The Composition and Performance

of Spatial Music

A dissertation submitted to the University of Dublin for the degree of Doctor of

Philosophy

Enda Bates
Trinity College Dublin, August 2009.

Department of Music &

Department of Electronic and Electrical Engineering
Trinity College Dublin

Declaration

I hereby declare that this thesis has not been submitted as an exercise for a degree at
this or any other University and that it is entirely my own work.
I agree that the Library may lend or copy this thesis upon request.

Signed,
___________________
Enda Bates

ii

Summary
The use of space as a musical parameter is a complex issue which involves a
number of different, yet interrelated factors. The technical means of performance, the
sonic material, and the overall musical aesthetic must all work in tandem to produce a
spatial impression in the listener which is in some way musically significant.
Performances of spatial music typically involve a distributed audience and often take
place in an acoustically reverberant space. This situation is quite different from the
case of a single listener at home, or the composer in the studio. As a result, spatial
strategies which are effective in this context may not be perceived correctly when
transferred to a performance venue. This thesis examines these complex issues in
terms of both the technical means of spatialization, and the compositional approach to
the use of space as a musical parameter. Particular attention will be paid to the
effectiveness of different spatialization techniques in a performance context, and what
this implies for compositional strategies which use space as a musical parameter.
Finally, a number of well known works of spatial music, and some original
compositions by the author, are analyzed in terms of the perceptual effectiveness of
the spatialization strategy.
The results of a large number of listening tests and simulations were analysed
to determine the fundamental capabilities of different spatialization techniques under
the less than ideal conditions typically encountered during a performance. This
analysis focussed on multichannel stereophony, Ambisonics, and Wavefield
Synthesis. Other methods which are orientated toward a single listener are not
addressed in this thesis. The results indicated that each spatialization scheme has
particular strengths and weaknesses, and that the optimum technique in any situation
is dependent on the particular spatial effect required. It was found that stereophonic
techniques based on amplitude panning provided the most accurate localization but
suffered from a lack of spaciousness and envelopment. Ambisonics provided an
improved sense of envelopment but poor localization accuracy, particularly with first
order Ambisonics systems. Consequently it would appear that stereophony is
preferable when the directionality and focus of the virtual source is paramount, while
Ambisonics is preferable if a more diffuse enveloping sound field is required.
Ambisonics was consistently preferred for dynamically moving sources as this
iii

technique eliminated the panning artefacts exhibited by amplitude panning as the

source moves from a position at a loudspeaker, to one inbetween a pair of
loudspeakers. The decoding scheme and order of the Ambisonics system also has a
significant effect on the perceptual performance of the system, particularly at offcentre listener positions. A single-band, max-rE decoding scheme was found to be the
most suitable approach for a distributed audience, and increasing the order of the
system was shown to improve the performance at all listener positions. It is
recommended that an octagonal array be adopted as a minimum standard for
performances of multichannel spatial music, as this arrangement can be utilized for
higher order Ambisonics and can also be readily implemented with digital audio
hardware.
Wavefield synthesis (WFS) was found to be quite distinct from multichannel
techniques such as stereophony or Ambisonics. The spatial aliasing frequency is a
critical aspect of any WFS system and localization errors and timbral distortions
significantly increase if this parameter is too low. The ability of WFS systems to
position virtual sources both behind and in front of the loudspeaker array was shown
to be extremely difficult to achieve, particularly if the listeners position is fixed or if
the performance space contains significant early reflections and reverberation.
In the latter half of this thesis, a number of landmark works of spatial music
were presented and analysed in terms of the perceptual validity of their approach to
spatialization. It was shown that many composers have used spatial distribution to
improve the intelligibility of different layers of material, and this approach was found
to agree with the findings of scientific research in the area of auditory cognition. The
use of recognizable spatial motifs was shown to be highly difficult to implement, and
complex, abstract spatial designs are only indirectly related to what is eventually
perceived by the audience. A gestural approach to spatial music has its origins in the
practice of diffusion, yet this approach is equally applicable to other aesthetics and
would seem to be highly suitable for mixed-media electroacoustic works. The use of
augmented instruments which map the actions of the performer to a spatialization
algorithm would seem to be well suited to performances of mixed-media spatial
music. In addition, the use of flocking algorithms to control spatialization and sound
synthesis also appears to be a novel and effective techniques for the creation of
spatially dynamic, electronic sounds.

iv

Acknowledgements
I am deeply grateful to my two supervisors, Dr. Dermot Furlong and Mr. Donnacha
Dennehy, for all their support and guidance over the past four years. Without their
incredible knowledge and encouragement, this thesis would not have been possible.
I would also like to particularly thank Dr. Fionnuala Conway and Gavin Kearney for
their help over the years, all those who took part in listening tests, and my colleagues
at the Spatial Music Collective.
Finally I am very grateful for the support of my family and friends, and all the
students and staff of the Music and Media Technology course in Trinity College.

Table of Contents
Summary ...................................................................................................................... iii
Acknowledgements........................................................................................................v
Table of Contents..........................................................................................................vi
List of Figures ................................................................................................................x
List of Tables ..............................................................................................................xiv
1 Introduction.................................................................................................................1
1.1 Spatial Music: A Personal Perspective ................................................................1
1.1.1 What Now in the Age of Disillusionment.....................................................2
1.1.2 Why Spatial Music?......................................................................................3
1.1.3 Why Talk About Spatial Hearing?................................................................5
1.1.4 The Imaginative Use of Empirical Thinking ................................................6
1.2 The Research Question ........................................................................................7
1.3 Aims and Objectives ............................................................................................8
1.4 Methodology ........................................................................................................8
1.5 Motivation............................................................................................................9
1.6 Outline................................................................................................................10
2 Spatial Hearing..........................................................................................................13
2.1 Directional Hearing............................................................................................14
2.2 Directional Hearing and Acoustics ....................................................................19
2.3 Distance Hearing & Moving Sources ................................................................22
2.3.1 Summary of Spatial Hearing.......................................................................25
2.4 Spatial Hearing with Multiple Sources ..............................................................26
2.4.1 The Limits of Auditory Perception .............................................................28
2.4.2 Spatial Hearing and Virtual Sources...........................................................29
2.4.3 Spatial Audio Techniques ...........................................................................30
3 Stereophony ..............................................................................................................32
3.1 Quadraphonic Sound..........................................................................................35
3.2 Cinema Surround Sound and 5.1 .......................................................................37
3.3 Multichannel Amplitude Panning Techniques...................................................38
3.4 Theories of Stereophonic Localization ..............................................................39
3.5 Critics of Summing Localization .......................................................................40
3.6 Meta-Theories of Localization...........................................................................41
vi

4 Sound Field Reconstruction......................................................................................43

4.1 Ambisonics ........................................................................................................43
4.1.1 Ambisonics and Microphone Directivity Patterns......................................44
4.1.2 Ambisonic Decoders...................................................................................47
4.1.3 Higher Order Ambisonics ...........................................................................48
4.1.4 Ambisonics in Practice ...............................................................................49
4.2 Wavefield Synthesis...........................................................................................55
4.2.1 WFS and Spatial Music ..............................................................................60
5 The Simulation of Distance ......................................................................................62
5.1 The Simulation of Moving Sources ...................................................................62
5.2 A General Model for Spatial Processing of Sounds ..........................................63
5.2.1 The Implications of Moores Model ...........................................................65
5.3 Ambisonics Distance Encoding .........................................................................66
5.4 Evaluating of Spatialization Techniques ...........................................................67
6 The Assessment of Spatial Audio .............................................................................68
6.1 Localization Accuracy under Ideal Conditions..................................................69
6.2 Assessing Stereophony ......................................................................................70
6.2.1 Discussion ...................................................................................................71
6.3 Assessing Ambisonics .......................................................................................72
6.3.1 Discussion ...................................................................................................75
6.4 Comparing Stereophony and Ambisonics .........................................................76
6.4.1 Discussion ...................................................................................................79
6.5 The Influence of the Listening Room ................................................................81
6.5.1 Spatialization Assessment in a Small Concert Hall....................................81
6.6 Artificial Reverberation and Spatial Dissonance...............................................88
6.7 Evaluating Wavefield Synthesis ........................................................................91
6.7.1 WFS Localization Accuracy and Spatial Aliasing...................................91
6.7.2 WFS in Real Rooms....................................................................................96
6.7.3 WFS Distance Effects ...............................................................................101
6.7.4 WFS Spectral Colouration ........................................................................105
6.7.5 WFS Conclusions and Discussion .........................................................106
6.8 Analysis of Results and Recommendations.....................................................108
6.8.1 Discussion .................................................................................................110
6.8.2 Implications...............................................................................................112
vii

7 Acoustic Spatial Music ...........................................................................................114

7.1 Early Spatial Music..........................................................................................115
7.2 The Co-Existence of Dissimilars .....................................................................118
8 Musique Concrte & Elektronische Musik.............................................................127
8.1 Stockhausen and the Serialization of Space.....................................................130
8.2 Architectural Spatial Music .............................................................................139
8.3 The Perception of Abstract Spatial Designs ....................................................144
8.4 Diffusion, and the legacy of Musique Concrte ..............................................150
8.5 Spectromorphology and the Gestural use of Space .........................................154
9 Electroacoustic Spatial Music.................................................................................159
9.1 Fixed Media Electroacoustic Spatial Music ....................................................159
9.2 Real-Time Electroacoustic Music....................................................................161
9.3 Real-Time Spatialization .................................................................................166
10 Spatial Music Composition...................................................................................169
10.1 Case Study I: Discordianism..........................................................................169
10.1.1 Analysis...................................................................................................170
10.1.2 Critical Evaluation ..................................................................................171
10.2 Case Study II: Sea Swell (for four SATB choirs)..........................................172
10.2.1 Analysis...................................................................................................174
10.2.2 Critical Evaluation ..................................................................................175
10.3 Case Study III: String Quartet No. I ..............................................................176
10.3.1 Analysis...................................................................................................176
10.3.2 The Spatialization Process ......................................................................179
10.3.3 Critical Evaluation ..................................................................................181
10.4 Case Study IV: Auto Harp .............................................................................182
10.4.1 Analysis...................................................................................................185
10.4.2 Critical Evaluation ..................................................................................188
11 Spatial Music Instruments.....................................................................................189
11.1 Augmented Strings.........................................................................................190
11.2 The Hexaphonic Guitar..................................................................................192
11.3 Case Study V: Etude No. 2 for Hexaphonic Guitar .......................................193
11.3.1 Analysis...................................................................................................194
11.3.2 Critical Evaluation ..................................................................................195
12 Behavioural Space ................................................................................................196
viii

12.1 The BOIDS Flocking Algorithm ...................................................................196

12.2 Case Study VI: Flock .....................................................................................198
12.2.1 Analysis...................................................................................................199
12.2.2 Critical Evaluation ..................................................................................200
12.3 Offline Spatialization with the Boids Algorithm ...........................................201
12.3.1 Granulation .............................................................................................203
12.3.2 Grainlet Additive Synthesis ....................................................................203
12.3.3 Glisson Synthesis ....................................................................................204
12.4 Case Study VII - Rise ....................................................................................205
12.4.1 Analysis...................................................................................................205
12.4.2 Critical Evaluation ..................................................................................206
13 Summary & Conclusion........................................................................................207
13.1 Spatialization Results & Recommendations..................................................207
13.2 Perception and Spatial Music Aesthetics.......................................................211
13.3 Conclusion .....................................................................................................214
14 List of Publications ...............................................................................................217
15 Summary of Published Articles ............................................................................218
Bibliography ..............................................................................................................220
Appendix A Scores .................................................................................................243

ix

List of Figures
Fig. 2.1 Spherical coordinate system ...........................................................................13
Fig. 2.2 Lateral source example...................................................................................15
Fig. 2.3 Zone-of-confusion example............................................................................17
Fig. 2.4 Impulse response of a room with a relatively short reverb time ....................18
Fig. 2.5 Source distance v sound intensity...................................................................23
Fig. 2.6 Spatial cues and stream segregation ...............................................................28
Fig. 3.1 Bell Labs stereophony, proposed (left) and implemented (right)...................32
Fig. 3.2 Blumleins coincident microphone arrangement............................................33
Fig. 3.3 Standard stereophonic arrangement................................................................34
Fig. 3.4 Lateral phantom source direction versus ILD for a quadraphonic layout ......36
Fig. 3.5 ITU 5.1 loudspeaker arrangement ..................................................................37
Fig. 4.1 Basic multichannel stereophony example ......................................................44
Fig. 4.2 Microphone responses derived from two figure-of-eight microphones .........45
Fig. 4.3 The Soundfield microphone ...........................................................................46
Fig. 4.4 Zero and first order spherical harmonics........................................................47
Fig. 4.5 Second (top) and third (bottom) order spherical harmonics...........................49
Fig. 4.6 First, second & third order microphones ........................................................49
Fig. 4.7 Ambisonic plane wave - theoretical (left) and real (right) sources ................53
Fig. 4.8 Directivity patterns of various ambisonic decoding schemes ........................54
Fig. 4.9 Illustration of the Huygens principle ..............................................................55
Fig. 4.10 The WFS method..........................................................................................56
Fig. 4.11 WFS reproduction of different wavefronts...................................................56
Fig. 4.12 WFS reproduction of two-channel stereo.....................................................57
Fig. 4.13 Truncation effects (a) and (b) 4ms later .......................................................58
Fig. 4.14 A distant source (a) below the aliasing freq & (b) above.............................59
Fig. 4.15 Optimised Phantom Source Imaging WFS...................................................59
Fig. 4.16 A WFS cinema system in Ilmenau ...............................................................60
Fig. 5.1 Moores spatial model (a) direct signal (b) reflected signal paths .................64
Fig. 6.1 Direction of phantom source versus ILD reported by Theile.........................70
Fig. 6.2 Naturalness responses reported by Guastavino ..............................................74
Fig. 6.3 Decoder criteria related to the size of the listening area.................................75
x

Fig. 6.4 Mean ratings as reported by Guastavino ........................................................77

Fig. 6.5 Localization accuracy results as reported by Capra .......................................78
Fig. 6.6 Ambisonic decoder directivity patterns..........................................................80
Fig. 6.7 Geometry of loudspeaker array and audience area.........................................82
Fig. 6.8 Listening tests loudspeaker configuration ......................................................82
Fig. 6.9 Reported (blue) and actual (orange) direction for a source at speaker 14......84
Fig. 6.10 Overall subjective localization performance ................................................85
Fig. 6.11 Ambisonics system evaluated by Frank .......................................................86
Fig. 6.12 Quality maps for 5ms (left) and 50ms (right) time difference .....................87
Fig. 6.13 Specular and diffuse reflections ...................................................................89
Fig. 6.14 Visibility of a virtual source in a WFS system.............................................92
Fig. 6.15 Subjective, scaled (1-5) judgments of locatedness reported by Huber.........94
Fig. 6.16 Standard deviation in horizontal localization reported by Huber.................94
Fig. 6.17 Test setup for Wittek's listening tests ...........................................................95
Fig. 6.18 Subjective assessment of locatedness reported by Wittek............................95
Fig. 6.19 Concert hall layout used by Start..................................................................97
Fig. 6.20 Perceived WFS virtual source directions reported by Start in......................98
Fig. 6.21 Virtual Source (a) behind the array & (b) in front of the array ....................99
Fig. 6.22 Test setup for Marentakiss listening tests .................................................100
Fig. 6.23 Aliasing frequency relative to listener position reported by Marentakis....100
Fig. 6.24 Technical setup for listening tests by Nogues ............................................102
Fig. 6.25 Distance of real (left) and virtual (right) sources reported by Kerber ........103
Fig. 6.26 WFS reproduction room reflections ...........................................................104
Fig. 6.27 WFS virtual source and simulated room acoustic ......................................104
Fig. 6.28 Perceived colouration for various WFS and OPSI systems .......................105
Fig. 7.1 Interior of St Marks Cathedral in Venice .....................................................114
Fig. 7.2 Echo effects in Giovanni Gabrielli's In Ecclesiis .........................................116
Fig. 7.3 Tuba Mirum section in Berliozs Requiem ..................................................117
Fig. 7.4 Spatial location of musicians in Brants Millennium II (1954)....................122
Fig. 7.5 Spatial patterns in Brants Millennium II (1954) .........................................123
Fig. 7.6 Sound axes in Brants Millennium II (1954)................................................123
Fig. 8.1 Pierre Henry performing with the potentiomtre d'espace, Paris, 1952 .......127
Fig. 8.2 Gesang der Jnglinge premiere at WDR Cologne, 1956..............................132
Fig. 8.3 Rehearsal of Gruppen in Cologne, March 1958 ...........................................134
xi

Fig. 8.4 Dress rehearsal of Carre in Hamburg, October 1960 ...................................134

Fig. 8.5 Spatial movement in Gruppen ......................................................................135
Fig. 8.6 Orchestra & audience layout in Stockhausens Carre ..................................136
Fig. 8.7 Spatial intervals and directions from Stockhausens Musik in Space ..........137
Fig. 8.8 Stockhausen with his rotating loudspeaker mechanism ...............................138
Fig. 8.9 The Philips Pavilion at the 1958 Worlds Fair in Brussels............................140
Fig. 8.10 String glissandi, bars 309-14 of Metastasis by Iannis Xenakis ..................140
Fig. 8.11 Orchestral disposition of Terretektorh........................................................142
Fig. 8.12 Karlheinz Stockhausen performing at OSAKA 70.....................................144
Fig. 8.13 Layout of instruments in Goreckis Genesis Cycle 1974 ...........................149
Fig. 8.14 An early concert with the Acousmonium ...................................................151
Fig. 8.15 Standard stereo setup (left), diffusion setup (right)....................................151
Fig. 8.16 The full BEAST system..............................................................................153
Fig. 8.17 Denis Smalleys perceived space ...............................................................156
Fig. 9.1 IRCAMs 4X (left) and Matrix 32 (right) digital processors .......................162
Fig. 9.2 Layout and spatialization diagram for Rpons .............................................163
Fig. 9.3 Amplitude envelopes for the six soloists in Rpons.....................................164
Fig. 9.4 The spatial switching mechanism used in Rpons .......................................165
Fig. 10.1 Sea Swell harmonic structure ..................................................................173
Fig. 10.2 Sea Swell rhythmic alliteration................................................................174
Fig. 10.3 Sea Swell notation example for opening sibilant section ........................174
Fig. 10.4 Sea Swell introductory section ................................................................175
Fig. 10.5 A Rosen autoharp....................................................................................182
Fig. 10.6 Auto Harp source material, grouped according to order of surrogacy .......183
Fig. 10.7 Loudspeaker envelopes in Auto Harp Franssen section.............................184
Fig. 10.8 Monophonic reduction of Auto Harp indicating section durations ............187
Fig. 10.9 The Penrose stairs.......................................................................................177
Fig. 10.10 Opening eight bars of String Quartet No. I .............................................177
Fig. 10.11 Large scale structure of String Quartet No. I............................................179
Fig. 10.12 Diffusion score example...........................................................................179
Fig. 10.13 Diffusion ranges in String Quartet No. I ..................................................180
Fig. 10.14 String Quartet No. I - diffusion score ......................................................180
Fig. 11.1 The Max Mathews augmented violin .........................................................191
Fig. 11.2 The Hypercello system ...............................................................................191
xii

Fig. 11.3 Roland 13-pin wiring diagram....................................................................193

Fig. 11.4 13-pin breakout box schematic...................................................................193
Fig. 11.5 Etude No. 3 tuning, spatialization and interval sequence..........................194
Fig. 12.1 Eric Singers Boids for Max MSP/Jitter.....................................................197
Fig. 12.2 Illustration of a flock avoiding a predator ..................................................197
Fig. 12.3 Max MSP Patch used to construct the composition Flock .........................198
Fig. 13.1 Octagonal loudspeaker layout with centre front (a) and without (b)..........209

xiii

List of Tables
Table 4.1 Furse-Malham set of encoding coefficients.................................................50
Table 4.2 Analysis of Ambisonics normalisation schemes .........................................51
Table 4.3 Summary of ambisonic decoding schemes..................................................53
Table 8.1 Auto Harp tuning scheme ..........................................................................185

xiv

1 Introduction
One of the defining characteristics of electroacoustic music has been the use of
technology to expand musical boundaries, whether through electronic processes in
conjunction with traditional instruments, or through electronic processes alone. This
allowed previously neglected musical parameters such as timbre to come to the fore,
while other parameters such as rhythm, melody and traditional harmony were
relegated to the background. The use of space as a musical parameter is another novel
aspect of this artistic movement. However, space is in many respects, fundamentally
different from these other parameters. Timbre relates to the spectral and temporal
relationship between the components of an individual sound object, while rhythm,
melody and harmony involve temporal and spectral relationships between sound
objects. Space as a musical parameter is, however, much broader and more difficult
to define. It incorporates the dimensions of individual sound objects, the relationships
between sound objects and even the relationship between the sound objects and the
acoustic space in which they are heard. Furthermore, audible space itself can only
become apparent through the temporal development of sounds themselves. Space as a
musical parameter is therefore all-encompassing and yet difficult to define. It
encompasses every aspect of a sound and yet the spatial aspects of a sound are often
not consciously perceived as separate from the sound itself. In our evolutionary
development, the where is dealt with immediately by our subconscious while the what
and why become the primary focus of conscious attention.

1.1 Spatial Music: A Personal Perspective

Men fight for freedom, then they begin to accumulate laws to take it away from
themselves.
Thomas Jefferson

The most difficult question I encounter when composing a piece of spatial

electroacoustic music is why, specifically, am I doing what Im doing? Why am I
placing a sound at this location, or why am I moving a sound in space in this
particular way? Why use a certain from of audio synthesis or this particular
spatialization technique? As an Irish composer in the early 21st century I see no

necessity to pledge allegiance to any one aesthetic or style. Indeed, as someone

whose musical origins are in punk rock, I find the idea of an overarching musical
ideology which orders and informs every composition to be distinctly unappealing.
The most attractive aspect of electroacoustic composition, at least to this composer, is
the complete freedom to write and perform music using any style or genre, any
instrument or object, and the most sophisticated and cutting-edge technology.
However, the fact remains that in the absence of an overriding ideology or aesthetic,
the questions presented at the start of this section prove troublingly difficult to
answer. Of course intuition and instinct are a critical and often underestimated aspect
of the compositional process. The initial choice of material and basic idea for a piece
are generally intuitively chosen, and many of the specific decisions made during the
composition may also be instinctual. However, the development of the initial idea
into the larger structure of the overall work generally requires some form of
intellectual process. When walking, each individual step may be taken intuitively, but
in order to get anywhere, we need to consciously think about where we are going, and
why.

1.1.1 What Now in the Age of Disillusionment

Order without liberty and liberty without order are equally destructive.
Theodore Roosevelt

The basic philosophy outlined in the previous Section is not in any sense
unusual and it could in fact be considered as quite symptomatic of our times. The title
of this section has been used to refer to various time periods following the First World
War, however this description is as appropriate now as in any other period of history.
The English speaking world has lost a great deal of faith in science, religion, politics,
and the media (often it should be said with very good reason) and while this can be in
many respects quite liberating, it may also lead to a certain loss of direction. In a
musical context, this dichotomy has been further magnified by the further
development of digital technology which fundamentally alters the relationship
between the composer and their instrument. In the past, the sheer expense of
hardware-based music technology meant that composers were by necessity limited in
the technological options available to them. However, this in turn forced composers
to be highly creative in their use of these limited devices, and to develop sophisticated

aesthetical frameworks which justified their particular mode of composition. In

contrast, when a modern composer starts an application such as Cycling74s
Max/MSP, Csound or Supercollidor, they see an empty screen, which may then be
filled in a bewilderingly variety of ways. Virtually any sound, acoustic of synthetic,
is readily and often freely available, and these sounds may be played or transformed
using virtually any technique ever invented. Complex sounds and processes can
therefore be rapidly produced simply by opening presets and randomly adjusting
parameters, however, this approach quickly leads to aimless composition and clich.
This dichotomy between freedom and order is also apparent in the contrast between
the creative possibilities offered by this software, and the logical and highly structured
approach required to actually write a program which can achieve these artistic goals.
The two sided nature of this process is mirrored in my own work as a composer. I
desire some form of logical framework to guide my inspiration to a finished piece of
music, but this framework is not some overarching philosophy, merely the particular,
and importantly, the most appropriate aesthetic for this particular piece.

1.1.2 Why Spatial Music?

Electronic music outside of the electroacoustic tradition consists of a vast
array of different styles and genres, many of which are strongly dance orientated. In
this context, the performers ability to work the crowd is perhaps as important as their
interaction with their musical instrument, which may be a laptop, hardware
controllers, or even turntables and vinyl records. In less dance-orientated styles, the
situation is much closer to that of a traditional performance, as the audience is more
concerned with observing the performer. However, there is a fundamental difference
between observing a performance by a violinist, and a performance by a laptop
musician. In the latter case, it is often difficult to relate the visible actions of the
performer to the audible result, as there is no clearly defined relationship between the
physical input and the resulting sound. The visual actions of the performer will be the
same if they are synthesizing and creating sounds in real-time, or if they are simply
triggering the playback of pre-recorded tracks. When I encountered spatial electronic
music for the first time, I was immediately struck by the contrast between the
dynamic movement of sounds in space in this music, and the static nature of the
sounds and the performers at the concerts of electronic music I had encountered up to

this point. I very quickly formed the opinion that spatialization was a critical and
necessary aspect of electronic music, as the spatial movement of sounds provided a
physicality and dynamism that greatly enhanced the listening experience. This
opinion was reinforced when I realized that the earliest practitioners of electronic
music had also believed in the absolute necessity of spatialization, particularly in a
performance context (see Chapter 8).
As with many other composers, this early enthusiasm was quickly tempered
by the experience of hearing my work performed for the first time outside of the
studio over a large loudspeaker array. The piece Discordianism, which is discussed in
detail in Chapter 10.2, is a clear example of a work which functions quite well for a
single listener, but rapidly degrades as the size of the listening area increases. This is
particularly true of the dynamically moving noise drone which is the focal point of the
third and final movement. Amplitude panning and a quadraphonic system simply
could not effectively reproduce this movement when the distance between the
loudspeakers was increased to cater for a larger audience. However, having created a
stereo mix of this piece for a CD release, I remained convinced of the benefits of
spatialization, as much of the dynamism and clarity of the work was lost in the
reduction to two channels. As I began to explore these issues it quickly became
apparent that a gap existed between empirical research on spatialization techniques
and the issues faced by a composer of spatial music during a live performance. The
majority of the scientific research conducted in this area has focused on the ideal
conditions of a single listener in an acoustically treated space. In addition, much of
the writing by composers of spatial music has discussed both the single listener and
performance contexts and this seemed somewhat anachronistic considering the drastic
decline in sales of music media in the modern era and the continued widespread
enthusiasm for live performances. These very practical concerns were the initial
motivation for much of the technical research discussed in this thesis. However, the
experience of composing Discordianism also raised other, broader questions about the
nature of spatial music composition. While the organization of material in this work
can be analysed using Denis Smalleys theory of spectromorphology (see Chapter 8.5)
this was not consciously implemented at the time. Certain aspects of the work are
quiet effective, such as the spatial distribution of different layers of rhythmic material
for example. However, the way in which spatial movement is used to support gestural
interplay between the different layers of material is quite inconsistent, as there was no
4

underlying rationale guiding the compositional process in this regard. In addition this
argument could perhaps also be applied to the mix of different aesthetics contained
within the piece. As I began to examine how other composers had dealt with the use
of space as musical parameter, it quickly became apparent that this issue could not be
treated in isolation, as the way in which these composers used space is intimately
connected to their overall musical philosophy. A study of the use of space in music,
rapidly lead to debates which are as old as electronic music itself. Spatial music is in
many respects a microcosm of electroacoustic music which can refer to many of the
different styles within this aesthetic, but is not tied to any one in particular. The study
of the aesthetics of spatial music and the musical use of space as a musical parameter
therefore appeared to be a good way to indirectly approach electroacoustic music
composition and the performance of electronic music in general.

1.1.3 Why Talk About Spatial Hearing?

While we are free to choose our actions, we are not free to choose the consequences
of our actions.
Stephen R. Covey

The two, seemingly distinct issues discussed in the previous Section, one quite
practical and technical, the other more conceptual and artistic, emerged at the same
time from a single piece of music, and have been ever present since. It therefore felt
entirely natural to me to approach the study of spatial music via these distinct, yet
interrelated topics. The musical use of space exploits a fundamental aspect of
hearing, namely our ability to locate sounds in space. However unlike other
parameters such as pitch, rhythm and timbre, the movement of sounds through space
is usually not an intrinsic aspect of the sound itself, but is instead an illusion created
through the careful blending of a number of static sources. A composer of spatial
music cannot therefore treat this parameter in the same way as pitch or rhythm, and
the technical details of how this illusion is created and maintained are clearly a
fundamental and necessary part of the composers craft. While practical problems
remain, there exists a clear empirical approach to solving these issues, and this will
allow us to impose some order on space by revealing just how far the illusion can be
pushed before it falls apart. This in turn will provide us with a firm basis upon which
we can reliably exploit space as a musical parameter. The empirical and systematic
examination of these techniques will indicate their particular strengths, capabilities
5

and weaknesses and will in effect function as a form of orchestration guide for spatial,
electroacoustic music

1.1.4 The Imaginative Use of Empirical Thinking

The composition of electronic music requires both imagination and technical
knowledge and therefore, is inherently both an empirical and a creative process. This
is equally true of spatial music, except in this case, it is not sufficient that the
composer just knows how to program the relevant software or hardware. As space is
so much broader and harder to define than other musical parameters, the composer
must also consider how the often illusory nature of this attribute will be perceived by
the audience. The first half of this thesis therefore focuses exclusively on the
perception of spatial audio and particularly on the weaknesses and limitations of the
techniques involved. Although much of this discussion may appear at first to be quite
negative, it can also be viewed in a much more positive light. The first act of any
composition is the rejection of certain instruments or aesthetics in favour of the
instrumentation or style chosen for this particular piece. In the same way, a clear
understanding of the limitations and weaknesses of each spatialization technique will
provide clear guidance as to the most appropriate technique for a particular piece of
spatial music, or indeed a particular type of movement within the piece. Rather than
being a negative finding, the discovery of the limitations of these techniques may in
fact provide some much needed direction for a composer of spatial music who must
choose between a myriad different techniques and applications. In addition, if we can
consider this empirical research as form of orchestration guide for spatial music, then
statements about the limitations of a particular technique are no more negative than a
statement that a violin cannot produce a pitch lower than G3!
The second half of this thesis concentrates on the artistic use of space and the
aesthetics of spatial music, however, this discussion is now informed by the results of
the empirical analysis conducted earlier. Various works of spatial music are assessed,
not in terms of their artistic validity, but in terms of the perceptual effectiveness of the
use of space as a musical parameter. The discussion is therefore primarily concerned
with examining whether the intentions of the composer in terms of the use of space
were realistic or achievable, and not with the artistic merit of the overall musical
aesthetic. In addition, a number of original compositions by the author are also

presented. Some of these works, particularly early pieces such as Discordianism, are
good illustrations of the practical and technical problems common to works of spatial
music and which are examined in detail earlier in the thesis. However, later works
such as Auto Harp and Rise hopefully illustrate the benefits of this research and a
more positive use of the empirical examination of spatialization techniques. The
algorithmic control of space using flocking algorithms represents an even tighter
merging of empirical and creative processes. In this case, it is the algorithm and not
the composer which dictates the movement of sound through space and works such as
Flock and Rise use this relatively new technique to explore the implications of this
ceding of control for the compositional process.

1.2 The Research Question

The way in which space is used as a musical parameter is influenced greatly
by a number of different, yet inter-related factors. The choice of medium, the means
of performance, the choice of material, and the overall musical aesthetic are
intertwined, and choices made in one area influence possible choices in other areas.
One significant choice is based on technological factors, namely whether to use a
collection of different pairs of loudspeakers, or a regular, symmetrical array of
matched loudspeakers. The use of a collection of pairs of different loudspeakers, a
loudspeaker orchestra, is primarily associated with the style of acousmatic music,
which has been extensively developed and refined through the years, with a particular
amount of work taking place in centres in France, the U. K. and Canada. Composers
such as Smalley, Emmerson, Wishart, Barrett et al have examined the use of space in
this aesthetic, often from the point of view of music analysis. As space is only
revealed by the temporal development of the sounds themselves, this analysis has
often focussed initially on the nature of the sound object itself, the intrinsic or
extrinsic connotations of the sound, before then moving onto the conceptual
spatialization and then finally the practical means of spatialization. The relegation of
the practical means of spatialization to the end of the discussion is understandable, as
the performance, in this case live diffusion to a loudspeaker orchestra, is highly
focussed on adapting the work to the particular venue, and preserving the stereo
image for the entire audience.

An alternate approach based on multiple tracks of audio played through a

symmetrical array of loudspeakers has its origins in a different aesthetic. This
approach allows for the creation of much more sophisticated spatial effects than can
be achieved with manual diffusion to a disparate orchestra of loudspeakers. However,
this approach often neglects to examine how these abstract spatial designs are
perceived by the audience. This problem arises from the significant differences
between the perception of a single listener, such as the composer in the studio, and a
distributed audience in a performance setting, and is often further exacerbated by a
lack of intervention (such as that of a diffusionist) during the performance. In
general, it is difficult to adapt these works for the acoustic or technical setup of a
particular performance space. The question which this thesis therefore attempts to
answer is, what are the limitations, strengths and weaknesses of the most commonly
used spatialization techniques, and what does this imply for the performance and
composition of spatial music?

1.3 Aims and Objectives

The main aims of this thesis are to examine the perceptual effectiveness of
various works of spatial music in terms of the technical means of spatialization, and
also the compositional approach to the use of space as a musical parameter. Particular
attention will be paid to the effectiveness of different spatialization techniques in a
performance context, and what this implies for compositional strategies which use
space as a musical parameter. In this way, the thesis may function as a sort of guide
to spatial orchestration, which covers both the technical operation of different
spatialization techniques and how this relates to the different aesthetics of spatial
music.

1.4 Methodology
This thesis begins with an examination of the perceptual mechanisms related
to spatial hearing and a scientific evaluation of the perceptual capabilities of the most
commonly used spatialization schemes, namely stereophony, Ambisonics and
wavefield synthesis (WFS). The perceptual performance of these systems under the
less than ideal conditions typically found in a performance is examined in detail

through a series of listening tests carried out by the author. The results of these tests
are then incorporated into a meta-analysis of the existing research in this area which
summarizes the results of a large number of other listening tests and simulations. The
conclusions drawn from this meta-analysis are then used to assess the validity of the
various spatial strategies adopted by composers of spatial music such as Charles Ives,
Karlheinz Stockhausen, Iannis Xenakis, Denis Smalley and Pierre Boulez. Finally,
this research was utilized in the composition of a number of original works of spatial
music, the development of a spatial music instrument and an implementation of the
Boids flocking algorithm for spatial music composition for the Csound synthesis
language.
This particular methodology was adopted so as to ensure that the real technical
and perceptual limitations of the practical means of spatialization are fully considered.
This emphasis on the limitations of these systems is perhaps somewhat negative, yet
if space is to be used effectively as a musical parameter then these practical issues
must be fully appreciated. This is particularly true in the case of spatial music
performances as while a particular technique may be effective for a single listener, it
may be much less effective for a distributed audience.

1.5 Motivation
The use of space in electroacoustic music composition and performance is
perhaps one of the most unique aspects of this artistic movement. However, the
dissemination of this music via fixed media and domestic audio technology is still a
significant challenge. This problem has been further exacerbated by the drastic
decline in sales of fixed media music and the increase in online distribution. Yet
despite these difficulties, the publics appetite for live musical performances is
undiminished, and in fact has significantly expanded over the last decade. The
musical and social experience of attending a concert is fundamentally different from
listening to music on an mp3 player, on the internet, or at home, and this is
particularly true of live performances of spatial electroacoustic music. The
experience of listening to music performed with a large and carefully configured
loudspeaker array or loudspeaker orchestra provides a unique selling point and an
experience which cannot be easily reproduced elsewhere. This aspect of
electroacoustic music is the primary motivation for this thesis, which concentrates
9

exclusively on the performance context. Binaural technology and other techniques

specifically for the delivery of spatial audio to a single listener will therefore not be
considered.

1.6 Outline
This thesis is broadly divided into two parts. The first part deals with auditory
perception and the various perceptual mechanisms related to directional hearing. This
provides a perceptual basis for an assessment of various spatialization techniques in
terms of their perceptual performance in a performance context. The second half of
this thesis examines the use of space as a musical parameter through the analysis of
various works of spatial music. In each case, the compositional and spatialization
strategy is assessed in terms of its perceptual effectiveness, based upon the findings
presented in the first half of this thesis.
Chapter Two summarizes the perceptual cues associated with spatial hearing.
The perceptual mechanisms which allow a listener to determine the direction and
distance of a source signal are presented and the effect of acoustic reflections and
reverberance on spatial hearing are discussed. Finally the perception of multiple,
concurrent sources is discussed in terms of Bregmans theory of auditory scene
analysis (ASA).
Chapter Three examines the technique of stereophony which for the first time
allowed for the creation of virtual sources that are not fixed at the physical location of
a loudspeaker. The development of this technique is presented from its conception in
the 1930s, to modern, multichannel formats such as 5.1. The perceptual basis of the
stereophonic principle is also investigated along with various theories of stereophonic
localization.
Chapter Four introduces more recently developed spatialization techniques
such as Ambisonics and wavefield synthesis (WFS). The development of
Ambisonics from Alan Blumleins work with coincident microphone techniques is
presented and the perceptual optimization of ambisonic decoders is discussed. The
final part of this chapter addresses the theoretical background of the new technique of
WFS and some of the practical issues related to this technically demanding method
are discussed.

10

Chapter Five investigates different approaches to the simulation of distance

and the dynamic motion of virtual sources. Well known algorithms by John
Chowning and F. R. Moore are discussed and compared with the perceptual
mechanisms involved in the localization of a real source, which were discussed earlier
in Chapter Two.
Chapter Six analyses the results of a wide variety of tests conducted with the
spatialization techniques presented in the preceding two chapters. Various
stereophonic systems are first evaluated in terms of localization accuracy and the
perception of dynamically moving stereophonic sources. Various ambisonic decoding
schemes are then evaluated, particularly in terms of their performance at off-centre
listener positions. A number of tests which compare and contrast stereophony and
Ambisoncs are then discussed in terms of the perceptual differences between these
two techniques. The effect of acoustic reflections and reverberance on source
localization is then examined and the results of a series of listening tests conducted by
the author are presented. Finally, the results of a number of listening tests carried out
with WFS systems are presented and the perceptual performance of this technique is
assessed.
Chapter Seven is the opening chapter of the second half of this thesis which
focuses on spatial music composition and aesthetics. The history and development of
European antiphonal choral music is discussed, and works of acoustic spatial music
by Charles Ives and Henry Brant are analysed. The use of spatial distribution to
increase the intelligibility of different independent layers of material is also discussed.
Chapter Eight charts the development of electronic music in mid-twentieth
century. Various landmark works by Karlheinz Stockausen, Iannis Xenakis and
Denis Smalley are analyzed along with two original compositions by the author. The
legacy of Musique Concrte and Elektronische Musik is assessed in terms of their
effect on the development of electroacoustic spatial music. Abstract compositional
systems such as serialism are assessed in terms of their perceptual effect and the
performance practice of diffusion is discussed. Finally, Smalleys theory of
spectromorphology and the gestural use of space is presented and its use as a
compositional structuring principle is assessed.
Chapter Nine focuses specifically on mixed media electroacoustic music, i.e.
music for live instruments and spatial electronic sound. Landmark works by

11

Stockhausen and Boulez are examined in terms of their technical and artistic approach
and the specific difficulties associated with this form of spatial music are discussed.
In Chapter Ten, a number of original works of acoustic, electronic and mixedmedia spatial music are presented and analyzed. An original work of choral spatial
music is discussed in terms of the spatial choral music discussed previously in
Chapter Seven. Two works of electronic spatial music are presented which illustrate
the divergent approaches to electronic music composition discussed in Chapter Eight.
Finally an original mixed-media composition by the author is analyzed and the
spatialization approach adopted for this work is evaluated.
Chapter Eleven examines various musical instruments which can be used for
the live performance of spatial music. Various augmented instruments such as the
hypercello are introduced, and the use of the hexaphonic guitar as a spatial music
instrument is discussed. Finally the technical implementation of a hexaphonic system
is presented along with an original electroacoustic composition for the hexaphonic
guitar.
Chapter Twelve discusses the use of flocking algorithms such as Boids for
sound spatialization and synthesis. Real-time and off-line applications are evaluated
along with two original compositions which illustrate these different approaches to
the use of flocking algorithms in electroacoustic composition.

12

2 Spatial Hearing
Traditional musical parameters such as pitch, rhythm and timbre are perceived
with a relatively high degree of accuracy. Various studies have shown that a change
in pitch of a fraction of a semitone is quite perceptible and our ability to temporally
segregate an audio signal is similarly precise. The cross-modal perception of spatial
locations is also relatively accurate but is reduced significantly when the visual
element is removed. Parameters such as pitch, timbre and rhythm are often directly
related to the physical structure of the instrument or the actions of the musician.
However, the use of space in music often relies on electronic processes which can
only simulate the effect of spatial movement. The ability of these processes to satisfy
the various perceptual mechanisms involved would appear to be crucial if this aspect
of the work is to be successful. However, before these spatialization processes can be
assessed it is first necessary to understand the various perceptual mechanisms
involved in normal spatial hearing. By necessity, the mechanisms which allow the
location of a real sound to be determined must be first understood, before the illusion
of sounds moving in space can be created.
In this thesis, spatial locations will be described using the spherical coordinate
system illustrated in Figure 2.1 [Blauert, 1997]. The azimuth angle indicates the
angular position in the horizontal plane (with zero degrees being straight ahead), the
elevation indicates the vertical angle of incidence in the median plane, and the
distance is measured from a point at the centre of the head, directly to the source.

Fig. 2.1 Spherical coordinate system

13

The study of spatial hearing can be broadly divided into two categories, which
typically focus on either certain characteristics of the source signal or of the acoustic
environment. A typical audio scene may contain sources that are relatively discrete
and localizable (localization refers to the perceived direction and distance of the
source signal), however, the same scene may also contain reflections which are
diffuse and not easily localizable and provide information about the acoustic
environment. This chapter will begin with the more straightforward scenario of a
single source under free-field conditions. This implies that the contribution of
reflections of the source signal from nearby surfaces is negligible and that the
influence of the acoustic environment can be ignored. While this rarely occurs under
normal circumstances, outdoor locations such as an open field or mountain top can be
considered as approximately equivalent to free-field conditions.

2.1 Directional Hearing

Under ideal conditions it has been shown that the region of most precise spatial
hearing lies in the forward direction with frontal hearing having a localization
accuracy of between 4.40 and 100 for most signal types [Blauert, 1997]. Accuracy
decreases as the source azimuth moves to the sides, with the localization blur at 900
being between three to ten times its value in the forward direction. For sources to the
rear of the listener, localization blur improves somewhat but is still approximately
twice that for frontal sources.
It is reasonable to assume that the following three factors must influence to some
extent our ability to localize the position of a sounding object;

The audio signal produced by the source

The body, head and ears of the listener

The acoustic environment containing both the source and the listener

It is also reasonable to assume that these three factors must interact in some fashion to
produce an impression in the listener of the spatial location of the source signal.
Consider now a simple example of a laterally displaced sound source
positioned in front of a single listener as shown in Figure 2.2. For now the discussion
is limited to the horizontal plane and free-field conditions. As the signal produced by
the source spreads out and progresses toward the listener the wavefront will first
arrive at the right ear, before then diffracting around the head to reach the left ear.

14

This ability to hear binaurally, i.e. with two ears, is an important aspect of the human
auditory system. Localization mechanisms are often distinguished as being either
interaural, i.e. related to the differences between the signals at the two ears, or
monaural which is related to attributes of the signal that are perceived equally with
both ears.

Fig. 2.2 Lateral source example

The preceding example illustrates how a laterally displaced source will result
in a time delay between the two signals arriving at the ears. This interaural time delay
(ITD) is one of the principal cues used to determine the source azimuth and an
extensive amount of experimental work has been carried out to examine its influence
on source localization. Various experiments have shown that both the spectral
content of the source signal and the attack portion of the signal envelope can produce
ITD cues, depending on the context [Blauert, 1997].
If the source in the preceding example consisted of a continuous signal then
the interaural delay will result in a phase shift between the two ear signals. This
localization cue will be referred to here as an interaural phase shift (IPD) to
distinguish it from the more general ITD cue. It should be noted that this phase shift
can only be determined by comparing the relative phase of the two ear signals. As the
frequency of the source signal increases it eventually produces a phase shift between
the two signals that is greater than 1800. At this point the IPD cue becomes a less

15

reliable indicator of azimuth as it is impossible to determine which signal is leading

and which is lagging. So clearly the IPD cue is frequency dependent and in practice
has been found to become unreliable at about 700-800Hz, and completely ineffective
above 1.6 kHz [Blauert, 1997].
ITD cues can also be derived from the attack portion of the wavefront at the
two ears even when phase differences are inconclusive. Experiments have shown that
the lowest frequency at which the attack portion of the signal provides useful
estimates of azimuth is proportional to the steepness of the attack [Elfner et al, 1968].
Generally the ITD cue is the more prominent temporal cue above 1.6kHz while the
IPD is more prominent at lower frequencies [Blauert, 1997].
The preceding example illustrated how a laterally displaced source arrives first
at the closest ear and then diffracts around the head to reach the other ear. The head
shadowing which arises as a result of this diffraction results in an amplitude
difference between the two ear signals which is also related to the azimuth of the
source. This localization cue is commonly referred to as the interaural level
difference (ILD). A signal whose wavelength is comparable to the size of an obstacle
will diffract easily around the obstacle. Therefore, when the frequency of the source
signal in our example is low enough so that its wavelength is comparable to the
spacing of the ears, little head shadowing will occur. Calculating this frequency using
an average head diameter of 180mm and a speed of sound in air of 340m/s results in a
figure of 1.89kHz [Wiggens, 2004]. Experimental results indicate that ILD values
change smoothly with increasing frequency and that the more high frequency content
in the signal, the greater the average level difference. In general it has been shown
that there is a strong weighting for ILDs with high frequency signals and poor
weighting of ILDs with low frequency signals. The converse is true for the ITD. For
wideband stimuli, the ITD has been shown to be the dominant localization cue
[MacPherson, 2002]. Neither cue is particularly strong in the region of 2kHz.
The preceding discussion describes the interaural localization cues which
provide information on the likely azimuth of the source. However, the example
shown in Figure 2.3 demonstrates that some other process must also be involved.
This example contains the same source as before with a duplicate source positioned
symmetrically to the rear. Clearly in this case, both sources would create exactly the
same ITD, IPD and ILD cues and that this information would not be sufficient to
resolve the two potential source directions. This cone-of-confusion [Begault, 1994] is
16

thought to be resolved using both head movements and monaural localization cues
created by the filtering effect of the shoulders and ears.

Fig. 2.3 Cone-of-confusion example

The complex shape of the ear pinnae acts as an acoustic filter which
introduces time and level differences between the individual spectral components of
each ear input signal. As the shape of the upper body and pinnae is irregular and
highly idiosyncratic, this filtering effect would change relative to the spatial location
of the source signal. The total filtering effect of the pinnae, head and shoulders is
often measured and characterized as the Head Related Transfer Function (HRTF). As
the ear pinnae are orientated toward the front, the two source signals in the preceding
example would be filtered in different ways, allowing the listener to distinguish
between the two source locations. It is still not entirely clear as to how the brain
distinguishes between the spectral content of the source signal and spectral changes
introduced by the HRTF. This could be achieved using slight movements of the head,
as this would introduce changes in the HRTF independently of the source signal and,
indeed, numerous experiments have shown that head movements allow listeners to
determine whether a source is positioned in front or behind or above or below
[Blauert, 1997; Spikofski, 2001]. Rotating the head toward the source will alter the
various interaural cues in such as way as to resolve the cone of confusion and the ITD
and ILD will decrease as the head turns towards a source positioned in front but

17

increase for a source to the rear. In the same way, head movements will result in a
change in the filtering effect of the head and pinnae which could also help to resolve
the cone of confusion. Of course, head movements can only be effective if the signal
duration is long enough to allow for their effect on the different cues to be evaluated.
In a number of experiments with narrowband sinusoidal signals, certain
frequencies were found to correlate with specific angles of elevation, independently of
the actual source position [Blauert, 1997]. This suggests that HRTF cues not only
help to resolve the cone of confusion in the horizontal plane, but also contribute to the
localization of elevated sources and, in fact, HRTF cues have been shown to be
crucial in determining the position of a source anywhere in the median plane. Studies
suggest that spectral cues in both low (below 2kHz) and high frequency (above 5kHz)
regions are used to resolve front-back confusion while the prominent spectral cues for
the judgement of elevation are derived solely from the high frequency components
(above 5kHz) [Asano et al, 1990].

Fig. 2.4 Impulse response of a room with a relatively short reverb time

The preceding discussion on directional hearing has focussed on free field

conditions where the effect of the acoustical environment is neglected. However, an
environment such as this is quite different from the acoustic spaces in which most

18

music is heard. In an enclosed space, a sounding object will produce a wavefront that
will travel directly to the ears and also indirectly via multiple reflections from the
floor, walls and ceiling of the space. This indirect signal is generally divided into
early reflections which arrive relatively shortly after the direct sound, and the later
arriving, more diffuse reflections or reverberance (see Figure 2.4 [Wiggens, 2004]).
Despite the presence of these additional wavefronts arriving from multiple different
directions it is still quite possible to accurately localize a sound in an enclosed space.
[Bates et al, 2007a] The law of the first wavefront, or precedence effect is thought to
account for this ability and as shall be seen later, is an important consideration in
multichannel loudspeaker systems.

