source room and in the receiving
room, since the measurement result is taken from the level difference. In
each room, the spatially averaged sound pressure level must be measured,
and this can be done in three different ways:

• Fixed microphone positions, at least five in each room with at least

0.7 m between positions (Figure 12.2). No two positions shall lie in
the same plane relative to the room boundaries.
• Mechanically moved microphone with radius at least 0.7  m
(Figure 12.3). The plane of the traverse shall be inclined at least 10°.
• Manually scanned microphone, following a certain path (Figure 12.4).

≥ 0.5 m

≥ 1.0 m ≥ 0.7 m
≥ 0.5 m ≥ 0.5 m
≥ 1.0 m

≥ 1.0 m
≥ 0.5 m

Figure 12.2 Sketch of a source position and microphone positions in the source room
with minimum distances indicated.
344  Sound Insulation in Buildings

r ≥ 0.7 m

θ ≥ 10°
≥ 1.0 m

≥ 0.5 m ≥ 0.5 m

Figure 12.3 Sketch of mechanically moving microphone with minimum radius, inclination

and distances from source and room boundaries.

220° rotation

m Microphone
.9
–0
m
0.3
2.0 m

0.5 m

Floor level

Figure 12.4 Manual scanning with the microphone mounted on a rod and following a cylin-
drical path. (Modified from ISO 16283-1, Acoustics – Field measurements of sound
insulation in buildings and of building elements – Part 1: Airborne sound insulation,
International Organization for Standardization, Geneva, Switzerland. Figure 1c.)

The intention is to achieve an equivalent of five uncorrelated microphone

positions for frequencies above 200 Hz. This is obtained for the mechan-
ically moved microphone with a radius of 0.7 m (see Chapter 13, Table
13.4). The averaging time shall equal the time for one rotation, which is at
least 30 s at frequencies 100 Hz to 5000 Hz and at least 60 s at low frequen-
cies 50 Hz to 80 Hz.
For the fixed positions, the minimum distance of 0.7 m corresponds to
uncorrelated positions above 250 Hz (the distance must be half a wave-
length). This means that with five fixed positions, the equivalent number
of uncorrelated positions is less than five at frequencies below 250 Hz.
Increasing the minimum distance between microphone positions to 0.85 m
would comply with uncorrelated positions above 200 Hz (Table 12.2). The
averaging time shall be:
 Measurement methods  345

≥ 15 s at frequencies 50 Hz to 80 Hz

≥ 6 s at frequencies 100 Hz to 400 Hz
≥ 4 s at frequencies 500 Hz to 5000 Hz

The manually scanning method was introduced as an option in ISO 16283-1.

Four different patterns are suggested, and the most efficient among these is
the cylindrical type, as shown in Figure 12.4.
The microphone is mounted at the end of a rod, 0.3 m to 0.9 m long. The
path starts about 0.5 m above the floor and approximately 90° to one side;
the rod is swept in a circular path parallel to the floor to cover an angle
of approximately 220°. The sweep continues vertically upwards along a
straight line until the microphone is about 2 m above the floor, but at least
0.5 m below the ceiling. Then, the movement continues with another cir-
cular sweep, before descending to the starting point. A speed of the micro-
phone shall be as constant as possible, with a maximum of 0.25 m/s. The
minimum averaging times are the same as for the mechanically moved
microphone, i.e. 30 s at 100 Hz and above and 60 s below 100 Hz. This pro-
cedure is equivalent to five uncorrelated positions at 100 Hz using a 0.3 m
rod, and eight positions using a 0.9 m rod (Hopkins, 2015, Figure 2).

12.2.3 Low-frequency procedure
According to ISO 16283-1, a special low-frequency procedure shall be used
for the 50 Hz, 63 Hz and 80 Hz one-third octave bands when the room
volume is smaller than 25 m3. This applies to both the source room and
the receiving room. However, it would make sense to apply this method
more generally to the frequencies that are below the Schroeder limiting fre-
quency. The principle is that measurements made in the central zone of the
room (the default procedure) are combined with the highest sound pressure
level that occurs in a room corner. The microphone position in the corner
is at a distance of 0.3 m to 0.4 m from each room boundary (Figure 12.5).

dz

dy
dx

Figure 12.5 Position of a microphone near a corner. The distances dx, dy and dz should
all be within 0.3 m to 0.4 m.
346  Sound Insulation in Buildings

Actually, measurements shall be made in four corners, two at floor level

and two at ceiling level, but only the highest of these four results is used.
Which position gives the highest results may change with frequency. If a
single loudspeaker is used, the measurements must be repeated with the
source moved to a different position, and the corner sound pressure level
in each frequency band is calculated as the energy-averaged sound pressure
level using Equation 1.34. The low-frequency energy-averaged sound pres-
sure level LLF is calculated in each frequency band by combining the sound
pressure level from the default procedure, L, and the highest sound pres-
sure level from the corners, L corner, using:

1 Lcorner
2
L

LLF = 10 lg  ⋅ 10 10 + ⋅ 1010  (dB) (12.7)
3 3 

LLF is determined for each loudspeaker position and the resulting DnT or R′
values for 50 Hz, 63 Hz and 80 Hz are averaged as for the default procedure.
However, the reverberation time is measured in the octave band 63 Hz, and
the result is applied in all three one-third octave bands 50 Hz, 63 Hz and
80 Hz. The reason for this pragmatic method is that the modal density may
be very low below 100 Hz, especially in small volumes and, thus, measure-
ments in one-third octave bands may be unreliable or even impossible.

12.3 WINDOWS AND FACADES

The purpose of these measurements can be either to evaluate the protection

of the façade against outdoor noise sources or to evaluate the transmis-
sion loss of a particular façade element, e.g. a window. So, ISO 16283-3
describes two methods, a global method and an element method.
The sound source used for the measurement can be either the existing
traffic noise or a loudspeaker. Traffic noise is the preferred source for the
global method, while the measurement using a loudspeaker is preferred for
the element method.

12.3.1 The global method

The outdoor sound pressure level is measured in a distance of 2.0 m from
the façade, and in the receiving room, the time- and space-averaged sound
pressure level is measured in five fixed positions or using a moving micro-
phone, exactly as when measuring the sound insulation between two
rooms. The measured level difference can be either normalized (adjusted
to 10 m 2 absorption area) or standardized (adjusted to 0.5 s reverberation
time). The standardized level difference is:

T
D2m,nT = L1, 2m − L2 + 10 lg (dB) (12.8)
T0
 Measurement methods  347

where
L1,2m is the sound pressure level 2.0 m in front of the façade.
L 2 is the spatially averaged sound pressure level in the receiving room.
T is the reverberation time in the receiving room (s).
T0 = 0.5 s is the reference reverberation time.

Equation 12.8 is used both for traffic noise and for loudspeaker measure-
ments. Using a loudspeaker, the position of the loudspeaker is in a distance
of r ≥ 7 m from the centre of the test specimen or at least D = 5 m in front of
the façade (Figure 12.6). The angle of incidence shall be 45° ± 5° and the
loudspeaker can be placed either on the ground (preferred) or as high over
the ground as possible.

12.3.2 The element method

The outdoor sound pressure level is measured close to the façade (Figure
12.7). If the sound source is a loudspeaker (the preferred method), it should
be placed so that the angle of incidence is close to 45°, as shown in Figure
12.6. The distance shall be r ≥ 5 m from the centre of the test specimen or at
least D = 3.5 m in front of the façade. The result of the measurement is the
apparent sound reduction index:

S
′ ° = L1,s − L2 + 10lg
R45 − 1.5dB (12.9)
A

where
L1,s is the sound pressure level on the test specimen.
L 2 is the spatially averaged sound pressure level in the receiving room.

θ = 45° ± 5°

Figure 12.6 Position of the loudspeaker in front of the façade. For the global method,
r ≥ 7 m and D ≥ 5 m. For the element method, r ≥ 5 m and D ≥ 3.5 m.
348  Sound Insulation in Buildings

(a) ≤ 10 mm (b) ≤ 3 mm

Figure 12.7 Position of the microphone on the test surface for the measurement
of L1,s. (a) Microphone with diaphragm perpendicular to the surface. (b)
Microphone with diaphragm parallel with the surface.

S is the area of the test specimen (m 2).

A is the equivalent absorption area of the receiving room (m 2).

If, for practical reasons, the element loudspeaker method cannot be used,
it is possible to use road traffic as a noise source. Then, it is assumed that
the angles of incidence cover a wide range and the formula for the apparent
sound reduction index is:

S
′ s = L1, s − L2 + 10 lg
Rtr, − 3 dB (12.10)
A

where the symbols are the same as in Equation 12.9.

A major problem with measurements using traffic noise as the sound
source is that the background noise in the receiving room can make mea-
surements impossible at high frequencies. This is a lesser problem using
a loudspeaker as the sound source, especially if the swept-sine method is
used. In that case, fixed microphone positions (at least five) must be used in
the receiving room.

12.4 IMPACT SOUND INSULATION

The impact sound insulation is measured in a building using a tapping

machine, as described in ISO 16283-2. The tapping machine is placed on
the floor under test in at least four positions, randomly distributed, but
not closer than 0.5 m to any wall. The tapping machine has five ham-
mers in a line, and in case of floor constructions with beams or ribs, the
line through the hammers should be oriented 45° to the direction of the
 Measurement methods  349

beams or ribs. If the floor under test is larger than 20 m 2 , the number of
tapping machine positions may be increased, and if the area of the floor
is larger than 50 m 2 , the number of tapping machine positions should be
eight.
The measurement of spatially averaged sound pressure level and the
position of microphones in the receiving room follow the same rules as
for airborne sound insulation (see Section 12.2). If fixed microphone posi-
tions are used, the number of positions shall equal the number of tapping
machine positions (or a multiple of the number of tapping machine posi-
tions). Measurements shall be made in at least two different microphone
positions for each tapping machine position, if the number of tapping
machine positions is less than six. If six or more tapping machine positions
are used, it is sufficient to measure in one microphone position for each
tapping machine position.
Measurements of impact sound insulation should include the low fre-
quencies down to the 50 Hz one-third octave band. For frequencies below
100 Hz, the low-frequency procedure using corner positions apply (see
Section 12.2). Averaging times for the measurements and correction for
background noise are also the same as for airborne sound insulation.
For each tapping machine position, the result is calculated either as the
standardized impact sound pressure level (Equation 12.3) or as the normal-
ized impact sound pressure level (Equation 12.4). Finally, the averaging
over all tapping machine positions is made as the energetic average using
one of these formulas:

 1 m 
L′ nT = 10 lg 
 m ∑10
j =1
Ln′ T , j /10
 (dB) (12.11)


 1 m 
∑
′

L′ n = 10 lg  10Ln, j /10  (dB) (12.12)

 m j =1


where
m is the number of tapping machine positions.
Ln′ T , j 
is the standardized impact sound pressure level for tapping
machine position j.
Ln,′ j is the normalized impact sound pressure level for tapping machine
position j.
The microphone shall be calibrated as for normal sound level measure-
ments before the measurements. This is particularly important for the mea-
surement of impact sound pressure level, as it is the actual sound pressure
level that is measured, not the difference between two levels as for airborne
sound insulation.
350  Sound Insulation in Buildings

12.5 REVERBERATION TIME

The measurement of reverberation time in the receiving room is a part of

most building acoustic measurements. The procedure is described in ISO
3382-2 and two different methods are available, the interrupted noise
method and the integrated impulse response method. Independent of the
method, the decay curve is evaluated between 5 dB and 25 dB below the
initial level, and the reverberation time is denoted as T20 in order to indicate
the evaluation range 20 dB. Since the reverberation time is defined at the
time needed for a 60 dB decay, T20 is three times the time interval between
the start and stop of the evaluation range.

12.5.1 Measurement positions
For an engineering grade measurement, the number of source-micro-
phone combinations should be at least six, e.g. two source positions
combined with three microphone positions. However, when the result is
used only as a correction term to other engineering-level measurements,
only one source position and three microphone positions are required.
Also, under these conditions, a rotating microphone boom may be used
instead of multiple microphone positions, but only for the interrupted
noise method.
ISO 3382-2 recommends that in small domestic rooms, one source posi-
tion should be in a corner. Microphone positions should be at least half a
wavelength apart and a quarter wavelength from the boundaries. If a rotat-
ing microphone boom is used, the radius of the boom shall be at least 0.7 m.

12.5.2 The interrupted noise method

The traditional method uses white or pink noise emitted from a loudspeaker
and interrupted after a stationary sound level has been built up in the room.
Since the signal is stochastic, the measurements must be repeated, and at
least two decays must be measured in each source-receiver combination.
Background noise may be a problem if the loudspeaker cannot provide a
sufficiently high level, which should be at least 35 dB above the background
noise in each frequency band. Although the use of broadband noise and
parallel filters for the analyses is very efficient, it may be necessary to use
band-limited noise in one-third octave bands in order to get sufficient signal-
to-noise level.

12.5.3 The integrated impulse response method

The integrated impulse response method has become a very attractive
alternative to the traditional method after the introduction of the swept-
sine technique. A single sweep can cover the entire frequency range and
 Measurement methods  351

the background noise can be suppressed as much as needed, simply by

choosing a sufficiently long sweep time. Since the signal is ergodic, there
is no need to repeat measurements (they will be identical); one measure-
ment in each position is sufficient. Only fixed source and microphone
positions can be used because the technique requires time-invariant
systems.
The practical application of the measurement method is discussed by
Christensen et al. (2013). They show signal-to-noise level of up to 100 dB
with very long sweeps. With 40-s sweep, they got nearly 90 dB signal-to-
noise level, which is more than sufficient.

12.5.4 Lower limits for reverberation time

There are several problems with measuring reverberation times at low
frequencies, especially the short reverberation times in typical furnished,
domestic rooms:

• ringing of the one-third octave band filters

• time constant of the detector
• low modal density, especially in small room volumes
• insufficient signal-to-noise ratio

Normally a filter with bandwidth B has an impulse response with a

reverberation time of its own and, of course, this may interfere with the
measured decay curve of the room. As shown by Jacobsen and Rindel
(1987), the following requirements should be fulfilled for the reverbera-
tion time T:

BT > 16 (T > 70/f for one-third octave bands) (12.13)

T > 2 ⋅ Tdet (12.14)

where f is the centre frequency of the filter and Tdet is the reverberation
time of the detector, assuming exponential averaging of the decaying sound
pressure. This implies a lower limit for measurable reverberation times,
which is 0.7 s at 100 Hz and 1.4 s at 50 Hz (Table 12.4). However, it was
suggested by Jacobsen and Rindel (1987) to overcome these problems by
using a time-reversal technique, either by recording the decay in the room
and playing it backwards through the filters or by constructing the filters
with a time-reversed impulse response meaning a slow built-up and fast
decay. With this technique, the minimum requirements for the reverbera-
tion time are changed to:

BT > 4 (T > 17.5 / f for one-third octave bands) (12.15)

352  Sound Insulation in Buildings

T > 1 / 4 ⋅ Tdet (12.16)

Thus, the lower limit at 100 Hz one-third octave band is reduced to 0.18 s
and 0.35 s at 50 Hz.
Some instrument manufacturers have overcome the problem by the use
of reverse filtering, i.e. the one-third octave band filters have been specially
crafted to provide short reverberation times and still comply with require-
ments for the frequency response. Thus, the lower limit for measurement
of reverberation time in one-third octave bands can be around 0.1 s in the
entire frequency range from 50 Hz to 5000 Hz.
Another problem in small volumes is that the modal density is very low. In
the one-third octave bands below 100 Hz, there may be only a few modes,
or none. With two or three modes within a band, the decay curve may suf-
fer from strong wavy behaviour due to interference, or double slope due to
very different attenuation of the modes. In either case, the evaluation of a
reverberation time is difficult and associated with a huge uncertainty. A
pragmatic solution to this problem was suggested by Hopkins and Turner
(2005), namely, to measure the reverberation time in the 63 Hz octave band
and apply this result for the 50 Hz, 63 Hz and 80 Hz one-third octave bands.
This idea has been adapted in the ISO 16283 series for sound insulation
measurements.
Finally, it is a problem with commonly used sound sources for building
acoustic measurements that the sound power in the 63 Hz octave is not
sufficient for measuring the reverberation time. The signal-to-noise level
should be at least 35 dB and the background noise is often quite high at
low frequencies. The best solution is probably to use the swept-sine tech-
nique in combination with a loudspeaker that has good performance at low
frequencies.

Table 12.4  M inimum measurable reverberation time (in s) using normal one-
third octave and octave band filters and either forward or time
reversed analysis
Forward Backward
f, Hz One-third octave Octave One-third octave Octave
50 1.40 0.35
63 1.11 0.36 0.28 0.09
80 0.88 0.22
100 0.70 0.18
125 0.56 0.18 0.14 0.04
160 0.44 0.11
200 0.35 0.09
 Measurement methods  353

12.5.5 Spatial averaging
The spatial averaging of reverberation time can be made in two different
ways. The reverberation times are evaluated in each pair of source-receiver
combinations, and the arithmetic average of reverberation times is calcu-
lated. Alternatively, the decay curves can be averaged, synchronizing the
beginnings and averaging the sound pressure squared in each time slot dur-
ing the decay. The sound power of the source shall be kept the same for
all measurements. The reverberation time is calculated from the ensemble-
averaged decay curve. This is the preferred method of spatial averaging in
ISO 3382-2.

12.6 VIBRATIONS

Vibrations in buildings are usually measured in the frequency range from

1 Hz to 80 Hz. The transducer can be an accelerometer of a geophone.
The accelerometer is an electromechanical transducer that produces
a voltage proportional to the acceleration of the vibrations. The most
common principle is the compression type piezoelectric accelerometer.
A heavy mass resists on two piezoelectric discs; the mass is preloaded
by a stiff spring and the whole assembly is mounted in a metal hous-
ing with a thick base. When the accelerometer is subjected to vibration,
the mass will produce a force on the piezoelectric discs, and the force
is proportional to the acceleration. A voltage will be developed across
the two piezoelectric discs, and the voltage is proportional to the force,
and thus also to the acceleration. The accelerometer has a mass-stiffness
resonance, which defines the upper limit of the usable frequency range.
The typical range of the resonance frequency is 12 kHz to 70 kHz, i.e. of
no practical importance.
The principle of a geophone is a magnetic mass mounted on a spring
and moving within a wire coil. It generates a voltage proportional to the
velocity of vibrations. The sensitivity can be very good and, thus, it can be
used for low vibration levels. The geophone has a mass-stiffness resonance,
typically in the range from 5 Hz to 50 Hz: Above this frequency, the veloc-
ity response is flat, but falls off towards lower frequencies. The resonance
is normally heavily damped, so that there is no resonance peak in the fre-
quency response. Still, a correction filter may be necessary to correct for the
non-flat response in the frequency range of interest, 1 Hz to 80 Hz.

12.6.1 Calibration
Before any vibration measurement, it is important to calibrate the equip-
ment. Some calibrators give a specified RMS acceleration while other cali-
brators give a specified peak acceleration. An example is the B&K type
354  Sound Insulation in Buildings

4291 calibrator that gives a peak acceleration level of 1 g = 9.81 m/s2 at

80 Hz. Using the reference acceleration a 0 = 10 −6 m/s2 , the calibration signal
corresponds to the acceleration level:

a 9.81
La = 20 lg = 20 lg −6
≅ 136.8 dB re 1 µm/s2 (12.17)
a0 2 ⋅ 10

For a harmonic vibration with the angular frequency ω, the velocity is

related to the acceleration by:

a a
v= = (12.18)
ω 2π f

Using the reference velocity v 0 = 10 −9 m/s and f = 80 Hz, the corresponding
velocity level of the calibration signal yields:

v  a 
Lv = 20 lg = La + 20 lg  0 
v0  ω ⋅ v0 
= La + 60 − 20 lgω ≅ 142.8 dB re 1 nm/s (12.19)

12.6.2 Measurement positions
Vibration measurements can be made both in vertical and horizontal direc-
tions. In buildings, the highest vibration levels are usually the vertical vibra-
tions on the floor. The vibration measurements should be made in positions
where the highest vibration level is obtained, which is in the middle of the
floor, or between the primary beams supporting the floor (Figure 12.8).
Additional measurement positions should be chosen more than 1 m from
the walls and from other positions.
The mounting of the accelerometer can be with bees’ wax on hard sur-
faces or using a device, as shown in Figure 12.9, in the case of carpets or
similar soft floor coverings.

12.6.3 Frequency weighting
Vibration measurements can either be performed in one-third octave
bands in the frequency range from 1 Hz to 80 Hz or using the standard-
ized frequency weighting curve (Figure 2.18). It is possible to calculate the
weighted acceleration level or the weighted velocity level from the one-third
 Measurement methods  355

Primary
beam

Secondary
beams or
joists

Structural wall
or primary beams

Figure 12.8 Examples of measurement positions where the highest level of vertical

vibration is expected. (Reproduced from Jakobsen, J., Measurement of vibra-
tion and shock in buildings for the evaluation of annoyance. Nordtest project
618–86. Technical Report No. 139, Danish Acoustical Institute, Denmark,
1987. With permission.)

60°)
=3
°( °
120
0
12
3×

130
0
15

Thread for mounting

of transducer
≈16
≈35
15

15°

Figure 12.9 
E xample of mounting device for measurement of vertical vibration on
carpet-covered floors. The weight of the device plus attached transducer
should be about 2.5 kg. Dimensions are in mm. (Reproduced from Jakobsen,
J., Measurement of vibration and shock in buildings for the evaluation of annoy-
ance. Nordtest project 618–86. Technical Report No. 139, Danish Acoustical
Institute, Denmark, 1987. With permission.)
356  Sound Insulation in Buildings

octave band spectrum using an equation similar to Equation 1.16 for the
A-weighted sound pressure level.

12.6.4 Integration time
The instantaneous maximum value of the acceleration level can be mea-
sured with an exponential averaging time constant of 1 s (“slow”) or with
a linear averaging over 2 s.

12.6.5 Noise floor
It is always recommended to ensure that the measured vibration values are
not influenced from the background noise of the measurement system. If
the accelerometer is placed on a heavy non-vibrating object, the apparent
vibration level should be at least 10 dB below the level of the real measure-
ments. If possible, baseline measurements should be obtained to verify the
background noise level.

12.6.6 Averaging of results
Whether it is relevant to average the measured results depends on the situa-
tion and the purpose of the measurements. In some cases, the spatial aver-
age over the measurement positions is calculated as an energetic average
(Equation 1.34). In other cases, the result must be reported for each mea-
surement position.
The Norwegian standard NS 8176 deals with vibrations in buildings due
to land, based transport, and the statistical maximum velocity must be cal-
culated in each position from at least 15 events of passing vehicles or trains.
According to this standard, the average maximum velocity is calculated
as the arithmetic average of the frequency-weighted maximum velocities
(not the squared velocities). Finally, the statistical maximum value (95 per-
centile) is calculated as the arithmetic average plus 1.8 times the standard
deviation of the frequency-weighted maximum velocities.

12.6.7 The vibration dose value

The Vibration Dose Value (VDV) is used in some countries for assessing
human reactions to intermittent vibration (see ISO 2631-1, 1997). Instead
of the common RMS acceleration, the root-mean-quad acceleration is
applied, which ensures that the VDV is more sensitive to peaks in the accel-
eration level:

VDV = 4
∫a
0
4
w dt (m/s1.75) (12.20)
 Measurement methods  357

where aw is the frequency-weighted acceleration as a function of time and T

is the total measurement period in seconds. Note the unusual unit of VDV.
An example of the practical application of VDV is found in the work of
Yang et al. (2014).

