Lecture Notes in Electrical Engineering

Volume 145
Abbas Mohammadi and Fadhel M. Ghannouchi

RF Transceiver Design
for MIMO Wireless
Communications

ABC
Authors
Prof. Abbas Mohammadi Prof. Fadhel M. Ghannouchi
Amirkabir University University of Calgary
Electrical Engineering Department Electrical and Computer Engineering
Tehran Intelligent RF Radio Laboratory
Iran Alberta
Canada

ISSN 1876-1100 e-ISSN 1876-1119

ISBN 978-3-642-27634-7 e-ISBN 978-3-642-27635-4
DOI 10.1007/978-3-642-27635-4
Springer Heidelberg New York Dordrecht London
Library of Congress Control Number: 2011945157

c Springer-Verlag Berlin Heidelberg 2012

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed. Exempted from this legal reservation are brief excerpts in connection
with reviews or scholarly analysis or material supplied specifically for the purpose of being entered
and executed on a computer system, for exclusive use by the purchaser of the work. Duplication of
this publication or parts thereof is permitted only under the provisions of the Copyright Law of the
Publisher’s location, in its current version, and permission for use must always be obtained from Springer.
Permissions for use may be obtained through RightsLink at the Copyright Clearance Center. Violations
are liable to prosecution under the respective Copyright Law.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
While the advice and information in this book are believed to be true and accurate at the date of pub-
lication, neither the authors nor the editors nor the publisher can accept any legal responsibility for any
errors or omissions that may be made. The publisher makes no warranty, express or implied, with respect
to the material contained herein.

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)
Preface

The multiple-input multiple-output (MIMO) technique provides higher bit rates

and better reliability in wireless systems. The efficient design of RF transceivers
has a vital impact on the implementation of this technique. This first book is com-
pletely devoted to RF transceiver design for MIMO communications. The book
covers the most recent research in practical design and applications and can be
an important resource for graduate students, wireless designers, and practical
engineers.
The book opens with an introduction to MIMO wireless communications,
where the main advantages of using MIMO technique are described. This is fol-
lowed by a discussion on the implementation techniques for MIMO modulators.
After describing the fundamental concepts for RF transceivers and power amplifi-
ers, the design and analysis methods for the RF section of MIMO transmitters and
receivers are presented. Furthermore, the RF impairments in MIMO and OFDM
systems, including nonlinearity, phase noise, I/Q imbalance and DC offset, are
discussed; and, their compensation methods are presented. Finally, the design
techniques for single RF front-end MIMO systems are described.

Audience
The book can be used by graduate students, researchers and design engineers in
microwave and wireless design areas. It is assumed that the reader has a funda-
mental knowledge of communication circuit design and communication systems
theory. The book may be used as a textbook for a graduate course on wireless
transceiver design techniques.

Acknowledgments
We would like to acknowledge the Informatics Circle of Research Excellence
(iCORE), Alberta, Canada, for their financial contribution to support the visiting
professor stay of the first author at iRadio Lab, University of Calgary, which
helped the completion of this book. We are grateful to our great students and re-
searchers in Microwave/mm-wave and Wireless Communications Lab of
Amirkabir University and iRadio Lab of the University of Calgary. This book
could not have been completed without their fruitful research. Their useful com-
ments, discussions, collaborations and help in producing many of the results pre-
sented in this book over the years are appreciated. In particular, we would like to
VI Preface

thank A. Abdipour, M. Heloui, R. Mirzavand, S.A. Bassam, M.R. Keshavarzi,

M.H. Madani, and I. D'Adamo. We also gratefully appreciate the careful and thor-
ough editing by Ms. Cathie Heys.
Finally, we would like to profoundly thank our spouses for their understanding
and patience throughout the many evenings and weekends taken to prepare this
book, as well as our parents for their encouragement and valuable support in our
early professional years as graduate students and young researchers.
Contents

Chapter 1: Introduction ..................................................................................... 1

1.1 Multiantenna Wireless Communications ................................................... 1
1.2 MIMO Wireless Transceivers.................................................................... 3
1.3 MIMO Techniques in Commercial Wireless Systems............................... 4
1.4 Organization of the Book and Future Challenges ...................................... 6
References ......................................................................................................... 7

Chapter 2: MIMO Wireless Communications ................................................. 9

2.1 MIMO System ........................................................................................... 9
2.2 MIMO Channel........................................................................................ 10
2.2.1 SISO Channel Model .................................................................... 10
2.2.2 MIMO Channel Modeling ............................................................ 11
2.3 MIMO Capacity....................................................................................... 12
2.3.1 SISO Capacity .............................................................................. 12
2.3.2 MIMO Capacity............................................................................ 13
2.4 MIMO Design Advantages ...................................................................... 17
2.4.1 Space-Time Codes for Diversity................................................... 17
2.4.2 Spatial Multiplexing ..................................................................... 19
2.4.3 Diversity-Multiplexing Tradeoff................................................... 21
2.5 MIMO Channel Models........................................................................... 22
2.5.1 Stanford University Interim Channel Models ............................... 23
References ....................................................................................................... 25

Chapter 3: Digital Modulation Techniques in MIMO Systems.................... 27

3.1 Criteria to Design Digital Modulators ..................................................... 27
3.1.1 Spectral Efficiency........................................................................ 28
3.1.2 Power Efficiency........................................................................... 29
3.1.3 Implementation Complexity ......................................................... 30
3.1.4 Performance Study in Wireless Channel ...................................... 30
3.2 Single-Carrier SISO Systems................................................................... 32
3.2.1 Single-Carrier Modulation ............................................................ 32
3.2.2 MQAM ......................................................................................... 33
3.2.2.1 MQAM Spectral Efficiency ............................................ 34
3.2.2.2 MQAM Power Efficiency Performance.......................... 37
3.2.2.3 MQAM Modulator Implementation................................ 38
3.2.3 MPSK ........................................................................................... 39
3.3 Single-Carrier MIMO Systems ................................................................ 41
3.3.1 QAM-MIMO ................................................................................ 42
VIII Contents

3.3.2 MPSK-MIMO............................................................................... 43
3.4 Multicarrier SISO Modulation Systems................................................... 44
3.5 OFDM Modulation Systems .................................................................... 45
3.5.1 OFDM Implementation Using IFFT/FFT ..................................... 47
3.5.2 Cyclic Prefix in OFDM Modulation ............................................. 47
3.5.3 RF Impairment in OFDM Systems ............................................... 48
3.5.3.1 Peak-to-Average Power Ratio in OFDM Modulation..... 48
3.5.3.2 Phase Noise in OFDM Modulation................................. 50
3.6 MIMO OFDM Systems ........................................................................... 52
References ....................................................................................................... 53

Chapter 4: Fundamental Concepts in Wireless Transceiver Design............ 55

4.1 Linear Systems and Transformations....................................................... 55
4.2 Nonlinear Systems ................................................................................... 56
4.2.1 One-dB Compression Point .......................................................... 57
4.2.2 Third-Order Intercept Point .......................................................... 58
4.2.3 Second-Order Intercept Point ....................................................... 60
4.2.4 Cross Modulation.......................................................................... 60
4.2.5 AM/AM and AM/PM Distortion .................................................. 61
4.2.6 Spectral Regrowth......................................................................... 62
4.3 Noise........................................................................................................ 63
4.3.1 Noise Factor of Cascaded Networks ............................................. 64
4.3.2 Capacity Reduction due to Noise Factor of an RF Receiver......... 66
4.4 RF Systems Design Parameters ............................................................... 67
4.4.1 Path Loss....................................................................................... 67
4.4.2 Link Budget .................................................................................. 67
4.4.3 Receiver Sensitivity ...................................................................... 67
4.4.4 Dynamic Range............................................................................. 68
4.4.5 ADC Selection .............................................................................. 69
4.4.6 Phase Noise................................................................................... 70
4.5 Modulation Accuracy .............................................................................. 71
4.5.1 Error Vector Magnitude (EVM) ................................................... 71
4.5.2 Effect of I/Q Imbalance on EVM.................................................. 71
4.5.3 Effect of Phase Noise on EVM ..................................................... 72
4.5.4 Effect of Nonlinearity on EVM .................................................... 73
4.5.5 Effect of Carrier Leakage on EVM............................................... 73
4.5.6 Total EVM .................................................................................... 74
References ....................................................................................................... 75

Chapter 5: RF Power Amplifier and Linearization Techniques .................. 77

5.1 Introduction ............................................................................................. 77
5.2 Transmitter Systems Parameters.............................................................. 78
5.2.1 Gain .............................................................................................. 78
5.2.2 Bandwidth..................................................................................... 78
5.2.3 Noise Figure.................................................................................. 78
5.2.4 Power Efficiency........................................................................... 78
Contents IX

5.2.4.1 Total Efficiency .............................................................. 79

5.2.4.2 Drain Efficiency.............................................................. 79
5.2.4.3 Power-Added Efficiency................................................. 80
5.2.5 P1dB ............................................................................................. 80
5.2.6 IP3................................................................................................. 81
5.2.7 PAPR ............................................................................................ 82
5.2.8 Power Back-Off ............................................................................ 82
5.2.9 ACPR............................................................................................ 82
5.2.10 EVM ........................................................................................... 83
5.2.11 Memory Effect............................................................................ 86
5.3 RF Power Amplifiers ............................................................................... 86
5.3.1 Linear PAs (Classes A, AB, B, C) ................................................ 87
5.3.1.1 Class A PA...................................................................... 90
5.3.1.2 Class B PA ...................................................................... 92
5.3.1.3 Class AB PA ................................................................... 95
5.3.1.4 Class C PA ...................................................................... 95
5.3.2 Switching-Mode PAs (Classes D, E, F) ........................................ 98
5.3.2.1 Class D PA.................................................................... 100
5.3.2.2 Class E PA .................................................................... 104
5.3.2.3 Class F and Inverse Class F PAs................................... 105
5.3.3 Comparison of the Classes of Operation in RF Pas .................... 108
5.4 Linearization of RF Power Amplifiers .................................................. 108
5.4.1 Feedback Linearization............................................................... 109
5.4.2 Feedforward Linearization.......................................................... 111
5.4.3 Digital Predistortion Linearization.............................................. 113
5.4.3.1 Look-Up-Table Model .................................................. 115
5.4.3.2 Memory Polynomial Model .......................................... 115
5.4.4 Analog Predistortion Linearization............................................. 116
5.4.5 Comparison of Linearization Techniques ................................... 119
5.5 RF Transmitter Architectures ................................................................ 120
5.5.1 Polar Transmitter Architecture ................................................... 120
5.5.2 LINC Transmitter Architecture................................................... 121
5.5.3 EER Transmitter Architecture .................................................... 124
5.5.4 Delta-Sigma Transmitter Architecture........................................ 125
References ..................................................................................................... 127

Chapter 6: Transmitter Design for MIMO Wireless Communications..... 129

6.1 Complexity and Cost in MIMO Systems............................................... 129
6.2 Transmitters Architectures..................................................................... 130
6.2.1 Direct Conversion Transmitter ................................................... 131
6.2.2 Superhetrodyne Transmitter........................................................ 132
6.3 Brief Overview of MIMO Transmission Schemes ................................ 133
6.3.1 FDM Technique.......................................................................... 133
6.3.2 TDM Technique.......................................................................... 133
6.3.3 CDM Technique ......................................................................... 133
6.4 MIMO Transceiver Architectures.......................................................... 134
X Contents

6.4.1 Antenna Selection Architecture .................................................. 134

6.4.1.1 Maximum Capacity Criterion ....................................... 135
6.4.1.2 Maximum Minimum Singular Value Criterion............. 136
6.4.1.3 Norm-Based Selection (NBS) Criterion........................ 136
6.4.2 Frequency-Division Multiplexing (FDM) Architecture.............. 136
6.4.3 Time-Division Multiplexing (TDM) Architecture...................... 136
6.4.4 Code-Division Multiplexing (CDM) Architecture...................... 139
6.5 Distortion and Impairment Compensation in MIMO Transmitters ....... 141
6.5.1 Antenna Crosstalk....................................................................... 142
6.5.2 Nonlinear RF Crosstalk .............................................................. 143
6.5.3 Effects of Nonlinear Crosstalk on DPD Extraction .................... 144
6.5.4 Impairment and Distortion Compensation .................................. 145
References ..................................................................................................... 148

Chapter 7: Receiver Design for MIMO Wireless Communications........... 151

7.1 Receiver Architectures........................................................................... 151
7.1.1 Superheterodyne Receiver .......................................................... 151
7.1.2 Direct Conversion Receiver ........................................................ 152
7.1.3 Low IF Receiver ......................................................................... 152
7.1.4 Image Rejection Receiver ........................................................... 153
7.1.5 Six-Port Receiver ........................................................................ 154
7.1.6 Five-Port Receiver ...................................................................... 156
7.2 Smart Antenna Receiver Architectures.................................................. 157
7.3 MIMO Receiver Architectures .............................................................. 161
7.3.1 Superheterodyne MIMO Front-End............................................ 162
7.3.2 Direct Conversion MIMO Front-End ......................................... 163
7.3.3 Six-Port MIMO Front-End ......................................................... 164
7.4 Capacity Reduction of MIMO System due to the Front-End................. 164
7.5 Radio Frequency Interference on MIMO Receivers.............................. 166
7.6 MIMO Testbed Design .......................................................................... 168
7.6.1 Testbed Architecture................................................................... 168
7.6.1.1 RF Transmitter .............................................................. 170
7.6.1.2 RF Receiver .................................................................. 171
7.7 Commercial MIMO Transceivers .......................................................... 173
7.7.1 UXA234 MIMO Transceiver from NXP-Philips........................ 173
7.7.2 MAX2842 MIMO Transceiver from MAXIM ........................... 174
7.7.3 PM8800 MIMO Transceiver from PMC-Seirra.......................... 175
7.7.4 AD9356 MIMO Transceiver from Analog Devices ................... 176
References ..................................................................................................... 178

Chapter 8: RF Impairments in OFDM Transceivers .................................. 181

8.1 OFDM Transceivers .............................................................................. 181
8.2 Noise in OFDM Transceivers ................................................................ 182
8.2.1 Phase and Amplitude Noise ........................................................ 183
8.2.2 Analysis of Phase and Amplitude Noise Impacts in OFDM....... 185
8.2.3 Impacts of Phase and Amplitude Noise on OFDM Systems....... 189
Contents XI

8.3 Nonlinearity in OFDM Transceivers ..................................................... 194

8.3.1 Analysis of Nonlinear Circuit Impact in OFDM......................... 196
8.4 Concurrent Analysis of Nonlinearity and Phase Noise in OFDM
Transceivers........................................................................................... 202
8.4.1 Theoretical Analysis ................................................................... 203
8.4.2 Performance Analysis ................................................................. 208
8.4.3 Fifth-Order Nonlinear Circuit and Phase Noise Modeling ......... 212
8.5 OFDM Signal with Phase Noise Passing through Dynamic Nonlinear
Circuits................................................................................................... 218
8.5.1 Dynamic Nonlinear Circuits ....................................................... 218
8.5.2 Modeling and Analysis ............................................................... 219
References ..................................................................................................... 224

Chapter 9: RF Impairments in MIMO Transceivers .................................. 227

9.1 Phase Noise in MIMO Transceivers ...................................................... 227
9.1.1 Phase Noise Model ..................................................................... 227
SISO Case Study......................................................................... 229
9.1.2 Impact of Phase Noise on MIMO Systems ................................. 230
MIMO Case Study ...................................................................... 233
9.1.3 Adaptive Modulation MIMO System ......................................... 233
9.1.4 BER of Adaptive Modulation MIMO System with Phase
Noise........................................................................................... 236
Case Study: Adaptive Modulation MIMO.................................. 237
9.2 DC Offset in MIMO Transceivers ......................................................... 239
9.2.1 DC Offset.................................................................................... 240
9.2.2 BER OF MQAM Modulation under Impact of DC Offset ......... 241
9.2.3 MIMO System Model................................................................. 243
9.2.4 BER of Adaptive Modulation MIMO under the Impact of
DC Offset.................................................................................... 244
9.2.5 BER Upper Bound of Adaptive Modulation under the Impact
of DC Offset ............................................................................... 245
9.2.6 Throughput Analysis................................................................... 247
9.3 I/Q Imbalance in MIMO Transceivers................................................... 254
9.3.1 I/Q Imbalance Model .................................................................. 254
9.3.2 MIMO System Model................................................................. 255
9.3.3 Impact of I/Q Imbalance on BER of Adaptive Modulation
MIMO......................................................................................... 257
9.3.4 I/Q Imbalance Compensation in Adaptive Modulation
MIMO Systems........................................................................... 258
9.3.5 BER Analysis.............................................................................. 260
References ..................................................................................................... 262

Chapter 10: Single RF Front-End MIMO Transceivers ............................. 265

10.1 RF Front-End MIMO Using Antenna Selection .................................. 266
10.2 Single RF Front-End MIMO Using FDM............................................ 268
10.3 Single RF Front-End MIMO Using TDM ........................................... 269
XII Contents

10.3.1 Input Signal............................................................................. 271

10.3.2 Switch Model.......................................................................... 271
10.3.3 Time Sampling of Received Signals ....................................... 272
10.3.4 Baseband Processing............................................................... 274
10.3.5 Experimental Studies .............................................................. 275
10.4 Single RF Front-End MIMO Using CDM ........................................... 279
10.5 Single RF Front-End MIMO Using a Parasitic Antenna ..................... 282
References ..................................................................................................... 287

Index .................................................................................................................. 289

Chapter 1
Introduction

Multiple input multiple output (MIMO) wireless systems provide many advantages
by using more than a single antenna. As such, it is considered in the current and
future wireless standards. However, radio frequency (RF) transceiver design for
MIMO wireless communications is a challenging task. This subject has been
attracting research attention in both academia and industry. This chapter provides a
general overview of MIMO systems and transceiver implementation.