2.2 Directional Hearing and Acoustics

As has been seen, in a reverberant environment, sounds reach the ears via
several different paths. Although the direct sound is followed by multiple reflections
which would be audible in isolation, the first-arriving wavefront dominates many
aspects of perception, including source localization [Litovsky et al, 1999]. This
theory, known as the precedence effect, or the law of the first wavefront, states that
when multiple coherent signals are presented at the ears, localization is primarily
based on the earliest arriving signal. One study [Begault, 1994] showed that for
delays up to approximately 0.6ms the source moved laterally toward the non-delayed
side of the listeners head. For delays greater than approximately 0.6ms and less than
35ms the source position remains unchanged but some timbral coloration may occur.
Even greater delay times result in two distinct sources as the delayed signal is
perceived as a distinct echo. It should be noted that the precise time values at which
these perceptual effects occur is highly dependent on the nature of the source signal
[Litovsky et al, 1999]. Other studies have also shown that the presence of significant
early reflections can have an effect on localization accuracy [Hartman et al, 1985].
Lateral reflections from side walls were found to be particularly detrimental for
azimuthal localization while the effect of early reflections from the floor and ceiling
was less conclusive. Hartmann speculates that as the angular direction of these
reflections matched that of the source signal, reflections from the floor and ceiling
may reinforce horizontal localization [Hartmann et al, 1985]. In general, it has been
found that signals with strong transient characteristics are localized independently of

19

the room reverberation time, but may depend on the specific room geometry which
can result in significant early reflections [Hartmann, 1983].
Early reflections and reverberation also has a significant effect on both the
perceived size of the source, and the spatial impression of the acoustic environment.
The term spaciousness has been used to refer to both these effects, and other terms
such as spatial impression, ambiance, apparent source width, immersion and
envelopment are also frequently used, sometimes interchangeably, and sometimes
with more specific definitions. A strong correlation has been found between the
degree of coherence between the two ear signals and the lateral component of the ear
signals, and this is though to influence the spatial impression of both the source and
environment [Blauert, 1997; Plenge et al, 1975]. Early studies of spaciousness
described this characteristic in terms of the lateral and frontal components of the ear
signals [Barron et al, 1981]. However, what is meant by spaciousness is different
depending on whether this lateral component is derived from early reflections alone,
or from both early reflections and later reverberation. The addition of lateral early
reflections results in a change in the perceived size of the source and the apparent
source width (ASW) is generally used to refer to this source-specific measure
[Beranek, 1996]. Early reflected energy arriving within approximately 80 ms of the
direct sound results in an increased ASW and the extent of this broadening effect
depends upon the ratio of the total energy to the energy of the lateral component
[Barron et al, 1981]. Blauert introduced the term locatedness as a measure of the
degree to which an auditory event can be said to be clearly in a particular location
[Blauert, 1997] and this is clearly related to ASW. Localization refers only to the
perceived direction and does not therefore directly relate to the locatedness or ASW,
although clearly the source direction will be difficult to precisely determine in the
case of a very large ASW. The important distinction between these two measures will
be discussed in more detail in Chapter Six.
Later arriving reverberation alters the spatial impression in a different way
which is primarily related to the acoustic environment. The term spaciousness is
often used in this case, as opposed to ASW which is primarily related to the spatial
impression of the source. The terms spaciousness and envelopment are often
considered equivalent, although occasionally slightly different definitions are given.
When distinguished, spaciousness is described as the sense of open space in which the

20

source is located while envelopment refers to the sense of immersion and involvement
in a reverberant sound field which fully surrounds the listener [Rumsey, 2001].
Numerous studies have shown that envelopment and spaciousness are
generally desirable qualities in a concert hall [Schroeder et al, 1974; Barron et al,
1981; Beranek, 1996] and many concert hall designs have attempted to increase the
amount of lateral reflected energy directed to the seating area for this reason. The
acoustician David Griesinger has suggested that the importance of the distinction
between early arriving lateral reflections and later reverberation is often overlooked in
this context [Griesinger, 2009]. It has already been shown that early arriving
reflections reduce localization accuracy and Griesinger suggests that an increase in
lateral energy of this sort will negatively impact clarity. This clearly contradicts
previous studies which have stressed the importance of ASW. Griesinger suggests
that the direct sound and later lateral reverberation should be emphasized to improve
clarity (which corresponds to improved localization) and also spaciousness (meaning
the desirable spatial characteristics of the hall).
Rumsey has similarly pointed out the difference between acoustic research,
which suggests that large values of ASW are preferable, and studies of spatialization
techniques which emphasize localization accuracy [Rumsey, 1998]. Griesinger has
noted a similar contradiction between the differing levels of reverberation typically
found in performances and recordings of classical music [Griesinger, 2009]. A clear
distinction can be made between the three-dimensional spatial experience of a live
instrumental performance in a concert hall and a two-channel stereo reproduction
which must reproduce both the direct and reflected sound from the front. Improved
localization accuracy is perhaps desirable in the latter case in order to distinguish
sources in the foreground from background reverberation [Rumsey, 1998]. However,
the situation is much more complicated in the case of electronic spatial music
performances which often utilize multi-channel loudspeaker arrays and electronic
spatialization techniques within a concert hall. The distinction between ASW,
spaciousness and envelopment introduced earlier may also be hard to maintain in this
situation, as a multichannel reproduction of a single source from multiple positions
around the audience will provide some sense of envelopment, but by the direct sound
and not the reverberant field. In this situation, the sense of spaciousness may be
associated with the source rather than the acoustic environment, and the distinction

21

between these terms becomes harder to define. In addition, it is hard to predict the
effect of a lateral source on perceptual aspects such as ASW or spaciousness.
The preceding discussion indicates the significant effect of acoustic reflections
on the perception of an auditory source. In addition, reverberation provides a
significant amount of information regarding the size and composition of the
environment within which the source is situated. The degree of attenuation of
acoustic reflections provides an indication of the nature of the reflecting surfaces
while the time and duration of the diffuse late reverberation can indicate the
dimensions of the space. Room reflections and reverberation can also provide
information on another highly important aspect of spatial hearing, namely, the
distance of the source from the listener and this will be examined in more detail in the
next section

2.3 Distance Hearing & Moving Sources

The most obvious clue to the distance of a sounding object is its amplitude. It
has been shown that for a constant point source in a free sound field, the sound
pressure level falls of by six decibels with every doubling of the distance [Blauert,
1997]. However, amplitude can only serve as an absolute measure of distance if the
listener is familiar with the source signal as the amplitude of an unfamiliar or
synthesized sound can only provide a relative estimate of a change in distance
[Mershon, 1997; Sheeline, 1983]. For example, it is easy to determine whether
someone is speaking quietly nearby or shouting loudly far away because of the
familiar and characteristic dynamics of the human voice, but when the source signal is
unfamiliar this distinction cannot be made. Various studies under free-field
conditions have found that for a source with constant amplitude, there is no
relationship between the actual and perceived distance [Nielsen, 1993; Mershon et al,
1975], indicating that, in this case, the signal amplitude can only indicate how the
relative distance is changing. For very close sources (< 1m), it appears that the
auditory system also uses additional binaural localization cues to estimate the source
distance. Brungart analysed a number of HRTF measurements which showed that
very near sources result in substantial changes in ILD [Brungart et al, 1999], although
the ITD remains largely unchanged.

22

It has long been known that the perception of distance is also influenced by the
effect of acoustic reflections in the listening environment. The level of the direct and
early reflected sound will change substantially as the distance from the source to the
listener changes. However, the level of diffuse reverberation is largely independent of
the position of the listener in the room. Therefore, as the source distance increases,
the direct sound will decrease while the reverberant sound remains constant. Beyond
a certain distance, the reverberant signal level will be greater than the direct signal
level, and the perceived distance becomes fixed and independent of the actual source
distance. This critical distance is indicated in Figure 2.5 [Howard et al, 1996].
Various studies have shown that in reverberant rooms, the perceived distance of a real
source is independent of the source level [Nielsen, 1993], which suggests that the
ratio between the direct and reverberant signals, the D/R ratio, is a significant distance
cue in real rooms. This theory was first proposed in the 1960s and this simple ratio is
still commonly used by sound engineers and producers to control the depth of
different sources in two-channel stereo mixes.

Fig. 2.5 Source distance v sound intensity

The D/R ratio can provide a relative sense of distance but this simple ratio
ignores the fine spatial and temporal structure of the reflected indirect signals.
Experiments by Kendall reported that a strong impression of distance was perceived
when listening to dry test signals augmented solely with a limited number of artificial
early reflections, even when these reflections were restricted to those which followed
the direct signal by 33ms or less [Kendall et al, 1984]. Michelsen carried out a
23

similar test which also found that a better distinction of distance was achieved when
simulated early reflections were added instead of solely diffuse reverberation
[Michelsen et al, 1997]. Neher investigated the perceptual effect of different early
reflection patterns and found that listeners were unable to distinguish between an
early reflection pattern comprised of accurately panned reflections, and one that was
physically identical except that each reflection was simply reproduced by the nearest
available loudspeaker. This suggests that although spatial differences in the early
reflections pattern are perceptually salient, the actual angles of incidence of
reflections may not be crucial [Neher, 2004].
Michael Gerzon presented a similar model of distance hearing, based on a
theory originally proposed by Peter Craven [Gerzon, 1992b]. The Craven hypothesis
assumes that the apparent distance of sounds is derived from the relationship between
the relative time delay and amplitude of the early reflections and the direct signal.
Gerzon and others have suggested that closely-spaced or coincident microphone
techniques having a substantially omnidirectional total energy response will
reproduce the absolute source distance better than microphones with a more
directional response [Gerzon, 1992b; Theile, 1991]. Gerzon also points out that
although it is now known that the simple direct/reverberant ratio does not provide an
absolute measure of distance, it is still a useful subsidiary cue for relative distance,
and is thus preferably made consistent with the apparent distance [Gerzon, 1992b].
In general it has been found that the perceived distance of a sound source in a
room is compressed, as it increases virtually linearly with source distance at short
range, but converges to a certain limit when the source distance is increased beyond
the critical distance [Mershon et al, 1975; Nielsen, 1993]. There is therefore a nonlinear relationship between the perceived and actual source distance. Bronkhurst
suggests that this non-linearity arises not only due to the different mechanisms
involved for different source distances, but also because the auditory system is not
always able to accurately separate the direct and reflected signals [Bronkhurst, 2002].
In a number of listening tests, the effect of early reflections on perceived source
distance was assessed in terms of the angle of incidence of the reflected sound. When
the lateral walls were made completely absorbent, the source was perceived to be
close to the head, virtually independently of the actual source distance. When lateral
reflections were introduced, the perceived distance more closely matched the actual
source distance, although larger distances were still underestimated [Bronkhurst,
24

2002]. This suggests that the direct to reverberant ratio is estimated by the auditory
system using directional binaural cues to separate the direct and reverberant signals,
although further tests are needed to confirm the validity of this hypothesis.
While clearly the amplitude of the source signal and the specific relationship
between the direct and indirect signals have been shown to be dominant cues in
distance perception, other secondary cues also provide some indication of relative
distance. In general, an increase in source distance results in a reduction of the high
frequency spectral content of the source signal due to the effect of air absorption.
This occurs at large distances outdoors, but also in rooms due the absorptive nature of
the boundary surfaces and the large overall distances travelled by the indirect signals
as they reflect around the room.
The perceived shift in frequency due to the movement of a source relative to
the listener, i.e. the Doppler effect, also provides an indication of the relative motion
of the source.

2.3.1 Summary of Spatial Hearing

The preceding section summarized the different localization mechanisms
which are used by the auditory system to determine the direction and distance of a
source signal. It is still unclear, however, as to how these multiple, different and
potentially conflicting localization cues are resolved into a single distinct spatial
impression. Most theories of auditory localization now propose that the perceived
source location is the one that satisfies as many of the localization cues as possible.
When multiple conflicting localization cues are present, a simple majority decision is
used to determine the location of the sounding object. Most real sources produce a
complex signal which will produce localization cues that support each other in the
frequency ranges at which they dominate. In addition, the corresponding visual
component of a real source will also support the direction suggested by the auditory
senses. Broadband signals are able to satisfy more localization mechanisms than
narrowband signals and are therefore, in general, easier to locate. In addition the
presence of onset transients greatly increases localization accuracy. Narrowband
signals such as sine tones, particularly in the 2kHz region, are difficult to localize
even under ideal conditions as no localization mechanism is particularly effective in
this region. Early reflections and reverberation can also have a significant effect on

25

the perceived size of the source, which can result in a corresponding decrease in
localization accuracy.

2.4 Spatial Hearing with Multiple Sources

In a complex auditory environment, sounds generated by numerous different
sources at different spatial locations all arrive together at the ears. Yet, most listeners
can readily segregate the audio scene into its multiple distinct components. This
process is often referred to as the cocktail party effect, where a listener can
consciously focus their attention on any one of a number of overlapping and
concurrent conversations. Albert Bregman attempted to explain this phenomenon
with the theory of auditory scene analysis (ASA) which was informed by Gestalt
psychology and describes this process whereby an audio scene is segregated into
multiple distinct streams [Bregman, 1990]. Auditory scene analysis describes the
relationship between the different elements in an auditory scene in terms of the
Gestalt principles of perceptual grouping, namely;

The principle of proximity: elements which are positioned close together in

space or time were probably generated by the same event

The principle of similarity: sounds with a similar timbre or frequency probably

belong to the same event

The principle of good continuation and completion: sounds generated by the

same event tend to be continuous and follow each other

The principle of common fate: sounds with similar frequency, dynamic or

rhythmic trajectories probably originated from the same event.

These grouping principles were described by Bregman as primitive segregation, as

these perceptual processes are involuntary and innate rather than learned. More
refined musical knowledge, and prior experience with different sounds is structured in
units referred to as schemas and these are voluntarily employed when attention is
being paid to a specific sound or piece of music. Schemas therefore operate over a
longer time-scale than primitive segregation as it involves prior experiences and also
expectations about the sounds to come. It would therefore appear that auditory
perception depends upon two concurrent perceptual processes in two overlapping
timescales. Primitive processes firstly perform a quick interpretation of the composite

26

ear signals, dividing it into separate streams, while at the same time longer schema
processes interpret each stream as it changes in terms of prior experience and
expectation of its future state. Interestingly, a similar two stage process is thought to
operate for the visual senses, as first uncovered in the 1959 paper "Receptive fields of
single neurons in the cat's striate cortex," by Hubel and Wiesel [Hubel et al, 1959].
Further research in machine vision has suggested that when the visual input enters the
brain from the eyes, it is immediately sent through two separate neurological
pathways [Goodale et al, 1992]. The fast path quickly transmits a rough, blurred
outline of the image to the frontal cortex while the second path performs a slower (the
slow image arrives in the prefrontal cortex about 50ms after the fast image) analysis
of the image, using prior experience and knowledge to fill out and refine the rather
crude initial impression.
It has been shown that spatial auditory cues are not a dominant factor in
determining the number of competing sound sources [Bregman, 1990]. However,
other studies have shown that spatial hearing is highly important for the intelligibility
of multiple, simultaneously presented speech signals [Shinn-Cunningham, 2003; Best,
2004] and that our ability to segregate an audio scene into multiple streams strongly
influences our perception of fundamental musical constructs such as melody and
rhythm [Bregman, 1990]. The concept of spatial masking, in which a listeners ability
to detect and understand the content of multiple signals is improved when the signals
are spatially separated is highly important in spatial music. As shall be seen later in
Chapter Seven, composers such as Charles Ives, Henry Brant and Karlheinz
Stockhausen regularly used multiple, spatially separated musicians to reinforce the
inherent polyphony in the musical work. Of course spatial masking can also be used
in the opposite way, to deliberately undermine the individuality of each source in an
attempt to create a single sound mass using multiple musicians positioned together on
stage. Works such as Atmospheres by Gyorgy Ligeti or Metastasis by Iannis Xenakis
are clear examples of orchestral works which deliberately utilise spatial masking in
this way. These two different uses of spatial masking can be further expanded in
purely electronic works, where the various spectral components of a signal can be
heightened through the dynamic spatial separation of the various components. The
electroacoustic composer Natasha Barrett illustrates this process with a number of
audio examples in a paper on spatio-musical composition strategies [Barrett, 1990].
In the first example, eight continuous sounds with a significant degree of temporal
27

and textural overlap are played simultaneously at static points in space while in the
second example each sound is dynamically moved in space. Barrett suggests that
only five distinct sources are perceivable when positioned at static spatial locations
but that this number increases when each source is provided with individual and
dynamic spatial trajectories.

2.4.1 The Limits of Auditory Perception

Although ASA clearly suggests that spatial cues are an important factor in
auditory perception, Bregman does include certain qualifications such as the
frequency dependence of the cocktail effect and the detrimental effect of
reverberation and reflections. Harley developed the table shown in Figure 2.6 which
summarizes the interaction between the various spatial cues and stream segregation
and illustrates the conflicts which may potentially occur [Harley, 1998a].

Fig. 2.6 Spatial cues and stream segregation

The limits of auditory perception can also be exceeded by electronic processes

such as, for example, in the speed of movement of a particular source. Blauert

28

reported that a cyclical lateral movement must take at least 172ms to occur if the
trajectory is to be accurately followed while 233ms is required for a front-rear
movement. Sounds following a circular path around a listener that exceed these limits
will at first appear to oscillate from left and right before stabilising at a central
position at even faster speeds [Blauert, 1997]. This effect was deliberately employed
by the composer Karlheinz Stockhausen in his eight-channel composition Cosmic
Pulses [Sonoloco, 2007]. The technical implementation of this work was carried out
by Joachim Haas and Gregorio Garcia Karman at the Experimental studio in Freiburg
between December 2006 and April 2007. Karman describes the perceptual effect of
the OKTEG (Oktophonic effect generator) system developed for this piece to
implement very high speed rotational effects, as follows;
Like in the Rotationsmhle a device used in the spherical auditorium at the
Worlds Fair in Osaka and later implemented as output stage of the Klangwandler the OKTEG provides the performer with manual control of rotation velocity, and
different routings are accomplished by means of matrix programs. The
Rotationstisch, first used as a spatialization instrument in KONTAKTE, was later
further developed for exploring the artefacts, which appeared at very high rotation
speeds. Following this idea, the OKTEG provides with sample accurate trajectories
and arbitrary high rotation speeds, assisting the exploration of a continuum linking
space and timbre. When sound trajectories get close to the upper velocity range of
16 rot/sec in the composition of COSMIC PULSES, the perception of movement is
gradually transformed into a diffuse and vibrating spatial quality. Higher rotation
frequencies manifest themselves as audible modulation effects. [Sonoloco, 2007]

2.4.2 Spatial Hearing and Virtual Sources

It has been shown that our ability to localize the position of a sounding object is
determined by the interaction of:

The audio signal produced by the source

The body, head and ears of the listener

The acoustic environment containing both the source and the listener

The preceding chapter summarized the effect of these parameters on our perception of
the direction and location of a real sounding object. The composition of spatial music
obviously involves the direct manipulation of the first parameter, however, the latter
two parameters can only be dealt with indirectly, if at all. Some spatial music,
particularly acoustic music for distributed musicians, can be understood relatively
simply in terms of the localization mechanisms discussed in this chapter. However,
most electronic spatial music does not confine itself to multiple independent signals

29

produced by individual loudspeakers. Instead, different electronic processes are used

to position and dynamically move virtual sources around the listening area using
multiple coherent loudspeaker signals.

2.4.3 Spatial Audio Techniques

A great many techniques have been developed over the past century for the
production of spatial audio. However, in general, the approach taken consists of
either:

The manipulation of level and/or time differences in pairs or multiple pairs of

loudspeakers.

The reconstruction of a sound field over a listening area using a loudspeaker

array.

The reconstruction of the ear signals using headphones or highly localized

loudspeaker signals.

The first approach of manipulating either phase/time or more usually level differences
between pairs of loudspeakers is often referred to as stereophony. The production of
ITD and ILD cues through the manipulation of these factors can be achieved both
acoustically through the use of different microphone arrays, and through electronic
processing techniques such as amplitude panning. Stereophony originally referred to
any method of reproducing a sound field using a number of loudspeakers, but is now
generally used to refer specifically to techniques based on the manipulation of level
and/or time differences in pairs or multiple pairs of loudspeakers, such as in twochannel stereo and 5.1 surround sound.
Ambisonics and Wavefield Synthesis are two techniques which attempt to
reconstruct a sound field within a listening area using loudspeaker arrays.
Ambisonics is a complete set of techniques for recording, manipulating and
synthesizing artificial sound fields [Malham et al, 1995] which has been regularly
used in spatial music and theatre for the past three decades. While never a
commercial success, Ambisonics has proved enduringly popular for spatial music
presentations for various reasons, such as its independence from a specific
loudspeaker configuration and its elegant theoretical construction which is based on
spherical harmonics.

30

Wavefield Synthesis (WFS) is a more recently developed technique which can

be considered as an acoustical equivalent to holography, or holophony [Berkhout,
1998]. The technique uses large numbers of loudspeakers arranged in linear arrays
and can theoretically recreate a sound field over a much larger listening area than is
possible with other sound field reconstruction techniques such as Ambisonics.
The third approach uses HRTF data to either record or synthesize spatial
auditory cues. This binaural approach is highly applicable for a single listener as it
requires a strict separation of the two ear signals, such as when listening with
headphones. However, it is much more difficult to extend this approach to large
groups of listeners and so will not be covered in this thesis.
Apart from potentially WFS, these techniques can really only recreate the
directional perceptual cues. The simulation of distance is often achieved using
additional processes which will be discussed later.

31

3 Stereophony

Fig. 3.1 Bell Labs stereophony, proposed (left) and implemented (right)

The earliest work on stereophony was carried out independently by both Bell
Laboratories in the United States and Alan Blumlein at EMI in the UK in the early
nineteen thirties. The approach adopted by Bell labs was based on the concept of an
acoustic curtain [Steinberg et al, 1934], namely that a sound source recorded by a
large number of equally spaced microphones could then be reproduced using a
matching curtain of loudspeakers (Figure 3.1 left). In theory, the source wavefront is
sampled by the microphone array and then reconstructed using the loudspeaker array.
In practice, this approach had to use a reduced number of channels, so a system was
developed using three matching spaced omni-directional microphones and three
loudspeakers placed in a front-left, centre and front-right arrangement (Figure 3.1b).
This approach was problematic however, as the reduction in channels distorted the
wavefront and audible echoes sometimes occurred due to the phenomenon of spatial
aliasing (see Section 3.2.2). Spaced microphone techniques such as this capture the
different onset arrival times of high frequency transients, and so capture the ITD
localization cues present in the original signal. However, this also makes it difficult
to process the audio afterward as unpredictable time differences are fixed in the
recording.

32

Fig. 3.2 Blumleins coincident microphone arrangement

At the same time, Alan Blumlein was developing various alternative

arrangements, such as the two coincident microphones with figure-of-8 directivity
patterns shown in Figure 3.2 [Wiggens, 2004]. This coincident microphone
arrangement records level differences which vary with the angular position of the
source, but as the microphones are coincident, time differences are not captured.
However, Blumlein realized that the resulting level differences would in fact result in
an IPD at low frequencies due to the unavoidable cross-talk between the
loudspeakers. To illustrate this, consider two sources radiating a low frequency signal
with no time difference, but with a greater amplitude signal radiating from the left
loudspeaker (Figure 3.3). The listener will receive at his left ear the louder signal
from the left loudspeaker, combined with the quieter signal from the right
loudspeaker, which is now delayed due to the greater distance travelled. The sum of
these two wavefronts will be a phase-shifted and amplified version of the louder
wavefront. A similar and inverse summing process occurs at the right ear and
Blumlein realized that the resulting difference in phase between the two ear signals
will produce an interaural time cue at low frequencies that is proportional to the
amplitude difference between the loudspeaker signals. In turn, at higher frequency
ranges, head-shadowing acts as a greater obstacle to the two wavefronts, so the
amplitude differences produce an effective ILD cue. Stereophony, therefore, to some
33

extent resembles natural hearing, as it produces IPD and ILD cues in the frequency
ranges at which these localization cues are most effective. It therefore uses the
unavoidable cross-talk between the loudspeakers as an advantage, as this cross-talk
produces an IPD which is related to the original source direction. Critics of this
approach of summing localization argue that level differences alone cannot produce
the ITD cues necessary for correct localization of onset transients [Thiele, 1980].
However, subjective listening tests have shown that this is not the case and that
transients can be clearly localized in Blumlein stereo recordings [Rumsey, 2001]. In
addition, this approach allows for the post-processing of the stereo image by adjusting
the combination of the two microphone signals. More recently alternative
microphone arrangements such as ORTF or the Decca tree have been developed
which represent a trade-off between the two approaches and reduce the conflicting
ITD cues that arise for transient and steady-state signals with purely coincident
techniques.

Fig. 3.3 Standard stereophonic arrangement

These microphone techniques can of course also be adapted to artificially

position a monophonic recording in a stereo field. The introduction of time
differences to a monophonic signal routed to two loudspeakers can be used to position
or pan the signal between the loudspeakers. This approach, however, can introduce
contradictory phantom image positions for the transient and steady-state parts of the
signal [Martin et al, 1999a], as additional phase differences are introduced by the
summing effect of the loudspeaker cross-talk. In addition, comb filtering can occur
when both channels are summed to a monophonic signal and the perceived source

34

position is highly dependent on the position and orientation of the listener [Rumsey,
2001]. Amplitude panning introduces level differences by simply weighting the
signal routed to each loudspeaker and this technique is quite effective when used with
a symmetrical pair of loudspeakers in front of a single, centrally positioned listener,
with an optimal separation angle of 30o. Amplitude panning can be considered as a
simplification of Blumleins coincident microphone technique shown in Fig. 3.2.
With this arrangement, a signal in the front left quadrant will arrive at the maximum
of the blue microphone response characteristic and at the null point of the red
microphone. Amplitude panning simplifies this idea so that a signal panned hard left
will only be produced by the left-most loudspeaker, and vice versa, while a signal
panned to the centre will be created as a phantom image by both loudspeakers. This
has the result that a slight yet perceptible change in timbre occurs when a signal is
panned from a loudspeaker position to a point in between.
The specific implementation of stereophony for two loudspeakers, i.e. twochannel stereo, is by far the most commonly used audio format in the world today.
However, as this format only utilises a pair of front loudspeakers, it must necessarily
reproduce both the direct source signal and reverberation from the front. One of the
earliest formal extensions of this method to more than two channels is the
Quadraphonic system, which is summarized in the next section.

3.1 Quadraphonic Sound

Quadraphonic sound was one of the earliest attempts to develop a standardized
system for the presentation of spatial audio. Introduced in the early seventies, the
system used four loudspeakers positioned equidistantly and symmetrically at the four
corner points of the listening space. Quadraphonic sound was a commercial failure
for various reasons such as a number of incompatible delivery formats and issues with
multiplexing four discrete channels into two-channel media. However, the
Quadraphonic format continued to be used for performances of spatial music for
many years and in fact, four equally spaced loudspeakers were used in electroacoustic
music long before the development of Quadrophonics as a commercial format.
Karlheinz Stockhausens landmark work Kontakte was written in 1960 for four
symmetrical loudspeakers, which were however, unlike Quadraphonics, placed at the
front, left, right and rear positions.

35

The standard Quadraphonic loudspeaker arrangement exhibits a number of

deficiencies when creating phantom images, even for a single listener. The extended
angular separation of 45o between each loudspeaker pair degrades the stereo image
significantly, as small changes in levels result in large changes in the perceived source
position [Thiele et al, 1976]. In addition, lateral and, to a lesser extent, rear pairs of
loudspeakers have been shown to be incapable of producing a reliable phantom image
as the perceived direction abruptly shifts from one loudspeaker to the next (Figure 3.4
[Ratliff, 1974]). This result is perhaps unsurprising as it has been shown that
stereophony depends upon the crosstalk between the loudspeaker signals at the two
ears to reproduce the correct IPD. In this situation, both loudspeaker signals will
arrive at the nearest ear at the same time and the stereophonic image will be distorted.

Fig. 3.4 Lateral phantom source direction versus ILD for a quadraphonic layout

The decision to adopt a four-channel format, such as Quadraphonics, was

almost certainly influenced by the difficulties in storing more than four channels on
analogue media and the prohibitive cost at that time of large numbers of amplifiers
and loudspeakers. Today, the storage and playback of many channels of audio can be
readily and inexpensively achieved with digital technology and the cost of amplifiers
and loudspeakers has also decreased significantly. These factors, combined with the

36

inherent perceptual defects of Quadraphonics, have resulted in the development of

similar formats which utilise a greater number of loudspeakers.

3.2 Cinema Surround Sound and 5.1

5.1 surround sound is a spatial audio format which is commonly used in
cinemas and also for music and films released for the domestic market on Digital
Versatile Disk (DVD) media. The loudspeaker layout and channel configuration for
5.1 surround is specified in ITU-R BS.775 and shown in Figure 3.5. The frontal bias
of this arrangement indicates its origins in cinema and explains its alternative name of
3-2 stereo. 5.1 is primarily intended to support a frontal visual image while
maintaining backward compatibility with traditional two-channel stereo. Hence, the
front left and right loudspeakers maintain an angular separation of +/- 60o, even
though the presence of the centre loudspeaker could enable the creation of a wider
frontal image. Similarly, the rear/side channels are only intended to be used to
produce reverberation and ambience in support of the primary frontal image. The
final .1 or low frequency effects (LFE) channel is used for sub-bass effects,
although in many systems this loudspeaker is used to produce all the low frequencies
in conjunction with five smaller loudspeakers.

Fig. 3.5 ITU 5.1 loudspeaker arrangement

37

While 5.1 surround can be highly effective when used for its intended purpose
as support for a frontal visual image, it is much less suitable for the presentation of
spatial music to a distributed audience. The front three loudspeakers allow for
accurate frontal images to be created in this direction but the problems with lateral
and rear virtual images discussed in the previous section are magnified. Of course the
commercial attractiveness of 5.1 cannot be understated. The means of production and
delivery are by now well-established and it is unlikely that any new format will
replace it in the near future. Some research has reported good results using 5.1 as a
delivery format for playback with an alternative loudspeaker arrangement and
encoding such as Ambisonics [Gerzon, 1992c; Bamford, 1995; Wiggins, 2004].

3.3 Multichannel Amplitude Panning Techniques

Numerous extensions of the basic stereo amplitude panning method have been
developed which can be utilized for non-standardized loudspeaker arrangements such
as a regular eight-channel array. Vector Base Amplitude Panning (VBAP) is a
popular technique developed by Ville Pulkki [Pulkki, 1997]. This vector-based
reformulation of the amplitude panning method can be used to extend the basic
stereophonic principle to an arbitrary number of loudspeakers. Once the available
loudspeaker layout is defined, virtual sources are positioned by simply specifying the
source azimuth. If a virtual source is panned to the same direction as any of the
loudspeakers, then only that loudspeaker will be used while sources panned to a point
between two loudspeakers will be produced using this loudspeaker pair and the
tangent panning law. VBAP also includes a spread function which ensures that the
same number of loudspeakers is used to produce the image regardless of the specified
position. This is highly advantageous for dynamically moving sources as it reduces
the timbral shift that occurs as the signal moves from a position at a loudspeaker, to a
position between a loudspeaker pair [Pulkki, 1999].
Spat is a real-time modular spatial-sound-processing system developed by
IRCAM and Espaces Nouveaux for the Max MSP environment [Jot, 1999]. The
system allows for the positioning and reverberation of audio sources in three
dimensions using a high level control interface based on a number of perceptual
parameters. The design of Spat is largely based on the spatial processing algorithms
developed by Chowning and Moore in the seventies and eighties [Chowning, 1971;

38

Moore, 1983] which will be discussed later in this Chapter. The supplied output
module can be configured for reproduction using standard two-channel stereo,
discrete intensity panning over various multichannel loudspeaker configurations (up
to a maximum of eight), Ambisonics B-format encoding, or binaural encoding for
reproduction over headphones.

3.4 Theories of Stereophonic Localization

After the development of stereophony, a number of theorists attempted to
develop a more complete mathematical framework to describe and predict the
behaviour of stereophonic phantom imaging. Experimental work with subjective
listening tests had raised questions about the effect of head movements in
stereophonic listening [Bauer, 1962] while other tests had shown that reliable
phantom images could not be created with lateral and rear loudspeaker pairs [Ratliff,
1974]. A mathematical description of the relationship between the ear signals and the
perceived location was needed in order to resolve these issues. However, the way in
which multiple coherent audio signals combine to produce a perception of direction is
quite complex. The term summing localization is used to describe this mode of
auditory localization which arises from multiple coherent loudspeaker signals.
Clark developed a mathematical framework which explained basic Blumlein
stereophony in terms of the manipulation of level differences to produce IPD cues at
low frequencies and ILD cues at mid to high frequencies [Clark et a, 1958l]. They
also discovered, however, that the basic stereophonic method produces slightly
different results at low and high frequencies. In a number of listening tests, the
perceived direction reported was proportional to the predicted angle in both frequency
ranges, yet the perceived direction was generally wider for high frequency signals.
Clark recommended that this difference be corrected by adjusting the weighting of the
difference and sum signals by a factor of 0.7 for frequencies above 700 Hz.
Bauer used phasor analysis to describe the stereophonic principle and
formulated the stereophonic law of sines from which the standard stereo panning law
is derived [Bauer, 1962]. Bauer also assumed that while head movements did indeed
alter the perceived source direction they did not contribute directly to localization
accuracy in stereophonic listening.

39

Makitas theory of stereophonic localization assumed that the perceived

direction of a sound source corresponds to the direction of propagation of the
wavefront produced by that source [Makita, 1962]. Essentially, it states that the
angular direction of the wave-front produces the interaural cues which then in turn
give rise to a perception of direction. Makita then went on to describe the wavefront
as a function of the direction, velocity, and variations in velocity of the summed
wave-fronts. A method of predicting the perceived direction of a stereophonic image
was created and verified experimentally for a centrally positioned listener. A number
of anomalies remained, however, as just like Clark, Makita found that the perceived
direction was frequency dependent. In addition, while Makita proposed that head
movements were indeed a factor in summing localization, experimental results were
inconclusive.

3.5 Critics of Summing Localization

Summing localization is generally regarded as the best current description of
the process whereby multiple coherent loudspeaker signals produce a single perceived
source direction. However, this theory is not without its critics and various
alternatives have been suggested. Gunther Thiele suggested that the localization of
stereophonic images does not occur due to the physical superposition of the
loudspeaker signals at the ears. Instead, the auditory system first detects the
the multiple loudspeaker signals at different spatial locations, and then merges them
together to a phantom source in a psychoacoustic process after their signal content
was detected to be congruent [Thiele, 1980]. Thieles Association Model therefore
supposes that phantom sources and natural sources are perceived differently and rely
on different processes. The ear signals which arise due to the superposition of
multiple loudspeaker signals cannot therefore provide all the information about the
properties of the phantom source. While Thieles Association Model can account for
certain aspects of stereophonic localization such as the timbral shift exhibited by
phantom sources, it cannot predict the perceived source direction, and has proved
difficult to verify experimentally. For these reasons, the theory of summing
localization and not Thieles Association Model is used throughout the rest of this
thesis.

40

3.6 Meta-Theories of Localization

While these localization theories had shown that stereophony was reasonably
reliable for two channels in front of a central listener, experience with Quadraphonics
had suggested that this principle was not necessarily the best approach for threedimensional audio. Michael Gerzon suggested an alternate approach which separates
the process into two distinct stages of recording/encoding and playback/decoding.
This division allowed for the design of a decoder that would be psychocoustically
optimized, meaning it would be designed to satisfy as many of the existing
localization theories and auditory cues as possible. Gerzon suggested that this
approach, which forms the basis of the Ambisonics system, would provide the best
performance that could be expected of any spatial audio reproduction system. His
meta-theory of localization [Gerzon, 1992a] incorporated various different models of
localization and is fundamental to the design of ambisonic decoders. He described a
hierarchy of localization models and for each derives a vector whose direction gives
the predicted direction of the sound, and whose magnitude R represents the stability
of the localization. A real source would therefore have a vector magnitude of one
with a positional angle described using spherical coordinates. The optimal decoder
design would therefore ensure that the localization vectors for each model agreed for
all frequencies, and that their magnitude was as large as possible, in every direction.
Gerzons theory emphasized the importance of the two models based,
respectively, on the velocity and energy flow of the sound field at the ears. The
theories of Clark, Bauer and Makita are described as special cases of a more general
description based on the velocity of the wavefront. This velocity model is primarily
concerned with the IPD cues produced at low frequencies. Gerzon showed how a
velocity vector equal to that produced by a real source can be produced by a
multichannel system and that this will ensure that the perceived source direction will
remain stable as the head is rotated. However, this can only be achieved at low
frequencies, as at higher frequencies, the signal wavelength becomes comparable to
the size of the human head and the effect of head-shadowing becomes more
pronounced. The frequency range in which the velocity model is applied, i.e. the
decoder cross-over frequency, is therefore related to the size of the effective listening
area. Above a certain limit, the size of the listening area is smaller than the human
head, and is therefore not useful. Choosing a frequency of approximately 700Hz
41

would produce an effective listening area suitable for a single listener, while 400Hz
[Gerzon, 1992a] would be suitable for a domestic listening situation with
approximately six listeners.
Above this cross-over frequency, the decoder emphasizes the ILD localization
cues which arise due to the directional behaviour of the energy field around the
listener. It can be shown mathematically that it is only possible to recreate the energy
field of a real sound source using a small number of loudspeakers, if the sound
happens to be at the position of one of the loudspeakers. Therefore, at mid and high
frequencies, not all of the ears localization mechanisms can be satisfied in a practical
reproduction system. The direction of the energy localization vector can, however, be
adjusted so it matches the velocity localization vector (E = V) for all frequencies up
to 4kHz. This is similar to the stereophonic approach recommended by Clark (see
Section 3.4) [Clark et al, 1958]. In addition, Gerzons design optimizes RE in all
directions, which necessarily compromises localization in the directions of the
loudspeakers in favour of making the quality of the localization uniform in all
directions [Benjamin et al, 2006]. This effectively eliminates the timbral change
which occurs with amplitude panning as the signal moves from a position at a
loudspeaker to one in between loudspeakers. This also means, however, that the
localization of sources positioned at a loudspeaker will be less than optimal. In
summary, therefore, Gerzon recommends that the following optimizations be
implemented when designing an ambisonic decoder:
-

The velocity and energy vector directions are the same up to 4kHz (E = V)

At low frequencies, the magnitude of the velocity vector should be near unity
for all directions (rV = 1)

At high frequencies, the energy vector magnitude should be maximised as

much as possible and made consistent in all directions (maximum rE )

The fundamental design theory of ambisonic decoders therefore concentrates on the

velocity and energy models while other models are only used to further refine the
design. It is clear therefore that the Ambisonics system was based on psychoacoustic
principles, however, this system also built upon the work of Alan Blumlein with
coincident microphone techniques and has been developed into a complete set of
recording and production techniques.

42

4 Sound Field Reconstruction

Two of the most common techniques that attempt to reconstruct a sound field
over a given area are Ambisonics and Wavefield Synthesis. While these techniques
were initially derived from quite distinct theoretical fundamentals, recent research has
shown that under certain conditions, both theories can be considered equivalent
[Daniel et al, 1999; Nicol et al, 1999].

4.1 Ambisonics
Ambisonics was developed by Michael Gerzon, Peter Fellgett and others in the
1970s as an alternative spatialization technique for surround and periphonic
loudspeaker arrangements [Gerzon, 1974a; Fellgett, 1975]. The basic approach taken
by Ambisonics is described by one of its author as follows.
For each possible position of a sound in space, for each possible direction and for
each possible distance away from the listener, assign a particular way of storing the
sound on the available channels. Different sound positions correspond to the stored
sound having different relative phases and amplitudes on the various channels. To
reproduce the sound, first decide on a layout of loudspeakers around the listener, and
then choose what combinations of the recorded information channels, with what
phases and amplitudes, are to be fed to each speaker. The apparatus that converts
the information channels to speaker feed signals is called a decoder, and must be
designed to ensure the best subjective approximation to the effect of the original
sound field [Gerzon, 1974b].

This approach differs from Blumlein stereophony in a number of ways. Firstly, the
initial encoding stage is removed from the eventual playback system, its sole aim
being to capture as much information about the sound scene as possible using a
certain number of channels. The decoding stage can now use the recorded spatial
information to determine the appropriate loudspeaker signals that will recreate this
spatial scene. Furthermore, as discussed in the last section, this decoding stage can be
psychacoustically optimized so that as many localization mechanisms as possible are
satisfied [Gerzon, 1992a]. The technical means by which this system was realised
built on the work by Blumlein with coincident microphone techniques. This chapter
will therefore begin with a summary of the Ambisonics system and the associated
encoding and decoding process based on microphone directivity patterns. This will
be followed with a more detailed description of the theoretical basis of the system and
the application of psychoacoustic principles to the decoding process.
43

4.1.1 Ambisonics and Microphone Directivity Patterns

Consider the arrangement shown in Fig. 4.1. A point source S is producing a
wavefront which travels from a frontal position to the centre listening position P.
Four directional microphones are positioned coincidentally at the centre point P such
that each microphone is pointing directly toward a loudspeaker positioned at the four
corners. The basic stereophonic approach would be to record the wavefront with the
four microphones and then replay the four signals from the four loudspeakers in order
to reproduce the original wavefront at the centre listening position P.

Fig. 4.1 Basic multichannel stereophony example

A unique characteristic of Blumleins coincident figure-of-eight arrangement

(shown in Fig. 4.2 [Wiggens, 2004]) is that the two response patterns can be
combined to generate a new virtual figure-of-eight response pattern pointing in any
direction in the same plane. The addition of a third vertically orientated figure-ofeight microphone extends this idea to three dimensions. Any figure-of-eight
microphone can be combined with an omni-directional microphone to produce a
response characteristic with increased directivity in a certain direction. So if an omnidirectional microphone is added to this coincident arrangement of three figure-of44

eight microphones, it will be possible to create almost any response pattern in any
direction. If the four microphones shown in Figure 4.1 are replaced with this new
arrangement, then four virtual response patterns corresponding to the original
loudspeaker arrangement can be derived. However, if the number of loudspeakers is
increased from four to eight, eight new virtual response patterns corresponding to this
new loudspeaker arrangement can also be derived. In fact, with this particular
arrangement of four microphones it is possible to create response patterns that
correspond to whatever loudspeaker arrangement is available. In addition, the
directivity of the response characteristic can be adjusted which is highly important in
the decoding process.

Fig. 4.2 Microphone responses derived from two figure-of-eight microphones

The process just outlined describes in broad terms the basic Ambisonics
encoding method. This particular microphone arrangement is called a Soundfield
microphone and is shown in Fig. 4.3. This apparatus contains the four described
microphone capsules and produces four audio signals which correspond to the four
microphone response patterns. The Ambisonics terminology for this set of four
signals is referred to as A-Format. Obviously it is not possible to have four
microphone capsules occupying the exact same point in space, however, Soundfield
microphones are able to overcome this problem with electronic equalization so that
the output produced is essentially coincident up to 10 kHz [Rumsey, 2001]. These
45

corrected signals are labelled W, X, Y and Z in reference to the four microphone

capsules and are generally collectively referred to as a B-format signal in Ambisonics
terminology.

Fig. 4.3 The Soundfield microphone

The preceding description of the Ambisonics system illustrates how it

developed from Blumleins research into coincident microphone techniques. It
describes how a relatively simple arrangement of four microphone capsules can be
used to capture a significant amount of directional information about a sound field.
The functionality of the Soundfield microphone readily lends itself to a description
based on spherical harmonics. Using this mathematical description, any function of
direction can be described as a function on the surface of a sphere, and expressed
uniquely as a sum of spherical harmonics. The order of the theory is the order of
spherical harmonics used by the model and a minimum of zero and first order
components are required to represent a three dimensional field. Figure 4.4 graphically
illustrates the zero and first order spherical harmonics [Wiggens, 2004], and their
relationship with the Soundfield microphone response characteristics is clear. Using
this mathematical description, it is also possible to encode an existing monophonic
signal into a B-format signal and position it at any location on a unit sphere
surrounding the central listening position. This technique forms the basis of
ambisonic panning and the encoding equations can be readily applied in software.
This mathematical description of the B-format encoded sound field also allows
operations such as rotations and translations to be applied to the whole sound field.
The tilt, tumble and dominance effects found in many ambisonic processors
implement these particular effects [Malham, 1998].

46

Fig. 4.4 Zero and first order spherical harmonics

4.1.2 Ambisonic Decoders

One of the unique advantages of the Ambisonics system is that the encoded
signal can be decoded for a variety of loudspeaker arrangements from mono to stereo,
horizontal arrays of loudspeakers or even periphonic (with height) arrays. It has been
shown how the decoder design is psychoacoustically optimized in terms of the
velocity and energy localization vectors described by Gerzon in his meta-theory of
auditory localization [Gerzon, 1992a]. The decoding is accomplished by calculating
the virtual response patterns needed for the available loudspeaker array and then
performing two different decoding strategies at low and high frequencies. This is
often implemented using shelf filters (consisting of 90o all-pass filters) which
smoothly alter the gain of W relative to X, Y and Z above and below the chosen
crossover frequency. In more recent designs, a cross-over filter network is used to
split the signal instead of the traditional shelf filters [Farina et al, 1998]. In practice,
optimal decoders can be readily designed for regular polygon loudspeaker
arrangements where the loudspeakers are placed in diametrically opposed pairs.
Optimal decoders for non-regular loudspeaker arrangements are much harder to
derive and are the subject of considerable research [Wiggens, 2004].
Gerzon defined the Ambisonics system in terms of the decoder design.
However, a number of different decoders have since been suggested by Gerzon and
others [Malham, 1995, Farina et al, 1998]. The original specification used shelf
filters to apply two different sets of weightings to the W and XYZ components at low
and high frequencies. This produces different directivity characteristics which in turn
maximizes the velocity and energy vectors rV and rE at the frequencies in which these
localization models dominate. As stated previously, the choice of cross-over
frequency is related to the size of the effective listening area, with Gerzon
recommending a figure of 700Hz for a single listener and 400Hz for a small group of
listeners [Gerzon, 1992c]. As the size of the required listening area increases further,

47

the appropriate cross-over frequency will continue to decrease. Intuitively, this

suggests that a single-band decode would be more suitable for very large areas and
indeed dual-band decoding using shelf filters is generally not recommended for large
listening areas [Malham, 1992]. This modification becomes necessary when large
numbers of listeners are seated away from the centre point, due to the way in which
the velocity vector rV is optimized in the original decoder specification. In order to
ensure that rV =1 at the centre point, the loudspeaker directly opposite to the source
direction must produce a signal with inverted phase relative to the loudspeaker signal
at the source direction. The phase relationship between the multiple loudspeaker
signals will therefore be distorted at off-centre listener positions. This is particularly
true for listeners seated near to a loudspeaker as these anti-phase components can
distort the perceived source direction toward the loudspeaker directly opposite the
intended direction. At higher frequencies, it can be shown that ILD localization cues
are produced by the temporally averaged directional behaviour of the energy field
around the listener [Gerzon, 1992a]. The energy model is therefore not dependent on
the phase relationship between the loudspeaker signals and should therefore be useful
for non-central listeners. Various decoders suitable for large listening areas have been
proposed. These implement a single frequency band decode which optimizes the
energy vector RE and reduces or completely eliminates anti-phase components. The
effectiveness of these decoders will be examined in more detail in the next chapter.

4.1.3 Higher Order Ambisonics

The original Ambisonics system was based on the zero and first order
decomposition of a sound field as this approach produced acceptable results with an
economical number of channels. It has been shown how a function of direction may
be described as a function on the surface of a sphere and expressed uniquely as a sum
of spherical harmonics. If the order of spherical harmonics is increased then the
accuracy of the encoded sound field also increases, as this effectively increases the
spatial resolution of the system. The order of the system reflects the order of
spherical harmonics used, so the traditional Ambisonics B-format signal represents a
first order system, while a second order system would also include the additional
spherical harmonics shown in Figure 4.5 [Wiggens, 2004]. Obviously it is extremely
difficult to create higher order Soundfield microphones as there is a limit to how

48

many capsules can be positioned at approximately the same point in space. This topic
is currently the focus of considerable research and various higher order microphone
arrangements are being examined (see Figure 4.6) [Bertet et al, 2009; Moreau et al,
2006]. However, there is no such problem when encoding monophonic recordings in
software and it is in this regard that higher order Ambisonics has proved most useful.
Various software implementations of high order ambisonic encoders and decoders are
now available and this has important consequences for situations involving multiple
listeners.