12.7 LOSS FACTOR

12.7.1 Structural reverberation time

The total loss factor of a building element can be estimated by measuring
the structural reverberation time using Equation 2.30:

2.2
ηtotal = (12.21)
f Ts

where f is the centre frequency of a one-third octave band and Ts is the

measured structural reverberation time.
The structural reverberation time is measured in accordance with ISO
10848-1 Section 7.3 and ISO 3382-2 using the integrated impulse response
method with point excitation and measurement of velocity or acceleration
in a number of positions. The excitation can be achieved using either a
hammer or a vibration exciter and the MLS technique. See ISO 18233 for
details on the latter. At least three excitation points are used, and at least
three transducer positions are used for each excitation point.

12.7.2 Upper limits for loss factor

The problems with measuring structural reverberation time are no less than
those with measuring reverberation time in a room. The ringing of the
filters means that it can be necessary to apply the time-reversed analysis.
As loss factor is reciprocal to reverberation time, the ringing of the filters
causes an upper limit of measureable loss factors. Equations 12.13 and
12.15 lead to upper limits of the loss factor that can be measured with one-
third or octave band filters (Table 12.5). Again, it is an advantage to use
time-reversed analysis.
The low modal density is another problem, which is more pronounced in
structures than in rooms. A low modal density is a precaution for using an

Table 12.5  Maximum measurable loss factors using normal one-third octave and
octave band filters and either forward or time-reversed analysis
Forward Backward
One-third octave Octave One-third octave Octave
0.03 0.10 0.13 0.39
358  Sound Insulation in Buildings

alternative method of measuring the bandwidth of single structural modes.

If a structural mode is separated in frequency from other modes, it is pos-
sible to estimate the loss factor at the natural frequency f 0 by measuring the
3 dB bandwidth Br and using Equation 2.52:

Br
ηtotal = (12.22)
f0

12.8 RADIATION EFFICIENCY

The radiation efficiency of a surface element can be measured if the element

can radiate sound to a receiving room with approximately diffuse sound
field, preferably in a laboratory. The radiating element should be excited to
vibrations, e.g. by a steel framework bolted to the lower part of the element,
as described in NT ACOU 033.
The average surface velocity level Lv is calculated as the energetic average
from:

∑
n
vi2
Lv = 10 lg i =1
2
(dB) (12.23)
nv 0

where n is the number of measurement positions on the radiating surface,

vi is the RMS velocity in the normal direction in position i and v 0 = 10 −9 m/s
is the reference velocity.
The radiated sound power from the surface to the receiving room is
determined by measuring the average sound pressure level and assuming a
diffuse sound field. The average sound pressure level Lp is calculated as the
energetic average from:

∑
m
pi2
Lp = 10 lg i =1
2
(dB) (12.24)
mp 0

where m is the number of measurement positions in the receiving room,

pi is the RMS sound pressure in position i and p 0 = 20 μPa is the reference
sound pressure.
At least two different positions of the exciter should be used. For each
exciter position, the vibration level should be measured in at least ten differ-
ent positions and the sound pressure level in at least three different positions.
The radiation efficiency σ is calculated from:

S
10 lg σ = Lp − Lv − 10 lg + 28 (dB) (12.25)
A
 Measurement methods  359

where S is the area of the radiating surface and A is the equivalent absorp-
tion area of the receiving room. This formula can be derived from Equations
6.24 and 4.29.

12.9 JUNCTION ATTENUATION

The measurement of transmission of vibrations through a junction in build-

ings can be made in accordance with the method in NT ACOU 090, which
has been adapted in ISO 10848-1.
The building element i is excited either with an electrodynamic exciter or
with a hammer. The vibration level is measured simultaneously on element
i and on the element j on the other side of the junction. Measurements are
made in both directions, and for each direction at least two exciter posi-
tions should be used. For each exciter position, the vibrations are measured
in at least eight positions distributed over elements i and j. If using excita-
tion with a hammer, at least ten hammer blows at different positions are
used for each of the eight pairs of vibration transducer positions. In addi-
tion, the structural reverberation time is measured in each of the two ele-
ments under test following ISO 3382-2 and using a 20 dB evaluation range.
The average vibration level difference is determined by:
1
Dv ,ij =
2
( Dv,ij + Dv, ji ) (dB) (12.26)
where
Dv,ij is the difference between the average velocity level of element i and
j, when element i is excited.
Dv,ji is the difference between the average velocity level of element j and
i, when element j is excited.

In case of transmission between single-layered structures, the vibration

reduction index Kij should be determined. Then, it is necessary to measure
the structural reverberation time of both structures. For each of the mea-
sured elements (i and j), the equivalent absorption length is estimated by:

2.2 π2 S j
aj = (m) (12.27)
f
Ts, j c
fref
where
Sj is the area of element j (m 2).
Ts,j is the structural reverberation time of element j (s).
c = 344 m/s is the speed of sound in air.
f is the frequency (Hz).
fref = 1000 Hz is a reference frequency.
360  Sound Insulation in Buildings

The vibration reduction index is calculated by:

lij
Kij = Dv ,ij + 10 lg (dB) (12.28)
ai a j
where
Dv ,ij is the direction-averaged vibration level difference (dB).
li,j is the length (m) of the junction between the elements i and j.
ai and aj are the equivalent absorption lengths (m) of elements i and j,
respectively.

REFERENCES

ANSI S1.1 (1994). American National Standard – Acoustical terminology, The

American National Standards Institute, Inc., New York.
ASTM E90-09 (2009). Standard test method for laboratory measurement of air-
borne sound transmission loss of building partitions and elements, American
Society for Testing and Materials, Conshohocken, PA.
ASTM E336-16 (2016). Standard test method for measurement of airborne sound
attenuation between rooms in buildings, American Society for Testing and
Materials, Conshohocken, PA.
ASTM E413-16 (2016). Classification for rating sound insulation, American Society
for Testing and Materials, Conshohocken, PA.
ASTM E492-09 (2016). Standard test method for laboratory measurement of impact
sound transmission through floor-ceiling assemblies using the tapping machine,
American Society for Testing and Materials, Conshohocken, PA.
ASTM E966-10 (2010). Standard guide for field measurements of airborne sound
insulation of building facades and facade elements, American Society for
Testing and Materials, Conshohocken, PA.
ASTM E989-06 (2012). Standard classification for determination of impact
­insulation  class (IIC), American Society for Testing and Materials,
Conshohocken, PA.
ASTM E1007-16 (2016). Standard test method for field measurement of tapping
machine impact sound transmission through floor-ceiling assemblies and
associated support structures, American Society for Testing and Materials,
Conshohocken, PA.
ASTM E1332-16 (2016). Standard classification for rating outdoor-indoor sound
attenuation, American Society for Testing and Materials, Conshohocken, PA.
ASTM E2235-04 (2012). Standard test method for determination of decay rates
for use in sound insulation test methods, American Society for Testing and
Materials, Conshohocken, PA.
C.L. Christensen, G. Koutsouris, J.H. Rindel (2013). The ISO 3382 parameters:
Can we simulate them? Can we measure them? Proceedings of ISRA 2013,
International Symposium on Room Acoustics, 9–11 June 2013, Toronto, ON,
Canada.
C. Hopkins and P. Turner (2005). Field measurement of airborne sound insulation
between rooms with non-diffuse sound fields at low frequencies. Applied
Acoustics 66, 1339–1382.
 Measurement methods  361

C. Hopkins (2015). Revision of international standards on the field measurements

of airborne, impact and façade sound insulation to form the ISO 16283 series.
Building and Environment 92, 703–712.
ISO 717-1 (1996). Acoustics – Rating of sound insulation in buildings and of building
elements – Part 1: Airborne sound insulation, International Organization for
Standardization, Geneva, Switzerland.
ISO 717-2 (1996). Acoustics – Rating of sound insulation in buildings and of build-
ing elements – Part 2: Impact sound insulation, International Organization for
Standardization, Geneva, Switzerland.
ISO 2631-1 (1997). Mechanical vibration and chock – Evaluation of human expo-
sure to whole-body vibration – Part 1: General requirements, International
Organization for Standardization, Geneva, Switzerland.
ISO 10052 (2004). Acoustics – Field measurements of airborne and impact sound
insulation and of service equipment sound – Survey method, International
Organization for Standardization, Geneva, Switzerland.
ISO 10140-2 (2010). Acoustics – Laboratory measurement of sound insulation
of building elements – Part 2: Measurement of airborne sound insulation,
International Organization for Standardization, Geneva, Switzerland.
ISO 10140-3 (2010). Acoustics – Laboratory measurement of sound insulation
of building elements – Part 3: Measurement of impact sound insulation,
International Organization for Standardization, Geneva, Switzerland.
ISO 10848-1 (2006). Acoustics – Laboratory measurement of the flanking transmission
of airborne and impact sound between adjoining rooms – Part 1: Frame document,
International Organization for Standardization, Geneva, Switzerland.
ISO 15186-1 (2000). Acoustics – Measurement of sound insulation in buildings and
of building elements using sound intensity – Part 1: Laboratory measurements,
International Organization for Standardization, Geneva, Switzerland.
ISO 15186-2 (2003). Acoustics – Measurement of sound insulation in buildings
and of building elements using sound intensity – Part 2: Field measurements,
International Organization for Standardization, Geneva, Switzerland.
ISO 15186-3 (2002). Acoustics – Measurement of sound insulation in buildings and of
building elements using sound intensity – Part 3: Laboratory measurements at low
frequencies, International Organization for Standardization, Geneva, Switzerland.
ISO 16032 (2004). Acoustics – Measurement of sound pressure level from service
equipment in buildings – Engineering method, International Organization for
Standardization, Geneva, Switzerland.
ISO 16283-1 (2014). Acoustics – Field measurements of sound insulation in build-
ings and of building elements – Part 1: Airborne sound insulation, International
Organization for Standardization, Geneva, Switzerland.
ISO 16283-2 (2015). Acoustics – Field measurements of sound insulation in build-
ings and of building elements – Part 2: Impact sound insulation, International
Organization for Standardization, Geneva, Switzerland.
ISO 16283-3 (2016). Acoustics – Field measurements of sound insulation in build-
ings and of building elements – Part 3: Façade sound insulation, International
Organization for Standardization, Geneva, Switzerland.
ISO 18233 (2006). Acoustics – Application of new measurement methods in building and
room acoustics, International Organization for Standardization, Geneva, Switzerland.
F. Jacobsen and J.H. Rindel (1987). Time reversed decay measurements. Journal of
Sound and Vibration 117, 187–190.
362  Sound Insulation in Buildings

J. Jakobsen (1987). Measurement of vibration and shock in buildings for the evalu-
ation of annoyance. Nordtest project 618–86. Technical Report No. 139,
Danish Acoustical Institute, Denmark.
J. Jakobsen (2001). Danish guidelines on environmental low frequency noise, infra-
sound and vibration. Journal of Low Frequency Noise, Vibration and Active
Control 20 (3), 141–148.
A. Løvstad, J.H. Rindel, V. Støen, A.T. Windsor, A. Negård (2014). Measurement and
simulation of high sound insulation and identification of sound transmission
paths in complex building geometries. Proceedings of Forum Acusticum 2014,
September 2014, Krakow, Poland.
M. Machimbarrena and F. Jacobsen (1999). Is there a systematic disagreement
between intensity-based and pressure-based sound transmission loss measure-
ments? Building Acoustics 6(2), 101–111.
NT ACOU 033 (1980). Bulkheads: Sound radiation efficiency – Laboratory test,
Nordtest, Helsinki, Finland.
NT ACOU 090 (1994). Building structures, junctions: Transmission of vibrations –
Field measurements, Nordtest, Helsinki, Finland.
NT ACOU 102 (1999). Building elements – Façade elements and façades: Field mea-
surements of airborne sound insulation – Loudspeaker method using MLS
noise signals, Nordtest, Helsinki, Finland.
Norsonic (2006). Swept sine measurement technique using the Nor121 analyser.
Application Note, Ed. 3, 10.06, Norsonic, Lierskogen, Norway.
NS 8176 (2005). Vibration and shock. Measurement of vibrations in buildings from
landbased transport and guidance to evaluation of effects on human beings,
2nd edition, Norwegian Standard, Oslo, Norway.
T. Poulsen and F.R. Mortensen (2002). Laboratory evaluation of annoyance of low
frequency noise. Working Report No. 1, Danish Environmental Protection
Agency, Copenhagen, Denmark.
F. Satoh, M. Sano, Y. Hayashi, J. Hirano, S. Sakamoto (2011). Sound insulation mea-
surement using 10 minute swept-sine signal. Proceedings of InterNoise 2011,
4–7 September 2011, Osaka, Japan.
C. Simmons (1999). Measurement of sound pressure levels at low frequencies in
rooms. Comparison of available methods and standards with respect to micro-
phone positions. Acustica – Acta Acustica 85, 88–100.
J. Takala and M. Kylliäinen (2013). Room acoustics and background noise levels in
furnished Finnish dwellings. Proceedings of InterNoise 2013, 15–18 September
2013, Innsbruck, Austria.
M.L.S. Vercammen (1992). Low-frequency noise limits. Journal of Low Frequency
Noise and Vibration 11, 7–13.
T. Yang, P. Liu, J. Yin (2014). Research on assessment of fourth power vibration dose
value in environmental vibration caused by metro. Proceedings of ICSV 21,
13–17 July 2014, Beijing, China.
Chapter 13

Measurement uncertainty

The sound pressure level in a room varies from one point to another, even in
an ideal diffuse sound field. These variations are particularly large when the
excitation is a pure tone or narrow band noise like a one-third octave band
and, thus, some kind of spatial averaging is necessary in order to get a reason-
ably accurate measurement result. Measurements at low frequencies in small
rooms need special consideration. This chapter presents an overview of the
theoretical background and methods for estimation of uncertainty in sound
pressure level measurements and reverberation time measurements in a room.
Chapter 4, Introduction to Room Acoustics, is a basis for this chapter.

13.1 SOME STATISTICAL CONCEPTS

The fundamental principles of measurement uncertainty in building

acoustics are found in the ISO Guide to the expression of Uncertainty in
Measurement (GUM) and in ISO 12999-1 (2014).

13.1.1 Uncertainty of a measurement result

The following concepts and definitions apply:

Measurand z: The measured quantity, e.g. sound pressure level or rever-

beration time.
Mean value ‹z›: The arithmetic average of all measured z. For sound
pressure levels in a room, the arithmetic average of the squared sound
pressure in all measurement positions.
Variance σ 2: The mean square deviation from the mean value.
Standard deviation σ: The square root of the variance. For sound pres-
sure levels in a room, the root mean square (RMS) deviation from
the mean value of the squared sound pressure in all measurement
positions. The standard deviation can either be determined experi-
mentally from the measured data, or it can be derived from assumed
probability distributions.

363
364  Sound Insulation in Buildings

Uncertainty Characterizes the dispersion of the result of a measurement

and may be the standard deviation or the half-width of an interval
having a stated level of confidence.
Standard uncertainty u Uncertainty of a result expressed as a standard
deviation. For a Gaussian distribution u = σ.
Expanded uncertainty U Quantity defining an interval about the result
of a measurement that may be expected to encompass a large fraction
of the distribution of values. U = k·u.
Coverage factor, k Factor used as a multiplier of the standard uncer-
tainty in order to obtain an expanded uncertainty. It depends on
the statistical distribution function, on the confidence level and on
whether a one-sided or a two-sided test is considered.
Confidence level, g The statistical probability that the result is within a
certain range, either on both sides of the mean value (two sided) or on
one side of the mean value (one sided).
A measurand z can be statistically characterized by the probability den-
sity function w(z), which in some cases can be a mathematical func-
tion of the mean ‹z› and standard deviation σ (or the variance σ 2).

13.1.2 The Gaussian distribution

The Gaussian distribution is the most commonly used statistical distribu-
tion and is also called the normal distribution. If the mean value is μ and
the standard deviation is σ, the probability density function is:
2
( z−µ )
1 −
w (z) =
2
e 2 σ (13.1)
σ 2π
This is a symmetric distribution and the probability for a random sample to
be within the range μ ± σ is 68 %. The probability to be within an extended
range of μ ± 2σ is 95 % (Figure 13.1).

13.1.3 Confidence levels
Assuming a Gaussian distribution, the measurement result is an estimate
of the true result, which will be within a range of <z> ± k(g)·σ(z), where σ(z)
is the standard deviation and k(g) is a coverage factor that depends on the
confidence level, g (Table 13.1). This is called a two-sided test and is applied
when a measurement result is reported without comparison with any target
value. For example, with 95 % probability, the true result is within a range
of ±1.96 times the standard deviation (the expanded uncertainty U).
When a measured result is compared with a requirement, the one-sided
test applies. If, for example, a 90 % confidence level is chosen, the coverage
factor for a one-sided test is k = 1.28. This means that the limit zmax is not
 Measurement uncertainty  365

1.0

0.8

0.6
Probability

0.4

0.2

0.0
–3 –2 –1 0 1 2 3
Measurand, z
w(z) P(z) 1 – P(z)

Figure 13.1 Gaussian distribution. Probability density function w(z), cumulative prob-

ability function P(z), and the opposite function 1 − P(z).

Table 13.1  Confidence levels g and the associated coverage factors k for one- and
two-sided tests
Coverage factor, k 1.00 1.28 1.65 1.96 2.58 3.29
Confidence level for 68 80 90 95 99 99.9
two-sided test, g (%)
Confidence level for 84 90 95 97.5 99.5 99.95
one-sided test, g (%)

exceeded with 90 % probability, if <z> + k(g)·σ(z) = <z> + 1.28·σ(z) < zmax. It

is noted that the coverage factor is smaller for a one-sided test than for a
two-sided test with the same confidence level. At the end of this chapter, we
shall return to this point with an example.

13.2 VARIANCE OF SOUND PRESSURE IN A ROOM

13.2.1 Pure tone excitation

It is assumed in this section that the frequency is above Schroeder’s limit-
ing frequency, so statistical methods apply to the sound field. Thus, the
excitation of sound with a single frequency will involve a number of room
modes having natural frequencies close to the frequency of excitation. In a
measurement point, the sound pressure is the result of contributions from
the room modes that have been excited, each mode being represented by its
amplitude and phase angle (see Equation 4.6).
366  Sound Insulation in Buildings

If a microphone is moved slowly in a room while a pure tone is emitted,

the sound pressure level will vary strongly, and the fluctuations look sto-
chastic (Figure 13.2a).
The range of the fluctuations can be described by the relative spatial vari-
ance, defined by:

(p ) = (p )
2
2
σ 2 2 − p2
ε R2 = 2 2 (13.2)
p2 p2

where σ (p2) is the standard deviation of the squared sound pressure and
<p2> denotes the spatial average. The sound pressure is the RMS value of a
time averaging, see Section 13.3.3 about sufficient measurement time.

(a) (b)
125 Hz sine 125 Hz noise
90 90
Sound pressure level (dB)

Sound pressure level (dB)

80 80

70 70

60 60

50 50

40 40
0 100 200 300 400 0 100 200 300 400
Distance (cm) Distance (cm)
(c) 125 Hz sine (d) 125 Hz noise

50 55 60 65 70 75 80 50 55 60 65 70 75 80

Figure 13.2 Measured distribution of sound pressure level in a 22 m2 furnished room.

(a) As a function of distance for a 125 Hz sinusoidal signal. (b) As a func-
tion of distance for a 125 Hz one-third octave band filtered pink noise. (c)
Histogram for a 125 Hz sinusoidal signal. (d) Histogram for a 125 Hz one-
third octave band filtered pink noise. The dashed lines in (c) and (d) indicate
the 10 percentile. (Adapted from Pedersen, S. et al., J. Low Freq. Noise, Vib.
Act. Control 26(4), 249–270, 2007.)
 Measurement uncertainty  367

13.2.2 Band-limited noise excitation

A diffuse sound field and frequencies above Schroeder’s limiting frequency
fg are assumed (Equation 4.21):

T
f g = 2000 (Hz) (13.3)
V

where T is the reverberation time in (s) and V is the room volume in (m3).
If the excitation of a room is with a band-limited noise signal (Figure 13.2b)
and the bandwidth is B (Hz), the relative spatial variance (Lubman, 1968)
is approximately:

2 1
ε R2 = arctan(Z) − 2 ln(1 + Z 2 ) (13.4)
Z Z

where Z = B/Br is the ratio between the bandwidth B of the noise signal and
the bandwidth Br of a typical room mode. With the reverberation time T,
the mode bandwidth is Br ≈ 2.2/T (see Equation 4.19). If the bandwidth B
is not very small, a good approximation for the relative spatial variance is:

−1 −1
 B   BT 
ε R2 ≅  1 + = 1 +  (13.5)
 π Br   6.9 

where T is the reverberation time. It is seen that the variance decreases with
increasing bandwidth of the noise signal and with increasing reverberation
time; thus, a wideband noise signal in a highly reverberant room yields very
small spatial variations of the squared sound pressure. The approxima-
tion (Equation 13.5) is valid within 2.5 % for BT > 20. Below that range,
Equation 13.4 is used.
In practice, most measurements are performed with a wide band noise
signal and a series of band-limiting filters after the microphone. However,
it does not matter where in the measurement chain the band-limiting filter
is located; the above equation is still a valid approximation. The most com-
mon filters are the octave band (B = 0.71 f) and the one-third octave band
(B = 0.23 f), where f is the centre frequency of the band.
In the extreme case of a pure tone (bandwidth B = 0 Hz), we have εR 2 = 1,
and this corresponds to a standard deviation on the sound pressure level
of 5.6 dB (Schroeder, 1969). Thus, the sound pressure level for a pure tone
excitation in a room will vary within a range of approximately 22 dB (with
95 % probability assuming Gaussian distribution). However, these results
refer to using the sound pressure level as the measurand, i.e. when the mean
value and the standard deviation are determined directly from the dB val-
ues. This introduces a bias error in the mean value compared to the correct
result calculated from the squared sound pressures. The bias error and the
368  Sound Insulation in Buildings

 ias error D and standard deviation σ of the sound pressure level in a

Table 13.2  B
diffuse sound field as functions of the product BT, B being the
measurement bandwidth and T the reverberation time
1 + BT/6.9 1 3 10 30 100 300 1000
BT 0 14 62 200 683 2063 6893
Bias D (dB) 2.51 0.76 0.22 0.07 0.02 0.00 0.00
Standard deviation σ (dB) 5.57 2.73 1.41 0.80 0.44 0.25 0.14
Source: Lubman, D., J. Sound Vib., 16, 43–58, 1971.


standard deviation in dB can be estimated from the BT product and some

values are given in Table 13.2.
The bias means that the correct energetic averaged sound pressure level
Lp2 (Equation 1.34) is greater than or equal to the simple arithmetic dB-
average L dB (Equation 1.31), and the difference is the bias D error:

Lp2 = LdB + D (dB) (13.6)

In Table 13.2 the bias error is 2.5 dB for a pure tone (B = 0) but decreases
with increasing bandwidth. For broadband noise excitation (BT > 200), the
bias error becomes negligible. The standard deviation of the sound pressure
level L from the mean value <L> is 5.6 dB for a pure tone in a diffuse sound
field, as first found by Doak (1959).