1.1 Multiantenna Wireless Communications

Wireless communication systems are realized by using a transmitter, receiver and

wireless channel. The classic implementation of these systems is usually a
transmitter and its single antenna on the transmitting side and a receiver and its
single antenna on the receiving side. The concept of using multiple antennas at
receiver was introduced in the early 1960s [1]. The basic idea is the provision of
multiple copies from a transmitted signal on the receiving side and their
combination to obtain a signal with better performance. This technique is called
the space diversity technique. To realize the receiver diversity technique, multiple
antennas are employed on the receiver side. The antennas’ output signals are then
combined by applying different weighting coefficients, according to different
diversity techniques.
Figure 1.1 shows the general diagram of the receiver diversity method. The
receiver diversity may be realized using three different methods, selection
combining, equal gain combining, and maximal ratio combining. In the selection
combining method, the receiver selects the strongest signals among the antennas’
outputs. This signal is processed in the single receiver. Maximal ratio combining
is realized by weighting both the gain and phase of the received signals to
maximize the signal-to-noise ratio (SNR) at the output of the receiver.
On the other hand, in the equal gain combining method, only the phases of the
channel using weighting coefficients are cancelled. The maximum ratio combiner
diversity technique provides the best performance to mitigate the wireless channel
impact, while the selection diversity offers the simplest implementation technique
among the different diversity methods [1].

A. Mohammadi et al.: RF Transceiver Design for MIMO Wireless Communications, LNEE 145, pp. 1–8.
springerlink.com © Springer-Verlag Berlin Heidelberg 2012
2 Chapter 1 Introduction

Element − 1

W1*
Element − 2

∑
Output
W2*
Element − n

Wn*
Fig. 1.1 Diagram of the receiver diversity technique

The receiver diversity technique is a special implementation for the use of multiple
antennas. The multiple antenna technique can be used in both the transmitter and the
receiver. This technique is generally known as a multiple input, multiple output
(MIMO) system. The various realization techniques for MIMO systems are shown in
Figure 1.2. A system that uses a single antenna on both the transmitting and receiving
sides is called a single input single output (SISO) system.

Bits SISO SISO Bits

Tx Rx

Bits SIMO SIMO Bits

Tx # Rx

Bits MISO MISO Bits

Tx Rx

Bits MIMO MIMO Bits

#
Tx Rx

Fig. 1.2 Block diagram of multiple antennas systems

1.2 MIMO Wireless Transceivers 3

On the other hand, a system that employs a single antenna in the transmitter and
multiple antennas in the receiver is called a single input multiple output (SIMO)
system. Indeed, this system realizes the receiver diversity scenario. The idea of
exploring the transmitter diversity was introduced in 1990s [2]. By using a similar
notation, a system that uses multiple antennas at the transmitter and a single
antenna in the receiver is called a multiple input single output (MISO) system.
Likewise, the system that employs multiple antennas in the transmitter and
multiple antennas in the receiver edges is the MIMO system.
A MIMO system with transmitter antennas and receiver antennas is
shown in Figure 1.3. The input data is transmitted through antennas after
processing on the transmitter side. The processing includes channel coding,
modulation, space-time encoding, spatial mapping, and RF up-conversion.
Each antenna transmits a signal through a wireless channel. Accordingly,
all simultaneous radiators operate as a transmitter.

Fig. 1.3 General block diagram of MIMO systems

1.2 MIMO Wireless Transceivers

The art of designing modern wireless transceivers has seen extensive progress
during the last two decades. The design of these transceivers generally must
follow three main factors. They should be high performance, low cost, and able to
handle the complex objectives of advanced wireless communication systems. A
general block diagram of a wireless transceiver using multiple antennas in both the
transmitter and the receiver is shown in Figure 1.4.
In recent transceivers, a great amount of signal processing is achieved in the
baseband using digital signal processors (DSPs), application specific integrated
circuits (ASICs), and field-programmable gate arrays (FPGAs). A signal at the
output of the baseband section is up-converted to RF frequency using a heterodyne
technique or a direct conversion method. A frequency synthesizer provides the
carrier signal for up-conversion. The power amplifiers play a crucial role in
providing the performance of the system. The RF front-end delivers the signal to the
transmitter antennas. After passing through a wireless channel on the receiver side,
the signals of the multiple antennas must be down-converted. This process includes
RF signal processing using low noise amplifiers (LNAs), frequency synthesizing,
and down conversion. Signals must then be delivered to the receiver baseband.
4 Chapter 1 Introduction

The excessive signal processing is performed in the receiver to detect the

signals. Likewise, on the transmitter side, digital signal processing is carried by
receiver digital circuits.

Bits baseband Modulation &

RF @ Tx
Space − Time coding @ Tx

Matching
Synthesizer Duplexer

Bits baseband Demodulation &

RF @ Rx
Space − Time decoding @ Rx

Fig. 1.4 Block diagram for MIMO wireless transceivers

1.3 MIMO Techniques in Commercial Wireless Systems

In a fading environment, the performance of a wireless link can be greatly improved
by using multiple antennas on both the transmitting and receiving sides. These benefits
include increased reliability as well as high data rates. A MIMO wireless
communication system can be designed to take advantage of reliability improvement
or increasing the data rate. The first improvement is called a diversity order
improvement, and the latter is called spatial multiplexing. However, it has been shown
that both types of gains can be simultaneously obtained using a fundamental tradeoff
between them [4-5]. The application of the MIMO technique in wireless
communications is aimed for both spatial multiplexing and diversity improvement.
Most wireless standards have been reconsidered regarding the usage of MIMO
systems. In the case of wireless networking standards (IEEE 802.11x), the IEEE
802.11n amendment introduces MIMO in the physical layer of the network to
improve the network throughput over previous standards. This increases the
maximum data rate from 54 Mbps for the IEEE 802.11a/g standards to 600 Mbps
for the IEEE 802.11n standard, using a 40 MHz bandwidth with MIMO capability.
The MIMO technique is used in IEEE 802.11n wireless local area network (LAN)
to provide spatial multiplexing.
The increased throughput comes from a mixture of changes to the way data
packets are sent, along with the use of sophisticated radio techniques that demand
high performance from the RF hardware. Multiple RF channels may be
implemented with integrated transceivers and the same local oscillator (LO) with a
separate front-end module. Figure 1.5 shows the main components for the IEEE
802.11n wireless LAN [3].
In addition, the World Interoperability for Microwave Access (WiMAX)
standard also supports MIMO for both fixed and mobile WiMAX (IEEE 802.16e
and IEEE 802.16m). Its MIMO feature is intended for high-speed applications,
and it covers both space-time and spatial multiplexing codes [10]. The functional
block diagram of a WiMAX IEEE802.16e from NXP Semiconductors (Philips) is
shown in Figure 1.6 [11]. By using two antennas, this system provides a better
performance for a wireless system.
1.3 MIMO Techniques in Commercial Wireless Systems 5

The MIMO technique is also considered in the 3G (third generation) Long

Term Evolution (LTE) standard, in order to increase the data rate to up to 277
Mbps in the downlink for a single user in an ideal radio link condition using a 4x4
MIMO system with a 20 MHz bandwidth [6]. Figure 1.7 shows a simplified block
diagram of a LTE transceiver using two-channel reception [9].

QAM Modulation

STBC (optional )
IFFT RFChain
Data

&Interleaving
Spatial
DeMux

Encoder
Mapping

IFFT RFChain

Fig. 1.5 Simplified block diagram of WLAN IEEE802.11n using MIMO

Fig. 1.6 Block diagram of WiMAX RF transceiver IEEE802.16e using MIMO from NXP
[11]
6 Chapter 1 Introduction

Fig. 1.7 Simplified transceiver block diagram of LTE [9]

1.4 Organization of the Book and Future Challenges

The book addresses some of the key design techniques for RF transceivers of
MIMO wireless systems. Chapter 2 presents the fundamentals for understanding
the multiple-antenna theory and advantages. This begins by presenting the MIMO
system model, and MIMO channel models are the briefly discussed. This is
followed by an examination of the MIMO capacity concept. The diversity gain
and spatial multiplexing concepts are also introduced.
Chapter 3 deals with digital modulation techniques and orthogonal frequency-
division multiplexing (OFDM) systems. After introducing the power efficiency and
bandwidth efficiency concepts in digital modulation, the impact of fading channels
on these modulators are discussed. The MIMO-OFDM technique and its
implementation process are also discussed in this chapter. Chapter 4 provides a
comprehensive overview of the fundamental concepts in wireless transceivers
design. Chapter 5 covers the different subjects on RF power amplifiers and
linearization techniques. The various power amplifier design techniques, including
linear, high-efficiency and broadband techniques, are described. This is followed by
a discussion on power amplifier linearization techniques. RF transceiver design for
MIMO wireless communications is an active research area where more suitable
architecture of MIMO transceiver is still an ongoing research topic [12-14].
Chapter 6 deals with transmitter design issues in MIMO wireless transceivers.
The various transmitter architectures, such as heterodyne and direct conversion,
are investigated. This is followed by the introduction of the transmitter design
technique for MIMO applications. Chapter 7 presents the receiver design
procedure for wireless MIMO communications. This chapter includes the
commercial receiver design techniques for MIMO wireless communications. RF
References 7

impairments in MIMO systems and their compensation techniques have been

attracting the research attention in both academia and industry [15-18].
Chapter 8 discusses RF impairments and compensation techniques for OFDM
wireless transceivers. The impairments include phase noise and power amplifier
nonlinearities. The impacts of the impairments are also examined in this chapter,
and the compensation techniques are presented. Chapter 9 discusses RF
impairments and compensation techniques for MIMO wireless transceivers. The
impairments include phase noise, DC offset, I/Q imbalance, and nonlinearities.
The last chapter addresses the different alternatives in designing single RF front-
end MIMO transceivers. The cost and complexity of the RF transceiver of a
MIMO system increases linearly with an increasing number of antennas. The
simple RF front-end design for MIMO systems is actively researched in both
academia and industry [18-24].

References
[1] Janaswamy, R.: Radiowave Propagation and Smart Antennas for Wireless
Communications. Kluwer Academic Publishers (2001)
[2] Paulraj, A., Nabar, R., Gore, D.: Introduction to Space-Time Wireless
Communications. Cambridge University Press (2003)
[3] Agilent Technologies, MIMO Wireless LAN PHY Layer RF Operation &
Measurement. Agilent Application Note #1509 (April 2008)
[4] Biglieri, E., Calderbank, R., Goldsmith, A., Paulraj, A., Vincentpoor, H.: MIMO
Wireless Communications. Cambridge University Press (2007)
[5] Ebrahimzad, H., Mohammadi, A.: Diversity-Multiplexing Tradeoff in MIMO
Systems with Finite SNR. In: European Conference on Wireless Technology,
Munich, pp. 146–149 (October 2007)
[6] Qualcomm Incorporated, LTE—A Well-Designed Mobile OFDMA IP Solution,
Qualcomm Incorporated Report (January 2008)
[7] Tsoulos, G.: MIMO System Technology for Wireless Communications. CRC Press
(2006)
[8] Kaiser, T., Bourdoux, A., Boche, H., Fonollosa, J.R., Andersen, J.B., Utschick, W.:
Smart Antennas—State of the Art. Hindawi Publishing Corporation (2005)
[9] Rumbey, M.: LTE and the Evolution to 4G Wireless: Design and Measurement
Challenges. Wiley (2009)
[10] Hanzo, L., Akhtitman, Y., Wang, L., Jiamg, W.: MIMO-OFDM for LTE, WiFi and
Wimax. Wiley (2010)
[11] Locher, M., Tomesen, M., Kuenen, J., Daanen, A., Visser, H., Essink, B., Vervoort,
P.P., Nijrolder, M., Kopmeiners, R., Redman-White, W., Balmford, R., El Waffaoui,
R.: A Low Power, High Performance BiCMOS MIMO/Diversity Direct Conversion
Transceiver IC for WiBro/WiMAX (802.16e). In: IEEE 2007 Custom Intergrated
Circuits Conference, CICC (2007)
[12] Domizioli, C.P., Hughes, B.L., Gard, K.G., Lazzi, C.: Optimal front-end design for
MIMO receivers. In: IEEE Global Communications Conference, Globcom 2008,
New Orleans (2008)
[13] Rafati, H., Razavi, B.: Receiver Architecture for Dual-Antenna Systems. IEEE
Journal Solid-State Circuits 42(6), 1291–1299 (2007)
8 Chapter 1 Introduction

[14] Eickhoff, R., Kraemer, R., Santamaria, I., Gonzalez, L.: Developing energy-efficient
MIMO radios. IEEE Vehicular Technology Magazine 4(1), 34–41 (2009)
[15] Madani, M.H., Abdipour, A., Mohammadi, A.: Analysis of performance degradation
due to non-linearity and phase noise in orthogonal frequency division multiplexing
systems. IET Communications Journal 4(10) (2010)
[16] Keshavarzi, M.R., Mohammadi, A., Abdipour, A., Ghannouchi, F.M.:
Characterization and Compensation of DC Offset on Adaptive MIMO Direct
Conversion Transceivers. IEICE Transactions on Communications (January 2011)
[17] Schenk, T.: RF Imperfections in High-rate Wireless Systems. Springer, Heidelberg
(2008)
[18] Horlin, F., Bourdoux, A.: Digital compensation for analog front-ends: a new
approach to wireless transceiver design. Wiley (2008)
[19] Kalis, A., Kanatas, A.G., Papadias, C.B.: A novel approach to MIMO transmission
using a single RF front end. IEEE Journal of Selected Areas of Communication 26(6)
(August 2008)
[20] Duman, T.M., Ghrayeb, A.: Coding for MIMO Communication Systems. Wiley
(2008)
[21] Molisch, A., Win, M., Winters, J.: Reduced-complexity transmit/receive-diversity
systems. IEEE Transactions on Signal Processing 51(11), 2729–2738 (2003)
[22] Alrabadi, O.N., Papadias, C.B., Kalis, A., Prasad, R.: A Universal Encoding Scheme
for MIMO Transmission Using a Single Active Element for PSK Modulation
Schemes. IEEE Transactions on Wireless Communications 8(10), 5133–5143 (2009)
[23] Tzeng, F., Jahanian, A., Pi, D., Heydari, P.: A CMOS Code-Modulated Path-Sharing
Multi-Antenna Receiver Front-End. IEEE Journal of Solid-State Circuits 44(5),
1321–1335 (2009)
[24] Lari, M., Bassam, A., Mohammadi, A., Ghannouchi, F.M.: Time-Multiplexed Single
Front-End MIMO Receivers with Preserved Diversity Gain, vol. 5(6), pp. 789–796
(2011)
Chapter 2
MIMO Wireless Communications

The multiple input multiple output (MIMO) technique provides the higher bit rate
and the better reliability in wireless systems. These advantages are achieved by
designing appropriate apace-time codes that provide diversity improvement, spa-
tial multiplexing gain, or a trade-off between diversity order and spatial multiplex-
ing. This chapter provides an overoview on MIMO wireless system concept and
its performance. Moreover, the MIMO channel models are discussed.

2.1 MIMO System

A multiple input multiple out (MIMO) system with M T transmitting antennas and
M R receiving antennas is shown in Figure 2.1. The input data are transmitted
through M T antennas after processing on the transmitter side. The processing
includes channel coding, modulation, space-time encoding, spatial mapping, and
radio frequency (RF) up-conversion. Each antenna transmits a signal through a
wireless channel. Accordingly, the M T antennas simultaneously operate as an en-
tire transmitter. The radiated signals are represented by a column vector ( x ) that
has M T × 1 dimensions. These signals, after passing through the wireless channel,
are received by M R receiving antennas.

Coding Channel mapping

Data Sequence Data Sequence
Modulation Demodulation
Mapping Decoding

Fig. 2.1 Block diagram of a MIMO system

A. Mohammadi et al.: RF Transceiver Design for MIMO Wireless Communications, LNEE 145, pp. 9–26.
springerlink.com © Springer-Verlag Berlin Heidelberg 2012
10 Chapter 2 MIMO Wireless Communications

2.2 MIMO Channel

The impulse response of a linear time-varying communication system between a
transmitter and a receiver is used to describe the effects of a linear transmission
channel.

2.2.1 SISO Channel Model

In the first step, a narrowband system using a single antenna at both the transmit-
ter and the receiver is assumed (single input single out – SISO). The symbol pe-
riod is assumed to be T. Moreover, the digital signal in discrete time may be
represented by the complex time series {xk}.
In this case, the transmitted signal is represented by:

∞
x(t ) = ∑ Es xmδ (t − mT ) (2.1)
m =−∞

where E s is the transmitted symbol energy, assuming that the average energy
constellation is normalized to unity. In a linear time-invariant (LTI) system, a
function h(t) as the time-invariant impulse response of the channel can be consi-
dered [1]. If the signal x(t) is transmitted, the received signal r(t) is given by

r (t ) = h(t ) * x(t ) + n(t ) (2.2)

where * denotes the convolution product, and n(t) is the additive noise of the sys-
tem. Therefore, the input-output relation is represented as:

∞
r (t ) = ∑ Es xm h(t − mT ) + n(t ) (2.3)
m =−∞

One obtains the discrete representation of the received signal by sampling the re-
ceived signal at the rate of T, (r(kT)) as:

∞
r ( kT ) = ∑ Es xm h[( k − m)T ] + n( kT )
m =−∞
(2.4)
∞
r[ k ] = ∑ Es xm h[ k − m] + n[ k ]
m =−∞

As may be seen, a time-invariant channel can be represented as an LTI system and

its sampled representation h[k] . The extension of this representation to a time-
varying channel is completely straightforward [2], [4]. Moreover, there are
2.2 MIMO Channel 11

different methods to extract the channel impulse response both in narrowband and
broadband transmissions [3]. The channel impulse response generally depends on
attenuation of the path loss term, shadowing, and multipath fading.