Fig. 4.5 Second (top) and third (bottom) order spherical harmonics

Fig. 4.6 First, second & third order microphones

4.1.4 Ambisonics in Practice

While the specification and design of first order B-format Ambisonics systems
is well established and standardized, the same cannot be said for higher order systems.
Although, Gerzon presented equations up to third order in his 1975 paper, these were
specified using Cartesian co-ordinates rather than the polar system which later
became standard in ambisonic systems. As a result, a defined terminology for higher
49

order systems, and a spherical harmonic formulation for the higher order channels in a
form which is consistent with B-format systems, has not yet been agreed upon
[Malham, 1999].
W
X
Y
Z
R
S
T
U
V
K
L
M
N
O
P
Q
where

=
(1 / 2)
=
cos(A) * cos(E)
=
cos(E) * sin(A)
=
sin(E)
=
(3*sin2(E)-1) / 2
=
cos(A) * sin(2E)
=
sin(A) * sin(2E)
=
cos2 (E) * cos(2A)
=
cos2 (E) * sin(2A)
=
sin(E) * (5*sin2(E)-3)/2
=
(135/256) * (cos(E)*cos(A)*(5*sin2(E)-1))
=
(135/256) * (cos(E)*sin(A)*(5*sin2(E)-1))
=
(27/4) *(cos(2*A)*sin(E)*cos2(E))
=
(27/4) *(sin(E)*sin(2*A)*cos2(E))
=
cos(3*A)*cos3(E)
=
sin(3*A)*cos3(E)
A = source azimuth &
E = source elevation
Table 4.1 Furse-Malham set of encoding coefficients

While the various formulations of the ambisonic equations are mathematically

equivalent, the same equations must obviously be used for encoding and decoding if
the sound field is to be reconstructed correctly. The Furse-Malham (FMH)
formulation of the ambisonic equations was developed to maintain compatibility with
older B-format hardware and software while extending the system specification to
third order spherical harmonics. The FMH set deviates from a strict mathematical
description of spherical harmonics as it applies weightings to the channels so that all
the harmonic coefficients have a maximum value of one [Malham, 1999]. In a
mathematically strict formulation, as the order M of the harmonics is increased, the
maximum value that each harmonic may attain increases. While this is not a problem
for diffuse sound fields, point sources produced by panning could result in excessive
signal levels in the higher order channels. The FMH set therefore applies weightings
to each channel to prevent this occurring. In addition, the FMH set reduces the zeroth
order W signal by 3dB (gain of 0.707). This gain factor was originally applied to
produce equivalent signal levels in the W and XYZ components and was included in
the FMH set to maintain compatibility with the original B-format specification used

50

in existing ambisonic hardware such as the Soundfield microphone. The FMH set of
coefficients for encoding a signal into third order Ambisonics is shown in Table 4.1.

Furse
Malham

Criteria

N3D N2D

Hybrid
W

SN3D

Extensible to higher orders in a straightforward manner. Hence able

to meet expanding users needs.

No

Yes

Yes

Yes

Yes

Consistent with 1st-order B-format recordings, equipment and

software.

Yes

No

Yes

Yes

No

Consistent with higher-order horizontal-only FMH format

recordings, equipment and software.

Yes

No

Yes

No

No

Consistent with the Ambisonics portion of MPEG-4

Yes

Yes

Yes

No

Yes

Consistent with FMH format recordings, equipment and software

having periphonic order P greater than 1.

Yes

No

No

No

No

Components of the same degree have the same 3D gains. So that

typical processing (such as rotation) can be coded reasonably
cleanly.

No

Yes

Yes

Yes

Yes

The mathematical formulation is consistent across the degrees. No

special cases.

No

Yes

Yes

No

Yes

The harmonics are orthonormal, which eases such things as

beamforming.

No

Yes

No

No

No

The scheme benefits from direct support in general maths libraries.

No

Yes

No

Yes

Yes

The scheme has mathematical grounding in the 3D world.

No

Yes

No

No

Yes

Higher-degree components have headroom for when the sound field

is not spherically isotropic. Means that 1st order monitoring can be
used over higher orders (with more accuracy).

Yes

No

Yes

No

Yes

Greatest dynamic range when the major sources are in/near the
horizontal plane

No

No

Yes

No

No

Greatest dynamic range when the major sources are distributed over
the sphere

No

Yes

No

No

No

Table 4.2 Analysis of Ambisonics normalisation schemes

The deviation of the FMH set from a strict mathematical spherical harmonic
formulation does, however, have some negative consequences. Firstly, while
decoding equations can be readily obtained up to third order, this is much more
difficult at higher orders. In addition, the calculation of matrices for tilt, tumble and
dominance effects has proven to be extremely difficult, and, the alphabetical naming
scheme is unpractical above third order. As it is much easier to design and implement
Ambisonics systems based on more mathematically pure formulations of the basic
equations, particularly at higher orders, a number of different formats have been
proposed. Efforts are being made by the Ambisonics community to standardize these
different formats and an analysis of the advantages and disadvantages of the

51

prominent formats is shown in Table 4.2. The necessary weightings to convert

between these different formats are also readily available [Malham, 1998; Ambisonics
Standards, 2009].
It should be noted that the Ambisonics system assumes that all loudspeakers
are at a great enough distance from the listener so that the signals arriving at the ears
can be approximated by plane waves. However, this may not be the case when there
is only a single listener within a relatively tightly spaced loudspeaker array. This
effect is illustrated in Figure 4.7 [Nicol et al, 1999] which compares the reproduction
of a plane wave using ideal and real loudspeakers. In this case nearfield
compensation must be applied to the decoder to negate the low frequency boost which
results due to the spherical radiation pattern of the nearby loudspeakers [Daniel,
2003].
The decoding equations for regular loudspeaker layouts are readily available
up to second order [Furse, 2000] and numerous higher order software decoders are
available in Steinbergs VST plugin format or as externals for packages like Max
MSP and Audiomulch. The decoding equations can be formulated in a number of
different ways, such as;
S = (2/N) *[W + 2(cos(a) * cos(e) * X) + 2(sin(a) * cos(e) * Y) + 2(sin(e) * Z)]
[Gerzon, 1992a]
S = (0. 5) *[(2 d)(2)(W) + d((cos(a) * cos(e) * X) + (sin(a) * cos(e) * Y) + (sin(e) * ))]
[Wiggens, 2004]
S = zerobase * (W) + firstbase*[(cos(a) * cos(e) * X) + (sin(a) * cos(e) * Y) + (sin(e) * Z)]
[Schacher et al, 2006]
where S is the loudspeaker signal
W,X,Y and Z are the B-format signals
N is the number of loudspeakers
d is the directivity factor (0 = omni, 1 = cardioid, 1.5 = hypercardioid, 2= Fig-of-8)
zerobase, firstbase = reformulated directivity factor

These formulations of the first order decoding equations differ in a number of ways.
The first difference relates to how the 3dB gain adjustment (factor of 2) is
distributed between the zero and first order components. Similarly, the gain
adjustments which are used to alter the directivity of the response characteristic can
be applied in a number of ways, as shown in the second two equations. Interestingly,
only the first equation includes a weighting factor due to the number of loudspeakers,

52

so if the second two equations are used the perceived volume will increase as more
loudspeakers are added. The first approach has some potential benefits as, if the
volume is automatically reduced when more loudspeakers are added, as in the first
equation, then, the signal to noise ratio (or the bit resolution in a digital
implementation) will also be reduced.

Fig. 4.7 Ambisonic plane wave - theoretical (left) and real (right) sources
Decoder Titles /
Descriptions

Approx. optimum
frequency ranges
Response
Characteristic
Effective Listening
Range

Strict,
Matched,
Velocity,
max-rV,
Furse-Malham,
Systematic,
Ideal
< 700 Hz
Hypercardioid, strong
anti-phase components
Single Listener

Energy,
max-rE,
Rationalised Energy
Decoder,
Rationalised Square
Decode

In-phase,
Controlled opposites,
Cardiod

500Hz - 5kHz

All

Hypercardioid, reduced
anti-phase components
Increased listening area,
still some anti-phase
components
Criteria (1st order)
rV = 1 rE = 0.667
rV = rE = 0.707
nd
Criteria (2 order)
rV = 1 rE = 0.8
rV = rE = 0.866
Criteria (3rd order)
rV = 1 rE = 0.857
rV = rE = 0.924
V = E
V = E
Crieria (all)
Table 4.3 Summary of ambisonic decoding schemes

Cardioid, no anti-phase
components
Large listening area but
reduced localization
accuracy.
rV = 0.5 rE = 0.667
rV = 0.667 rE = 0.8
rV = 0.75 rE = 0.857
V = E

The precise decoder weightings, which correspond to the directivity

characteristic, can be derived mathematically in various different ways. In general
there are four particularly common decoding schemes, which are summarized in
Table 4.3. The original dual-band decoder proposed by Gerzon optimizes the velocity
and energy vectors in different frequency ranges. Each of these two decoding
schemes can of course also be applied across the entire frequency range. The final
53

common decoding scheme is the in-phase decode proposed by David Malham for
large areas. Figure 4.8 illustrates the directivity patterns of these different ambisonic
decoding schemes [Daniel, 2000].
It should also be pointed out that there are a minimum number of speakers
needed to successfully reproduce each ambisonic order, and that this number is
always greater than the number of channels available for the decoder [Wiggens,
2004]. The number of loudspeaker required can be calculated from the following
equation, where M is the order of the system, and N is the number of loudspeakers.
N > 2M+1
M > (N+1)

for horizontal arrays only

2

for periphonic arrays.

Fig. 4.8 Directivity patterns of various ambisonic decoding schemes

The preceding discussion illustrates the many different implementations of the

Ambisonics system that have been developed since its conception in the 1970s. In
many respects the Ambisonics system is defined in terms of the decoder, and the
design of the decoder is perhaps one of the most crucial aspects of any Ambisonics
system. Various objective and subjective assessments have been carried out in an
attempt to evaluate the perceptual effect of these different decoding schemes, and
these will be summarized in the next Chapter.

54

4.2 Wavefield Synthesis

Fig. 4.9 Illustration of the Huygens principle

In much the same way that Ambisonics can be considered as an extension to

the theories of Alan Blumlein, Wavefield Synthesis (WFS) can be considered as an
extension of the acoustic curtain approach developed at Bell Labs [Steinberg et al,
1934]. The theoretical basis of WFS was developed by Berkout at TU Delft in the
eighties. Berkout, whose background was in seismology, used Huygens wave theory
as the basis for a system of acoustical holography [Berkout, 1998]. Huygens
principle states that any wavefront can be regarded as a superposition of elementary
spherical waves. Therefore, any wavefront can be synthesized by a number of
secondary point sources, as illustrated in Figure 4.9. Berkout used Kirchoffs
quantified version of this theory to create a generalised spatialization system using
linear arrays of loudspeakers (see Figure 4.10). In effect, it allowed for the
calculation of loudspeaker signals which would produce the same wavefront in the
listening area as would have been produced by a real source in any position behind, or
even in front of the array. While the particular integral equation involved is quite
complex it essentially operates by adjusting the delay and gains of the signal produced
by each loudspeaker so that they combine to produce the required wave. This
equation includes a term which must be adapted to the actual directivity
characteristics of the loudspeakers in the array [de Vries, 1995].
The WFS approach has a number of potential advantages over other
spatialization techniques such as stereophony or Ambisonics. Firstly, as WFS

55

effectively samples and reconstructs the desired wavefront, it can potentially

reproduce the correct wavefront over an extended listening area. This has enormous
potential for the presentation of spatial music and is quite different from the single
listener orientated formats such as stereophony or Ambisonics. In addition, WFS can
potentially position point sources both behind and in front of the array as illustrated in
Figure 4.11

Fig. 4.10 The WFS method

Plane wave sources can also be achieved with WFS, as shown in the middle
diagram in Figure 4.11. This type of reproduction mimics the effect of very distant
sources and can potentially be used to extend the suitable listening area for standard
two-channel stereo reproduction. Figure 4.12 illustrates how generating the two
stereo signals as plane waves can potentially maintain the angular relationship
between the two wavefronts across a greater area [Franco et al, 2004].

Fig. 4.11 WFS reproduction of different wavefronts

56

While WFS appears to be a promising technique there are a number of

problems which have proved quiet difficult to overcome. Many of these issues arise
from the simplifications which must be applied to the Kirchoff-Helmholtz integral
equation so that it can be used with practical loudspeaker arrays. Firstly, the
Kirchoff-Helmholtz integral describes the behaviour of spherical wavefronts in three
dimensions, so the restriction to the horizontal plane introduces amplitude errors in
the reconstructed wavefront. WFS can be implemented in three dimensions using a
loudspeaker wall [Komiyama et al, 1991; Ono et al, 1997], however, this drastically
increases the required processing power and number of loudspeakers. For this reason,
typical WFS setups use linear arrays which are restricted to the horizontal plane and
include additional processes in an attempt to reduce the errors introduced by this
simplification.

Fig. 4.12 WFS reproduction of two-channel stereo

In WFS theory, the new wavefront is constructed by summing an infinite

number of secondary sources, whereas in reality only a finite number of loudspeakers
are available. This simplification places boundaries on the effective listening area and
can also potentially introduce echo signals due to diffraction effects at the edges of the
array. These truncation effects can be clearly seen in Figure 4.13 which shows a
source placed two metres behind the centre of the array and the two echo signals
which arise from the edges of the array [Franco et al, 2004]. Truncation effects can
be reduced by applying a tapering windowing function to the loudspeaker signals in
order to reduce the weighting of the loudspeakers signals at the edges of the array

57

[Boone et al, 1995] However, this solution does reduce the effective listening area
[Sonke, 2000].
Perhaps the most significant issue with WFS is the problem of spatial aliasing
which arises due to the discrete nature of the loudspeaker array. Essentially, the interloudspeaker spacing places an upper limit on the frequencies which can be accurately
reconstructed with the array. This effect is somewhat analogous to Nyquists
sampling theorem which states that the sampling rate must be twice the highest
frequency to be reproduced. A WFS system effectively samples a virtual wavefront,
and in this case it is the inter-loudspeaker spacing which places an upper limit on the
frequencies which can be accurately sampled and reproduced.

Fig. 4.13 Truncation effects (a) and (b) 4ms later

This frequency limit, the so-called aliasing frequency, is also dependent on the
angular position of the source and listener, relative to the array. Figure 4.14 shows a
distant source positioned at 30 o to the left behind the array for two frequencies above
and below the aliasing frequency [Franco et al, 2004]. Above this limit, the original
shape of the wavefront is not reproduced and the perceived direction and spectral
coloration of the source will be highly dependent on the listening position.
The aliasing frequency can obviously be increased by using a greater number
of loudspeakers placed closer together. However, in practical WFS systems this is
often not possible. One solution is to randomize the time-offsets of all of the high
frequency source content in an attempt to reduce the extent of the artefacts which
occur due to spatial aliasing [Start, 1997]. This approach does, however, sacrifice the
spatial accuracy of the sound field at high frequencies and has proven to be quiet

58

difficult to implement. One promising solution to this problem is to use standard

stereophonic phantom imaging above the spatial aliasing frequency, and WFS below.
Figure 4.15 illustrates this approach of Optimized Phantom Source Imaging (OPSI)
where the three blue loudspeakers are used to create phantom images which
correspond to the direction of the virtual source generated by the WFS system
[Wittek, 2002]. Experimental evaluation of this approach has been promising and
suggests that it does not negatively impact on perceived source localization accuracy,
even over large listening areas [Huber, 2002].

Fig. 4.14 A distant source (a) below the aliasing freq & (b) above

Fig. 4.15 Optimised Phantom Source Imaging WFS

59

The influence of acoustic reflections of the loudspeaker signals within the

listening room can also potentially degrade the perceptual accuracy of the reproduced
wavefronts to a significant degree. While traditional physical acoustic absorption
techniques in the listening room can obviously be used, active cancellation using the
reproduction loudspeaker array is also under investigation [Spors et al, 2003]. Some
authors have suggested that the negative influence of the listening room acoustic is
one of the most serious problems in practical WFS systems [Wittek, 2003]

Fig. 4.16 A WFS cinema system in Ilmenau

4.2.1 WFS and Spatial Music

While the WFS approach has numerous potential benefits, it is extremely
demanding in terms of hardware and processing power. The relationship between the
size and spacing of the array and the effective listening area means that large numbers
of loudspeakers are required to create an effective listening area sufficient for a large
audience. For example, in order to obtain a spatial aliasing frequency of 1kHz over a
circular listening area with a diameter of two metres, a WFS system with 32
loudspeakers is required [Theile et al, 2003]. It is clear therefore that extending this
system to cover a listening area suitable for a reasonably sized audience is logistically
challenging, and for this reason, WFS has not to date been widely used for the
performance of spatial music. However, the potential of WFS to create sources
within the array and to work over an extended area, means that this technique will
continue to be an intriguing prospect for composers of spatial music. WFS is also the
subject of considerable research and a number of commercial WFS systems are now

60

available. One system has been successfully operating in the Ilmenau cinema since
2003 (see Figure 4.16). In addition, the WONDER system, developed by Marije
Baalman for the Linux environment, is a free open-source WFS program developed
specifically for the production of spatial music [Baalman, 2006]. While WFS is
certainly a highly promising technique, the large amount of hardware required for its
implementation will most likely restrict its use to large dedicated venues for spatial
music production, at least for the near future.
WFS is one of the few spatialization techniques which can theoretically
recreate the correct wavefield over an extended listening area and its ability to
simulate virtual sources within the listening area is certainly unique. Other techniques
such as stereophony and Ambisonics can only reproduce sources at the distance of the
loudspeakers, and additional processes must be used to simulate distance. These
processes attempt to simulate the perceptual cues related to distance hearing presented
in Section 2.2, and are discussed in the next Section.

61

5 The Simulation of Distance

As discussed previously in Section 2.3, the perception of distance is thought to
largely depend on the following parameters;
-

The relative amplitude of the source signal

The spatio-temporal pattern of early reflections

The ratio of the direct and reverberant signals

The attenuation of high frequencies due to air absorption

The Doppler effect

The familiarity of the listener with the source signal

While recorded sounds may retain many of these perceptual cues, the same cannot be
said for synthesized sounds. A method of simulating distance through the
manipulation of these perceptual cues was first proposed by John Chowning in 1971.
This system, although originally specified for a Quadraphonic system, can readily be
adapted to other loudspeaker arrangements and directional panning techniques. When
this system was first proposed, the influence of early reflections on distance
perception was not yet fully understood. Chowning's system therefore used the direct
to reverberant signal ratio as the primary distance cue, with some modifications.

5.1 The Simulation of Moving Sources

Chownings model for the simulation of moving sources [Chowning, 1971]
used amplitude panning to position a source at a certain azimuth within a
Quadraphonic loudspeaker array. At the same time, the source signal is also routed to
a multichannel reverberation unit. The perceived distance of the source is then
controlled by simply adjusting the relative amplitude of the direct and reverberant
signals. Chowning proposed that the amplitude of the direct signal should be set to
1/distance while the reverberant signal is set to 1/distance. This configuration will
result in the intensity of the direct sound reaching the listener falling off more sharply
with distance than the reverberant sound, such as would occur with a real source.
However, this approach has the result that at large enough distances, the reverberant
signal will become greater than the direct signal, and the source direction will become
ambiguous. To overcome this, Chowning proposed that the reverberation be divided
into the global reverb signal which is reproduced from all loudspeakers and is
62

proportional to 1/distance, and the local reverberation which is produced in the same
direction as the direct signal and made proportional to 1-(1/distance). Therefore, as
the distance increases, the reverberation becomes increasingly localized,
compensating for the decreasing direct signal level. Chowning suggests that this is a
fair approximation of a real acoustical situation as when the distance of a sound
source increases, the distance to nearby reflecting surface decreases, thereby giving
the reverberation some directional emphasis [Chowning, 1971]. More recent research
has suggested that the directional aspect of the reverberant signal, particularly of the
lateral early reflections, is used by the auditory system to estimate the egocentric
distance to the source [Bronkhurst, 2002; Gerzon, 1992b]. The directional behaviour
of the reverberant signal in Chowning's system is somewhat similar to this situation.
However, as the precise spatio-temporal pattern of the reflected signal is not
recreated, it is likely that this system will not provide an absolute sense of distance.
The advantage of this system lies primarily in its efficiency and Chowning went on to
develop a real-time digital implementation of the system using a standard
Quadraphonic loudspeaker array. The inclusion of the Doppler effect and other
secondary cues such as high frequency air absorption, when combined with
independent control of the direct and reverberant signals resulted in an effective
simulation of movement which was highly sophisticated for its time and indeed this
approach is still widely used today, particularly for real-time applications.
Chowning illustrated this new technique, along with a number of FM synthesis
algorithms (also developed by Chowning) in the landmark work, Turenas, which was
completed in 1972.
A significant increase in computational processing power was required if
multiple discrete reflections, as well as the diffuse global reverberation, were to be
accurately simulated. Indeed, it was more than a decade after Chowning's 1971 paper
before such a spatialization system was proposed.

5.2 A General Model for Spatial Processing of Sounds

F. R. Moore developed Chownings algorithm into a more general conceptual
spatial processing model which was implemented in the Cmusic sound synthesis
language [Moore, 1983]. This approach draws upon previous work on artificial
reverberation techniques [Schroeder, 1962; Moorer, 1979] and psychoacoustics

63

[Blauert, 1997; Gardner, 1962]. Moores model represents the performance and
illusory acoustic spaces as two rooms, namely;
-

The listening area or inner room whose boundary is marked by the

surrounding loudspeaker array, and

The illusory acoustic space or outer room from which sounds emanate.

Fig. 5.1 Moores spatial model (a) direct signal (b) reflected signal paths

In this model, the loudspeakers function as acoustic windows which transmit

information from the illusory space outside the listening area perimeter. Increasing
the number of loudspeakers therefore increases the spatial resolution of the illusory
space. The paths of both direct and reflected signals are calculated using this two
room model with the assumption that the boundaries of the inner room/listening area
are completely absorptive. This has the effect that any paths which cut through the
inner room are completely attenuated. Figure 5.1 (a) shows the direct signal path
from a front-right source to a quadraphonic array. In this instance the back-left
loudspeaker would be completely attenuated due to the absorptive characteristics of
the inner room boundary. Figure 5.1 (b) illustrates sixteen potential reflected paths,
one from the source to each wall and to each loudspeaker [Moore, 1983]. Each of
these paths are evaluated in terms of the following parameters;
-

Attenuation due to distance travelled

Frequency dependent attenuation due to air absorption

Frequency dependent attenuation due to reflection and absorption by the

reflecting surface

64

Absorption due to collision with the outer walls of the inner room (these are
modelled as being completely absorptive)

Time delay due to the finite speed of sound transmission

The various attenuation factors due to reflections and absorption are derived from
measurements of real acoustic spaces and surfaces. A reverberation algorithm based
on the overall size and shape of the outer room and the reflective properties of its
walls is used to generate the diffuse global reverberation. Multiple delay lines are
used to model the transmission times of the direct and reflected signals from the
source to the listener. A change in source distance will therefore result in a change in
the amount of delay, producing a pitch shift which is very similar to the Doppler
Effect.
Real-time implementations of Moores model were later developed [Ballan et
al, 1994] including a version for the Max/MSP environment [Yadegari et al]. The
simplifications introduced in order to produce a real-time configuration included the
down-sampling of source paths, interpolating delay times, and improvements in the
inner room ray intersection detection algorithm [Yadegari et al, 2002]. The real-time
modular spatial-sound-processing system Spat (discussed previously in Section 3.1.3)
is also based on the algorithms developed by Chowning and Moore [Jot, 1999]. More
recent adaptions of this model for the Csound processing language have also been
developed [Hofmann, 2008] and will be discussed later in Chapter Twelve

5.2.1 The Implications of Moores Model

The spatial audio model developed by Moore differs from other spatialization
techniques in that it emphasizes the simulation of distance rather than direction.
Systems like stereophony and Ambisonics attempt to reproduce the psychoacoustic
cues predominantly associated with directional hearing to the best possible extent, in
every direction. Moore argued that this approach cannot be applied in a concert
situation as different perceptual cues would be required at each listening position in
order to produce the same perceived source direction for each listener. Moores
model therefore focussed on the physical characteristics of the real or imaginary space
or spaces to be simulated, while ignoring the psychoacoustic parameters which vary
for different listener positions. The contributions of the real listening environment are
simply discounted and the composition is specified solely in terms of the virtual

65

space. Moore argues that although each listener will hear something different, all
listeners will in general perceive different perspectives of the same illusion. He
states;
Depending on proximity to the loudspeakers, each listener will hear these
sounds from a slightly different perspective. Information is presented at each
loudspeaker about all sound sources in the virtual outer room. The
differences in perception among listeners in the inner room are analogous to
perspective distortion. [Moore, 1983]
In effect, Moore assumes that each listener will localize the virtual source to a
different direction, but that the relative source distance and the broad overall
trajectory will still be apparent to each listener.

5.3 Ambisonics Distance Encoding

As discussed in Chapter Three, modifications must be made to an Ambisonics
system when the loudspeakers are very near the listener. In this case, nearfield
compensation is applied to the decoder to negate the low frequency boost which
results due to the spherical, rather than the assumed planar, radiation pattern of the
nearby loudspeakers. This is somewhat similar to the proximity effect which occurs
with velocity microphones. Daniel proposed a new encoding format called NearField Compensated Higher Order Ambisoncs (NFC-HOA) which compensates for
this bass boost in the encoding stage [Daniel, 2003]. Daniel also suggested that this
effect could be used for the simulation of distance. Sources at the loudspeaker
distance are simply uncompensated, as they will reproduce the distance related
wavefront curvature anyway. Sources beyond the array will have reduced low
frequency content to compensate for the nearfield distortion, while sources inside the
array can in theory be simulated by increasing the proximity bass boost. As this
filtering is applied in the encoding stage, the size of the loudspeaker array now has to
be known in advance. This is a significant disadvantage as the clear separation of the
encoding and decoding stages is one of the principal benefits of Ambisonics.
However, it may be possible to apply additional filtering to compensate for any
differences between the array size assumed during encoding and the actual array
specified at the decoding stage [Daniel, 2003]. It has also been suggested that
focussed ambisonic sources (similar to WFS) can also be created inside the array, but
this approach has yet to be experimentally verified [Ahrens et al, 2009].

66

5.4 Evaluating of Spatialization Techniques

The preceding two chapters outlined the various perceptual mechanisms
involved in spatial hearing and also the different ways spatialization techniques
attempt to reproduce these auditory cues. A spatial location can be defined in terms
of the angular direction and distance from the listener to the source, and different
electronic processes have been developed to simulate these two parameters. WFS is
the one exception, as this system can theoretically recreate the exact wavefield, and
hence both the directional and distance cues. The effectiveness of this, or any
approach to spatial audio can only really be evaluated through empirical research.
This is particularly true for situations involving multiple listeners as many of these
spatialization systems were primarily designed for a single or small group of listeners.
The following chapter will therefore summarize the results of a number of
experiments on the perceptual performance of various spatialization techniques, using
both objective and subjective criteria. It is hoped that the analysis of these results will
provide some insight into the capabilities and limitations of these techniques, which
will in turn provide some indication of the perceptual validity of various
compositional strategies in a performance setting.

67

6 The Assessment of Spatial Audio

Spatial audio production techniques have been evaluated using both subjective
listening tests, and objective tests based on acoustic measurements and/or data
simulations. Many of these tests focus on angular localization accuracy, particularly
in the horizontal plane, and use this as the primary measure of the effectiveness of the
spatialization technique(s) under examination. The source angle can be measured
using subjective data, with listening tests for example, or with objective data
generated from recordings or data simulations. The angular position of a static source
in a fixed location is perhaps one of the easiest parameters to assess in spatial audio.
However, it by no means follows that this is the only important factor. Naturalness is
another highly important aspect of any sound reproduction system and this highly
subjective parameter is often used to determine the listeners impression of the
realism or trueness to nature of the auditory scene [Rumsey, 2001]. Rumsey
suggests that "out of phase effects, such as when the phase of one loudspeaker or
headphone is inverted, can contribute to negative judgements of naturalness as this
effect never occurs under natural circumstances [Rumsey, 2001].
The perceived distance, and in the case of dynamically moving sources, the
smoothness of the trajectory, are both important parameters in spatial music
composition. Is it preferable that every listener perceives the same trajectory, even if
the movement crudely jumps from loudspeaker to loudspeaker, or is it better if every
listener hears a natural and smooth movement, even if the precise trajectory varies
from listener to listener? These broad questions are difficult to assess rigorously in an
unbiased subjective examination and the wide range of parameters to be controlled
means that data modelling or simulations can only provide limited results. For this
reason, many researchers focus either on specific criteria, such as angular location,
distance or spatial trajectory, or alternatively on broader, more subjective attributes
such as envelopment or naturalness, which are often assessed with free-response or
scaled judgements between different reproduction techniques. Concert hall acoustic
studies in particular focus on the sense of spaciousness or envelopment which arises
due to the diffuse, ambient sound and is desirable in this context (see Chapter Two).
Localization studies are instead primarily used to examine the perceptual mechanisms
related to directional hearing, or the performance of electronic spatialization
68

techniques and so tend to focus on measurements of the size and location of the
source signal. However, while this distinction serves a reasonable purpose in this
case, it is not clear that such a distinction can be made in the area under discussion in
this thesis. A complex sound scene will typically contain multiple sound objects
which may overlap or even merge depending upon the context. This may result in a
diffuse source which leads to a sense of spaciousness without a clearly perceived
direction. In addition localization measurements only record the perceived source
direction and do not indicate whether the source is sharply defined and focussed or
widely dispersed around a particular direction. Blauerts measure of locatedness
which was discussed earlier in this thesis will be an important measure in this regard
[Blauert, 1997].
A common approach in this area is to restrict the discussion to the ideal
situation of a single, centrally positioned listener in an anechoic or near-anechoic
room. While this approach may provide useful information about the relative
strengths of different techniques it does not address their performance under less than
ideal conditions. For performances of spatial music, it is critical to also assess the
influence of off-centre listener positions and the effect of the acoustic reflections in
the playback room.

6.1 Localization Accuracy under Ideal Conditions

Numerous studies have been carried out to determine the localization accuracy
of the human auditory system [Blauert, 1997; MacPherson, 2002]. It has been shown
that the region of most precise spatial hearing lies in the forward direction with frontal
hearing having an accuracy of between 4.40 and 100 for different source signals.
Localization ability decreases as the source azimuth moves to the sides, with the
localization blur at 900 being between three to ten times its value for the forward
direction. For sources to the rear of the listener, localization blur improves somewhat
but is still approximately twice that for frontal sources [Blauert, 1997]. The
assessment of dynamically moving sources is often measured in terms of the
minimum audible movement angle (MAMA), which is the smallest difference in
source direction perceptible by the listener. Studies have found that the MAMA is
independent of the source trajectory but that its resolution is a U-shaped function of

69

source velocity which degrades as the velocity increases or decreases beyond the
optimum range [Saberi et al, 1990].

6.2 Assessing Stereophony

As discussed in earlier in Chapter Three, the basic stereophonic principle
cannot reliably produce phantom images to the sides or rear of a listener. To produce
a predictable phantom image, the listener must be facing a pair of equidistant
loudspeakers separated by not more than 600. For a lateral loudspeaker pair, small
changes in signal level result in large changes in perceived direction and localization
is highly unstable (see Figure 3.4). The basic Quadraphonic system, which extends
the loudspeaker angle to 900, cannot therefore produce reliable phantom images, even
for a single listener. Theile recommended a six channel system to overcome these
problems as this arrangement uses an additional loudspeaker at +/-900, and hence
maintains the ideal loudspeaker separation angle of 600 [Theile et al, 1976]. The
results of listening tests carried out under anechoic conditions with this arrangement
and a single, central listener are shown in Figure 6.1. Clearly this arrangement
improves localization accuracy for lateral sources. However, distortions still occur
for phantom images positioned at either side of the loudspeakers at 900.

Fig. 6.1 Direction of phantom source versus ILD reported by Theile

Ville Pulkkie carried out a number of experiments to assess the localization

accuracy of phantom images created using the VBAP system [Pullkie et al, 2001]. A
number of different loudspeaker arrangements were examined in a large anechoic

70

chamber using both listening tests and a binaural auditory model which calculated
localization cues from the signals arriving at the ear canals. The results indicated that
the panning direction accurately matched the perceived direction when the centre of
the loudspeaker pair is near the median plane, but degrades as the loudspeaker pair is
displaced laterally. In general, a bias toward the median plane was reported for
sources produced by laterally biased loudspeaker pairs. Similar results were obtained
for a variety of different source signals [Pulkki, 2002]. The results of tests using
vertically orientated loudspeaker pairs suggest that the perception of the elevation of
amplitude panned phantom images varies widely from person to person. Only if a
virtual source is positioned at the same elevation as a loudspeaker can it be reliably
localized in that direction.
Grohn also carried out listening tests to assess the localization of both static
and dynamic sources in an immersive virtual environment [Grhn, 2002]. The VBAP
system was used with fourteen loudspeakers which were placed in a non-symmetrical
arrangement due to the presence of various graphical displays. A variety of source
signals were presented while subjects were asked to indicate the position and
movement of the virtual source using a tracked baton pointer method. As expected,
localization was best when the source was at a loudspeaker position, and worst when
panned between loudspeakers. The trajectory of dynamic sources therefore tended to
bend toward the loudspeaker positions and similar results were reported for all test
signals. In an additional test, it was found that a second distracting stimulus
decreased localization accuracy in line with other similar experiments [Blauert, 1997],
but only if the distracting signal was at least 10-15dB less than the target signal.

6.2.1 Discussion
The results of the tests discussed in the previous Section clearly suggest that a
minimum of six loudspeakers is required to ensure reasonably accurate localization
for a single, central listener. In addition, the quality of lateral sources in quadraphonic
systems decreases significantly as the size of the array is increased to accommodate a
distributed audience (see Section 10.2). Other tests suggest, perhaps unsurprisingly,
that localization accuracy increases as more and more loudspeakers are added [Ballas
et al, 2001].

71

The performance of stereophonic systems in terms of localization is highly

dependent on the position of the phantom image. If the source is panned to a
loudspeaker position, then only that loudspeaker will be used, ensuring that
localization is as accurate as possible. However, when the source is panned between a
loudspeaker pair, localization accuracy degrades significantly, especially in lateral
directions. The use of a greater number of loudspeakers would reduce the extent of
the degradation in localization accuracy but this solution will not eliminate the
resulting panning artefacts which arise for dynamically moving sources. In this case,
as the number of contributing loudspeakers is constantly changing, a small yet clearly
perceptible timbral shift occurs, which tends to emphasize the loudspeaker positions
and distort the perceived trajectory. This effect is almost certainly due to the
increased ASW exhibited by stereophonic phantom images compared to a source
produced by a single loudspeaker. This occurs due to the slight difference in source
direction which is perceived in stereophonic phantom images at low and high
frequencies (see Chapter 3, Section 3.4).

6.3 Assessing Ambisonics

J. Daniel carried out a number of experiments to investigate different first and
second order ambisonic decoding schemes [Daniel et al, 1998]. The decoders were
assessed using both HRTF measurements and binaural simulations, and listening tests
with four, five and six channel loudspeaker arrays. Under ideal conditions with a
single, centrally positioned listener a dual-band decoding scheme was found to be
preferable in terms of localization accuracy for the first order system. Similar results
were achieved with the second order system using a single band decode which
optimized rV. The results indicated that decoders which perform well in non-ideal
conditions (such as an off-centre listener position) are not the optimal low frequency
solutions (rV =1) designed for ideal conditions, but rather the max rE and in-phase
decoders. The tests also indicated that a homogeneous sound field can only be
created if there are sufficient loudspeakers for the particular ambisonic order, as
discussed in Section 4.4.
Benjamin conducted a number of listening tests to examine the differences
between different ambisonic decoders and loudspeaker layouts [Benjamin et al,
2006]. First order Ambisonics material consisting of Soundfield recordings, and

72

various encoded test signals were played back using square, rectangular and
hexagonal arrays in a medium-sized, acoustically treated listening room.
Interestingly, the initial tests were carried out in an ordinary untreated room but were
abandoned as extremely poor localization was achieved. Four different decoding
schemes were examined, a full-band velocity decode, a full-band energy decode, a
full-band in-phase decode, and a dual-band energy/velocity decode based on Gerzons
original design and with a transition frequency of 400Hz. Subjects were free to move
around within the array and were asked to listen for attributes such as directional
accuracy, perspective, timbral changes, artefacts and loudspeaker proximity effects.
Overall, the majority of test subjects preferred the hexagonal array with dual-band
decoding. The square layout was least preferred due to poor lateral imaging and
spectral changes for different source directions. The rectangular layout was found to
work well for frontal sources with rear ambience. In terms of the decoding scheme,
the velocity and in-phase decoders were least preferred for opposing reasons. The
velocity decoder produced uncomfortable in-head imaging and comb-filtering,
probably due to the high frequency anti-phase components which would be present in
a full-band velocity decode. The in-phase decoder on the other hand was judged to be
much too diffuse and reverberant, although comb filtering and artefacts due to listener
movement were eliminated. The full-band energy decoder was judged to provide a
balance between these two extremes and was found to work well at off-centre listener
positions. However, the shelf filter decoder produced more defined sources as it
appeared to pull the various spectral components of the signal to the same perceived
direction. An interesting general finding was that the loudspeaker layout is
significantly more important than the choice of decoder. The results of the initial
failed test would suggest that the acoustics of the listening room are also highly
important.
Guastavino conducted a number of listening tests which compared 1D, 2D and
3D ambisonic presentations [Guastavino et al, 2004]. In an acoustically treated room
containing six symmetrically arranged loudspeakers in a hexagonal formation with
two sets of three loudspeakers arranged above and below. The twenty-seven expert
listeners subjects were first asked to rate various ambient recordings made with a
Soundfield microphone decoded using a full-band in-phase decoding scheme. The
test results show a strong preference for the 2D, hexagonal layout in terms of
naturalness, source distance, envelopment and timbral coloration. The 3D schemes
73

were described as sounding further away, indistinct and less enveloping while the 1D
scheme was found to be the most stable with listener movement. In a second
experiment, a more directive decoding scheme (similar to a max rE scheme) was used
as this provided a better balance between localization accuracy and sensitivity to
listener position [Guastavino et al, 2004]. Similar results were achieved in both tests
and an analysis of the results suggests that the preferred layout, at least in terms of
naturalness, is dependent on the source material (see Figure 6.2 [Guastavino et al,
2004]). The 3D layout appeared to be preferred for indoor environments, while the
2D layout was preferred for outdoor scenes and the 1D scheme for frontal music
scenes.

Fig. 6.2 Naturalness responses reported by Guastavino

Kratschmer conducted informal listening tests with a number of ambisonic

decoding schemes and a forty-eight loudspeaker array [Kratschmer et al, 2009]. The
results suggest optimal performance in terms of localization accuracy is achieved
when the number of loudspeakers is matched to the Ambisonics order, using the
formula given earlier in this section. The results of this test and others [Bertet et al,
2009] suggest therefore that the performance of an Ambisonics system decreases
significantly when the number of loudspeakers greatly exceeds the minimum number
required for that particular order.

74

David Malham published an informal paper outlined his experience working

with large area Ambisonic systems for theatre and music performances [Malham,
1992]. He notes that positioning an audience within a periphonic array can be
problematic due to acoustic screening of the lower loudspeakers by other audience
members. He also suggests that the decoding scheme and loudspeaker layout often
needs to be manually adjusted to compensate for the effect of the room acoustic.

6.3.1 Discussion
The results of these tests confirm Gerzons original proposal in that a dualband decoder which optimizes the velocity and energy vectors is preferred when there
is a single listener. However, when off-centre listener positions are taken into
account, decoders which optimize rV are least preferred due to the significant antiphase components which are required to maximize the velocity component. The inphase decoding scheme eliminates these anti-phase components entirely and so is
very stable across a wide listening area, but is also very diffuse. The max-rE decoder
represents a good compromise between these two extremes, particularly at higher
orders. As with stereophony, it appears that Ambisonics requires a minimum of six
loudspeakers for optimum performance.

Fig. 6.3 Decoder criteria related to the size of the listening area

Daniel proposes that for a given order and distance from the centre of the
array, the max-rE decoding scheme is most suitable [Daniel, 2000]. If the listening
area extends beyond this distance, or to the loudspeaker periphery, then the in-phase
scheme is preferred (see Figure 6.3 [Daniel, 2000]). He proposes a tri-band decoding

75

scheme which applies the basic, max-rE and in-phase decoders in three consecutive
frequency bands, based upon the size of the listening area.

6.4 Comparing Stereophony and Ambisonics

Bamford used various analytical techniques to assess the ability of stereo,
5.1 and first and second order Ambisonics to recreate a plane wave [Bamford, 1995].
In general, it was found that the second order Ambisonics system produced the best
results and was more consistent for different source angles. The results indicate that
the upper frequency at which a plane wave is accurately reconstructed at the centre of
the array, is increased when as the order of the Ambisonics system is raised. The
performance of both Ambisonics systems deteriorated at higher frequencies but this
effect was reduced with the higher order system.
Martin investigated the focus or source width of phantom images produced
using amplitude panning, first and second order Ambisonics and some non-standard
amplitude panning functions [Martin et al, 199b]. Listening tests were carried out in
an acoustically treated, medium sized room with a small number of test subjects and a
symmetrical eight channel array which was also used to produce interaural crosscorrelation measurements. Overall, pair-wise amplitude panning was found to
produce the most focussed source as it used a single loudspeaker when the source is
positioned at a loudspeaker position, and a maximum of two loudspeakers in other
directions. On the other hand, this produced a noticeable change in timbre as the
source direction moves across a loudspeaker position. The second order Ambisonics
system and a polarity restricted cosine panning function produced the most consistent
images across all angles. The polarity restricted cosine function is essentially
amplitude panning with a consistent number of loudspeakers, similar to the spread
function in Ville Pulkkis VBAP. Martin recommended this algorithm as it produced
the most consistent results and appeared less sensitive to listener position than the
second order Ambisonics system. It should be noted that very little detail is given in
terms of the precise ambisonic decoding scheme adopted. However, Martin mentions
that anti-phase components were a disturbing factor in the listening tests with
Ambisonics, which suggests that a basic decoding scheme was used which optimizes
the velocity component rV. It is perhaps not surprising, therefore, that the restricted

76

polarity cosine function was preferred, as this method contains no disturbing antiphase components.
Dickins analysed the performance of first and second order Ambisonics and a
custom non-negative least squares (NNLS) panning algorithm by measuring the
directional energy vector magnitude rE [Dickins et al, 1999]. Two listening locations
have been considered, one in the sweet spot at the centre of the array, and another
towards the rear of the array. The tests were carried out in an acoustically treated
listening room. Martin suggests that in general a compromise must be made between
optimizing the directionality of the source, and minimising panning artefacts as the
source moves. The NNLS algorithm is therefore similar to Martins polarity
restricted cosine function and Ville Pulkkis VBAP in that it allows a trade-off
between maximum directivity at loudspeaker positions and a more diffuse panning
which is homogeneous in all directions. The NNLS algorithm was preferred to the
second order Ambisonics system as it functioned well at off-centre listening positions
and could be extended to non-symmetrical loudspeaker arrays. However, as with the
tests carried out by Martin, little detail is given regarding the ambisonic decoding
scheme used. A strict decoding scheme which optimized rV would be expected to
function poorly away from the sweet spot, while a decode which optimized rE would
be much more similar to the NNLS algorithm and would provide a better comparison.

Fig. 6.4 Mean ratings as reported by Guastavino

Guastavino conducted a number of listening tests which compared twochannel stereo, transaural and b-format Ambisonics using six symmetrically arranged
77

loudspeakers within an anechoic chamber [Guastavino et al, 2007]. A free

verbalization task and a multiple comparison task was conducted with eleven
experienced listeners. The subjects were first asked to rate recordings made with a
Soundfield microphone, an ORTF stereo pair and binaural microphone in terms of
various subjective measures such as envelopment, readability (meaning how well
different sources in the scene can be distinguished) and naturalness. and an overall
rating. The Soundfield recording was decoded using a single band in-phase decoding
scheme. In a second experiment, subjects were asked to rate various monophonic
signals positioned at various angles using amplitude panning, a Soundfield recording
and a customised Transaural system. The results of the two experiments, shown in
Figure 6.4 [Guastavino et al, 2007], indicate a strong contrast between Ambisonics
and the other two techniques in that stereophony and transaural provide precise
localization and a good readability but a lack of immersion and envelopment while
Ambisonics provides a good sense of immersion and envelopment but poor
localization accuracy and readability of the scene. In a similar study conducted by
Capra (see Figure 6.5) slightly better results were achieved for an Ambisonics system
implemented using a dual-band shelf filter decoder [Capra et al, 2007].

Fig. 6.5 Localization accuracy results as reported by Capra

Pulkki carried out a number of listening tests in order to assess the validity of
a binaural auditory model [Pulkki et al, 2005]. The experiment was conducted in an
anechoic chamber using a symmetrical eight-channel array, first and second order

78

Ambisonics, amplitude panning (VBAP) and a spaced microphone technique.

Although the primary aim of the experiment was to verify the binaural model, some
interesting results were obtained. Lateral sources produced with the first order
Ambisonics system were found to be consistently biased toward the median plane and
exhibited a strong frequency dependence. The second order system exhibited almost
no bias, reduced frequency dependence and increased localization accuracy. The best
average localization accuracy was obtained with amplitude panning, although the
results for second order Ambisonics were equally good for frontal sources.
Jot compared a number of different amplitude panning and ambisonic
techniques in terms of a number of objective localization criteria [Jot et al, 1999].
The comparisons were based on a single listener at the centre of a hexagonal
loudspeaker array. Noticeable improvements are obtained when raising Ambisonics
from first-order to second order, with an average performance comparable to that of
the six-channel pair-wise panning but more uniform across all azimuths. While
amplitude panning produced more stable and accurately localized images at
loudspeaker positions, it was less uniform and tended to reveal the loudspeaker
locations. No technique produced particularly accurate localization cues at high
frequencies.

6.4.1 Discussion
The results of the tests presented in the preceding Section indicate that
Ambisonics is consistently preferred to amplitude panning for dynamically moving
sources as it produces a more uniform phantom image and hence disguises the
loudspeaker position. However, amplitude panning was also consistently preferred
for static sources as this method uses fewer loudspeakers and so reduces the
localization blur. This would seem to support Martins view that in general a
compromise must be made between optimizing the directionality of the source, and
minimising panning artefacts as the source moves [Martin et al, 1999a]. The results
which indicated that Ambisonics produced a more diffuse enveloping sound field but
less tightly focussed sources is arguably another interpretation of the same
fundamental difference between the two spatialization techniques.
A number of alternative amplitude panning techniques were presented which
attempt to reduce the timbral changes produced when a source is dynamically panned

79

to different positions [Pulkki, 2005; Martin, 1999a, Dickins et al, 1999]. These
techniques are similar in that they can control the number of contributing
loudspeakers independently of the source azimuth. In this way, they can ensure that a
minimum number of loudspeakers is always used, which then smoothes the perceived
trajectory. This is clearly very similar to the Ambisonics approach of optimizing the
energy vector rE for all directions, at the cost of reducing the maximum localization
accuracy that could be achieved at the loudspeaker positions. Pernaux point out that
amplitude panning algorithms like VBAP can be considered as analogous to a local
ambisonic velocity decode whereby only the loudspeakers closest to the source
direction are used, and the requirement to optimize the velocity component (rV = 1) is
dropped [Pernaux et al, 1998]. They go on to develop a dual-band Vector Based
Panning algorithm (VBP) which uses VBAP at low frequencies and a Vector Based
Intensity panning algorithm (VBIP) at high frequencies, which like Ambisonics,
ensures that E = V. The significant advantage of this system over Ambisonics is that
appropriate decoding factors can be more readily obtained for non-regular
loudspeaker arrays.

Fig. 6.6 Ambisonic decoder directivity patterns where M is the system order

These advanced amplitude panning schemes are highly reminiscent of

Ambisonics, and particularly higher order Ambisonics systems which are optimal for
larger listening areas. Higher order systems increase the directivity of the response
characteristic (see Figure 6.6 [Daniel, 2000]) which in turn reduces the number of
contributing loudspeakers. When this is combined with a decoder which reduces anti80

phase components Ambisonics becomes highly similar to these advanced amplitude

panning techniques such as VBP. However, some tests have indicated that
Ambisonics was still preferred to VBP for moving sources [Pernaux et al, 1998].

6.5 The Influence of the Listening Room

The vast majority of the tests discussed so far have taken place in acoustically
treated listening rooms or anechoic chambers, thereby eliminating the potentially
detrimental influence of the room acoustic on the recreated sound field. This is
particularly relevant in the context of spatial music as many performance spaces
contain significant lateral reflections and reverberation, and will also contain
numerous listeners seated away from the centre of the loudspeaker array. It is known
that reverberation and reflections can reduce localization accuracy with real sources
[Hartmann, 1983; Hartmann, 1985] and it is reasonable to assume that virtual source
localization will be similarly affected. Furlong examined the influence of the
listening room and loudspeakers on sound field reconstruction using a coupled room
acoustic simulator which allowed user control over such parameters as listening room
geometry and reflection coefficients, listener position, and loudspeaker number,
position, orientation, and directivity [Furlong et al, 1992]. The simulations suggested
that placing the loudspeakers too close to either the listeners or the room boundaries
will have a significant effect on the perceptual performance of multichannel systems
like Ambisonics. In order to verify these simulated results, a series of listening tests
was carried out by the author in a small concert hall using various spatialization
techniques such as VBAP, Delta Stereo and first and second order Ambisonics.

6.5.1 Spatialization Assessment in a Small Concert Hall

In these listening tests various test signals were presented at eight discrete
locations for multiple listeners in an attempt to ascertain the relative level of
localization blur [Bates et al, 2007b]. The tests were carried out in a small concert
hall (shown in Figure 6.7) with a reverberation time (RT60) of 0.9secs at 1kHz, using
a loudspeaker array consisting of sixteen Genelec 1029A loudspeakers arranged
around a nine listener audience area. In a control test, monophonic signals were
played back using the eight black loudspeakers shown in Figure 6.8 while the other

81

dummy loudspeakers were used to increase the choice of angle for the listeners
[Bates et al, 2007b].