13.2.3 Variance at low frequencies

Below Schroeder’s limiting frequency, the modal overlap index is small (less
than 3) and the room modes may be too spread to fill the frequency band
that is measured. Thus, the efficient bandwidth of the modes may be less
than the measurement bandwidth (see Figure 4.7). For this case, Lubman
(1974) suggested to replace the measurement bandwidth B by the effective
noise bandwidth B eff calculated as the number of modes within B multi-
plied with the bandwidth of the modes Br:

∆N
Beff = B ⋅ ⋅ Br ⋅ κ = ∆N (B) ⋅ Br ⋅ κ (13.7)
∆f

where ΔN(B) is the number of modes within the bandwidth B. The overlap
factor κ (≤1) is included in order to take account of the fact that the modes
have some degree of overlap. Even if the modal overlap index is M < 3, there
is still more than one mode within the bandwidth of the modes (M > 1) in a
certain frequency range extending at least one octave below fg.
So, the relative spatial variance is calculated with Z = B eff /Br = ΔN(B)·κ
in Equation 13.4. Using instead the approximation (Equation 13.5) yields:
 Measurement uncertainty  369

−1
 ∆ N (B) ⋅ κ 
ε R2 ≅  1 +  for f <f g (13.8)
 π

The overlap factor κ has been estimated by comparison with empirical data
like those in Figure 13.3, and the result is κ = 0.37 = π/8.5. The suggested
approximation for the relative spatial variance below the limiting frequency is:

−1
 ∆ N (B) 
ε R2 ≅  1 +  for f < f g (13.9)
 8.5 

where ΔN(B) is the number of room modes within the measurement band-
width B. This approximation was first suggested by (Rindel, 1981).
Figure 13.3 shows a comparison of measured spatial standard deviations
in a room compared with the estimations from Equation 13.9 below the
limiting frequency and Equation 13.5 above the limiting frequency. The
two theoretical curves cross at the limiting frequency.
In the common case of measurements in one-third octave bands, Equations
13.5 and 13.9 can be modified with insertion of B = 0.23 f and the modal
density (Equation 4.13). Using the approximation (Equation 4.14), the rela-
tive spatial standard deviation yields, below and above fg, respectively:

−1/ 2
 V ⋅f3 
εR ≅ 1 + for f < f g (13.10)
 2.9 ⋅ c 3 

−1/ 2
 T ⋅f 
εR ≅ 1 +  for f ≥ f g (13.11)
 30 

These are displayed graphically in Figure 13.4. The relative spatial stan-
dard deviation is the higher one in the two Equations 13.10 and 13.11. At
­frequencies below the Schroeder limiting frequency, the room volume is the
important parameter, whereas at frequencies above the limiting ­frequency,
the reverberation time is the important parameter. The approximation
above the limiting frequency (Equation 13.11) is good, and that under the
limiting frequency (Equation 13.10) is also quite good, because the error
by using the simple Equation 13.5, instead of Equation 13.4, to some
extent, compensates for the error due to the approximate modal density
(Equation 4.14) instead of the more correct one (Equation 4.13).
If the volume and reverberation time are known, it is possible to estimate
the relative spatial standard deviation for the curves in Figure 13.4. The
curve with the highest standard deviation is always the valid one, and the
cross-over is at the Schroeder limiting frequency. It should be noted that
reducing the reverberation time in a room makes things worse because the
standard deviation is increased at frequencies above the limiting frequency.
370  Sound Insulation in Buildings

Measured
Relative spatial standard deviation

0.8
Below fg
Above fg
0.6

0.4

0.2

0
100 200 500 1000 2000
One-third octave center frequency (Hz)

Figure 13.3 M
 easured and estimated relative spatial standard deviation in one-third
octave bands for a 70 m3 room with reverberation time approximately 1.0 s.
The Schroeder limiting frequency is 240 Hz. (Measured data from Jacobsen,
F., The diffuse sound field. Statistical considerations concerning the rever-
berant field in the steady state. Report No. 27, The Acoustics Laboratory,
Technical University of Denmark, 1979.)

1.0

25 V(m3), below fg
100
Relative spatial standard deviation

0.8
400
1600
0.6
T(s), above fg
10 5 2 1 0.5
0.4

0.2

0.0
50 100 200 500 1000 2000 5000
One-third octave center frequency (Hz)

Figure 13.4 Curves for estimation of the relative spatial standard deviation on one-third
octave band measurements in a room. At low frequencies, the volume is the
determining parameter, at high frequencies it is the reverberation time. For a
given volume and reverberation time, the cross-over is at the Schroeder limiting
frequency.
 Measurement uncertainty  371

It is sometimes claimed that it is preferable to have a low limiting frequency

and that the reverberation time should not be too long for that reason;
however, it is seen from Figure 13.4 that for a fixed volume, a lower limit-
ing frequency goes hand in hand with a higher spatial standard deviation.
At very low frequencies, the number of room modes is small and the one-
dimensional axial modes tend to dominate. There may be none or very few tan-
gential or oblique modes, depending on the ratio between room dimensions.
The calculated frequency response of a room at low frequencies is shown
in Figure 13.5. The room is rectangular with dimensions (lx, ly, lz) = (4.32 m,
3.38 m, 2.70 m) and volume 39 m3. The frequency response is shown for a
source in one corner and three different receiver positions; in the opposite
corner, in the centre of the room, and in a point halfway between the centre
and a corner. All modes are at maximum in the corner position while in the
centre position, all modes with an uneven number nx or ny or nz are elimi-
nated; so, the first peak in the centre is at the mode (2,0,0) around 80 Hz.
In the very low frequency range 0 Hz < f < f 2,0,0, the following observa-
tions are made:

• The modal density is very low; in some one-third octave bands, there
may be a single mode: in other bands, the sound pressure rides on the
skirts of the frequency response of one or two modes with natural
frequencies outside the band.
• The range of variation of the sound pressure level, i.e. the difference
between the maximum and minimum curves, is high in the region
above the first axial mode (1,0,0), but decreases towards zero when
the frequency drops towards the cavity mode (0,0,0).

50

40

30
SPL (dB)

20

10 Pos (0, 0, 0)
Pos (lx/4, ly/4, lz/4)
Pos (lx/2, ly/2, lz/2)
0
20 40 60 80 100 200
Frequency (Hz)

Figure 13.5 Calculated frequency responses in three receiver positions at low frequen-

cies in a 39 m3 room. The source is in a corner.
372  Sound Insulation in Buildings

• The probability function is asymmetric; in the position halfway

between corner and centre, the level is only a few dB lower than in
the corner. This reflects the shape of the cosine-shaped axial mode,
having maximum at the walls and a node in the centre.
• The spatial variation of the sound pressure is deterministic. Thus,
the normal statistical methods applied to a stochastic system are
not usable.

These observations lead to the result shown in Figure 13.6. The frequency
range is divided into three, and the corresponding limiting frequencies are
f 2,0,0 (the frequency of the second axial mode in the longest room dimen-
sion) and fg (the Schroeder limiting frequency).
The course of the standard deviation in region 1 has been determined on
the basis of a number of measured data from some cases: see examples in
Figures 13.7 and 13.8. The empirical equation is:

lx ⋅ f
εR ≅ for f < f2,0,0 (13.12)
c

where lx is the longest dimension of the room (assuming a box-hyphen room)

and f 2,0,0 = c/lx is the natural frequency of the room mode (2,0,0). Below this
frequency, it can be assumed that there are very few room modes, and
they are dominated by the three lowest axial modes (1,0,0), (0,1,0) and

Region 1 Region 2 Region 3

1
Relative spatial standard deviation (εR)

Below f (2,0,0)
0.8 Between f (2,0,0) and fg
Above fg
0.6

0.4

0.2

f2,0,0 fg
0
10 20 50 100 200 500 1000 2000 5000
One-third octave centre frequency (Hz)

Figure 13.6 C
 ourse of the relative spatial standard deviation, and division into three
frequency regions.
 Measurement uncertainty  373

(0,0,1). The spatial standard deviation in the three regions is determined by

Equation 13.12 in region 1, Equation 13.10 in region 2, and Equation 13.11
in region 3, respectively.
The spatial standard deviation in dB is determined as:

p2 + σ (p2 )
σ (Lp ) ≅ 10 lg = 10lg(1 + ε R )
p2

≅ 10 lg(e) ⋅ ε R ≅ 4.34 ⋅ ε R for ε R  1 (13.13)

The results shown in Figure 13.7 were derived from measurements in a

148 m3 reverberation room with hard surfaces and average reverberation
time 2.1 s. Measurement positions were located in a mesh of 7 × 10 posi-
tions in two different heights above the floor, 0.7 m and 1.7 m. The distance
between the horizontal positions was 0.7 m. Region 2 is between the fre-
quencies 45 Hz and 238 Hz. The very different course in the three regions
is clearly seen in this example, and the peak is noticed at 45 Hz.
The results shown in Figure 13.8 were derived from measurements in a 31 m3
bedroom with furniture and average reverberation time 0.2 s. Measurement
positions were located in a mesh of 4 × 5 positions in two different heights
above the floor, 0.7 m and 1.7 m. The distance between the horizontal positions

5
Standard deviation of SPL in 1/3 octave bands (dB)

Measured
4 Prediction

0
20 50 100 200 500 1000 2000 5000
One-third octave centre frequency (Hz)

Figure 13.7 Measured and estimated spatial standard deviation in a 148 m3 reverberation

room (7.5 m × 5.5 m × 3.5 m, T = 2.1 s). (Measured data from Simmons, C.,
Noise Control Eng. J., 60, 405–420, 2012.)
374  Sound Insulation in Buildings

5
Standard deviation of SPL in 1/3 octave bands (dB)

Measured
4 Predicted

0
20 50 100 200 500 1000 2000 5000
One-third octave centre frequency (Hz)

Figure 13.8 Measured and estimated spatial standard deviation in a 31 m3 bedroom (3.8
m × 3.3 m × 2.5 m, T = 0.2 s). (Measured data from Simmons, C., Noise Control
Eng. J., 60, 405–420, 2012.)

was 0.7 m. In this example, the volume is small and reverberation time is very
short. The Schroeder limiting frequency is 160 Hz and f2,0,0 = 90 Hz, so region
2 is quite narrow in this case. The peak is noticed around 90 Hz.

13.3 MEASUREMENT UNCERTAINTY OF SOUND

PRESSURE LEVEL MEASUREMENTS

Measurement uncertainty in building acoustics is addressed in ISO 12999-1

(2014), and it is based on the assumption of a Gaussian (normal) distribu-
tion of the sound pressure level when measured in a room. Although this
is not correct, it may be a sufficiently good approximation at frequencies
above the Schroeder limiting frequency. The obvious advantage of this
assumption is that the expanded uncertainty can be calculated for a given
confidence level.

13.3.1 Probability density functions

In a diffuse sound field, or in practice at frequencies above the Schroeder
limiting frequency, it has been shown both theoretically and experimen-
tally that the RMS deviation of the sound pressure level L from the spatial
average <L> is 5.6 dB (Doak, 1959; Schroeder, 1969).
 Measurement uncertainty  375

However, the probability density function is not the symmetric Gaussian

distribution. Instead, the mean squared sound pressure in a room with a pure
tone excitation is known to follow a Poisson distribution. The probability
density function for the sound pressure level relative to the spatial average
(Pierce 1989, Equations 6 and 7.16) can be mathematically expressed as:

w (z) = ez− e
z
(13.14)

where

ln (10)
z= (L − L0 ) ≅ 4.34 (L − L0 ) (13.15)
10

Here, L is the sound pressure level in an arbitrary point and L 0 is the sound
pressure level corresponding to the spatial average of the squared sound
pressure, p2 .
The Poisson distribution is an asymmetric function (Figure 13.9). This
implies that the spatial averaged sound pressure level (L) is 2.5 dB lower than
L 0. The probability for a measured sound pressure level L to be less than L 0
is 63 %, whereas the probability to exceed L 0 is only about half of that, 37 %.
Other characteristic measures that can be extracted from the cumulative
probability function of the Poisson distribution are that the 50 % (median)
level is L 0 − 1.6 dB, the 68 % confidence interval is (−7.6 dB; 2.7 dB) and the

1
0.9
0.8
0.7
0.6
Probability

0.5
0.4
0.3
0.2
0.1
0
–20 –15 –10 –5 0 5 10
L–L0 (dB)

w (z) P (being less) P (exceeding)

Figure 13.9 Probability functions for the sound pressure level in a diffuse sound field.
(Adapted from Pierce 1989, Figures 6.16.)
376  Sound Insulation in Buildings

95 % confidence interval is (−16 dB; 5.7 dB). So, if a microphone is moved in

a room excited by a pure tone, the sound pressure level fluctuates more than
20 dB, but the maxima will be only about 6 dB above the spatial average L 0.
Experimental results for a pure tone excitation in a room are shown in Figure
13.2a and c.
If a diffuse sound field cannot be assumed, which is the case at frequen-
cies below Schroeder’s limiting frequency, the sound field can be analysed
by assuming a rectangular room with dimensions lx ≥ ly ≥ lz. Then, the
squared sound pressure is known as a function of position (x, y, z) at each
mode (nx, ny, nz) from Equation 4.4:

 x  y  z
p2 = p02 ⋅ cos2  π nx  ⋅ cos2  π ny  ⋅ cos2  π nz  (13.16)
 lx   l y  l z

It follows that in an oblique mode (all n ≠ 0), the spatial average covering
the entire volume is:
lx ly
 x  y
p 2 2
0
∫

0
2
lx  ∫
= p ⋅ cos  π nx  d x ⋅ cos2  π ny  d y
0
 ly 

lz
 z 1

∫
⋅ cos2  π nz  d z = ⋅ p02
0
 lz  8
(13.17)

This implies that in each of the eight corners, the sound pressure level is
10 lg (8) = 9 dB higher than the sound pressure level L 0 of spatial averaged
squared sound pressure. In a similar way, it is found that in the tangential
modes (two n ≠ 0), the highest sound pressure level is 6 dB above the room
average and in the axial modes (one n ≠ 0), the highest sound pressure level
is 3 dB above the room average.
The probability density functions of the various types of room modes
have been derived by Waterhouse (1970). The cumulative probability
curves for the various types of room modes, oblique, tangential, axial and
the cavity mode, are shown in Figure 13.10.
The statistics of the oblique and tangential modes have some similari-
ties; about 40 % of all positions in the volume have sound pressure levels
below L 0 and about 60 % of all positions have a higher sound pressure
level. The median level is 1.5 dB to 1.7 dB higher than L 0. These are the
most important modes in the frequency range between f 2,0,0 and fg (region
2 in Figure 13.6). Although the sound field is not diffuse because of insuf-
ficient modal overlap, the sound field is still three-dimensional and several
modes will occur together within the one-third octave bands. Thus, it is
impossible to predict where the minima will occur.
In the extreme low frequency range – below, the first room mode
(1,0,0) – the sound in the room excites the first room mode as well as the
 Measurement uncertainty  377

1
0.9 Oblique
0.8 Tangential
Axial
Cumulative probability

0.7
Cavity
0.6
0.5
0.4
0.3
0.2
0.1
0
–20 –15 –10 –5 0 5 10
L–L0 (dB)

Figure 13.10 Cumulative probability functions for the sound pressure level in various
types of room modes.

c­ avity mode (0,0,0). The latter has a very simple probability density func-

tion because all parts of the volume have equal sound pressure, so the vari-
ance is zero and the cumulative probability curve is collapsed into a vertical
line at L 0 (Figure  13.10). If the frequency decreases from f 1,0,0 towards
0 Hz, the cumulative p ­ robability curve gradually changes from that of the
pure axial mode to that of the cavity mode. This means that the variance
decreases towards zero. This effect is clearly seen in Figure 13.5.
The statistics of the axial modes is different from that of the other
mode types; the median level equals L 0, so half of all positions in the
room will have sound pressure levels below L 0, and the other half have
levels above L 0. However, the probability density function is strongly
asymmetric; minimum levels are very low in the centre of the room,
whereas maximum levels are only 3 dB higher than L 0. In the low fre-
quency range (region 1 in Figure 13.6), the axial modes (1,0,0), (0,1,0)
and (0,0,1) dominate, and in each mode the sound field is one-dimen-
sional and spatially well defined.
In Table 13.3, the median level and the confidence intervals associated
with the various spatial distribution functions for pure tone excitation of
a room are stated. They are all asymmetric, in contrast to the Gaussian
distribution, which is included for comparison.

13.3.2 Zones to be avoided
Below the Schroeder limiting frequency, the sound field in a room is not
stochastic and it is possible to choose measurement positions that minimize
378  Sound Insulation in Buildings

Table 13.3  Measurement uncertainty (dB), single frequency excitation, for

different confidence levels, calculated for some probability functions
Confidence level (%)
1-sided low
2-sided 1-sided high Axial Tangential Oblique Poisson Gaussian
Displayed in figure 13.10 13.10 13.10 13.9 13.1
Median (50 %) 0.0 1.7 1.5 −1.6 0.0
68 16 −9.2 −4.6 −5.4 −7.6 −4.3
84 2.7 5.0 6.2 2.7 4.3
80 10 −13.0 −6.9 −7.6 −9.8 −5.5
90 2.9 5.4 7.1 3.6 5.5
90 5 −18.9 −10.1 −10.8 −12.9 −7.1
95 3.0 5.7 7.9 4.8 7.1
95 2.5 −24.8 −13.2 −14.0 −16.0 −8.5
97.5 3.0 5.9 8.3 5.7 8.5
99 0.5 −37.5 −20.5 −21.5 −23.2 −11.1
99.5 3.0 6.0 8.7 7.2 11.1

the measurement uncertainty. Figure 13.11 shows the sound pressure level
along the length of a room; the L 0 level occurs in the distance lx /4 from the
walls, and in distances greater than lx /3, the sound pressure level drops more
than 3 dB below L 0. As indicated in Figure 13.11, the centre region should

6
Zone to be avoided
3
L–L0 (dB)

–3

–6

–9

–12
0 1/4 1/3 1/2 2/3 3/4 1
x / lx

Figure 13.11 S patial variation of sound pressure level in a room when the first axial mode
is excited. Measurements in the central zone more than 1/3 room dimen-
sion from the walls should be avoided.
 Measurement uncertainty  379

be avoided in order to achieve a low measurement uncertainty. The conclu-

sion is that measurements at low frequencies are preferably made in the cor-
ner regions of the room where the spatial variation is limited. The precise
distance from the room boundaries is not important.
Above the Schroeder limiting frequency, the picture is different, almost
the opposite, in fact. From Chapter 3, Section 3.4, we know that the sound
pressure level in a diffuse sound field is influenced by interference effects
near the boundaries. With a pure tone excitation, the interference patterns

(a)
6
Zone to be avoided

3
L –L0 (dB)

–3
0 0.25 0.5 0.75 1 1.25 1.5
x/λ
(b)
9
Zone to be avoided
6

3
L –L0 (dB)

–3

–6

–9

–12
0 0.25 0.5 0.75 1 1.25 1.5
r/λ

Figure 13.12 Spatial variation of sound pressure level in a near room boundaries in a dif-
fuse sound field excited by a pure tone. (a) Near a wall a zone < 0.2 λ should
be avoided. (b) Near a corner a zone < 0.7 λ should be avoided.
380  Sound Insulation in Buildings

near a wall and a corner are shown in Figure 13.12. The size of the interfer-
ence zones that should be avoided is 0.2 λ from the wall and 0.7 λ from the
corner. Measurements are preferably made in the central zone of the room,
but in practice it is often impossible to avoid the interference zones unless
the room is big.

13.3.3 The influence of the measurement time

The measurement time t may also give a contribution to the measurement
uncertainty. For band-limited noise with the bandwidth B, the relative vari-
ance of the time-averaged squared sound pressure p 2 is:

2
ε t2 = (13.18)
Bt

where t is the measurement time (or the integration time). The relationship
is also shown in Figure 13.13 in terms of the expanded uncertainty in dB.
The conversion to dB is used (Equation 13.13).

U (Lp )
σ (Lp ) = 4.34 ⋅ ε t = (13.19)
k(g)

1000

Expnaded uncertainty, U(Lp)

for confidence level 95 %
5 min
0.1 dB
Integration time t (s)

2 min
100 0.2 dB

1 min

0.5 dB

1 dB

10
20 50 100 200 500 1000 2000 5000 10000
One-third octave center frequency (Hz)

Figure 13.13 Necessary integration time for measurements in one-third octave bands in

a room. The parameter on the curves is the expanded uncertainty in dB
for 95 % ­confidence level.
 Measurement uncertainty  381

Together with Equation 13.18, we get the measurement time:

2
2  4.34 ⋅ k(g) 
t=  (13.20)
B  U (Lp ) 

In Figure 13.13, two-sided test and k(g) = 1.96 for 95 % confidence level is
assumed (Table 13.1).
The total relative spatial variance is:

ε 2 = ε t2 + ε R2 (13.21)

However, in many cases, it is no problem to avoid the uncertainty contri-

bution due to the measurement time. From Equations 13.5 and 13.18, it
follows that εt2 ≪ εR 2 if the integration time t is:

2
t  0.29 ⋅ T + (13.22)
B
where T is the reverberation time of the room and B is the bandwidth of the
measurement. As the last term typically is not significant, a simple rule of
thumb is that the integration time in each position should be at least three
times the reverberation time of the room.

13.3.4 The integrated impulse response method

New and very efficient measuring methods use either the maximum length
sequence method or the swept-sine method to generate an impulse response,
which is integrated to get the total sound pressure level (see ISO 18233).
There are two main advantages of these methods; the signal-to-noise ratio
is increased and the stochastic nature of the traditional broad-band noise
excitation is avoided. The latter means that the influence of the measure-
ment time on the uncertainty can be neglected (i.e. ε t2  = 0). These new mea-
surement methods correspond to measurements with an infinite integration
time using noise signals.

13.4 SPATIAL AVERAGING

13.4.1 Correlation coefficient
The correlation coefficient R between the sound pressures p1 and p2 taken
in two different positions is defined as:

p1 ⋅ p2
R= (13.23)
p12 ⋅ p22
382  Sound Insulation in Buildings

where the horizontal bars mean time averaged values.

In a three-dimensional diffuse sound field, the correlation coefficient (see
Kuttruff, 1979) is:

sin(kx)
R= (13.24)
kx

where k = 2π f/c is the wave number and x is the distance between the two
positions. The shape of the function is shown in Figure 1.4. It is noted that
R = 0 for kx = π, and R is relatively small for kx > π.
For practical applications, it can be concluded that

λ
R ≅ 0 for x ≥ (13.25)
2

The distance between the microphone positions in a room must be at least

half a wavelength in order to obtain uncorrelated measurement positions.
However, this may be difficult or impossible to fulfil at low frequencies,
especially in small rooms.
The above consideration may also be applied to the distance from the
microphone positions to reflect room surfaces. The reflection from a hard
surface can be interpreted as a sound wave that continues through the surface
and reaches the image of the microphone that is created by a mirror in the
plane of the surface. A measurement position near a reflecting surface should
not be correlated with the image position behind the wall. Thus, in order to
fulfil Equation 13.25, the distance from the surface must be at least λ/4.