2.2.2 MIMO Channel Modeling

In MIMO systems, both the transmitter and receiver have several antennas. A
MIMO system with M T transmitting antennas and M R receiving antennas is
shown in Figure 2.2. In this system, the channel between each transmitting-
receiving antenna pair can be modeled as a SISO channel. Accordingly, the chan-
nel matrix (H) for a MIMO system with MT transmitting antennas and MR receiv-
ing antennas may be obtained. By arranging all inputs and outputs in vectors,
x[k ] = [ x1, k ,..., xM T ,k ]T and r[k ] = [r1,k ,..., rM R ,k ]T , the input-output relationship at
any given time instant (k) is obtained as:

r[k ] = Es H[k ]x[k ] + n[k ] (2.5)

where H[k] is defined as the M R × M T MIMO channel matrix in sampling time

kT, and n[k ] = [n1,k ,..., nM R ,k ]T is the sampled noise vector, containing the noise
contribution at each receive antenna, such that the noise is white in both time and
spatial dimensions as:

E {n[k ]n[m]} = σ n2 I M R δ [k − m] (2.6)

where σ n2 is the Gaussian noise variance and I M R is the identity matrix. By nor-
malizing the transmit symbol energy to unity, the received signal is shown to be:

r = Hx + n (2.7)

where r is the received column vector signal with dimensions of M R × 1 com-

posed of the received signal, r j ; n is an M R × 1 column vector composed of the
noise components, n j ; and, H is the M R × M T channel matrix with jith com-
ponent being the channel coefficient, h j ,i . Accordingly, a MIMO system can be
represented using the following matrix equation:

⎡ r1 ⎤ ⎡ h1,1 " h1, M T ⎤ ⎡ x1 ⎤ ⎡ n1 ⎤

⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢ # ⎥=⎢ # % # ⎥⎢ # ⎥+⎢ # ⎥ (2.8)
⎢r ⎥ ⎢h ⎥ ⎢ x ⎥ ⎢n ⎥
⎣ M R ⎦ ⎣ M R ,1 " hM R , M T ⎦ ⎣ MT ⎦ ⎣ M R ⎦
12 Chapter 2 MIMO Wireless Communications

h1,1
hM R ,1

Space − Time h1, MT Space − Time

Encoder Decoder
hM R , MT

Fig. 2.2 MIMO channel

Example 2.1: Independent and Identically distributed Rayleigh MIMO Channel

If each individual channel coefficients is a zero-mean complex, circularly symme-
tric Gaussian variable or, equivalently, a complex variable whose amplitude and
phase are Rayleigh and uniformly distributed, respectively, it is called a wide-
sense stationary uncorrelated scattering homogeneous (WSSUSH) Rayleigh fad-
ing channel [4]. When the antenna spacing on both sides of the link is large
enough, the various channel correlations become very small and can be assumed
to be equal to zero [2]. The typical antenna spacing for negligible correlation is
about λ / 2 , where λ is the wavelength. Furthermore, if all individual channels
coefficients are characterized by the same average power, the channel matrix is
represented as Hw. This is a random fading matrix with unit variance and circular-
ly symmetric complex Gaussian entries. This channel, which uses the indepen-
dent and identically distributed (i.i.d.) assumption, is represented by as CN (0,1) .

2.3 MIMO Capacity

The channel capacity is a fundamental limit on the rate of error-free messages that
can be transmitted through a communication channel. In this section, the capacity
of SISO and MIMO channels under fading are discussed.

2.3.1 SISO Capacity

The channel capacity for SISO communication systems over additive white Gaus-
sian noise (AWGN) channels has been extracted by Shannon as:
S
C = B log 2 (1 + ) (2.9)
N
where C is the channel capacity in bit per second, B is the channel bandwidth in
Hz, S is the transmitted power, and N is the noise power in the channel bandwidth.
2.3 MIMO Capacity 13

Defining γ  S / N , the channel capacity for fading channels can be obtained as

[5], [6]:
∞
C = ∫ B log 2 (1 + γ ) pγ (γ )d γ (2.10)
0

where pγ (γ ) is the probability density function of fading channel.

2.3.2 MIMO Capacity

MIMO systems with multiple antennas on both the transmitting and receiving
sides have been considered. The capacity of MIMO systems can be expressed in
bits per second per hertz (bps/Hz) as [4], [7]:

γ
C = max{log 2 det( I M R + HRss H H )}
MT (2.11)
Tr ( Rss ) = M T

where I M R is the identity matrix, H is the channel matrix with HH being its trans-

pose conjugate, Rss = E{ss } is covariance matrix of the transmit signal, γ gives
H

the average signal-to-noise ratio (SNR) per receiver branch, and Tr (.) represents
the trace of a matrix. As can be seen, the channel capacity may be reached by
choosing the optimal covariance structure for the transmitted signals. If the chan-
nel is unknown on the transmitter side, the transmitted signal can be considered
spatially white, e.g., Rss = I MT . Accordingly, the MIMO channel capacity for a
sample deterministic realization is given by:

γ
C = log 2 det( I M R + HH H ) (2.12)
MT

If the eigendecomposition of HH H is represented as QΛQ , where Q is an

H

M R × M R matrix satisfying QQ H = Q H Q = I M R and Λ = diag{λ1 , λ2 ,..., λM R }

with λi ≥ 0 , equation (2.12) can be written as:

γ
C = log 2 det( I M R + Q ΛQ H ) (2.13)
MT
14 Chapter 2 MIMO Wireless Communications

By using the identity of det( I m + AB) = det( I n + BA) for Amn , Bnm and using
QQ H = Q H Q = I M R , the capacity relation may be written as:

γ
C = log 2 det( I M R + Λ) (2.14)
MT

This relation equivalently may be represented as:

r
γ
C = ∑ log 2 (1 + λi ) (2.15)
i =1 MT

where r is the rank of channel and λi ≥ 0 are the eigenvalues of HH H [4].

Equation (2.12) can be also used to obtain the MIMO capacity in the fading
channel. Extending equation (2.10) for an MIMO channel, the capacity of the
MIMO channel under fading can be represented by:

γ
C = E{log 2 det( I M R + HH H )} (2.16)
MT

where the average is over the distribution of the elements of H.

In order to provide a clear idea about MIMO capacity, the asymptotic MIMO
capacity should be examined. Assuming the spatially white Guassian channel (
CN (0,1) ) and using the law of large numbers considering large
M ( M T = M R = M ) [14], [7] yields:

1
HW HWH → I M M →∞ (2.17)
M

where HW is the spatially white Gaussian channel. Accordingly, the capacity can
be written as:

C = M log 2 (1 + γ ) (2.18)

This interesting result shows that the capacity increases linearly with an increasing
number of antennas.
The capacity of a single input multiple out (SIMO) system, which is also
known as the receiver diversity, can be expressed as [7]:

CSIMO = log 2 (1 + γ M R ) (2.19)

2.3 MIMO Capacity 15

As can be seen, only the logarithmic improvement in capacity with increasing

M R is achievable in the SIMO system. For a random channel, this relation is va-
lid for both the known channel state information (CSI) at the transmitter and the
unknown CSI. This is due to the implementation of a single transmitting antenna.
The capacity of the multiple input single output (MISO) channel when no CSI is
provided at transmitter is obtained as:

CSIMO = log 2 (1 + γ ) (2.20)

The capacity of the MISO system with a known CSI at transmitter is expressed as:

CMISO = log 2 (1 + γ M T ) (2.21)

Example 2.2: Capacity Estimation for 2X2 MIMO System

The capacity of a 2X2 MIMO system is presented in this example [25]. As shown
in Figure 2.3, a MIMO system with polarized matched transmitting and receiving
antennas is assumed. The antennas gains are 0 dBm.

r11 = 1000
x1 (t ) r21 y1 (t )
r12
d x2 (t ) r22 = 1000 y2 (t ) d

( a)
r11 = 1000
r21
r12
d r22 d

s = 100

r/2 r/2

(b)

Fig. 2.3 Capacity estimation for 2X2 MIMO channel: a) without scattering b) with single
scattering [25]
16 Chapter 2 MIMO Wireless Communications

The communications system operates at 1.9 GHz with a 200 KHz bandwidth and a
transmitting power of 1 mW. The noise temperature is 300 K. The capacity is cal-
culated in the following cases:

Case I: SISO Capacity

The received power due to the propagation path loss is a function of frequency op-
eration and is obtained as:

16π 2 d n
Pr = PG
t t Gr ( )
λ2

where Pr , Pt are the receiving and transmitting powers; Gr , Gt are the receiving
and transmitting antenna gains; d is the distance between the transmitter and the
receiver; n is the path loss exponent, which is usually between 2 and 6; and, λ is
the free space path length. In this example, it is assumed that n = 2. Using the
above relations and equation (2.20):

Pr = 1.58 × 10−13 W (−98dBm)

N o = kTe B = 8.28 × 10 −16 W (−120.8dBm)
γ = 22.8dBm
C1×1 = log 2 (1 + γ ) = 7.6 bps / Hz

Case II: MIMO Capacity without Scattering

As illustrated in Figure 2.3(a), by calculating H and using equation (2.15):

⎛ e− jkr11 e− jkr12 ⎞
H = ⎜ − jkr ⎟
⎜ e 21 e− jk11r22 ⎟⎠
⎝

where k = ω μ0ε 0 .
The capacity is obtained as:
2
γ
if d = .5 → λ1 = 2.0, λ2 = 0 → C2×2 = ∑ log 2 (1 + λi ) = 7.6 bps / Hz
i =1 2
2
γ
if d = 10 → λ1 = 1.41, λ2 = .6 → C2×2 = ∑ log 2 (1 + λi ) = 12.95 bps / Hz
i =1 2
2.4 MIMO Design Advantages 17

Case III: MIMO Capacity with Single Scattering

Similarly, as shown in Figure 2.3(b), by calculating H and using equation (2.15),

the capacity is obtained as:

2
γ
if d = .5 → λ1 = 2.14, λ2 = .2 → C2×2 = ∑ log 2 (1 + λi ) = 12.01 bps / Hz
i =1 2

Case IV: MIMO Capacity in a Rich Scattering Environment

By using equation (2.18) and assuming a scenario of rich scattering, the capacity
is obtained as 15.2 bps/Hz.

2.4 MIMO Design Advantages

MIMO systems can be designed either to provide maximal diversity to increase
transmission reliability or to achieve maximal multiplexing gain to support high
data rates [8]. Equation (2.15) shows that the channel capacity of a MIMO system
can be much higher than that of a SISO system. This performance of a MIMO sys-
tem is quantified with spatial multiplexing gain. On the other hand, if the signal
copies are transmitted from multiple antennas or received at more than one anten-
na, the multiple antenna systems can provide a gain that can improve reliability of
a wireless link. This gain is called diversity gain.
Furthermore, MIMO systems can be used to simultaneously provide both diver-
sity and multiplexing gain. However, there is a fundamental tradeoff between
them, either in large SNR values [9] or finite SNR values [11].

2.4.1 Space-Time Codes for Diversity

In this design, the signal copies are transmitted from multiple antennas or received
at more than one antenna in space-time multi-antenna systems. The average sym-
bol error probability of a MIMO communications system for maximum likelihood
(ML) detection is has an upper boundary in high SNR values as [4]:

γ d min
pe ≤ N e ( )− M (2.22)
4M

where Ne is the number of nearest neighbors in a scalar constellation, d min is the

minimum distance of separation of the scalar constellation, and M=min{MR,MT}.
As can be seen in Figure 2.4, increasing the number of antennas increases the
slope of the bit error rate (BER) curves and improves the reliability of wireless
communications. Two general techniques to generate the space-time codes for
18 Chapter 2 MIMO Wireless Communications

diversity are space-time block codes and space-time trellis codes [4], [7]. Al-
though the space-time trellis codes were introduced earlier [12], the space-time
block codes have been the preferred coding techniques in practice, due to the ease
of their implementation.
log[ BER]

Slope = −1
Slope = − M

SNR[dB]

Fig. 2.4 Space-time diversity gain in MIMO systems

 Example 2.3: Alamouti Space-Time Coding

The Alamouti code is an orthogonal space-time block code (O-STBC). It is im-
plemented by using two antennas at the transmitter and an arbitrary number of the
antennas at the receiver [10]. The code words for multiple antennas are written as:

1 ⎛ χ1 − χ2* ⎞
Χ= ⎜ ⎟ (2.23)
2 ⎜⎝ χ2 χ1* ⎟⎠

The Alamouti transmitter is shown in Figure 2.5. The Alamouti code has a spa-
tial multiplexing rate equal to one, as two symbols are transmitted over two sym-
bol durations. The performance of the Alamouti code is presented in Figure 2.6.

− χ 2* χ1
χ1* χ 2
T T
Alamouti − Coder
Information Bits Constellation
⎛ χ1 − χ 2* ⎞
Source Mapper ( χ1 χ2 ) → ⎜ ⎟
⎝ χ2 χ1* ⎠

Fig. 2.5 The Alamouti space-time coding

2.4 MIMO Design Advantages 19

Fig. 2.6 The performance of the Alamouti space-time code with different numbers of an-
tennas at the receiver

2.4.2 Spatial Multiplexing

In the layered space-time codes for spatial multiplexing, the sequence of informa-
tion bits is divided into certain sets of substreams. The substream where the signal
processing is conducted is referred as a layer. Hence, this is known as the layer
space-time (LAST) technique. As shown in Figure 2.7, MT independent substreams
are transmitted through MT antennas. In this technique, the number of receiving an-
tennas must be equal to or larger than the number of transmitting antennas [7].
The process of dividing the sequence of information into substreams is done by
a demultiplexer. The demultiplexing process may be applied to bits or symbols.
Hence, the encoding can be realized in three different ways according to the de-
multiplexer position in the transmitter chain and the direction of the layer [1].
These encoding processes are referred to as horizontal, vertical and diagonal en-
coding. The different encoding techniques are shown in Figure 2. 8.
In horizontal encoding, the data bits are demultiplexed into MT substreams that
are independently encoded, interleaved and modulated. On the other hand, in the
vertical realization, the data stream is encoded, interleaved and modulated; and,
the resulting symbols are then demultiplexed into MT substreams. The process to
realize the diagonal spatial multiplexing is similar to horizontal encoding with the
only difference being that, after the final stage, the frames of symbols undergo a
stream interleaver, which rotates the transmitted frames [7], [8].
20 Chapter 2 MIMO Wireless Communications

Fig. 2.7 Spatial multiplexing with three antennas at the transmitter and three antennas at the
receiver [8]

Bits Encoder
Interleaver
Bits DeMux
Modulator

Bits Encoder
Interleaver
Modulator

Encoder
Bits Symbol
Interleaver DeMux
Modulator

Bits Encoder Symbol

Interleaver
Bits DeMux
Modulator Stream
Bits Encoder Symbol Interleaver
Interleaver
Modulator

Fig. 2.8 Spatial multiplexing encoding: a) horizontal encoding, b) vertical encoding, c) di-
agonal encoding
2.4 MIMO Design Advantages 21

2.4.3 Diversity-Multiplexing Tradeoff

The channel capacity of a MIMO system can be considerably higher than that of a
SISO system using layered space-time coding. This performance of a MIMO
channel provides spatial multiplexing gain. On the other hand, MIMO systems can
improve the reliability of a wireless link; and, this performance provides diversity
gain. Traditionally, MIMO systems have been designed to extract either maximal
diversity or spatial multiplexing gain.
MIMO systems, however, can provide both diversity gain and multiplexing gain.
In [9], an optimal tradeoff curve between the asymptotic diversity and the multiplex-
ing gain has been derived at infinite SNR. The diversity gain is obtained as:
log Pout
d = lim − (2.24)
ρ →∞ log ρ
The spatial multiplexing gain is defined as:
R
r = lim (2.25)
ρ →∞ log ρ
where d is the diversity gain, r is the multiplexing gain, Pout is the outage probabil-
ity, R is the data rate in bps/Hz, and ρ is average SNR per antenna. It has been
shown that the most famous space-time coding schemes, such as space-time block
code (STBC) and Bell layered space-time (BLAST), are not optimal with respect
to the diversity-multiplexing tradeoff (DMT) criteria. The asymptotic diversity
multiplexing tradeoff and the finite SNR diversity multiplexing tradeoff for
MISO/SIMO and MIMO are presented in Figures 2.9 and 2.10 [14].

2
SNR =0 dB
1.8 SNR= 10 dB
SNR= 30 dB
1.6 Infinite SNR

1.4
Diversity Gain

1.2

0.8

0.6

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1
Multiplexing Gain

Fig. 2.9 Diversity-multiplexing tradeoff curve for SIMO (dashed lines) and MISO (solid
lines) using two antennas
22 Chapter 2 MIMO Wireless Communications

4
SNR=10 dB
SNR=20 dB
3.5
SNR=30 dB
SNR=50 dB
3 Infinite SNR

2.5
Diversity Gain

1.5

0.5

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Multiplexing Gain

Fig. 2.10 Diversity-multiplexing tradeoff curve for MIMO (2 X 2)

2.5 MIMO Channel Models

There are different classifications for MIMO channels [2], [16]. As shown in Ta-
ble 2.1, a classification based on the modeling approach divides the models into
physical models and analytical models. Physical channel models characterize a
channel on the basis of electromagnetic wave propagation. They explicitly model
wave propagation parameters, such as the complex amplitude and delay. More so-
phisticated models also incorporate polarization and time variation. Physical mod-
els are independent of antenna configurations (antenna pattern, number of anten-
nas, array geometry, polarization, mutual coupling) and system bandwidth [16].
Physical MIMO channel models can further be split into deterministic models,
geometry-based stochastic models, and non-geometric stochastic models. Deter-
ministic models characterize the physical propagation parameters in a completely
deterministic manner, e.g. ray tracing. Using geometry-based stochastic channel
models, the impulse response is characterized by the laws of wave propagation
applied to specific transmitters (Tx) and receivers (Rx) and scatterer geometries,
which are chosen in a stochastic (random) manner. In contrast, non-geometric sto-
chastic models describe and determine physical parameters in a completely sto-
chastic way by prescribing underlying probability distribution functions without
assuming an underlying geometry [17].
In contrast to physical models, analytical channel models characterize the im-
pulse response of the channel between the individual transmitting and receiving
antennas in a mathematical/analytical way without explicitly accounting for wave
2.5 MIMO Channel Models 23

propagation. The individual impulse responses are subsumed in a MIMO channel

matrix. Analytical models are very popular for synthesizing MIMO matrices in the
context of system and algorithm development and verification.