Fig. 6.7 Listening tests loudspeaker configuration

Fig. 6.8 Geometry of loudspeaker array and audience area

Although many spatialization techniques are optimized for regular equidistant

loudspeaker arrays, this can be difficult to implement in concert halls which are often
rectangular in shape. When the loudspeaker array is squashed to maximize the

82

available seating area in a rectangular shaped room (such as the one used for this test),
delay and gain adjustments must be made to displaced loudspeakers. In this test, the
appropriate delay was applied to each of the two lateral loudspeakers when encoding
the test signals. The gain adjustments were applied to these two loudspeakers by
calibrating each loudspeaker in the array to 70dBA at the centre listening position.
This approach is preferable to using the inverse square law when operating in a
reverberant acoustic environment, due to the superposition of the direct and
reverberant sound affecting the total SPL.
In order to assess the effect of various stimuli, users were presented with one
second unfiltered recordings of male speech, female speech, Gaussian white noise and
a music sample containing fast transients. The results indicate that for most
combinations of listening and source position, the influence of the room is not enough
to cause a listener to incorrectly localize a monophonic source away from the desired
location when using an asymmetrical 8-speaker array. However, for extreme cases
such as a front-corner listening position with a source positioned to the rear, the
degree of accuracy becomes heavily dependent on the nature of the source signal.
In a second test, first and second order Ambisonics, VBAP and Delta
Stereophony were assessed using the same loudspeaker configuration and a forcedchoice, speaker identification method. The first order Ambisonics was implemented
using IRCAMs Spat software and a traditional, dual-band decoding scheme. The
second order system was implemented using the ICST externals for Max/MSP and a
max-rE decoding scheme devised by David Malham specifically for horizontal eight
channel arrays [Schacher et al, 2006]. Delta Stereophony (DSS) is a sound
reinforcement system which has been used successfully in large auditoria. The main
objective of DSS is to reinforce the original direct sound while also ensuring accurate
sound source localization. This can be achieved if the listener at any place in the
room receives the first wavefront from the direction of the sound event being
reinforced, rather than from any of the other loudspeaker positions [Fels, 1996].
The results of the listening tests were verified using calculated ITDs inferred
from high resolution binaural measurements recorded in the test environment. An
equivalent acoustic model was also implemented to investigate specific aspects of the
effect of the room acoustic. The results indicate that neither amplitude panning or
Ambisonics can create consistently localized virtual sources for a distributed audience
in a reverberant environment. Source localization at non-central listener positions is
83

consistently biased toward the nearest contributing loudspeaker in the array,

irrespective of the spatialization technique or source stimulus. Both B-format and
Second Order Ambisonics exhibited a consistent localization bias towards the nearest
loudspeaker and due to the greater number of contributing loudspeakers in these
systems, this resulted in significant distortions in the perceived source angles. This
can be clearly seen in the results for a source position at loudspeaker 14 which are
shown in Figure 6.8.

Fig. 6.9 Reported (blue) and actual (orange) direction for a source at speaker 14

An equivalent acoustically modelled environment was implemented using the

EASE developer package in order to look at the influence of room acoustical
response. The constructed model closely represents the absorption and reverberation
characteristics of the real hall. Impulse response measurements were taken using
maximum length sequence (MLS) noise at particular points in the hall to verify the
accuracy of the model. The measured pressure levels and arrival times of the direct
sound and early reflections were found to be comparable to the simulations in EASE.
All systems showed small areas where the SPL drops significantly (more than 10dB).

84

This can be attributed to the superposition of loudspeakers presenting the source

material whilst significantly exciting the room modes. Furthermore, each system
exhibits large areas where the level of the reverberation is greater than that of the
direct sound, which could also adversely affect localization. In all cases, the
localization accuracy increases with distance from the source which clearly shows the
importance of maintaining sufficient distance between the listeners and the
loudspeaker array.
Each system was also analyzed in terms of its subjective hit rate, calculated by
correlating the ideal localization histogram with the observed results. This measure,
shown in Figure 6.9 expresses the percentage localization accuracy over all source
positions and stimuli and provides a measure of the overall performance of each
spatialization technique. It is clear that the localization performance of each system
does not achieve that of monophonic, point sources. Overall, the intensity panning
systems perform better for front and back sources, with VBAP providing 12.7%
higher localization accuracy over DSS for rear sources. Second Order Ambisonics
performs consistently better than B-format at all source positions and its performance
is comparable to VBAP for lateral rear sources.

Fig. 6.10 Overall subjective localization performance

The results of this test suggest that accurate localization of phantom sources is
difficult to achieve for multiple listeners in a reverberant environment. While

85

acoustic reflections and reverberation in the listening room probably contribute to the
overall reduction in localization accuracy, particularly at certain listener locations, the
results shown in Figure 6.10 would seem to indicate that this is not the primary
negative influence on the performance of the various systems. At locations away
from the centre of the array, the temporal relationship between the contributing sound
waves is distorted and so the precedence effect effectively dominates source
localization. The proximity of the loudspeaker array to certain listeners results in
phantom images which are consistently localized to the nearest contributing
loudspeaker. The best results are therefore achieved with VBAP, as this system only
ever uses a maximum of two loudspeakers, while the worst results were achieved with
b-format as this system uses nearly all of the loudspeakers to produce a phantom
image. The increased directivity of the second order Ambisoncs, max rE decoding
scheme performs better than the first order system particularly at off-centre listener
positions.

Fig. 6.11 Ambisonics system evaluated by Frank

Frank et al investigated the subjective localization accuracy of a virtual source

created using the twelve channel, nearly circular 2D Ambisonics system shown in
Figure 6.11 [Frank et al, 2008]. The room used for the test is a medium sized 12m x
10m room with some acoustic treatment. First, third and fifth order Ambisonics
decoders were examined using basic, max-rE, and in-phase decoding schemes. As
shown, test subjects were placed at two different listening positions. The results
86

indicated that localization improves as the order increases, independently of the

listener position. As expected, localization at the central listening position is more
accurate than at the off-centre position. The max rE decoder was found to provide the
best overall performance while the in-phase decoder was least preferred. Localization
was biased toward the left side, presumably due to the closer proximity of the listener
to the loudspeakers at the off-centre listener position. This proximity bias was
significantly reduced as the order increased. Interestingly, the delay adjustments
applied to compensate for the non-equidistant loudspeaker layout were found to be
detrimental to the perceived performance of the system. The authors suggest that this
unexpected result occurred due to pronounced phase distortions outside the sweet
spot. However, the exact reason why the uncompensated system performed better is
still uncertain.

Fig. 6.12 Quality maps for 5ms (left) and 50ms (right) time difference

A large number of experiments have been carried out to examine the influence
of the precedence effect [Wallach et al, 1949; Litovsky et al, 1999]. It has been found
that a number of correlated sound waves arriving in close succession will be fused
together and perceived as a single sound at a single location. The duration over which
fusion will take place is highly dependent on the nature of the source signal, but
studies have found an approximate lower limit of 5ms for transient sounds and an
upper limit of 50ms for broadband sounds [Litovsky et al, 1999]. Moore used these
limits to evaluate a number of listener positions with the ITU 5.1 layout in terms of
the precedence effect. By checking the time difference between each loudspeaker
pair, a mark out of ten was given for each position in the listening area allowing
performance to be quantified at different positions in the reproduction area [Moore et
al, 2007]. The results for a reproduction area of 20m2 are shown in Figure 6.12.
87

They indicate that transient signals will only be correctly localized in a relatively
small central listening area due to the large travel distances involved in such a large
array. A large effective listening area is shown for a 50ms time difference and the
author suggests that the effective listening area could be extended using a form of
transient suppression applied to the output signals [Moore et al, 2007].
It has been suggested that altering the directivity characteristic of the
loudspeakers can help to reduce the detrimental effect of the room acoustic on
stereophonic source localization. The results of a series of listening tests carried out
by Harris suggested that diffuse acoustic radiators such as Distributed-Mode
Loudspeakers (DML) reduce the degradation caused by room acoustics on
stereophonic localization [Harris et al, 1998].

Zacharov examined the effect of

loudspeakers with different directivity characteristics in terms of source localization,

envelopment and naturalness [Zacharov, 1998]. In a number of listening tests more
directional loudspeakers were preferred for all parameters and for all listening
positions. Zacharov suggests that the increased directivity of these loudspeakers
reduces the excitation of the listening room modes and hence reduces the negative
impact of the room acoustic on stereophonic reproduction. However, further tests are
required to see if a similar approach could improve the performance of Ambisonics
under similar conditions.
It is clear that unless a certain minimum distance is maintained between the
listeners and the loudspeakers, the precedence effect will ensure that localization will
collapse to the nearest contributing loudspeaker.

While it is clear that the

reverberation and early reflections in the listening room also influence the perception
of virtual sources, the extent and exact nature of this influence is more difficult to
define as this will entirely depend on the exact dimensions and layout of the particular
space.

6.6 Artificial Reverberation and Spatial Dissonance

The preceding chapter illustrated how the simulation of distance is primarily
achieved through the addition of artificial reverberation and early reflections. It has
been shown that the acoustics of the listening room affect auditory perception and so
it must be expected that the addition of artificial reverberation will also have some
influence on the perceived source. Begault investigated the effect of synthetic

88

reverberation on three dimensional audio samples synthesized using HRTF data

[Begault, 1992]. He found that the error between perceived and actual source angle
increased with the addition of reverberation and suggested that this was due to the
effect of early reflections, in line with earlier studies [Hartman, 1985]. The
directionality of early reflections was also investigated but the results found no
consistent relationship between the angle of early reflections and perceived source
angle. A consistent relationship between reverberation and source distance was found
and the addition of artificial reverberation dramatically increased the effective
externalization of the synthesized sound scene. Although the results of this study
using binaural techniques cannot be directly compared to multi-channel loudspeaker
systems, it does seem to agree with the results of other studies and suggests that there
is a certain trade-off between the simulation of distance cues and the production of a
reliable source direction.
Martin investigated the influence of the angular reflection pattern of artificial
early reflections and reverberation [Martin et al, 2001]. The reflections of any wave
can be classified into two groups depending on the precise angle of reflection. If the
reflective surface is large and flat relative to the wavelength of the reflected sound,
then specular reflections will be produced according to a predictable angle as shown
in Figure 6.13 (left). If the surface is irregular, then the incident wave is scattered and
multiple diffuse reflections are produced instead of a mirror image reflection, as
shown in Figure 6.13 (right).

Fig. 6.13 Specular and diffuse reflections

The results of an initial test indicated that there was an easily recognizable difference
between signals processed using a specular and diffused reflection model, whether a
reverberant tail was included or not. In general, a mix of specular and diffused
reflection models was found to be preferable to a typical perfectly specular reflection
model. Martin also suggests that the inclusion of a height component in a synthetic
89

room model improved the attractiveness and realism of the resulting sound, even
when reproduced using a two-dimensional loudspeaker configuration [Martin et al,
2001].
It is still unclear how the addition of artificial reverberation affects auditory
perception at non-central listener positions. Lund developed a 5.1 surround sound
spatialization system based on amplitude panning with suitable spatially distributed
early reflections and diffuse reverberation [Lund, 2000]. He suggests that the
localization of virtual sources will be made more robust and accurate if the desired
source location is supported by accurately modelled early reflections and
reverberation. The results of a listening test with a very small number of participants
(five listeners) seem to support this view as the system produced more consistently
localized phantom sources and greatly increased the effective listening area [Lund,
2000]. However, further tests with a greater number of subjects are required to fully
verify these results.
Michelsen found that the addition of artificial reverberation and early
reflections produced a clear and unchanged sense of distance in both an anechoic and
reverberent listening room [Michelsen et al, 1997]. This suggests that the listening
room acoustic is not significantly detrimental to the artificial simulation of distance
cues. However, the perceptual effect of combining artificial reverberation with the
natural reverberation of the listening room itself is unknown. The composer Denis
Smalley suggested that this would produce a negative cognitive effect, which he
refers to as spatial dissonance, as the auditory system would be presented with
localization cues that suggest two, potentially conflicting, acoustic spaces. However,
very few listening tests have been undertaken to ascertain how detrimental this effect
actually is. In the authors experience, the layering of two synthetic reverberation
effects does not create two clearly perceivable acoustic spaces but instead results in
one, slightly indistinct space. Clearly, the simulation of distance is highly dependent
on the use of artificial reverberation and it is hard to see how else this effect could be
implemented other than the physical placement of loudspeakers at different distances.
As discussed in Chapter Three, wavefield synthesis differs from stereophony and
Ambisonics in that it can theoretically recreate virtual sources at different distances by
synthesizing the correct wavefront over an extended area. The evaluation of this
system and the veracity of these claims will be examined in the next section.

90

6.7 Evaluating Wavefield Synthesis

It is well known that the distance between adjacent loudspeakers in a WFS
system is directly related to the spatial aliasing frequency [Berkout, 1998]. This
frequency is one of defining characteristics of any WFS system and represents the
upper frequency limit for accurate reconstruction of the synthesized wavefront.
Clearly, determining a suitable lower limit for the spatial aliasing frequency is an
important area for investigation as this directly influences the necessary size of the
loudspeaker array, and hence the listening area. In addition, while various techniques
exist to reduce high frequency distortion in WFS virtual sources, it is important to
assess the perceptual effects of these techniques. For example, while certain
techniques reduce the timbral coloration introduced by this system, this can also
influence the directional localization.
The ability to position virtual sources behind and in front of the loudspeaker
array is an often lauded feature of WFS and this inherent ability to reproduce distance
is unique. However, much of these claims are based on simulations, such as the
example shown in Figure 3. 11 and subjective listening tests are required to verify
these claims.
Finally, the influence of the listening room acoustic has had a significant and
detrimental effect on other spatialization systems and it is expected that WFS will be
no different.

6.7.1 WFS Localization Accuracy and Spatial Aliasing

One of the primary goals of any spatialization system is to ensure that each
listener can clearly and accurately perceive the direction of the virtual source, in any
direction. Most practical systems eliminate the height component and so simply
require a number of loudspeakers to be placed around the audience. WFS differs
from other systems in that it is based on linear arrays of loudspeakers rather than a
small number of distributed loudspeakers. A WFS system which can position a
virtual source in any direction in the horizontal plane requires therefore a very large
number of loudspeakers placed in a linear array surrounding the audience. This was
demonstrated by Verheijen who found that accurate synthesis is not possible for
sources which do not lie in a straight line from the listener to the source, through the
array [Verheijen, 1998]. The only area in which a virtual source can be correctly
91

synthesized is between two lines from the listener to the edges of the array, as shown
in Figure 6.14. This area is further reduced if tapering windows are applied to the
array to reduce truncation effects (see Section 3.2.2).

Fig. 6.14 Visibility of a virtual source in a WFS system

WFS can only accurately synthesize a sound field up to the spatial aliasing
frequency, so clearly the high frequency content will be distorted in some way. Start
reported that the imperfect reconstruction of the sound field above the spatial aliasing
frequency gives rise to an increase in the apparent source width (ASW) due to the
uncertain directionality of the high frequency content [Start, 1997]. Wittek points out
that this must be taken into account when evaluating the localization accuracy of any
WFS system [Wittek, 2003]. For example, a measure of the standard deviation in the
reported directional data will not indicate whether listeners perceive a tightly focussed
source within the range of directions reported, or a broad diffuse source distributed
between the range of reported angles. As discussed earlier in Chapter Two,
locatedness is a measure of the degree to which an auditory event can be said to be
clearly in a particular location [Blauert] and this parameter is sometimes used in the
assessment of WFS systems to determine the focus or apparent width of the virtual
source.
Some of the earliest perceptual experiments with WFS systems were carried
out by Vogel [Vogel, 1993]. In an experiment with an array of twelve loudspeakers
each separated by 45cm, he found that correct directional localization was maintained
92

despite the very low spatial aliasing frequency of 380Hz of this system. However,
Wittek points out that as this system can only correctly synthesize frequencies below
380Hz, it cannot be assumed that WFS is responsible for the correct localization
[Wittek, 2003]. For a WFS virtual source positioned behind the array, the
loudspeaker nearest to a direct line from the listener to the source will be producing
the earliest and often the loudest signal. The precedence effect would therefore
provide a localization cue at all frequencies, which, in this case, coincides with the
source position specified by the WFS system. The mean directional error reported in
Vogels test is no lower than what would be expected if the precedence effect was
dominating localization and so these results do not indicate that localization accuracy
is improved by this particular WFS system. This situation does not occur with
focussed sources in front of the array, as in this case the first wavefront does not
arrive from the same direction as the virtual source [Wittek, 2003].
Since these early tests further experiments have been carried out with
loudspeaker arrays of greater size and resolution [Vogel, 1993; Huber, 2002]. The
results of these tests demonstrated that localization accuracy was greater than the
actual physical resolution of the loudspeaker array, no doubt due to the increased
resolution of the array and the associated increase in the spatial aliasing frequency.
The results of both these tests clearly indicate the importance of the spatial aliasing
frequency in terms of the performance of the WFS system.
Start compared the minimal audible angle (MAA) of real sources and virtual
sources produced using a WFS system [Start, 1997]. He found no difference between
the MAA of a real source and the WFS source for a spatial aliasing frequency of
1.5kHz, for both broadband and low-pass-filtered noise signals. When the spatial
aliasing frequency was reduced to 750Hz however, the MAA increased somewhat.
Start suggested that this result implied that a spatial aliasing frequency of 1.5kHz
would ensure that the dominant low frequency localization cues are satisfied and so
the source will be accurately localized.
Huber conducted listening tests in an anechoic chamber to compare real
sources, two-channel stereo, WFS with loudspeaker spacings of 4cm and 12cm, and
an augmented WFS system based on Optimized Phantom Source Imaging (OPSI)
[Huber, 2002]. OPSI uses amplitude panning to position the portion of the signal
which lies above the sampling aliasing frequency [Wittek, 2002], thereby reducing the
artefacts which occur due to the incorrect reconstruction of the high frequency signals
93

in WFS systems. Figure 6.15 shows the scaled judgements of locatedness (which
Huber refers to as localization quality) for each of the five systems. Clearly none are
able to match the real source in terms of locatedness, however, a significant
improvement is apparent when the loudspeaker spacing is reduced to 4cm (which
results in a aliasing frequency of 3kHz). The worst results were achieved with stereo
while the hybrid OPSI method was found to produce approximately the same results
as the normal WFS system.

Interestingly, the standard deviation in localization

accuracy shown in Figure 6.16 does not indicate any differences between the real and
WFS sources, which indicates the importance of assessing the perceptual sense of
locatedness as well as the perceived direction.

Fig. 6.15 Subjective, scaled (1-5) judgments of locatedness reported by Huber

Fig. 6.16 Standard deviation in horizontal localization reported by Huber

94

A similar comparison test was carried out by Wittek on the localisation

performance of various WFS systems compared to a real source, amplitude panning
and a hybrid OPSI system (see Figure 6.17) [Wittek, 2007]. For the WFS systems, a
virtual source was positioned one metre behind the array using loudspeaker arrays
with a spatial aliasing frequency of 2.5kHz (labelled WFS12) and 7.5kHz (labelled
WFS4) respectively. The OPSI signals were produced using the WFS12 system and
three loudspeakers at spacings of 76cm to produce the phantom source with a
crossover frequency of 2kHz.

Fig. 6.17 Test setup for Wittek's listening tests

Fig. 6.18 Subjective assessment of locatedness reported by Wittek

95

The results were similar to Huber's in that none of the systems matched the
performance of a real source in terms of locatedness, but better results were reported
with the WFS system than with amplitude panning (see Figure 6.18). As with Huber,
better results were achieved when the spatial aliasing frequency was increased from
2.5kHz to 7.5kHz. Similarly, these differences are not evident when only the standard
deviations of the measured auditory event directions are considered. No degradation
in localization quality was found using the hybrid OPSI method.
Sanson examined localization inaccuracies in the synthesis of virtual sound
sources using WFS at high frequencies [Sanson et al, 2008]. Objective and
perceptual analyses were carried out through a binaural simulation of the WFS array
at the ears of the listener using individual head related transfer functions (HRTFs).
The array could be configured for loudspeaker spacing of 15cm, resulting in an
aliasing frequency around 1500Hz, or a loudspeaker spacing of 30cm results in an
aliasing frequency around 700Hz. Two listener positions were evaluated, one central
and one laterally displaced to the right by 1m. The results of the test indicated that
localization accuracy was dependent on the listening position, the source position and
the frequency content of the source signal. Localization accuracy decreased as the
listener position moved laterally away from the centre point, As the source cut-off
frequency was increased, localization at the off-centre position degraded, but not at
the central listening position. The authors suggest that this is due to the unequal
distribution of high frequency content at either ear when the listener is positioned at a
non-central location. This would provide a conflicting localization cue relative to the
low frequency content which is accurately reproduced by the WFS system. Clearly,
technical solutions to the distorted high frequency content in WFS systems must
address localization for off-centre listener positions.

6.7.2 WFS in Real Rooms

Start carried out a number of experiments with a WFS system in various different
rooms such as an anechoic chamber, a medium sized auditoritum and a large concert
hall (see Figure 6.19) [Start, 1997]. The test signals consisted of speech, broadband
noise and low and high pass filtered noise (below and above the spatial aliasing
frequency). The author made the following observations based upon the results of a
number of listening tests in each room:

96

 The averaged RMS error for the low-pass-filtered noise signal is almost
identical for the real and virtual sources in the anechoic room but the results
for a real source are somewhat better in the auditorium and concert hall.
The best results were achieved with the speech signal
The averaged RMS error for the high-pass-filtered noise signal is much larger
for synthesized sources.
The localization performance of the WFS system is worse in the concert hall
while real sources are localized to the same degree of accuracy as the other
rooms.

Fig. 6.19 Concert hall layout used by Start

Start suggests that the reduction in localization accuracy of the WFS system in
the concert hall is primarily due to the lower spatial aliasing frequency (750Hz) of this
system. The standard deviation of the averaged perceived direction for each room is
shown in Figure 6.20. In the anechoic room and the auditorium a clear difference can
be seen in the results for the low and high pass filtered noise. In each case, the low
frequency signal (<1.4kHz and <1.2kHz, respectively) whether real or synthesized, is
localized more accurately than the high-pass-filtered signal. The localization
accuracy with synthesized sources is particularly worse with the high-pass-filtered
signal. However, this is not the case in the concert hall as here the worst results are
97

achieved with the low-pass-filtered signal (now <750Hz) for both real and
synthesized sources. These results suggest that the WFS system is not working
correctly in this particular room. Start suggests that this is solely due to the decrease
in the spatial aliasing frequency. However, the drastic reduction in performance for a
low-pass-filtered noise signal results suggest that the performance of the WFS system
is also significantly affected by the reproduction room acoustic.

Fig. 6.20 Perceived WFS virtual source directions reported by Start in

a) Anechoic Chamber
b) Auditorium, Delft University of Technology
c) Concert hall De Doelen, Rotterdam [Start]

98

Fig. 6.21 Virtual Source (a) behind the array & (b) in front of the array
gray bars: virtual sources (loudspeaker spacing = 11cm)
black bars: virtual sources (loudspeaker spacing = 22cm)
white bars: real sources [Verheijen]

A similar experiment was carried out by Verheijen using two different arrays
with spatial aliasing frequencies of 0.75kHz and 1.5kHz respectively, and virtual
sources behind and in front of the array [Verheijin, 1998].

The mean standard

deviation results, shown in Figure 6.21(left), indicate an improvement with the

increased array resolution in the anechoic room but not in the reproduction room.
This again seems to suggest that the influence of the reproduction room acoustic is as
significant a factor as the spatial aliasing frequency. The results for two test subjects
and a virtual source in front of the array are also shown in Figure 6.21(right). This
test was carried out in the anechoic room and a comparison of the results with a
source behind the array suggests that localisation accuracy decreases when the source
is positioned in front of the array.
Marentakis examined localization accuracy with WFS in a variable-acoustics
concert hall using the Minimum Audible Angle (MAA) as a measure of localization
performance [Marentakis et al, 2008]. The MAA was estimated for different listener
positions, listener orientations and varying acoustical conditions. The large hall could
be configured to adjust the reverberation time from 0.4 to 4sec and in this test was set
to an absorptive setting with a relatively low reverberation time, and a reflective
setting which increased the energy of side reflections and the overall reverberation
time. A 48-channel WFS system was placed in the hall at 7m from the rear wall and
4m height, as shown in Figure 6.22. A double logarithmic loudspeaker spacing was
99

utilized so that central loudspeaker spacing (24cm) was greater than lateral
loudspeakers (13cm). This unusual arrangement results in an aliasing frequency
which varies with the source angle and listener position, as shown in Figure 6.23.

Fig. 6.22 Test setup for Marentakiss listening tests

Fig. 6.23 Aliasing frequency relative to listener position reported by Marentakis

100

The results illustrated the general and expected trend that localization ability
decreases as the listener orientation changes from frontal to lateral orientation. This is
particularly evident for the listener position closest to the loudspeaker array, which
suggests that there is a limit to how close to the loudspeakers listeners can be placed
without a significant decrease in localization accuracy. The best results were obtained
for the central listener position while a slight reduction in accuracy was reported for
the listener position furthest from the array. As expected localization accuracy
generally decreased in the more reflective room but, in general, good localization was
achieved. However, Marentakis points out that signals with strong transient
characteristics, such as the enveloped white noise signals used in this test, are
localized independent of the room reverberation [Hartmann, 1983].

6.7.3 WFS Distance Effects

WFS differs from other spatialization techniques as it can theoretically
reproduce the wavefront curvature associated with a virtual source in front of, or
behind the loudspeaker array. As WFS recreates the sound field over an extended
area, the amplitude and location of a virtual source should therefore change in a
realistic fashion with listener movement. It has been suggested that this idea of
motion parallax, where the perspective changes naturally as the listener moves, can
provide an indication of the distance of a WFS virtual source.
Boone assessed the intelligibility of speech by comparing noise and speech
signals played back using a single loudspeaker, and two WFS virtual sources which
differed only in terms of their distance [Boone et al, 2003]. The results showed an
improvement in intelligibility when the speech and noise source were separated using
WFS which would seem to indicate that there is at least some perceptual difference
between two WFS virtual sources differing only in distance. However, Wittek points
out that this perceptual difference could occur due to differences in the way a single
loudspeaker and an array of loudspeakers interact with the acoustic of the
reproduction room [Wittek, 2003].
Nogues conducted an experiment which provided better evidence that the
movement of the listener through a WFS sound field does indeed provide some
indication of distance [Nogus et al, 2003]. They asked the subjects to control the
distance of a WFS virtual source so that the source distance matched the distance of

101

two other simultaneously reproduced virtual reference sources (see Figure. 6.24). The
subjects were asked to move around in the listening area throughout each test. In the
first experiment, the subjects could only manipulate the WFS source distance while
other parameters such as the direct to reverberant energy ratio and signal level were
kept constant. The results showed that the subjects were indeed able to position the
middle guitar in between the other two solely using the perspective cue of the WFS
sound field. This result suggests that the perception of distance due to the virtual
source position in a WFS system can be perceived independently of the subjective
distance impression.

Fig. 6.24 Technical setup for listening tests by Nogues

In a second experiment, other distance cues such as the direct to reverberant

ratio were included. A number of sources were synthesized at different distances and
the subjects were asked to adjust the direct to reverberant ratio of each source so that
the perceived distance matched that of a pair of reference sources. The results
showed that the perception of distance was primarily influenced by the direct to
reverberant ratio, and that the wavefront curvature is a weak localization cue which is
easily overridden by other cues.
Kerber implemented listening tests to compare the perceived distance of real
and virtual focussed dry sources. The results for a real source shown in Figure 6.25
support the results of other tests in that only small distances ( < 1m) are perceived
accurately and that large distances are consistently underestimated. Focussed WFS
sources do not seem to be able to produce the same distance perception as real

102

sources. Wittek suggests that these results indicate that at a fixed listening position,
the curvature of the wavefront of a dry WFS virtual source does not support distance
perception. However, in spite of not being a crucial cue, a correct wavefront
curvature (and thus a consistency between curvature and actual distance) may support
the perception of distance, particularly if the listener can move [Wittek, 2003]

Fig. 6.25 Distance of real (left) and virtual (right) sources reported by Kerber

Usher similarly found that in the absence of any indirect sound, when a source
is positioned beyond a certain distance using a WFS system, the curvature of the
wavefront does not seem to be used to determine the distance of the virtual source, but
rather the timbre of the perceived source dominates [Usher et al, 2004]. It would
seem, therefore, that in order to accurately produce WFS virtual sources at different
distances, some form of artificial reverberation is needed to provide additional
distance cues. However, Wittek points out that disturbing reflections caused by the
WFS array itself may in fact also hinder the perception of the distance of virtual
sources in front of or behind the array. It is not the case that a dry WFS virtual source
will automatically produce a natural reflection pattern in the reproduction room
[Wittek, 2003] and this is illustrated in Figure 6.26 which shows a WFS system with a
virtual source (blue dot) positioned in front of the array. The correct reflections that
would arise if a real source was at this position are indicated by the green dots, while
the actual reflections that arise are shown as orange dots. Clearly both the timing and
direction of the actual reflections do not correspond to the desired source position, but
rather to the distance of the loudspeaker array itself. The virtual source distance will
therefore most likely be perceived to be the distance of the loudspeaker array rather
103

than the specified source distance. This again indicates the significant influence of
the reproduction room acoustic on the performance of WFS systems. Various
listening room compensation schemes have been proposed which can actively cancel
early reflections in the horizontal plane [Corteel et al, 2003; Spors et al, 2003]. The
results of simulations suggest that these techniques could help to reduce the
detrimental effect of early reflections in the listening room over a large area.

Fig. 6.26 WFS reproduction room reflections

Fig. 6.27 WFS virtual source and simulated room acoustic

104

Or course, WFS can also simulate different source distances through the
addition of early reflections and reverberation. Figure 6.27 illustrates a scheme
proposed by Caulkins to artificially inject early reflections which are otherwise absent
in the WFS reproduction of focussed sources [Caulkins et al, 2003]. As with other
spatialization techniques, more listening tests are required to fully determine the
perceptual effect of combined and potentially conflicting virtual and real acoustic
reflections.

6.7.4 WFS Spectral Colouration

Perceptible changes in the timbre and spectral content of the source can occur
in WFS systems because of distortions in the high frequency content caused by spatial
aliasing and diffraction. Start found that the colouration due to spatial aliasing
changes rapidly with listener movement, and is very distinct for broadband signals
[Start, 1997]. In addition, the dynamic movement of a virtual source also introduced
spectral changes, but this effect was much less noticeable with speech signals.
Interestingly the tests found that source colouration was significantly reduced in the
reverberant reproduction room compared to the anechoic room. Start recommended a
high-frequency optimization scheme to reduce these colouration artefacts. However,
this approach does have a negative influence on localization accuracy.

Fig. 6.28 Perceived colouration for various WFS and OPSI systems

105

Wittek carried out a series of listening tests to determine the relative level of
colouration of real, WFS, WFS+OPSI and stereophonic sources in an acoustically
treated listening room [Wittek, 2007]. The results, shown in Figure 6.28, indicate that
the amount of signal colouration increases with spatial aliasing, and that the OPSI
method can significantly reduce the perceived colouration in comparison to the WFS
system. The author suggests that the non-zero result for the real reference source was
due to the non-individualised HRTF used in the experiment. Interestingly, the lowest
level of colouration was reported with the standard stereo system.

6.7.5 WFS Conclusions and Discussion

The results of the various tests discussed in the previous chapters illustrate the
demanding nature of the WFS system. Clearly, the spatial aliasing frequency is a
crucial parameter, and the results of Start, Huber, Wittek illustrate the degradation in
localization accuracy and timbre that occurs when this factor is too low.
Consequently, a very large number of loudspeakers are required if a WFS system is to
be successfully implemented for a large group of listeners. Start presented a WFS
system with an inter-loudspeaker spacing of 0.25m and a spatial aliasing frequency of
1360Hz, which, for now will be taken as a minimum system specification. A circle
with a 2m radius has a circumference of 12.56m which, based on our chosen system,
equates to approximately 50 loudspeakers. Assuming a single seated listener occupies
an area of approximately 0.5m2 (1m x 0.5m) and that all listeners must be seated at
least 0.75m from the array, it can be shown that approximately ten listeners will be
accommodated by this system. Clearly, implementing a full 3600 WFS system for a
larger audience will be a significant logistical challenge. Techniques to increase the
perceived spatial aliasing frequency without increasing the number and resolution of
the loudspeaker array are currently being investigated [Corteel et al, 2008], although
it appears this can only be achieved over a limited listening area.
Much work remains to be done in terms of the evaluation of WFS, as it
appears that the perceptual effects suggested by simulations may not be as significant
in practice. In terms of localization accuracy it seems that, providing the spatial
aliasing frequency is high enough, WFS can indeed create well-localized virtual
sources, at least under anechoic conditions. However, it is still unclear whether the
precedence effect is also contributing to localization, particularly in reverberant

106

conditions. This view seems to be supported by the results of tests which found that
localization accuracy is decreased for focussed sources in front of the array
[Verheijen, 1998] or for non-central listener positions [Marentakis et al, 2008]. The
quality of localization or locatedness is another important factor, as measures such as
the standard deviation in source angle may suggest good localization was achieved
when in fact significant source broadening occurred. Further subjective tests are
required to fully determine the localization accuracy of WFS systems for focussed
sources in front of the array and for non-central listener positions.
It has been suggested that WFS can be used to enlarge the effective listening
area of other stereophonic techniques, and this appears to be true for certain
applications. Various tests [Corteel et al, 2004] have found that WFS does
significantly increase the listening area for two-channel stereo as sources are less
likely to collapse to the nearest loudspeaker when listeners are displaced laterally
from the sweet spot. WFS has similarly been successfully used in a domestic
situation as a flexible and robust method for 5.1 reproduction as it is less sensitive to
listener position and non-standard loudspeaker positions [Corteel et al, 2004]. Other
studies have found that plane wave reproduction with a WFS system can be used to
diffuse the rear channels in a 5.1 reproduction set-up, again increasing the effective
listening area [Boone et al, 1999]. The use of WFS as a flexible reproduction format
for domestic cinema and audio applications would therefore seem to be one of the
most promising applications of this technique. The CARROUSO Project (Creating,
assessing and rendering in real-time of high quality audio-visual environments in
MPEG-4 context) has attempted to merge WFS with the flexible MPEG-4 standard
for the transfer of both recorded and synthesized sound fields between different
reproduction systems. Much of this work has concentrated on developing practical
implementations of WFS suitable for the domestic market, such as flat-panel,
distributed mode loudspeakers [Farina et al, 2000].
The reproduction of virtual sources at different distances is one of the most
widely lauded features of the WFS method. However, it appears that when the
listener position is fixed, the correct wavefront curvature produced by a WFS virtual
source does not provide any perceptible distance information [Kerber et al, 2004]. In
addition, reflections from the array and reproduction room walls will tend to pull the
perceived source distance to the distance of the loudspeakers [Wittek, 2003].
Although it has been suggested that this weak cue might help to support other more
107

dominant distance cues [Nogus et al, 2003], there is little evidence to support this
claim. The one significant exception is when the listener can move, and in this
instance the ability of WFS to reproduce a sense of changing perspective or motion
parallax has been shown to support the estimation of distance.

6.8 Analysis of Results and Recommendations

The experiments discussed in the preceding sections illustrate the wide range of
interrelated factors which influence the perceived performance of spatialization
techniques such as stereophony, Ambisonics and WFS. Virtual sources positioned
using multichannel stereophonic techniques have been found to be unstable if the
source is not positioned at a loudspeaker. This is true at all listener positions for
lateral and rear sources but also for front sources if the listener is displaced laterally
closer to one of the loudspeakers. However, although source localization often
collapses to the nearest loudspeaker, this will always be to the loudspeaker pair about
the desired source position. Therefore, increasing the number of loudspeakers will
also increase the overall localization accuracy and six loudspeakers seems to be the
minimum number of channels required for reasonable accuracy in all directions, at
least for a single listener. The differences between virtual sources positioned at, or
between loudspeakers also has an effect on dynamically moving sources and a
number of studies have found that trajectories created with stereophony tend to
therefore highlight the positions of the loudspeakers.
It is clear that for Ambisonics, the number and arrangement of loudspeakers in
the array is particularly important. A regular hexagonal array was found to produce
much better results than either square or rectangular layouts, particularly for lateral
sources, and so can be taken as a minimum specification, as with stereophony.
Likewise, the precedence effect influences localization accuracy in much the same
way as with stereophony. The results of a number of tests clearly indicate that
different decoder designs are optimal depending on whether playback is for a single
listener or a group of listeners. A full-band max rE (or in-phase, if the audience is
very near the array) has been shown to be the optimal solution for larger listening
areas as these decoders reduce (or eliminate, in the case of in-phase) the anti-phase
components which aid localization at the centre point but significantly distort it at
other listener positions. The results of a number of studies also seem to confirm that

108

localization accuracy is improved when the order of the Ambisonics system is

increased [Pullki et al, 2005; Jot et al, 1999; Daniel, 2000]. This result was largely
expected as an increase in order represents an increase in the spatial resolution of the
spherical harmonic representation of the reconstructed sound field. The directional
information represented by the spherical harmonics is therefore more accurately
represented and, consequently localization accuracy improves. A number of studies
have also clearly demonstrated that the size of the effective listening area also
increases with the order of the system [Bates et al, 2007b; Frank et al, 2008]. A
number of theoretical studies indicate that Ambisonics can only perfectly reconstruct
a sound field in a very small area at the centre of the array, and this is only possible up
to a certain frequency [Poletti, 1996; Daniel et al, 1998; Bamford, 1995]. Daniel
suggests that the reconstructed sound field becomes increasingly and linearly distorted
away from the centre point. Consequently, if the system order (and hence the
reconstruction frequency limit at the centre point) is increased, this will also increase
the accuracy of the reconstructed sound field at other, off-centre positions [Daniel et
al, 1998].
Ambisonics is generally preferred to stereophony for dynamically moving sources
as it produces smooth trajectories that do not highlight the positions of the
loudspeakers. Consequently, Ambisonics will never produce a source using a single
loudspeaker and so unlike stereophony, cannot produce the most tightly focussed
virtual image possible. Increasing the order the system reduces this effect as the
increase in directivity reduces the number of loudspeakers which are active at any one
time. There is some evidence that a similar trade-off is apparent with recorded sounds
as in one test, Soundfield microphone recordings were found to be more spacious and
enveloping than stereophonic recordings, but also less accurate in terms of directional
localization.
The simulation of distance appears to be largely dependent on the addition of
artificial reverberation. Other processes such as the Doppler effect, air absorption, or
wavefront curvature (as in WFS) do not seem to be able produce a reliable perception
of distance on their own, but are important as secondary distance cues. While a
straightforward ratio between the direct and diffuse reverberant signals can produce a
relative sense of distance, the results of a number of tests indicate that early
reflections are more important than the diffuse reverberant signal in this regard. Some
tests have found that a mixture of directionally accurate specular and diffuse early
109

reflections is preferable to purely specular reflections [Martin et al, 2001]. The

addition of artificial reverberation clearly supports distance modelling but its effect on
directional localization is unclear. Some have suggested that these additional indirect
signals reduce localization accuracy especially at off-centre listening positions
[Begault, 1992], while others have suggested the exact opposite and argue that the
increased realism of such a sound scene benefits localization [Lund, 2000].
WFS is in many respects very different from stereophony or Ambisonics and is
certainly much more demanding in technical terms. While the results of various
listening tests seem to suggest that well-localized virtual sources can be created with
WFS this is very much dependent on the spatial aliasing frequency, and hence on the
size of the array. In addition, questions remain as to how focussed these virtual
sources are, and the contribution of the precedence effect in localization with WFS.
While it appears that WFS can be used to increase the effective listening area of other
spatialization techniques such as two-channel stereo or 5.1 surround sound, extending
this to a full, large scale system is difficult, if for no other reason than the many, many
loudspeakers which would be required to surround a large audience. In addition, the
results of listening tests do not seem to support the claim that WFS can position
sources behind or in front of the array through the reproduction of the correct
wavefront curvature. The one notable exception to this is when the listener can move
through the listening area, otherwise WFS systems must use artificial reflections and
reverberation to simulate different source distances, in much the same way as other
spatialization techniques. These results indicate that WFS is perhaps not, at the
moment at least, the most suitable system for the presentation of spatial music as the
perceptible benefits do not seem to justify the vastly increased technical requirements.

6.8.1 Discussion
The preceding discussion illustrates the difficulties in the presentation of
spatial audio to multiple listeners. The influence of the precedence effect is
particularly noticeable for off-centre listeners and it appears that a high degree of
directional localization accuracy can only really be achieved for every listener if a
single loudspeaker is used. Spatialization techniques such as pair-wise amplitude
panning, and to a lesser extent, higher order Ambisonics, produce the next best
results, as the number of contributing loudspeakers is restricted and are situated in the

110

same approximate direction as the source. As lower order Ambisonics systems

generally utilize every loudspeaker to produce the virtual image, these are particularly
susceptible to localization distortion due to the precedence effect. WFS, while
appropriate for certain applications, does not seem to be viable as of yet for
presentations of spatial music due to the technical and logistical restraints and the
limited benefits.
It is also clear that Ambisonics is consistently preferred to stereophony for
moving sources as it disguises the positions of the loudspeakers which results in a
smoother trajectory. A number of non-standard amplitude panning techniques have
been developed which attempt to overcome this problem through increasing the
number of loudspeakers which are used at any one time. While these techniques
certainly appear to improve matters for dynamically moving sources, it is not clear if
they provide any advantage over a high order Ambisonics system other than the fact
that they can readily collapse the virtual image to a single loudspeaker. Max Re and
in-phase decoding schemes, without shelf filtering, appear to be the optimal decoding
schemes for larger listening areas.
The results presented in this Chapter suggest that a minimum of six
loudspeakers is required to produce optimal results for a single listener with either
stereophony or Ambisonics. An eight channel system would therefore seem to
represent an acceptable minimum layout for larger number of listeners as it contains
eight discrete spatial locations to which sources will be localized with a good degree
of accuracy, it is sufficient for third order Ambisonics, and is reasonably achievable in
terms of hardware. It has been the experience of the author that quadraphonic
systems produce extremely poorly localized lateral virtual sources when extended for
larger numbers of listeners and movements from front to back instead abruptly switch
between each position (see Section 10.2). While an additional pair of lateral
loudspeakers alleviates this issue somewhat, the wide angle between lateral
loudspeaker pairs is still problematic. An eight channel system contains a pair of
lateral loudspeakers and this provides a more useful degree of discrimination in lateral
positions and movements, while still being reasonably efficient and economical. For
these reasons, the author has adopted an eight channel loudspeaker array as a standard
system for performances of spatial music.

111

6.8.2 Implications
The results presented in the preceding section suggest that it is very difficult to
produce spatial locations and trajectories which are unambiguously perceived by
every listener, in the same way. Even in the case of point sources which are clearly
localized, each listener will be orientated differently with regards to the loudspeaker
array, and so will have a different perspective on the spatial layout. As noted earlier,
directional localization accuracy is the main topic under investigation in many of
these tests, but this is not necessarily the only way in which space can be utilized in a
musical composition. The results presented earlier suggest that this may in fact be a
necessity. However, it is just as important to know if these other uses of space are
clearly perceptible to an audience, and if so, which spatialization technique can
achieve this most effectively, if at all? Clearly, Ambisonics is the preferred
spatialization technique for dynamically moving sources. However, it is also clear
that the precise trajectory perceived by each listener will be strongly influenced by
their position within the array.
If a recorded sound is to be used in spatial music composition, the ambisonic
Soundfield microphone represents the most flexible recording option if an enveloping
sound field is required. However, if a more directional diffusion is required, then
monophonic or stereophonic microphone techniques are perhaps more applicable as
although multi-channel microphone techniques can be very effective, they are tied to
a specific reproduction layout.
While many composers continue to utilize various multi-channel techniques,
others have adopted an entirely different approach based upon a single two-channel
stereo source and a large, disparate collection of spatially distributed pairs of
loudspeakers, i.e. a loudspeaker orchestra. This aesthetic represents a very different
approach to the multi-channel techniques discussed in the preceding chapters.
However, the art of diffusion is admirably focussed on the perception of the audience
and the real technical problems which arise in these kinds of performances, something
which is often lacking in multi-channel tape compositions.
The second half of this thesis will focus on spatial music composition via the
analysis of a number of different composers and aesthetics, and some original
compositions by the author. Different approaches to the use of space as a musical
parameter will be assessed in terms of the technical and perceptual research presented

112

in the preceding chapters. Inevitably, greater emphasis will be placed on music from
the twentieth century as many significant aspects of spatial music are dependent on
technical developments from this era, however, spatial music is not solely a twentieth
century phenomenon. The spatial distribution of performers has been used for
centuries in European religious choral music, and this antiphonal style is itself derived
from the even more ancient call-and-response form. The next chapter in this thesis
will examine this early form of spatial music and investigate the development of
acoustic spatial music in the first half of the twentieth century, prior to the
development of recording and amplification technology and electronic spatialization
techniques.

113

7 Acoustic Spatial Music

Spatial music is often closely associated with technological developments in
the twentieth century, yet the use of space as a musical parameter is much older. Calland-response patterns can be found throughout history in many different cultures and
musical traditions. In this dialogue form, musical material is divided between two
groups, which will necessarily be situated at two different spatial locations. Call-andresponse patterns therefore represent the most basic form of spatial music and they are
a fundamental aspect of the earliest formalized system of spatial music, antiphonal
choral music.

Fig. 7.1 Interior of St Marks Cathedral in Venice

114

7.1 Early Spatial Music

In the early Christian tradition, the term antiphon was used to denote a sung
response to a psalm during a religious service, and this gave rise to the antiphonal
style of singing which was widely used in medieval church music. The polyphonic
choral music of the sixteenth century and the Renaissance retained this older
technique, often using a smaller capella choir in conjunction with the main ripieno
choir. The popularization of this technique is generally attributed to the Venetian
School founded by the Flemish composer Adrian Willaert, who became maestro di
cappella of St. Mark's Cathedral in Venice in 1527 (see Figure 7.1). The existing use
of alternating choirs was facilitated here by the unique interior of the Cathedral of St.
Marks which contains two spatially separated organs and choir lofts. Composers
began to take advantage of the increased distance between the groups and use the
spatial separation as a special effect. Bryant points out however that it cannot be
assumed that the spatial separation of the choirs, or cori spezzati, was an integral part
of early music of the Venetian School. The historical records suggest that this was
instead an alternative arrangement which was occasionally used as a special effect
based on the preference of a performances music director or special guest [Bryant,
1981]. Indeed, as antiphonal music is largely based on successive, often contrasting
phrases which alternate between the two choirs, a greater spatial separation is not,
strictly speaking, necessary, as the separation is inherent to the musical material.
Willaerts eight-part Vespers, composed in 1550 is one of the earliest examples of this
music and features the echo and dialogue effects between the two spatially separated
groups which are typical of this aesthetic.
By the early seventeenth century the music of the Venetian school of
composers had developed beyond this strict ritual alternation between two choirs to a
rapid, more sophisticated dialogue between multiple choirs and instrumental groups.
Students of Willaerts such as Andrea and Giovanni Gabrieli composed large scale
works for multiple groups and choirs and were also the earliest composers to include
dynamic markings and specific instrumentation in the score. Bryant suggests that the
larger choir would have remained at floor level as they had other ceremonial duties
during the mass [Bryant, 1981]. The instrumental groups, who had no active role in
the ceremony, would have been placed in an inconspicuous position, most likely in
the two elevated organ lofts, while the smaller cappella choir would most likely have
115

been situated at some other point away from the main choir. This theory is supported
by historical accounts from the time, one of which details the method of
synchronizing the various spatially distributed groups. It appears that two additional
conductors were employed to relay the beat indicated by the principal conductor,
situated at floor level with the main choir, to the musicians in each organ loft.
Giovanni Gabrielli's famous work In Ecclesiis is an excellent example of this form of
polychoral music, using four separate groups of instrumental and singing performers
accompanied by organ and basso continuo to create a spatial dialogue and echo
effects (see score extract in Figure 7.2).

Fig. 7.2 Echo effects in Giovanni Gabrielli's In Ecclesiis

The Venetian school was highly influential across Europe, helped in part no
doubt by the invention of the printing press a century before. The English composer
Thomas Tallis composed Spem in Alium in 1573 for forty separate vocal parts
arranged in eight choirs while Orazio Benevolis Festal Mass was written for fiftythree parts, two organs and basso continuo. Over the next four hundred years, the use
of antiphony became rarer with some notable exceptions such as the antiphonal choral
effects of J. S. Bachs St. Matthew Passion (1729), and the motivic interplay of
spatially separated groups of Mozarts Serenade in D for four Orchestras (1777). In
the Romantic era, composers occasionally placed groups of musicians away from the
main orchestra for dramatic effect. One example is Berliozs Requiem (1837), which

116

at its premiere included four brass ensembles positioned at the four cardinal points,
along with a massive orchestra of singers, woodwinds, horns, strings, and percussion.
Berlioz was aware in advance that this work would be premiered in Les Invalides, the
gigantic domed cathedral of the military hospital in Paris, and he exploited the
characteristics of this space in this new commission. In the famous Tuba Mirum
section, the invocation of Gods fury with the damned is invoked through the
consecutive entrance of the four brass ensembles which gradually builds to a dramatic
climax of massed timpani and voices (see Figure 7.3). 1

Fig. 7.3 Tuba Mirum section in Berliozs Requiem

Although Berlioz was clearly thinking about the use of space in music (he
referred to it as architectural music), the spatial distribution of the performers in this
case is largely for dramatic effect and is not a critical feature of the work, and this is
true of most of the historical examples discussed in this Section. In the early part of
the twentieth century, space was sometimes used to create a sense of perspective by
contrasting the orchestra on stage with more instruments placed at a distance off1

Interestingly, during the first performance, the conductor Habeneck (a rival of Berlioz) is alleged to
have broken for a pinch of snuff during the critical entrance of the brass ensembles, requiring Berlioz
himself to leap in and take over the conducting duties [Bromberger, 2007].