13.4.2 Discrete spatial averaging

Because of the spatial variations within a room, the sound pressure mea-
surements are usually taken as the average of the squared sound pressure
over a number of measurement positions, N. If the distance between the
positions is sufficiently large (Equation 13.25), the measurements can be
considered uncorrelated and, thus, the relative variance of the averaged
result is:

ε R2
ε2 = (13.26)
N

If for any reason, the distance between the microphone positions cannot
fulfil the minimum requirement (Equation 13.25), it is possible to calcu-
late the equivalent number of uncorrelated positions according to Lubman
(1969):
 Measurement uncertainty  383

N
Neq = (i ≠ j) (13.27)
1+
1
N ∑ ∑ R (kx )
i j
2
ij

Here, N is the actual number of microphone positions and xij denotes the
distance between positions i and j. So, all possible pair combinations are
considered in the denominator and R can be calculated from Equation
13.24 if the sound field can be assumed a diffuse sound field. It is seen that
in general Neq ≤ N. It is also noted that Neq need not be a natural number.
Figure 13.14 shows examples of calculated equivalent number of mea-
surement positions as a function of the frequency and for different distances
between the actual positions. At high frequencies, it appears that Neq ≈ N,
but at low frequencies, the positions are not fully independent, and Neq < N.
With the positions equally distributed on a circle with radius r as in the
example in Figure 13.14, the equivalent number of uncorrelated positions
can be estimated by this empirical formula:
  3 −1/3 

  4π  
Neq ≅ min {N , Nmax } =min N ,  1 +  rf    (13.28)
   c   
where f is the centre frequency of the measurement band.

Neq
N = 12

10
N=9

N=6

N=3

0
50 100 200 400 800 1600 Hz 3150

Figure 13.14 Equivalent number of uncorrelated measurement positions per third octave

band in a diffuse sound field. The parameter is the actual number of fixed
positions, evenly distributed on a circle with radius of 1.5 m. (Adapted from
Tuominen, H. T., Proceedings of NAS-80, Åbo, Finland, 67–70, 1980. With
permission.)
384  Sound Insulation in Buildings

At very low frequencies in small rooms, it is impossible to obtain suf-

ficient distance between the microphone positions. If the exact positions
in the room are not known, it may be sufficient to apply a general rough
estimate. A realistic number of equivalent uncorrelated positions is esti-
mated from Equation 13.28 together with an estimated maximum possible
radius in the room. For this purpose, it is assumed that the room has a
rectangular floor plan with length/width ratio 1.25 and that the circle with
measurement positions keeps at least 0.7 m from the walls. This leads to the
maximum radius in a room with volume V and height h:

1 V
rmax = − 0.7 m (13.29)
2 1.25 ⋅ h

As an example, a room with volume 50 m3 and height 2.3 m is considered.

This yields the radius rmax = 1.39 m and Neq,max = 5 at 100 Hz. So, if five fixed
microphone positions are used, they actually count at five independent posi-
tions at frequencies above 100 Hz. However, another example room has a
volume of only 26 m3 and height 2.3 m. In this case, the radius rmax = 0.8 m
and Neq,max = 3 at 100 Hz. Even if the number of microphone positions is
more than five, they will only count as 3 at 100 Hz.

13.4.3 Continuous spatial averaging

As an alternative to a number of fixed microphone positions, it is also pos-
sible to use continuous spatial averaging. This is usually done by mounting
the microphone on a stand that can rotate the microphone slowly with a
certain radius, and preferably in a plane that is not parallel to any of the
room surfaces.
Assuming a three-dimensional diffuse sound field, i.e. by applying the cor-
relation coefficient (Equation 13.24), the equivalent number of uncorrelated
positions has been calculated by Lubman et al. (1973). The results are shown
in Figure 13.15 for a straight line segment and for a circular path. The length
of the path is L, which in the case of the circle means L = 2πr, where r is the
radius. It appears that for the circular path, a good approximation is:

 λ
 1 for L≤
 2
Neq ≅  (13.30)
 2L λ
for L>
 λ 2

Some formulas to estimate the equivalent number of positions for various

kinds of continuous spatial averaging are collected in Table 13.4. The circle
path is often used for sound insulation measurements, especially in labora-
tories, (see Figure 12.3).
 Measurement uncertainty  385

30

10

Neq
Neq = 1+2L/λ

3 Straight line segment Neq = 2L/λ

Circular path

1
0.1 0.3 1 3 10
L/λ

Figure 13.15 Equivalent number of uncorrelated measurement positions in a diffuse

sound field using continuous spatial averaging. Results are shown for
a straight line segment as well as for a circular path. (Reproduced from
Lubman, D. et al., J. Acoust. Soc. Am., 53, 650–659, 1973, Figure 3. With the
permission of the Acoustical Society of America).

Table 13.4  Approximate formulae to estimate the equivalent number of

measurement positions by continuous averaging over various
paths or areas in a diffuse sound field
Region of averaging Approximate Neq
Line, length L 1 + 2L/λ
Circle, circumference L 2L/λ
Circular disc, area S 1 + S/(λ/2)2
Rectangular surface, area S S/(λ/2)2
Rectangular box, volume V V/(λ/2)3
Source: Lubman, D., J. Sound Vib., 16, 43–58, 1971.


13.5 MEASUREMENT UNCERTAINTY ON
REVERBERATION TIME MEASUREMENTS

13.5.1 General
The uncertainty of decay rate measurements has been studied by Davy
et al. (1979) and Davy (1980). The decay rate d in dB per second is related
to the reverberation time T in seconds by d = 60/T.
386  Sound Insulation in Buildings

The measurements can be repeated in the same position, and n denotes

the number of decays measured in each position. The ensemble variance in
one position is εe2(d).
The spatial variance is εs2(d). The number of independent measurement
positions is denoted N. Thus, the estimated relative standard deviation of
the average reverberation time is (T):

σ (T ) 1 ε s2 (d) ε e2 (d) (13.31)
= +
T N d2 n d2

13.5.2 The interrupted noise method

In the work of Davy et al. (1979), the ensemble variance and the spatial variance
were derived for the case of measurements using the interrupted noise method.
By insertion of the results from this reference in Equation 13.31, we get:

σ (T ) 1+ H / n
=G (13.32)
T N BT
where
T is the reverberation time, in seconds.
σ (T) is the standard deviation of T, in seconds.
G and H are constants that depend on the evaluation range.
n is the number of decays measured in each position.
N is the number of independent measurement positions.
B is the bandwidth, in Hz.

The constants G and H depend on the evaluation range D and a parameter

γ = T/Tdet, which is the ratio of measured reverberation time and the rever-
beration time inherent in the measuring apparatus. Tdet is the reverberation
time of the averaging detector. For some typical values of D and γ, the val-
ues of the constants G and H can be taken from Table 13.5.
The standard ISO 3382-2 defines three levels of accuracy with a corre-
sponding minimum number of measurement positions and decays in each
position (Table 13.6).

Table 13.5  Values of the constants G and H for different evaluation

ranges, D
H
Evaluation range, D (dB) G γ = 3 γ = 5 γ = 10
10 1.75 2.67 3.32 3.87
20 0.88 1.72 1.90 2.04
30 0.55 1.42 1.52 1.59
 Measurement uncertainty  387

Table 13.6  M inimum number of positions N and decays n for each of the three
quality levels of measurements according to ISO 3382-2
Survey Engineering Precision
No. of positions, N 2 6 12
No. of decays, n (interrupted noise method) 1 2 3
Equivalent no. of decays, n (integrated 10 10 10
impulse response method)

0.20

s(T20)
0.15 s(T30)
Standard deviation (s)

0.10

0.05

0.00
50 100 200 500 1000 2000 5000
One-third octave center frequency (Hz)

Figure 13.16 Standard deviation of measured reverberation time in one-third octave

bands if the reverberation time is 1.0 s, using average of six positions and
the interrupted noise method with two decays in each position. Results are
shown for T20 and T30.

An example of the application of Equation 13.32 together with Table 13.5

is the estimation of the standard deviation of T20 with the parameter γ = 5.

1 + (1.90 / n)
σ (T20 ) = 0.88 ⋅ T20 (13.33)
N BT20

This and the analogue result for σ(T30) are shown as functions of the fre-
quency in Figure 13.16 for the example of T20 = T30 = 1.0 s.

13.5.3 The integrated impulse response method

For the integrated impulse response method, the ensemble variance is theo-
retically ε e2 (d) = 0. This corresponds to the averaging of an infinite number
of excitations in the same position, if the interrupted noise method had
388  Sound Insulation in Buildings

been used (see Davy, 1980). For the estimation of the standard deviation of
a measurement result, Equation 13.32 may be used with a value of n = 10.

13.6 MEASUREMENT UNCERTAINTY ON
SOUND INSULATION MEASUREMENTS

13.6.1 Combined measurement uncertainty

The sound reduction index is determined from three sets of measurements:
the average sound pressure level in the source room, the average sound
pressure level in the receiver room and the average reverberation time in the
receiver room. The result is calculated from:

S ⋅T
R = Lp1 − Lp2 + 10 lg (dB) (13.34)
0.16 ⋅ V

where S is the area of the partition wall or floor (m 2) and V is the volume of
the receiver room (m3). It is noted that the uncertainty calculation will be
the same in case of DnT, as the difference is only whether the partition area
and the room volume are included or not.
The standard deviation of the last term is estimated from:

 σ (T )  σ (T )
σ (10 lg T ) = 10 lg  1 +  ≅ 4.34 ⋅ (dB) (13.35)
 T T

where σ(T) is from Equation 13.32. The approximation is correct within

10 % for σ (T )/T < 0.6.
The standard deviation for each of the two sound pressure levels is cal-
culated from Equations 13.10, 13.11 and 13.12 and using the equivalent
number of positions (Equations 13.28 and 13.29). For the source room,
this yields:

ε R,1
σ (Lp1) = 4.34 ⋅ (dB) (13.36)
Neq,1

and similarly for the receiver room. Thus, the standard deviation of the
reduction index can be estimated from:

2
 σ (T ) 
σ (R) = σ 2 (Lp1) + σ 2 (Lp2 ) +  4.34 ⋅  (dB) (13.37)
 T 

Similarly, the standard deviation of the impact sound pressure level can be
estimated from:
 Measurement uncertainty  389

2
 σ (T ) 
σ (Ln,T ) = σ 2 (Lp2 ) +  4.34 ⋅  (dB) (13.38)
 T 

The measurement uncertainty of impact sound has been studied by

Hagberg and Thorsson (2010) with special emphasis on the low frequen-
cies. They found that the reverberation time had nearly normal distribution
above 100 Hz, but at lower frequencies the distribution functions were not
symmetric.
For a more detailed description, see Machimbarrena et al. (2015).
Analysing a large set of measurement data from field measurements of
partition walls, it is concluded that the measurement uncertainty of the
airborne sound insulation, to some extent, depends on the type of wall
construction. Thus, individual uncertainty calculations should be made
for each measurement situation instead of applying generalised uncertainty
data as given in ISO 12999-1.

13.6.2 Example of airborne sound

insulation measurement
Source room, L p1: V = 500 m3, h = 4 m, T = 1 s, 2 source pos. × 5 mic. pos. = 10
positions
Receiver room, L p2: V = 50 m3, h = 2.3 m, T = 0.5 s, 2 source pos. × 5 mic.
pos. = 10 positions
Receiver room, T20: T = 0.5 s, 2 × 3 = 6 positions, interrupted noise, two
decays per position
The results are shown in Figure 13.17 and Table 13.7 for the frequency
range from 50 Hz to 5 kHz.
The maximum number of independent microphone positions in the
receiver room (50 m 3) is only about N = 2.5 at 50 Hz. Combined with
two source positions, this means that in this room N = 2 × 2.5 = 5 at
50 Hz, increasing to N = 10 at 125 Hz and higher frequencies. In the
source room, N = 10 at all frequencies because the volume is sufficiently
large.

13.6.3 Measurement uncertainty on
single-number quantities
Wittstock (2007, 2009) studied the statistical distribution of measured air-
borne sound insulation, and he found that the distribution function of the
transmission loss R′ is very close to the Gaussian distribution, much closer
than the distribution function of the transmission coefficient τ. So, it makes
sense to apply the statistical methods for uncertainty directly to the trans-
mission loss in dB.
390  Sound Insulation in Buildings

4
Standard deviation σ(R) (dB)

0
50 100 200 500 1000 2000 5000
Centre frequency, one-third octave bands (Hz)

Figure 13.17 Calculated standard deviations in one-third octave bands for the example
case of measured airborne sound insulation.

The standard deviation for in situ measurements of airborne sound insu-

lation can be estimated to σ (Rw′ ) = 0.9 dB to 1.3 dB (ISO 12999-1, Annex
B). Similarly, for field measurement of impact sound insulation, the esti-
mated standard deviation is σ (Ln,w
′ ) = 1.0 dB. However, these estimates may
be a little too optimistic. Wittstock and Scholl (2009) have analysed the
uncertainty of transmission loss measurements and compared them with
calculated uncertainties for the same cases. For 24 heavyweight construc-
tions, the uncertainty was u(Rw′ ) = 1.4 dB to 2.1 dB, whereas for 24 light-
weight constructions, the uncertainty was u(Rw′ ) = 1.5 dB to 2.7 dB.

13.6.4 Application of uncertainties in relation

to requirements for sound insulation
The application of measurement uncertainty is highly relevant in connec-
tion with field measurements of sound insulation when the purpose, is to
prove fulfilment of sound insulation requirements. For this purpose, we
will use an example, adapted from Wittstock (2009).
In this case, the airborne sound insulation shall fulfil the limit
′  = 53 dB. Then, the measured transmission loss Rw′ must fulfil the
Rw,req
equation:

′ (13.39)
Rw′ − k ⋅ u ≥ Rw,req
 Measurement uncertainty  391

Table 13.7  C
 alculated standard deviations in one-third octave bands for each of
the terms in Equation 13.37
Frequency (Hz) σ(Lp1) (dB) σ(Lp2) (dB) σ(10 log T) (dB) σ(R) (dB)
50 1.0 3.3 0.9 3.6
63 0.9 4.1 0.8 4.3
80 0.7 3.9 0.7 4.0
100 0.7 3.5 0.6 3.6
125 0.6 3.0 0.6 3.1
160 0.5 2.3 0.5 2.5
200 0.5 2.1 0.5 2.2
250 0.5 1.9 0.4 2.0
315 0.4 1.8 0.4 1.8
400 0.4 1.6 0.3 1.6
500 0.3 1.4 0.3 1.5
630 0.3 1.3 0.3 1.3
800 0.3 1.1 0.2 1.2
1000 0.2 1.0 0.2 1.1
1250 0.2 0.9 0.2 1.0
1600 0.2 0.8 0.2 0.9
2000 0.2 0.7 0.1 0.8
2500 0.1 0.7 0.1 0.7
3150 0.1 0.6 0.1 0.6
4000 0.1 0.5 0.1 0.5
5000 0.1 0.5 0.1 0.5
Note: Input data according to the example.


According to ISO 12999-1, Table 3, the standard uncertainty for in situ

measurements can be estimated to u = σsitu = 0.9 dB. The coverage factor k
is found in Table 13.1 for a one-sided test, but we have to agree on which
confidence level is relevant for the case. If the confidence level is 90 %, the
coverage factor is k = 1.28. Therefore, the extended uncertainty is U = k × u 
= 1.28 × 0.9 ≈ 1.2 dB.
A measurement is made and the result is Rw′  = 53.7 dB. However, it can-
not be decided whether the requirement is fulfilled or not because Rw′  − 1.2
dB = 52.5 dB < 53 dB.
A second independent measurement is made, i.e. with different equip-
ment used by different staff. The results is Rw′  = 54.1 dB. Still, this is not
sufficient by itself (54.1 dB – 1.2 dB = 52.9 dB < 53 dB). However, with
independent measurements, it is possible to use the mean value of the trans-
mission loss, and the combined uncertainty is reduced by √n, where n = 2 is
the number of independent measurements. The result is:
392  Sound Insulation in Buildings

k ⋅ u 53.7 + 54.1 1.2

′
Rw,mean − = − = 53.9 − 0.8 = 53.1 dB > 53 dB
n 2 2

The requirement is found to be fulfilled with 90 % confidence level.

REFERENCES

J.L. Davy, I.P. Dunn, P. Dubout (1979). The Variance of Decay rates in Reverberation
Rooms. Acustica 43, 12–25.
J.L. Davy (1980). The variance of impulse decays. Acustica 44, 51–56.
P.E. Doak (1959). Fluctuations of the sound pressure level in rooms when the receiver
position is varied. Acustica 9, 1–9.
GUM (1995). Guide to the Expression of Uncertainties in Measurement, 2nd
Edition, International Organization for Standardization, Geneva, Switzerland.
K. Hagberg and P. Thorsson (2010). Uncertainties in standard impact sound measure-
ment and evaluation procedure applied to light weight structures. Proceedings
of ICA 2010, August 23–27, 2010, Sydney, Australia.
ISO 3382-2 (2008). Acoustics – Measurement of room acoustic parameters – Part
2: Reverberation time in ordinary rooms, International Organization for
Standardization, Geneva, Switzerland.
ISO 12999-1 (2014). Acoustics – Determination and application of measurement
uncertainties in building acoustics – Part 1: Sound insulation, International
Organization for Standardization, Geneva, Switzerland.
ISO 18233 (2006). Acoustics – Application of new measurement methods in building
and room acoustics, International Organization for Standardization, Geneva,
Switzerland.
F. Jacobsen (1979). The diffuse sound field. Statistical considerations concerning the
reverberant field in the steady state. Report No. 27. The Acoustics Laboratory.
Technical University of Denmark.
H. Kuttruff (1979). Room Acoustics, 2nd Edition, Applied Science Publishers,
London.
D. Lubman (1968). Fluctuations of sound with position in a reverberant room.
Journal Acoustical Society America 44, 1491–1502.
D. Lubman (1969). Spatial averaging in a diffuse sound field. Journal Acoustical
Society America 46, 532–534.
D. Lubman (1971). Spatial averaging in sound power measurements. Journal of
Sound Vibrations 16, 43–58.
D. Lubman (1974). Precision of reverberant sound power measurements. Journal of
Acoustical Society America 56, 523–533.
D. Lubman, R.V. Waterhouse, C.S. Chien (1973). Effectiveness of continuous spa-
tial averaging in a diffuse sound field. Journal Acoustical Society America 53,
650–659.
M. Machimbarrena, C.R.A. Monteiro, S. Pedersoli, R. Johansson, S. Smith (2015).
Uncertainty determination of in situ airborne sound insulation measurements.
Applied Acoustics 89, 199–210.
 Measurement uncertainty  393

S. Pedersen, H. Møller, K. Persson Waye (2007). Indoor measurements of noise at

low frequencies – Problems and solutions. Journal of Low Frequency Noise,
Vibration and Active Control 26(4), 249–270.
A.D. Pierce (1989). Acoustics, An Introduction to Its Physical Principles and
Applications. 2nd Edition. Acoustical Society of America, New York.
J.H. Rindel (1981). Stationære lydfelter i rum. (Stationary sound fields in rooms, in
Danish). Lecture Note S. The Acoustics Laboratory. Technical University of
Denmark.
M.R. Schroeder (1969). Effect of frequency and space averaging on the transmis-
sion responses of multimode media. Journal Acoustical Society America 46,
277–283.
C. Simmons (2012). Uncertainties of room average sound pressure levels mea-
sured in the field according to the draft standard ISO 16283-1. Noise Control
Engineering Journal 60, 405–420.
H.T. Tuominen (1980). Practical methods for predicting and checking the confidence
intervals of sound insulation measurements. Proceedings of NAS-80, June
10–12, 1980, Åbo, Finland, 67–70.
R. Waterhouse (1970), Sampling statistics for an acoustic mode. Journal Acoustical
Society America 47, 961–967.
V. Wittstock (2007). On the uncertainty of single-number quantities for rating air-
borne sound insulation. Acta Acustica – Acustica 93, 375–386.
V. Wittstock (2009). Uncertainties in applied acoustics – Determination and han-
dling. Proceedings of NAG/DAGA 2009, March 23–26, Rotterdam, The
Netherlands, 22–29.
V. Wittstock and W. Scholl (2009). Determination of the uncertainty of predicted
values in building acoustics. Proceedings of NAG/DAGA 2009, March 23–26,
Rotterdam, The Netherlands, 920–923.
Chapter 14

Noise effects and subjective

evaluation of sound insulation

In this chapter, the sound insulation in buildings is analysed from the

­point  of view of the users. The discussion is restricted to dwellings and,
in particular, multi-unit houses. The most serious problem may be the
exposure to traffic noise and the need for sound insulation of windows
and façades. Noise from neighbours and the need for sound insulation of
internal walls and floors are also very important and lead to the question:
Which level of sound insulation should be required to obtain satisfactory
acoustical conditions?

14.1 NOISE AND HEALTH

14.1.1 Noise descriptors
International standard ISO 1996-1 defines noise descriptors applied to
environmental noise, for example, for noise regulations and noise map-
ping. They are all some kind of time-averaged, A-weighted sound pressure
level in the free field. The latter means that even if the sound is measured or
calculated close to a building, the result is corrected for the possible influ-
ence of reflections from that building.
If we do not take into account the distribution over day and night, the
sound energy is integrated over 24 h and we get Lp,A,24 h. However, preferred
descriptors do take into account this distribution. The day-evening-night
level L den is defined as:

 1
(
 24 h tday ⋅ 10
0.1 Lday,12
+ tevening 

Lden = 10 lg   , dB (14.1)

 

 ⋅100.1 (Levening,4 +5 dB) + t night ⋅ 100.1 (Lnight,8 +10 dB) ) 


where tday, tevening, tnight are the number of hours of the periods (24 h in total),
and L day, L evening and L night are the A-weighted, time-averaged sound pres-
sure levels in the same periods. Note the penalty of 5 dB in the evening and

395
396  Sound Insulation in Buildings

Table 14.1  Relation between some noise descriptors valid for road
traffic noise with typical example of traffic distribution
78 % (day, 12 h), 11 % (evening, 3 h) and 11 % (night, 9 h)
Level (dB) Difference (dB)
Lp,A,24 h 55 0
Lday,12 h 57 2
Lday,15 h 57 2
Levening,3 h 54 −1
Lnight,9 h 50 −5
Lden 58 3
Ldn 58 3

10 dB in the night. The definition of the three periods may differ from one
country to another; for example, the EU member states use 3 h evening and
9 h night periods. Sometimes, the evening period is not used at all, and we
get the day-night level L dn defined as:

 1
Ldn = 10 lg 
 24 h
( ) 
tday ⋅ 100.1 Lday,15 + t night ⋅ 100.1 (Lnight,9 +10 dB)  dB (14.2)


For road traffic noise having typical traffic distribution, the relation
between the noise descriptors is shown in Table 14.1.

14.1.2 Sleep disturbance
The World Health Organization has identified neighbour noise as a health
problem based on the findings from a large survey in eight European cities
from 2002 to 2003, the Large Analysis and Review of European hous-
ing and health Status (LARES) project (WHO, 2007). Information was
collected in 3382 households from 8539 people by interviews and health
questionnaire. Of all responding residents, 24 % reported that noise expo-
sure at night was a main reason for sleep disturbance. Traffic noise was
the dominant cause of sleep disturbance, closely followed by noise from
neighbouring flats (Figure 14.1).
While traffic noise has since long been known to be a serious source of
annoyance, the LARES investigation showed that also chronic annoyance
due to neighbour noise is associated with hypertension, depression, and
migraine (Maschke and Niemann, 2007). They concluded that neighbour
noise annoyance is a highly underestimated risk factor for healthy housing.
Noise from neighbours may affect people in different ways. Figure 14.2
illustrates how some people may react with curiosity, changing to annoy-
ance and anger and, in severe cases, ending with hatred and other similar
 Noise effects and subjective evaluation of sound insulation  397

50

45
Percentage with sleep disturbance
40

35

30

25

20

15

10

0
Traffic Neighbour flat Parking Surrounding Animals Staircase use
areas

Figure 14.1 Sleep disturbance in adult residents and main sources of disturbing noise.
Total for eight European cities (n = 8539). (Adapted from WHO, Large Analysis
and Review of European Housing and Health Status (LARES). Preliminary over-
view, WHO Regional Office for Europe, Copenhagen, Denmark, 2007, Figure
37. With permission.)