Table 2.1 MIMO Channel Modeling Classification

Physical Models Analytical Models

Deterministic, e.g. ray tracing Correlation based
Geometry-based Stochastic Propagation motivated
Non-geometrical stochastic,
e.g. Saleh-Valenzuela’s method
[17], [19]

Analytical models can be further subdivided into correlation-based models and

propagation-motivated models. Correlation-based models characterize the MIMO
channel matrix statistically, in terms of the correlations between the matrix entries.
The propagation-motivated models characterize the channel matrix via propaga-
tion parameters.
On the other hand, empirical methods can be extracted from analytical and
physical models based on experimental results [18], [19]. They either generalize
the tap-delay line concept to include the directional domains or use a combination
of tap-delay lines and geometry-based models with prescribed parameters. These
models are very useful for simulation purposes. They also constitute the back-
ground of many standardized channel models, such as 3GPP, COST 259, COST
273, IEEE 802.16a, IEEE 802.16e, and IEEE 802.11n.

2.5.1 Stanford University Interim Channel Models

In this section, the Stanford University Interim (SUI) channel models of MIMO
systems are briefly discussed [2], [19], [20], [24]. SUI channel models have been
developed for fixed wireless access networks at 2.5 GHz [16]. The general archi-
tecture of the SUI model is presented in Figure 2.11.
The input-mixing matrix models transmitter antenna correlations, and the out-
put-mixing matrix presents the receiver antenna correlations. The SUI channel
models basically use a three tap-delay line to model the multipath fading [4], [20].
Antennas are assumed to be omnidirectional on both the transmitting and receiv-
ing sides. The Doppler spectrum is assumed to be given by:
υ 2 υ
S (υ ) = 1 − 1.720( ) + 0.785( ) 4 υ ≤ υm (2.26)
υm υm
where υ m is the maximum Doppler frequency.
24 Chapter 2 MIMO Wireless Communications

Input Output
Vector Tap − Delay − Line Vector
Input Mixing Mixing
Matrix Output
Matrix Matrix

Fig. 2.11 Schematic of a SUI channel [4]

The models describe six types of channels, numbered from 1 to 6 [21]. As an

example, the parameters of SUI channel 1 are presented in Table 2.2.

Table 2.2 The channel parameters for SUI-1 model

SUI-1 Model Tap 1 Tap 2 Tap 3

Delay [microsec] 0.0 0.4 0.9
Power [dB] 0 -15 -20
99% K-factor 4 0 0
Doppler Frequency [Hz] 0.4 0.3 0.5
Envelope Antenna Correlation 0.7 0.7 0.7

Example 2.4: IEEE 802.16d/e Models

These models are intended for macro-cellular fixed wireless access [24]. The tar-
geted scenario is as follows:
• The cell size is less than 10 km in radius,
• The user’s antenna is installed under the eaves or on the rooftop, and
• The base station height is 15 to 40 m.
References 25

In fact, the IEEE 802.16 model is an improved version of the SUI channel
models and is valid for both omnidirectional and directional antennas. The use of
directional antennas cause the global K-factor to increase, while the delay spread
decreases. As an example, Table 2.3 indicates how SUI channel 4 is modified
when the terminal antenna has a 30-degree beamwidth. An additional feature of
the IEEE 802.16 standard is a model for the narrowband Ricean K-factor:

K = K0FsFhFbRγu (2.26)

where Fs is a seasonal factor, Fs = 1.0 in summer (leaves) and 2.5 in winter (no
leaves); Fh is the receiving antenna height factor, Fh = 0.46(h/3) (h is the receiving
antenna height in meters); Fb is the beamwidth factor, Fb = −0.62(b/17) (b in de-
grees); K0 and γ are regression coefficients, K0 = 10 and γ = −0.5; and, u is a log-
normal variable, i.e. 10log10 (u) is a zero-mean normal variable with a standard
deviation of 8 dB.

Table 2.3 IEEE 802.16 Model

IEEE 802.16 Model Tap 1 Tap 2 Tap 3

Delay [microsec] 0.0 1.5 4
Power [dB] 0 -10 -20
99% K-factor 1 0 0
Doppler Frequency [Hz] 0.2 0.15 0.25
Envelope Antenna Correlation 0.3 0.3 0.3

References
[1] Carlson, A.B., Crilly, P.B., Rutledge, J.C.: Communication Systems: An Introduction
to Signal and Noise in Electrical Communications, 4th edn. McGraw Hill (2001)
[2] Oestges, C., Clerckx, B.: MIMO Wireless Communications: From Real World Prop-
agation to Space Time Code Design. Academic Press (2007)
[3] Pahlavan, K., Levesque, A.: Wireless Information Networks, 2nd edn. John Wiley
and Sons (2005)
[4] Paulraj, A., Nabar, R., Gore, D.: Introduction to Space-Time Wireless Communica-
tions. Cambridge University Press (2003)
[5] Lee, W.C.Y.: Estimation of channel capacity in Rayleigh fading environment. IEEE
Transactions on Vehicular Technology 39(3), 187–189 (1990)
[6] Mohammadi, A., Kumar, S.: Characterization of Adaptive Modulators in Fixed Wire-
less ATM Networks. IEEE/KICS Journal of Communications and Networks 6(2),
123–132 (2004)
[7] Tsoulos, G.: MIMO System Technology for Wireless Communications. CRC Press
(2006)
[8] Gesbert, D., Shafi, M., Shiu, D., Smith, P.J., Naguib, A.: From Theory to Practice: An
Overview of MIMO Space–Time Coded Wireless Systems. IEEE Journal on Selected
Areas in Communications 21(3), 281–302 (2003)
26 Chapter 2 MIMO Wireless Communications

[9] Zheng, L., Tse, D.N.C.: Diversity and multiplexing: a fundamental tradeoff in mul-
tiple antenna channels. IEEE Transactions on Information Theory 49, 1073–1096
(2003)
[10] Alamouti, S.M.: A simple transmit diversity technique for wireless communications.
IEEE Journal on Selected Areas in Communications 16(10), 1451–1458 (1998)
[11] Ebrahimzad, H., Mohammadi, A.: On Diversity-Multiplexing Tradeoff in MIMO
channel at Finite SNR. IEICE Transactions on Fundamentals of Electronics, Commu-
nications and Computer Sciences E93-A(11), 2057–2064 (2010)
[12] Tarokh, V., Seshadri, N., Calderbank, A.R.: Space-time Codes for High Data Rate
Wireless Communication: Performance Criterion and Code Construction. IEEE
Transactions on Information Theory 44(2), 744–765 (1998)
[13] Gershman, A.B., Sidiropoulos, N.D.: Space-Time Processing for MIMO Communica-
tions. Wiley (2005)
[14] Ebrahimzad, H., Mohammadi, A.: Diversity-Multiplexing Tradeoff in MIMO Sys-
tems with Finite SNR. In: European Conference on Wireless Technology, Munich,
pp. 146–149 (October 2007)
[15] Papoulis, A., Pillai, S.U.: Random Variable Variables and Stochastic Process, 4th
edn. McGraw Hill (2002)
[16] Almers, P., Bonek, E., Burr, A., Czink, N., Debbah, M., Degli-Esposti, V., Hofstetter,
H., Kyosti, P., Laurenson, D., Matz, G., Molisch, A.F., Oestges, C., Ozcelik, H.: Sur-
vey of Channel and Radio Propagation Models for Wireless MIMO Systems.
EURASIP Journal on Wireless Communications and Networking 2007, article ID
19070, 19 pages (2007)
[17] Saleh, A.M., Valenzuela, R.A.: A statistical model for indoor multipath propagation.
Journal on Selected Areas in Communications 5(2), 128–137 (1987)
[18] Rappaport, T.S.: Wireless Communications: Principles and Practice, 2nd edn. Pren-
tice Hall (2002)
[19] Wallace, J.W., Jensen, M.A.: Modeling the indoor MIMO wireless channel. IEEE
Transactions on Antennas and Propagation 50(5), 591–599 (2002)
[20] Baum, D.S., Gore, D.A., Nabar, R.U., Panchanathan, S., Hari, K.V.S., Erceg, V.,
Paulraj, A.J.: Measurement and characterization of broadband MIMO fixed wireless
channels at 2.5 GHz. In: Proceedings of the International Conference on Personal
Wireless Communications (ICPWC 2000), India (December 2000)
[21] Erceg, V., et al.: IEEE p802.16 – channel models for fixed wireless applications
(ieee802.16.3c-01/29r4) (2001)
[22] Correia, L.M.: COST 259 – Wireless flexible personalized communications. Wiley,
London (2001)
[23] Correia, L.M.: COST 273 – Towards mobile broadband multimedia networks.
Elsevier, London (2006)
[24] IEEE P802.16 e /D12, IEEE Standard for Local and metropolitan area networks Part
16: Air Interface for Fixed and Mobile Broadband Wireless Access Systems Amend-
ment for Physical and Medium Access Control Layers for Combined Fixed and Mo-
bile Operation in Licensed Bands (October 2005)
[25] Flaviis, F.D., Jofre, L., Romeu, J., Grau, A.: Multiantenna Systems for MIMO Com-
munications. Morgan & Craypool Publishers (2008)
Chapter 3
Digital Modulation Techniques in MIMO
Systems

The basic work related to digital communication techniques was carried out by
Shannon [1]. Since then, the theory and applications of digital communication
systems have been greatly advanced [2-6]. With the introduction of mobile and
cellular communications, digital communication techniques have been
investigated for wireless systems [7-10]. In addition, a new stream of research on
and implementation of digital communication techniques has been established
with the advent of multiple input multiple output (MIMO) communication systems
during the last decade, [11-14].
This chapter provides an overview of different digital modulation techniques
and their applications in single input single output (SISO) and MIMO wireless
systems with an emphasis on their implementation techniques. The fundamentals
of digital modulation techniques are presented. Then, the analysis and the
modeling of the various modulation schemes both in single carrier, multicarrier
and orthogonal frequency-division multiplexing (OFDM) systems are described.
The last section of the chapter is devoted to the accuracy metrics of digital
modulation implementation.

3.1 Criteria to Design Digital Modulators

The design of any digital communication system is initiated with a description of

the channel (received power, available bandwidth, attenuation, noise statistics, and
other impairments such as fading) and a definition of the system requirements
(data rate and error performance). Design choices that match the channel and meet
the performance requirements then need to be determined [4], [15]. The general
criteria to select a digital modulation method are spectral efficiency, bandwidth
efficiency, and implementation complexity. These criteria are even more critical in
wireless communications, due to spectrum limitations. In addition, the proper
operation of a digital modulation scheme in a wireless channel is an important
factor that must be considered.

A. Mohammadi et al.: RF Transceiver Design for MIMO Wireless Communications, LNEE 145, pp. 27–54.
springerlink.com © Springer-Verlag Berlin Heidelberg 2012
28 Chapter 3 Digital Modulation Techniques in MIMO Systems

3.1.1 Spectral Efficiency

The spectral efficiency of a digital modulator is defined as the number of bits per
second that can be transmitted in one Hertz of system bandwidth. A suitable
starting point to describe the spectral efficiency is the Shannon capacity theorem,
which can be stated as:
S
C = W log 2 (1 + ) (3.1)
N
where S / N is the ratio of the average received signal power to the noise power,
and W is the channel bandwidth. The capacity, C, is given in bit/sec. The capacity
of a channel defines the maximum number of bits that can be reliably sent per
second over the channel. Accordingly, the maximum spectral efficiency, η max , is
defined as:
C S
η max = log 2 (1 + ) (3.2)
W N
Relationship (3.1) can be written as:

C S
= log 2 (1 + ) (3.3)
W N oW

where N o is the power spectral density of white noise.

On the other hand, for any digital communication system, the relationship
between the received S / N and the received bit energy to noise-power spectral
density, Eb / N o , is as follows [4],[15]:

S E
= b rb (3.4)
N No

where rb is the data rate in bits per second.

Figure 3.1 shows the relationship between rb / W and Eb / N o for the case
where the data (information) rate, rb , is equal to C. As can be seen, a curve
separates a region of practical communication systems from a region where such
communication systems cannot operate reliably. In the spectral efficient region, a
modulation scheme can be designed to transmit a number of bits in one hertz using
a practical Eb / N o .
3.1 Criteria to Design Digittal Modulators 229

Fig. 3.1 The bandwidth efficciency of various digital modulation systems[15]

3.1.2 Power Efficieency

The power efficiency of a digital modulator is defined as the required signal-too-
noise ratio (SNR) for a certain
c bit error rate (BER) probability over an additivve
white Gaussian noise (AAWGN) channel. The power spectral efficiency can bbe
explained with the use off Figure 3.2. This figure shows the waterfall curves foor

Fig. 3.2 The BER curves veersus Eb / N o for Binary PAM and 4-ary PAM modulation in
AWGN channel
30 Chapter 3 Digital Modulation Techniques in MIMO Systems

two modulation schemes – binary pulse amplitude modulation (B-PAM) and 4-ary
PAM. The B-PAM modulation scheme requires less Eb / N o to provide a certain
BER probability compared to the 4-ary PAM modulation scheme. Therefore, B-
PAM is more power efficient than 4-ary PAM.

3.1.3 Implementation Complexity

The complexity of the modulator implementation has a considerable impact on the
performance and cost. As can be seen from Figure 3.2, the M-ary modulation
techniques provide either better bandwidth efficiency, e.g., MQAM (M-ary
quadrature amplitude modulation) and MPSK (M-ary phase-shift keying), or
better power efficiency, e.g., MFSK (multiple frequency-shift keying). However,
their implementations are complex and costly. On the other hand, some constant
envelope modulation techniques, such as Guassian minimum-shift keying
(GMSK), can be easily implemented, but their bandwidth efficiencies are usually
low. In wireless applications, although multicarrier modulations, such as OFDM,
require complex and costly implementations, they provide better performance in
multipath fading channels [7], [9].

3.1.4 Performance Study in Wireless Channel

Multipath channels are characterized by coherence bandwidth. The coherence
bandwidth is the maximum channel width that can be used and is extracted from
the power delay profile of a wireless channel [7], [8]. To estimate the capacity in a
multipath fading channel, the Shannon capacity is examined first. The Shannon
capacity for an AWGN channel is obtained using (3.1). For a fading channel, the
Shannon limit can be modified as [5], [16]:

≺C
= ∫ log 2 (1 + γ ) p (γ )d γ (3.5)
W γ

where γ = S / N and pγ (γ ) is the probability density functions of the received

signal.
In a fading channel, the received signal power varies randomly over distance or
time in a fading channel. Accordingly, γ is0020a random variable with a distribution
of pγ (γ ) . Therefore, the capacity in a fading channel decreases by a factor:

≺C
∫ log 2 (1 + γ ) p(γ )d γ
γ
ξc = = (3.6)
C log 2 (1 + γ )

where ξc is called the capacity reduction factor.

3.1 Criteria to Design Digital Modulators 31

On the other hand, if the signal fade level is constant at about one symbol time,
the average error probability is a suitable performance metric [8]. The average
error probability is computed by integration of the error probability in the AWGN
channel over the fading distribution. It is obtained as:

P s = ∫ Ps (γ ) pγ (γ )d γ (3.7)
γ

where Ps (γ ) is the probability of the symbol error in AWGN with a SNR equal to
γ . The symbol error rate for coherent modulation can be shown as a generic
relationship [8]:

Ps (γ ) ≈ α M Q( β M γ ) (3.8)

where α M and β M depend on the modulation and its approximation and

∞
1
Q( z ) = ∫ e− x / 2 dx . This generic relationship is depicted in Table 3.1 for
2

z 2π
QAM modulation.

Table 3.1 Approximate Symbol and Bit Error Probabilities for Bandwidth Efficient
Modulation Schemes [8]

Modulation Ps (γ s ) Pb (γ b )
MPAM
2( M − 1) 6γ s 2( M − 1) 6γ b log 2 M
Ps = Q( ) Pb ≈ Q( )
M M 2 −1 M log 2 M M 2 −1
MPSK π 2 π
Ps ≈ 2Q ( 2γ s sin( )) Pb ≈ Q( 2γ b log 2 M sin( ))
M log 2 M M
MQAM
3γ s 4 3γ b log 2 M
Ps ≈ 4Q ( ) Pb ≈ Q( )
M −1 log 2 M M −1

On the other hand, for a Rayleigh fading channel, the probability density
function of the received SNR is given by:

1 −γ / Γ
pγ (γ ) = e (3.9)
Γ

where Γ is the average power of γ .

32 Chapter 3 Digital Modulation Techniques in MIMO Systems

Thus, according to (3.7), the average probability of error in Rayleigh fading can be
approximated as [8]:

1 −γ / Γ α
P s = ∫ Ps (γ ) pγ (γ )d γ = ∫ α M Q( β M γ ) e dγ ≈ M (3.10)
γ γ
Γ 2β M Γ

The last approximation is more accurate for higher SNRs. The symbol error
rate in a Rayleigh fading channel decreases linearly with an increasing average
SNR. However, the symbol error rate in an AWGN channel decreases
exponentially with increasing SNR.
Similarly, the average capacity concept can be extended to MIMO channels.
The average capacity of a MIMO channel in Rayleigh fading can be represented
as [17], [18]:
∞
≺C 1 s −1
k ! yt −s
= ∫ Ln(1 + ρ y )e − y ∑ [ Ltk− s ( y)]2 dy (3.11)
W Ln(2) 0 k = 0 ( k + t − s )!

where t = max( Lt , Lr ), s = min( Lt , Lr ), ρ is the normalized transmitting SNR per

branch, and Lan (.) is the generalized Laguerre polynomial which is expressed as:

x −α e x d n − x n +α
Lαn ( x) = (e x ) (3.12)
n ! dx n

3.2 Single-Carrier SISO Systems

3.2.1 Single-Carrier Modulation

In digital communication systems, each symbol in an M-ary alphabet is related to
a unique sequence of k bits, expressed as:

M = 2k k = log2 M (3.13)

where M is the size of the alphabet. In the case of digital transmissions, the term
“symbol” refers to the member of the M-ary alphabet that is transmitted during
each symbol duration, Ts . In order to transmit the symbol, it must be mapped onto
an electrical voltage or current waveform. Since one of M symbols or waveforms
is transmitted during each symbol duration, Ts , the data rate, rb, in b/s, can be
expressed as [4], [6]:
1 k log 2 M
rb = = = bit / sec (3.14)
Tb Ts Ts
3.2 Single-Carrier SISO Systems 33

From (3.10), one can write that the effective time duration, Tb , of each bit in
terms of the symbol duration, Ts , or the symbol rate, Rs , is:

1 Ts 1
Tb = = = (3.15)
rb k kRs
Using (3.9) and (3.11) .we can express the symbol rate, Rs , in terms of the bit
rate, rb , as follows:

rb
Rs = Symbol / sec (3.16)
log 2 M

From (3.10) and (3.11), one can see that any digital scheme that transmits k =
log2M bits in Ts seconds using a bandwidth of W Hertz, operates at a bandwidth
efficiency of:
rb log 2 M 1
η = = = (3.17)
W WTs WTb

where Tb is the effective time duration of each data bit [4].