117

stage. The Austrian composer Gustav Mahler often used off-stage musicians in
addition to the main orchestra, such as, for example, the brass and percussion in the
fifth movement of Symphony No. 2 (1894) or the off-stage snare drum in Symphony
No. 3 (1896). Other significant composers of the era also used similar effects,
although not as frequently as Mahler. Igor Stravinsky made use of tubas dans le
couline (in the corridor i.e. in the wings) in the ballet score of Firebird (1910) and
Strauss featured six trombones auf dem Theater in Die Frau ohne Schatten (1919).
This period was one of great upheaval in Western Art music as many composers
begin to move away from the strict functional tonality and defined meters of previous
eras. While composers like Stravinsky and later Schoenberg experimented with the
basic foundations of musical structure, others retained aspects of traditional music
practice but utilized them in a very different way. The American Charles Ives is one
such composer who regularly combined traditional tonality with the then new ideas of
musical quotation, polyrhythms and meters, and spatial effects. Ives was a
contemporary of Mahler (both produced most of their music within the same thirty
year period from 1888 to 1918) and although their music is quiet disparate, and
derived from very different musical traditions, both composers do exhibit certain
similarities [Morgan, 1978]. For example, both composers were interested in the
quotation of other musical material within their own compositions and both regularly
combined and juxtaposed layers of different and contrasting material. Both
composers also retained aspects of functional tonality in their work and made
extensive use of overlapping yet unrelated tempi. While both composers used the
spatial distribution of performers in their work, it would be Ives who developed this
practice further and have the most lasting effect on the development of spatial music,
particularly in America.

7.2 The Co-Existence of Dissimilars

Charles Ives was born in Connecticut in 1874 and although he received little
acclaim in his lifetime, he is now regarded by many as one of the most innovative
composers of his generation. Throughout his life, Ives talked about the influence of
his father, a musician and band-leader, on the development of his unorthodox
approach to music. He stated that his father had once directed his two sons to sing a
tune in one key while he accompanied them on piano in another, while numerous

118

accounts exist of his experiments with the spatial effect of simultaneous marching
bands [Ross, 2009; Mortenson, 1987]. This influence can clearly be seen in the music
of Charles Ives, which often makes use of overlapping keys and meters and the
combination of European Art music with American popular and church music. Over
the course of his life, Ives would go on to explore many of the musical innovations
which would become associated with modern contemporary music such as
polytonality and polyrhythm, tone clusters and microtonality, musical quotation and
collage, and also spatial music. Much of Ives music involves the juxtaposition of
various disparate elements and his compositional use of space generally reflected this.
The Unanswered Question (1908), one of his most famous compositions, uses the
spatial distribution of musicians to highlight the three distinct layers of strings,
woodwinds and brass. The three layers operate independently at their own tempo and
key and this musical separation is further accentuated through the placement of the
string orchestra off-stage, the woodwind ensemble on stage, and the solo trumpet
positioned at some other distant position, such as a balcony. Throughout the piece the
string orchestra performs slow, sustained tonal triads which are punctuated by short
trumpet phrases (the question), which are in turn answered in an increasingly
incoherent fashion by the woodwinds. The symbolism of this work is beautifully
supported by the spatial distribution and layering of the musical material. The slowmoving tonal strings represent the natural world which surrounds the audience and
remains in constant yet slow motion, undisturbed by the question and answer dialogue
of the trumpet and woodwinds. The trumpet sounds out the question with a clear,
atonal phrase which is then answered by the woodwind ensemble. In contrast to the
clear question of the solo trumpet, the woodwinds respond with multiple, overlapping
and seemingly unrelated phrases and so, no answer is found to the eternal question of
the title.
In much of Ives music, space is used to clarify and define the various
overlapping yet independent musical layers, and other composers at the time were
also beginning to experiment with the layering of disparate musical material. Italian
futurists like Luigi Russolo were exploring collage form and noise while Darius
Milhauds 1918 ballet LHomme et son desir used multiple, spatially distributed
ensembles playing independently of each other, sometimes in different metres
[Zvonar, 2004]. However, Ives fourth and last symphony (1910-1916) is certainly
one of the most ambitious works that made use of this technique. This work, for a
119

gigantic orchestra with additional off-stage ensembles, was not performed in full until
nearly a decade after Ives death. The second movement juxtaposes so much different
thematic material that a second conductor is generally required while the final
movement contains a contrasting dialogue between discordant and tonal material. In
the conductor's note Ives wrote [Johnson, 2002];
"As the distant hills, in a landscape, row upon row, grow gradually into the horizon,
so there may be something corresponding to this in the presentation of music. Music
seems too often all foreground even if played by a master of dynamics
It is difficult to reproduce the sounds and feeling that distance gives to sound wholly
by reducing or increasing the number of instruments or by varying their intensities.
A brass band playing pianissimo across the street is a different sounding thing than
the same band playing the same piece forte, a block or so away. Experiments, even
on a limited scale, as when a conductor separates a chorus from the orchestra or
places a choir off the stage or in a remote part of the hall, seem to indicate that there
are possibilities in this matter that may benefit the presentation of music, not only
from the standpoint of clarifying the harmonic, rhythmic, thematic material, etc., but
of bringing the inner content to a deeper realization.

Clearly Ives used the spatial separation of performers to create a sense of distance and
perspective, in much the same way as European composers such as Mahler.
However, Ives also used this spatial distribution to clarify different layers of
independent and potentially dissonant musical material and to facilitate the
performance of overlapping yet unrelated musical layers, often at different tempi or
metres. While the spatial separation of musical material at different tempi obviously
has practical benefits for its performers, the above quote also indicates that Ives had
intuitively realised that this spatial separation also benefited the listener. It is
unknown whether this insight was informed by the recent development of Gestalt
psychology in Germany, or derived from Ives own experience. However, it has since
been shown by Bregman and others that our ability to segregate an audio scene into
multiple streams strongly influences our perception of musical parameters such as
melody and rhythm [Bregman, 1990]. Bregmans work on Auditory Scene Analysis
(see Chapter Two, Section 2.3) emphasized the importance of spatial cues in the
segregation of audio streams and he suggested that the spatial separation of a
multiplicity of sounds prevents the auditory system from computing dissonances
between them. Other studies have also found that a listeners ability to detect and
understand the content of multiple signals is improved if the signals are spatially
separated signals [Shinn-Cunningham, 2003; Best, 2004] and this would also appear
to support Ives use of space to clarify the harmonic, rhythmic and thematic

120

material. Although largely ignored for much of his career, Charles Ives would
eventually be recognized as a highly creative and innovative composer, and his
experiments with spatial music would be an important influence on a number of
American composers.
The composer Henry Brant (1913-2008) is one such composer who was
influenced by Ives. Over the course of his career, Brant wrote seventy-six works of
spatial music (along with fifty-seven non-spatial works [Harley, 1997]) and has
become one of the most famous composers of orchestral spatial music. Brants use of
space was clearly influenced by Ives, as illustrated by this extremely admiring 1954
description by Brant of Ives The Unanswered Question.
This unique, unprecedented little work, written in 1908, presents, with extraordinary
economy and concentration, the entire twentieth-century spatial spectrum in music, and
offers guidelines for solving all the practical problems involved. The spatialcontrapuntal-polytemporal principles so brilliantly exemplified in this piece are the basis
for the more complicated spatial superimpositions present in all my own recent largescale works [Brant, 1967].

Brant composed and wrote extensively on the use of space in orchestral music and
his first spatial work, Antiphony I (1953) contains many of the core ideas which the
composer would continue to use throughout his career. In this piece, the orchestra is
divided into five groups which are placed at different parts of the auditorium and
perform material in contrasting tempi, meter and harmonies. Although the entrance of
each group is cued, they then proceed independently at their own speed. The
composer states that a purposeful lack of relationship between the intervals,
phrasing, note-values, tone-quality and sonorities of the various lines will necessarily
produce a complex result as soon as the lines are combined [Harley, 1997]. Brant
presented his ideas about the compositional use of space in a brief article written in
1967 [Brant, 1967] which is summarized as follows;
-

The perception of different layers of musical material can be enhanced by the

spatial separation of the performers.

Exact rhythmic coordination is difficult to achieve when musicians are

spatially separated by large distances.

Spatial separation is equivalent to pitch separation but allows for greater

complexity as material in the same harmonic range which would merge if
produced from the same location, can be separated into distinct musical lines
if the sources are separated in space.
121

Each situation and listening position is different and there is no single

optimum listening position.

Brant is clearly influenced by Ives in his approach, which is based on completely

contrasting material, spatially separated and without any rhythmic synchronization
between the groups. The use of strings as a continuous static layer in contrast with
other musical lines at different spatial locations represents another similarity between
this work and The Unanswered Question. However, unlike Ives, Brant specified the
precise location of each ensemble, something he would continue to do throughout his
career. This allowed Brant to explore other spatial effects as illustrated by the spatial
distribution used for Millenium II (1954) which is shown in Figure 7.4[Harley, 1997].
The piece uses a spatial assembly of ten trombones and ten trumpets, positioned along
the side walls of the hall, as well as on-stage percussion and brass, and a single voice,
preferably positioned to the rear and above (in a balcony, for example). At the
beginning of the piece, the trumpets and trombones enter one-by-one in numerical
order, each playing in a different melody in a different key. Brant describes the effect
as follows;
there is a compelling impression of the hall tangibly filling up with sound,
principally along the walls, but also with some feeling of the centre being
progressively saturated, especially as the accumulation proceeds towards its
maximum point. The impression of the sound travelling gradually down the first wall
is very strong; this impression of moving direction becomes less well defined as the
further entrances and accumulations occur [Brant, 1967].

Fig. 7.4 Spatial location of musicians in Brants Millennium II (1954)

122

Fig. 7.5 Spatial patterns in Brants Millennium II (1954)

Fig. 7.6 Sound axes in Brants Millennium II (1954)

123

This piece also includes other spatial effects such as the stepwise introduction
of trombone/trumpet pairs, which are described by Harley as sound axes [Harley,
1997]. This spatial movement is linked to pitch, as the entries begin with a very high
trumpet note paired with a very low note on the trombone directly opposite, and end
with a convergence around middle C (see Figures 7.5 and 7.6). This movement is
quite different from the successive entry of the spatial assembly instruments as this
time a relationship exists between the material, instead of each instrument having its
own independent melody, key and meter. Brant used the term spill to describe the
effect of spatially distributed musicians performing similar, harmonically related
material. He uses the Tuba Mirum section of Berliozs Requiem (see Figure 7.3) as
an example and describes how the common tonality and tone quality of the four
groups causes the resulting musical texture to extend from the corners to fill the room
[Brant, 1967].
Throughout his long career, Brant continued to compose spatial works of
greater and greater scale, culminating in Bran(d)t aan de Amstel, the massive
spectacle of spatial music which encompassed most of the city of Amsterdam in the
1984 Holland Festival [Harley, 1997]. This huge work involved a colossal number of
musicians including numerous bands in public squares, a youth jazz band, two
choruses, two brass bands, four street organs and four boatloads of performers moving
through the citys canals.
Ives influence can also be seen in the spatial distribution scheme adopted by
the early composers of electronic music in America in the early 1950s. The relatively
recent development of magnetic tape was quickly adopted by composers as it greatly
facilitated the editing and splicing together of different sounds. The Project for Music
for Magnetic Tape was established in New York in the early 1950s and over the next
two years this group produced three new electronics works, Cages Williams Mix
(1952), Earle Brown's Octet (1952) and Morton Feldman's Intersection (1953). Each
piece was realized using eight unsynchronized monophonic tapes positioned
equidistantly around the auditorium. The spatial separation of multiple independent
musical layers, in this case electronically generated, is clearly reminiscent of the
approach taken by Ives and Brant. In a lecture on experimental music given in 1957,
Cage described this approach as follows;
Rehearsals have shown that this new music, whether for tape or for instruments, is
more clearly heard when the several loudspeakers or performers are separated in

124

space rather than grouped closely together. For this music is not concerned with
harmoniousness as generally understood, where the quality of harmony results from
a blending of several elements. Here we are concerned with the coexistence of
dissimilars, and the central points where fusion occurs are many: the ears of the
listeners wherever they are. This disharmony, to paraphrase Bergsons statement
about disorder, is simply a harmony to which many are unaccustomed [Cage, 1957].

In a 1992 interview Brant stated that the main function of space in music is to
make complexity intelligible [Harley, 1997] and Cages co-existence of dissimilars
is very reminiscent of this. Brant did, however, distinguish his music from Cage,
stating that his approach is opposed to what later came to be termed aleatoric or
indeterminate music, in which accident and chance are looked upon as fundamental
musical parameters. When uncoordinated rhythmic layers are combined with spatial
distribution, accident is no more a factor than it is in the performance of rubato in a
complex Chopin ratio [Brant, 1967].
The music of Ives, Brant and Cage uses space as a fundamental musical
parameter, as in this case, the spatial separation of the different musical lines is crucial
for the correct perception of the work. However, although space is a critical aspect of
this music, it is nevertheless used in a somewhat limited way, namely just to separate
and clarify the various musical lines. While composers like Cage adopted a strictly
indeterminate approach, others began to develop more formal systems to organize the
spatial aspects of a work. Brant was well aware of the dangers in this approach,
saying that schemes for spatial distribution that are conceived in terms of their visual
expressiveness on paper cannot be expected to produce any effect on the aural
mechanism. He also discussed the degradation in spatial impression that occurs after
a certain level of musical activity is reached, stating that the impression of the sound
travelling gradually down the first wall is very strong; this impression of moving
direction becomes less well defined as the further entrances and accumulations occur
[Brant, 1967]. However, Brant did use geometrical patterns to map out spatial
trajectories, such as the sound axes of Millennium discussed earlier, and in the years
to follow many more composers would also turn to geometric abstraction in an
attempt to systematically organize the spatial relationships within a composition. This
approach was undoubtedly influenced by the development of electronic music as now,
for the first time, sounds could be entirely removed from their source and reproduced
at will. The different aesthetics of electronic music which developed in the middle of

125

the twentieth century would profoundly influence the use of space, not only in
electronic music, but also in the way composers employed space in orchestral music.

126

8 Musique Concrte & Elektronische Musik

Fig. 8.1 Pierre Henry performing with the potentiomtre d'espace, Paris, 1952

The development of recording and amplification technology allowed

composers to, for the first time, completely separate a sound from its physical source,
and hence its physical location in space. The French composer and engineer Pierre
Schaeffer conducted some of the earliest experiments with music based on recorded
sounds while working at the French national broadcaster Radiodiffusion Franaise
(RDF) in 1936. He continued this work after the war and by 1948 Schaeffer had
completed a number of works in this new style of Musique Concrte. Compositions
such as Etude Violette (1948) and Aux Chemin de Fers (1948) were broadcast on
French national radio. In 1949, Schaeffer met the composer Pierre Henry and the two
went on to found the first purpose-built electroacoustic music studio at the Groupe de
Recherche de Musique Concrte (GRMC) at RDF in 1951. The two also collaborated
on a number of compositions such as Symphonie pour un homme seul (1950) and
Orphee 51 ou Toute la Lyre (1951), initially using turntables and then later magnetic

127

tape. In 1952, Schaeffer published la Recherche d'une Musique Concrte (The

Search for a Concrete Music) which summarized the basic principles and working
methods of Musique Concrte at that point. In this text, Schaeffer introduced the idea
of the movement of sound along trajectories sonores (sonic trajectories) and the
creation of spatial relief through the contrast of static spatial locations and dynamic
mobile sound sources which are controlled manually by the performer. This new
approach was demonstrated for the first time in a performance in Paris in 1951 which
featured multiple monophonic turntables routed to four loudspeakers positioned to the
left, right and rear of the stage, and a fifth loudspeaker placed overhead. During the
performance, four tracks were routed to each loudspeaker while a fifth was spatially
diffused live by a performer using the potentiomtre d'espace, a highly theatrical
system which controlled the spatial distribution of the fifth track (see Figure 8.1).
The aesthetic of Musique Concrte is often discussed solely in terms of its use
of recorded sounds. However, this term reflects the compositional working method as
much as the source material. The traditional compositional process begins with an
abstraction, the score, which is then eventually realized by the performer. Musique
Concrte inverts this approach, as the composer now begins with real, concrete
sounds which must be then manipulated to create an abstract piece of music. For
Schaeffer, it is the sound itself which is of utmost importance and all of the activities
of the composer and the listener should be directed towards the inherent properties of
the sound itself. Schaeffers compositional aesthetic was therefore incompatible with
traditional approaches to composition, or with the principles of serialism, which more
and more composers began to adopt in the post-war period. Serialism developed
initially from the twelve-tone music of composers such as Arnold Schoenberg, Alban
Berg and Anton Webern and was adopted by many composers as a means of
departing from the functional tonality of previous eras. Later composers generalized
this idea and began to use a series of ordered elements, such as a twelve-tone pitch
set, as a structuring principle to order and manipulate other parameters within the
composition such as rhythm and dynamics. Since serialism presented itself as a denial
of tonality, Schaeffer saw no point in applying serial methods to concrete material
[Palombini, 1993], as this organizing principle is imposed upon the sounds, instead of

being derived from them. Nevertheless, Schaeffers ideas were well received at a
lecture at Darmstadt festival in 1951 and, perhaps due to the influence of visiting
composers like Boulez, Messiaen and Stockhausen, serialist tendencies, although
128

resisted by Schaeffer, began to emerge within the GRMC [Palombini, 1993]. This
eventually lead to the resignation of Schaeffer, Henry and others from the GRMC in
1958 and the founding of a new collective, the Groupe de Recherches Musicales
(GRM), which was later joined by composers such as Luc Ferrari, Iannis Xenakis,
Bernard Parmegiani, and Franois Bayle.
Meanwhile in Germany, another electronic music studio was established
which would be far more amenable to the tenants of serialism. The physicist Werner
Meyer-Eppler had published a thesis in 1949 on the production of electronic music
using purely electronic processes [Meyer-Eppler, 1949] and in 1951, Meyer-Eppler,
with Robert Beyer and Herbert Eimert, established a new studio for this purpose in
Cologne at the Nordwestdeutscher Rundfunk (NWDR). Their approach to
Elektronische Musik, differed from Musique Concrte in its use of synthesized sounds
rather than recorded acoustic sounds. However, both aesthetics also differ
fundamentally in terms of their basic approach to electronic music composition.
Composition with sound synthesis is inherently more suited to abstract structuring
principles such as serialism, as the material can be deliberately generated to fit the
preconceived score. This is much more difficult with the complex acoustic sounds of
musicque concrte and indeed Schaeffer struggled to find a formal compositional
system which originated from the intrinsic properties of the sounds. Schaeffers
disappointment is evident in this quote from an interview in 1986;
"I fought like a demon throughout all the years of discovery and exploration in
musique concrte. I fought against electronic music, which was another approach, a
systemic approach, when I preferred an experimental approach actually working
directly, empirically with sound. But at the same time, as I defended the music I was
working on, I was personally horrified at what I was doing. . . . . . I was happy at
overcoming great difficulties - my first difficulties with the turntables when I was
working on Symphonie pour un Homme Seul, my first difficulties with the tape
recorders when I was doing Etude aux objects - that was good work, I did what I set
out to do. My work on the Solfege - it's not that I disown everything I did - it was a
lot of hard work. But each time I was to experience the disappointment of not
arriving at music. I couldn't get to music, what I call music. I think of myself as an
explorer struggling to find a way through in the far north, but I wasnt finding a way
through. " [Hodgkinson, 2001]

The dichotomy between these two approaches is characteristic of the divergent

approaches to electronic and electro-acoustic music composition and performance that
has existed throughout its fifty year history, and this is equally true of their respective
approach to the use of space. Is the focus on the sound itself or the relationship

129

between sounds, does the spatial movement or distribution arise from the sound itself
or is it imposed upon the sound?
The NWDR studio in Cologne swiftly became one of the most well-known
electronic music studios in the world, helped in no small part by the growing fame of
the composer Karlheinz Stockhausen, who had joined the studio in 1953.
Stockhausen had worked briefly with Schaeffer in 1952 and produced a single
Konkrete Etde in 1952. Stockhausens approach to composition at this time was
based upon the ideas of total serialism, meaning the application of a measurable scale
of proportions (the series), not only to pitches, but also to non-pitched parameters
such as timbre, rhythm and also space. Schaeffers musicque concrte would
eventually lead to an aesthetic in which space is used in a performance to highlight
and exaggerate the pre-existing content of the music, and this will be looked at in
more detail later in this chapter. The alternative legacy of composers like
Stockhausen can be seen in the large number of abstract schemes which composers
began to use to organize the spatial distribution of material.

8.1 Stockhausen and the Serialization of Space

The German composer Karlheinz Stockhausen (1928-2007) is widely regarded
as one of the most important and influential composers of the twentieth century.
Throughout his career, Stockhausen composed numerous works of spatial music and
also published various treatises on electroacoustic composition and the use of space in
music. Although often a controversial figure, Stockhausen was highly influential,
particularly in the development of electronic music, and was one of the few avantgarde composers to become well known to the general public.
Stockhausen completed his undergraduate studies at the Cologne Conservatory
of Music in 1951 and went on to study serialist techniques at the Darmstadt
composition school. After studying for a short time in Paris with Olivier Messiaen
and Pierre Schaeffer, Stockhausen returned to Germany in 1953 and took a position as
an assistant to Herbert Eimert at the NWDR studio in Cologne. In his early career,
Stockhausen primarily composed "totally organized" serial music, in which each
musical parameter is controlled by a series of numerical proportions. Stockhausen
believed that the precedent for this extension to the serial principle could be found in
the works of Webern and he made the following comment about Webern's Concerto,

130

Opus 24, "What is essential is not a uniquely chosen gestalt (theme, motive), but a
chosen sequence of proportions for pitch, duration and volume [Morgan, 1991].
Electronic music would be the ideal medium to implement these ideas as it allowed
the composer to create the musical material and hence organize timbre and space
according to a series of proportions. These ideas were implemented by Stockhausen
in the electronic work Gesang der Jnglinge (1955/1956) which has been described as
the first masterpiece of electronic music [Simms, 1996] and was also the first piece
to serialize the projection of sound in space [Smalley J., 2000]. In this piece,
Stockhausen attempts to forge a connection between recordings of a boy soprano and
electronically synthesized sounds ranging from white noise to sine tones.
Stockhausen categorized the vocal recordings into basic phonetic components such as
noise-like plosive consonants and vowel sounds which resemble pure tones. These
were then combined with artificial consonant and vowel-like sounds created from
layered sine waves and filtered white noise to produce a range of material which fills
a continuum of timbres from pure tones to noise. This positioning of material within
a continuum of timbres illustrates Stockhausens conception of serialism as a
graduated scale of values between two opposing extremes which he later described in
an interview in 1971;
"Serialism is the only way of balancing different forces. In general it means simply
that you have any number of degrees between two extremes that are defined at the
beginning of a work, and you establish a scale to mediate between these two
extremes. Serialism is just a way of thinking." [Cott, 1973]

Although both recorded and synthesized sounds are used in Gesang der
Jnglinge, every sonic event and parameter (including spatial locations and the
comprehensibility of the text) was controlled and organized serially. The original
work was performed using five groups of loudspeakers at its premiere in 1953 but was
subsequently mixed down to four tracks. No five-track tape machines existed at this
time, so a four-track machine was used to feed four loudspeakers positioned around
the audience while an additional tape machine was used to feed a fifth loudspeaker
positioned on stage (see Figure 8.2 [Smalley J., 2000]). Although it is known that
Stockhausen attempted to organize space in this work in the same way as every other
parameter, the exact means by which serial techniques were applied to the spatial
distribution and movement is not entirely known [Smalley J., 2000]. The spatial
distance, especially of the boys voice, is clearly an important aspect of this work, and

131

the composer relates this parameter to the comprehensibility of the voice. When the
voice is positioned close-by with little reverb, the text is clearly comprehensible and is
the primary focus of attention. However, as the voice recedes into the distance and
the level of reverberation increases, comprehensibility decreases and the complex
interaction of the voice with itself and other similar electronic elements becomes the
primary focus of concern [Moritz, 2002].

Fig. 8.2 Gesang der Jnglinge premiere at WDR Cologne, 1956

Stockhausen's theoretical writings from this time are primarily concerned with
the difficulties in applying serialist proportions to non-pitched parameters such as
space and timbre. Stockhausen was well aware of the difficulties inherent in this
approach, particularly in terms of timbre, writing that "we only hear that one
instrument is different from another, but not that they stand in specific relationship to
one another", and it is clear that a similar problem exists in terms of the perception of
spatial locations and movements. Stockhausen's solution is described in the wellknown text, Wie die Zeir Vergeht (how time passes) which was published in 1955
[Stockhausen, 1959]. In this essay, Stockhausen presents a theory of the unity of
musical time in which every musical parameter is considered at different temporal

132

levels. So, for example, the macro-level rhythm of an individual musical phrase can
be related to the specific micro-level subdivision of the spectrum by its harmonic
structure, which is also similarly related to timbre [Morgan, 1991]. At even greater
durations, the entire composition can be considered as a timbre with a spectrum
derived from the pitch, rhythm and dynamic envelope of the combined individual
musical phrases. This new approach moved the focus away from the pointillistic note
relationships of early serialism and onto what Stockhausen referred to as "group
composition", which emphasized the overall character of large groups of
proportionally related material, rather than on the relationship between individual
pitches. This approach is demonstrated in Gruppen (1955-57), for three orchestras
positioned to the left, in front and to the right of the audience. The spatial separation
of the three orchestras was initially devised to clarify the carefully constructed
relationships between the three layers of material, which is clearly reminiscent of the
approach adopted by Brant and Ives. However, in certain passages, musical material
is passed from one group to another and Stockhausen notes that similar orchestration
was deliberately used for each of the three groups in order to achieve this effect. The
spatial movement was produced using overlapping crescendos and decrescendos (see
Figure 8.5) which are clearly reminiscent of stereophonic panning and illustrates how
Stockhausens experience with electronic music composition influenced his
composing for instruments. Indeed, Stockhausen originally intended to write
Gruppen for both orchestral and electronic forces but the electronic element was
eventually abandoned due to practical and economic constraints [Misch et al, 1998].
The composer describes the spatial aspects of this work as follows;
The spatial separation of the groups initially resulted from the superimposition of
several time layers having different tempi which would be unplayable for one
orchestra. But this then led to a completely new conception of instrumental music in
space: the entire process of this music was co-determined by the spatial disposition of
the sound, the sound direction, sound movement (alternating, isolated, fusing,
rotating movements, etc. ), as in the electronic music Gesang Der Jnglinge for five
groups of loudspeakers, which was composed in 1955/56 [Moritz, 2002].

The spatial segregation of the three groups clearly helps to distinguish the
different layers of material but the more elaborate spatial effects also seem to be quite
effective. Various reviews of performances of this work have commented on the
dramatic effect of a single chord in the brass instruments travelling around the hall,
and this spatial movement is also quite apparent in stereo recordings of this piece (in

133

which the three groups are generally panned hard left, centre and hard right). The
approach to space used in Gruppen was further developed in Carre (1959-1960)
which was composed by Stockhausen and his assistant at the time, the British
composer Cornelius Cardew. This piece was composed for four orchestras which as
the name suggests (Carre literally means square in French) were arranged in a square
around the audience, as shown in Figure 8.6 As with Gruppen, similar orchestration
is used for each group with the addition of a mixed choir of eight to twelve singers.
The vocal text consists largely of phonemes and other vocal sounds chosen for their
relationship to the instrumental sounds.

Fig. 8.3 Rehearsal of Gruppen in Cologne, March 1958

Fig. 8.4 Dress rehearsal of Carre in Hamburg, October 1960

134

Fig. 8.5 Spatial movement in Gruppen

135

Fig. 8.6 Orchestra & audience layout in Stockhausens Carre

In 1958 Stockhausen published two articles in the German music journal Die
Riehe entitled Electronische und Instrumentale Musik (Instrumental and Electronic
music) [Stockhausen, 1975a] and Musik in Raum (Music in Space) [Stockhausen,
1975b]. In the latter essay, Stockhausen briefly discusses the history of Western
spatial music before going on to outline the functional use of space in his own work.
The composer describes two difficulties which often occur in serialist music. Firstly,
the extensive use of serialist processes results in music in which every parameter is
constantly changing, and this often results in a rather static, pointillistic texture.
Secondly, in order to articulate longer time-phrases, one parameter must remain
constant and dominate, but this is in direct contradiction with the serialist aesthetic.
Stockhausen suggests that the spatial distribution of sounds can be used to articulate
longer time-phrases and structure the material, and also to clarify the complex
relationships between the different layers [Stockhausen, 1975b]. Stockhausen also
discusses the perception of distance in terms of air absorption, reverberation and early
reflections which, despite the typically characteristic terminology, compares relatively
well to the current research reviewed earlier in this thesis. Stockhausen discusses the
relative nature of the perception of distance, and notes the importance of source
recognition in this context and the difficulty in estimating the distance of unknown,
synthesized sounds. Consequently, Stockhausen concluded that the distance of a
source is a secondary musical parameter which is determined by the timbre and
loudness of the source signal, and that the direction of the source should be the
primary spatial parameter which is related to overall serial structure. Interestingly

136

Stockhausen suggests that proportional changes in location should be used instead of

fixed scale steps and the appropriate minimum change in value should be derived
experimentally.

Fig. 8.7 Spatial intervals and directions from Stockhausens Musik in Space

Stockhausens next work, Kontakte (1958-60), would complete the idea for a
combination of instrumental and electronic music originally considered for Gruppen.
This famous work can be performed as a four-channel tape piece or as a mixed-media
work for four-channel tape, live piano and percussion. As the title suggests, this work
explores the various points of contact between static instrumental sounds and dynamic
electronic textures, between different spatial movements, and also the temporal
relationship between pitch and rhythm. At one point in this piece, Stockhausen
dramatically illustrates his theory of the unity of musical time as a high-frequency,
clearly pitched tone smoothly transforms into a slower rhythmical procession of
clicks. Although direction, or more precisely changes in direction, is the primary
spatial element in Kontakte, distance is also used to support dynamic aspects of the
music, such as the fast oscillating spatial movements which emphasize certain loud
dynamic sounds. In addition, spatial depth is suggested through the contrast of a
dense layer of material in the foreground which is suddenly removed to reveal another
137

more distant layer in the background. In order to generate dynamic spatial trajectories
and rotation effects, Stockhausen developed a rotating loudspeaker mechanism which
is surrounded by four microphones (see Figure 8.8). Sounds played back with the
rotating loudspeaker are recorded and reproduced using four corresponding
loudspeakers positioned around the audience. The physical rotation of the
loudspeaker results in a Doppler shift, time varying filtering, phase shifts and other
distortions which are difficult to accurately reproduce electronically [Roads, 1996].
Six distinct spatial movements are used in Kontakte, namely rotation, looping,
alternation, a static distribution with a duplicated source, a static distribution with
different sources, and single static sources. These spatial motifs are implemented at
different speeds and directions and interact with each other, and with the static
instrumental performers. This series of spatial movements is treated in the same way
as pitch and rhythm using a system based on the change in speed and angular
direction. In fact, Stockhausen equates the spatial movements with rhythm, arguing
that changes in spatial location can articulate durations in exactly the same way [Cott,
1973]. The interaction between the instruments and the electronics is one of the most
important aspects of this work and, as in Gesange der Junglinge, Stockhausen
effectively positions the concrete sounds, now a piano and percussion instead of a boy
soprano, within the electronic texture. Dramatic crescendos in the instruments seem
to trigger electronic textures while at other times the pitch and rhythmic motion of the
instrumental parts reflect the spatial motion of the electronic sounds.

Fig. 8.8 Stockhausen with his rotating loudspeaker mechanism

138

8.2 Architectural Spatial Music

New halls for listening must be built to meet with demands for spatial music. My idea
would be to have a spherical chamber, fitted all round with loudspeakers. In the
middle of this spherical chamber, a platform, transparent to both light and sound,
would be hung for the listeners. They could hear music, composed for such adapted
halls, coming from above, from below and from all directions. [Stockhausen, 1975b]

One of the high points of avant-garde electronic music was undoubtedly the
extravagant international expositions of the late 1950s and 60s which featured
multimedia art works and music and performances by such prominent composers such
as Karlheinz Stockhausen, Edgard Varse and Iannis Xenakis. An elaborate
multimedia environment was commissioned by Philips for the 1958 Brussels World's
Fair to showcase their new advances in audio and visual technology. The pavilion
(see Figure 8.9) was designed by the architect and composer Iannis Xenakis in the
form of an electronic poem [Zvonar, 2004]. It contained within its unusual
hyperbolic paraboloid structure a long unbroken projection surface with elaborate
lighting and projection equipment and an eleven channel multi-track tape system
which could be routed to over four hundred loudspeakers. The main musical element
of the program consisted of Edgard Varse' tape piece Pome Electronique, which
was synchronized to the visual effects and dynamically distributed through the space
in nine different sound routes using a switching system controlled with an
additional tape track.
The architect and designer of the Philips Pavilion, Iannis Xenakis, was also a
composer and his tape piece Concrte PH was played as an interlude between shows.
Xenakis had moved to Paris from his native Greece in 1947 and took a position as an
assistant to the architect Le Corbusier. In his spare time, Xenakis studied composition
and produced his first major work, Metastasis in 1954 after studying with Olivier
Messiaen. An architectural influence is evident in the score for this work which
features massed string glissandi in which each individually notated instrument is
arranged according to the large scale, deterministic structure (see Figure 8.10). In a
1954 article, Xenakis describes this approach as follows, "the sonorities of the
orchestra are building materials, like brick, stone and wood. . . the subtle structures of
orchestral sound masses represent a reality that promises much [Hoffman, 2001] .
Xenakis went on to use the structural design of Metastasis as the basis for the wall

139

curvature of the Philips Pavilion and a similar hyperbolic paraboloid structure is

evident in both the Pavilion (Figure 8.9) and the score (Figure 8.10).

Fig. 8.9 The Philips Pavilion at the 1958 Worlds Fair in Brussels

Fig. 8.10 String glissandi, bars 309-14 of Metastasis by Iannis Xenakis

140

Many of Xenakis compositions make use of abstract mathematical theories

such as probability theory, game theory or the kinetic theory of gases. These
mathematical functions were used to structure individual elements within the overall
design, which Xenakis often considered in spatial, almost architectural terms.
Xenakis rejected serialism as a compositional principle, but his mathematically
formalized organization of pitch, duration, timbre, dynamic and spatial location into a
unifying overall structure [Xenakis, 2008] is reminiscent of total serialism. However,
his use of formalized mathematical techniques to design large-scale geometrical
sound masses is very different from the moment to moment form of serialist
composers like Stockhausen. Xenakiss geometrical conception of spatial music was
perhaps influenced by his collaborator at the 1958 Worlds Fair, Edgard Varse, who
made the following comment about the future of spatial music in 1936;
When new instruments will allow me to write music as I conceive it, the movement of
sound-masses, of shifting planes, will be clearly perceived in my work, taking the
place of the linear counterpoint. When these sound-masses collide, the phenomena of
penetration or repulsion will seem to occur. Certain transmutations taking place on
certain planes will seem to be projected onto other planes, moving at different speeds
and at different angles. There will no longer be the old conception of melody or
interplay of melodies. The entire work will be a melodic totality. The entire work
will flow as a river flows [Varse, 1998].

Xenakis continued to create large-scale multimedia art works after the

Brussels Worlds Fair and he referred to this new form of multimedia art as the
polytype, from the Greek polys (many, numerous) and topos (place, space, location)
[Harley, 1998b]. The polytype therefore refers to the many small elements of light
and sound which create and articulate a larger space such as the elaborate audiovisual
system in the Philips Pavilion. Xenakis also suggested that geometrical shapes could
be created through the sonic projection of sounds in defined trajectories around a
loudspeaker array. So for example, a single sound played successively and with a
slight overlap through a circular array of loudspeakers would produce a circle, while
many short impulses played through a group of loudspeakers would produce a sonic
surface. Xenakis utilized Peirre Schaefers concept of spatial relief and divided the
spatial distribution between stereophonie statique (sound emanating from numerous
static points in space) and stereophonie cinematique (sounds produced by multiple,
mobile sources) [Harley, 1998b]. The composer suggested that the different spatial
patterns and distributions could then be structured and related to create a form of
spatial counterpoint.
141

Xenakis attempted to implement these spatial designs in the orchestral work

Terretektorh (1965-66). In this work, the musicians are divided into eight groups and
arranged in six concentric circles (as shown in Figure 8.11 [Hoffman, 2001]) while
the audience is unusually distributed amongst the musicians. Various geometrical
patterns and movements are implemented within the five main spatial distributions
which are used to structure the piece, namely;

stochastically distributed points

sound-planes with internal movement
static sounds
densely woven individual lines
continuous glissandi

Fig. 8.11 Orchestral disposition of Terretektorh

Xenakis would go on to compose numerous large scale multimedia works in

which abstract geometrical designs are used to control the distribution of both
instrumental and electronic sounds, and lights and projected visuals. Hibiki-hana-ma
142

for multi-channel tape was composed for the 1970 Worlds Fair in Osaka and was
performed in the Japanese pavilion through eight hundred loudspeakers arranged
above, around and under the audience. Stockhausen was also present at Osaka 70 and
his music was performed twice a day for 183 days in the German Pavilion by twenty
instrumentalists and singers (see Figure 8.12). The composer himself controlled the
sound projection from a position in the centre of the spherical venue which contained
fifty-five loudspeakers arranged in seven rings, and six small balconies for the
musicians. The design of the auditorium was largely based on Stockhausens
specifications, which are clearly expressed in the quote at the beginning of this
section. As with his earlier works, Stockhausen again used spatial trajectories to
articulate different musical events. However, the spatial distribution was now
implemented in real-time. This approach is somewhat similar to the live spatial
diffusion practised by Pierre Schaefer, albeit with a greater emphasis on geometrical
paths. Stockhausen describes the experience as follows:
I can create with my hand up to six or seven revolutions a second. And you can
draw a polyphony of two different movements, or let one layer of sound stay at the left
side, then slowly move up to the centre of the room, and then all of a sudden another
layer of sound will start revolving like mad around you in a diagonal circle. And the
third spatially polyphonic layer will just be an alteration between front-right and
back-left, or below me and above me, above me and below me, alternating. This
polyphony of spatial movements and the speed of the sound become as important as
the pitch of the sound, the duration of the sound, or the timbre of the sound. [Cott,
1973].

Although Stockhausen would continue to argue for the construction of concert

halls specifically for the performance of spatial music, such as the pavilion in Osaka,
the general trend from the seventies onwards was toward more generalized
multichannel solutions such as Quadraphonics and Ambisonics. Stockhausen was one
of the first composers to adopt an arrangement of eight loudspeakers and he used this
arrangement extensively for the rest of his career. His first composition for this
arrangement, Sirius (1975-77) utilized manual diffusion of electronic sounds, live
instrumentation in the form of trumpet, soprano, bass clarinet, and bass, and an eightchannel version of the rotational table used in Kontakte. In an interview after the
premiere of Sirius, the composer described the use of rotational effects in this work.
Sirius is based entirely on a new concept of spatial movement. The sound
moves so fast in rotations and slopes and all sorts of spatial movements that it seems
to stand still, but it vibrates. It is [an] entirely different kind of sound experience,
because you are no longer aware of speakers, of sources of sound the sound is
everywhere, it is within you. When you move your head even the slightest bit, it

143

changes color, because different distances occur between the sound sources.
[Felder, 1977]

Fig. 8.12 Karlheinz Stockhausen performing at OSAKA 70

8.3 The Perception of Abstract Spatial Designs

The predominance of abstract designs over audibility of their sonorous results is the
basic problem in post-war modernist music [Harley, 1998a].

The development of electronic processes which could dynamically move

sounds through space encouraged composers such as Stockhausen to attempt similar
dynamic processes in instrumental spatial music. In Musik in Raum, Stockhausen
suggests that spatial distribution could be used to structure and articulate different
layers of material, thereby avoiding the homogenous pointillistic textures common to
many twelve-tone compositions. In many respects, this is similar to Brants
conception of spatial music, yet the American was highly dismissive of Stockhausen

144

and he criticized Gruppen as not really being spatial because all the orchestras have
brass, woodwinds, and percussion, so the direction and the tone quality cannot
indicate the source of the material [Harley, 1997]. However, this criticism is not
entirely justified as Stockhausen is clearly using the spatial distribution of the
musicians to articulate different layers of material, albeit with similar instrumentation.
While the delineation of spatial locations with different timbres may be necessary
when many disparate layers of material are operating concurrently, such as in the
music of Henry Brant, this is not necessarily required when each group produces
intermittent, rhythmically and harmonically distinct passages of music, as in the
moment to moment form typically employed by Stockhausen.
One of the most contentious aspects of Stockhausens treatise on spatial music
was his proposal for the serialization of direction based upon the division of a circle
into a series of proportions. Critics argue that the absolute directions specified on
paper will not be perceived by the audience [Harrison, 1999; Harley, 1998a], and the
tests discussed earlier in this thesis would seem to support this view. The results of
these tests illustrate the significant variations in perceived source direction which
occur with electronic spatialization systems and suggest that a reliable and absolute
perception of direction is difficult to achieve electronically. This problem is
exacerbated in a performance setting when the audience is seated at different locations
within the loudspeaker array. The apparent contradiction between the compositional
design and the perceptible results in the music of Stockhausen, and other composers,
is not only evident in their use of space, but also in other areas, as illustrated by the
quote at the beginning of this section. Some composers questioned whether the equal
division of parameters such as pitch, motion, duration and form on paper, translates to
an equal division in the perception of the listener [Bernard, 1987]. The twelve-tone
music of the Vienna school which preceded serialism had primarily focussed on a
serialized pitch row which was, in theory at least, evident in the resulting pointillistic
texture. By 1955, however, Stockhausen had begun to use these serialist procedures
to control the overall character of large groups of material, in what he referred to as
"group composition". In this aesthetic, it is questionable whether the audience is
intended to perceive the absolute and tightly controlled internal relationships within
the composition. Instead perhaps, it is the overall result of these procedures which is
important. Serialism began as a negation of tonality, and serialist procedures serve to
eliminate any perceivable melody or theme, or any defined tempo or rhythm. These
145

purely negative goals could also be achieved by simply randomizing every parameter,
yet the controlled bursts of activity in Kontakte or Gruppen sound anything but
random. In effect, the serialist procedures eliminate any repetition, of pitches,
rhythms or indeed spatial movements, while also maintaining a definite coherence in
the audible material. In these works, direction is therefore divided into a series of
proportions in the same way as every other parameter, and it is the consistency of this
approach which produces the non-repeating, yet entirely coherent material in these
works. The precisely specified spatial locations presented in Musik in Raum are
therefore perhaps not intended to be perceived in an absolute sense, but rather as a
way to remove any recognizable or re-occurring spatial motifs. Stockhausen therefore
uses space to create a sense of perspective in the listener which is not in fact fixed, as
in the classic concert listening experience, but varying. This approach ties in with the
composers overall aesthetic in which the music is similarly removed from a definite
tonal, harmonic or rhythmical centre [Cott, 1973].
Stockhausens use of space, particularly in his early career, suggests that the
composer was well aware of the perceptual issues with his approach. Musik in
Raum contains a detailed and relatively accurate assessment of the perception of
auditory distance and also suggests that a perceptually appropriate scale of directions
should be determined by listening tests. In a later interview, Stockhausen clearly
recognizes the importance of the source material in terms of its perceived spatial
movement:
I can say in general that the sharper the sound and the higher the frequency, the
better it moves and the clearer its direction. But Id also say that the more a sound is
chopped, the better it moves in space. And the sharper the attack of each segment of
a sound event, the better it moves. Whereas a continuous or harmonious sound or a
spectrum of vowels doesnt move very well. [Cott, 1973]

Stockhausen integrated certain spatial parameters, such as direction, speed and

angular movement into the overall serial structure but other parameters such as
distance are used in a much more intuitive and indeed dramatic manner. For example,
while the directional movements in Kontakte are serially controlled, distance is used
to dramatically highlight certain climactic passages. Stockhausens theatrical and
dramatic use of space is also evident in the Helicopter String Quartet (1995), which
uses the sound of the quartet (who are distributed amongst the helicopters) is mixed
with the sound of the helicopter rotors. Stockhausens sophisticated use of spatial
depth and distance is sometimes ignored in certain ideologically driven criticisms (to
146

which Stockhausen was no stranger) which concentrate solely on the serial control of
direction discussed in Musik in Raum. Consider the following statement:
"Stockhausen dismisses the use of space by Gabrieli, Berlioz, and Mahler as being
too theatrical, and argues instead that direction is the only spatial feature worthy of
compositional attention because it could be serialized. [Harley, 1998a]

Although Stockhausen does state that direction is the only spatial parameter suitable
for serialization, he clearly does not consider this to be the only spatial feature
worthy of compositional attention [Harley, 1998a] . As discussed previously in this
Chapter, distance is used extensively in Kontakte and Gesang der Jnglinge but, as
this parameter is highly subjective, and dependent on multiple parameters such as
amplitude, timbre and the nature of the source signal, it is not suitable for serial
control.
Stockhausens acoustic spatial music also displayed an awareness of the
limitations of the medium. His use of spatial movement in Gruppen was one of the
first attempts to replicate stereophony using acoustic instruments and is particularly
successful in this regard due to the relative restraint displayed by the composer.
Spatial movements are implemented, often in isolation, as short, distinct gestures, and
this helps to clarify the perceived movement. A more complex spatial movement is in
fact implemented only once in the dramatic and famous passage shown in Figure 8.4
where a single travelling chord rotates around the distributed brass instruments. The
results of experiments presented in Chapter Six indicate that the presence of an
additional distracting stimulus reduced the localization accuracy for the primary
source [Blauert, 1997]. This suggests that when these spatial movements are isolated
in this way, they are much more likely to be clearly perceived by the audience.
However, if multiple complex trajectories are occurring simultaneously, it will be
much harder for the listener to determine the precise trajectory of each source. In
addition, the fragility of movements created using stereophonic processes has been
clearly demonstrated and this too imposes perceptual limitations on the complexity of
the spatialization process. It would therefore appear to be quite difficult to justify the
elaborate and visually-orientated spatial schemes implemented by composers such as
Varse and Xenakis. Xenakis appeared to reach the same conclusion much later in
his career, as evident in the following comment made by the composer in 1992.
In reality, sound movements are usually more complex and depend on the
architecture of the performance space, the position of the speakers and many other
things. When you want to reproduce such a complicated phenomenon with live

147

musicians playing one after another with amplitude changing in the same way that
you change the levels in a stereo sound projection, sometimes it will work and
sometimes it will not. It depends on the speed of the sound as well as on the angle of
two loudspeakers or musicians, that is, on the relative position of the listener. These
two considerations are equally important.
Xenakis Music, Space and Spatialization, 1992: 67)

While the spatial movement of sound obviously lends itself to graphical

geometrical representation, there is a danger in simply equating our visual perception
with our audible perception. A visual representation of a circle is outside of time, in
that the entire shape can be comprehended all at once, however, this is not the case
when a sound is moved around the audience in a circular fashion. In this case, the
perception of the sound shape will only become clear once the sound has moved
through a full 360 degrees rotation. Creating a spatial counterpoint between different
trajectories based on a graphical representation is therefore difficult, as the precise
shape is not instantaneously clear as it is in the visual realm. Harley illustrates this
problem in an analysis of the Genesis cycle (19623), a large-scale orchestral work by
Henryk Gorecki. The chaotic and highly dissonant material in this piece is organized
spatially into blocks of material, in a similar fashion to Brant and the group
compositions of Stockhausen. However, this piece also attempts to create geometric
spatial trajectories using triangular, polygonal and rectangular arrangements of the
orchestra, as shown in Figure 8.13. This elaborate seating plan necessitated the
repositioning of the audience and performers before each of the three movements.
However, Harley correctly suggests that the slight effect of the different layouts is far
less important than the symmetrical movement of material from left to right, or from
front to back [Harley, 1998a], which could be achieved with any of these
arrangements. The highly visual conception of spatial music envisaged by composers
like Edgard Varse and Iannis Xenakis is perhaps achievable when combined with a
synchronized visual component. However, it is difficult to see how distinct sonic
shapes can be reliably produced even in isolation, with either acoustic or electronic
techniques. The creation of a spatial counterpoint between multiple shapes and
trajectories appears even less realizable.