Level 1 Level 2 Level 3

Noise begins or Noise continues or Noise gets worse
is relatively becomes louder still
quiet

Annoyance, anger, Hatred, hostility, desire

bitterness, feeling for revenge, bloodshed
Curiosity... (I could murder,
aggrieved
strangle, kill...)

... Or irritation ... Or tension, pressure, Depression, tiredness,

frustration, taking it out on
resignation, feeling others, feeling upset
fraught and anxious and frightened

Figure 14.2 Noise reaction process. (Reproduced with permission from Grimwood,

C.J., Effects of environmental noise on people at home. BRE information
paper IP 22/93, 1993. Public sector information licensed under the Open
Government Licence v3.0.)

reactions. Other people may react with irritation, growing to tension and
depression. Noise annoyance is related to health (Niemann et al., 2006):
annoyance → negative emotional reaction → neuro-vegetative-hormonal
regulatory disturbances → illness.
398  Sound Insulation in Buildings

14.1.3 Dose-effect relationships
Risk ratio (RR) and odds ratio (OR) are used in statistics, for instance,
to calculate dose-effect relationships. With reference to Table 14.2, the
investigated people are divided into groups (e.g. based on different level
of noise exposure), and the data are collected to show how many exhibit
a certain effect or not (e.g. stroke). Group 0 is a reference group, in which
the dose-effect relationship is assumed to be negligible. In each group, the
risk of being affected is a number between 0 and 1, and the RR is the ratio
between the risk in that group and the risk in the reference group. For
example, RR = 1.05 means that there is an increased risk of 5 % compared
to the reference group.
The odds for being affected in a group are simply the ratio between the
number being affected and the number not affected. The OR in a group is
then the odds in that group divided by the odds in the reference group. If
the effect is sufficiently rare (Yes << No), the risk in all groups may be much
less than unity and then RR ≈ OR.
Annoyance data for 8.6 million people living in London between 2003
and 2010 were analysed in relation to road traffic noise exposure in noisy
areas, L day (07–19) > 60 dB relative to quiet areas L day < 50 dB (Halonen
et al., 2015). Deaths were 4 % more common among adults and elderly
in noisy areas compared with quiet areas (Table 14.3). Adults living in
noisy areas were 5 % more likely to be admitted to hospital for stroke
compared to those living in quiet areas and went up to 9 % in the elderly
population.
In addition to annoyance, sleep disturbance and health problems, traffic
noise may have negative effects on children’s behaviour. In a large study, in
Denmark, 46 904 children with behavioural problems at the age of 7 years
were identified, and address from birth to 7 years was used to model the
traffic noise exposure. It was found that road traffic noise in early childhood

Table 14.2  Calculation of RR and OR for being affected (symptom, disease etc.)
when belonging to a certain group (e.g. based on level of noise exposure)
Effect
Yes No Risk Odds RR OR
Y0 Y0
Group 0 (reference) Y0 N0 1 1
(Y0 + N0 ) N0
Y1 Y1 Y1(Y0 + N0 ) Y1N0
Group 1 Y1 N1
(Y1 + N1) N1 Y0 (Y1 + N1) Y0 N1
Y2 Y2 Y2 (Y0 + N0 ) Y2 N0
Group 2 Y2 N2
(Y2 + N2 ) N2 Y0 (Y2 + N2 ) Y0 N2
 Noise effects and subjective evaluation of sound insulation  399

 isk ratio of death or stroke due to road traffic noise L day > 60 dB
Table 14.3  R
relative to L day < 50 dB
Risk ratio Confidence
(RR) interval 95 %
Death Adults (>25) 1.04 (1.00–1.07)
Elderly (>75) 1.04 (1.00–1.08)
Stroke Adults (>25) 1.05 (1.02–1.09)
Elderly (>75) 1.09 (1.04–1.14)
Source: Halonen, J.I. et al., Eur. Heart J., 36(39), 2653–2661, 2015.

(0–7 years) may be associated with hyperactivity and inattention symptoms

(Hjortebjerg et al., 2016: Figure 14.3). The results were adjusted for several
possible confounders.
The results shown in Figure 14.3 are divided into groups of noise
­exposure, L den (<50, 50–55, 55–60, 60–65, ≥65) dB, the first one being
the reference group. There is a monotonic increase of the odds ratio until
60 dB to 65 dB, but the curve levels out and decreases for the highest
exposure group. For noise levels due to road traffic between 50 dB and
65 dB, the relation with the children’s behavioural problems is strong and
significant.
A possible explanation for the weaker relationship above 65 dB may be that
the noise exposure in this investigation was calculated without screens or noise

1.6

1.5

1.4

1.3
Odds ratio

1.2

1.1

0.9

0.8
50 55 60 65
Exposure to road traffic noise, Lden (dB)

Figure 14.3 Relation between exposure to road traffic noise at childhood and abnor-
mal scores on total difficulties shown as odds ratio with 95 % confidence
intervals in four noise exposure categories. The reference category is Lden
< 50 dB. (Reproduced from Hjortebjerg, D. et al., Environ. Health Perspect.,
124(2), 228–234, 2016. With permission.)
400  Sound Insulation in Buildings

barriers; but, in fact, such measures for noise protection of housing areas with
high noise exposure are common in Denmark since the 1970s. In the act of
environmental protection, it was stated that the number of dwellings exposed
to noise levels more than 65 dB from road traffic should be reduced to a mini-
mum of 50 000 before 2010. The typical attenuation of road traffic noise due
to a noise barrier is 8 dB to 15 dB. If the result in the highest noise group in
Figure 14.3 is shifted downwards by 12 dB, it fits precisely on the curve.

14.2 NOISE FROM NEIGHBOURS

14.2.1 Acoustical comfort
The concept of acoustical comfort in relation to buildings was first used by
Cummins (1978). Acoustical comfort is characterised as follows (Rindel
and Rasmussen, 1995):

• Absence of unwanted sound

• Presence of wanted sound of desired level and quality
• Opportunities for acoustic activities without annoying or disturbing
other people and without being heard by unauthorised persons

In order to achieve acoustical comfort in a building, certain requirements

have to be fulfilled concerning the airborne and impact sound insulation
and the noise level from traffic and building services.
It is important to observe that acoustic comfort for a person is related to
the person not only as a receiver of sound but also as a source of sound. It
can be annoying to be exposed to noise from the neighbours, but it can be
equally annoying to know that your activities can be heard by other people
and may cause annoyance. Poor sound insulation between dwellings can be
a cause of conflicts and a cause of restraints of activities.
Playing a musical instrument is an interesting example. Many countries
have music schools for children, and it is important that children have the
possibility to learn to play a musical instrument. However, if the family
lives in a multi-unit house with poor sound insulation, it can be a serious
problem to let the children practice their instruments at home.
Another example is a family with small children living in an apartment
house with mediocre sound insulation. Normal activity of children includes
running, jumping, shouting and crying. However, a noise-sensitive neigh-
bour may bang the wall or the ceiling to signalize disturbance by the noise.
So, the parents will try to keep the children quiet. If this becomes a concern
all the time, the family can eventually feel forced to move and find another
place to live. While sound conditions usually play a minor role when choos-
ing a new home, they can often be the main reason for moving away.
A study was undertaken in England and Wales 1992–1994 in order to
investigate complaints about poor sound insulation between dwellings
 Noise effects and subjective evaluation of sound insulation  401

(Grimwood, 1997). 40 cases of complaints were examined by sound insu-

lation measurements and interviews with complainants and adjacent resi-
dents. Most common types of noise sources heard, and the percentage of
cases where reported were: Music, television, radio (98 %), footfalls on
floors (95 %), voices (78 %), doors and cupboard doors (68 %).
In a social survey of noise conditions in dwellings in Norway (Løvstad
et al., 2016), some questions were related to whether people were concerned
about disturbing their neighbours when having visitors or children playing,
when listening to music or playing a musical instrument, and when walk-
ing on the floor. Other questions were dealing with noise annoyance. The
results are shown in Figure 14.4.
The main sources of noise annoyance are traffic noise and impact
sound (footfall), both with 20 % annoyed (moderately + very + extremely).
Airborne noise from neighbours (speech, TV, music, etc.) caused around
10 % annoyance. However, comparing the degree of annoyance with
the degree of concern about disturbing, it is remarkable that more than
30 % were concerned (to some extend or very much) about disturbing
their neighbours with speech, children, music, and other airborne sounds

35

30
Percentage annoyed (moderately + very + extremely)

25

20

15

10

0
Speech, etc. Music, etc. Footfall Traffic noise Noise (all
sources)

Figure 14.4 Percentage of people being annoyed by various noise sources in their

home (n = 702). (Adapted from Løvstad et al., Sound quality in dwellings
in Norway  – A socio-acoustic investigation. Proceedings of BNAM 2016,
Stockholm, Sweden, 2016.)
402  Sound Insulation in Buildings

40

35

30

25
Percentage

20

15

10

0
Speech, etc. Music, etc. Footfall
Annoyed (moderately + very + extremely) Concerned to disturb neighbours

Figure 14.5 Percentage of people being annoyed by various noise sources in their home,
and percentage of people being concerned that they may disturb their neigh-
bours by various activities (n = 702). (Adapted from Løvstad et al., Sound
quality in dwellings in Norway – A socio-acoustic investigation. Proceedings
of BNAM 2016, Stockholm, Sweden, 2016.)

(Figure  14.5). More than 15 % were concerned (to some extent or very
much) about disturbing their neighbours due to footfall noise. These find-
ings clearly demonstrate that the constraints dwellers put on their own
activities due to insufficient sound insulation are at least as important as
the annoyance from noise. Noise from neighbours is like a troll with two
heads; the fear of disturbing the neighbour is half the problem.

14.2.2 Sound insulation and social disruption

An interesting survey was carried out in Canada (Bradley, 1983). The
investigation comprised 49 party walls and interviews of the persons living
on either side of the walls, in total 98 persons distributed over 11 locali-
ties, mainly in two-story terraced houses, 16 % multi-storey dwellings.
Measured sound insulation of the party walls varied from 39 dB to 60 dB.
The measured values were sound transmission class that is approximately
equal to R′w (see Section 5.8.1). The measurements included the record-
ing of A-weighted noise levels for one 24 h period in each subject’s liv-
ing room. The survey illuminates many interesting aspects of the identified
noise problems. For example, it appears from the investigation that the
residents’ perception of whether a neighbour shows consideration or not
 Noise effects and subjective evaluation of sound insulation  403

has a significant correlation with the measured sound insulation, but not
with the actual noise level at the neighbours. Bradley (1983) states:
‘This is strong evidence that in this study at least the inadequacy
of the party wall was a source of social disruption in that neigh-
bors were thought to be inconsiderate when it was really the
party wall that was at fault.’

14.2.3 Influence of age and lifestyle

Noise protection in multi-family houses from the point of view of the resi-
dent is the title of a survey in Germany (Weeber et al., 1986). The investi-
gation comprised 16 housing estates of varying ages, where measurements
were carried out and 471 residents were interviewed. The airborne sound
insulation R′w varied between 42 dB and 61 dB and the impact sound insu-
lation L′n,w of the separating floors varied between 70 dB and 26 dB. The
houses were of different age, built between 1921 and 1982. The acoustical
conditions in new and older dwellings were assessed using very different cri-
teria. It was found that the age and condition of the building had i­ nfluences
on the expectations and thereby also on the residents, assessment.
The main conclusion from this investigation is that to achieve conditions
with no disturbance in normal situations, the airborne sound insulation
of the floors should be R′w ≥ 60 dB, whereas R′w = 50 dB would cause about

100

80
Percentage dissatisfied

60

40

20

0
40 45 50 55 60 65
Vertical sound insulation, R΄w (dB )

Figure 14.6 Proportion of residents in multi-storey flats who rate the sound insulation
as poor, presented as a function of the measured airborne sound insulation
through the separating floors. Average for 16 housing estates and regression
line (n = 471). (Weeber, R. et al., Schallschutz in Mehrfamilienhäusern aus der Sicht
der Bewohner (Sound insulation in multi-family houses from the point of view
of the resident, in German). F 2049, IRB Verlag, Stuttgart, Germany, 1986.)
404  Sound Insulation in Buildings

30 % of the residents to be frequently disturbed by the neighbour’s TV.

Impact sound insulation L′n,w of the separating floors at L′n,w = 50 dB would
cause about 15 % to be frequently disturbed by the neighbour walking on
the floor. Figure 14.6 shows the relation between the percentage of people
finding the sound insulation insufficient and the airborne sound insula-
tion of the separating floors. Each dot represents one of the 16 houses.
Three outliers were identified (the open dots) and could be explained as
follows: One site with higher dissatisfaction than the main trend was an
old, refurbished house with new, young tenants with children; little out-
door noise. The other two outliers with less dissatisfaction were houses
with elderly tenants, no children, and in addition relatively high levels of
outdoor noise.

14.2.4 Willingness to pay for good sound insulation

The Canadian investigation mentioned above (Bradley, 1983) also included
a question about the money, and the result showed a high degree of cor-
relation between the measured sound insulation and the amount of money
which residents were prepared to pay extra per month to eliminate disturb-
ing noise. The regression line found reaches zero only at a sound insulation
of 60 dB.
In 1995, an investigation was made in Sweden in order to find what
level of sound insulation new dwellings should have (Wibe, 1997). The
minimum requirements for sound insulation in Sweden at that time were
R′w ≥ 55 dB (row houses), 53 dB (flats, vertical), 52 dB (flats, horizontal)
and L′n,w ≤ 58 dB (between dwellings), 64 dB (from stairways and corri-
dors). A total of 2 322 questionnaires were used for the analysis: 65 %
of the p­ articipating people lived in multi-family houses, 20 % in single
houses, 10 % in row houses and 5 % in other kinds of houses. One of
the main ­questions was about the ­willingness to pay a higher rent if the
sound ­insulation of the ­apartment could be significantly improved. The
average answer was about 2 500 (Sweden Kroner) per year. In summary, it
was found that if the sound ­insulation of the dwelling could be improved,
around 60 % of the ­population were willing to pay on average a 10 %
higher rent.
A survey in Norway 2015 also included questions about willingness
to pay for better sound insulation (Løvstad et al., 2016). Among 702
replies, around 25 % would not pay extra, but around 30 % were will-
ing to pay from 1 200 to 12 000 (Norway Kroner) extra per year for
better sound insulation. The corresponding minimum requirements for
sound insulation in Norway were R′w ≥ 55 dB and L′n,w ≤ 53 dB. To the
opposite question, less sound insulation for a reduced monthly payment,
66 % replied that this would be out of question; only 10 % would accept
lower sound insulation against a lower payment around 12 000 NOK
per year.
 Noise effects and subjective evaluation of sound insulation  405

14.3 DOSE-RESPONSE FUNCTIONS

14.3.1 Statistical methods
Two different statistical models are usually considered, probit or logit. The
probit model is based on the normal distribution (the Gaussian d ­ istribution:
see also Chapter 13). However, when cumulative probability functions
are evaluated and adapted to observed data, it is a disadvantage that these
­functions are in integral forms. Instead, the logit model is based on slightly
different distributions often assumed in social surveys. The principle behind
the logit model is to consider the data to be analysed in ­proportions, ­preferably
with equal uncertainty attached to each proportion. The logit transforma-
tion achieves this. It is defined by Altman (1991, p. 145) as follows:

 p 
t = logit(p) = ln  (14.3)
 1 − p 
et
⇒ p= (14.4)
1 + et
where p is a proportion and t is the logit. The logit transformation stretches out
the lower and upper parts of the distribution in the same way as in the Gaussian
distribution (Table 14.4). They both belong to the group of sigmoid functions
(S-shaped curves), which are used to model dose-response relationships.
The logistic function has the probability density function:
z− µ
−
e s
w(z) = (14.5)
z− µ 2
 − 
s 1 + e s 
 
where the mean value is μ and s is a parameter proportional to the standard
deviation σ. The cumulative distribution function is:

Table 14.4  E
 ffect of logit transformation
of a proportion p
p logit(p)
0.01 −4.60
0.05 −2.94
0.10 −2.20
0.25 −1.10
0.50 0.00
0.75 1.10
0.90 2.20
0.95 2.94
0.99 4.60
406  Sound Insulation in Buildings

(a) (b) 1.0

0.5 0.8
0.4

0.3 0.6
w(z)

0.2

P(z)
0.1 0.4
0.0
–3 –2 –1 0 1 2 3
Logistic Gauss 0.2
Measurand, z

0.0
–3 –2 –1 0 1 2 3
Logistic Gauss
Measurand, z

Figure 14.7 Logistic function and comparison with the Gaussian distribution.

(a) Probability density functions; (b) cumulative distribution functions.

1 1 1  z − µ  (14.6)
P(z) = = + tanh 
−
z− µ
2 2  2s 
1+ e s

1 2 2
This is a symmetric distribution with the variance σ = s π . The func-
2
3
tions are shown in Figure 14.7, and it is seen that the cumulative distribution
is very close to that of a Gaussian distribution. An advantage of the logistic
function is that the hyperbolic tangent can be used for calculations, whereas
the Gaussian cumulative distribution can only be expressed in integral form.
With the logistic function, the probability for a random sample to be
within the range μ ± σ is 72 %, i.e. somewhat higher than with the Gaussian
distribution (68 %). The probability to be within an extended range of
μ ± 2σ is 95 % with both distributions (Table 14.5).
The proportion of the distribution in the extreme ranges is called
the tails of the distribution. The proportion outside ±3 standard devia-
tions is 0.9 % in the logistic distribution, but only 0.3 % in the Gaussian

Table 14.5  C
 omparison of the logistic and the Gaussian distributions
Range ±1 σ ±2 σ ±3 σ
Logit (%) 72.0 94.8 99.1
Gaussian (%) 68.3 95.4 99.7
Note: Probability of being within a range of ±1, ±2 or ±3 standard deviations.
 Noise effects and subjective evaluation of sound insulation  407

distribution (Table 14.5). Thus, the logistic distribution has heavier tails
than the Gaussian distribution, and this makes it more robust to inaccura-
cies in the model or to errors in the data. This is another advantage of the
logit model compared to the probit model.

14.3.2 Traffic noise annoyance

For road traffic noise, it is well established that an outdoor noise level Lp,A,24 h
below 55 dB in a housing area means that ~ 15 % to ~ 20 % of the people
living in the houses are annoyed by the noise. An early investigation gave the
results shown in Figure 14.8. The investigation is described below as a good
example of the challenges we are facing in this kind of social surveys. The fact
that it was made as early as 1972 has the unique advantage that the sound
insulation of windows and façades can be assumed the same in all the inves-
tigated houses. The use of special sound insulating windows in houses with
high noise exposure started shortly after 1972. The same kind of investigation
is not possible today because houses exposed to high levels of external noise
have windows with better sound insulation than houses in quiet areas.
Twenty-eight housing areas in Copenhagen were selected in pairs, so half
of them were clearly exposed to road traffic noise and the other half were
in more quiet areas. The pairing also considered the age, quality, and price
100

90
Percentage annoyed (>2 complaints of 18)

80

70

60

50

40

30

20

10

0
45 50 55 60 65 70 75
Lp,A,24 h (dB)

Figure 14.8 Percentage of people annoyed by noise in their dwellings as function

A-weighted sound pressure level of outdoor noise from road traffic (free
field values). Results from 28 locations in Copenhagen 1972. Statistical anal-
ysis (n = 960), fraction of explained variance 0.91. (Kragh, J., Analysis of the
Correlation between Reactions to Road Traffic Noise and Physical Indices of the
Noise. Report No. 5 (In Danish). The Acoustical Laboratory, The Danish
Academy of Technical Sciences, Lyngby, Denmark, 1977.)
408  Sound Insulation in Buildings

level of the houses. It was thought to be very important for the later statisti-
cal analysis that there was a sufficient spread in the noise exposure, and that
there were none in the middle range. Noise measurements were made with
24 h integration time and the microphone on a building façade. Free field
Lp,A,24 h results were derived by subtraction of 5 dB due to the microphone
position. This is the noise parameter shown on the abscissa in Figure 14.8.
In total, 960 personal interviews were made among the dwellers in August
and September 1972. The purpose was to find out how traffic noise might
influence the psychical and physical health of the inhabitants, but this purpose
was hidden. The investigation contained in total 75 questions, out of which 18
were related to annoyance due to road traffic noise, e.g. if it was difficult to fall
asleep in the evening, or if they did not open the windows because of the noise.
In a preliminary analysis of the results, it was found that a good indicator of
noise annoyance could be the number of complaints among the 18 relevant
questions and the median of all questionnaires was two complaints out of 18.
Thus, the annoyance parameter on the ordinate shown in Figure 14.8 is the
percentage in each housing area that gave more than two complaints.
The annoyance data were averaged for each housing area (on average 34
interviewed persons in each area). A regression analysis was made using
the results from the 28 housing areas, as plotted in Figure 14.8. A sigmoid
function was assumed, and the arcsine function was used as an approxima-
tion. The curve shows, for example, that if the free field Lp,A,24 h increases
from 55 dB to 65 dB, the percentage annoyed increases statistically from
14 % to 55 %. In the middle range between 20 % and 80 % the slope of the
dose-response curve is ~4 % points per dB.
The dose-response relationships can also be derived from laboratory
experiments. The advantages are that the parameters can be chosen to cover
a wide range of exposure, which is important for a good statistical analysis.
The connection between exposure and respondent can be controlled, and
there is little risk of confounders that could lead to errors. The drawbacks
of laboratory experiments include lack of realism, only short-time expo-
sure is possible, and the number of test persons is normally more limited
than in a field investigation. Figure 14.9 is an example of a dose-response
relationship found in a laboratory experiment, where a recording of road
traffic noise was used at seven different levels in 5 dB steps from 40 dB to
75 dB. The test persons were (among many other things) asked whether the
noise would be acceptable if they worked in an office with that noise. Probit
analysis was used to estimate the dose-response curve. The results show
that in an office, 10 % would be dissatisfied if the noise level exceeds 45 dB.
The average slope of the curve in the middle range is 4 % points per dB.