From (3.13), the smaller the WTb product, the more bandwidth efficient is any
digital communication system. Thus, signals with small WTb products are often
used in systems of limited bandwidth. MQAM and MPSK modulation schemes
are two examples of bandwidth efficient modulation techniques that are often used
in practical radio systems. The symbol and bit error probabilities for the
bandwidth efficient modulation schemes based on γ b = Eb / N o and γ s = Es / N o
are presented in Table 3.1 [8]. The γ s is equal to the average energy per symbol.
In the following subsections, the MQAM and MPSK modulation techniques are
explained in more detail. The different spectral efficient modulation techniques
are discussed extensively in literature [2-10].

3.2.2 MQAM
M-ary quadrature amplitude modulation (MQAM) is a two-dimensional linear
modulation scheme and it may be represented as:

sm ( t ) = Amc g ( t ) cos2π fct − Ams g ( t ) sin2π fct, m = 1,2,…, L (3.18)

where g ( t ) is a signal pulse shape, f c is carrier frequency, Amc and Ams are the
in-phase (I) and quadrature (Q) signal amplitudes, and L = M .
The constellation diagram of MQAM modulation in the I-Q plane is illustrated
in Figure 3.3 [6].
34 Chapter 3 Digital Modulation Techniques in MIMO Systems

Fig. 3.3 The constellation diagram of MQAM modulator with different constellation sizes

3.2.2.1 MQAM Spectral Efficiency

The low-pass equivalent of an MQAM signal is given by:

v ( t ) = ∑( an + jbn ) g ( t − nTs ) ; an , bn = ±1, ±3,… , ± L − 1 (3.19)
n

where Ts is the symbol rate, and g ( t ) is the pulse shape. The band pass signal,
sm ( t ) , can be related to the low-pass signal, v ( t ) , through:

sm ( t ) = Re ⎡⎣ v(t )e j 2π fc t ⎤⎦ (3.20)

The autocorrelation function of sm ( t ) is given by:

ϕss (τ ) = Re ⎡⎣φvv (τ ) e j 2π f t ⎤⎦
c
(3.21)

where ϕvv (τ ) is the autocorrelation function of the equivalent low-pass signal,

v (t ) .
3.2 Single-Carrier SISO Systems 35

The power spectral density of sm ( t ) is then given by:

1
фss ( f ) = ⎡ фvv ( f − f c ) + фvv ( − f − f c ) ⎤⎦ (3.22)
2⎣

where фvv ( f ) is the power spectral density of v (t ) . It can be shown [2] that the
power spectral density of v (t ) is given by:

1
фvv ( f ) = G ( f ) фii ( f )
2
(3.23)
2

where G ( f ) is the Fourier transform of g(t) and фii ( f ) is the power spectral
density of the information sequence.
The QAM signal may be considered to consist of two pulse amplitude
modulation (PAM) signals conveyed by the cosine and sine carriers. The spectral
efficiency of the QAM will be twice that of its PAM components in the I and Q
channels. Substitution for фii ( f ) in the above equation [2], gives the power
spectral density signal carriers by the cosine carrier as:
2
σ2 ⎛m⎞
μ2 ∞
m
фvv ( f ) = G ( f ) + 2 ∑ G ⎜ ⎟ δ ( f − ),
2
(3.24)
Ts T m =−∞ ⎝ Ts ⎠ T

where μ and σ 2 are the mean and variance of the information sequence.
It is known that the mean value of the PAM symbols is zero ( μ = 0) , and, thus,
this equation can be simplified as:

σ2
фvv ( f ) =
2
G( f ) (3.25)
Ts

Hence, the spectral efficiency of the PAM signal is controlled by the pulse shape,
g ( t ) . The raised cosine pulse is commonly used as a pulse shape for the digital
modulations in bandwidth-limited channels. The signal g ( t ) for this pulse shape
is given by:
πt πα t
sin( ) cos( )
Ts Ts
g (t ) = (3.26)
πt 2α t 2
1− ( )
Ts Ts

where α is called the roll of factor.

36 Chapter 3 Digital Modulation Techniques in MIMO Systems

The Fourier transform g ( t ) is given by:

⎧ 1−α
⎪ T if 0 ≤ f ≤
2Ts
⎪
⎪⎪ T πT 1−α 1−α 1+ α
G ( f ) = ⎨ s [1 + cos( s ( f − ))] if ≤ f ≤ (3.27)
⎪ 2 α 2T s 2 s 2Ts
⎪ 1+α
⎪ 0 if f ≥
⎪⎩ 2Ts

The normalized raised cosine filter in the frequency and time domains are
illustrated in Figure 3.4. Thus, the bandwidth efficiency of a QAM modulator can
be adjusted by parameter α .

1.2

0.8
Amplitude

0.6

alpha=.5 alpha=.3
0.4

alpha=.7
0.2

-0.2
-4 -3 -2 -1 0 1 2 3 4
Normalized Time

1.2

1
Alpha=0
0.8
Alpha=.5
0.6 Alpha=1

0.4

0.2

0
-1.5 -1 -0.5 0 0.5 1 1.5
Normalized Frequency

Fig. 3.4 The pulse shape and the spectrum of a raised cosine pulse shape
3.2 Single-Carrier SISO Systems 37

As can be seen from Figure 3.4, the channel bandwidth, W, is approximately equal
1+ α 1
to ; and, since = rb / log 2 ( M ) , we obtain the result of:
Ts Ts
rb log 2 M
η=
= , (3.28)
W 1+ α
where η is the QAM spectral efficiency, and R is the QAM transmission rate.

3.2.2.2 MQAM Power Efficiency Performance

As explained, the power efficiency is a measure of the received power needed to

achieve a specified BER (bit error rate). For an AWGN channel, the probability of
bit error in the MQAM system for the coherent modulation is obtained as [19]:

1
4(1 − )⎧
M ⎪ ⎛ 3γ s ⎞
Pb , MQAM = ⎨Q ⎜⎜ ⎟⎟ +
log 2 M ⎪⎩ ⎝ M − 1 ⎠ (3.29)
⎛ 3γ s ⎞ ⎛ 3γ s ⎞ ⎪⎫
+Q ⎜ 3* ⎟⎟ + Q ⎜⎜ 5* ⎟⎟ + ...⎬
⎜ M −1 ⎠ M −1 ⎠
⎝ ⎝ ⎪⎭

It is usually a good approximation to compare the first terms, if one ignores the
other terms. Accordingly, the BER of QAM modulation can be written as [20]:
4( M − 1) ⎛ 3γ s ⎞
Pb ≈ Q⎜ ⎟ (3.30)
M log 2 ( M ) ⎜⎝ M − 1 ⎟⎠
The BER of MQAM modulation schemes, assuming perfect clock and carrier
recovery, are shown in Figure 3.5 [20]. The symbol error rate of an ideal
rectangular QAM modulation can be obtained as [2]:

(
PM = 1 − 1 − 2 P M + ( P M )2 ) (3.31)
= 2 P M − ( P M )2

where
⎛ 1 ⎞ ⎛ 3γ s ⎞
P = 2 ⎜1 − ⎟ × Q ⎜⎜ ⎟⎟ (3.32)
⎝ M −1 ⎠
M
⎝ M ⎠

Similarly, if one can ignore the second term, in order compare to the first terms in
(3.31), the symbol error rate of an ideal rectangular MQAM modulation can be
obtained as:
⎛ 1 ⎞ ⎛ 3γ s ⎞
PM 2 P M = 4 ⎜1 − ⎟ × Q ⎜⎜ ⎟⎟ (3.33)
⎝ M ⎠ ⎝ M −1 ⎠
38 Chapter 3 Digital Modulation Techniques in MIMO Systems

Fig. 3.5 Bit error rate versus Eb/No for ideal MQAM modulation in an AWGN channel

3.2.2.3 MQAM Modulator Implementation

The QAM modulator has been traditionally realized using the heterodyne method
shown in Figure 3.6. The QAM mapper in the figure divides a binary sequence
with a rate of fb into two binary symbol streams, each with a rate of fb/2.
A 2-to- L = M level baseband converter is used to convert the streams into
L = M -level baseband PAM signals in the I or Q paths. These symbols are
shaped by raised cosine filters. The filtered I and Q baseband signals are used to
modulate the I and Q outputs of an intermediate frequency (IF) oscillator. The
modulated signal at the IF is up-converted to the desired transmitting frequency in
one or more steps. Mixers are commonly used in this realization. A power
amplifier is used to boost the modulated signal at the transmitting frequency to the
required power.
3.2 Single-Carrier SISO Sy
ystems 339

It should be emphasizeed that, if the IF frequency is low, the classic heterodynne

design described above suffers
s from disadvantages, such as design complexitiees,
high costs, radio frequenccy (RF) filter requirements, and data rate limitations.
Direct implementation n methods are often preferred to the classic design iin
wireless systems, where these
t limitations can be effectively overcome. The basic
idea of direct microwavee implementation is the use of the baseband signal tto
modulate a carrier at the t desired transmitted frequency. This technique is
examined in detail in the following
f chapters.
The MQAM receiver can c be also implemented using either superheterodyne oor
direct conversion method ds. A general block diagram of an MQAM receiver usinng
a superheterodyne method d is shown in Figure 3.7.

Fig. 3.6 A superheterodyn

ne implementation of an MQAM transmitter

vth1
LPF
Digital fb /2
vth2 Logic

BPF AMP Data

Vth-n combiner
90

L-to 2 level converter in Q

LPF
channel
fb /2

Fig. 3.7 A superhetrodyne im

mplememntation of an MQAM receiver

3.2.3 MPSK
In the MPSK modulation scheme, all of the information is encoded in the phase oof
the transmitted signal. Th
herefore, the transmitted signal over symbol period, Ts ,,is
[23], [8]:
40 Chapter 3 Digital Modulation Techniques in MIMO Systems

2 Es ⎛ 2π ( m − 1) ⎞
sm ( t ) = cos ⎜ 2π f c t + ⎟⎟ 0 ≤ t ≤ Ts m = 1, 2,...., M (3.34)
Ts ⎜ M
⎝ ⎠

∞
where Es = ∫ xm2 (t )dt .
−∞
The most common MPSK technique is quadrature phase-shift keying (QPSK)
where M = 4. In this modulation, the symbols can have four different values. The
modulated signals are:

2 Es ⎡ π⎤
xm ( t ) = cos ⎢( 2m − 1) ⎥ cos ( 2π f c t )
Ts ⎣ 4⎦
(3.35)
2 Es ⎡ π⎤
− sin ⎢( 2m − 1) ⎥ sin ( 2π f c t ) 0 ≤ t ≤ Ts , m = 1, 2,3, 4
Ts ⎣ 4⎦

Hence, the modulated signal can be shown as:

⎡ π⎤ ⎡ π⎤
xm ( t ) = Es cos ⎢( 2m − 1) ⎥ φ1 ( t ) − Es sin ⎢( 2m − 1) ⎥ φ2 ( t ) (3.36)
⎣ 4 ⎦ ⎣ 4⎦

where

⎧ 2
⎪φ1 ( t ) = cos ( 2π f c t ) 0 ≤ t ≤ Ts
⎪ Ts
⎨
⎪ 2
⎪φ2 ( t ) = T sin ( 2π f c t ) 0 ≤ t ≤ Ts
⎩ s

The constellation diagrams for QPSK and 8PSK modulation schemes are
illustrated in Figure 3.8. Figure 3.8(a) shows the QPSK constellation diagram in
two-dimensional signal space. The decision regions are selected as [23]:

sˆ = "00" if r ∈ D1 ;
sˆ = "01" if r ∈ D2 ;
sˆ = "11" if r ∈ D3 ;
sˆ = "10" if r ∈ D4 ;

The BER and symbol error rate (SER) of QPSK modulation based on Eb/No is
presented in Figure 3.9. As can be seen from Figures 3.5 and 3.9, the BER of
QPSK modulation is similar to that of 4QAM modulation.
3.3 Single-Carrier MIMO Systems 41

ϕ 2 (t )
ϕ 2 (t )

D2 x2 x1 D
1

"01" "00"
2 Eb ϕ1 (t )
ϕ1 (t )

"11" x3 x4"10"
D3 D4

Fig. 3.8 a) QPSK modulation and decision regions [23], b) 8PSK modulation

0
10
Theoretical SER
Theoretical BER
-1
10

-2
10
SER and BER

-3
10

-4
10

-5
10

-6
10
0 1 2 3 4 5 6 7 8 9 10
Eb/No (dB)

Fig. 3.9 BER and SER of QPSK modulation

3.3 Single-Carrier MIMO Systems

A flat-fading MIMO channel with NT transmitter (Tx) antennas and NR receiver
(Rx) antennas is assumed. In a single-carrier MIMO system considering perfect
synchronization, the input-output relationship is given by:

y=Hx+w (3.37)
42 Chapter 3 Digital Modulation Techniques in MIMO Systems

where y is an NR×1 vector of the received signal, H denotes the NR×NT channel
matrix, x is an NT×1 vector of the transmitted symbols, and w is an NR×1 AWGN

{hμυ }μ
N ,N
R T
vector. The channel coefficients, =1,υ =1
, are statistically independent and
identically distributed complex-valued Gaussian random variables with a variance
of 1. The components of the noise vector have the same distribution with a
variance of N0.

3.3.1 QAM-MIMO
As discussed in Chapter 2, the objective of MIMO system design can be either
improvement of the diversity order using space-time codes or increasing the data
rate using layered space-time codes for spatial multiplexing. In the first design, the
space-time codes are used to obtain a SER in a fading channel similar to the SER
in an AWGN channel. It can be proven that, by increasing the number of antennas
in the transmitter and receiver and using a suitable space-time code, the SER
based on SNR curves in a fading channel get closer to this type of curve in an
AWGN channel [14], [11]. Space-time block codes are often used for this
application. However, the theory of the space-time trellis codes are also well
developed [11], [12].Figure 3.10 illustrates a general block diagram of QAM-
MIMO implementation using a space-time coder.
On the other hand, in a spatial multiplexing scenario, the objective is the
increase of data. In this application, the main target is the sending of more bits per
second per hertz compared to a SISO system. The SER (or BER) performance of
this design is similar to the waterfall curves in a fading channel. Figure 3.11 shows
a BER curve based on SNR for QAM-MIMO using four antennas in both the
transmitter and receiver in a Rayleigh fading channel. The measurement results
are also included to evaluate the analytical results.

x1 y1

x2 y2

x Nt y Nr

Fig. 3.10 Block diagram of a QAM-MIMO system

3.3 Single-Carrier MIMO Systems 43

Fig. 3.11 QAM-MIMO spatial multiplexing in Rayleigh fading

3.3.2 MPSK-MIMO
MPSK-MIMO is also used in design scenarios, either for the diversity order
improvement or spatial multiplexing. A general block diagram of an MPSK-
MIMO system is presented in Figure 3.12. This system also uses space-time

x1 y1

x2 y2

x Nt yNr

Fig. 3.12 Block diagram of an MPSK-MIMO system

44 Chapter 3 Digital Modulation Techniques in MIMO Systems

codes for diversity order improvement and layered space-time codes for spatial
multiplexing purposes. As discussed in Chapter 2, the simplest space-time codes
are Alamouti codes [23]. The SER performance of the Alamouti scheme with
8PSK modulation is presented in Figure 3.13 [14].