148

Fig. 8.13 Layout of instruments in Goreckis Genesis Cycle 1974

As mentioned earlier, the aesthetics of Musique Concrte and Elektronische

Musik which developed in the 1950s can be used to illustrate two divergent
approaches to electronic music, and also to the use of space in music. The latter
approach, which was discussed in the preceding Section, represents a continuation of
traditional modes of composition in which abstract structuring systems are used to
organize sounds and their position in space. The alternative approach, which has its
origins in the work of Pierre Schaeffer, Pierre Henry and Musique Concrte, uses the
sounds themselves as a starting point, and the large scale structure and spatial
distribution is derived from the intrinsic properties of the sounds. Over the past fifty
years, this alternate approach has developed into a set of performance practices based
upon the manual diffusion of a stereo source to a loudspeaker orchestra and this
aesthetic will be examined in the following section.
149

8.4 Diffusion, and the legacy of Musique Concrte

The legacy of Musique Concrte can still be seen in the international
community of composers who create stereo electroacoustic works specifically for live
diffusion to a loudspeaker orchestra. The term acousmatic music, meaning sound
which is heard without a corresponding visible cause, is often used to describe this
aesthetic. However, the exact definition of this term is disputed as it may simply refer
to the means of production, i.e. music for loudspeakers alone, or to a particular style
of composition [McFarlane, 2001].
Spatial diffusion was used in the very earliest performances of Musique
Concrte. Pierre Schaeffer had introduced the idea of the movement of sound along
sonic trajectories and the creation of spatial relief through the contrast of static spatial
locations and dynamic sources which would be controlled manually by a performer
using the potentiomtre d'espace. Over the next two decades two-channel stereo and
magnetic tape were adopted as the medium of choice, and composers, led by Pierre
Henry, began to formulate a performance practice based on the diffusion of a stereo
source to a large number of loudspeakers using a special mixing desk. Unlike the
multichannel approach adopted by Stockhausen and others, this loudspeaker orchestra
consists of a diverse range of speakers, chosen for their specific tonal characteristics.
This aesthetic focuses on the temporal, spectral and spatial development of sounds
rather than the relationship between these parameters, and the diffusion process is
therefore used to exaggerate the dynamic, spectral and spatial content of the musical
material already present in the work. In early performances, the intervention of an
engineer was already required due to the technical limitations of magnetic tape (such
as tape hiss and a limited dynamic and spectral range) and in time, these technical
considerations would come to include the diffusion process. The first formalized
system based on this approach was the Acousmonium at the Groupe de Recherches
Musicales (GRM), the collective founded by Schaeffer and Henry in 1958. The
Acousmonium was developed by the composer and technician Franois Bayle, who
took charge of the GRM in 1966, and the engineer Jean-Claude Lallemand [Gayou,
2007] and was conceived as a continuation of Jacque Pullins work on the
potentiomtre despace. The first concert was held in Paris in 1974 and featured a
performance of Bayles Experience Acoustique.

150

Fig. 8.14 An early concert with the Acousmonium

Fig. 8.15 Standard stereo setup (left), diffusion setup (right)

The approach developed by Henry and Bayle was adopted by other composers
and institutions such as the Groupe de Musique de Bourges in Belgium. The
Birmingham ElectroAcoustic Sound Theatre (BEAST) at the University of
Birmingham is another important centre for research in this area. Although these

151

groups differ in terms of the precise technical setup and loudspeaker layout, they use
many similar working methods and techniques. The source material is generally a
stereo CD, and is often the commercial CD release of the piece. The stereo track is
routed to a special mixing desk which allows the diffusion engineer to control the
routing of the stereo track to different loudspeaker pairs. Often this will be a
commercial desk, reverse-engineered so that it takes a stereo signal as the input, and
each individual fader channel feeds a different loudspeaker pair.
The diffusion process is as much concerned with adapting the material for the
particular performance space as it is with the spatial articulation of the material. For
this reason, the loudspeaker pairs are often specifically arranged in an attempt to
preserve the stereophonic image for as much of the audience as possible. Jonty
Harrison, who works at the University of Birmingham, illustrates this approach using
the two layouts shown in Figure 8.15 [Harrison, 1999]. In a normal two-channel
stereo system, only listeners seated in the sweet spot (point (a) in Figure 8.15 (left))
will perceive the stereo image correctly. At point (b), the stereo image will collapse
to the left speaker due to the precedence effect, while at point (c) listeners will
experience a significant hole-in-the-middle effect due to the wide loudspeaker angle at
this location close to the loudspeakers. Meanwhile a distant listener at point (d) will
perceive a drastically narrowed image. Diffusion systems attempt to overcome these
problems through the introduction of additional pairs of loudspeakers and a typical
layout is shown in Figure 8.15 (right). In this arrangement, the main stereo pair has
been narrowed to reduce the hole-in-middle effect for listeners close to the stage.
This is supported by another pair of similar loudspeakers positioned at a wider angle
which can be used to increase the image width as necessary. Additional distance
effects are supported through the use of a distant loudspeaker pair, positioned at the
back of the stage and angled across the stage. Finally, a rear pair is added so that the
stereo image can be extended out from the stage and around the audience. This group
of loudspeakers is described as the "main eight" in the BEAST system and this layout
is described by Harrison as the absolute minimum arrangement required for the
playback of stereo tapes [Harrison, 1999]. More loudspeakers are often added to the
main eight to further extend the capabilities of the system. For example, additional
side-fill loudspeakers are often added to facilitate the creation of smoother movements
from front to back. Various elaborate systems have been developed, such as the

152

BEAST system in Birmingham (shown in Figure 8.16) and the Acousmonium at

GRM.

Fig. 8.16 The full BEAST system

Modern diffusion practice also attempts to articulate musical material by

performing different passages through different sounding pairs of speaker.
Composers do not generally provide a score or notation that indicates how the work
should be diffused, instead it is up to the diffusion engineer to interpret the work and
perform the piece in a way that highlights the musical content and also adapts it to
the specific loudspeaker array and acoustics of the particular performance venue. The
composer and theorist Denis Smalley describes this approach as follows;
Sound diffusion is the projection and the spreading of sound in an acoustic space for
a group of listeners as opposed to listening in a personal space (living room, office
or studio). Another definition would be the sonorizing of the acoustic space and the
enhancing of sound-shapes and structures in order to create a rewarding listening
experience [Smalley et al, 2000].

153

Diffusion has traditionally been controlled manually but some work has been carried
out with automated processes and digital spatialization techniques [Truax, 1999;
Moore, 1983]. Harrison suggests that this approach reflects the primarily manual
processes used in the studio by composers in the early days of Musique Concrte
[Harrision, 1999]. Others, like Smalley have compared the spatial gestures produced
by the diffusion engineer to the physical gestures of traditional instrumental
performers [Smalley, 2007]. However, although the diffusion process clearly adds
something to the performance, the diffusionist can only emphasize the pre-existing
material in the work. Clearly then for a successful performance of a diffusion piece,
the composer must organize the material in a way that supports its eventual
performance. Denis Smalleys theory of spectromorphology details how a gestural
approach can be used as a structuring principle in electroacoustic, and particularly
acousmatic music.

8.5 Spectromorphology and the Gestural use of Space

Denis Smalley developed the theory of spectromorphology as a means of
describing and analysing the electroacoustic listening experience. Smalley studied
with Franois Bayle at GRM and his compositional aesthetic and theoretical writings
developed from the theories of Musique Concrte described by Pierre Schaeffer in
Traite des objects musicaux [Smalley, 1997]. Schaeffers approach emphasized
reduced listening, that is the conscious focus on the intrinsic properties of the sound
and the ignoring of extrinsic, referential properties. However, this form of listening is
highly unnatural and difficult to maintain, as the natural instinct of any listener is to
identify the source and cause of the sound. Smalley uses the term source-bonding to
describe this natural tendency to relate sounds to supposed sources and causes, or to
relate sounds to each other due to a shared origin. In instrumental music, this process
is determined by the physical interaction of the performer with the instrument as the
spectromorphological profile of the sound indicates how the instrument is excited by
the performer. Smalley suggests that in the case of acousmatic music, both the source
and cause of the perceived sound may be detached from a directly experienced or
known physical gesture, a concept he refers to as gestural surrogacy. First order
surrogacy refers to sounds produced by human gestures such as traditional
instrumental music and singing. Second order surrogacy retains some aspects of

154

human gesture but the resulting spectromorphology is not identifiable as a known

musical instrument. Remote surrogacy applies when the sound is not related to
human gestural activity or any other known source, such as unusual synthesized
sounds. Smalley suggests that the way electroacoustic music uses technology to
develop musical gesture beyond the note-based instrumental paradigm is one of its
greatest achievements [Smalley, 2007]. He describes gesture as an energy-motion
trajectory which can be shaped by the composer through the manipulation of the
attack and decay envelope of the sound object. The movement of the sound in space
should therefore support and emphasize the inherent spectromorphological profile of
the sound object. Smalley suggests that weak gestures which are stretched out in time
or evolve slowly are detached from human physicality and are instead perceived in a
more environmental sense. This results in a shift in attention away from the forward
motion of a distinct gesture to a static environmental texture in which the internal
activity of the sound object is the primary focus of attention [Smalley, 2007]. A
composition could therefore utilize either a gesture-carried structure which implies a
degree of forward motion, triggered by some external impetus, or a texture-carried
structure which focuses more on the internal activity of the sound which appears to
act without any obvious external stimulus. In this way Smalley suggests that gesture
and texture can be used as forming principles in a composition and this process is
illustrated in Smalleys acousmatic composition Empty Vessels (1997). This work is
entirely constructed from environmental recordings made by the composer in his
garden. The microphones were placed inside a number of large garden pots so the
resulting recordings capture both the internal resonances of the pots and also the
external environmental sounds of birds, rain and planes. Additional recordings were
also made of just the environmental sounds without the filtering effect of the pots.
The work begins with a struck chord of unknown origin which slowly transforms into
the recorded environmental sounds. This smooth transformation is achieved through
the careful matching of the initial struck chord with the resonant drone of the garden
pots present in the original recordings. The descending glissandos of planes flying
overhead are also smoothly transformed into abstract processed drones which again
are related back to the environmental recordings via the internal resonances of the
garden pots, the empty vessels of the title.
Smalley recognizes that the perception of the listener will be influenced not
only by the spectromorphological profile created by the composer but also by the
155

acoustic of the listening room. He therefore divides the perceived space into the
composed space (which contains the spatial cues created by the composer), and the
listening space (the space in which the composed space is heard), as shown in Figure
8.17. The composed space consists of both the internal space of the sounding object,
such as the resonances of a struck enclosure, and the external space of the
environment containing the sound object, which is made apparent through reflections
and reverberation. The listening space is also subdivided into the personal space,
which relates to the precise position of the listener within the venue, and the diffused
space created by the various loudspeakers distributed around the venue. Smalley
suggests that, in his experience, the perception of a number of parameters will be
different when the work is transferred from a single listener in a studio to a large
performance space [Smalley, 1997]. Spatial depth which is captured using stereo
recording techniques or synthesized using artificial reverberation can easily create
images which appear to originate from outside the array for a single listener.
However, in a larger space these images may instead become superimposed within the
space, rather than beyond the physical loudspeaker positions. Similarly, creating a
sense of spatial intimacy [Smalley et al, 2000] becomes much more difficult as the
size of the listening area increases. The spatial variants suggested by Smalley are
shown in Figure 8.17 [Smalley, 1997].

Fig. 8.17 Denis Smalleys perceived space

Although spectromorphology clearly originated from the acousmatic tradition

and stereo diffusion, this approach can be applied to other compositional aesthetics

156

and technical formats. Spectromorphology can be used to relate and structure a wide
variety of sounds as either gesture or texture, performance or environmental. Human
activity, whether in the form of traditional instrumental performance or the
manipulation of non-traditional instruments (such as the garden pots in Empty
Vessels), can be related to synthesized sounds via the shaping of dynamic spatial and
spectromorphological gestures. In addition, environmental recordings can be related
to continuous synthesized or processed sounds without accented attack and decay
envelopes. Soundscape composition [Truax, 2008] is a compositional aesthetic which
utilizes the latter approach and is predominantly based upon environmental sounds
and textures. Luc Ferraris Presque Rien No. 1 (1970), which consists solely of
layered recordings from a day at the beach with a minimum level of manipulation of
the material, is a well known example of this style of composition. The Vancouver
Soundscape (1970) by the World Soundscape Project similarly consists of direct
reproductions of natural soundscapes with a minimum level of human intervention.
Since the seventies, soundscape composition has developed beyond this minimalist
approach to include digital synthesis and multichannel techniques and the Canadian
composer Barry Truax is one of the chief exponents of this aesthetic. His early work
used granular synthesis and a quadraphonic speaker system to create highly textural
works with a strong environmental character. In later works such as Pacific (1990),
or Basilica (1992), Truax used granulated environmental recordings with an
octophonic, eight-channel array and multiple decorrelated granular streams to create
an immersive sonic environment. Truax has argued that the avoiding the
representational meaning of environmental sounds is difficult, stating that
environmental sound acquires its meaning both in terms of its own properties and in
terms of its relation to context [Truax, 1996]. Despite this difficulty, environmental
recordings have also been used in a more symbolic fashion in which the different
recorded sounds and spaces are used to tell a sort of narrative. Various composers
and theorists have explored this compositional aesthetic and suggested various
symbolic interpretations of different spaces and movements [Trochimczyk, 2001;
Wishart, 1985].
The gestural use of space suggested by Denis Smalley in his theory of
spectromorphology originated in the acousmatic tradition of stereo diffusion, but this
idea is equally applicable to compositions for multi-channel loudspeaker arrays, or
mixed-media electroacoustic works. The notion of gestural shaping also suggests an
157

obvious approach to works which combine acoustic and spatialized electronic sounds
as the spectromorphological profile of the synthetic sounds can be deliberately
designed to match or mimic the instrumental gestures of the performers.

158

9 Electroacoustic Spatial Music

The addition of live performers introduces a number of difficulties in the
performance of spatial music. The spatial distribution of performers around the
audience can be very effective but is logistically challenging and highly dependent on
the specific layout of the performance space. Compositions for live performers on
stage and spatialized audio are particularly challenging as the static location of the
live performers provides a frontally-biased visual and audible focus which can
conflict with the non-visual and spatialized audio. Linking the musical gestures of the
instrumental performer with the spatial gestures of the electronic part is a significant
challenge, as the spatialization process is often not related to the musical instrument in
any obvious way. In addition it is rarely practical or possible for a single performer to
concurrently play and diffuse a musical instrument
In general, electroacoustic music of this sort has taken one of two different
approaches. The first and most common approach uses traditional instruments in
conjunction with a previously prepared electronic track and this mixed-media
approach was really the only technologically feasible solution prior to the
development of digital computing technology. As digital processing power increased,
the processing and spatialization of instruments in real-time became feasible and by
the end of the 1980s various composers had completed major works for live
instruments and real-time electronic processing.
The alternative approach, which will be looked at later in this chapter,
attempts to incorporate the spatialization process into the design of entirely new
musical instruments. This approach has a number of potential benefits but is highly
dependent on specific, idiosyncratic devices.

9.1 Fixed Media Electroacoustic Spatial Music

Stockhausens Kontakte (1958-60) is one of the earliest and most successful
electroacoustic compositions to include live instrumental performers and spatialized
electronic sounds, although this piece can be performed either as a tape piece or as a
mixed-media work. Stockhausen overcomes the conflict between the static
performers and the spatially dynamic electronic part in various ways. Firstly, the
scored instrumental material is tightly integrated with the synthesized electronic
159

sounds within the overall serialist structure, as discussed in detail in Chapter Eight. In
addition, various points of contact between the static instrumental sounds and
dynamic electronic textures are created through crescendos in one part which provoke
a response from the other. Harrison suggests that the unusual loudspeaker
arrangement also contributes to the success of this piece. As the four loudspeakers are
placed in a cruciform arrangement rather than in the corners of the space, this means
that three of the four loudspeakers and the live instruments (piano and percussion) are
in front or to the sides of listeners with only one loudspeaker behind [Harrison, 1999].
Therefore, the instrumental sounds and most of the electronic parts are produced
largely from in front of the audience, this helps to unify the two parts into an
integrated whole.
Although Kontakte successfully integrates the static instrumental performers
and the dynamic spatialized electronic parts within the work as a whole, in much of
the piece the static spatial locations of the performers is actually deliberately
contrasted with the dynamic spatial trajectories in the electronic part. The various
points of contact between the two spaces is then heightened for dramatic effect in
much the same way as other elements in this piece such as pitch and rhythm. The
composer himself wrote that in works which combine electronic and instrumental
music, there remains the problem of "finding the larger laws of connection beyond
that of contrast, which represents the most primitive kind of form [Stockhausen,
1975a]".
In electroacoustic works such as Kontakte, the performers must perform while
listening to a click track in order to synchronize with the tape part. However, some
composers were uncomfortable with the rigidity of this arrangement, and the French
composer Pierre Boulez was a particularly vocal critic, stating that give and take
with the tempo of a piece is one of the basic features of music. It is perhaps
unsurprising that Boulez was so attuned to this particular limitation considering his
career as a conductor and he would go on to compose for one of earliest systems for
the real-time processing and spatialization of sounds, an approach which would
capture all the spontaneity of public performance, along with its human
imperfections [Boulez et al, 1988].

160

9.2 Real-Time Electroacoustic Music

The French composer Pierre Boulez was, along with Karlheinz Stockhausen,
one of the chief exponents of serialism and modernist contemporary music. Like
Stockhausen, Boulez was profoundly influenced by the twelve-tone compositions of
Anton Webern and the total serialism of Olivier Messiaen. Boulez did not at first
share the Germans interest in electronic or spatial music, preferring instead to
compose acoustic works in a totally serialized, punctual style. By the late fifties and
early sixties however, Boulezs compositional approach had become less strict and his
music began to incorporate electronic sounds and spatial effects. He was particularly
interested in the combination of electronic and acoustic sounds, the fertile land from
his 1955 article on the future of electronic music, la limite du pays fertile (At the
edge of the fertile land) [Husler et al, 1985]. His first electroacoustic work,
Posie pour Pouvoir (1958), was scored for three orchestras and five-track tape with
an electronic part constructed from recorded and processed instrumental sounds. This
arrangement was perhaps influenced by Stockhausens Gruppen, as Boulez was one
of the three conductors at the premiere of this piece. As with many of his
compositions, Posie pour Pouvoir was later withdrawn by Boulez due to his
dissatisfaction with the electronic part. Boulez was particularly unhappy with the
limitations imposed by the click track, which was required to synchronise the human
performers with the tape. In an interview in 1985, Boulez described this issue as
follows;
I have conducted pieces of other composers who use tape and have learned that one
is at a disadvantage because one must constantly take the synchronization into
account, leaving no room for the coincidences of interpretation. I don't mean this
only in the sense that one can select structures at will, but that the gesture of
interpretation as I would like to call it is completely paralyzed, because one
must pay attention to too many things that have little to do with the actual
performance. The time of a tape recording is not psychological time but rather
chronological time; by comparison, the time of a performer a conductor or an
instrumentalist is psychological, and it is really almost impossible to interconnect
the two. [Husler et al, 1985]

Faced with these difficulties, Boulez' abandoned his work with electronic sounds and
it would be over twenty years before Boulez would return to this area. His
opportunity to do so came in 1970 when he was asked by the French president
Georges Pompidou to create and head a new institute for musical research and
composition, the Institut de Recherche et Coordination Acoustique/Musique
161

(IRCAM). The institute opened in Paris in 1977 and, along with the Centre for
Computer Research in Music and Acoustics (CCRMA) at Stanford, IRCAM quickly
became one of the most important centres for research in digital music technology.

Fig. 9.1 IRCAMs 4X (left) and Matrix 32 (right) digital processors

By 1980 IRCAM had developed the first prototype of a real-time digital

system which was expressly designed for the live performance and spatialization of
electroacoustic music. The final version of this system was completed in 1984 and
consisted of the 4X processor (Figure 9.1 left) and the Matrix 32 (Figure 9.1 right).
The 4X was a real-time digital signal processor for the analysis, transformation and
synthesis of sounds in real-time which was designed at IRCAM by Giuseppe Di
Giugno and Michel Antin [Boulez et al, 1988]. The 4X could store, manipulate and
recall up to four seconds of digital audio using various software modules, or patches,
and a real-time operating system and event scheduler (developed by Miller Puckette,
Michel Fingerhut and Robert Rowe). The Matrix 32 was designed by Michel Starkier
and Didier Roncin and functioned as a programmable and flexible audio-signal traffic
controller for the dynamic routing of audio signals between the instrument
microphones, the 4X and the loudspeakers. Boulez would go on to use both devices
in the electroacoustic composition Rpons, a flagship project for IRCAM, which was
premiered as a work-in-progress in October 1982 at the Donaueschingen Festival in
Germany [Husler et al, 1985]. Rpons is scored for a 24-piece orchestra, six solo
instruments (harp, glockenspiel, vibraphone, cimbalom and two pianos), six
loudspeakers and IRCAMs real-time digital signal processing system. Boulez was

162

assisted by Andrew Gerzso in the design of software patches for the 4X processor and
several technicians were required during the performance to manage the various
devices. Boulez positioned the instruments and loudspeakers in Rpons using a very
similar layout to that of the earlier work, Posie pour Pouvoir (see Figure 9.2).
According to the composer, the un-amplified and unprocessed orchestra is situated in
the centre of the auditorium in order to provide a clear visual focus for the audience
[Boulez et al, 1988]. The audience is therefore seated in-the-round, surrounding the
orchestra, and within the circle formed by the six soloists and the loudspeaker array.
The title of this piece comes from a medieval French term for the antiphonal choral
music discussed earlier in this thesis. This call-and-response process is implemented
throughout Rpons and the dialogue between the soloists and the main orchestra is
particularly reminiscent of medieval choral antiphony. Boulez relates this process of
multiple responses (the orchestra) to a single question (the soloist) to the notion of the
multiplication and the proliferation of sounds. This idea is also implemented
electronically as various processes multiply single instrumental notes into a multitude
of notes, chords or timbres, all of which are related to the original in a clear way.

Fig. 9.2 Layout and spatialization diagram for Rpons

After an opening movement performed solely by the central orchestra, the six
soloists enter dramatically with different but concurrent, short arpeggios. These
instrumental arpeggios are captured by the 4X system and circulated around the
loudspeaker array in the trajectories shown as coloured arrows in Figure 9.2. The
composer describes the effect of this passage as follows;
163

The attention of the audience is thereby suddenly turned away from the centre of the
hall to the perimeter, where the soloists and the speakers are. The audience hears the
soloist's sounds travelling around the hall without being able to distinguish the paths
followed by the individual sounds. The overall effect highlights the antiphonal
relation between the central group and the soloists by making the audience aware of
the spatial dimensions that separate the ensemble from the soloists and that separate
the individual soloists as well. [Boulez et al, 1988]

The various spatial trajectories are therefore not intended to be perceived directly but
are instead used to distinguish the different electronic gestures associated with each of
the six soloists. This effect is further exaggerated through the use of different
velocities for each spatial trajectory, which are determined by the amplitude envelope
of the original instrumental gesture. Each instrument has a similar amplitude
envelope (see Figure 9.3) consisting of a sharp attach and an slowly decreasing decay,
however, the duration of the decay depends both on the pitch of the note and also the
instrument on which it is played [Boulez et al, 1988]. Therefore, the velocity of each
trajectory decreases proportionally to the decay of the associated instrumental
passage, linking the instrumental gesture of the soloist to the spatial gesture of the
electronic part.

Fig. 9.3 Amplitude envelopes for the six soloists in Rpons

The dynamic spatialization is implemented using an unusual switching mechanism

rather than a panning scheme such as stereophony or Ambisonics, perhaps to ensure
that the system responds immediately to the live input without any delay, or latency.
The 4X system controls the switching of the input signal from loudspeaker to
loudspeaker with a series of flip-flop modules. These modules are controlled via a
timing signal whose frequency changes in proportion to changes in the amplitude of
the sounds waveform envelope, which is continuously generated by an envelope
follower module [Boulez et al, 1988]. Therefore, as the input signal decays, each flipflop module holds the signal at each speaker for longer before switching to the next

164

loudspeaker, slowing the spatial trajectory. This process illustrated in Figure 9.4 for a
system with four loudspeakers, and hence four flip-flop units.

Fig. 9.4 The spatial switching mechanism used in Rpons

The themes of multiplication and displacement are further developed through the
recording, processing and playback by the 4X processor of arpeggiated chords
performed by each soloist. Boulez developed the idea of an arpeggio as a
displacement in time and pitch which may be applied, not just to the notes of a chord,
but also to the spreading of the electronic material in pitch, time and space. Boulez
describes this process as an arpeggio (the multiple recorded and processed copies
created by the 4X) of an arpeggio (the sequentially played notes of the instrumental
chord) of an arpeggio (the individually cued soloists) [Boulez et al, 1988]. The
speaker switching mechanism could also be interpreted as a form of spatial arpeggio
and this is perhaps another reason why this switching mechanism was used instead of
a standard panning algorithm.
As with many large-scale works of spatial music, Rpons has been performed
relatively infrequently. As always, finding a venue which can accommodate the
various distributed musicians and loudspeakers is difficult, and the dependence of this
piece on highly specific hardware is an additional limitation in this case. IRCAMs

165

real-time processor was, however, highly prescient in its combination of a general

purpose digital processor and customised software to implement the required
processing. As computing power became more readily available and affordable,
IRCAM moved away from hardware devices such as the 4X toward the development
of software for general purpose personal computers. The real-time scheduler and
operating system developed for the 4X was later developed by Miller Puckette into
the graphical programming language Max/MSP which is now used extensively in
electroacoustic performances. Various other software packages have also been
developed by IRCAM for audio analysis and synthesis including the spatialization
package Spat, which was discussed previously in Chapter Three.

9.3 Real-Time Spatialization

Although modern personal computers are much more powerful than IRCAMs
4X processor, many of the same issues still apply. The interaction between the
instrumental performer and the electronic part is crucial in any electroacoustic
composition, and creating a clearly perceivable relationship between these two
mediums remains a significant challenge. While a clear contrast between these two
elements can be used as a compositional element, this approach is somewhat limited.
Dialogue de l'ombre double (1985) was completed by Boulez shortly after Rpons
and this work for solo clarinet and pre-recorded tape illustrates how the contrast
between these two mediums can be successfully used as a compositional device. The
dialogue of the title is between the live clarinettist and the pre-recorded and
electronically processed clarinet on tape, and Boulez dramatically heightens the
contrast between these two elements using lighting and spatial effects. During
instrumental passages, the un-amplified clarinettist is positioned in the middle of the
hall and audience while the electronic responses on tape are produced using six
loudspeakers positioned around the audience. Lighting is used to further heighten this
contrast as during instrumental passages the clarinettist is brightly lit with a spotlight
which is turned off for the electronic response of the tape part.
The composer Denis Smalley, when asked how his approach to spatial
diffusion changes when additional human performers are added gave this response.

166

"The same factors are present, but they are considered differently because the focus
of a live performance visually and musically - is the performer. So, I don't want to
use as full a diffusion system as I do for tape pieces, because overdoing the diffusion
will tend to undermine the carefully considered musical relationships between the live
performer and the content of the acousmatic domain. - If you have more than one
performer, there are different considerations, because the visual and sonic weight in
front, on stage, is increased. Therefore, I think that there is more for the eye to focus
on and follow, and the acousmatic possibiities become reduced. For example, one
can't have a lot of visual silences on stage, where people are sitting doing nothing. "
[Smalley et al, 2000]

Smalley suggests that this problem is apparent in performances of Edgar

Varses Deserts (1950-54) for wind, percussion and electronic tape [Smalley]. In
two sections of this work, the musicians on stage sit quietly, doing nothing and the
electronic tape part dominates. Smalley suggests that this creates a visual silence on
stage that contrasts uncomfortably with the required acousmatic perception of the
tape, as the human performers provide an unavoidable visual focus for the audience
which conflicts uncomfortably with the acousmatic electronic part. The extent to
which this is a problem in a particular work is however somewhat subjective, and
other composers such as Natasha Barrett have suggested that this represents a broader
dichotomy between the sonic demands of the composition, and its performance.
Barrett: Sometimes. For example, unless I am composing a purely theatrical work,
my initial agenda is sound, whereas the performers primary agenda is performance,
and they are not necessarily the same thing. This is, of course, not always the case,
but either way we need to integrate the demands of both performance and sound. For
example, in Symbiosis, there istoward the enda stretch of three minutes for solo
computer.
Otondo: There are also parts where it is almost purely instrumental.
Barrett: Exactly, and in Symbiosis, this is one way in which I balance the elements: to
give both parts solos and to have both coming together at important meeting points.
In those three minutes of solo computer, the performer has to find a way to
compose herself. Everybody is looking at her, and she is a performer having to
listen. In that section the theater emerges as she demonstrates the act of listening and
not just sitting there waiting for the next thing to happen. You see a string quartet
play a piece with computer or tape, particularly when they play with a click track,
and they look like they are thinking, When do I come in next? What time is it?

[Barrett et al, 2007]

In the opinion of the author, the contrast between the visible and static performer, and
acousmatic, spatially dynamic sounds can be problematic, but only if this issue is not
addressed by the composer within the overall aesthetic. In addition, this is much less
of a problem when the performers are distributed around the audience, or when
smaller numbers of musicians are involved.

167

As stated earlier, perhaps the biggest issue is forging some connection

between these two elements beyond that of contrast. Smalley suggests that the
invisible acousmatic content in a mixed-media work must share gestural or
behavioural characteristics with the instrumental part if both elements are to be
regarded as equally significant and since the acousmatic component is competing with
the visible actions of the performer, it must have a strongly articulated presence
[Smalley, 1996]. While this is certainly true, the amplification of the instrumental
parts must be considered as an equally significant issue. In smaller venues, the
acoustic instruments require little or no amplification, and in this case the positioning
of loudspeakers close to the performers with a small amount of amplification of the
instruments, can be sufficient to forge a connection with the amplified, electronic
sounds. In larger venues, the greater issue is often the disconnect between the
perceived position of the amplified, instrumental sound (which often collapses to the
nearest PA loudspeaker), and the visual position of the performer on stage. If this
issue is not addressed (perhaps through the positioning of performers close to the
loudspeakers which are being used to amplify their particular instruments), then a
connection between the acousmatic and instrumental sound may be achieved but a
connection to the visual performer may not.

168

10 Spatial Music Composition

In this Chapter, a number of original works of acoustic, electronic and mixedmedia spatial music will be discussed and analyzed. These compositions are
presented chronologically, beginning with Discordianism, the piece which as
discussed previously, provided much of the impetus for this thesis.

10.1 Case Study I: Discordianism

Discordianism is the earliest piece of spatial music written by the author
which is included in this thesis and as we shall see, the use of amplitude panning and
a quadraphonic array proved to be highly problematic and a relatively inexperienced
choice. This quadraphonic tape piece was not composed with Denis Smalleys theory
of spectromorphology in mind, however, this theory is a useful way of classifying
different types of material and is referenced extensively in the following discussion.
The primary source material consists of freely improvised material produced
using an electric guitar and a wide assortment of effects pedals. Various traditional
instrumental gestures such as plucked harmonics were extracted from the original
guitar recording (the first order surrogacy material using Smalleys terminology). A
sample of a single note on a prepared piano which is introduced later on in the piece
also provides a similar degree of first-order surrogacy. This straightforward
instrumental material is slightly abstracted using various processes such as multichannel granulation. This material, although abstracted, still retains aspects of the
original instrumental gesture and so would be classified as second order surrogacy.
However, much of the original recording is also second-order as this material was not
produced directly with the guitar but instead by manually controlling the effects
pedals to produce bursts of feedback and other sustained tones. For example, a
feedback pedal (which essentially loops the output of an effects pedal(s) back into its
input) allowed some highly unusual effect such as controlling the pitch of feedback
using the tone control on the guitar. The resulting sounds therefore contain a degree
of human gestural activity, but the spectromorphological profile is not particularly
identifiable as a musical instrument. The granulated material fulfils a similar function
as the temporal reordering and pitch shifting produce bubbling, irregular textures
which still retain a degree of human gesture. Finally, granular time-stretching and
169

resonant filters are used to create a variety of sustained drone including a heavily
distorted and sustained drone taken untreated from the original performance. The
various drones represent the most abstract material, and hence remote surrogacy.

10.1.1 Analysis
The title of this piece is a reference to the modern, chaos-based religion
founded in either 1958 or 1959 which has been described as both an elaborate joke
disguised as a religion, and as a religion disguised as an elaborate joke. The number
five is highly important in this parody religion and Discordianism is structured
accordingly. It is in 5/4 time, it is composed of five, twenty-five bar sections, and it is
approximately five minutes long. It also uses elements of chaos theory in its
construction, specifically in the granular synthesis algorithm which incorporates the
classic population-growth model, and this chaotic process, combined with the
preciso e meccanico rhythmical activity is intended to reflect the irreverent
philosophy referred to in the title.
The opening section of this work is marked by repeated sequences (in
multiples of five) of a plucked guitar harmonic which cycle around the four channels
and builds in dynamic to a sudden decrescendo. This first-order surrogacy gesture
interacts with the other material in various ways, sometimes triggering a short burst of
feedback, sometimes effecting the activity or density of the bubbling, spatial
granulated texture operating in the background, or sometimes provoking simply
silence. Gradually, the density and activity of the granular texture increases until at
the one minute mark, a burst of feedback triggers a heavy, distorted feedback drone in
both front channels. The sustained nature of this drone extends the second-order
surrogacy even further, although occasional human gestural activity from the original
performance remains. The short feedback bursts now function in a structural sense,
marking out five bar intervals and triggering the introduction of new material such
rapid granulated tremoloed notes and waves of white noise produced by resonant
filters. These additional drones slowly drift around the four channels, in contrast to
the static distorted drone locked to the front channels. The dynamic and density of the
material steadily increases before a brief decrescendo which is immediately followed
by a sudden, highly dynamic reintroduction of all the material, triggered again by a
short burst of feedback.

170

This highly dramatic point signals the start of the third and final section of this
work and the gradual introduction of four layers of percussive material proceeding at
different tempos from the four corners the array. It also forcefully reintroduces both
first-order and remote surrogacy and finally confirms the connection between them in
the form of spatially distributed layers of clearly instrumental material, operating at
different tempi, and a granulated version of the distorted drone from the second
section. This new, heavily distorted drone now moves dynamically in space with
dynamic and irregular lateral movements created by the stereo spreading of the
granulation process and a rapid back and forth movement between the front and rear
channels using amplitude panning. The final section steadily increases in dynamic
and intensity as more layers of rhythmic and granulated material are added building to
a sudden crescendo and rapid cut-off before a brief, swinging spatial movement of the
distorted drone which is finally, cut off by one last burst of feedback.

10.1.2 Critical Evaluation

This early work was written for a quadraphonic tape concert, and then later
remixed to two-channel stereo for a CD release. As such, it provides an excellent
opportunity to assess various issues related to the performance and indeed the
function of space in music.
During performances of this piece, the opening section was relatively robust in
terms of the spatialization. However, this is perhaps to be expected as in this section
each channel is primarily just used as a single point source. The sustained drones in
the second section also seem to function relatively well, although the degree of
envelopment was certainly reduced in comparison to other eight channel works which
were performed at the same concert. The amplitude panned drone which is highly
prominent in the third and final section appeared to be the most affected by the change
in listening environment. The spatial movement of this material was relatively clear
in a studio environment but the increased reverberation, greater inter-loudspeaker
distance, and distributed listening positions significantly distorted the perceived
trajectory to the point where no clear direction of front-back movement could be
distinguished in the performance venue. These, admittedly highly subjective
experiences, would seem to agree with the results of more formal evaluations
discussed earlier in this thesis.

171

In this piece, spatialization is primarily used to clarify different rhythmical

material and to a less extent, to articulate the various gestures. As noted earlier, this
work was composed during the very early stages of this thesis and the lack of an
underlying rationale as to the approach to spatial movement is perhaps all too clear.
Clear gestural relationships between the repeated plucked harmonic and granulated
drone textures are created, however these relationships could have been further
heightened through the dynamic movement of the drone material in response to the
triggering plucked harmonic. The rhythmical approach adopted here was influenced
more by Gyrgy Ligeti than Henry Brant, yet the use of spatial distribution to clarify
and distinguish different layers of electronic drones and rhythmic material is highly
effective.
The two channel stereo reduction also provides an interesting point of
comparison as to the benefit of the spatialization process. The stereo version, which
was necessary for a CD release, was professionally mastered and its overall dynamic
range is therefore reduced in comparison to the four channel mix used for
performances. Upon listening to the four channel mix after the stereo version, the
first obvious difference is that the different layers of material are much easier to
distinguish when spatially distributed among a greater number of channels. The
degree of intelligibility in the stereo mix could perhaps be increased using
equalization to carve out individual spectral space for each layer. Yet much of the
interest in this piece rests in the similarities between the spectromorphological
profiles of the different surrogate layers and this would presumably be affected by the
spectral adjustments necessary. Most of all, the dramatic and expressive benefits of
spatialization are clearly demonstrated by the dynamically moving, heavily distorted
drone in the final section which is significantly static in the two channel reduction.

10.2 Case Study II: Sea Swell (for four SATB choirs)
The initial inspiration for this work came from some experiments carried out
by the author with granulated white noise. When a very large grain duration was
employed the granulation algorithm produced a rolling noise texture which rose and
fell in a somewhat irregular fashion and was highly reminiscent of the sound of
breaking waves. This noise texture was originally intended to accompany the choir,

172

however in practice this proved to be unnecessary as the choir could easily produce
hissing, noise textures which achieved the same effect.
The choir is divided into four SATB groups positioned symmetrically around
the audience and facing a central conductor. The piece utilizes many of Brants ideas,
albeit with a much less atonal harmonic language. In particular, spatial separation is
used to clarify and distinguish individual lines within the extremely dense, sixteen
part harmony, which is shown in Figure 10.1. If this chord was produced from a
single spatial location, it would collapse into a dense and largely static timbre.
However, this problem is significantly reduced when the individual voices are
spatially distributed. Certain individual voices are further highlighted through offset
rhythmic pulses and sustained notes, as shown in the score excerpt in Figure 10.2.

Fig. 10.1 Sea Swell harmonic structure

Exact rhythmic coordination is difficult to achieve when the musicians are

spatially separated and hence a slow, regular tempo is used throughout. The beat is
further delineated through the use of alliteration in the text as the unavoidable sibilant
s sounds mark the rhythmic pulse (see Figure 10.2). These sibilant sounds are also
used in isolation at the beginning of the piece, to create a wash of noise-like timbres
which mimic the sounds of crashing waves suggested by the title. Each singer
produces a continuous unpitched sssssssssss sound which follows the dynamic
pattern shown in Figure 10.3. The angled lines indicate the dynamics which are
divided into three levels, below the bar line indicating fading to or from silence, above
the bar line indicating maximum loudness and on the bar line indicating a medium
volume. The note durations are provided to indicate durations for the dynamic
changes and do not indicate pitch. Instead a continuous ssssss sound is produced
for each entire phrase.

173

Fig. 10.2 Sea Swell rhythmic alliteration

Fig. 10.3 Sea Swell notation example for opening sibilant section

10.2.1 Analysis
Brant used the term spill to describe the effect of overlapping material
produced by spatially distributed groups of musicians. However, Brant was also
aware of the perceptual limitations of this effect and he noted that precise spatial
trajectories became harder to determine as the complexity and density of the spatial
scene increases. The opening of this piece will contain a significant degree of spill
due to the common tonality of each part and will thus be perceived as a very large and
diffuse texture. Large numbers of these sibilant phrases are overlapped in an irregular
fashion to deliberately disguise the individual voices and create a very large, spatially
complex timbre of rolling noise-like waves. This opening section can be seen in full
in Figure 10.4.
The loose ABA structure of this work ends with a shortened refrain of the
opening section discussed earlier. The B section consists of a lyrical and highly
melodic progression of layered cannons. The lyrics consist of some fragments of
poetry by Pablo Neruda as well as some original lines by the author. Three different
lines are cannoned between the sopranos, altos and tenors in a form of spatial and

174

melodic counterpoint, while the basses reintroduce the sibilant sounds from the
introduction.

Fig. 10.4 Sea Swell introductory section

10.2.2 Critical Evaluation

The opening hiss section of this piece has worked extremely well in practice,
particularly once all of the sixteen voices have entered. The significant amount of
reverberation and the similar timbre of each part is such that the individual voices
largely disappear into a very diffuse and spatially dynamic noise texture which is
highly enveloping and quite evocative. The use of alliterative s sounds is also quite
effective as a rhythmical device as these sibilants are quite pronounced and
prominent. While this is highly useful rhythmically, care must be taken when hissing
sounds are used in conjunction with normal vocal parts (e.g. bars 45-66) to ensure that
the level of the hiss sounds does not dominate. The spatial separation of the dense
harmonies and multiple canons is also quite beneficial as this separation really helps
to clarify and separate the multiple, layered cannons.

175

10.3 Case Study III: String Quartet No. I

The first part of this piece to be composed was the granulated and timestretched recordings of whole-tone trills used in the electronic part. The spatially and
spectrally dynamic drone possessed a clear timbral relationship to the acoustic sound
of a quartet, and the dynamic spatialization of this part could clearly enhance the
existing dynamic of acoustic quartet. However, this scenario left little room to freely
compose a string quartet and this approach indeed proved to be exceedingly difficult
in practice as it took a period of six months before any progress was made with the
instrumental part.
The issues raised by Smalley and discussed in the previous section will clearly
be of concern in a work of electroacoustic spatial music written for string quartet. A
string quartet will provide a strong, frontally biased visual focus for the audience
which can only be avoided if the musicians are distributed around the audience. In
this work, the amplified string quartet (using the front loudspeakers, either the house
PA or the front loudspeakers of the surround array) remains on stage while the
spatialized electronic part is used to extend and widen the acoustic sound of the
quartet from the stage and out into the hall. However, due to the unavoidable frontal
bias of this layout, the electronic part only rarely moves entirely to the rear, and
functions more as a dynamic, spatial and timbral extension of the acoustic sound of
the quartet.

10.3.1 Analysis
The primary concept behind this piece is of an endless ascent, inspired by the
optical paradox of the Penrose stairs (see Figure 10.9) or its auditory equivalent, the
shepard tone. This concept is implemented using an ascending dominant seventh
scale on G, which is equivalent to the Mixolydian mode. This mode uses the same
series of intervals as the major scale (C major in this case), except the fifth is taken as
the tonic. Although this rather straightforward progression is the basis of the entire
work, it is generally disguised and distorted in different ways. At the beginning of the
piece, the cello begins with the tonic G, the first note in the progression, however, the
other three parts respectively start on the fifth, sixth and seventh notes in the series.
Similarly, different rhythmic groupings are used for each part and this is particularly
evident in the opening ascent which is shown in Figure 10.10. The cello holds each
176

note for twelve crotchets, the viola for ten, the second violin for eight and the first
violin for six, and in addition, the higher three parts begin after a single crotchet rest,
in order to reduce the number of times more than one part changes note
simultaneously and disguise the exact harmonic root of the progression.

Fig. 10.5 The Penrose stairs

Fig. 10.6 Opening eight bars of String Quartet No. I

The electronic part is constructed from granulated and time-stretched

recordings of whole-tone trills played on the individual instruments of the quartet.
The granulation process transformed each monophonic recording into a wide,
stereophonic drone and the time-stretching process lengthened the original rapid trill
to a slow and irregular glissando. Composers such as Barry Truax [Truax, 1990] and
Curtis Roads [Roads, 2004] have noted that when sounds are granulated using longer
177

grain durations, the original timbre is much more apparent, and this approach was
adopted here. As a result, the granulated drone has a similar timbre to the original
recordings, and therefore to the acoustic quartet. In addition, the bowing action which
is clearly audible in the original recordings is also apparent in the granulated drone.
Using Smalleys terminology, the drone retains aspects of the spectromorphological
profile of the original instrumental gestures, which will in turn bond with the live
instrumental gestures of the quartet. Multiple layers of these drones were finally
superimposed to create a very dense drone which nonetheless retains the
spectromorphology of the original instrumental gesture.
The large scale structure of this piece is delineated by a series of six
crescendos, which is clearly illustrated in Figure 10.11. At first, the electronic part is
positioned to the front, with the quartet, and is harmonically restricted to the same
dominant seventh scale. As the dynamic increases, the electronic drone becomes
increasingly chromatic as more notes, not present in the original scale are introduced,
and by the first crescendo, all twelve pitches are present in the electronic part. At this
point, the electronics move from the front to a wider, 90o spread. As the quartet begin
the following syncopated section (beginning at bar 45) the electronic part mimics the
short, stabbing rhythms of the instruments and extends them spatially. The
descending progression which ends at the third crescendo is reminiscent of the
opening ascent, but now the tone intervals of the original progression are filled,
resulting in a more chromatic descent by the quartet which reflects the corrupting
influence of the electronic part. The spatialization of the electronics matches the
descending progression of the quartet as it collapses from a wide spatial distribution
to a narrow, frontal position matching the position of the quartet.
The second section (beginning at bar 98) opens with a single, granulated cello
trill on the root G, diffused to the rear of the hall. This is the first time that the tonic is
clearly revealed, however, exact harmonic root is still slightly uncertain due to the
slow and irregular glissandoing of the time-stretched trill. The quartet explores this
idea with a series of slow, descending dominant seventh progressions which are
continually interrupted by glissandos and the introduction of chromatic notes by the
quartet and the electronics. By the fourth, and lowest crescendo, the quartet has
resolved to a somewhat stable harmonic centre, but this quickly disintegrates into the
penultimate ascent which is primarily driven by a two and a half octave upward
glissando on the cello. The fifth and final ascent is essentially a refrain of the initial
178

progression with a significantly increased dynamic. The final eight bars again reveal
the slow glissando in the electronic part which is finally echoed and resolved by the
cello and viola.

Fig. 10.7 Large scale structure of String Quartet No. I

Fig. 10.8 Diffusion score example

10.3.2 The Spatialization Process

The electronic part in this quartet is organized in a number of stereo files
which are cued and triggered by the diffusionist using a Max/MSP patch. As the
electronic parts are non-rhythmical, no click track is required, so the diffusionist must
manually trigger the various sound files as indicated in the score. The patch contains
two controls which sequentially trigger the audio files associated with the stereo
channels A and B. The cues for the electronic part are indicated on a single line stave
with a capital letter A or B and the associated file number, as shown in Figure 10.12.
The azimuth control is indicated by a straight line arrow, above the stave indicates a
narrow frontal azimuth (Figure 10.13 (b)), at the stave indicates a wide azimuth
(Figure 10.13 (c)) while below the stave indicates a rear orientated azimuth (Figure

179

10.13 (d)). The example in Figure 10.12 indicates that the azimuth should move from
approx 10-20o to 90o over the course of ten bars.

(a)

(b)

(c)

(d)

Fig. 10.9 Diffusion ranges in String Quartet No. I

Fig. 10.10 String Quartet No. I - diffusion score

The two channels can be spatialized directly from the patch to a multichannel
array using higher order Ambisonics, or alternatively, a stereo feed can be routed
directly from the patch into a mixing desk for manual diffusion to a loudspeaker
orchestra. If the spatialization is to be implemented using Ambisonics, then a MIDI
controller is also required to adjust the distance and two azimuth controls during the
180

performance. This controller (or the computer keyboard) is also used to trigger
playback of the various audio files.

10.3.3 Critical Evaluation

In practice, the overall dynamic of the electronic part was restrained so that it
coloured and shaded the quartet, without dominating, however, the dynamics of the
electronic part should be pushed for dramatic effect at the main crescendo points in
the piece. The electronic part primarily acts to extend and widen the quartet sound,
and so is never entirely spatially separated from the frontal quartet. It is for this
reason that the patch is arranged with stereo channels which are adjusted using the
distance and azimuth controls. Azimuth in this case indicates the stereo spread and so
acts as the primary spatial control for each channel, where as distance acts as the
primary amplitude control. A suggested diffusion strategy is indicated in the
diffusion score shown in Figure 10.14.
At the time of writing, this piece has been performed three times with the
author as diffusionist in each case. The first two performances took place in quite
large and reverberant acoustics (namely SS Michael and Johns Hall, Dublin, Ireland
and Muziekgebouw aan t IJ, Amsterdam, Netherlands) using an eight channel array
and Ambisonics, while the third performance took place in a medium sized theatre
venue (the Beckett Theatre, Dublin, Ireland) using manual diffusion to three pairs of
loudspeakers at varying distances from the audience. Perhaps surprisingly, for this
piece the more diffuse sound of the larger venues and Ambisonics was preferable as
this enhanced the blending of the synthetic and acoustic sounds which was one of the
primary goals of this piece. In the smaller and less reverberant theatre, the
loudspeaker positions (particularly the lateral loudspeakers) became much more
apparent and again Ambisonics was beneficial in overcoming this problem.
Performing this piece also revealed the significant skill and experience required to
successfully and transparently diffuse sounds to a diverse range of loudspeakers.
While faders on a mixing desk provide much more sensitive control than MIDI faders
and ambisonic panning, this physical benefit is more than offset by the skill required
to smoothly move sounds using fader movements alone.