14.3.3 Socio-acoustic surveys
In a socio-acoustic survey, the questions asked about noise problems are very
important for the quality of the results. The way in which such questions are
 Noise effects and subjective evaluation of sound insulation  409

100

80
Percentage dissatisfied

60

40

20

0
35 40 45 50 55 60 65 70 75 80
Noise level, Lp,A,T (dB )

Figure 14.9 Percentage dissatisfied as a function of function A-weighted sound pressure

level of road traffic noise. Laboratory experiment on indoor climate in an
office (n  = 112). The curve is the result of probit analysis. (Adapted from
Clausen, G. et al. Indoor Air, 3, 255–262, 1993.)

asked and the wording of the possible answers should follow the recommenda-
tions in Technical Specification ISO/TS 15666. Questions and possible answers
about how much the person is bothered/disturbed/annoyed can be like this:
“Thinking about the last 12 months, when you are at home, how much
does noise from traffic annoy you? Not at all – Slightly – Moderately –
Very – Extremely?” A numerical rating scale may also be used, e.g. with
numbers from 0 to 10, where 0 means not at all and 10 means extremely.
In addition, there must be a possibility for answering “do not know”.
Researchers in this field do not agree on using a numerical scale. However,
ISO/TS 15666 is considered the best practice agreed on. The standard also
contains an Annex B with recommended wordings of the questions in sev-
eral languages. The words used for the answers have been balanced in such
a way that the rating between the words is about 25 %.
An example is shown in Figure 14.10, which is one of the results from
a socio-acoustic survey (Amundsen et al., 2011). From the replies of 738
people in combination with known traffic noise exposure, the four curves
in Figure 14.10 were derived by logit analysis. Five regions are indicated cor-
responding to the distribution of the replies from the interviews. For example
at a noise level of 75 dB, 10 % were not annoyed, 20 % slightly annoyed, 30 %
moderately annoyed, 31 % very annoyed, and 9 % extremely annoyed (Table
14.6). The often used outdoor noise limit Lp,A,24 h = 55 dB corresponds approxi-
mately to 60 % not annoyed and 15 % annoyed (moderately + very + extremely).
Each of the four curves have the same shape, since a logistic distribution was
410  Sound Insulation in Buildings

100
1
90

80
2
Cumulative proportion (%)

70

60
3
50

40

30
4
20

10
5
0
40 45 50 55 60 65 70 75 80
Lp,A,24 hOutdoor (dB)

Figure 14.10 Cumulative annoyance curves for road traffic noise in a dwelling as a func-
tion of the outdoor A-weighted sound pressure level. The curves are the
result of logit analysis (n  = 738). Region 1: Not annoyed. Region 2: Slightly
annoyed. Region 3: Moderately annoyed. Region 4: Very annoyed. Region 5:
Extremely annoyed.

Table 14.6  A
 nnoyance due to road traffic noise heard in the dwelling at selected
levels of the outdoor noise
Outdoor sound pressure level Lp,A, 24 h
Degree of annoyance 45 dB (%) 55 dB (%) 65 dB (%) 75 dB (%)
1 Not annoyed 83 59 29 10
2 Slightly annoyed 12 56 32 20
3 Moderately annoyed 3 10 23 30
4 Very annoyed 1 4 13 31
5 Extremely annoyed 0 1 3 9
Source: Amundsen, A.H. et al., J. Acoust. Soc. Am., 129, 1381–1389, 2011.
Note: Data from logit analysis in Figure 14.10.

assumed, and the curves are shifted along the dB-axis by approximately 10 %,
10 % and 15 %, respectively. The slope of the curves in the middle range is
~3 % points per dB.
In the same investigation, the indoor noise level due to road traffic noise
was estimated and the result of the analysis is seen in Figure 14.11. The
often used indoor limit of Lp,A,24 h = 30 dB corresponds approximately to
40 % not annoyed and 30 % annoyed (moderately + very + extremely).
 Noise effects and subjective evaluation of sound insulation  411

100

90

80

70
Cumulative proportion (%)

60

50

40

30

20

10

0
15 20 25 30 35 40 45 50
Lp,A,24 h Indoor (dB)
Slightly annoyed Moderately annoyed
Very annoyed Extremely annoyed

Figure 14.11 Cumulative annoyance curves for road traffic noise in a dwelling as a func-
tion of the indoor A-weighted sound pressure level. The curves are the
result of logit analysis (n  = 738). (Adapted from Amundsen, A.H. et al., J.
Acoust. Soc. Am., 129, 1381–1389, 2011.)

14.3.4 Different traffic noise sources

On the basis of a large amount of data collected from various socio-­acoustic
surveys, an analysis based on data collected by Miedema (2004) is pre-
sented in ISO 1996-1 (2016), Annex H. In contrast to the method used in
logit and probit analyses, these data have been analysed by using a model
where annoyance is assumed to be ­proportional to loudness times dura-
tion. Annoyance is known to be proportional to sound pressure squared
raised to the power of 0.3. Using the day-night level L dn as input, the gen-
eral equation for percentage highly annoyed becomes:

(
pHA = 100 ⋅ exp  − 100.1 (Ldn −Lct +5.3 dB) )  , % (14.7)
−0,3

 

where the community tolerance level L ct is defined as the day-night sound

level at which 50 % of the people are predicted to be highly annoyed by
noise exposure (Schomer et al., 2012; ISO 1996-1, 2016). The mean value
412  Sound Insulation in Buildings

of the community t­olerance level for different traffic noise sources is given
in Table 14.7, together with the standard deviations. The 95 % prediction
intervals can be estimated as ±2 standard deviations. The dose-response
curves for aircraft, road traffic, and railroad are shown in Figure 14.12.
Because of the different character of the noise sources (especially different
variation with time), the annoyance score is quite different for the same
day-night level. Aircraft is more annoying than road traffic (correspond-
ing to +5 dB), while railroad is less annoying (corresponding to −9.5 dB).
However, in cases where noise from railroad is accompanied by strong

Table 14.7  Community tolerance level L ct (mean and standard deviation) for
different traffic noise sources
Railroad (low Railroad (high
Aircraft Road traffic vibration) vibration)
Mean Lct, dB 73.3 78.3 87.8 75.8
Standard deviation, dB 7.1 5.1 3.5 4.2
Difference from road 5.0 0.0 −9.5 2.5
traffic, dB
Source: ISO 1996-1, Acoustics – Description, measurement and assessment of environmental
noise – Part 1: Basic quantities and assessment procedures. International Organization
for Standardization, Geneva, Switzerland, 2016.

60

50
Percentage highly annoyed

40

30

20

10

0
45 50 55 60 65 70 75
Ldn (dB)
Aircraft Road traffic Railroad (low vibration)

Figure 14.12 Dose-response curves for aircraft, road traffic and railroad (low vibration).
(Adapted from ISO 1996-1, Acoustics – Description, measurement and assess-
ment of environmental noise – Part 1: Basic quantities and assessment procedures.
International Organization for Standardization, Geneva, Switzerland, 2016.)
 Noise effects and subjective evaluation of sound insulation  413

vibrations, the annoyance score is much higher, between that for aircraft
and road traffic (not shown in Figure 14.12).
It is interesting to compare the curve for road traffic in Figure 14.12 with the
previous Figures 14.8 and 14.10. First, we recall that for a typical day-night dis-
tribution of traffic, the outdoor Lp,A,24 h = 55 dB corresponds to Ldn = 58 dB (Table
14.1). Figure 14.12 then shows 6 % highly annoyed from road traffic. In compar-
ison, Figure 14.8 shows 15 % annoyed. Figure 14.10 also shows 15 % annoyed
(moderately + very + extremely) but only 5 % highly annoyed (very + extremely);
the latter being close to the 6 % highly annoyed in Figure 14.12. This demon-
strates clearly the importance of how we define ‘annoyed’ or ‘highly annoyed’.
In future socio-acoustic i­nvestigations, this problem may be avoided if
the method in ISO/TS 15666 for such surveys is applied.

14.3.5 Airborne sound insulation

In the previously mentioned Norwegian socio-acoustic investigation on
noise annoyance in dwellings (Løvstad et al., 2016) measured sound insu-
lation data were collected from the actual addresses where the interviewed
persons live. Airborne sound insulation was measured between dwellings
in horizontal and vertical directions. The questionnaire included two ques-
tions related to airborne sound insulation: one about annoyance due to
speech, TV, etc., and another about annoyance due to music with bass
and drums. Table 14.8 gives an overview of the results. Both normalized
(R′w) and standardized (DnT,w) results were calculated without and with the
50 Hz spectrum adaptation term (indicated by index 50).

Table 14.8  Data from measured airborne sound insulation and degree of
annoyance due to speech and music
Horizontal Vertical
R′w R′w,50 DnT,w DnT,w,50 R′w R′w,50 DnT,w DnT,w,50
Measurements
Number, N 355 346 296 296 394 354 366 349
Min, dB 46 45 45 44 50 50 52 51
Max, dB 64 63 65 64 69 68 70 69
Mean, dB 56.8 53.8 58.6 55.4 61.6 58.7 61.0 57.8
Standard dev., dB 3.0 2.8 4.0 3.0 4.3 2.6 4.2 2.4
Annoyed due to 19.9 20.1 20.0 19.9 22.7 22.0 21.9 21.8
speech etc., (%)
Annoyed due to 27.8 26.4 25.4 24.8 27.3 25.4 25.4 26.5
music with bass
(%)
Source: Rindel, J.H. et al., Dose-response curves for satisfactory sound insulation between
dwellings, Proceedings of ICBEN 2017, Zürich, Switzerland, 2017.
Note: Percentage annoyed includes all levels (extremely, very, moderately, slightly).
414  Sound Insulation in Buildings

The upper part of Table 14.8 shows the number of measurements, the
range of variation, the mean value, and the standard deviation. It is noted
that airborne sound insulation is generally better in the vertical direction
than in the horizontal direction; mean values of R′w and R′w,50 are about
5 dB better for floors than for walls. This suggests that annoyance from
speech or music should be related mostly to the horizontal sound insula-
tion. If this is correct, it means that the dose-response results for airborne
sound insulation in the vertical direction cannot be reliable. In the subjec-
tive response from the questionnaire, there is no distinction between hori-
zontal or vertical direction of the annoying sound.
The degree of annoyance from speech is less than that from music. It is
also noted that the annoyance from speech seems to be higher in the vertical
direction than in the horizontal direction. However, this is simply a conse-
quence of the fact, that the average sound insulation of the floors is about
5 dB better than that of the walls. The degree of annoyance (all levels) corre-
sponding to the mean value of sound insulation is on average 26 % for music
and 21 % for speech. The cumulative annoyance curves were quite similar
for speech and music, and only the latter are shown here in Figure 14.13.

100

90

80
Cumulative proportion (%)

70

60

50

40

30

20

10

0
45 50 55 60 65
R΄w Horizontal (dB)
Slightly annoyed Moderately annoyed
Very annoyed Extremely annoyed

Figure 14.13 Cumulative annoyance curves due to music with bass as a function of the
normalised level difference between dwellings in the horizontal direction.
The curves are the result of logit analysis (n  = 702). (Adapted from Rindel,
J.H. et al., Aiming at satisfactory sound conditions in dwellings – The use of
dose-response curves. Proceedings of BNAM 2016, Stockholm, Sweden, 2016.)
 Noise effects and subjective evaluation of sound insulation  415

14.3.6 Impact sound insulation

The above-mentioned investigation also included impact sound. One
­question was about annoyance from the neighbour above due to walk-
ing, running, jumping, moving furniture, etc. Correlation was found with
measured impact sound pressure level, either normalised or standardized,
but only when the 50 Hz adaptation term was included. Impact sound
omitting the frequencies below 100 Hz had no correlation with the subjec-
tive response. The result for normalised impact sound is shown in Figure
14.14. The range of variation of L′n,w + C I,50–2500 was 44 dB to 61 dB, and
mean ± standard deviation was 53.7 dB ± 3.0 dB. This implies that 95 %
of the data can be assumed within a range of 12 dB. The results for the
­standardised impact sound L′nT,w + C I,50–2500 is not shown here; the correla-
tion with subjective data was equally good, but the mean ± standard devia-
tion of the data was 50.1 dB ± 2.0 dB, i.e. a very narrow range of variation;
the 95 % range being only 8 dB.

100

90

80
Cumulative proportion (%)

70

60

50

40

30

20

10

0
40 45 50 55 60
L΄n,w + CI,50–2500 (dB)
Slightly annoyed Moderately annoyed
Very annoyed Extremely annoyed

Figure 14.14 Cumulative annoyance curves due to footfall noise from above as func-
tion of the normalized impact sound pressure level including the 50 Hz
adaptation term. The curves are the result of logit analysis (n  = 702).
­
(Adapted from Rindel, J.H. et al., Aiming at satisfactory sound conditions
in dwellings – The use of dose-response curves. Proceedings of BNAM 2016,
Stockholm, Sweden, 2016.)
416  Sound Insulation in Buildings

14.3.7 Vibrations
A survey on vibrations in dwellings due to traffic (road and rail) was carried
out in Norway (Klæboe et al., 2003; Turunen-Rise et al., 2003). About 1500
people were interviewed and one question was: “Can you in your dwelling
notice shaking or vibration caused by <source>? If yes, is the shaking/these
vibrations highly annoying, somewhat annoying, a little annoying or not
annoying for you?”
Measurements of vibrations were performed in the dwellings and out-
side on the ground. A semi-empirical model was used to determine the
statistical maximum RMS vibration velocity using the Wm weighting filter
(ISO 2631-2) (see Figure 2.18). The statistical maximum is defined as the
95 percentile, i.e. the value not exceeded in 95 % of the time. The results
of the logit analysis using the velocity level 20 lg (v/vref) dB are shown in
Figure 14.15.
The curves show that about 7 % will be highly annoyed at a weighted vibra-
tion velocity of 0.3 mm/s. At a velocity of 0.5 mm/s, 10 % will be  highly
annoyed and about 40 % will annoyed slightly + moderately + highly.

0.1 0.32 1.0 3.2 mm/s

100

90

80
Cumulative proportion (%)

70

60

50

40

30

20

10

0
90 100 110 120 130 140
Velocity level, Lv,w,95, dB re 1 nm/s
Perceives vibrations Slightly annoyed
Moderately annoyed Highly annoyed

Figure 14.15 Cumulative annoyance curves for vibrations in the dwelling due to road and
rail traffic. The curves are the result of logit analysis (n  = 1427). (Adapted
from Turunen-Rise, I.H. et al., Appl. Acoust., 64, 71–87, 2003.)
 Noise effects and subjective evaluation of sound insulation  417

14.4 LOW-FREQUENCY NOISE FROM NEIGHBORS

Problems with insufficient low-frequency sound insulation and music played

on powerful hi-fi equipment were described by Craik and Stirling (1986). The
background was a very strong increase in the number of complaints about
music from neighbours and, in particular, from discotheques. About 25 % of
all complaints about noise were concerning amplified music. Measurements
and investigations in 40 selected dwellings showed that the degree of annoy-
ance was related to the signal-to-noise level, and even when the music was
below the background noise, the dwellers could hear the music clearly. Still,
at a level 5 dB below the background noise, 20 % were bothered by the music.
An investigation on sound in dwellings was performed by Adnadevic
et al. (2011). Measurements showed that the highest sound levels were pro-
duced by music, singing, playing a musical instrument, or music repro-
duction using modern audio equipment. The Lp,A,T of these sources was
between 80 dB and 90 dB. The measured maximum values were around
88 dB to 92 dB for listening to music, and around 91 dB to 98 dB for watch-
ing opera on TV. Loud music listening was found to be the only source with
significant energy below 100 Hz and the measured spectra in the source
rooms were close to spectrum 2 of ISO 717-1 (the traffic noise spectrum).
The importance of music as a sound source in residential buildings was
studied by Lang and Muellner (2013). They made a questionnaire investiga-
tion in Austria, Germany and Switzerland, which showed that music, espe-
cially pop and rock, is an important sound source in residential buildings. It
was found that high volume of the music is highly preferred, and 85 % uses
bass loudspeaker or subwoofer to enforce low frequencies. The preferred
type of music has a nearly flat frequency response down to 50 Hz and, thus,
the authors find it reasonable that the frequency range to be observed for
sound insulation has to include the 50 Hz to 80 Hz one-third octave bands.
A laboratory experiment has been conducted with the purpose of sys-
tematically investigating the influence of the low-frequency content in noise
from neighbours (Mortensen, 1999). The experiment was carried out in
the listening room at the Technical University of Denmark. Three sound
signals were used: music from a neighbouring room, footfall noise from a
male walker in the room above and footfall noise from two children run-
ning and playing catch in the room above.
The frequency spectrum of each of the three sound signals was modi-
fied in order to simulate five different types of building constructions. For
airborne sound, the slope of the spectrum between 50 Hz and 160 Hz was
varied in order to simulate the sound transmission through different con-
structions ranging from light to heavy. Above 160 Hz, the spectrum was
kept constant. For impact noise, a similar procedure was used; only the
slope of the spectrum between 50 Hz and 125 Hz was varied. The sound
examples were presented to 25 test persons through loudspeakers. Some of
the main results are shown in Tables 14.9 and 14.10.
418  Sound Insulation in Buildings

Table 14.9  Results of laboratory experiment with airborne sound and simulated
constructions with different sound insulation below 160 Hz
Music, percent
Type of construction R′w (dB) R′w + C50–3150 (dB) Lp,A,T (dB) annoyed
Light 56 49 48 98
Light-medium 56 53 43 90
Medium 57 55 37 80
Medium-heavy 57 56 36 83
Heavy 57 56 35 83
Source: Mortensen, F.R., Subjective Evaluation of Noise from Neighbours with Focus on Low
Frequencies. Main report. Publication no. 53, Department of Acoustic Technology,
Technical University of Denmark, 1999.

Table 14.10  R
 esults of laboratory experiment with impact sound and simulated
constructions with different sound insulation below 125 Hz
Walker, Children,
Type of L′n,w L′n,w + CI,50– Walker, Children, percent percent
construction (dB) 2500 (dB) Lp,A,T (dB) Lp,A,T (dB) annoyed annoyed
Light 55 62 39 41 71 81
Light-medium 55 58 32 35 51 78
Medium 55 56 27 30 36 47
Medium-heavy 55 55 25 28 28 51
Heavy 54 54 24 25 20 47
Source: Mortensen, F.R., Subjective Evaluation of Noise from Neighbours with Focus on Low
Frequencies. Main report. Publication no. 53, Department of Acoustic Technology,
Technical University of Denmark, 1999.

From the results of this investigation, it was concluded that the spectrum
adaptation terms in ISO 717-1 and ISO 717-2 with extended frequency
range down to 50 Hz give a better correlation between subjective and objec-
tive evaluation of sound insulation.
While the importance of the extended frequency range down to 50 Hz
is clear in relation to impact sound, and other researchers have come to a
similar conclusion (Warnock, 1998; Späh et al., 2014), the case of airborne
sound insulation is not so clear. Only when the sound is music with strong
contents of low frequencies, the 50 Hz adaptation term gives an improved
correlation with subjective impressions; but for other kinds of noise from
neighbours like speech, TV, etc., there is no need to include the frequencies
below 100 Hz for airborne sound insulation (Kylliäinen et al., 2016).
The spectrum and the frequency range are very different in music and
speech. A comprehensive investigation based on listening tests using
speech and music in combination with 20 partition walls with quite dif-
ferent acoustic characteristic was performed (Park et al., 2008; Park and
 Noise effects and subjective evaluation of sound insulation  419

Bradley, 2009a,b). Three music samples were selected for the tests, all of
them being potentially annoying but of different musical styles. It was
found that the optimum range of included frequencies in the single-number
quantity is different for music sounds and speech sounds. Some measures
were strongly related to ratings of speech sounds and others to ratings of
music sounds, but none were highly successful for both types of sounds.
The best correlation with annoyance from music sounds was a parameter
defined as the A-weighted sound pressure level difference with pink noise,
63 Hz to 6300 Hz (R 2 = 0.978). For speech sounds, the best correlation
with annoyance was a parameter defined as the A-weighted sound pres-
sure level difference with pink noise, 200 Hz to 6300 Hz (R 2 = 0.977). It
was also found that the signal-to-noise ratio was well related to annoy-
ance ratings and audibility ratings of transmitted music sounds. Almost
all subjects found that sounds were just audible at signal-to-noise levels as
low as −10 dB. The thump of the rhythmic beating, which is typical in loud
music, may explain this finding.
From the point of view that sound insulation requirements are meant to
ensure a reasonable degree of protection against all typical sounds from
neighbours, the one-third octave bands 50 Hz to 80 Hz are important and
should be included in the evaluation of airborne sound insulation in dwell-
ings (Rindel, 2017).

14.5 DESIGN CRITERIA FOR SATISFACTORY

SOUND INSULATION

In many countries, the building regulations for sound insulation were

established a long time ago (1940–1960) and mainly from a consideration
of the traditional building practice at that time, rather than from an analy-
sis of the needs for sound insulation. Although the purpose of the building
regulations was to avoid new buildings with clearly unacceptable acousti-
cal conditions, the result was very often that these minimum requirements
were considered a satisfactory goal for the acoustic design. After decades
with a very stable situation in building acoustics, changes in this field in
many countries started in the 1990s, and the development seems to go in a
direction where acoustic conditions are taken more seriously as a relevant
and important design criterion for modern buildings. So, the hope for the
future is that the building technology will change in such a way that houses
can offer acoustical comfort to the people they are built for.
In the national building codes, it is often stated as a basic requirement
for new dwellings that the acoustical conditions shall be satisfactory. So,
the big question is how to define ‘satisfactory’. In recent years, the method
of socio-acoustic investigations has matured, the methodology has been
standardised (ISO/TS 15666, 2003; ISO 1996-1, 2016), and results from
several new investigations have been published. We know that the degree of
420  Sound Insulation in Buildings

annoyance or satisfaction is described by a sigmoid function, which means

that we can never expect 100 % satisfaction. So, it is merely a political deci-
sion as to what is satisfactory. Is it enough if 50 % are not annoyed, or should
we aim at 70 % or 80 % being satisfied (not annoyed)? Or should we rather
look at the percentage of those who are highly annoyed/extremely annoyed?
Since the survey paper by Schultz (1978), it has been common practice to
use the fraction “highly annoyed” for predicting the community annoyance
to noise, and this is also used in ISO 1996-1 (2016). However, the dose-
response curves, of which we have seen several examples in this chapter,
have a very flat slope with values close to 0 % if we look at the lowest curve
that gives the percentage indicating extremely annoyed persons. The next
curve for very + extremely annoyed also has a flat slope. This means that it is
very uncertain which dB-value actually corresponds to a certain percentage
highly annoyed, e.g. 3 %, 5 % or 7 %. The result will be much more pre-
cise if we look at the upper curve, which normally has the steepest slope in
the range of interest; this is the curve representing all degrees of annoyance
(slightly, moderately, very, and extremely). In addition, the proportion above
this curve represents the percentage who is satisfied (not annoyed). This is
suggested as a better and more precise way to read the dose-response curves.
Table 14.11 presents relations between the expected percentage satisfied/
annoyed and acoustic criteria for sound insulation and noise. In order to
create this table, the dose-response results from various investigations have
been slightly modified in such a way that the slope of the cumulative logis-
tic distribution is 4 % points per dB in the middle range, as suggested by
Rindel (1999) and Rindel et al. (2016).

Table 14.11  Acoustic design criteria for dwellings and the corresponding
expected percentage of people finding conditions satisfactory (not annoyed) or
not satisfactory (annoyed to some degree)
Satisfied
90 80 70 60 50 40 30 20
(not annoyed), %
Annoyed
(moderately + 
3 5–8 9–12 15–18 20–25 27–35 37–45 50–55
very + 
extremely), %
Airborne R′w, dB 64 59 56 53.5 51 48.5 46 43
sound R′w,50, dB 62 57 54 51.5 49 46.5 44 41
insulation
DnT,w, dB 66 61 58 55.5 53 50.5 48 45
DnT,w,50, dB 64 59 56 53.5 51 48.5 46 43
Impact sound L′n,w,50, dB 42.5 47.5 50.5 53 55.5 58 60.5 63.5
level L′nT,w,50, dB 39 44 47 49.5 52 54.5 57 60
Noise indoor, Lp,A,24 h, dB 16 21 24 26.5 29 31.5 34 37
road traffic
Noise outdoor, Lp,A,24 h, dB 47 52 55 57.5 60 62.5 65 68
road traffic Lden, dB 50 55 58 60.5 63 65.5 68 71
 Noise effects and subjective evaluation of sound insulation  421

The following abbreviations are used in Table 14.11: R′w,50 = R′w + C 50–3150,

DnT,w,50 = DnT,w + C 50–3150, L′n,w,50 = L′n,w + C I,50–2500.
The current minimum requirement in the Nordic countries (Table 5.2)
is close to the column with 60 % satisfied. This level of acoustic qual-
ity may be labelled ‘acceptable’, while we should aim somewhat higher
for ‘­satisfactory’ conditions. The values corresponding to 80 % satisfied
(and  5  % to 8 % annoyed) may be candidates for ‘satisfactory’ acousti-
cal conditions and thus suggested as a target for future development of
better houses.