0
10
8PSK Modulation & Number of receive antennas=2

-1
10
Symbol Error rate

-2
10

-3
10

-4
10

-5
10
4 6 8 10 12 14 16 18 20
SNR[dB]

Fig. 3.13 Symbol error rate of Alamouti space-time codes with 8PSK modulation

3.4 Multicarrier SISO Modulation Systems

In a wireless multipath channel, the coherence bandwidths may be smaller than
the channel bandwidth. If the channel bandwidth is smaller than the coherence
bandwidth, the channel can be assumed a flat-fading channel [7]. This results in a
frequency selective channel response. The main idea of a multicarrier modulation
technique is the division of the channel bandwidth to a number of orthogonal
subchannels, where the number of subchannels is chosen to make the subchannel
bandwidth to be smaller than the coherence bandwidth. Each subchannel carries a
fraction of data rate. Indeed, multicarrier modulation is conceptually designed
based on frequency-division multiplexing (FDM) [8]. However, unlike a
conventional FDM system, the overlapping between adjacent subcarriers is
possible. A generic diagram of a multicarrier modulator and demodulator is shown
in Figure 3.14. The high-speed data stream, x(n), is divided into N slower streams
and modulated using a prescribed signal constellation [22]. The modulated
streams, d(k)(n), are then up-sampled by a factor. This results in signals y(k)(n).
They are then filtered by a bank of synthesis filters, g(k)(n), and the filtered signals
are summed. The transmitted signal, s(n), has the form:
3.5 OFDM Modulation Systems 45

N −1 ∞
s ( n) = ∑ ∑ g (k ) (l ) y(k ) (n − l ) (3.38)
k = 0 l =−∞

At the receiver, a bank of filters, fˆ (n) , are used to separate the composite
(k )

(k )
signal, r(n), into N individual signals, yˆ (n) . These signals are down-sampled by
factor N, yielding dˆ (n) . To remove the distortion introduced by the channel,
(k )

ˆ ( k ) (n) , k = 0,1,2, …, N-1, are employed on a per subcarrier basis.

equalizers, w

d (0) (n) y (0) (n)

↑N g (0) (n)
x ( n) Bit Substream
SUM
DeMux Modulator d ( N −1) (n) y ( N −1) (n)

↑N g ( N −1) (n)

dˆ (0) (n)
yˆ (0) (n)
f (0) ( n) ↓N w(0) (n)

Substream Bit xˆ(n)

dˆ ( N −1) (n) Demodulator Mux
yˆ ( N −1) ( n)
f ( N −1) (n) ↓N w( N −1) (n)

Fig. 3.14 Multicarrier modulation systems: a) transmitter b) receiver

3.5 OFDM Modulation Systems

Orthogonal frequency-division multiplexing (OFDM) is realized using the substream
spectrum overlapping concept, where each substream is transmitted through a
subchannel; and, the substreams are still orthogonal, and they can be separated in a
demodulator. Moreover, it is implemented using a fast Fourier transformation (FFT)
and inverse FFT (IFFT) technique and contains a cyclic prefix. OFDM is basically a
multicarrier modulation technique, and its operation can be described based on the
concept of the overlapping multicarrier modulation technique [8].
Assume a communication system with a baseband bandwidth, B, (passband
bandwidth, 2B) and a data rate, R. Moreover, it is assumed that the coherence
bandwidth for the channel is Bc <B and that the whole bandwidth is divided into N
subchannels, where BN = B/N << Bc. Using these assumptions, flat fading on
each subchannel can be considered [8]. The input data is also divided into N
substreams. The center frequency of the overlapping channels are:
46 Chapter 3 Digital Modulation Techniques in MIMO System
ms

f n = f c + nBN n = 0,..., N − 1 (3.399)

where fc is the carrier freq
quency.
The data rate for each substream is RN = R/N. If a raised cosine pulse shape is
assumed, the symbol timee for each substream is obtained as:
1+ α
TN = (3.400)
BN

where α is the roll-off faactor.

The modulated signal related
r to all subcarriers is presented as:
N −1
s(t ) = ∑ si g (t ) cos(2π fi t + ϕi ) (3.41)
i =0

where g (t ) is the pulse-sshaping filter, si is the complex symbol associated witth

ith subcarrier, and ϕ i is the phase of the ith carrier. The subcarrierrs,
cos(2π ( f c + i / TN )t + ϕi ) , provide a set of orthonormal basis functions over thhe
interval [0, TN]. It is a strightforward exercise that shows that the minimum m
frequency separation req quired for subcarriers to remain orthogonal over thhe
symbol interval [0, TN] is 1/TN. [8].
The passband bandw width of each subchannel is 2BN; thus, the passbannd
subchannels in this sy ystem would overlap. Therefore, these overlappinng
subchannels are demodulaated using the receiver structure shown in Figure 3.15.

Fig. 3.15 Block diagram of OFDM communication system

3.5 OFDM Modulation Systems 47

3.5.1 OFDM Implementation Using IFFT/FFT

OFDM is usually implemented using FFT and IFFT. This realization of an OFDM
system is illustrated in Figure 3.15 and can be described as follows. The input
data is first de-multiplexed into N parallel substreams, and each one is mapped to
a real or complex symbol stream using some modulation constellation, e.g., QAM
or PSK [35], [26]. An IFFT is computed on each set of symbols, giving a set of
complex time-domain samples. These samples are then quadrature mixed to
passband in the standard way. The real and imaginary components are first
converted to the analog domain using digital-to-analog converters (DACs); and,
the analog signals are then used to modulate cosine and sine waves at the carrier
frequency, fc, respectively. These signals are then summed to give the transmission
signal, s (t ).
The receiver picks up the signal r (t ) , which is then quadrature mixed down to
baseband using cosine and sine waves at the carrier frequency. This also creates
signals centered on 2fc, so low-pass filters are used to reject these. The baseband
signals are then sampled and digitized using analog-to-digital converters (ADCs),
and a forward FFT is used to convert back to the frequency domain. This returns
N parallel streams, each of which is converted to a binary stream using an
appropriate symbol detector. These streams are then re-combined into a serial
stream, which is an estimate of the original binary stream at the transmitter.

3.5.2 Cyclic Prefix in OFDM Modulation

The filters of OFDM have a non-ideal spectral transfer function. Due to the
dispersive effects of the wireless channel, intersymbol interference (ISI) of the
adjacent OFDM symbol results and degrades the error performance of the system.
Therefore, a cyclic extension is appended to the OFDM symbol to compensate for
these effects. Although a buffer of zeros would suffice in preventing the ISI, the
use of cyclic extension has more advantages [22].
A representation of the use of a cyclic extension is shown in Figure 3.16. This
cyclic extension in called the cyclic prefix (CP). As can be seen, the CP is created
by copying the end of the OFDM symbol and placing that copy at the beginning of
the symbol, for every symbol. The length of the CP is selected according to the
impulse response of the wireless channel. As result of the channel dispersion, the
samples of symbol M-1 spreads to samples of symbol M, while the samples of
symbol M are spreading to the samples of symbol M+1. The CP must have
enough length to capture the interference due to channel dispersion.
48 Chapter 3 Digital Modulation Techniques in MIMO Systems

Fig. 3.16 Cyclic prefix in OFDM modulation: a) addition of CP to OFDM symbols, b)

transmission through the channel, c) removal of CP at receiver

3.5.3 RF Impairment in OFDM Systems

3.5.3.1 Peak-to-Average Power Ratio in OFDM Modulation

The ratio between the peak and average powers of a signal is called the peak-to-
average power ratio (PAPR). Here, the PAPR is extracted for multicarrier
modulation using N carriers [24]. Hence, this calculation may also be used to
obtain PAPR for OFDM modulation.
A multicarrier signal may be considered as the summation of N independent
random voltages (as random variables), each with an average voltage, Vrms , and
2
average power, Vrms . The total normalized average power of a multicarrier signal
is given by:

Pave = ∑Vrms
2
(3.42)
N

Assuming unmodulated sinusoidal carriers with same amplitude, this equation

may be written as:
2
⎛V ⎞ N 2
Pave = ∑ ⎜ max ⎟ = Vmax (3.43)
N ⎝ 2 ⎠ 2
On the other hand, the peak power is the power of the sum of the in-phase voltage
of all the carriers. It can be represents as:

Ppeak = (V1max + V2 max +…+VNmax ) 2 = N 2Vmax

2
(3.44)
3.5 OFDM Modulation Systems 49

The ratio of the signal peak power to its average power is obtained as:
Ppeak
= 2N (3.45)
Pave

The average power of the envelope of the signal is more common in multicarrier
modulation. A multicarrier signal with N unmodulated carriers with uniform
frequency spacing, Δω , each having unity amplitude; and, the zero phase offset
can be written as:
N −1
V = ∑ Vmax Cos(ωt + nΔωt ) (3.46)
n =0

This equation can be also wriiten as [24]:

N Δωt
sin( )
( N − 1)Δωt 2
V = Vmax Cos (ωt + ] (3.47)
2 Δωt
sin( )
2
Therefore, the envelope of a multicarrier modulation is obtained as:

N Δω t
Sin( )
Venv = 2 Vmax (3.48)
Δωt
Sin( )
2

The average power of the envelope is obtained by integration over the envelope
period:

2
⎡ N .Δω.t ⎤
1 ⎢ ⎥
T Sin( )
Penv − ave = ∫⎢ 2 Vmax ⎥ dt = NVmax
2
(3.49)
T 0⎢ Δω.t ⎥
Sin( )
⎢⎣ 2 ⎥⎦

Hence, the envelope peak-to-average ratio is obtained as:

Ppeak
PAPR = =N (3.50)
Penv − ave

The signal peak-to-average ratio (dB) and envelope peak-to-average ratio (dB)
as functions of the number of unmodulated carriers are illustrated in Figure 3.17.
50 Chapter 3 Digital Modulation Techniques in MIMO Systems

30

25
Peak-to-average ratio[dB]

20

15

10
Envelop PAPR
Signal PAPR
5

0
0 50 100 150 200 250 300
Number of carriers

Fig. 3.17 Signal and envelope peak-to-avearge ratios as functions of the number of
unmodulated carriers.

3.5.3.2 Phase Noise in OFDM Modulation

An oscillator is a system that generates a periodic signal with a specified or

controllable frequency. There is phase noise in the output of an RF oscillator. The
output voltage of an oscillator can be represented by [28]:

Vout (t ) = ( A0 + aN (t )).Cos(2π f 0 t + ϕ N (t )) (3.51)

where A0 is the average amplitude of the output signal, f 0 is nominal frequency

of oscillation, and a N (t ) and ϕ N (t ) represent the time-varying components of the
amplitude and phase, respectively. These components are considered as noise,
because an ideal oscillator would have a constant amplitude, A0 , and a phase that
varied at a constant rate, 2π f 0 .
If the amplitude noise can be neglected, the main noise contribution in the
OFDM signal is due to phase noise. Therefore, the SNR performances in an
OFDM system with a noisy oscillator that can be calculated by integrating its
phase noise power spectral density (PSD) function over the channel bandwidth.
The phase noise PDF of an oscillator is approximated by a Lorenzian function
with uniform phase distribution [29], [30]. It is parameterized by its total
3.5 OFDM Modulation Sysstems 551

integrated phase noise, K,

K and a -3dB bandwidth, B, to which we superimpose a
noise floor, L0 (set at 35d
dB above the thermal noise) [27]:

1 K .B
L( f ) 2 = . + L0 (3.522)
π f 2 + B2

Figure 3.18 shows phase noise PSDs for different K and a -3dB bandwidth of 1000
Hz. Therefore, the phase noise
n variance is:
Bs Bs
2 2
1 KB 2K B 1
σϕ = 2 ∫ L( f ) df = 2 ∫ + L0 ) df = arctg ( s ) + L0 Bs (3.533)
2 2
(
0
π 0 f +B22
π 2 B π

Fig. 3.18 Phase noise PDF for

f different K, B = 100Hz

Noise includes componentts at many frequencies; therefore, its phase, with respect to
the main carrier, is rando
om. Figure 3.19 shows noise added to the carrier phasoor.
Random amplitude and ph hase variation due to noise changes the constellation of aan
MQAM modulated signal and degrades the performance of the systems [26].
52 Chapter 3 Digital Modulation Techniques in MIMO System
ms

Phase noise degrades the performance of OFDM signals. Indeed, the OFDM M
signal is impaired by non n-ideal oscillators in both the transmitter and the receiveer.
The first impact of this im mpairment is the common phase error. Moreover, phasse
noise also results in inter--subchannel interference (ICI). This interference developps
from the mutual effects off adjacent channels. This leads to loss of the orthogonalitty
between subchannels. Theese effects are studied in detail in chapter 8.

Fig. 3.19 Noise added to the carrier phasor

3.6 MIMO OFDM Systems

OFDM is effectively used to deal with transmission over frequency-selectivve
fading channels and to av
void relatively complex signal processing in the receiveer.
Accordingly, the idea off multicarrier modulation, or OFDM, can be applied tto
MIMO frequency-selectiv ve channels as shown in Figure 3.20.

x1 y1
I
IFFT1
FFT1

x2 y2
Demod./
Encoder Decoder

xNt yNr
I
IFFTNt FFTN r

Fig. 3.20 Block diagram of a MIMO-OFDM communication system

For each element of the transmitting antenna element, the IFFT of M-arry
symbols must be obtained. This is followed by adding a cyclic prefix to each oof
these sequences, and the resulting signal is transmitted using a suitable RF fronnt-
References 53

end. For each of the received signals, the FFT of the aggregate received signal
(superposition of all the OFDM words from each of the transmitted antennas) is
calculated; and, the cyclic prefix is removed. The resulting set of signals is then
used for demodulation/decoding [14].
As mentioned, the main advantage of OFDM system is the avoidance of
expensive equalization algorithms and achievement of signal transmission at a
reasonable complexity. However, there are some challenges with the use of
MIMO-OFDM. It should be emphasized that MIMO-OFDM symbols are typically
long. This leads to some channel variation over one OFDM word. Hence, this may
destroy the orthogonality of the subcarriers at the receiver and may cause inter-
channel interference, degrading the system performance [14]. Moreover, MIMO-
OFDM suffers from all RF impairments as discussed in the previous section.

References
[1] Shannon, C.E.: A Mathematical Theory of Communication. Bell System Technical
Journal 27, 379–423, 623–656 (1948)
[2] Proakis, J.G., Salehi, M.: Digital Communications, 5th edn. McGraw-Hill, New York
(2008)
[3] Wozencraft, J.M., Jacobs, I.M.: Principles of Communication Engineering (1965)
[4] Sklar, B.: Digital Communications: Fundamentals and Applications. Prentice-Hall
(2001)
[5] Wolfowitz, J.: Coding Theorem of Information Theory, 2nd edn. Springer,
Heidelberg (1964)
[6] Nguyen, H.N., Shwedyk, E.: A First Couse in Digital Communications. Cambridge
University Press (2009)
[7] Rappaport, T.: Wireless Communication, Principles and Practice, 2nd edn. Prentice-
Hall (2002)
[8] Goldsmith, A.: Wireless Communications. Cambridge University Press (2005)
[9] Pahlevan, K., Leveque, A.H.: Wireless Information Networks, 2nd edn. (2005)
[10] Stuber, G.L.: Principle of Mobile Communications. Kluwer Academic Press (2001)
[11] Paulraj, A., Nabar, R., Gore, D.: Introduction to Space-Time Wireless
Communications. Cambridge University Press (2003)
[12] Biglieri, E., Constantinides, A., Goldsmith, A., Paulraj, A., Vincentpoor, H.: MIMO
Wireless Communications. Cambridge University Press (2007)
[13] Tsoulos, G.: MIMO System Technology for Wireless Communications. CRC Press
(2006)
[14] Duman, T.M., Ghrayeb, A.: Coding for MIMO Communication Systems. Wiley
(2008)
[15] Sklar, B.: Defining, designing, and evaluating digital communication systems. IEEE
Communications Magazine 31(11), 91–101 (1993)
[16] Mohammadi, A., Kumar, S.: Characterization of Adaptive Modulators in Fixed
Wireless ATM Networks. KICS/IEEE Journal of Communications and Networks
(JCN) 6(2), 123–132 (2004)
[17] Foschini, G.J., Gans, M.J.: On limits of wireless communications in a fading
environment when using multiple antennas. Wireless Personal Communications 6(3),
311–335 (1998)
54 Chapter 3 Digital Modulation Techniques in MIMO Systems

[18] Simon, M.K., Alouini, M.S.: Digital Communication over Fading Channels, 2nd edn.
John Wiley & Sons (2005)
[19] Lu, J., Letaief, K.B., Chuang, J.C.-I., Liou, M.L.: M-PSK and M-QAM BER
Computation Using Signal-Space Concepts. IEEE Transactions on
Communications 47, 181–184 (1999)
[20] Keshavarzi, M.R., Mohammadi, A., Abdipour, A., Ghannouchi, F.M.:
Characterization of DC Offset on Adaptive MIMO Direct Conversion Transceivers.
IEICE Transactions on Communications (1) (2011)
[21] Mark, J.W., Zhuang, W.: Wireless Communications and Networking. Prentice-Hall
(2003)
[22] Wyglinski, A.M., Nekovee, A., Hous, Y.T.: Cognitive Radio Communications and
Networks, Principle and Practice. Academic Press (2010)
[23] Alamouti, S.M.: A simple transmit diversity technique for wireless communications.
IEEE Journal of Selected Areas of Communications 16(10), 1451–1458 (1998)
[24] Pothecary, N.: Feedforward Linear Power Amplifiers. Artech House (1999)
[25] http://en.wikipedia.org/wiki/Orthogonal_frequency-
division_multiplexing
[26] Madani, M.H., Abdipour, A., Mohammadi, A.: Analysis of performance degradation
due to nonlinearity and phase noise in orthogonal frequency division multiplexing
system. IET Communications 4(10), 1226–1237 (2010)
[27] Madani, M.H., Abdipour, A., Mohammadi, A.: Analytical Performance Evaluation of
the OFDM Systems in the Presence of Jointly Fifth Order Nonlinearity and Phase
Noise. Journal of Analog Integrated Circuits and Signal Processing (2010)
[28] Keshavarzi, M.R., Mohammadi, A., Abdipour, A.: Characterization of Adaptive
Modulation MIMO Systems in the Presence of Phase Noise. In: Wireless
Communications & Mobile Computing Conference, Leipzig, Germany (2009)
[29] Come, B., Ness, R., Donnay, S., Van der Perre, L.: Impact of front-end non-idealities
on bit error rate performance of WLAN-OFDM transceivers. In: IEEE Radio and
Wireless Conf., Rawcon, pp. 91–94 (September 2000)
[30] Schiek, B., Rolfes, I., Siweris, H.J.: Noise in highfrequency circuits and oscillators.
John Wiley & Sons, NY (2006)
Chapter 4
Fundamental Concepts in Wireless Transceiver
Design

The fundamentals related to transceiver design for wireless communications are

presented in this chapter. The main parameters in the transceiver design are
discussed in order to provide a performance metric in the evaluation of wireless
communication systems. The modulation accuracy is evaluated using error vector
magnitude (EVM) to measure the modulation and demodulation accuracy. Some
practical considerations are also discussed.

4.1 Linear Systems and Transformations

A linear time-invariant (LTI) system is characterized by its impulse response. The
output of the LTI system can be expressed in terms of input and impulse response
as a convolution relation:
∞
y (t ) = x(t ) * h(t ) = ∫ x(t − τ )h(τ )dτ (4.1)
−∞

where h(t ) is the impulse response of an LTI system.

The Fourier transform of the impulse response is called the frequency response,
which can be shown as:
∞
− j 2π f c t
H ( f ) = F {h(t )} = ∫ h(t )e dt (4.2)
−∞

As shown in Figure 4.1, the relation between the input and the output of an LTI
system in the frequency domain is presented as:

Y ( f ) = H ( f )X ( f ) (4.3)

where Y ( f ), X ( f ) are the spectra of the input and output signals, respectively.