181

10.4 Case Study IV: Auto Harp

Fig. 10.11 A Rosen autoharp

Auto Harp is an eight channel acousmatic work constructed from recordings

of strummed arpeggios and individual plucked strings on the autoharp shown in
Figure 10.5. This particular type of autoharp was manufactured in the former East
Germany in the 1950s and 60s and second-hand instruments are now widely and
cheaply available. East German factories were not allowed to put their own
factory/family name on instruments during this time and so no identifying marks
appear on these instruments apart from the distinctive rose decal. Hence these
instruments have come to be collectively known as Rosen autoharps, from the
German word for roses [White, 2008]. Autoharps are played by strumming with a
pick or finger. Pressing one of a number of levers (six in this case) causes certain
strings to be muted, leaving a chord. This results in an instrument which is very easy
to play, but is also quite limited, as picking and muting individual notes is difficult
due to the tight spacing of adjacent strings. The autoharp has therefore been
predominantly used as an accompaniment instrument, particularly in American folk
music.
The acousmatic piece, Auto Harp, is an attempt to extend the musicality of
this simple instrument beyond its physical limitations. Smalleys ideas of gestural
surrogacy therefore play an important part in this work as at various times the focus
shifts from clearly instrumental gestures (first-order surrogacy) to static

182

environmental textures (second-order and remote surrogacy) created from the

complex, sustained resonance which results from a strummed arpeggio through all
twenty-four undampened strings of the autoharp. Various strummed arpeggios and
picked sequences were recorded using both monophonic, stereo and Soundfield
microphone arrangements, and these arpeggios form the basic building blocks of the
entire work. These clearly instrumental passages, presented in static spatial locations,
represent the first order surrogacy material, which is then abstracted to second-order
and remote surrogacy using a variety of techniques. Each individual string was also
recorded in isolation, and these samples are used with various algorithmic techniques
to generate new sequences. Although these algorithmic passages are still clearly
instrumental in origin, the rapid spatial distribution and complicated picked sequences
produced are somewhat removed from the basic first order gestures produced initially
from static spatial locations. More remote levels of surrogacy are attained through the
use of granulation time-stretching techniques which emphasize and sustain the
complex resonance of the vibrating strings. The different material used in this piece is
arranged in terms of first-order, second-order, and remote surrogacy in Figure 10.6.
The only material which is entirely synthetic, and hence displays remote surrogacy, is
the layered sine waves used in the two Franssen sections that open and close the
piece. The title of this section comes from the psychoacoustical illusion, the Franssen
Effect, which is used throughout the opening section of this work.
FIRST ORDER

SECOND ORDER

REMOTE SURROGACY

Strummed arpeggios
mono /stereo
Picked Theme
Mono

Franssen Intro/Outro
eight channel
Time-stretched Arpeggio
eight channel

Programmed Theme
Stereo

Triad Drone
eight channel

Programmed Sequences
stereo / eight channel
Strummed arpeggios
b-format
Fig. 10.12 Auto Harp source material, grouped according to order of surrogacy

183

The Franssen effect is an auditory illusion which produces large and

unpredictable localization errors in human listeners [Franssen, 1962]. This illusion
was first reported by N. V. Franssen in his 1962 thesis on stereophonic localization.
In the initial experiment, two loudspeakers were arranged symmetrically in front of a
single listener. A sine tone with a sharp onset transient is routed to one loudspeaker
and immediately begins to exponentially decay over a fixed time period, t, while
concurrently, the same signal exponentially rises in the other loudspeaker, also over
the same period, t, as shown in Figure 10.7. Franssen found that listeners consistently
localized the source signal to the initial loudspeaker, even after the signal had entirely
moved to the other loudspeaker, and this illusion has since been referred to as the
Franssen effect. This illusion was investigated further by Hartmann and Rakerd in
1989 in a series of detailed tests which revealed the nature and limitations of this
unusual effect [Hartmann et al, 1989]. They discovered that the Franssen effect is
entirely based on the inability of a listener to accurately localize a steady-state sine
tone in a reverberant space. The source is first perceived at the location of the initial
onset transient, and as the following steady-state sine wave provides no further
reliable localization cues, the perceived source location remains unchanged, even after
the real physical location of the source has moved to the other loudspeaker.
Hartmanns experiments revealed that the Franssen effect is highly dependent on
various factors such as the source signal, the transition time, t, and the listening
environment [Hartmann et al, 1989]. Firstly, Hartmann reported that the illusion fails
entirely for broadband noise signals, as these spectrally rich signals produce viable
localization cues even in the absence of a sharp onset transient. Secondly, the illusion
fails entirely under anechoic conditions as in this case the steady-state localizations
are sufficient for accurate localization.

Fig. 10.13 Loudspeaker envelopes in Auto Harp Franssen section

184

String No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

Harmonic
1
1
3
7
2
9
5
11
3
7
2
9
5
21
3
7
15
17
19
5
23
3
7
2

Freq
261. 6
261. 6
392. 4
457.8
523. 3
294.3
327.0
359.7
392. 4
457.8
523. 3
294.3
327.0
343. 4
392. 4
457.8
490. 5
278.0
310.7
327.0
376.1
392. 4
457.8
523. 3

Table 10.1 Auto Harp tuning scheme

10.4.1 Analysis
The introductory section of Auto Harp uses the loudspeaker amplitude
envelope shown in Figure 10.7 in conjunction with twenty-four sine tones which
correspond to the twenty-four strings of the autoharp. Each string is tuned to a
different harmonic (octave adjusted) of the lowest note (C 261.6Hz) as shown in
Table 8.1. Each harmonic is adjusted so it falls within a two octave range. This
spectral tuning produces a complex resonance with very little beating due to the close
harmonic relationship between each string. Each sine tone onset is presented in
sequence from the four loudspeakers to the right, moving from front to back, and is
shifted to the matching loudspeaker on the left using a transition time t of 60ms.
After all twenty-four sine tones have been introduced, reversed recordings of each
individual string are gradually introduced at the same real spatial locations as the
associated sine tone, and the entire sequence builds to a sharp crescendo. The overall
effect of this sequence is to delay and disguise the movement of the layered sine tones
from right to left, as the initial onsets are clearly localized to the right, while the final

185

reversed autoharps recording clearly crescendo on the left side. Localization during
the middle steady-state section will therefore vary depending on the acoustical
environment, the individual listener and the associated strength of the Franssen effect,
introducing an element of variability in a composition for fixed media.
The overall structure of Auto Harp is marked on the waveform of a
monophonic reduction of the eight-channel work in Figure 10.8. The crescendo
which marks the end of the initial Franssen section is articulated with a spatial sweep
from front to back on the left side of the array. This is immediately responded to by a
sweep through each individual note in a corresponding spatial motion from back right
to front right. The entire Play section which follows uses this type of spatial dialogue
between static arpeggio recordings and dynamic, algorithmic sequences. In each case,
a recorded arpeggio (mono, stereo or b-format) produces an algorithmic response,
which is constructed from the individual string recordings, pitch shifted up an octave.
The spectromorphological profile of each recorded arpeggio was carefully applied to
the algorithmic response by matching the duration of each sequence. Karlheinz Essls
Real-Time Composition Library for Max/MSP was used to create these algorithmic
responses [Essl, 1995]. The series object was used to create a random sweep through
all twenty-four notes, while the sneak object randomly selects notes but only chooses
notes adjacent to the currently selected note. This object was used to create the
distinctive and rapid sequences of notes which are prominent in this section.
The next section is introduced by a mono recording of the plucked sequence
which functions as the main melodic theme. The theme is presented four times in this
section, initially as the performed sequence in the original mono recording. This is
followed by a sequenced recreation of the theme using the individually recorded
strings which are spread across the front pair of loudspeakers. The final two iterations
of the theme use both the original progression, and various sequenced progressions
which are distributed across each of the four loudspeaker pairs. In this case, the
sequenced progressions are also transposed to harmonize with the initial progression.
A granulated time-stretched version of this progression is also introduced in this
section, shifting the focus away from the highly gestural material which has
dominated up to this point. The spatial distribution and dynamic of this drone alters
in response to the instrumental gestures and at times the drone itself displays strong
spatial gestures. This is evident in the crescendo that ends this section, and also at

186

348, when the diffuse drones builds and then suddenly collapses to low frequency
drone statically positioned in the front loudspeaker pair.

Fig. 10.14 Monophonic reduction of Auto Harp indicating section durations

The Triad Drone section is the first to move away completely from the highly
gestural arpeggios and sequences which dominated the opening half of this piece.
The first harmonies which are not contained within the tuning system shown earlier
are also introduced in this section. Multiple stereo channels of a spectrally rich drone
were constructed from a recorded arpeggio which was time-stretched using the
granule opcode in the Csound processing language (Csound is discussed in more
detail in chapter 12.3). Two of these decorrelated stereo channels quickly begin a
slow upward glissando before eventually settling at two pitches to form the triad
drone which gives this section its name. The two pitch-shifted drones were tuned by
ear to produce a relatively consonant harmony. The tuning of these two new drones
corresponds approximately to the fifth and ninth harmonics of the original root note.
The penultimate section is, as its name suggests, primarily constructed from a
time-stretched arpeggio constructed using a granulation algorithm in Max/MSP. The
resulting drone quickly separates into a dense cloud of clicks created from the initial
plucking action and a thick, mid-frequency drone created from the resonance of the
instrument. This spatially diffuse sequence decays slowly and then builds again to a
final crescendo which is punctuated by occasional unprocessed arpeggios. The piece
ends with a reversal of the initial Franssen section, as the drone fades into associated
sine tones which collapse spatially in the same way as they appeared.

187

10.4.2 Critical Evaluation

The use of the Franssen effect in the opening section of this work has proven
to be highly effective, although the precise effect perceived is highly dependent on
both the acoustic of the performance space and the position of the listener within the
array. The onset of each tone is generally well localized to the right of the array, and
the perceived source direction then gradually shifts from right to left before collapsing
to a spatial arpeggio down the left hand side of the array. However, due to the
inherent difficulty in localizing sine tones in a reverberant environment, the speed at
which the sine tones move from right to left in this section is quite unpredictable,
which introduces a nice element of variability into this fixed media piece.
The autoharp is in many respects an ideal source for a work of spatial music as
each plucked note contains a sharp, percussive transient that is highly localizable,
followed by a more diffuse, sustained pitch. In this way, each note can play the role of
either an instrumental gesture (for example, the programmed spatial arpeggios) or
environmental texture (the time-stretched, sustained drones). Furthermore, the
instrumental gestures can be divided into first order surrogacy, recorded instrumental
gestures, or second order surrogacy, programmed gestures, and moving between these
different levels of surrogacy provides much of the impetus for this piece. The inherent
and quite dramatic nature of the basic gesture, i.e. the plucked arpeggio, is also clearly
related to the programmed arpeggios, while the sustained pitch of the initial gesture
also blends very well with the time-stretched, environmental like drones. Finally, the
use of both Ambisonics and amplitude panning is a highly effective means of spatially
segregating the different layers of material. This is particularly true of the closemiked, soundfield microphone recordings which are spatially quite distinct from the
surrounding and immersive drone material.

188

11 Spatial Music Instruments

The preceding chapter explored how different composers have attempted to
connect the musical gestures of an instrumental performer to the spatial gestures of
the electronic part. While some composers have turned to synchronised tape parts or
real-time electronic processing to achieve this connection, others have attempted to
design new instruments which incorporate the spatialization processes. While various
new musical interfaces have been developed which feature some form of spatial
control, the idiosyncratic nature of these devices mean they are unlikely to be widely
adopted. Augmented instruments, i.e. traditional musical instruments featuring
additional hardware or sensors, represent a possible solution to this problem, as they
can potentially combine existing and sophisticated instrumental practice with spatial
or timbral processing algorithms. The mapping of musical gestures to electronic
processes is a critical issue in the design of any augmented instrument. The composer
David Wessel suggests that musical control intimacy and virtuosity require both
spatial and temporal precision in the sensing of gestures (control intimacy refers to a
tight connection between a body movement and change in an auditory feature)
[Wessel, 2006]. One form of augmentation specific to stringed instruments is the use
of polyphonic pickups which produce a separate audio signal for each string. The
discrete multi-channel output of these instruments would seem to be very suitable for
spatialization to a multi-channel loudspeaker array. By linking the spatial location to
the choice of string, the spatialization process can be synchronized to the physical
performance of the instrument. In addition, spatialization algorithms and other
processes can be applied to each individual string as required. This signal processing
approach to instrument augmentation has the advantage that the performer does not
need to learn any new gestures or instrumental techniques. In addition, the necessary
hardware has become widely available, particularly for electric guitar, and can often
be retrofitted non-destructively to an existing instrument. The electronic violin
developed by Max Mathews [Boulanger, 1986] and the Hyperinstruments developed
by Todd Machover [Machover, 1992] are two examples of augmented stringed
instruments which incorporate polyphonic pickup technology. The polyphonic output
of these instruments was used extensively by the composers who wrote for them. As
many manufacturers now produce electric instruments with polyphonic outputs, this

189

approach represents a viable and generalised solution to stringed instrument

augmentation that is not tied to specific hardware.

11.1 Augmented Strings

The electronic violin developed by Max Mathews in 1986 (see Figure 11.1)
was an attempt to use the considerable dynamic, expressive timbral range and highly
evolved instrumental technique of the violin as the musical source, modifier, and
controller of various real-time performance networks [Boulanger, 1986]. The
electronic violin used four contact microphones inserted into the bridge to pickup the
sound of the strings. Mathews noticed that when the outputs from the four strings
were combined and passed through a single amplifier and speaker, nonlinearities
resulted in unpleasant combination tones. To eliminate this problem, Matthews used
a separate amplifier and loudspeaker for each string. When the composer Richard
Boulanger later came to compose a piece for the instrument, he commented that the
discrete four-channel output of the instrument significantly directed the composition
in a number of ways [Boulanger, 1986], specifically toward a composed spatial
component. In the resulting piece Three Chapters from the book of Dreams
Boulanger supports and contrasts the linear counterpoint and compound melody with
a concurrent spatial counterpoint. He comments By assigning the output of each
string to a separate speaker, the audience is given the unique sense that they are seated
within the violin body. This spatial component is based on the inherent design of the
instrument and its antiphonal treatment is at the same time quite old and quite new.
[Boulanger, 1986]
The Hyperinstrument group at MIT Media Lab have been researching and
developing augmented instruments since the late eighties [Machover, 1992]. The first
of these, the Hypercello (shown in Figure 11.2), was completed in 1991 and combined
an electroacoustic cello with additional sensors to provide data on various
performance parameters such as bow position, placement and pressure, string and
finger position and pitch tracking. The composer Tod Machover worked with the
renowned cellist Yo-Yo Ma to create Begin Again Again, an interactive composition
in which different playing techniques such as tremolo, bow bounce, pizzicato and
legato are mapped to various electronic processes including spatial movement.

190

Fig. 11.1 The Max Mathews augmented violin

Fig. 11.2 The Hypercello system

191

11.2 The Hexaphonic Guitar

Polyphonic pickups (also called divided or split pickups) have been widely
used over the past three decades to detect and convert the pitch coming from
individual guitar strings into MIDI messages. The majority of these systems however
do not provide split signals for external processing, preferring instead to multiplex the
signals into a single cable which can then be easily connected to a MIDI converter.
This emphasis on MIDI capability is changing however and some dedicated systems
such as the Gibson HD. 6X-Pro have been developed specifically for polyphonic
processing. The initial prototyping of the instrument was carried out by Adrian Freed
and the Guitar Innovation Group at UC Berkeleys Centre for New Music and Audio
Technologies (CNMAT) [Yeung, 2004]. Further research has since been carried out
at CNMAT on polyphonic guitar effects based on vocal-tract modeling, frequency
localized distortion and coordinated equalization [Jehan et al, 1999]. The co-director
of CNMAT, David Wessel has also conducted research on augmented instrument
design and interactive computer music with a particular focus on the live performance
of improvised computer music. Wessel had used polyphonic pickups in performances
such as Situated Trio, an interactive live performance for a hexaphonic guitarist and
two computer musicians with expressive controllers [Wessel et al, 2002]. In this
piece, the polyphonic guitar signal is processed by the two computer musicians using
various algorithms such as granulation, looping, cross-synthesis and spatialization.
The polyphonic guitar signal is also converted to MIDI to provide a high level
discrete event representation of the guitarists performance for triggering automated
computer based processes.
The adaptation of existing MIDI guitar systems is an alternative and cost
effective method for deriving a polyphonic signal from an electric guitar. Popular
MIDI guitar systems by Roland, RMC and AXON use a 13-pin connector which
carries the six individual signals from the hexaphonic transducer, a mono audio feed
from the guitars normal pickups, and some controls specific to the Roland system.
The wiring diagram shown in Figure 11.3 indicates the individual string signals from
the 13-pin connector which can be accessed simply by wiring the appropriate pins to
several 1/4 inch jack connectors. A schematic for such a breakout box is shown in
Figure 11.4 [Berg, 2007].

192

Fig. 11.3 Roland 13-pin wiring diagram

Fig. 11.4 13-pin breakout box schematic

11.3 Case Study V: Etude No. 2 for Hexaphonic Guitar

Writing for a hexaphonic guitar in which each string is individually spatialized
requires the composer to carefully consider which strings will be used to produce the
chosen pitches. In this scenario, the routing and particular spatialization method used
will inform the composition in a very real and tangible way. In addition, the
relationship between the visible and spatially static guitarist and the spatially
distributed sounds must be carefully considered. In the beginning of this piece, the
spatial distribution of the guitar strings is disguised and this helps to forge a
connection between the visible performer and the diffused sound. Once this
connection is made, each distinct spatial location is then gradually introduced and

193

layered to eventually create a large, enveloping drone which gradually overcomes the
live performer and closes the piece.

11.3.1 Analysis
In this etude for hexaphonic guitar the strings are consecutively routed in pairs
through three instances of Ambience, the freeware reverb VST plugin by
Smartelectronix (see Figure 11.5). Various plucked harmonics and intervals are
sustained using the reverb hold function which is triggered with a MIDI foot
controller. After each group of three harmonics/intervals have been layered, the
combined drone is then dynamically routed to a spatialization algorithm (2nd order
Ambisonics using the ICST externals for Max/MSP) which chaotically pans the drone
around the entire array. This entire process is then repeated three times with
increasing intensity, culminating in a very loud drone which dominates the
instrumental part. This process is highlighted during the final section as the performer
remains motionless while the vast and spatially dynamic drone continues, culminating
in a loud crescendo.

Fig. 11.5 Etude No. 3 tuning, spatialization and interval sequence

194

11.3.2 Critical Evaluation

The Ambience VST plugin produces a highly synthetic reverberation that is
much less realistic than many other reverberation algorithms, however, artificial
reverberation is used in this piece more as an effect than as a realistic simulation of an
acoustic space. This becomes particularly obvious when the sustain function of this
reverb effect is repeatedly engaged and the various drones are reproduced using each
of the three pairs of loudspeakers.
The percussive sounds at the start of the piece are produced by striking the six,
muted strings with the heel of the hand behind the bridge and the resulting sound is
therefore distributed relatively equally among the six loudspeakers. The staccato,
rhythmical nature of the material and the enveloping, nondirectional nature of the
spatial distribution helps to create an initial connection between the instrumental
player and the spatialized audio. The following introduction of single, sustained notes
is perhaps the first time that the spatial distribution of the guitar strings is made clear
and this gradual introduction of the spatialization scheme helps to reduce the contrast
between the location of the performer and the resulting sounds. As this process is
repeated in different ways and with an increasing dynamic, the layered drones
gradually build to a vast, spatially dynamic noise drone which becomes increasingly
remote from the actions of the performer. In the very final section, this separation is
completed and the performer rests, motionless while the spatial drone plays for
approximately 20-30 seconds, before triggering the end of the piece at whatever point
they choose.

195

12 Behavioural Space
Modern computer technology has now advanced to such a degree that an
ordinary laptop computer and suitable audio interface is capable of processing and
spatializing numerous channels of audio in real-time. These developments have
allowed composers to utilise more advanced algorithms to control the spatialization of
sounds. Flocking algorithms have proved to be particularly popular in this regard and
conceptually, the notion of flocking, or swarming, contains obvious connotations in
terms of spatial motion. Flocking algorithms such as Craig Reynolds Boids
[Reynolds, 1987], model the overall movement of groups of autonomous agents, such
as a flock of birds or a shoal of fish, and have been widely used in computer graphics
and animation. Recently, these algorithms have also been adapted for musical
applications including spatialization.

12.1 The BOIDS Flocking Algorithm

In his investigation into the motion of flocks of birds, Reynolds discovered
that the aggregate motion of the flock as a whole can be simulated through the
interaction of autonomous individual agents following a number of simple rules
[Reynolds, 1987]. He identified the three primary factors of separation, alignment
and cohesion which govern the aggregate motion of the flock. These factors are
represented as a set of behavioural rules for each individual agent, whereby each
agent avoids colliding with nearby agents, attempts to match the velocity and heading
of nearby agents, and attempts to move toward the centre of the flock.
The 3D implementation of Boids for Max/MSP/Jitter [Singer, 2008] (shown in
Figure 12.1) generates a visual representation of the flock and outputs the current and
previous position of each individual agent at a specified period as Cartesian
coordinates. The behaviour of the flock is modified using controls such as speed
range, strength of centring instinct, neighbour speed matching, avoidance and
preferred distance and the strength of attraction to a specified attract point, which
allows the user to direct the overall movement. Other implementations of the
Reynolds flocking algorithm include a predator which the flock will fly around and
avoid, as shown in Figure 12.2 [Martin, 2005].

196

Fig. 12.1 Eric Singers Boids for Max/MSP/Jitter

Fig. 12.2 Illustration of a flock avoiding a predator

Flocking algorithms such as Craig Reynolds Boids [Reynolds] have been

adapted for various music applications such as interactive audio installations [Unemi
et al, 2005], real-time music improvisation by a computer [Blackwell et al, 2004] and
real-time algorithmic composition [Uozumi, 2007]. Flocking algorithms have also
been used in conjunction with granular synthesis and both techniques would seem to

197

share many characteristics. Both use large numbers of individual agents or grains,
both produce macro-level complexity through relatively simple micro-level structures,
and both are indirectly controlled through various micro and macro level parameters.
Flocking algorithms would therefore appear to be highly suitable as a control
mechanism for granular synthesis and also suggests an obvious spatialization scheme
where each grain is positioned in space according to the location of a single agent in
the algorithmically generated flock.

12.2 Case Study VI: Flock

The acousmatic composition Flock was constructed in the Max/MSP environment
using the following processes;

Eric Singers implementation of Craig Reynolds flocking algorithm, Boids

Nobuyasu Sakonda's granulation algorithm, granular 2.5

ICST Higher Order Ambisonics Externals

IRCAMs Le Spatialisateur

These processes were combined so that each grain corresponds to a single element in
each flocking algorithm. The flock of individual grainboids are then spatialized using
one of the two spatialisation algorithms.

Fig. 12.3 Max/MSP Patch used to construct the composition Flock

198

The source material used in this piece consists of noise/click samples and prerecorded tremoloed notes on viola and cello. The overall movement of each flock can
be controlled with a user controlled attract point. While the individual spatial
movement of each grainboid is controlled by the flocking algorithm, the overall
movement of each flock can be adjusted by the user in a number of ways. Firstly a
bounding box can be created which will contain the flock within a certain specified
area. Alternatively, the strength of the flocks attraction to the user-specified attract
point can be used to direct the flock to follow a particular point in space. In this
piece, noise/click samples and pre-recorded tremoloed notes on viola and cello are
associated with a particular flock of eight grainboids using the Max/MSP pitch shown
at the start of this section in Figure 12.3. Each flock follows a slightly different
overall trajectory, using the associated attract point with a medium attraction strength
setting so that the specified path is only loosely followed.

12.2.1 Analysis
The nature of the source signal has a large bearing on the effectiveness of the
perceived spatialization (see Chapter two) and the source material for this piece was
chosen with that in mind. Localization accuracy has been found to improve when the
source signal contains plenty of onset transients, and when broadband signals are
used. The initial testing of this system was carried out using a noise/click sample for
this reason and this material was then also used in the final composition. Tremoloed
string samples were chosen as they provide pitched material which also contains lots
of irregular transient information due to the rapid bowing action. In addition, the fast
irregular bowing motion recorded in the original tremolo string samples is well
matched to the granulation process.
Two spatialisation methods were used in order to investigate the perceptual
difference between these two techniques, namely amplitude panning and Higher
Order Ambisonics. A subtle yet distinct difference was perceived between these two
spatialization schemes, however, this difference is certainly also influenced by the
artificial reverberation and early reflections added by IRCAMs Le Spatialisateur
program. This perceptual difference, although relatively subtle, does help to spatially
distinguish the various flocks and heightens the spatial micro-complexity of the piece.
The Doppler Effect was utilized throughout the work, perhaps surprisingly

199

considering the presence of pitched material. However, the view was taken that the
pitch variations introduced by the spatial movement and resulting Doppler shift was
aesthetically pleasing and fitted the micro-variations in pitch resulting from the rapid
bowing action of the original samples, as well as supporting the perceived movement
of each flock.

12.2.2 Critical Evaluation

The flocking algorithm spatialization scheme used in this piece produced a
complex spatial motion which is rich in micro-level detail. The precise motion of
each individual grainboid is not clearly perceptible, however the overall spatial
movement is highly coherent and very evocative. The complex spatial motion
produced by the flocking algorithm also has a significant effect on other parameters
such as timbre and particularly, dynamics, as this parameter is closely linked to the
distance of each flock from the centre point. The precise harmonies are therefore
somewhat indeterminate, as the flocking algorithms control the precise timing of the
movement of each flock through the centre point. Overall, the piece follows a loose
ABA structure which is delineated by two low register pedal tones underneath the
changing harmonies produced by the other pitched flocks. The main harmonic change
at the start of the B section occurs at the approximate halfway point (515) while the
return to the final A section occurs gradually as the new pedal tone flock is gradually
replaced by the original.
The loose, almost drone-like aesthetic utilized in this composition was greatly
influenced by the real-time operation of the synthesis and spatialization methods.
While real-time control of these processes is highly suitable for performance or
experimentation, it is perhaps less suitable for the creation of longer works due to the
very large number of parameters involved. In addition, the real-time operation of this
system placed significant limitations on the amount of material which can be
generated concurrently. It was found that a maximum of thirty-two grainboids in four
flocks (as shown in Figure 12.3) could be realised using Ambisonics encoding, while
IRCAMs Le Spatialisateur processor could at most realise a single flock of only eight
grainboids (based on a 2GHz Pentium Processor with 2GHz of RAM). The finished
work therefore had to be constructed from individual layers of pre-composed material
which were then later sequenced together. The Boids algorithm is not itself

200

computationally expensive and flocks of many hundreds of agents can be generated,

controlled and visualized in real-time. However, when combined with advanced
audio synthesis and spatialization techniques, computational limitations restrict what
can be achieved in real-time. Some of these issues have been investigated by the
Interactive Swarm Orchestra (ISO) project at the Institute for Computer Music and
Sound Technology (ICST) in Zurich (the ICST has also released a popular set of
Higher Order Ambisonics externals for Max/MSP). The ISO project similarly
employs flocking algorithms to control real-time computer sound synthesis and 3D
sound positioning, using a collection of generic open-source C++ libraries [Bisig et al,
2008]. Some encouraging preliminary work has been carried out at the ICST on
Ambisonics Equivalent Panning (AEP), a computationally efficient implementation of
Ambisonics suitable for large numbers of sources [Neukon et al, 2008]. However, the
addition of distance effects in the form of artificial reverberation is still too
computationally demanding to achieve in real-time for a large number of sources. In
addition, this real-time, performance-oriented approach also inevitably restricts the
complexity and sophistication of compositions realized with these systems. While the
various real-time implementations of this technique are highly suitable for
performance applications, an offline approach is in many ways much more suitable
for the creation of longer works and provides a much greater degree of control over
the synthesized material.

12.3 Offline Spatialization with the Boids Algorithm

Csound [Boulanger, 2000] is a well established synthesis language, widely
used in computer music composition, which supports a variety of granular synthesis
techniques [Roads, 2004]. In Csound, a synthesis method, defined in an instrument in
the orchestra file, produces macro-level sound events according to note statements in
an associated score file. In the case of granular synthesis, each note statement
corresponds to a micro-level event (although higher level granular synthesis opcodes
are also available). This approach trades off instrument complexity for score
complexity and provides a great deal of control over the synthesis, however, it also
requires a very large score file. Granular event generator programs such as CMask
[Bartetzki, 2008] can be used to create these micro-level score files using various
macro-level probability and stochastic functions. For this piece, a number of Csound

201

granular event generators and instruments, entitled AmbiBoid were developed by the
author to implement various granular synthesis techniques such as granulation,
glisson synthesis and grainlet additive synthesis. Each grain is then spatialized using
the Boids algorithm and second order Ambisonics with distance modelling. 2
The score file generators read spatial co-ordinates from text files generated by
a Max/MSP patch containing Singers Boids object. Due consideration was given to
coding the Boids algorithm within either Csound or C. However, generating the
spatial trajectories in real-time has a number of benefits, namely;

it provides a clear visual representation of the spatial movement during the

composition stage.

it provides immediate feedback on the effect of changing flocking parameters.

it reduces Csound rendering time.

In addition, this approach allows the same recorded trajectories to be applied to

different synthesis methods, at different speeds, and with different dynamically
changing parameters, so they can be used to generate musical building blocks, at a
macro level. The Max/MSP patch converts the Cartesian coordinate pairs (previous
and current value) of each agent to azimuth, elevation and distance values which are
then written to three ASCII text files. The patch currently records a maximum of 128
agents but its modular design can be readily extended to larger numbers. The Boids
object generates new values at a regular user-defined interval. In real-time
applications this interval is often set equal to the grain duration so that the Boids and
grain parameters are only updated when the grain envelope is at a zero point.
However, with this approach a single grain duration must be used for all grain
streams, which rules out many of the more sophisticated granular synthesis
algorithms. The Ambiboid score generators therefore map each grain stream to an
individual agent in the flock and linear interpolation to ensure that streams with
different grain durations read through the spatial data at the same rate, preserving the
flocking behaviour. The overall rate at which spatial coordinates are read is set by the
minimum grain duration. Therefore, if the overall tempo of the spatial movement
generated in Max/MSP is to be preserved, the boids objects clock period should be
set equal to the intended minimum grain duration.
2

The Ambiboib source code, along with various audio examples, can be downloaded at
www.endabates.net/ambiboid.

202

Once the desired spatial trajectories have been performed and recorded in
Max/MSP, the AmbiBoid utilities can be used to generate a Csound score file. Each
utility/instrument pair implements a specific form of granular synthesis with shared
parameters such as overall duration, number of grain streams, global reverb, etc. The
density is controlled via the inter-grain spacing which sets the amount of silence
between each consecutive grain as a multiple of that streams grain duration. The
density can be either fixed or dynamically varying between specified values at five
hit-points equally distributed across the total duration. The grain duration can also be
fixed or dynamically varying in a similar fashion. Various windowing functions such
as Gaussian, Hamming and Hanning windows can also be chosen. Three different
granular synthesis techniques were implemented, namely;

12.3.1 Granulation
The AmbiBoidgranule score generator and instrument performs straight
forward granulation of a monophonic sample. The grain durations can either be a
single fixed value for all streams, a single dynamically varying value for all streams,
or a randomly chosen duration for each stream varying between a given maximum
and minimum value. The playback position within the file for each grain is randomly
chosen and the playback rate of the audio sample is defined in the score file.

12.3.2 Grainlet Additive Synthesis

Grainlet synthesis is a combination of granular synthesis and wavelet synthesis
developed by Curtis Roads [Roads, 2004]. In this synthesis method, the duration of
each grain is related to the frequency of the source signal within that grain. In this
additive synthesis version, each grain stream corresponds to a single partial whose
grain duration is related to the frequency of that partial. The user specifies the
fundamental frequency and the number of grain streams/partials. The grain duration
consists of one hundred cycles of the signal, which results in a grain duration of one
second for the fundamental, 31ms for the 32nd harmonic, 15ms for the 64th harmonic,
etc. The source signal consists of either a sine wave or a set of harmonically related
sine tones.

203

12.3.3 Glisson Synthesis

In glisson synthesis, each grain, or glisson, has an independent frequency
trajectory, an ascending or descending glissando [Roads, 2004]. The perceived
relationship between spectral space and physical space from such terms as low and
high pitches, or ascending or descending melodies. The mapping of spatial elevation
to source frequency would therefore seem to be a natural and intuitive choice for this
combination of glisson synthesis and the three dimensional spatial movement of the
flocking algorithm. In this implementation, the source signal frequency can be
mapped to spatial elevation in two ways. The maximum and minimum elevation
angles (90o) can be linearly mapped to a specified frequency range, or alternatively
mapped as octaves so that each glisson contains an octave glissando whose direction
depends on the spatial movement. The usual density and grain durations controls are
also available.
Each instrument uses standard Csound opcodes to generate the window
function and source signal but most of the instrument code is concerned with the
spatialization algorithm. The current and previous spatial coordinates are changed
into a dynamic k-rate variable which is used to spatialize each individual grain with
distance effects, early reflections and global reverberation. These distance effects are
modelled using Chownings algorithm for the simulation of moving sources,
discussed previously in Chapter Three. In this model, the intensity of the direct sound
reaching the listener falls off more sharply with distance than the reverberant sound.
The loss of low intensity frequency components of a sound with increasing distance
from the listener is modelled by mapping distance to the cut-off frequency of a low
pass filter. The Doppler Effect is implemented using Csound code developed by
Christopher Dobrian [Dobrian, 2008]. Each grain is sent to an interpolating delay line
whose delay time is mapped to the distance of the source. The changing distance and
delay time produces a pitch shift in the audio signal which mimics the Doppler Effect.
Early reflections and global reverberation are applied to each grain using Csound code
developed by Jan Jacob Hofmann [Hofmann, 2008] for an ambisonic spatialization
instrument. Specular and diffuse reflections are generated according to the distance
from the source to the listener, and to the specified dimensions of the virtual space. A
global reverb is implemented using an eight delay line feedback delay network reverb
with a feedback matrix based upon the physical modelling scattering junction of eight

204

lossless waveguides of equal characteristic impedance [Hofmann, 2008; Smith, 1985].

Finally, all the audio signals are encoded using the Furse-Malham-Set (FMH) of
encoding equations [Malham, 1999] for Second Order Ambisonics. The final output
can be decoded for various loudspeaker configurations such as stereo, quadraphonic,
octagon or a cube, using various ambisonic decoding schemes.

12.4 Case Study VII - Rise

An 8-channel composition, Rise, was realized using the three synthesis
methods discussed in the preceding section. A number of spatial trajectories were
recorded in Max/MSP and applied to the different synthesis techniques in various
ways. These spatially dynamic textures were then contrasted with static decorrelated,
monophonic signals played directly through the eight-channel array. Glisson
synthesis is used to mimic birdsong and, indeed, the entire piece can be thought of as
a sort of virtual soundscape with echoes of natural sounds and spaces.

12.4.1 Analysis
Two distinct trajectories were used throughout this piece;

a fast spiralling trajectory with lots of vertical movement around the centre
point

a slow front-back-front trajectory passing through the centre point

The first trajectory was created specifically for use with glisson synthesis as it
contains significant vertical movement. Elevation values are mapped to the frequency
and direction of the glissando with the glissando start frequency restricted to linear
harmonic intervals, producing a more melodic texture. This can be clearly heard in
isolation at the very beginning of the piece. The second section begins with
synthesized tones created using the grainlet additive synthesis instrument and the slow
front-back trajectory. These synthesized tones provide the underlying harmony
which is derived from a two chord guitar progression (a granulated version of which
can be faintly heard later on). These strongly tonal textures are contrasted with
wideband noise textures which also follow the same front-back trajectory. These
rolling waves of noise were created using the granulation instrument and a heavily
distorted version of the original guitar progression. The underlying tonality of these
noise bursts becomes more apparent as the piece progresses and the grain duration

205

increases. These contrasting textures are accompanied by reiterations of the opening

glisson movement. As the glisson texture becomes increasingly dense and unifies as a
high frequency tone it leads into a noisy texture presented as eight decorrelated
signals played directly from the loudspeaker array. These static sources provide a
sense of spatial perspective as their static spatial location contrast with the
approaching and receding tonal and noise signals. The piece ends with a variation of
the initial glisson texture, this time decreasing in elevation and density.

12.4.2 Critical Evaluation

While each specific granulation algorithm is quite distinct in terms of its sonic
output, the spatialization algorithm results in some behaviour which is common to all
of the synthesis methods. Dynamic variation in the density of the granular texture has
a significant influence on the perceived timbre. At high densities the flocking
behaviour manifests largely as variations in the spatial, spectral and amplitude
characteristics of the overall texture, while at lower densities the individual voices and
their spatial motion become more readily apparent. In other words, at higher densities
a more unified tone is perceived while at lower densities, the signal tends to split into
multiple distinct voices. The spatial distribution of the flock as a whole also has a
significant influence on the perceived timbre. A flock with a strong centring instinct
and a low minimum separation distance will be tightly bunched together spatially and
when this is combined with a high density it results in more unified tone. The
distance of the flock from the centre point is another highly important parameter with
even very dense and tightly gathered flocks separating into individual voices as they
pass through the central listening position.

206

13 Summary & Conclusion

The trouble with space is that its the whole piece. Its the sounds and everything.
The impressions of space are created through the types of sounds and their temporal
experience. Space is the whole thing. It is not usually something that people perceive
as separate from the sounds themselves, although the composer might consider space
separately. For the listener, theyre all moulded into one. Thats why we end up
talking about the piece as a whole, because the whole is the space or spaces of the
piece
Denis Smalley [Smalley et al, 2000].

The above quote illustrates the vast range of issues related to the production,
composition and performance of spatial music which this thesis has attempted to
address. A composer of spatial music must consider specific aspects of the work such
as the types of sounds and the particular spatialization process involved, while also
remembering that many of these specific aspects may not be perceived directly by the
listener. As Smalley states the piece as a whole must be looked at, because the
whole is the space or spaces of the piece. The success of any work of spatial music
can therefore only be considered in terms of the overall compositional strategy which
describes the relationship between space and every other musical parameter.
This thesis has attempted to examine all of these issues in terms of auditory
perception as ultimately, the success of any work of spatial music depends upon the
perception of the listener. A number of different spatialization schemes were
analysed and a meta-analysis of the results of a large number of other listening tests
and simulations were presented. The results indicate that each spatialization scheme
has particular strengths and weaknesses, with the result that the most applicable
technique in any situation is dependent on the particular spatial effect required.

13.1 Spatialization Results & Recommendations

The review of various spatialization schemes presented earlier in this thesis
indicated the particular capabilities of each technique and that the most suitable
spatialization scheme in any situation is entirely dependent on the particular context.
What may be the most suitable approach in one case may not be the best approach in
another, and this applies to the different types of sounds and spatial movements within
a single composition as much as it applies to different musical aesthetics and
performance practices.

207

The practice of stereo diffusion to a loudspeaker orchestra is clearly quite

distinct from other approaches to spatialization. The real-time and manual nature of
the diffusion process indicates the intuitive nature of this approach which is as much
concerned with adapting the work to the acoustic of the particular performance space
as it is with generating spatial gestures or locations. Technically speaking, the
approach in this case is therefore relatively clear and well understood, and the
technical and artistic practice of diffusion has already been covered in detail by
composers and writers such as Jonty Harrison [Harrison, 1999], Denis Smalley
[Smalley et al, 2000] and others [MacDonald, 1995].
The relatively new technique of wavefield synthesis would seem to also be
quite distinct from multichannel techniques such as Ambisonics or VBAP, or the
practice of diffusion. The spatial aliasing frequency is crucial in any WFS system and
the results of a number of tests [Start, 1997; Huber, 2002; Wittek, 2007] illustrate the
degradation in localization accuracy and timbre that occurs when this parameter is too
low. However, due to the direct relationship between the spatial aliasing frequency
and the physical gap between each loudspeaker, a very large number of loudspeakers
will be required to cover a listening area suitable for a musical performance. One of
the most often cited innovations of WFS is the ability to position virtual sources both
behind and in front of the loudspeaker array. Numerous simulations have illustrated
the correct wavefront curvature produced by such a source but the results of listening
tests [Kerber et al, 2004; Wittek, 2007] suggest that in reality this effect does not
provide a strong perceptual cue. This is particularly true if the listeners position is
fixed or if the performance space contains significant early reflections and
reverberation. The only scenario where the wavefront curvature produced by a WFS
virtual source does seem to support the perception of distance is when the listener is
free to move around the listening area. In this case the movement of the listener
produces a sense of changing perspective or motion parallax due to correct curvature
of the virtual source over an extended area, and this has been shown to support the
perception of distance. While WFS is potentially highly suitable for certain
applications, clearly more research is required to overcome these significant technical
and logistical issues if WFS is to become more prevalent in performances of spatial
music.
Multichannel spatialization schemes such as Ambisonics and stereophony
would appear to be the most flexible approach, even though the performance of these
208

techniques is significantly affected when the listening area is expanded to surround a

group of listeners, or when significant early reflections and reverberation are present.
Although the majority of tests carried out with these systems have focussed on a
single listener in an acoustically treated space, these results do provide some
indication of the minimum technical requirements of any performance system.
Experience with the quadraphonic format revealed that a minimum of six
loudspeakers is required to produce a consistent localization with amplitude panning
[Thiele et al, 1976] and the results of various other tests indicate that a hexagonal
loudspeaker arrangement is also generally preferred for Ambisonics [Benjamin et al,
2006; Guastavino et al, 2004]. If a minimum of six loudspeakers is required for a
single listener then it is reasonable to assume that a greater number will be required
for an expanded listening area. Eight loudspeakers would therefore seem to be a
suitable minimum requirement and indeed, this particular arrangement has
increasingly become the standard for performances of electroacoustic music in recent
years. An octagonal array is particularly suitable as it allows for up to third order
Ambisonics, can be readily implemented with most digital audio hardware, and
provides eight distinct monophonic sources. Although this arrangement has become
increasingly common, no clear preference has emerged for either of the two possible
configurations shown in Figure 12.1. An octagonal layout with a centre-front
channel (Figure 12.1 left) is perhaps preferable in the case of a rectangular venue, as
potentially only channels four and eight would need to be repositioned. In addition,
this layout can be more readily adapted for other schemes such as quad or 5.1.
However, the alternative configuration (Figure 12.1 right) is perhaps more suitable for
the presentation of multiple two-channel stereo tracks.

Fig. 13.1 Octagonal loudspeaker layout with centre front (a) and without (b)

209

Even with this increase in the number of loudspeakers, reliable localization is

difficult to achieve for off-centre listeners with either stereophony or Ambisonics
[Bates et al, 2007b]. As amplitude panning techniques will only ever use a maximum
of two loudspeakers to produce a virtual source, this approach would seem to be
preferable if horizontal localization is a critical parameter. Higher order Ambisonic
decoding schemes which are optimized for extended listening areas (such as max-rE)
significantly reduce the number of contributing loudspeakers, however, stereophonic
virtual sources can readily collapse to a single loudspeaker, which will provide the
highest degree of localization accuracy possible. This feature is highly beneficial
when the source position is static but introduces significant artefacts when the source
is in motion as the number of contributing loudspeakers changes with the position of
the virtual source. A number of different tests [Pulkki et al, 2005; Martin et al, 1999;
Dickins et al, 1999] have shown that spatialization schemes which utilize a consistent
number of loudspeakers, irrespective of source position, provide a much smoother
trajectory than straight forward stereophonic panning. In addition to Ambisonics, a
number of alternative amplitude panning techniques have been developed (VBAP for
example) which can control the number of contributing loudspeakers independently of
the source azimuth. This is clearly very similar to the Ambisonics approach of
optimizing the energy vector rE for all directions, at the cost of reducing the maximum
localization accuracy that could be achieved at the loudspeaker positions. These
techniques may be beneficial if a source must change from a static position at a single
loudspeaker to a dynamic trajectory around the array, and are easier to implement for
certain non-symmetrical loudspeaker arrays. Indeed amplitude panning algorithms
such as these are in many ways analogous to a local ambisonic velocity decode
whereby only the loudspeakers closest to the source direction are used, and the
requirement to optimize the velocity component (rV = 1) is dropped. However, in
general, Ambisonics would seem to be more flexible, particularly in terms of
loudspeaker layout, and there is some evidence to suggest that Ambisonics is still
preferred for dynamically moving sources [Pernaux et al, 1998]. Other tests found
that stereophony provided precise localization and good readability but a lack of
immersion and envelopment, while Ambisonics provides an improved sense of
immersion and envelopment but poor localization accuracy and readability of the
scene [Gaustavino et al, 2007]. Consequently it would appear that stereophony is
preferable when the directionality and focus of the virtual source is paramount, while
210

Ambisonics is preferable if a more diffuse enveloping sound field is required or to

minimize panning artefacts in the case of dynamic sources. Additional distance cues
such as the direct to reverberant ratio, air absorption and the Doppler effect can also
improve dynamic trajectories and are relatively straightforward to implement in both
stereophonic and Ambisonics systems. There is evidence to suggest that the addition
of spatially accurate early reflections will improve the perception of distance but their
effect on horizontal localization is not entirely clear, particularly for off-centre
listeners. The results of tests carried out by Neher indicate that while a spatial early
reflection pattern is important, the precise angle of each early refection is not hugely
relevant [Neher, 2004].
The results presented in this section suggest an approach to spatialization
which is optimized in terms of the perceptual capabilities of these different
techniques. If a source must smoothly change from a static, tightly focussed position
to a dynamic and smooth trajectory then an extended amplitude panning technique is
perhaps preferable. However, if static and dynamic sources are to be contrasted than
a combination of mono or stereophony (for the static sources) and Ambisonics (for
dynamic sources) is preferable as the perceptible differences between these techniques
will support the contrast between the two different types of sources.

13.2 Perception and Spatial Music Aesthetics

In the latter half of this thesis, a number of landmark works of spatial music
were discussed and analysed in terms of the perceptual validity of their approach to
spatialization. Although many differences exist between these different compositions,
some general trends are evident. For example, a number of different composers used
spatial distribution to improve the intelligibility of different layers of independent and
potentially dissonant musical material; the co-existence of dissimilars described by
John Cage. It is entirely possible that this spatial distribution of the performers was
first implemented merely to facilitate the performance of overlapping yet unrelated
musical layers at different tempi or metres. However, there is now a significant
amount of scientific research to show that a listeners ability to detect and understand
the content of multiple signals is indeed improved when the signals are spatially
separated [Bregman, 1990; Shinn-Cunningham, 2003; Best, 2004]. Composers such
as Charles Ives, Henry Brant and Karlheinz Stockhausen were clearly well aware of

211

this fact, and they made extensive use of spatial distribution in their work for this
reason.
The presentation of any work of spatial music to a distributed audience will
necessarily result in certain members of the audience being situated outside the sweet
spot. This will result in an unavoidable distortion in the spatial trajectory which is
perceived by each member of the audience, and this distortion depends upon the
position of the listener within the array. Moore suggested that the differences in
perception among listeners is analogous to perspective distortion in photography or
cinematography [Moore, 1983]. Although this is perhaps unsurprising, it raises
significant questions about spatial music compositions which attempt to create and
relate recognizable sound shapes, such as in certain works by Iannis Xennakis for
example. This form of spatial counterpoint, which McNabb describes as the motivic
use of space assumes that spatial locations and trajectories are clearly and
unambiguously perceived by every listener [McNabb, 1986]. However the results of
the listening tests presented in this thesis show that this is extremely difficult to
achieve. Even in the case of point sources which are clearly localized, each listener
will be orientated differently with regards to the loudspeaker array, and so will have a
different perspective on the spatial layout. In this case it is very hard to see how
spatial motifs can be clearly and unambiguously perceived unless they are restricted
to very rudimentary movements. In the words of the composer Henry Brant,
Ideas of that kind seem to me more an expression of hope than of reality. . It is
hard enough to make the sounds do what you want in sound without saying that the
sound should be shaped like a pretzel or something like that [Brant, 1967]

It is difficult now to reliably gauge the perceptual effectiveness of vast

multimedia extravaganzas such as the Phillips Pavilion at Brussels in 1958 or the
Japanese Pavilion at Osaka in 1970. This form of audiovisual spatial music, referred
to by Xenakis as the polytype, contains a visual element which may support the
auditory spatial trajectory. Although there is very little perceptual evidence to suggest
that the abstract geometrical designs suggested by Varse and Xenakis can be reliably
achieved using sounds alone, this may be possible using a multimodal presentation
such as the polytype.
Stockhausen's serialization of angular direction in Kontakte is an often cited
example of an approach to spatialization which is not perceptually justified. While it
is certainly true that the carefully calculated directions worked out by Stockhausen
212

will not be precisely perceived by most, if not all the members of the audience, this is
perhaps not as critical as it first appears. Spatial movement in Kontakte is controlled
serially in the same way as rhythm and pitch. The precise angular location is not
intended to be accurately perceived by each member of the audience, instead, a
change in spatial location is used to indicate a certain temporal duration. So even
though each listener will perceive a slightly different spatial movement, they will all
perceive the same spatio-temporal movement. The abstract designs of these serialist
composers differ therefore from the abstract graphical designs discussed earlier in that
the specific trajectories and movements are not intended to be perceived by the
audience. The following quote from Pierre Boulez represents a very different
conception of spatial music to the highly visual form suggested by Xennakis and
Varse.
However, I do not want to suggest that one thereby plays as in tennis or car racing,
back and forth or in a cycle with everything following exactly. For me, sound
mixtures are far more interesting; one feels that they rotate somehow, but one can't
have the picture of an exactly comprehensible movement. With these different rates
and with these different times, one gets an impression, which is much richer for me
than a more normal circle or a loop to the left or right. For me, this is too anecdotal.
[Boulez et al, 1988]

Abstract and complex spatial designs can perhaps therefore be effective when used
indirectly in this fashion. This approach is somewhat reminiscent of other abstract
processes which have been used to indirectly create and control complex textures in
orchestral music. Consider this quote by the composer Gyorgy Ligeti from an
interview in 1978.
Technically speaking, I have always approached musical texture through partwriting. Both Atmospheres and Lontano have a dense canonic structure. But you
cannot actually hear the polyphony, the canon. You hear a kind of impenetrable
texture, something like a very densely woven cobweb. The polyphonic structure does
not come through, you cannot hear it, it remains hidden in a microscopic, underwater
world, to us inaudible [Bernard, 1987].

The dynamic spatial textures created using granular synthesis and flocking algorithms
would seem to function in the same way, as in this case only the overall motion and
the motion of sounds relative to each other will be perceived, while the individual
trajectories followed by each grainboid are not generally clearly perceptible. In this
case, it is the overall motion, rather than the specific location or direction, which is
important and hence, Ambisonics would seem to be a suitable spatialization scheme
for this purpose.