REFERENCES

M. Adnadevic, M. Mijic, D. Sumarac-Pavlovic, D. Masovic (2011). Noise in dwellings

generated in normal home activities – general approach. Proceedings of Forum
Acusticum 2011, 27 June – 1 July 2011, Aalborg, Denmark, 1335–1340.
D.G. Altman (1991). Practical Statistics for Medical Research. Chapman and Hall,
London.
A.H. Amundsen, R. Klæboe, G.M. Aasvang (2011). The Norwegian Façade Insulation
Study: The efficacy of façade insulation in reducing noise annoyance due to
road traffic. Journal of the Acoustical Society of America 129, 1381–1389.
J.S. Bradley (1983). Subjective rating of party walls. Canadian Acoustics 11(4),
37–45.
G. Clausen, L. Carrick, P.O. Fanger, S.W. Kim, T. Poulsen, J.H. Rindel (1993). A
comparative study of discomfort caused by indoor air pollution, thermal load
and noise. Indoor Air, 3, 255–262.
R.J.M. Craik and J.R. Stirling (1986). Amplified music as a noise nuisance. Applied
Acoustics 19, 335–356.
D.E. Cummins (1978). Classes of acoustical comfort in housing. Proceedings of
Inter-Noise 78, 8–10 May 1978, San Francisco, USA, 631–636.
C.J. Grimwood (1993). Effects of environmental noise on people at home. BRE
information paper IP 22/93.
C.J. Grimwood (1997). Complaints about poor sound insulation between dwellings
in England and Wales. Applied Acoustics 52, 211–223.
J.I. Halonen, A.L. Hansell, J. Gulliver, D. Morley, M. Blandiardo, D. Fecht, M.B.
Toledano, S.D. Beevers, H.R. Anderson, F.J. Kelly, C. Tonne (2015). Road traffic
noise is associated with increased cardiovascular morbidity and mortality and
all-cause mortality in London. European Heart Journal 36(39), 2653–2661.
D. Hjortebjerg, A.M. Nybo Andersen, J. Schultz Christensen, M. Ketzel, O. Raaschou-
Nielsen, J. Sunyer, J. Julvez, J. Forns, M. Sørensen (2016). Exposure to road
traffic noise and behavioral problems in 7-year-old children: A cohort study.
Environmental Health Perspectives 124(2), 228–234.
ISO/TS 15666 (2003). Acoustics – Assessment of noise annoyance by means of
social and socio-acoustic surveys. Technical specification. Organization for
Standardization, Geneva, Switzerland.
ISO 1996-1 (2016). Acoustics – Description, measurement and assessment of
environmental noise – Part 1: Basic quantities and assessment procedures.
International Organization for Standardization, Geneva, Switzerland.
422  Sound Insulation in Buildings

ISO 717 (1996). Acoustics – Rating of sound insulation in buildings and of building
elements. Part 1: Airborne sound insulation. Part 2: Impact sound insulation.
International Organization for Standardization, Geneva, Switzerland.
R. Klæboe, I.H. Turunen-Rise, L. Hårvik, C. Madshus (2003). Vibration in dwell-
ings from road and rail traffic – Part II: exposure-effect relationships based on
ordinal logit and logistic regression models. Applied Acoustics 64, 89–109.
J. Kragh (1977). Analysis of the Correlation between Reactions to Road Traffic Noise
and Physical Indices of the Noise. Report No. 5 (In Danish). The Acoustical
Laboratory, The Danish Academy of Technical Sciences, Lyngby, Denmark.
M. Kylliäinen, J. Takala, D. Oliva, V. Hongisto (2016). Justification of standard-
ized level differences in rating of airborne sound insulation between dwellings.
Applied Acoustics 102, 12–18.
J. Lang and H. Muellner (2013). The importance of music as sound source in resi-
dential buildings. Proceedings of Inter Noise 2013, 15–18 September 2013,
Innsbruck, Austria.
A. Løvstad, J.H. Rindel, C.O. Høsøien, I. Milford, R. Klæboe (2016). Sound qual-
ity in dwellings in Norway – A socio-acoustic investigation. Proceedings of
BNAM 2016, 20–22 June 2016, Stockholm, Sweden.
C. Maschke and H. Niemann (2007). Health effects of annoyance induced by neigh-
bour noise. Noise Control Engineering Journal 55(3), 348–356.
H.M.E. Miedema (2004). Relationship between exposure to multiple noise sources
and noise annoyance. Journal of the Acoustical Society of America 116,
949–957.
F.R. Mortensen (1999). Subjective Evaluation of Noise from Neighbours with Focus
on Low Frequencies. Main report. Publication no. 53, Department of Acoustic
Technology, Technical University of Denmark.
H. Niemann, X. Bonnefoy, M. Braubach, K. Hecht, C. Maschke, C. Rodrigues, N.
Röbbel (2006). Noise-induced annoyance and morbidity results from the pan-
European LARES study. Noise & Health 8(31), 63–79.
H.K. Park, J.S. Bradley, B.N. Gover (2008). Evaluating airborne sound insulation
in terms of speech intelligibility. Journal of the Acoustical Society of America
123, 1458–1471.
H.K. Park and J.S. Bradley (2009a). Evaluating standard airborne sound insulation
measures in terms of annoyance, loudness, and audibility ratings. Journal of
the Acoustical Society of America 126, 208–219.
H.K. Park and J.S. Bradley (2009b). Evaluating signal-to-noise ratios, loudness, and
related measures as indicators of airborne sound insulation. Journal of the
Acoustical Society of America 126, 1219–1230.
J.H. Rindel (1999). Acoustic quality and sound insulation between dwellings. Journal
of Building Acoustics 5, 291–301.
J.H. Rindel (2017). A comment on the importance of low frequency airborne sound
insulation between dwellings. Acta Acustica/Acustica 103, 164–168.
J.H. Rindel, A. Løvstad, R. Klæboe (2016). Aiming at satisfactory sound conditions
in dwellings – The use of dose-response curves. Proceedings of BNAM 2016,
Stockholm, Sweden.
J.H. Rindel, A. Løvstad, R. Klæboe (2017). Dose-response curves for satisfactory
sound insulation between dwellings. Proceedings of ICBEN 2017, 18–22 June
2017, Zürich, Switzerland.
 Noise effects and subjective evaluation of sound insulation  423

J.H. Rindel, B. Rasmussen (1995). Buildings for the future: The concept of acoustical
comfort and how to achieve satisfactory acoustical conditions with new build-
ings. (18 pages). COMETT-SAVOIR Course, CSTB, Grenoble.
P. Schomer, V. Mestre, S. Fidell, B. Berry, T. Gjestland, M. Vallet, T. Reid (2012). Pole
of community tolerance level (CTL) in predicting the prevalence of the annoy-
ance of road and rail noise. Journal of the Acoustical Society of America 131,
2772–2786.
T.J. Schultz (1978). Synthesis of social surveys on noise annoyance. Journal of the
Acoustical Society of America 64, 377–405.
M. Späh, K. Hagberg, O. Bartlomé, L. Weber, P. Leistner, A. Liebl (2014). Subjective
and objective evaluation of impact noise sources in wooden buildings. Noise
Notes 13, 25–42.
I.H. Turunen-Rise, A. Brekke, L. Hårvik, C. Madshus, R. Klæboe (2003). Vibration
in dwellings from road and rail traffic – Part I: A new Norwegian measurement
standard and classification system. Applied Acoustics 64, 71–87.
A.C.C. Warnock (1998). Floor research at NRC Canada. Proceedings of COST
Action E5 Workshop, Acoustic performance of medium-rise timber buildings,
3–4 December 1998, Dublin, (10 pages).
R. Weeber, H. Merkel, H. Rossbach-Lochmann, K. Gösele, E. Buchta (1986).
Schallschutz in Mehrfamilienhäusern aus der Sicht der Bewohner (Sound
insulation in multi-family houses from the point of view of the resident, in
German). F 2049, IRB Verlag, Stuttgart, Germany.
WHO (2007). Large Analysis and Review of European Housing and Health
Status (LARES). Preliminary overview. WHO Regional Office for Europe,
Copenhagen, Denmark.
S. Wibe (1997). Efterfrågan på tyst boende (The Demand for Silent Dwellings, in
Swedish). Anslagsrapport A4:1997. Byggforskningsrådet, Stockholm, Sweden.
Chapter 15

Experimental buildings with

high sound insulation

In the previous chapter, the question was which level of sound insulation
should be required to obtain satisfactory acoustical conditions for dwell-
ings in multi-unit houses. In this final chapter, we shall look at some exam-
ples of experimental buildings with particularly good sound insulation.
The examples are restricted to multi-storey buildings with dwellings. Of
course, the idea of making experimental buildings with very good sound
insulation comes from the ambition that dwellings for the future should
have satisfactory conditions, also for acoustics.
In some cases, the purpose has been to demonstrate in full scale that it
is possible to give priority to sound insulation in the building design and
to achieve practical experience with the suggested technical solution. It is
remarkable that the very high acoustic quality achieved in these projects
does not necessarily lead to higher construction costs, and one of the solu-
tions was actually developed with the purpose of saving construction time.

15.1 ‘THE QUIET HOUSE’ LUND, SWEDEN

15.1.1 General
The building was erected 1988–1989 in Kv. Bollen, Lund, Sweden. It has
three stories and 29 apartments. Builder: JM Byggnads och Fastighets
AB. Consultants: Sterner Akustik AB, Bentech Ingenörs AB. The project
was supported by The Swedish Council for Building Research and The
Development Fund of the Swedish Construction Industry.
Figure 15.1 shows a part of the façade of the experimental building.

15.1.2 Purpose of experiment
The project was initiated by Lars Holmgren, Cementa AB. The purpose
was to find out what could be achieved acoustically with a traditional
heavy-weight building construction. Well-known solutions and construc-
tion details were applied, but with an unusual focus on the acoustical needs.

425
426  Sound Insulation in Buildings

Figure 15.1 External view of a part of façade of the experimental building. (Adapted

from Byggforskningsrådet., Det tysta huset. Bätre ljudkomfort i nyproducerade
bostäder (The quiet house. Improved acoustical comfort in new-built
dwellings, in Swedish), Statens råd för byggforskning, T9:1991, Stockholm,
Sweden, 1991.)

The project report from Byggforskningsrådet (1991) contains a catalog of

technical details and acoustical solutions.
In the initial stage of the project, quite ambitious requirements for air-
borne and impact sound insulation were established. This was done by
definition of special reference curves in the extended frequency range from
50 Hz to 5000 Hz at a level about 10 dB better than the usual minimum
requirement (see Figure 15.2 and 15.3). The evaluation method of a mea-
surement result is the same as for the ISO reference curve; i.e. the sum of
­unfavourable deviations must not exceed 32 dB.
At the end of the project, the extra costs of future ‘quiet houses’ built as
this experimental building was estimated to be 2 % to 3 % compared to a
traditional building.

15.1.3 Building concept
The main constructions are 240 mm concrete walls and 290 mm concrete
slabs cast on site. The floors were either carpet or wood on cork. This is
type A in Figure 11.1. Flanking transmission is possible in all directions,
but of little importance due to the high weight of the constructions.
The project report mentions, as a drawback to the building concept, that
the time needed for the concrete to harden is longer than usual because of
the thickness of walls and slabs.
 Experimental buildings with high sound insulation  427

15.1.4 Acoustical results
An overview of the measurement results is presented in Table 15.1.
The following abbreviations are used in Table 15.1: R′w,50 = R′w + C 50–5000,
L′n,w,50 = L′n,w + C I,50–2500.
Examples of measured airborne sound insulation in the horizontal and
vertical directions are shown in Figure 15.2. In both directions, the trans-
mission loss is very high; at frequencies above 250 Hz, the transmission loss
is better in the vertical direction than in the horizontal direction.
The measured impact sound insulation in the vertical direction is
shown in Figure 15.3. The results are very good and the special project

Table 15.1  Single-number results of measured sound insulation in the ‘Quiet House’
Airborne, Airborne, Impact, vertical
Frequency range (Hz) horizontal (dB) vertical (dB) (dB)
100–3150 R′w = 62 R′w = 65 L′n,w = 43
50–5000 R′w,50 = 61 R′w,50 = 61 L′n,w,50 = 46
Note: Same as in Figures 15.2 and 15.3.

100

90

80
Transmission loss, R΄ (dB)

70

60

50

40

30

20
50 100 200 500 1000 2000 5000
Frequency (Hz)
Minimum requirement Project requirement Horizontal Vertical

Figure 15.2 Measured airborne sound insulation in the horizontal and vertical directions.
The full line is the project requirements in the extended frequency range. The
dashed line is the normal minimum requirement at the time of construction.
428  Sound Insulation in Buildings

80

70
Impact sound pressure level, L΄n (dB )

60

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
Minimum requirement Project requirement Measured vertical

Figure 15.3 Measured impact sound insulation in the vertical direction. The full line is
the project requirements in the extended frequency range. The dashed line
is the normal minimum requirement at the time of construction.

requirements are fulfilled (recalling that the reference curve may be passed
with up to 32 dB of unfavourable deviations).

15.2 ‘OPEN HOUSE URBAN BUILDING

SYSTEM’, DENMARK

15.2.1 General
The ‘OPEN HOUSE Urban Building System’ has been developed by
Dominia in cooperation with the architect firm Landskronagruppen,
Landskrona, Sweden.
The building concept has been used in several places in Denmark. The
experimental buildings described here were erected 1993 and are located at
Færgeparken III, Frederikssund, Denmark. There are three 3-storey build-
ings; in total, 36 dwellings. Architect: Mangor & Nagel, Frederikssund.
Builder: Arbejdernes Kooperative Byggeforening (a cooperative building
association). Consultant: Dominia, Copenhagen.
 Experimental buildings with high sound insulation  429

15.2.2 Purpose of experiment
The purpose was to gather experience with the newly developed building
system, which is a light-weight system with short building time. Improved
sound insulation was not a purpose of the experiment.

15.2.3 Building concept
The ‘OPEN HOUSE Urban Building System’ is a building system based
on factory-produced room modules and a site-produced concrete ­skeleton.
Each room module consists of a steel framework enclosing one or more
rooms. Walls, floors, and ceilings consist of gypsum boards mounted
on steel ribs. Compared to conventional concrete housing, the weight is
approximately 35 %, the building time at site is 25 % and the building time,
including production of modules, is 70 %. Different stages in the building
process are shown in Figure 15.4.
The construction principle is a column-beam skeleton with light-
weight walls and floors, type D in Figure 11.1. The concrete columns
are cast on site, while the concrete beams are prefabricated (Figures 15.5
and 15.6).

Figure 15.4 Schematic view of different stages in the building process. (From Dominia.
With permission from Dominia, Copenhagen, Denmark.)
430  Sound Insulation in Buildings

Prefab concrete beam

Concrete

Ste
el ske
leto
Corner n
profile

Figure 15.5 Detail of a junction. The steel profiles at the edges of the prefabricated
room modules create a form for in situ casting of the concrete columns. The
prefabricated concrete beams are fixed in the steel profiles. (With permis-
sion from Dominia, Copenhagen, Denmark.)

15.2.4 Acoustical results
An overview of the measurement results is presented in Table 15.2. The
measured sound insulation results are 8 dB to 10 dB better than the mini-
mum requirements in Denmark at the time of building. Measurements did
not include frequencies below 100 Hz.
The measured airborne sound insulation is shown in Figure 15.7.
Above 200 Hz, the transmission loss in the horizontal direction is higher
than that in the vertical direction, and the frequency course is typical
for a double wall, decreasing below 200 Hz towards a mass-spring-mass
resonance frequency that may be just below 100 Hz. Although no mea-
surement results are available below 100 Hz, it seems that the airborne
sound insulation at low frequencies may be an issue, both vertically and
horizontally.

Table 15.2  S ingle-number results of measured sound insulation in the ‘OPEN

HOUSE Urban Building System’
Airborne, horizontal (dB) Airborne, vertical (dB) Impact, vertical (dB)
R′w = 62 R′w = 61 L′n,w = 49
 Experimental buildings with high sound insulation  431

300

Acoustic sealing

Gypsum board

30 39 30
Sealing with

242
252
mineral wool 50 50
10 6 10

39

6 mm neoprene

9 mm gypsum 9 mm gypsum

Figure 15.6 Detail of a junction. (With permission from Dominia, Copenhagen, Denmark.)

The measured impact sound insulation is shown in Figure 15.8. The fre-
quency course suggests that there may be an issue with the impact sound
below 100 Hz. Otherwise, the results are quite satisfactory.

15.3 ‘ECO-HOUSE’, DENMARK

15.3.1 General
The ECO-House building system was developed in connection with
a competition from 1987 on the question: ‘How can Danish building
industry make the next great jump forward?’ Due to the falling price
on steel, it was considered advantageous to develop an alternative to the
432  Sound Insulation in Buildings

100

90

80
Transmission loss, R΄ (dB)

70

60

50

40

30

20
50 100 200 500 1000 2000 5000
Frequency (Hz)
Horizontal Vertical

Figure 15.7 Measured airborne sound insulation in horizontal and vertical directions in

the ‘OPEN HOUSE Urban Building System’.

concrete-based prefabricated building methods that had been dominat-

ing since the 1960s. The companies behind the new building system were
architect: Skaarup & Jespersen A/S, builder: KKS Enterprise A/S, and con-
sultant: Birch & Krogboe A/S (KKS Enterprise, 1992; Gath and Karstoft,
1992). The first versions of the building system were called ‘Scanhouse’.
The building concept has been used in several places in Denmark. The
experimental buildings described here were erected in 1992 and are located
in Egeskoven, Glostrup, Denmark. These are three-storey buildings, 49
apartments in total.

15.3.2 Purpose of experiment
The basic idea was to maximize prefabrication in order to minimize the
building time without compromising economy or quality. Acoustically,
there was no ambition to exceed the Danish minimum requirements for
sound insulation at the time of construction.

15.3.3 Building concept
The supporting structure is a framework of steel profiles. This is type D in
Figure 11.1. Floor constructions and walls are prefabricated elements of gyp-
sum board and mineral wool in thin-plate steel profiles. On the building site,
 Experimental buildings with high sound insulation  433

80

70
Impact sound pressure level, L΄n (dB)

60

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
Measured vertical

Figure 15.8 Measured impact sound insulation in the vertical direction in the

‘ECO-HOUSE’.

the steel framework is erected first. Then, the floor and façade elements and
the roof are mounted. Building materials needed for the internal parts are
brought into the building before the raw house is closed. The rest of the build-
ing process can take place in a dry environment protected from rain and wind.
The total building time is 70 % of that for more traditional building systems.
A construction detail is seen in Figure 15.9. The steel hat profile is a
structural beam that also provides support for the floor elements.

15.3.4 Acoustical results
An overview of the measurement results is presented in Table 15.3. The
average of measured sound insulation results is 5 dB to 10 dB better than the
minimum requirements in Denmark at the time of building. The measured

Table 15.3  S ingle-number results of measured sound insulation in the ‘ECO-HOUSE’

Airborne, horizontal (dB) Airborne, vertical (dB) Impact, vertical (dB)
R′w = 58–63 R′w = 56–61 L′n,w = 46–50
Source: Gath J., Karstoft, N.O., ECO-HOUSE. Et nyt dansk byggesystem. (ECO-HOUSE. A
new Danish building system, in Danish). Dansk ståldag 1992, Birch & Krogboe A/S.
(9 pages), 1992.
434  Sound Insulation in Buildings

300
150 150

13 13 248 13 13

2 mm × 13 mm 45 mm 2 mm × 13 mm gypsum
gypsum fixed
mineral
wool
Bottom profile
SKF 70/100 with felt 16 mm wood on felt
3 mm × 13 mm gypsum
20 mm trapez plate

Element
25 162 25

Neoprene

2 mm × 15 mm special gypsum
50 mm × 50 mm angle profile

Figure 15.9 Detail of the junction between floor and partition wall in the ‘ECO-HOUSE’.

impact sound insulation in the vertical direction is shown in Figure 15.10.

There is no available information about the airborne sound insulation in
the low frequency range from 50 Hz to 80 Hz.

15.4 ‘BRF KAJPLATSEN’, STOCKHOLM, SWEDEN

15.4.1 General
The building was erected in 1994 in BRF Kajplatsen, Kv Trålen, Stockholm,
Sweden. It has seven stories and 26 apartments. The experimental building
 Experimental buildings with high sound insulation  435

80

70
Impact sound pressure level, L΄n (dB)

60

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
Measured vertical

Figure 15.10 Measured impact sound insulation in vertical direction in the ‘ECO-

HOUSE’. The single-number values for this measurement are L­′n,w = 46 dB
and L′n,w + CI,50–2500 = 49 dB.

was part of a larger building project with a total of 176 apartments.

Architect: Åsberg & Buchmann Arkitekter AB. Builder: JM Byggnads och
Fastighets AB. Consultants: IIC, NCAB, Scandiakonsult Inst. AB, Plan
& Mark AB, J A Jonsson., Bröderna Skoogs Vent AB, Björn Lindström
Elektriska. Figure 15.11 shows the façade of the experimental building and
one of the other buildings.

15.4.2 Purpose of experiment
Ljunggren (1993, 1995) has shown theoretically that it should be possible
to increase the sound insulation in the vertical direction significantly by
using a column-plate structure with a large slab supported by load-bearing
columns instead of load-bearing walls (type B in Figure 11.1). The idea
was that the flanking transmission through the slab could be turned into
an advantage by distributing the energy of structural vibrations in a plate
with a large area. This implies that the walls should be light-weight double
constructions so that the junction attenuation would be negligible. The
principle is explained by Statistical Energy Analysis (SEA) in Section 7.6.
436  Sound Insulation in Buildings

1994

Figure 15.11 External view of two buildings, one of them being the experimental building
on ‘BRF Kajplatsen’.

The purpose of the experimental building was to prove that the area effect
could work in practice. The building costs are estimated not to exceed
those of a conventional building by more than 1 %.
The theory of this project was developed by Professor Sten Ljunggren
in research projects supported by The Swedish Council for Building
Research and The Development Fund of the Swedish Construction
Industry.

15.4.3 Building concept
Three similar buildings were erected, two of them with traditional load-
bearing walls of 180 mm concrete and 240 mm thick concrete slabs. The
experimental building has 260 mm thick concrete slabs supported by
­100  mm  × 100 mm steel columns in the façade and 400 mm × 400 mm con-
crete columns hidden in the walls inside the apartments (Figure 15.12). The
double walls used as partition walls between the apartments are shown in
Figure 15.13. All buildings have concrete walls around the stairwell and
elevator in the middle of the building. The experimental building is type B
in Figure 11.1, while the other buildings are type A.
The acoustical principle is illustrated in Figure 15.14. The total area of
the concrete slab is 300 m 2 . The area of the floor in one on the bedrooms
 Experimental buildings with high sound insulation  437

Figure 15.12 Plan of the experimental building on ‘BRF Kajplatsen’. The small black
squares are the columns, six from concrete inside the building and 25 from
steel in the façade. (Reproduced from Ljunggren, S., wksb, Zeitschrift für
Wärmeschutz, Kälteschutz, Schallschutz, Brandschutz, Heft 38, 1–4, 1996.
With permission.)

is 12.4 m 2 . Thus, if we refer to the theoretical result (Equation  7.42)

the area effect in this case might be up to 14 dB increased transmission
loss.