A. Mohammadi et al.: RF Transceiver Design for MIMO Wireless Communications, LNEE 145, pp. 55–76.
springerlink.com © Springer-Verlag Berlin Heidelberg 2012
56 Chapter 4 Fundamental Concepts in Wireless Transceiver Design

If the system is considered linear time-variant (LTV), the output is obtained as:
∞
y (t ) = x(t ) * h(t ,τ ) = ∫ x(t − τ )h(t ,τ )dτ (4.4)
−∞

where h(t ,τ ) is the impulse response of an LTV system.

x (t )
y (t ) = ∫ x(t − τ )h(τ )dτ
−∞
h (t )

(a)
∞

x (t )
y (t ) = ∫ x(t − τ )h(t ,τ )dτ
−∞

h (t , τ )

(b)

Fig. 4.1 Impulse responses of (a) an LTI system, (b) an LTV system

4.2 Nonlinear Systems

The output of a linear and memoryless system is expressed as:
y (t ) = ax(t ) (4.5)

where a is the linear system gain, x(t) and y(t) are the input and output,
respectively.
For memoryless nonlinear systems, the input-output relation is:
y (t ) = a0 + a1 x(t ) + a2 x 2 (t ) + a3 x3 (t ) + ... (4.6)
where parameters ai are time dependent for time-varying systems.
If a sinusoidal signal, ( x(t ) = A cos(ωo t )) , is applied to a nonlinear system, the
output is expressed as:

a2 A2 3 a A2
y(t ) = a0 + + (a1 A + a3 A3 ) cos(ωo t ) + 2 cos(2ωo t ) + ... (4.7)
2 4 2
4.2 Nonlinear Systems 57

4.2.1 One-dB Compression Point

If higher order terms are neglected rather than the first three terms, we may define
the gain of a nonlinear system as [1]:

3
g  a1 + a3 A2 (4.8)
4
The gain is reduced as a function of A, if a3 is negative. Figure 4.2 shows the
output of a typical nonlinear system.
The 1-dB compression point quantifies this gain reduction factor. This point is
either defined as the input signal level at which g is reduced by 1-dB compared to
the linear gain (input 1-dB compression point, PI −1dB ) or the point at which the
linear output power and the output power of the nonlinear amplifier differ by 1 dB
(output 1dB compression point, PO −1dB ).

Fig. 4.2 One-dB compression point of a power amplifier

According to Figure 4.2, this point is obtained as:

a1
A1dB = .38 (4.9)
a3
58 Chapter 4 Fundamental Concepts in Wireless Transceiver Design

If one defines the root mean square (RMS) of the input power, with respect to a
source resistor, RS, as [2]:

1 A2
PI −1dB = 10log( ) dBm (4.10)
2 Rs

The input power at which the 1-dB input compression point occurs for a single-
tone input is then given as:

.0725 a1
PI −1dB = 10 log( ) dBm (4.11)
Rs a3

4.2.2 Third-Order Intercept Point

When signals of different frequencies are applied to the input of a nonlinear system,
the output contains the combination terms, where are called intermodulation. If we
consider the input signal as:
x(t ) = A cos(ω t ) + B cos(ω t ) (4.12)
1 2

the desired and intermodulation terms are generated at the output of the nonlinear
system (4.6) as:

A2 B 2 3 3
y (t ) = a2 [ + ] + (a1 + a3 A2 + a3 B 2 ) A cos(ω1t )
2 2 4 2
3 3 A2 B2
+ (a1 + a3 B 2 + a3 A2 ) B cos(ω2 t ) + a2 [ ]cos(2ω1t ) + a2 [ ]cos(2ω2 t )
4 2 2 2 (4.13)
3
+ a2 AB cos((ω1 ± ω2 )t ) + a3 A B cos((2ω1 ± ω2 )t ) +
2
4
3 a A3 a B3
a3 B 2 A cos((ω1 ± 2ω2 )t ) + 3 cos(3ω1t ) + 3 cos(3ω2 t )
4 4 4

The term ω1 ± ω2 is second-order intermodulation, and the terms 2ω1 ± ω2 and

ω1 ± 2ω2 are third-order intermodulation terms. If we assume, A=B in (4.11), it
can be seen that the output power of the third-order products increases with the
cube of the input signal, whereas the desired output is proportional to the input
signal. A hypothetical intersection point when the first-order power product (P0)
and the third-order power product (P0IM3) are equal is called the third order
intercept point. This is shown in Figure 4.3.
4.2 Nonlinear Systems 59

POIP 3

PIIP 3

Fig. 4.3 Input and output third-order intercept point

The input amplitude of the third-order intercept point is extracted as:

a1
AIIP 3 = 2 (4.14)
3a3

Using AIIP 3 in relation to A1− dB , a relation between these parameters can be

extracted as:

a1
2
AIIP 3 3a3
= = 3.04 → 9.44dB (4.15)
A1− dB a
.38 1
a3

On the other hand, if two tones are applied to obtain the 1-dB compression point,
it can be shown that the AIIP 3 and A1− dB relation is expressed as [2]:

a1
2
AIIP 3 3a3
= = 5.25 → 14.4dB (4.16)
A1− dB a
.22 1
a3
60 Chapter 4 Fundamental Concepts in Wireless Transceiver Design

The equivalent of the input third-order intercept point (IIP3) of an n-stage

cascaded network is obtained as:

1 1 a12 a12 b12

2
= 2
+ 2
+ 2
+ .... (4.17)
AIIP 3 AIIP 3,1 AIIP 3,2 AIIP 3,3

where AIIP 3,1 ,… are the third-order input intercept amplitude and a1, and b1 are
linear gain coefficients of the corresponding block.

4.2.3 Second-Order Intercept Point

Second-order intermodulation distortion is important in differential circuits [1],
[3]. This parameter is obtained from a hypothetical intersection point of the first-
order product, a1A, and the second-order product, a2A2 , of a nonlinear system.
The amplitude of the second-order distortion is expressed as:
a1
AIIP 2 = (4.18)
a2
The cascaded input second-order intercept point (IIP2) can be written as:
1 1 a1 a1b1
2
= 2
+ 2
+ 2
+ .... (4.19)
AIIP 2 AIIP 2,1 AIIP 2,2 AIIP 2,3

4.2.4 Cross Modulation

Suppose a single-tone sinusoidal signal is added to an amplitude-modulated signal
with modulation index m as:

V (t ) = A cos(ω1t ) + B (1 + m(t )) cos(ω2 t ) (4.20)

If V(t) is applied to the input of a nonlinear system, the output contains the
combination terms. One of the third-order terms can be presented as:

3
AB 2 (1 + 2 m(t ) + m 2 (t )) cos(ω1t ) (4.21)
2

The distorted version of the modulation signal of the carrier of ω 2 has been
transferred to the carrier of ω1 . This is called a cross modulation effect and is due
to a nonlinear circuit.
4.2 Nonlinear Systems 61

4.2.5 AM/AM and AM/PM Distortion

The relation between the input and the output of a linear system in time domain is
represented as [4]:
y (t ) = ax(t − τ ) (4.22)
where a is a constant and τ is a time delay.
The output amplitude has a linear relation with the input amplitude, and the
delay is constant to all inputs. On the other hand, for memoryless nonlinear
systems, the input-output relation is:
y (t ) = a0 + a1 x(t ) + a2 x 2 (t ) + a3 x3 (t ) + ... (4.23)
where parameters ai are time dependent for time-varying systems.
For a third-order nonlinearity, the output signal for an input signal,
( x(t ) = A cos(ωo t )) , is:
3
y (t ) = (a1 A +a3 A3 ) cos(ωo t ) (4.24)
4
So far, it has been assumed that the coefficients of the memoryless system are real.
In reality, most systems exhibit a complex behavior with complex coefficients,
where the nonlinear circuits create nonlinear distortions in both amplitude
(AM/AM conversion) and phase (AM/PM conversion). It is straightforward to
show that a nonlinear circuit causes a deviation in linear characteristics, both in
amplitude and phase. Deviations in amplitude create AM/AM distortion, and
deviation in phase result in AM/PM distortion. Figure 4.4 shows typical AM/AM
and AM/PM conversions.

Example 4.1: Nonlinear Model for PA

Memoryless power amplifier nonlinearity are commonly modeled using
amplitude-dependent AM/AM and AM/PM distortion models. In a memoryless
amplifier, the results of the above conversion only depend on the current value of
a signal. Saleh’s model is a simple model that characterizes the nonlinearity of a
power amplifier [5]. The model is experimentally verified for traveling wave tube
amplifiers (TWTA). In this model, the input and output of an amplifier are:
x(t ) = ρ e jϕ
y (t ) = F ( ρ )e j (θ ( ρ ) +ϕ )
where ρ is the input amplitude, ϕ is the input phase, F ( ρ ) is the output
amplitude, and θ ( ρ ) + ϕ is the output phase. According to Saleh’s model, F ( ρ )
and θ ( ρ ) are obtained as:
α1 ρ β1 ρ 2
F (ρ ) = θ ( ρ ) =
1 + α2 ρ 2 1 + β2 ρ 2
where α1 = 2.1587, α 2 = 1.1517, β1 = 4.0330, and β 2 =9.1040.
62 Chapter 4 Fundamental Concepts in Wireless Transceiver Design

50
HB Software
ADS
45

40
Output Power(dBm)

35

30

25

20

15
0 5 10 15 20 25 30 35
Input Power(dBm)

(a)
-100
HB Software
ADS
-110

-120
Output Phase(Degree)

-130

-140

-150

-160

-170

-180
0 5 10 15 20 25 30 35
Input Power(dBm)

(b)
Fig. 4.4 (a) AM/AM conversion and (b) AM/PM conversion of a typical Doherty power
amplifier obtained by ADS software from Agilent and the in-house harmonic balance
method

4.2.6 Spectral Regrowth

If a power amplifier exhibits nonlinearity, the shape of the input signal at the
output of the transmitter is not preserved. In this case, the spectrum of the output
signal is not limited to the desired bandwidth. This effect is called spectral
regrowth, which can be quantified by the adjacent channel power ratio (ACPR).
4.3 Noise 63

B
S( f )
Ba

fo fc f

Fig. 4.5 Input and output third-order intercept point

In wireless communications, each user is assigned a specific channel width in

the system bandwidth. The interference power from the adjacent channels in the
main channel causes adjacent interference. Interference produced in the adjacent
channel is characterized by ACPR, which is the power in the main channel divided
by the power in the lower or upper adjacent channels.
Suppose S ( f ) denotes the power spectral density of the transmitted signal, B
is the bandwidth of the main signal, Ba is the bandwidth of the adjacent channel,
f c is the center frequency of the main signal, and f o is the offset frequency of the
adjacent signal. According to Figure 4.5, the ACPR is obtained as:

⎧ fc − fo + Ba / 2 S ( f )df ⎫
⎪ ∫ f − f − B /2 ⎪
ACPR = 10log ⎨ c f o+ B /a2 ⎬ (4.25)
⎩ ∫ fc − B /2
⎪ ⎪
c
S ( f )df
⎭

Example 4.2: ACPR of TETRA

The channel width of the Terrestrial Trunked Radio (TETRA) standard is 25 kHz.
The specifications of the ACPR at the adjacent channel is -60 dBc, measured at 25
kHz off from the transmitter center frequency, -70 dBc at 50 kHz, and -74 dBc at
1 MHz. The power is integrated using a bandpass root raised cosine filter with
α = 0.35 and a bandwidth of 18 kHz [6].

4.3 Noise
All signals are contaminated by noise, which degrades the system performance.
The noisiness of a signal is usually specified in terms of the signal-to-noise ratio,
SNR = S / N , where S is the signal power and N is the noise power. This
parameter is generally a function of frequency. The reduction in SNR throughout a
two-port network is characterized by the noise factor.
64 Chapter 4 Fundamental Concepts in Wireless Transceiver Design

SNRI
F= (4.26)
SNRo

where SNRI and SNRO are the input and output signal-to-noise ratios,
respectively.
Figure 4.6 shows a noisy two-port network. An expression for noise factor of a
two-port network, which is related to the reflection coefficients of a two-port
network, is [1]:
2
R Γ s − Γ opt
F = Fmin +4 N 2
(4.27)
Z o (1 − Γ 2 ) 1 + Γ
s opt

where Fmin is the minimum noise factor, RN is the equivalent noise resistance,
and Γ opt is the optimal reflection coefficient corresponding to the optimal source
admittance that provides the minimum noise factor. These parameters are the
device characteristics. In addition, Γ s is the source reflection coefficient.

Zs , NI

Z L , So , N o
SI

Fig. 4.6 A noisy two-port network

4.3.1 Noise Factor of Cascaded Networks

The block diagram of a cascaded connection of the system stages is shown in
Figure 4.7. In this diagram, gi and Fi are the power gain and noise factor,
respectively, of the corresponding stages. The equivalent noise factor for the
cascaded connection of the stages is given by [5]:
⎛ ⎞
L ⎜ ⎟
F − 1 F3 − 1 F −1 ⎟
FT = F1 + 2 + + ... = F1 + ∑ ⎜ l −l1 (4.28)
g1 g1 g 2 ⎜ ⎟
⎜ ∏ gn ⎟
l =2

⎝ n =1 ⎠
where the noise factor is often expressed in decibels. In this case, it is called the noise
figure (NF) and is defined as NF = 10 log( F ) . Since the minimum value for the
noise factor is equal to one, the noise figure of an ideal noiseless network is 0 dB.
4.3 Noise 65

SI , N I F1 F2 FL So , N o
g1 g2 gL

Fig. 4.7 A cascaded connection of the stages with specific power gains and noise factors

On the other hand, the noise temperature is also a measure of the noise of a
system. The parameters of a two-port network are the power, gain g , bandwidth
B , and noise temperature, TE . The noise temperature of the source is considered
to be T0 . By assuming power matching at the input of the network, the input noise
power at temperature T0 may be shown as N I = kT0 B , where k is Boltzmann’s
constant. At a room temperature, T0 = 298K , kT0 = 4 ×10−21 W / Hz or SNRdmin .
Therefore, the output noise power is obtained as:

N o = kB(T0 + TE ) g (4.29)

Accordingly, the relationship between the noise factor, F, and the equivalent noise
temperature can be obtained as:

TE = ( F − 1)T0 (4.30)

Similarly, the equivalent noise temperature of TE of an n-stage cascaded system

is obtained as:
T T
TE = TE1 + E 2 + E 3 + ... (4.31)
g1 g1 g 2

Example 4.3: Cascaded Noise Figure of an LTE System

The components and their gain and noise figures of an LTE direct conversion
receiver is shown in Table 4.1. By using (4.28), the cascaded system noise figure
was obtained to be 9.48 dB [2].

Table 4.1 The Parameters of a Typical LTE Receiver

Parameter Band- Duplexer LNA Mixer Roofing/ VGA

switch Filter
Gain(dB) -1.47 -1.8 16 2.7 32.2 10
NF(dB) 1.47 1.8 2.4 6.8 22.43 19.35
66 Chapter 4 Fundamental Concepts in Wireless Transceiver Design

4.3.2 Capacity Reduction due to Noise Factor of an RF Receiver

A typical radio frequency (RF) receiver, including an antenna, an RF front-end
and a load, is illustrated in Figure 4.8 [11], [12]. The antenna is modeled by a
Thevenin equivalent circuit with impedance z A = rA + jx A . The front-end may
include amplifiers, mixers, and filters. In this model, u is the antenna open-circuit
voltage containing both signal and noise. It can be represented as:
u = hx + n (4.32)
o

where x is the transmitted signal, h is the channel coefficient, and no is the thermal
noise as no ~ CN (0, 4kTo B.rA ) , B is bandwidth in Hertz, and rA = Re[ z A ] . The
output signal may be related to input signal u as:
r ∝ hx + n
where n = no+z, and z is the noise contribution from the front-end. The noise
factor of the front-end can be defined as:
2 2
E[ n ] E[ z ]
F= 2
= 1+ (4.33)
E[ n ] 4kTo B.rA
z =0

The SNR at the output of the front-end is reduced by factor F. The SNR at the
output is obtained as:
S (4.34)
SNRo =
F .N I

where N I = 4kTo B.rA .

z1 z2
ZA
u ⎛ z11 z12 ⎞
⎜ ⎟ ZL
⎝ z21 z22 ⎠

Fig. 4.8 A circuit model for a receiver front-end

This relation can also be presented as:

SNRo = SNRI − NF (4.35)

where NF is the noise figure of the front-end system.

4.4 RF Systems Design Parameters 67

S
According to (3.1), C = W log 2 (1 + ) , which is the capacity is related to the
N
SNR ratio. Therefore, the noisy front-end reduces the system capacity.
In designing a front-end, the general objective is usually to minimize the noise
figure of the front-end.

4.4 RF Systems Design Parameters

A wireless link analysis is performed in the design process to determine some link
parameter, e.g., received power, considering the limitations from signal
specifications and the standardization bodies. For instance, one may obtain the
range of a specific radio using the parameters of transmitter powers, antenna
gains, path loss and required SNR.

4.4.1 Path Loss

The propagation path loss is a function of the frequency operation and is obtained
as:
16π 2 d n
L = 10 log10 ( ) (4.36)
λ2
where L is the propagation path loss; d is distance between the transmitter and the
receiver; n is the path loss exponent, which is usually between 2 and 6; and, λ is
the free space path length.

4.4.2 Link Budget

Certain values for the different parameters are required in a radio link, in order to
achieve a specific quality. The link budget calculation provides an initial level of
estimation of link performance. The link budget equation is expressed as follows:

SNRdmin [dB] = PT + GT + GR − L − 10log10 (kT ) − 10log10 ( B) − NF (4.37)

where PT is the transmit power, GT is the transmitter antenna gain, GR is the

receiver antenna gain, L is the path loss, k = 1.38 × 10−23 and is Boltzmann’s
constant, T is room temperature in Kelvin, B is the receiver bandwidth, and NF is
the noise figure of the receiver.