213

In general, two distinct approaches to the use of space as a musical parameter

have developed from the aesthetics of Musique Concrte and Elektronische Musik
which were prevalent in the mid-twentieth century. This dichotomy of an abstract
syntax or an abstracted syntax [Emmerson, 1986], an organic structure or an
architectonic structure [Harrison, 1999], has been discussed extensively by composers
and theorists for the past fifty years, and is also clearly evident in much of the spatial
music discussed in this thesis. In terms of spatial music, an abstract syntax involves
the application of predefined spatial effects to different sources which may have been
created specifically for this purpose, such as the serialist compositions of Stockhausen
for example. In music which follows an abstracted syntax, such as diffusion concerts
and Smalleys spectromorphological compositions, a spatial strategy is derived from
the sonic attributes (timbre, spectromorphological profile, etc) of the source signal.
This form of spatial music would seem to engage much more directly with the
perceptual difficulties that arise in presentations of spatial music to multiple listeners.
Although diffusion is in many respects quite limited in what it can achieve spatially, it
is admirably focussed on adapting each work for the particular performance space.
Smalley's theory of spectromorphology has its origins in the practice of diffusion, yet
this theory can equally be applied to other types of performances and aesthetics and a
gestural approach would seem to be highly applicable when live instrumental
performers and spatial audio are combined. The use of augmented instruments which
map the actions of the performer to a spatialization algorithm would also seem to be
very suitable for spatial electroacoustic performances, however, the necessity for
specialized and expensive hardware is a significant obstacle. Multi-transducer
devices such as the hexaphonic pickup are one possible solution to this problem as
they may be non-destructively adapted for existing instruments.

13.3 Conclusion
This thesis has attempted to examine the various interrelated factors which
influence the musical use of space in a performance context. It has been shown that
the optimum spatialization technique in any situation is highly dependent on other
factors, such as the overall compositional aesthetic. Various landmark works of
spatial music have been assessed in terms of the perceptibility of the particular
spatialization scheme adopted for that work, and also in terms of how the various

214

spatial attributes involved were incorporated into the overall compositional

framework. One strategy which has been adopted by many, quite different,
composers is the use of spatial distribution to improve the intelligibility of multiple
layers of independent musical material. In this instance, space is undoubtedly a
fundamental aspect of the composition which is crucial if the work is to be perceived
correctly. Although this is clearly an important aspect of spatial music, it is also quite
rudimentary and straightforward in terms of the actual spatialization scheme adopted.
Compositions which make use of more elaborate spatialization schemes are more
difficult to assess, however, one feature which is common to many of the
compositions discussed in this thesis is the way in which space is absorbed into the
overall compositional aesthetic. This stands in marked contrast to surround mixes of
popular music, an area which has received increased attention in recent years due to
the emergence of the DVD and 5.1 surround sound as a consumer format. Despite the
level of activity in this area, there is still little consensus as to why certain instruments
or sounds should be positioned at certain locations. While the spatial distribution of
parts will help to separate and clarify the different parts, it will also significantly alter
the rhythmic and harmonic relationship between them, and this may or may not be of
benefit, depending on the musical context.
The same cannot be said of such well regarded compositions such as the
Unanswered Question, Kontakte, Empty Vessels or Rpons as in these works, spatial
effects are tightly integrated within the overall compositional strategy and are not
simply applied for their own sake. This implies that if space is to be successfully used
as a musical parameter, it must be supported by other aspects of the work. This
conception of space is difficult to reconcile with suggestions that space can become
the primary carrier of meaning [Henriksen, 2002] in a musical composition or that
spatial structures can become more significant than the actual sounds themselves. The
composer Johannes Goebel suggests in the following quote that space can function in
a similar fashion to dynamic differences in volume which seems to agree with many
of conclusions of this thesis.
In my opinion the spatial placement of sounds, whether instrumental or
electronically, has about the same potential for aesthetic differentiation as loudness.
Compared to pitch and timbre, localization yields far less potential for aesthetic
differentiation, but on the other hand no one would deny that in quite a few pieces
loudness is an important, highly sophisticated, composed part of music. And the
same could be said for the distribution of sound in space.

[Goebel, 2001]
215

The artistic control of volume can undoubtedly add expressiveness to a musical

performance and the importance of dynamics has long been an essential part of music
composition. In the opinion of the author, this is equally true for the musical use of
space, and this is particularly true in the case of electronic music. Performances of
electronic music often lack the visual component of traditional instrumental music
performances. During an instrumental performance visible actions by the performer
produces an audible result and it is this multi-modal feedback process that makes a
performance live. The dynamic spatialization of electronic sounds, particularly the
gestural/environmental structuring principle suggested by Denis Smalley, can
potentially replace or substitute for this missing visual component and provide a
necessary and potentially thrilling physicality.

216

14 List of Publications
P1

Bates, E. & Furlong, D., Score File Generators for Boids Based Granular
Synthesis in Csound, Proceedings of the 126th Audio Engineering Society
Convention, Munich, Germany, May 2009.

P2

Bates, E., Furlong, D. & Dennehy, D. Adapting Polyphonic Pickup

Technology for Spatial Music Performance, Proceedings of the International
Computer Music Conference, Belfast, Northern Ireland, August 2008.

P3

Bates, E., Kearney, G., Furlong, D. & Boland, F. Localization Accuracy of

Advanced Spatialization Techniques in Small Concert Halls, Proceedings of
the 153rd Meeting of the Acoustical Society of America, Salt Lake City, USA,
June 2007.

P4

Bates, E., Kearney, G., Furlong, D. & Boland, F. Monophonic Source

Localization for a Distributed Audience in a Small Concert Hall,
Proceedings of the 10th International Conference on Digital Audio Effects,
Bordeaux, France, September 2007.

217

15 Summary of Published Articles

P1

In this paper we present a set of score file generators and granular synthesis
instruments for the Csound language. The applications use spatial data
generated by the Boids flocking algorithm along with various user-defined
values to generate score files for grainlet additive synthesis, granulation and
glisson synthesis instruments. Spatialization is accomplished using Higher
Order Ambisonics and distance effects are modelled using the Doppler Effect,
early reflections and global reverberation. The sonic quality of each synthesis
method is assessed and an original composition by the author is presented.

P2

This paper describes how polyphonic pickup technology can be adapted for
the spatialization of electric stringed instruments such as the violin, cello and
guitar. It is proposed that mapping the individual strings to different spatial
locations integrates the spatial diffusion process with the standard musical
gestures of the performer. The development of polyphonic guitar processing
is discussed and a method of adapting MIDI guitar technology for this purpose
is presented. The compositional and technical strategies used with various
augmented instruments is presented along with an analysis of three
compositions by the author for spatialized hexaphonic guitar.

P3

A comparison of spatialization schemes is presented in terms of their

localization accuracy under the non-ideal listening conditions found in small
concert halls. Of interest is the effect of real reverberant conditions, noncentral listening positions and non-circular speaker arrays on source
localization. The data is presented by comparison of empirical binaural
measurements to perceptual listening tests carried out using Ambisonics,
Vector Base Amplitude Panning (VBAP), Spat (with B-format encoding) and
Delta Stereophony (DSS) systems. The listening tests are conducted by
comparing the localization of phantom sources generated by the spatialization
systems, to monophonic sources generated by reference loudspeakers. The
reference and phantom sources are presented at front, side and back locations
about a 9 listener audience, and the systems are tested in a random order with
218

a calibrated 16 loudspeaker array situated around the audience area. The

binaural recordings are compared to the subjective measurements of
localization accuracy through the inter-aural time difference (ITD) cues at
each listener position.
P4

The transfer of multichannel spatialization schemes from the studio to the

concert hall presents numerous challenges to the contemporary spatial music
composer or engineer. The presence of a reverberant listening environment
coupled with a distributed audience are significant factors in the presentation
of multichannel spatial music. This paper presents a review of the existing
research on the localization performance of various spatialization techniques
and their ability to cater for a distributed audience. As the first step in a major
comparative study of such techniques, the results of listening tests for
monophonic source localization for a distributed audience in a reverberant
space are presented. These results provide a measure of the best possible
performance that can be expected from any spatialization technique under
similar conditions.

219

Bibliography
[Ahrens et al, 2009] Ahrens, J. and Spors, S., Spatial Encoding and Decoding of
Focussed Virtual Sound Sources, Graz Ambisonics Symposium, 2009.
[Ambisonics Standards, 2009] Analysis of Normalisation Schemes: A supporting
document for the "Universal Ambisonic Standard", Ambisonics Google Groups,
http://docs.google.com/Doc?id=df4dtw69_66d4khc7fb&hl=en
[Asano et al, 1990] Asano, F., Suzuki, Y. and Sone, T., Role of spectral cues in
median plane localization, Journal of the Acoustical Society of America, Vol. 88,
pp. 159-168, 1990.
[Baalman, 2006] Baalman, M., swonder3Dq: Auralisation of 3D objects with Wave
Field Synthesis, Proceedings of the Linux Audio Conference, 2006.
[Ballan et al, 1994] Ballan, O., Mozzoni, L. and Rocchesso, D., Sound spatialization
in real time by first reflection simulation, Proceedings of the International Computer
Music Conference, pp. 475-476, 1994.
[Ballas et al, 2001] Ballas, J. A., Brock, D., Stroup, J. and Fouad, H., The Effect of
Auditory Rendering on Perceived Movement: Loudspeaker Density and HRTF,
Proceedings of the 2001 International Conference on Auditory Display, 2001.
[Bamford, 1995], Bamford, J. S. An Analysis of Ambisonic Sound Systems of First
and Second Order, Doctoral Thesis, University of Waterloo, Canada, 1995.
[Barrett, 1990] Barrett, N., Spatio-musical Composition Stratagies, Organized
Sound, Vol. 7(3), 2003.
[Barrett et al, 2007] Barrett, N. and Otondo, F. Creating Sonic Spaces: an interview
with Natasha Barrett, Computer Music Journal, Vol. 31(2), pp. 10-19, 2007.
[Barron et al, 1981] Barron, M. and Marshall, A. H., Spatial Impression due to
220

Early Lateral Reflections in Concert Halls: The Derivation of a Physical Measure,

Journal of Sound and Vibration, Vol. 77(2), pp. 211-232, 1981.
[Bartetzki, 2008] Bartetzki, A., Csound score generation and granular synthesis
with cmask www.kgw.tuberlin.de/abart/cmaskpaper/cmaskarticle.html, retrieved on
07-07-2008.
[Bates et al, 2007a] Bates, E., Kearney, G., Boland, F. and Furlong, D., Monophonic
Source Localization for a Distributed Audience in a Small Concert Hall, 10th
International Conference on Digital Audio Effects (DAFX-07), 2007.
[Bates et al, 2007b] Bates, E., Kearney, G., Boland, F. and Furlong, D., Localization
Accuracy of Advanced Spatialization Techniques in Small Concert Halls, 153rd
Meeting of the Acoustical Society of America, 2007.
[Bauer, 1962] Bauer, B. B., Phasor Analysis of Some Stereophonic Phenomena,
The Journal of the Acoustical Society of America, Vol. 33(11), pp. 1536-1962.
[Begault, 1992] Begault, D., "Perceptual Effects of Synthetic Reverberation on ThreeDimensional Audio Systems", Journal of the Audio Engineering Society, Vol. 40(11),
1992.
[Begault, 1994] Begault, D. R., 3-D Sound for Virtual Reality and Multimedia,
Academic Press Inc, pp. 45-46; pp. 51-56, 1994.
[Benjamin et al, 2006] Benjamin, E. M., Lee, R. and Heller, A. J., Localization in
Horizontal-Only Ambisonic Systems, Proceedings of the 121st Convention of the
Audio Engineering Society, 2006.
[Beranek, 1996], Beranek, L., Concert and opera halls. How they sound, A.I.P.
Press, 1996.
[Berg, 2007] Berg, J., Midi breakout box, www.unfretted.com, 2007, retrieved on
01/01/2009.
221

[Berkhout, 1998] Berkhout, A. J., A Wavefield Approach to Multichannel Sound,

Proceedings of the 104th Convention of the Audio Engineering Society, Preprint 4749,
1998.
[Bernard, 1987] Bernard, J. W., Inaudible Structures, Audible Music: Ligeti's
Problem, and His Solution, Music Analysis, Vol. 6(3), pp. 207-236, 1987.
[Bertet et al, 2009] Bertet, S., Daniel, J., Parizet, E. and Warusfel, O, Influence of
Microphone and Loudspeaker Setup on Perceived Higher Order Ambisonics
Reproduced Sound Field, Proceedings of the Graz Ambisonics Symposium, 2009.
[Best, 2004] Best, V. A., Spatial Hearing with Simultaneous Sound Sources: A
Psychophysical Investigation, Doctoral Thesis, University of Sydney, 2004.
[Bisig et al, 2008] Bisig, D., Neukom, M. and Flury, J. Interactive Swarm Orchestra
- An Artificial Life Approach to Computer Music, Proceedings of the International
Computer Music Conference, 2008.
[Blackwell et al, 2004] Blackwell, T. and Young, M., Self-organised music,
Organized Sound, Vol. 9, pp. 123136, 2004.
[Blauert, 1997] Blauert, J., Spatial Hearing: The Psychophysics of Human Sound
Localisation, pp. 3-4, pp. 38-39; pp. 47-48 pp. 118-119, pp. 140-154, pp173-190;pp.
348-358; MIT Press; 1997.
[Boone et al, 1995] Boone, M. M., Verheijen, E. N. G. and van Tol, P. F., "Spatial
sound field reproduction by wave field synthesis, Journal of the Audio Engineering
Society, Vol. 43(12), pp. 1003-1012, 1995.
[Boone et al, 1999] Boone, M. M., Horbach, U. and de Bruijn, W. P. J., Virtual
Surround Speakers with Wave Field Synthesis, Proceedings of the 106th Convention
of the Audio Engineering Society, Preprint 4928, 1999.

222

[Boone et al, 2003] Boone, M. M. and de Bruijn, W., "Improved speech intelligibility
in teleconferencing using Wave Field Synthesis", Proceedings of the 114th Convention
of the Audio Engineering Society, Preprint 5800, 2003.
[Boulanger, 1986] Boulanger, R., Toward a new age of performance: Reading the
book of dreams with the Mathews electronic violin, Perspectives of New Music, Vol.
24(2), pp. 130155, 1986.
[Boulanger, 2000] Boulanger, R., The Csound book, MIT Press, 2000.
[Boulez et al, 1988] Boulez, P. and Gerzso, A., Computers in Music, Scientific
American, Vol. 258(4), 1988.
[Brant, 1967] Brant, H., "Space as an essential aspect of musical composition",
Contemporary Composers on Contemporary Music, Holt, Rinehart and Winston, pp.
221-242, 1967.
[Bregman, 1990] Bregman, A. S., Auditory Scene Analysis, 196-200, pp. 204-250;
pp. 455-490, MIT Press; 1990.
[Bromberger, 2007] Bromberger, E., Berlioz Requiem - San Diego Symphony
Orchestra Program Notes, 2007, www.sandiegosymphony.
com/uploads/pdfs/pdfmw060713.pdf, retrieved on 06/06/2009.
[Bronkhurst, 2002] Bronkhorst, A. W., "Modeling auditory distance perception in
rooms, Proceedings of the EAA Forum Acusticum Sevilla, 2002.
[Brungart et al, 1999] Brungart, D. S. and Rabinowitz, W. M., Auditory localization
of nearby sources. Head-related transfer functions, Journal of the Acoustical
Society of America, Vol. 106(3), 1999.
[Bryant, 1981] Bryant, D., The cori spezzati of St Mark's: myth and reality, Early
Music History, Vol. 1, pp. 165-186, 1981.

223

[Cage, 1957] Cage, J., Experimental Music, Address given to the convention of
Music Teachers National Association, Chicago, 1957. Reprinted in Silence, MIT
Press, 1961.
[Capra et al, 2007] Capra, A., Fontana, S., Adriaensen, F., Farina, A. and Grenier, Y.,
Listening Tests of the Localization Performance of Stereodipole and Ambisonic
Systems, Proceedings of the 123rd Convention of the Audio Engineering Society,
Preprint 7187, 2007.
[Caulkins et al, 2003] Caulkins, T., Corteel, E. and Warusfel, O., Wave Field
Synthesis Interaction with the Listening Environment, Improvements in the
Reproduction of Virtual Sources Situated inside the Listening Room, Proceedings of
the 6th International Conference on Digital Audio Effects (DAFX-03), 2003.
[Chowning, 1971], Chowning, J., The simulation of moving sound sources, Journal
of the Audio Engineering Society, Vol. 19, pp. 26, 1971.
[Clark et al, 1958] Clark, H. A. M., Dutton, G. F., Vanderlyn, P. B., The
'Stereosonic' Recording and Reproducing System: A Two-Channel Systems for
Domestic Tape Records, Journal of the Audio Engineering Society, Vol. 6(2), pp.
102-117, 1958.
[Corteel et al, 2003] Corteel, E. and Nicol, R., "Listening room compensation for
Wave Field Synthesis. What can be done?", Proceedings of the 23rd International
Conference of the Audio Engineering Society, 2003.
[Corteel et al, 2004] Corteel E., Caulkins T., Warusfel O., Pellegrini R., Rosenthal M.
and Kuhn C., Sound Scene Creation using Wave Field Synthesis, Actes Journes du
Design Sonore, 2004.
[Corteel et al, 2008] Corteel, E., Pellegrini, R. and Kuhn-Rahloff, C., Wave Field
Synthesis with increased aliasing frequency, Proceedings of the 124th Convention of
the Audio Engineering Society, Preprint 7362, 2008.

224

[Cott, 1973], Cott, J., Stockhausen; conversations with the composer, Simon and
Schuster, 1973.
[Daniel et al, 1998] Daniel, J., Rault, J. and Polack, J. D., Ambisonics Encoding of
Other Audio Formats for Multiple Listening Conditions, Proceedings of the 105th
Convention of the Audio Engineering Society, Preprint 4795, 1998.
[Daniel et al, 1999], Daniel, J., Nicol, R. and Moreau, S, Further Investigations of
High Order Ambisonics and Wavefield Synthesis for Holophonic Sound Imaging,
Proceedings of the 114th Convention of the Audio Engineering Society, Preprint
5788, 1999.
[Daniel, 2000] Daniel, J., Reprsentation de champs acoustiques, application la
transmission et la reproduction de scnes sonores complexes dans un contexte
multimdia, Doctoral thesis, Universit Paris, 2000.
[Daniel, 2003] Daniel, J., Spatial Sound Encoding Including Near Field Effect:
Introducing Distance Coding Filters and a Viable, New Ambisonic Format,
Proceedings of the 23rd International Conference of the Audio Engineering Society,
2003.
[de Vries, 1995] de Vries, D., "Sound enhancement by wave field synthesis: adaption
of the synthesis operator to the loudspeaker directivity characteristics", Proceedings
of the 98th Convention of the Audio Engineering Society, Preprint 3971, 1995.
[Dickins et al, 1999] Dickins, D., Flax, M., Mckeag, A. and Mcgrath, D., Optimal
3D speaker panning, Proceedings of the 16th International Convention of the Audio
Engineering Society, 1999.
[Dobrian, 2008] Dobrian, C., Doppler shift in a stereo Csound instrument,
http://music.arts.uci.edu/dobrian/dsp2001/csound/dopplertheory.htm, retrieved on
07/07/2008.

225

[Elfner et al, 1968] Elfner, L. F. and Tomsic, R. T., Temporal and intensive factors
in binaural lateralization of auditory transients, Journal of the Acoustical Society of
America, Vol. 43, pp. 747-751, 1968.
[Emmerson, 1986] Emmerson, S., The relation of language to materials, The
Language of Electroacoustic Music, pp. 20-24, Macmillan Press, 1986.
[Essl, 1995] Essl, K., Lexikon-Sonate. An Interactive Realtime Composition for
Computer-Controlled Piano, Proceedings of the 2nd Brazilian Symposium on
Computer Music, 1995.
[Farina et al, 1998] Farina A. and Ugolotti E., Software Implementation of B-Format
Encoding and Decoding, Proceedings of the 104th International AES Convention,
1998.
[Farina et al, 2000] Farina, A., Zanolin, M. and Crema, E., Measurement of sound
scattering properties of diffusing panels through the Wave Field Synthesis approach,
Proceedings of the 108th International Convention of the Audio Engineering Society,
2000.
[Felder, 1977] Felder, D., An interview with Karlheinz Stockhausen, Perspectives of
New Music, Vol. 16(1), pp. 85101, 1977.
[Fellget, 1975] Fellgett, P., Ambisonics. Part One: General System Description,
Studio Sound, Vol.17, pp.2022;40, 1975.
[Fels, 1996] Fels, P., 20 Years Delta Stereophony System - High Quality Sound
Design, Proceedings of the 100th Convention of the Audio Engineering Society,
Preprint 4188, 1996.
[Franco et al, 2004] Franco, A. F., Merchel, S., Pesqueux, L., Rouaud, M. and
Soerensen, M. O., "Sound Reproduction by Wave Field Synthesis", Project Report,
Aalborg University, 2004.

226

[Frank et al, 2008] Frank, M., Zotter, F. and Sontacchi, A., Localization
Experiments Using Different 2D Ambisonics Decoders, 25th Tonmeistertagung
VDT International Convention, 2008.
[Franssen, 1962] Franssen, N. V., Stereophony, Philips Technical Library, The
Netherlands, English translation (1964).
[Furlong et al, 1992] Furlong, D. J. and MacCabe, C. J., Loudspeakers, Listening
Rooms, and Effective Synthetic Auralization, Proceedings of the 93rd Convention of
the Audio Engineering Society, Preprint 3445, 1992.
[Furse, 2000] Furse, R., First and Second Order Ambisonic Decoding Equations,
http://www.muse.demon.co.uk/ref/speakers.html, retrieved on 03/06/2009.
[Gardner, 1962] Gardner, M. B., "Binaural Detection of Single- Frequency Signals in
Presence of Noise", Journal of the Acoustical Society of America, Vol. 34, pp. 18241830, 1962.
[Gayou, 2007] Gayou, E., The GRM: landmarks on a historic route, Organised
Sound, Vol. 12(3), pp. 203211, Cambridge University Press, 2007.
[Guastavino et al, 2004] Guastavino, C and Katz, B. F. G., Perceptual evaluation of
multi-dimensional spatial audio reproduction, Journal of the Acoustical Society of
America, Vol. 116(2), 2004.
[Guastavino et al, 2007] Guastavino, C., Larcher, V., Catusseau, G. and Boussard, P.,
Spatial Audio Quality Evaluation: Comparing Transaural, Ambisonics and Stereo,
Proceedings of the 13th International Conference on Auditory Display, 2007.
[Gerzon, 1974a] Gerzon, M. A., Sound Reproduction Systems, Patent No. 1494751.
[Gerzon, 1974b] Gerzon, M. A., Surround Sound Psychoacoustics", Wireless World,
Vol. 80, pp. 483-486, 1974.

227

[Gerzon, 1992a], Gerzon, M. A., General Methatheory of Auditory Localisation,

Proceedings of the 92nd International Audio Engineering Society, Preprint 3306, 1992.
[Gerzon, 1992b] Gerzon, M. A., "The Design of Distance Panpots", Proceedings of
the 92nd Audio Engineering Society Convention, Preprint 3308, 1992.
[Gerzon, 1992c], Gerzon M. A., Psychoacoustic Decoders for Multispeaker Stereo
and Surround Sound, Proceedings of the 93rd International Audio Engineering
Society Convention, Preprint 3406, 1992.
[Goebel, 2001] Goebel, J., Letters, Computer Music Journal, Vol. 25(4), pp. 5-11,
MIT Press, 2001.
[Griesinger, 2009], Griesinger, D., The importance of the direct to reverberant ratio
in the perception of distance, localization, clarity, and envelopment (auditory
engagement), Proceedings of the 126th Audio Engineering Society Convention,
2009.
[Grhn , 2002] Grhn, M., Localization of a Moving Virtual Sound Source in a
Virtual Room, the Effect of a Distracting Auditory Stimulus, Proceedings of the 2002
International Conference on Auditory Display, 2002.
[Harley, 1997] Harley, M. A., An American in Space: Henry Brant's "Spatial
Music", American Music, Vol. 15(1), pp. 70-92, 1997.
[Harley, 1998a] Harley, M. A., Spatiality of Sound and Stream Segregation in 20thCentury Instrumental Music, Organized Sound, Vol. 3(2), pp. 147-166, 1998.
[Harley, 1998b] Harley, M. A., Music of Sound and Light: Xenakis's Polytopes,
Leonardo, The MIT Press, Vol. 31(1), pp. 55-65, 1998.

228

[Harris et al, 1998] Harris, N. and Flanagan, S., Stereophonic Localization in

Rooms, Comparing Conventional and Distributed-Mode Loudspeakers, Proceedings
of the 105th Convention of the Audio Engineering Society, Preprint 4794, 1998.
[Harrison, 1999] Harrison, J., Sound, space, sculpture: some thoughts on the what,
how and why of sound diffusion, Organised Sound, Vol. 3(2), pp. 11727,
Cambridge University Press, 1999.
[Hartmann, 1983] Hartmann, W. M., Localization of Sound in Rooms, Journal of
the Acoustical Society of America, Vol. 74, pp. 1380-1391, 1983.
[Hartmann et al, 1985] Hartmann, W. M. and Rackerd, B., Localization of Sound in
Rooms II: The Effects of a Single Reflecting Surface, Journal of the Acoustical
Society of America, Vol. 78, pp524-533, 1985.
[Hartmann et al, 1989] Hartmann, W. M. and Rackerd, B., Localization of sound in
rooms IV: The Franssen effect, Journal of the Acoustical Society of America, Vol.
86(4), pp. 1366-1373, 1989.
[Husler et al, 1985] Husler, J. and Boulez, P., Pierre Boulez on Rpons, Teilton,
Schriftenreihe, der Heinrich-Strobel-Stiftung des Sdwestfunks, Vol. 4, pp. 7-14,
Kassel, Brenreiter, translated from the German by Clemens Merkel and Christina
Unger, 1985.
[Henriksen, 2002] Henriksen, F. E., Space in Electroacoustic Music: Composition,
Performance and Perception of Musical Space, Doctoral thesis, City University,
London, 2002.
[Hodgkinson, 2001] Hodgkinson, T., An Interview with Pierre Schaeffer, The Book
of Music & Nature, Wesleyan University Press, pp. 34-44, 2001.
[Hoffman, 2001] Hoffmann, P., "Iannis Xenakis", The New Grove Dictionary of
Music and Musicians, Vol. 27, pp. 607-608, Macmillan Press, 2001.

229

[Hofmann, 2008] Hofmann, J. J., Ambisonics tutorial,

http://www.sonicarchitecture.de, retrieved on 07/07/2008.
[Howard et al, 1996] Howard, D. M. and Angus, J., Acoustics and
Psychoacoustics, pp. 257-262, Focal Press, 1996.
[Huber, 2002] Huber, T., Zur Lokalisation akustischer Objekte bei
Wellenfeldsynthese, M.Phil thesis, IRT/ TU Munich, 2002, www.hauptmikrofon.
de/wfs.html, retrieved on 06/06/2009.
[Hubel et al, 1959] Hubel, D. H. and Wiesel, T. N., Receptive fields of single
neurons in the cats striate cortex, Journal of Physiology, Vol. 148, pp. 547-579,
1959.
[Jehan et al, 1999] Jehan, A., Freed, T. and Dudas, R., Musical applications
of new filter extensions to max/msp, Proceedings of the International Computer
Music Conference, pp. 504507, 1999.
[Johnson, 2002] Johnson, M. E., Charles Ives's (Utopian, Pragmatist, Nostalgic,
Progressive, Romantic, Modernist) Yankee Realism, American Music, Vol. 20(2),
pp. 188-233, University of Illinois Press, 2002.
[Jot, 1999] Jot, J., "Real-time spatial processing of sounds for music, multimedia and
interactive human-computer interfaces", Multimedia Systems, Vol. 7(1), pp. 55-69,
1999.
[Jot et al, 1999] Jot, J., Larcher, V. and Pernaux, J. M., A Comparative Study of 3-D
Audio Encoding and Rendering Techniques, Proceedings of the 16th International
Conference of the Audio Engineering Society, 1999.
[Kendall et al, 1984] Kendall, G. S. and Martens, W. L., Simulating the Cues of
Spatial Hearing in Natural Environments, Proceedings of the 1984 International
Computer Music Conference, 1984.

230

[Kerber et al, 2004] Kerber, S., Wittek, H., Fastl, H. and Theile, G., Experimental
investigations into the distance perception of nearby sound sources: Real vs. WFS
virtual nearby sources, Joint Meeting of the German and the French Acoustical
Societies (CFA/DAGA), 2004.
[Komiyama et al, 1991] Komiyama, S., Morita, A., Kurozumi, K. and Nakabayashi,
K., Distance control system for a sound image, Proceedings of the 9th International
Conference of the Audio Engineering Society, pp. 233239, 1991.
[Kratschmer et al, 2009] Kratschmer, M. and Rabenstein, R., Implementing
Ambisonics on a 48 Channel Circular Loudspeaker Array, Proceedings of the Graz
Ambisonics Symposium, 2009.
[Ligeti, 1958] Ligeti, G., Metamorphoses of Musical Form, Die Reihe, Vol. 7, pp.
5-19, 1958.
[Litovsky et al, 1999], Litovsky, R.Y., Colburn, H. S, Yost, W. A., Guzman, S. J.,
The Precedence Effect, Journal of the Acoustical Society of America, Vol. 106(4),
pp. 1633-1654, 1999.
[Lund, 2000] Lund, T., Enhanced Localization in 5.1 Production, Proceedings of
the 105th Convention of the Audio Engineering Society, Preprint 5243, 2000.
[MacDonald, 1995] MacDonald, A., Performance Practice in the Presentation of
Electroacoustic Music, Computer Music Journal, Vol. 19(4), pp. 88-92, 1995.
[Machover, 1992] Machover, T., Hyperinstruments - MIT Media Lab Research
Report, 1992.
[MacPherson, 2002] MacPherson, E. A. and Middlebrooks, J. C., Listener weighting
of cues for lateral angle: The Duplex Theory of sound localization revisited, Journal
of the Acoustical Society of America, Vol. 111, pp. 22192236, May 2002.

231

[Makita, 1962], On the Directional Localization of Sound in the Stereophonic Sound

Field", E. B. U. Review, part A no. 73, pp. 102-108, 1962.
[Malham, 1992] Malham, D. G., Experience with large area 3-D Ambisonic sound
systems, Proceedings of the Institute of Acoustics, Autumn Conference on
Reproduced Sound, Vol. 8, 1992.
[Malham et al, 1995], Malham, D. G. and Myatt, A. "3-D Sound Spatialization using
Ambisonic Techniques", Computer Music Journal, Vol. 19(4), pp. 58-70, 1995.
[Malham, 1998] Malham, D. G. Sound Spatialisation, Proceedings of the First
COST-G6 Workshop on Digital Audio Effects (DAFX98), pp. 62-70, 1998.
[Malham, 1999] Malham, D. G., Higher order Ambisonic systems, abstracted from
Space in Music - Music in Space, M. Phil Thesis, University of York, 2003.
[Marentakis et al, 2008] Marentakis, G., Corteel, E. and McAdams, S., Wave Field
Synthesis Evaluation using the Minimum Audible Angle in a Concert Hall,
Proceedings of the 124th Convention of the Audio Engineering Society, Preprint 7360,
2008.
[Martin et al, 1999a] Martin, G., Woszczyk, W., Corey, J. and Quesnel, R., Sound
source localization in a five channel surround sound reproduction system,
Proceedings of the 104th Convention of the Audio Engineering Society, Preprint 4994,
1999.
[Martin et al, 1999b] Martin, G., Woszczyk, W., Corey, J. and Quesnel, R.,
Controlling Phantom Image Focus in a Multichannel Reproduction System,
Proceedings of the 107th Conference of the Audio Engineering Society, Preprint 4996,
1999.
[Martin et al, 2001] Martin, G., Depalle, P., Woszczyk, W., Corey, J. and Quesnel, R.,
A Hybrid Model for Simulating Diffused First Reflections in Two-Dimensional

232

Synthetic Acoustic Environments, Proceedings of the 19th Conference of the Audio

Engineering Society, 2001.
[Martin, 2005] Martin, A., Algorithms for Sonification: Smoke, Boids and the Game
of Life, M. Phil Thesis, Trinity College Dublin, 2005.
[McFarlane, 2001] Mc Farlane, M. W., "The Development of Acousmatics in
Montral", Journal of the Canadian Electroacoustic Community, eContact!, Vol. 6(2),
2001.
[McNabb, 1986] McNabb, M. Computer Music: Some Aesthetic Considerations,
The Language of ElectroAcoustic Music, pp. 146-149, Macmillan Press, 1986.
[Goodale et al, 1992] Goodale, M. A. and Milner, A. D. Separate visual pathways
for perception and action, Trends in Neurosciences, Vol. 15(1), pp. 20-25, 1992.

[Mershon et al, 1975] Mershon, D. H. and King, L. E. Intensity and reverberation

as factors in the auditory perception of egocentric distance, Perception &
Psychophysics, Vol. 18, pp. 409-415, 1975.
[Mershon, 1997] Mershon, D. H. Phenomenal geometry and the measurement of
perceived auditory distance, Chapter 13 in Binaural and Spatial Hearing in Real
and Virtual Environments, Lawrence Erlbaum Assoc. pp. 257-274, 1997.
[Meyer-Eppler, 1949] Meyer-Eppler, W., Elektronische Klangerzeugung:
Elektronische Musik und synthetische Sprache, Bonn: Ferdinand Dmmlers, 1949.
[Michelsen et al, 1997] Michelsen, J. and Rubak, P., Parameters of Distance
Perception in Stereo Loudspeaker Scenario, Proceedings of the 102nd Convention of
the Audio Engineering Society, Preprint 4472, 1997.

233

[Misch et al, 1998] Misch, L., Hentschel, F. and Kohl, J. On the Serial Shaping of
Stockhausen's Gruppen fr drei Orchester, Perspectives of New Music, Vol. 36(1),
pp. 143-187, 1998.
[Moore, 1983], Moore, F. R., A general model for spatial processing of sounds,
Computer Music Journal, Vol. 7, pp. 615, 1983.
[Moore et al, 2007] Moore, J. D. and Wakefield, J. P., Surround Sound for Large
Audiences: What are the Problems?, Annual University of Huddersfield
Researchers' Conference. 2007.
[Moorer, 1979] Moorer, J. A., "About This Reverberation Business", Computer Music
Journal, Vol. 3(2), pp. 13-28, 1979.
[Moreau et al, 2006] Moreau, S., Daniel, J. and Bertet, S., 3D Sound Field
Recording with Higher Order Ambisonics Objective Measurements and Validation
of a 4th Order Spherical Microphone, Proceedings of the 120th Convention of the
Audio Engineering Society, 2006.
[Morgan, 1978] Morgan, R. P., Ives and Mahler: Mutual Responses at the End of an
Era, 19th-Century Music, Vol. 2(1), pp. 72-81, University of California Press, 1978.
[Morgan, 1991] Morgan, R. P., Stockhausen's Writings on Music, The Musical
Quarterly, Vol. 75(4), pp. 194-206, Oxford University Press, 1991.
[Moritz, 2002] Moritz, A., Stockhausen: Essays on the Works,
http://home.earthlink.net/~almoritz/, retrieved on 01/05/2009.
[Mortenson, 1987] Mortenson, G. C., Father to Son: The Education of Charles
Ives, Music Educators Journal, Vol. 73(7), pp. 33-37, 1987.
[Neher, 2004] Neher, T., Towards A Spatial Ear Trainer, Doctoral Thesis,
University of Surrey, pp. 126-164, 2004.

234

[Neukon et al, 2008] Neukom, M. and Schacher, J. C., Ambisonics Equivalent

Panning, Proceedings of the International Computer Music Conference, 2008.
[Nicol et al, 1999], Nicol, R. and Emerit, M., 3-D-sound reproduction over an
extensive listening area: A hybrid method derived from holophony and ambisonic,"
Proceedings of the 16th International Conference of the Audio Engineering Society,
1999.
[Nielsen, 1993] Nielsen, S. H., Auditory distance perception in different rooms,
Journal of the Audio Engineering Society, Vol. 41, pp. 755-770, 1993.
[Nogus et al, 2003] Nogus M., Corteel E. and Warusfel O., Monitoring Distance
Effect with Wave Field Synthesis, Proceedings of the 6th International Conference on
Digital Audio Effects (DAFX-03), 2003.
[Ono et al, 1997] Ono, K., Komiyama, S. and Nakabayashi, K., A method of
reproducing concert hall sounds by loudspeaker walls, Proceedings of the 102nd
Convention of the Audio Engineering Society, Preprint 4490, 1997.
[Palombini, 1993] Palombini, C., Machine Songs V: Pierre Schaeffer: From
Research into Noises to Experimental Music, Computer Music Journal, Vol. 17(3),
pp. 14-19, MIT Press, 1993.
[Pernaux et al, 1998] Pernaux, J., Boussard, P. and Jot, J., Virtual Sound Source
Positioning and Mixing in 5.1 Implementation on the Real-Time System Genesis,
Proceedings of the International Conference on Digital Audio Effects (DAFx-98),
1998.
[Plenge et al, 1975], Plenge, G., Lehmann, P., Wettschurek, R. and Wikens, H., New
methods in architectural investigations to evaluate the acoustic qualities of concert
halls, Journal of the Acoustical Society of America, Vol. 57, pp. 1292-1299, 1975.
[Poletti, 1996] Poletti, M., The Design of Encoding Functions for Stereophonic and
Polyphonic Sound Systems, Journal of the Audio Engineering Society, Vol. 44(11),
pp. 948-963, 1996.
235

[Pulkki, 1997] Pulkki, V., Virtual sound source positioning using Vector Base
Amplitude Panning, Journal of the Audio Engineering Society, Vol. 45, pp. 456
466, 1997.
[Pulkki, 1999] Pulkki, V., "Uniform spreading of amplitude panned virtual sources",
Proceedings of the 1999 IEEE Workshop on Applications of Signal Processing to
Audio and Acoustics, Mohonk Mountain House, 1999.
[Pulkki et al, 2001] Pullki, V. and Karjalainen, M. Directional Quality of 3-D
Amplitude-Panned Virtual Sources, Proceedings of the 7th International Conference
on Auditory Display (ICAD2001), pp. 239-244, 2001.
[Pulkki, 2002] Pulkki, V., Compensating displacement of amplitude-panned virtual
sources, Proceedings of the 22nd International Conference of the Audio Engineering
Society, pp. 186-195, 2002.
[Pulkki et al, 2005] Pulkki, V. and Hirvonen, T., Localization of Virtual Sources in
Multichannel Audio Reproduction, IEEE Transactions on Speech and Audio
Processing, Vol. 13(1), 2005.
[Ratliff, 1974] Ratliff, P. A., "Properties of Hearing Related to Quadrophonic
Reproduction", BBC Research Department Report, BBC RD 38, 1974.
[Reynolds, 1987] Reynolds, C. W., Flocks, herds, and schools: A distributed
behavioral model, Computer Graphics, Vol. 21(4), pp. 2534, 1987.
[Roads, 1996] Roads, C., The Computer Music Tutorial, pp. 470-472, MIT Press,
1996.
[Roads, 2004] Roads, C., Microsound, MIT Press, 2004.
[Ross, 2009] Ross, A., The Rest is Noise, Farrar, Straus and Giroux, 2007.

236

[Rumsey, 1998], Rumsey, F. Subjective assessment of the spatial attributes of

reproduced sound," Proceedings of the Audio Engineering Society 15th International
Conference, pp. 122-135, 1998.
[Rumsey, 2001] Rumsey, F., Spatial Audio, pp. 10-12, pp. 21-39, pp. 52-60, pp.
111-116, pp. 204-208, Focal Press, 2001.
[Saberi et al, 1990] Saberi, K. and Perrott, D. R., Minimum audible movement angles
as a function of sound source trajectory Journal of the Acoustical Society of
America, Vol. 88(6), 1990.
[Sanson et al, 2008] Sanson, J., Corteel, E. and Warusfel, O., Objective and
subjective analysis of localization accuracy in Wave Field Synthesis, Proceedings of
the 124th Convention of the Audio Engineering Society, Preprint 7361, 2008.
[Schacher et al, 2006] Schacher, J. C. and Kocher, P. Ambisonics Spatialization
Tools for Max/MSP, Proceedings of the International Conference on Computer
Music, 2006.
[Schroeder, 1962] Schroeder, M. R., "Natural Sounding Artificial Reverberation",
Journal of the Audio Engineering Society, Vol. 10, pp. 219-223, 1962.
[Schroeder et al, 1974] Schroeder, M. R., Gottlob, D. and Siebrasse, K. F.,
Comparative study of European concert halls: Correlation of subjective preference
with geometric and acoustic parameters, Journal of the Acoustical Society of
America, Vol. 56, pp. 1195-1201, 1974
[Sheeline, 1983], Sheeline, C. W., An investigation of the effects of direct and
reverberant signal interaction on auditory distance perception, Doctoral Thesis,
Stanford University, 1983.
[Shinn-Cunningham, 2003] Shinn-Cunningham, B. G., "Spatial hearing advantages
in everyday environments", Proceedings of ONR workshop on Attention, Perception,
and Modelling for Complex Displays, 2003.

237

[Simms, 1986] Simms, B. R., Music of the Twentieth Century: Style and Structure,
Schirmer Books, 1986.
[Singer, 2008] Singer, E., Boids for Max MSP/Jitter, http://www. ericsinger. com,
retrieved on 07/07/2008.
[Smalley, 1996] Smalley, D., The Listening Imagination: Listening in the Electroacoustic Era, Contemporary Music Review, Vol. 13(2), pp. 77-107, Harwood
Academic Publishers, 1996.
[Smalley, 1997] Smalley, D., Spectromorphology : Explaining Sound-Shapes,
Organised Sound, Vol. 2(2), pp. 107-126, Cambridge University Press, 1997.
[Smalley, 2007] Smalley, D., Space-Form and the Acousmatic Image,
Organised Sound, Vol. 12(1), pp. 35-58, Cambridge University Press, 2007.
[Smalley J., 2000] Smalley, J. Gesang der Jnglinge: History and Analysis,
Masterpieces of 20th-Century Electronic Music: A Multimedia Perspective, Columbia
University Computer Music Centre,
www.music.columbia.edu/masterpieces/notes/stockhausen, retrieved 02/07/2009.
[Smalley et al, 2000] Smalley, D. and Austin, L., Sound Diffusion in Composition
and Performance: An Interview with Denis Smalley, Computer Music Journal, Vol.
24(2), pp. 10-21, MIT Press, 2000.
[Smith, 1985] Smith III, J. O., A new approach to digital reverberation using closed
waveguide networks, Proceedings of the International Computer Music Conference,
pp. 4753, 1985.
[Sonke, 2000] Sonke, J. J., "Variable Acoustics by wave field synthesis", Doctoral
Thesis, Delft University of Technology, 2000.

238

[Sonoloco, 2007], Stockhausen Edition no. 91 (Cosmic Pulses), Sonoloco Record

Reviews, http://home.swipnet.se/sonoloco25/stockhausen/91.html, retrieved
02/07/2009.
[Spikofski, 2001] Spikofski, G. and Fruhmann, M., Optimisation of Binaural Room
Scanning (BRS): Considering inter-individual HRTF-characteristics, Proceedings of
the 19th International Conference of the Audio Engineering Society, June 2001.
[Spors et al, 2003] Spors, S., Kuntz, A. and Rabenstein, R., An Approach to
Listening Room Compensation with Wave Field Synthesis", Proceedings of the 24th
International Conference of the Audio Engineering Society, 2003.
[Start, 1997] Start,E., "Direct Sound Enhancement by Wave Field Synthesis",
Doctoral Thesis, Delft University of Technology, 1997.
[Steinberg et al, 1934] Steinberg, J. and Snow, W., Auditory Perspective Physical
Factors, Electrical Engineering, Vol. 53(1), 1934, Reprinted in An Anthology of
articles on stereophonic techniques, Audio Engineering Society, 1986.
[Stockhausen, 1959] Stockhausen, K., ". . . How Time Passes . . , English translation
by Cornelius Cardew, Die Reihe, Vol. 3, pp. 10-40, Universal Edition Publishing,
1959.
[Stockhausen, 1962] Stockhausen, K., "The Concept of Unity in Electronic Music,
English translation by Elaine Barkin, Perspectives of New Music, Vol. 1(1), pp. 3948, 1962.
[Stockhausen, 1975a] Stockhausen, K.,Electronic and Instrumental Music, English
translation by Ruth Koenig, Die Reihe, Vol. 5, pp. 59-66, Universal Edition
Publishing, 1975.
[Stockhausen, 1975b] Stockhausen, K.,Music in Space, die Reihe, Vol. 5, pp. 6782, English translation by Ruth Koenig, Universal Edition Publishing, 1975.

239

[Theile et al, 1976], Theile, G. and Plenge, G., "Localization of lateral phantomsources". Journal of the Audio Engineering Society, Vol. 25, pp, 196-200, 1976.
[Theile, 1980] Theile, G., On the Localisation in the Superimposed Soundfield,
Doctoral Thesis, Berlin University of Technology, 1980.
[Theile, 1991] Theile, G. On the Naturalness of Two-Channel Stereo Sound,
Journal of the Audio Engineering Society, Vol. 39, 1991.
[Theile et al, 2003] Theile, G., Wittek, H. and Reisinger, M., Potential Wave Field
Synthesis Applications in the Multichannel Stereophonic World, Proceedings of the
24th International Conference of the Audio Engineering Society, 2003.
[Trochimczyk, 2001] Trochimczyk, M., From Circles to Nets: On the Signification
of Spatial Sound Imagery in New Music, Computer Music Journal, Vol. 25(4), pp.
39-56, MIT Press, 2001.
[Truax, 1990] Truax, B., "Composing with Real-Time Granular Sound", Perspectives
of New Music, Vol. 28(2), 1990.
[Truax, 1996] Truax, B., "Soundscape, Acoustic Communication & Environmental
Sound Composition", Contemporary Music Review, Vol. 15(1), pp. 49-65, 1996.
[Truax, 1999] Truax, B., "Composition and diffusion: space in sound in space",
Organised Sound, Vol. 3(2), pp. 141-6, 1999.
[Truax, 2008] Truax, B., Soundscape Composition as Global Music: Electroacoustic
Music as Soundscape, Organised Sound, Vol. 13(2), pp. 103-109, 2008.
[Unemi et al, 2005] Unemi, T. and Bisig, D., Music by interaction among two
flocking species and human, Proceedings of the 3rd International Conference on
Generative Systems in Electronic Arts, pp. 171179, 2005.
[Uozumi, 2007] Uozumi, Y., Gismo2: An application for agent-based composition,
Proceedings of the 2007 EvoWorkshops, pp. 609616, 2007.
240

[Usher et al, 2004] Usher, J., Martens, W. L. and Woszczyk, W., The influence of
the presence of multiple sources on auditory spatial imagery as indicated by a
graphical response technique, Proceedings of the 18th International Congress on
Acoustics (ICA 2004), 2004.
[Varse, 1998] Varse, E., "The Liberation of Sound", Contemporary Composers on
Contemporary Music, pp. 197, Perseus Publishing, 1998.
[Verheijen, 1998] Verheijen, E., "Sound reproduction by Wave Field Synthesis",
Doctoral Thesis, Delft University of Technology, 1998.
[Vogel, 1993] Vogel, P., "Application of wave field synthesis in room acoustics",
Doctoral Thesis, Delft University of Technology, 1993.
[Wallach et al, 1949] Wallach, H., Newman, E. B. and Rosenzweig, M. R., The
precedence effect in sound localization, American Journal of Psychology, Vol. 57,
pp. 315-336, 1949.
[Wessel et al, 2002] Wessel, M., Wright, D. and Schott, J., Situated Trio - an
interactive live performance for a hexaphonic guitarist and two computer musicians,
Proceedings of the 2002 Conference on New Instruments for Musical Expression
(NIME-02), pp. 171-173, 2002.
[Wessel, 2006] Wessel, D., An Enactive Approach to Computer Music
Performance, Le Feedback dans la Creation Musical, pp. 93-98, 2006.
[White, 2008] White, N. S., The Rosen - a pretty thing, but is it a serious
autoharp?, Autoharp Notes, www.ukautoharps.org.uk, retrieved on 01/05/2009.
[Wiggins, 2004] Wiggins, B., An Investigation into the Real-time Manipulation and
Control of Three-dimensional Sound Fields, Doctoral Thesis; University of Derby,
2004.
[Wishart, 1985] Wishart, T., On Sonic Art, Imagineering Press, 1985.
241

[Wittek, 2002] Wittek, H., Optimised Phantom Source Imaging (OPSI) of the high
frequency content of virtual sources in Wave Field Synthesis,
www.hauptmikrofon.de/wfs.html, retrieved 03/06/2009.
[Wittek, 2003] Wittek, H., Perception of Spatially Synthesized Sound Fields, 2003,
www.hauptmikrofon.de/wfs.html, retrieved 03/06/2009.
[Wittek, 2007] Wittek, H., Perceptual differences between wavefield synthesis and
stereophony, Doctoral Thesis, University of Surrey, 2007.
[Xenakis, 2008] Xenakis, I., Music and Architecture, Architectural Projects, Texts,
and Realizations Compiled, translated and commented by Sharon Kanach, pp. 146,
Pendragon, 2008.
[Yadegari et al, 2002] Yadegari, S., Moore, F. R., Castle, H., Burr, A. and Apel, T.,
Real-Time Implementation of a General Model for Spatial Processing of Sounds,
Proceedings of the International Computer Music Conference, pp. 244-247, 2002.
[Yeung, 2004] Yeung, B., Guitar dreams, an interview with Adrian Freed,
www.sfweekly.com, 2004, retrieved on 01/04/2007.
[Zacharov, 1998] Zacharov, N., "Subjective Appraisal of Loudspeaker Directivity for
Multichannel Reproduction", Journal of the Audio Engineering Society, Vol. 46(4),
1998.
[Zvonar, 2004] Zvonar, R., A History of Spatial Music, eContact, Vol. 7(4),
Canadian Electroacoustic Community, 2004.

242

Appendix A Scores

243