15.4.4 Acoustical results
An overview of the measurement results is presented in Table 15.4.
The measured airborne sound insulation between two rooms in the ver-
tical direction is shown in Figure 15.15, both for the experimental build-
ing and the traditional building. The difference is around 8 dB to 10 dB
in the entire frequency range. The thickness of the slab is not the same
in the experimental building and the traditional building, but this cannot
account for more than 1 dB difference. So, the efficiency of the principle in
the experimental b ­ uilding is clearly demonstrated. In the horizontal direc-
tion, the airborne sound ­insulation is also very high. At the low frequencies
50 Hz to 80 Hz, the average transmission loss is around 35 dB, which is
quite good for a light-weight double wall.
438  Sound Insulation in Buildings

Independent steel
studs

3 mm × 70 mm
mineral wool
3 mm × 13 mm
gypsum board

Sealing

Figure 15.13 Section of the double walls used between the apartments on ‘BRF
Kajplatsen’. (Reproduced from Ljunggren, S., wksb, Zeitschrift für
Wärmeschutz, Kälteschutz, Schallschutz, Brandschutz, Heft 38, 1–4, 1996.
With permission.)

Figure 15.14 Principle of the area effect. The area of the slab is much larger than the
area separating the source room and receiver room. (Reproduced from
Ljunggren, S., wksb, Zeitschrift für Wärmeschutz, Kälteschutz, Schallschutz,
Brandschutz, Heft 38, 1–4, 1996. With permission.)

Table 15.4  Single-number results of measured sound insulation on ‘BRF Kajplatsen’

Airborne, horizontal Airborne, vertical Impact, vertical
(dB) (dB) (dB)
Experimental R′w = 61 R′w = 68 L′n,w = 46
Traditional – R′w = 59 L′n,w = 51
 Experimental buildings with high sound insulation  439

80

70

60
Transmission loss, R´ (dB)

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
Experimental Traditional Avg. meas.

Figure 15.15 Measured airborne sound insulation in the vertical direction on ‘BRF

Kajplatsen’. The crosses are the average result at 50 Hz to 80 Hz in the
experimental building.

80

70
Impact sound pressure level, L΄n (dB)

60

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
Experimental Traditional

Figure 15.16 Measured impact sound insulation in the vertical direction on ‘BRF

Kajplatsen’.
440  Sound Insulation in Buildings

The measured impact sound insulation in the vertical direction is shown

in Figure 15.16. The advantage of the area effect is also obtained here, but
it is less than for the airborne sound insulation, around 5 dB to 8 dB in
most of the frequency range. At the high frequencies, the effect decreases,
but fortunately it is the low frequencies that matters here. The explana-
tion for this behaviour is that with impact excitation, the radiated sound
from the slab consists of a near field radiation as well as resonant ­radiation
(see  Equation 10.14) The relative contribution of the near field increases
with frequency and, thus, the area effect for impact sound decreases at high
frequencies.

15.5 ‘YLÖJÄRVI’ APARTMENTS, FINLAND

15.5.1 G eneral
Three apartment houses were erected in 1996 on the occasion of a Finnish
Housing Fair in Ylöjärvi. The houses have three stories and 20 apartments
in total. The building concept was developed by Tampere University of
Technology on the initiative of Skanska Oy and Finnforest Oy as part of
the Nordic Wood Project (see next Section 15.6).

15.5.2 Purpose of experiment
The purpose was to find a method to improve the sound insulation in
wooden apartment buildings by reduction of flanking transmission. The
problem to be solved was the development of a resilient structural joint
with which the non-bearing walls could be joined to the bearing beam-to-
column frames.

15.5.3 Building concept
The building concept is a column-beam structure made from laminated
veneer lumber (LVL) wood. This is type D in Figure 11.1. The prefabri-
cated ribbed slab elements are made from 51 mm thick LVL with 299 mm
high ribs having a spacing of 300 mm. The slabs are 1800 mm wide and
can have a possible span of 7 m. Details of the floor and the junctions are
shown in Figure 15.17. More information is found in the work of Keronen
and Kylliäinen (1997).

15.5.4 Acoustical results
The main problem in the research and development project was finding
a solution with sufficient impact sound insulation at low frequencies, i.e.
below 100 Hz. Impact sound pressure levels were measured from 50 Hz
to 3150 Hz, and in addition to the ISO 717-2 reference curve, Bodlund’s
 Experimental buildings with high sound insulation  441

Separating wall/floor External wall/floor

Lag screw M10 c/c 1000 Mineral wool 13 mm

Hole ø 12 through runner Density 100 kg/m3

51

299
Mineral wool 15 mm – Floor surface
– Mineral wool 30 mm
Acoustic steel Desity 30 kg/m3 (Density 100 kg/m3)
– Ribbed slab 299 mm,
Profiles 25 mm mineral wool between
c/c 400 mm ribs (density 30 kg/m3)
– Steel profiles 25 mm
– Plasterboard 13 mm
225 – Plasterboard 13 mm 150
277

Figure 15.17 Construction details of the floor and the junctions to wall and façade. The
building in Ylöjärvi has two kinds of floor surface: vinyl on 3 mm× 13 mm
plasterboard (30.8 kg/m2) or birch boarding on plastic foam layer on
2 mm× 13 mm plasterboard (18.8 kg/m2). (Reproduced from Keronen, A.,
Kylliäinen, M., Sound Insulation Structures of Beam-to-Column Framed Wooden
Apartment Buildings. Publication 77 Structural Engineering, Tampere
University of Technology, Finland, 1997. With permission.)

Table 15.5  S ingle-number results of measured sound insulation in ‘Ylöjärvi’

apartments

Airborne, horizontal Airborne, vertical Impact, vertical

(dB) (dB) (dB)
Vinyl/birch R′w = 64 R′w = 67/62 L′n,w = 44/39
floor
R′w,50 = 59 R′w,50 = 63/58 L′n,w,50 = 47/47
Source: Keronen, A., Kylliäinen, M., Sound Insulation Structures of Beam-to-Column Framed
Wooden Apartment Buildings. Publication 77 Structural Engineering, Tampere
University of Technology, Finland, 1997.
Note: Results in the vertical direction both for vinyl floor and birch floor are indicated by a
slash: */*.

reference curve was applied (see Section 5.8.3). An overview of the

­measurement results is presented in Table 15.5. Examples of the detailed
measurement results are presented in Figures 15.18 and 15.19. The results
are very satisfactory.
The following abbreviations are used in Table 15.5: R′w,50 = R′w + C 50–3150,
L′n,w,50 = L′n,w + C I,50–2500.
442  Sound Insulation in Buildings

80

70

60
Transmission loss, R΄ (dB )

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
Horizontal Vertical, birch Vertical, vinyl

Figure 15.18 Measured airborne sound insulation in the horizontal direction and in the
vertical direction both with vinyl and wooden flooring in ‘Ylöjärvi’ apart-
ments. (Adapted from Keronen, A., Kylliäinen, M., Sound Insulation Structures
of Beam-to-Column Framed Wooden Apartment Buildings. Publication 77
Structural Engineering, Tampere University of Technology, Finland, 1997.)

80

70

60
Impact sound pressure level, L΄n, dB

50

40

30

20

10

0
50 100 200 500 1000 2000 5000
Frequency (Hz)
Birch boarding floor Vinyl floor covering

Figure 15.19 Measured impact sound insulation in the vertical direction with two dif-
ferent floorings, vinyl or wood in ‘Ylöjärvi’ apartments. (Adapted from
Keronen, A., Kylliäinen, M., Sound Insulation Structures of Beam-to-Column
Framed Wooden Apartment Buildings. Publication 77 Structural Engineering,
Tampere University of Technology, Finland, 1997.)
 Experimental buildings with high sound insulation  443

15.6 NORDIC WOOD PROJECT

15.6.1 General
In the years 1993–1997, the so-called Nordic Wood project has lead to a
series of experimental wooden buildings being developed and built in the
Nordic countries (Nordic Wood, 1997; Hveem et al., 2000).

15.6.2 Purpose of experiment
The emphasis was on wood as an attractive building material. However,
the two obvious challenges in wooden multi-storey houses were fire protec-
tion and sound insulation. Thus, the purpose was to initiate research and
give support to experimental buildings that could provide possible solu-
tions to these problems.

15.6.3 Acoustical results
The greatest acoustical challenge in multi-storey wooden houses is the
impact sound at low frequencies. An overview of measurement results
of impact and airborne sound insulation from some of the experimental
buildings is found in Table 15.6. All impact sound insulation measure-
ments include frequencies down to 50 Hz, and in addition to the ISO 717-2
reference curve, Bodlund’s reference curve was applied (see the explana-
tion in Section 5.8.3).

Table 15.6  Single-number results of measured sound insulation of floor

constructions in the Nordic Wood project
Thickness R′w,50
Place (mm) L′n,w (dB) L′n,w,50 (dB) L′B (dB) R′w (dB) (dB)
Hørsholm, DK 375 44–47 52–54 58–62 59–63 –
Ylöjärvi, FI 487 36–44 41–47 49–56 62–67 –
Vik/Vikki, FI 400 48–53 54–58 62–67 58–62 –
Uleåborg/Oulo, 407 44–49 49–54 55–62 61–65 –
FI (RL-slab)
Uleåborg/Oulo, 505 51–54 54–57 62–66 61–65 –
FI (I-beam)
Solbakken, N 555 46–48 58–60 65–66 62–65 58–59
Wälludden, S 439 50–52 52–54 60–62 56–58 55–56
Orgelbäken, S 530 48–51 50–52 57–59 60–63 56–59
Source: Hveem, S. et al., Trehus i flere etasjer. Lydteknisk prosjektering. (Wooden multi-story
houses. Acoustical design, in Norwegian). Anvisning 37, Norges byggforskningsinsti-
tutt, Oslo, 2000.
444  Sound Insulation in Buildings

The following abbreviations are used in Table 15.6: L′n,w,50 = L′n,w + C I,50–

2500, L′B is impact sound evaluated with Bodlunds reference curve,
R′w,50 = R′w + C 50–5000.
The experience from the Nordic Wood project is that good sound
­insulation is possible for airborne sound and impact sound above 100 Hz,
but impact sound insulation below 100 Hz is an issue. The best result in
this aspect is the solution in Ylöjärvi, which is described in the previous
Section 15.5.

REFERENCES

Byggforskningsrådet (1991). Det tysta huset. Bätre ljudkomfort i nyproducerade

bostäder. (The quiet house. Improved acoustical comfort in new-built dwell-
ings, in Swedish). Statens råd för byggforskning, T9:1991, Stockholm, Sweden.
J. Gath and N.O. Karstoft (1992). ECO-HOUSE. Et nyt dansk byggesystem. (ECO-
HOUSE. A new Danish building system, in Danish). Dansk ståldag 1992,
Birch & Krogboe A/S. (9 pages).
S. Hveem, P. Hammer, A. Homb, A. Keronen, J.H. Rindel (2000). Trehus i flere etas-
jer. Lydteknisk prosjektering. (Wooden multi-story houses. Acoustical design,
in Noregian). Anvisning 37, Norges byggforskningsinstitutt, Oslo 2000.
A. Keronen and M. Kylliäinen (1997). Sound Insulation Structures of Beam-to-
Column Framed Wooden Apartment Buildings. Publication 77 Structural
Engineering, Tampere University of Technology, Finland.
KKS Enterprise (1992). Scanhouse – et nyt dansk byggesystem. (Scanhouse - a new
Danish building system, in Danish). Eight pages leaflet, KKS Enterprise A/S,
Hvidovre, Denmark.
S. Ljunggren (1993). Sound insulation in buildings of concrete with large span floors.
Proceedings of Noise 1993, 31 May – 3 June 1993, St. Petersburg, Russia,
83–88.
S. Ljunggren (1995). A new quiet house in Stockholm. Acta Acustica 3, 283–286.
S. Ljunggren (1996). Eine neue kosteneffektive Lösung zur Verbesserung der
Schalldämmung von Wohngebäude. (A new cost-efficient solution to improv-
ing the sound insulation of dwellings, in German). wksb, Zeitschrift für
Wärmeschutz, Kälteschutz, Schallschutz, Brandschutz, Heft 38, 1–4.
Nordic Wood (1997). Flervånings trähus. (Multi-unit wooden houses, in Swedish).
T. Hansson (Editor). Nordic Timber Council and Träinformation, Stockholm.
Index

A wave, 8–9
wave equation, 10, 205
AAD (aggregated adverse deviation), Bias error, 368
146 Bodlund-weighted impact
Absorption area, 82 sound pressure level, 144–145
Acceleration level, 2 Boundary absorption in a plate,
Accelerometer, 45, 353 see Structural absorption
Acoustical comfort, 400 coefficient
Acoustically thick plate, 216, 218 Boundary connections, 259
Airborne sound insulation Box-in-box system, 41
dose–response function, 413 Building materials, 13
measurement, 340–347 Bulk modulus, 4, 9
measurement uncertainty, 388–389
Airborne sound reduction index Ia, 145
Angular frequency, 3 C
Angular wave number, 4
Aperture, 104–106 Cardinale sine, 15
Apparent sound reduction index, see Cavity mode, 377
Transmission loss, apparent Centre frequency, 4
Area effect, 438 exact, 5
Averaging time, 337, 356, 380–381 nominal, 5
Averaging Characteristic dimension of a plate,
arithmetic, 14 167
energetic, 14 Combination of airborne and impact
geometric, 14 transmission, 321
harmonic, 14 Confidence level, 364
Axial modes, 70, 376–378 Core materials, 244
Corner position of microphone,
B 345
Correlation coefficient, 381
Background noise, 337 Coupling factor, 263
Bandwidth Coupling loss factor, 191
effective, 368 Coverage factor, 364
half-power, 32 Crest factor, 340
relative, 5 Criteria for sound insulation,
resonant system, 31 419–420
Bending Critical damping, 23
stiffness, 10 Critical frequency, 12, 112

445
446 Index

Crossover frequency, Floors

double wall, 124 floating, 283–296
shear waves, 12 heavyweight, 304
thick wall, 109 lightweight, 304
Cumulative probability function, 365 locally reacting, 292
massive, 276
D massive wood 297
point-supported, 295
Damping resonantly reacting, 290
coefficient, 24 surface-supported, 287
hysteretic, 37 Forced transmission, 197, 211,
viscous, 36 203–204
Dashpot, 19 Forced vibration, see Vibration, forced
Decay curve, 96–98 Frame absorber, 252–254
Decibel, 2 Frequency band, 4
Deflection, static, 43 Frequency response of room, 371
Diffuse sound field, 54, 56, 79–81, 86 Frequency weighting, sound, 336
Dilatational resonance, 240–243 Frequency weighting, vibration, 45, 354
Directivity factor, 87
Directivity of resonant radiation, 181 G
Dispersion, 9, 11
Dispersion curve, 12, 237 Gas filling in windows, 269
Displacement, static, 19 Gaussian distribution, 364, 406
Dose-effect relationships, 398 Geophone, 45, 353
Dose–response functions, 405–416 Global method, 346
Double leaf construction, 121 Grazing incidence, 210
Drywalls, 265–269 Group speed, 324–325
Dynamic stiffness of mineral
wool, 288 H

E Harmonic vibration, 3, 20
Hearing threshold, 7
Elastic connections, 262, 326–328 Heavyweight wall, 223
Element method, 347
Enclosures, 135 I
Energy balance, in a resonant system,
189 IIC (impact insulation class), 143
Energy balance, in a room, 82 Imaginary unit, 3
Energy density, 51–52, 54, 56 Impact sound
Equal loudness contour, 6 attenuation, 138, 280–297, 302
Equivalent absorption length, 325, 360 complaints, 305
Equivalent number of uncorrelated dose–response function, 415
positions, 383–385 heavyweight floors, 307–309
Excitation, acoustic, 198 index Ii, 145
External transmission loss, 226 insulation, 275
lightweight floors, 306–308
F measurement of, 348–349
Impact sound pressure level, 136, 275
Flanking transmission, 131, 313–332 Impedance
airborne sound, 313–317 characteristic, 107
building types, 314 hammer, 285
impact sound, 318–320 mechanical, 21, 27, 262
Floor covering, 280 point, 21, 276
 Index 447

surface, 290 Measurement positions, 338, 341, 350,

transfer, 21 354
Insertion loss, 38–39 Mechanical impedance, see
Integrated impulse response method, Impedance, mechanical
96, 350, 381, 387 Microphone, 335
Integration time, see Averaging time MLS (maximum length sequence), 96,
Interference pattern, 61–67 340
Interrupted noise method, 96, 350, 386 Mobility, see Impedance, mechanical
Modal density, 74–79, 192–194
J Motion equation, 19, 23
Multi-element partition, 104–105
Junction attenuation, 321, 324
Junction attenuation, measurement of, N
359
Natural frequency, 25, 31, 70
K Node, 71
Noise and health, 395–399
Knee frequency Noise
double wall, 250 descriptors, 395
floating floor, 290 excitation, 367
from neighbours, 400–404
L Non-diffuse rooms, 94
Normal modes in a plate, 155–160
Laminated glass, 117, 272 Normalized impact sound pressure
Lightweight wall, 224 level, 137
Line connections, 257 Normalized level difference, 104
Live walker, 302–304
Logarithmic decrement, 25 O
Logistic function, 405–406
Logit analysis, 404–406 Oblique modes, 71, 376–378
Longitudinal wave, 9 Octave band, 5
Loss factor, 26, 33, 190, 210, 212 One-sided test, 364–365
measurement of, 357 One-third octave band, 5
upper limit, 357 OR (odds ratio), 398
Loudness level, 6 Orthotropic plate, 229
Low frequency
impact sound, 304–310 P
noise from neighbours, 417–418
procedure, 345 Particle velocity, 8, 49, 56
sound, 6, 336 Period duration, 3, 25
variance, 368 Phase speed, 53, 324
Phasor, 20
M Phon, 6
Plane wave, 8
Mach number, 12, 177 Point connections, 255
Manual scanning, 344 Point excitation, 276
Mass law, 109, 204 Point impedance, see Impedance, point
Mass-air-mass resonance, see Poisson distribution, 375
Resonance frequency, Poisson ratio, 9
mass-air-mass Power input, 34, 285
Massive wood, see Floors, massive Probability density function, 364, 374
wood Probit analysis, 404
Mean free path, 90–91 Pure tone excitation, 365
448 Index

Q S
Q-factor, 37, 308, 310 Sabine’s formula, 83–85, 103
Sandwich plate, 234
R forced transmission, 240
resonant transmission, 238
Radiation efficiency, 163 Schroeder’s limiting frequency, 79,
Radiation efficiency, measurement of, 367, 370
358 SEA (statistical energy analysis),
Radiation loss factor, 210 189–201
Rayleigh’s method, 164 Shear modulus, 10
Reciprocity principle, 193, 227, Shear wave, 9
317–318 Sigmoid functions, 405
Reference curve, airborne sound Sinc, see Cardinal sine
insulation, 141 Sine-sweep, 96, 340, 342
Reference curve, impact sound, Single leaf construction, 106
144 Single-number quantity, measurement
Reflection uncertainty, 389
density, 92 Single-number rating, 140
influence of bandwidth, 61 Sleep disturbance, 396
normal incidence, 49 Slits, 128–131
oblique incidence, 52 SNR (signal-to-noise ratio), 341
random incidence, 54 Socio-acoustic surveys, 408–410
Refraction effect, 184 Sound absorption in air, 91
Requirement for sound insulation, 150, Sound bridges, double walls,
390, 419 254–265
Resonance, 30 Sound bridges, floating floors, 296
Resonance frequency, 30, 42 Sound insulation,
floating floor, 288, 291 airborne, 101
floor covering, 280 homogeneous, isotropic plate,
mass-air-mass, 123, 249 203
suspended ceiling, 297 impact, 136
Resonance peak, floating floor, 308 low frequencies, 221
Resonant system, 189 oblique incidence, 226
Resonant transmission, see Sound subdivided structures, 224
transmission, resonant thick plates, 218–221
Resonant vibration, see Vibration, thin plates, 216–218
resonant Sound power level, 2, 85
Reverberation distance, 85–88 Sound pressure level, 2
Reverberation formula, general, 89 A-weighted, 6
Reverberation time, 26, 83 measurement of, 335–339
calculation, 93 measurement uncertainty,
correction for, 339 374–381
evaluation range, 97 Sound propagation in plates, 8
lower limits, 351 Sound radiation, 160–181
measurement of, 96–98, 350–353 directivity, 181–186
measurement uncertainty, forced vibrations, 166–173
385–387 near field, 278
modal, 72 resonant vibrations, 174–181
RMS (root-mean-square), 2 Sound reduction index, see
Room constant, 87 Transmission loss
Room modes, 70–74 Sound transmission coefficient, see
RR (risk ratio), 398 Transmission coefficient
 Index 449

Sound transmission loss, see Time factor, 4

Transmission loss Time reversed analysis, 351–352,
Sound transmission 357
forced, 109–116 Traffic noise annoyance, 407, 411
inside to outside, 228 Traffic noise spectrum, 146–149
outside to inside, 227 Transfer function, 74
resonant, 115, 194–197, 203–204, Transmissibility, 35, 37
214 Transmission coefficient, 101
via cavity, 250, 252 Transmission coefficient, structural,
via sound bridges, 254, 264 321, 324
Sound build-up, 84–85 Transmission loss, 101
Sound decay, 83–85 apparent, 103, 317
Spatial averaging, 338, 341, 353, random incidence, 111
381–384 Transverse wave speed, 11, 235
continuous, 344, 382 Two-sided test, 364–365
discrete, 343, 382
Spatial standard deviation, 372–374 U
Spatial variance, 367
Spectrum adaptation term, Uncertainty in dB, 380
146–149 Uncertainty of a measurement result,
Speed of sound, 4 363
Speed of transverse waves, see Uncertainty, expanded, 364
Transverse wave speed Uncertainty, standard, 364
Standard deviation, 363
Standardized impact sound pressure V
level, 137
Standardized level difference, 104 Variance, 363
Standing wave, 51, 69, 69–74 Variance of sound pressure,
Stationary sound field, 85–89 365–373
STC (sound transmission class), VDV (Vibration dose value), 356
142 Velocity level, 2
Stiffness, dynamic, 20, 288–290 Vibration, 19
Stiffness, static, 19 dose–response function, 416
Structural absorption coefficient, evaluation of, 46–47
210–211 forced, 20, 26–29, 33, 184
Structural reverberation time, 357 free, 23
Structural transmission coefficient, human response, 44
see Transmission coefficient, insulation, 33, 40
structural measurement of, 353–356
Structure-borne vibration, 40–42 resonant, 20, 184
Sum of airborne and impact Vibration isolator, 35, 37, 42
sound insulation, 138, Vibration reduction index, 326
321, 332
Supersonic bending waves, 174 W
Suspended ceilings, 297
Wall impedance, 109, 113, 205–209,
T 236, 248
Waterhouse correction, 58, 88
Tangential modes, 71, 376–378 Wave equation, 8, 69
Tapping machine, 136, 276, Wave speed, 9
302–304 Wavelength, 4
Three-layer units, 271 Weighted apparent sound reduction
Time constant, 25 index, 142
450 Index

Weighted impact sound pressure level, Willingness to pay, 404

142–144 Windows, 117, 126–127, 269
Weighted sound reduction index,
140 Y
Weighted standardized level
difference, 142 Young’s modulus, 9