4.4.3 Receiver Sensitivity

The receiver sensitivity is defined as the minimum received signal power needed
to successfully demodulate the received signal for a certain bit error rate (BER)
68 Chapter 4 Fundamental Concepts in Wireless Transceiver Design

requirement. To be more specific, the minimum signal level at the receiver’s

input, Ssense , which can be detected with an acceptable, SNRdmin is called the
sensitivity of a system. According to the definition of noise factor:

SNRImin [dB] = SNRdmin [dB] + NF (4.38)

where SNRImin [dB] = Ssense [dB] − N I [dB] .

Assuming input power matching and having N I [dB] = −174 + 10 log10 ( B ) , the
sensitivity can be written as:

Ssense [dB] = −174 + 10log10 ( B) + NF + SNRdmin (4.39)

Example 4.4: Receiver Sensitivity for Digital Demodulation

According to (3.4), the follwing relationship governs a digital modulation system:
S E R
= b
N No B

Using this relation in (4.35), the receiver sensitivity for a specific digital
modulation system is obtained as:
Eb
S sense [dB ] = −174 + 10 log10 ( Rb ) + NF +
No min

Eb
where is the minimum value of energy of bit per power spectral density of
No min
noise that provides an acceptable BER for a digital system.

4.4.4 Dynamic Range

The difference between the maximum and minimum signals that can be processed
by a receiver is called the dynamic range. The linear dynamic range is expressed
as a difference between the input signal level that results in 1-dB compression and
the minimum signal level that can be distinguished from the noise. This definition
is more suitable for power amplifier applications. However, for low-noise
amplifiers and mixers, the minimum noise ends at a boundary; and, the maximum
power level for which distortion becomes unacceptable is the other boundary.
In general, the spurious free dynamic range (SFDR) definition is used, which is
the range between the receiver sensitivity and the maximum power level for which
the output third-order intermodulation product is equal to the output noise level.
4.4 RF Systems Design Parameters 69

The lower level depends on the minimum detectable signal. The minimum level is
a signal power that allows for detection with a desired SNR, in order to obtain a
desired BER. This is shown in Figure 4.9. This parameter can be expressed as [1]:

2
SFDR[dB] = ( IIP3[dB] + 174 − 10log10 ( B) − NF ) − SNRdmin (4.40)
3

POIP 3

PO −1dB

SNRdmin

PI −1dB PIIP 3

Fig. 4.9 Dynamic range analysis

4.4.5 ADC Selection

Analog-to-digital converters (ADCs) and digital-to-analog converters (DACs) are
the main blocks of a wireless transceiver. The SNR of an ideal N-bit ADC is given
by [7], [8]:
SNR = 6.02 N + 1.76 dB (4.41)
For practical applications, the difference between ideal and practical SNR values
may be pronounced [10]. If one considers the measured SNR values, the effective
number of bits for a practical ADC is obtained as:

SNRmeasured − 1.76
N= (4.42)
6.02
70 Chapter 4 Fundamental Concepts in Wireless Transceiver Design

4.4.6 Phase Noise

An oscillator is a system that generates a periodic signal with a specified or
controllable frequency. In a practical oscillator, the power distributed in frequency
bands on both sides of the oscillator frequency, rather than all power being
concentrated at the oscillation frequency. These unwanted frequency components
are called oscillator noise. There are the two kinds of noise in an oscillator output,
namely amplitude noise and phase noise. The output voltage of an oscillator can
generally be represented by[7], [8]:

Vout (t ) = ( A0 + aN (t )).Cos(2π f 0 t + φN (t )) (4.43)

where A0 is the average amplitude of the output signal; f 0 is the nominal

frequency of oscillation; and, aN (t ) and φ N (t ) represent the time varying
components of the amplitude and phase, respectively.
Oscillator noise has negative impacts on both the transmitter and receiver sides.
In the transmitter, oscillator noise causes the signal to be transmitted below or
above the desired frequencies. In the receiver, the oscillator phase noise can mix
with unwanted interferences and generate unwanted signals in the intermediate
frequencies. This process is called reciprocal mixing [7].
Reciprocal mixing is shown in Figure 4.10. As can be seen, the desired signal,
A, and the adjacent signal, Aa, are both converted to the same frequency. It is
usually possible to keep the amplitude of the oscillator relatively constant. A
simple phase noise model that agrees with measurements is Leeson’s model [9].
Based on this model, the normalized single-sideband noise spectral density is
obtained as:
FkT fo 2 f
L( f m ) = 10 log10 { [( ) + 1]( c + 1)} (4.44)
PS 2QL f m fm

where fm is the offset frequency in Hz, L(fm) is the noise level at fm in dBc/Hz, f0 is
the center frequency in Hz, QL is loaded quality factor of the oscillator, F is the
noise factor, T is the absolute temperature, PS is the carrier power in dBm, and fc is
the corner frequency for flicker noise in Hz. The L(fm) is specified at a particular
offset frequency from the carrier frequency.

Aa
VRF [Volt ]

fo fo + f m f RF f RF + f m

Fig. 4.10 Reciprocal mixing due to oscillator phase noise

4.5 Modulation Accuracy 71

Leeson’s model tells us that phase noise (at a given offset) improves as both the
carrier power and Q increase. These dependencies make sense. Increasing the
signal power improves the ratio simply because the thermal noise is fixed, while
increasing Q improves the ratio quadratically because the tank’s impedance falls
off as 1/QLfm .

4.5 Modulation Accuracy

4.5.1 Error Vector Magnitude (EVM)

The error vector magnitude (EVM) is a metric that measures the modulation and
demodulation accuracy. It is defined as:

∫ y I ( t ) − y (t )
2

EVM  (4.45)
∫ yI ( t )
2

where y (t ) is the actual signal, and yI ( t ) is the ideal signal. If the difference
between actual signal and ideal signal is assumed due to noise, it is obvious that
the EVM can be expressed as:
1
SNRdB = 10log10 ( ) (4.46)
EVM 2
It should be emphasized that EVM is an overall measure of accuracy including in-
phase / quadrature (I/Q) imbalance, phase noise, carrier leakage, ripple in gain,
nonlinear degradation, and frequency error [13]. The modern wireless systems
have EVM specifications for specific modulation schemes. For example, the
required EVM for wideband code division multiple access (WCDMA) using
quadrature phase-shift keying (QPSK) is less than 17.5% [13].

4.5.2 Effect of I/Q Imbalance on EVM

I/Q imbalance can be modeled using a local oscillator (LO) signal. In this model,
the LO signal is as shown in Figure 4.11. The output signal in this model can be
represented as:

rRF (t ) ⎡ e j 2π f LOt (1 + α e jθ ) + e− j 2π f LOt (1 − α e− jθ ) ⎤

1 (4.47)
rRF (t ) rLO (t ) =
2 ⎣ ⎦
where
rRF (t ) = I (t )Cos(2π f RF t ) − Q(t ) Sin(2π f RF t ) (4.48)

rLO (t ) = Cos(2π f LO t ) + jα Sin(2π f LO t + θ ) (4.49)

72 Chapter 4 Fundamental Concepts in Wireless Transceiver Design

θ
(1 + α )cos(2π f Lo t + )
2 rI
s (t )
rQ
θ
−(1 − α )sin(2π f Lo t − )
2

Fig. 4.11 I/Q imbalance modeling

I/Q imbalance results in image signals. Accordingly, the image suppression, IS,
(or its inverse, image rejection, IRR) is related to EVM as:
(4.50)
EVM = 10 IS /10
where

1 1 − α e − jθ 1 + α 2 − 2α Cosθ
= IRR = = (4.51)
IS 1 + α e jθ 1 + α 2 + 2α Cosθ

4.5.3 Effect of Phase Noise on EVM

Assuming the desired output signal is yI ( t ) = Acos (ωt ) and the actual output
signal is y ( t ) = Acos(ωt + ϕ (t )) , it is straightforward to show that EVM due to
phase noise [14], [15] is:

1 σ2 (4.52)
EVM rms = + 2 − 2exp(− )
SNR 2
where σ is the standard deviation of phase noise.
σ2
Approximation of exp(− ) with Taylor series, (3.61) can be written as:
2
1 (4.53)
EVM rms = + σ2
SNR
EVM variation with phase noise is presented in Figure 4.12.
4.5 Modulation Accuracy 73

13.5
SNR=20 dB

13

12.5
RMS Value of EVM(%)

12

11.5

11

10.5

10
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
RMS value of LO Phase Error(deg)

Fig. 4.12 EVM variation with LO phase noise

4.5.4 Effect of Nonlinearity on EVM

To estimate the effect of nonlinearity on EVM, a histogram of the transmitted
power is usually needed [13l]. However, if the power amplifier of the transmitter
operates well below its 1-dB compression point, the third-order intermodulation
interference will have a major contribution to the EVM among the nonlinear
effects. Accordingly, third-order intermodulation distortion (IMD3) may be used
to calculate the EVM using the approximate expression of [15]:
(4.54)
EVM IM 3  10( PIM 3 − PRF )/10
where PRF is the desired RF power in dBm.

4.5.5 Effect of Carrier Leakage on EVM

Carrier leakage results in DC offset in receiver and can be shown as:

s ( t ) = AI ( t ) cos(θ ( t )) cos(ωc t) - AQ ( t ) sin(θ ( t )) sin(ωc t) (4.55)

+ d I ( t ) cos(ωc t)- dQ (t)sin(ωc t)
74 Chapter 4 Fundamental Concepts in Wireless Transceiver Design

where AI ( t ) and AQ ( t ) the quadrature amplitudes of the I and Q signals, and

θ ( t ) is the corresponding phase angle. The d I ( t ) and dQ (t) are the DC offsets
at the output of the I and Q DACs, respectively. The carrier suppression can be
defined as the average carrier leakage to the average power of signal:
PC − Leak (4.56)
R = 10log10 ( )
Ps
The degradation in terms of EVM is given by:
(4.57)
EVM = 10 R /10

4.5.6 Total EVM

If the different factors that contribute to the degradation of the modulation
accuracy can be considered uncorrelated, the total EVM of the transmission signal
can then be obtained as:
EVM Total = ∑ EVM k2 (4.58)
k

where EVMk is the individual EVM due to any impairment factor that degrades the
signal performance.

Example 4.5: EVM and RCE in WiMAX Systems

The relative constellation error (RCE) is often used instead of EVM in WiMAX
(Worldwide Interoperability of Microwave Access) systems [16]. In general, RCE
and EVM are nearly interchangeable terms. However, RCE describes an EVM
measurement that is calculated over an entire Fixed WiMAX frame. There are two
WiMAX classifications, Fixed and Mobile, based on a subset of the IEEE 802.16
standards and are defined by the WiMAX forum. More specifically, Fixed
WiMAX is based on the orthogonal frequency-division multiplexing (OFDM)
physical layer of the 802.16-2004 specifications, which are sometimes called
IEEE 802.16d [17].
Mobile WiMAX is based on the orthogonal frequency-division multiplexing
access (OFDMA) physical layer of the 802.16e-2005 standard, which is a revision
of the original Fixed WiMAX standard [18]. Mobile WiMAX provides added
functionality, such as base station handoffs, multiple input multiple output
(MIMO) transmit/receive diversity, and scalable fast Fourier transform (FFT)
sizes. Table 4.2 shows a high-level side-by-side comparison of the Fixed and
Mobile WiMAX standards.
The IEEE 802.16-2004 standard prescribes that a Fixed WiMAX transmitter
must have a minimum RCE per each modulation scheme for either a subscriber
unit or a base station, as shown in Table 4.3. The IEEE 802.16e-2005 standard
(8.4.12.3) prescribes that a Mobile WiMAX transmitter must have a minimum
RCE (for either a subscriber unit or a base station) per each modulation scheme, as
shown in Table 4.4.
References 75

Table 4.2 Fixed Versus Mobile WiMAX

Fixed WiMAX Mobile WiMAX

Standard IEEE 802.16 IEEE 802.16e-2005

Multiplexing OFDM OFDMA
FFT size 256 Scalable (512, 1024, …)
Duplexing mode TDD, FDD TDD
Modulation scheme BPSK, QPSK, 16-QAM, and QPSK, 16-QAM, and 64-
64-QAM QAM
Subcarrier spacing 15.625, 31.25, 45 kHz 10.94 kHz
Signal bandwidths 3.5, 7, and 10 MHz 5, 7, 8.75, and 10 MHz
Spectrum 3.5 and 5.8 GHz 2.3, 2.5, and 3.5 GHz

Table 4.3 802.16d Minimum RCE for Various Modulation Types [3], [4]

Burst Type SS RCE (dB) BS (RCE)

BPSK-1/2 -13 -13
QPSK-1/2 -16 -16
QPSK-3/4 -18.5 -18.5
16-QAM-1/2 -21.5 -21.5
16-QAM-3/4 -25.0 -25.0
64-QAM-2/3 -29.0 -29.0
64-QAM-3/4 -30.0 -31.0

Table 4.4 802.16e-2005 Minimum RCE for Various Modulation Types [4]

Burst Type SS RCE (dB) BS (RCE)

QPSK-1/2 -15.0 -15.0
QPSK-3/4 -18.0 -18.0
16-QAM-1/2 -20.5 -20.5
16-QAM-3/4 -24.0 -24.0
64-QAM-1/2 -26.0 -26.0
64-QAM-2/3 -28.0 -28.0
64-QAM-3/4 -30.0 -30.0

References
[1] Tasic, A., Serdijn, W.A., Long, J.R.: Adaptive Low Power Circuits for Wireless
Communications. Springer, Heidelberg (2006)
[2] Rouphael, T.J.: RF and Digital Signal Processing for Software Defined Radio.
Newnes Publisher (2009)
[3] Shaeffer, D.K., Lee, T.: The Design and Implementation of Low-Power CMOS Radio
Receivers. Kluwer Academic Press (1999)
[4] Giannini, F., Leuzzi, G.: Nonlinear Microwave Circuit Design. John Wiley (2004)
76 Chapter 4 Fundamental Concepts in Wireless Transceiver Design

[5] Saleh, A.A.M.: Frequency-independent and frequency-dependent nonlinear models of

TWTA amplifiers. IEEE Transactions on Communications 29(11), 1715–1720 (1981)
[6] Luzzatto, A., Shirazi, G.: Wireless Transceiver Design. Wiley (2007)
[7] Lee, T.H.: The Design of CMOS Radio Frequency Integrated Circuits, 2nd edn.
Cambridge University Press, New York (2004)
[8] Smith, J.R.: Modern Communication Circuits, 2nd edn. McGraw Hill (1998)
[9] Leeson, D.B.: A simple model of feedback oscillator noise spectrum. Proceedings of
IEEE 54, 136–154 (1966)
[10] Kenington, P.: RF and Baseband Techniques for Software Defined Radio. Artech
House (2005)
[11] Rothe, H., Dahlke, W.: Theory of noisy fourpoles. In: Proceedings IRE, vol. 44,
pp. 811–818 (1956)
[12] Domizioli, C.P., Hughes, B.L., Gard, K.G., Lazzi, G.: Optimal Front-End Design for
MIMO Receivers. In: IEEE Global Telecommunications Conference (2008)
[13] Rouphael, T.J.: RF and Digital Signal Processing for Software Defined Radio, A
Multi-Standard Multi-Mode Approch. Elsevier Newnes Publications (2009)
[14] Georgiadis, A.: Gain, phase imbalance, and phase noise effects on error vector
magnitude. IEEE Transactions on Vehicular Technology 53(2), 443–449 (2004)
[15] Liu, R.F., Li, Y.M., Chen, H.Y., et al.: EVM estimation by analyzing transmitter
imperfections mathematically and graphically. Analog Integrated Circuits and Signal
Processing 48(3), 257–262 (2006)
[16] National Instruments, Introduction to WiMAX Transmitter Measurements (2009)
[17] IEEE Standard for Local and Metropolitan Area Networks: IEEE Std. 802.16-2004
[18] IEEE Standard for Local and Metropolitan Area Networks: IEEE Std. 802.16e-2005
Chapter 5
RF Power Amplifier and Linearization
Techniques

The radio frequency (RF) power amplifier (PA) is one of the most critical compo-
nents in designing transmitters in wireless communication systems, and its effi-
ciency dominates the overall efficiency of the transmitter. On one hand, the PA is
most power consuming ; for example, in a cellular phone, the battery life is largely
determined by the power efficiency of the PA. On the other hand, it is desirable to
have the ability to transmit data at the highest possible rate for a given channel
bandwidth, i.e., to have high spectral efficiency [1].

5.1 Introduction
There is considerable industrial interest in producing RF PAs with good linearity
and power efficiency. These two contradictory requirements can be achieved by
using external circuitry to linearize an efficient amplifier [2]. Linearity is the abili-
ty of an amplifier to amplify all signals’ parts by the same amount, so that all sig-
nals are amplified equally and is the most important characteristic of a PA. PAs
must be linear to minimize interference and spectral regrowth. However, PAs gen-
erally have nonlinear behavior and are basically the main sources of distortion and
nonlinearities in the RF transmitter.
Researchers have been focusing on designing more efficient power amplifica-
tion techniques. It has been shown that the power efficiency of traditional amplifi-
cation techniques could be improved at a cost of linearity degradation, which may
not be tolerable by the standards’ requirements. The current state of the art is the
design of a moderately linear PA with the additional implementation of a lineari-
zation technique [3]. To maximize power efficiency, the amplifier should operate
as close to saturation as possible, with the linearization system maximizing the
spectral efficiency in this near-saturated region.
In this context, behavioral modeling and linearization of PAs are critical steps
in designing high-performance power amplification systems for modern wireless
communications infrastructure. Indeed, behavioral modeling of PAs is vital for
performance estimation and system level simulation of the transmitter, which pro-
vides a time and computationally efficient alternative to the physics based model-
ing approach [4].

A. Mohammadi et al.: RF Transceiver Design for MIMO Wireless Communications, LNEE 145, pp. 77–128.
springerlink.com © Springer-Verlag Berlin Heidelberg 2012
78 Chapter 5 RF Power Amplifier and Linearization Techniques

5.2 Transmitter Systems Parameters

The quality of a PA can be characterized by a number of specifications, many of
which are provided in detail in the following sections.

5.2.1 Gain
The gain of an amplifier is the ratio of the output amplitude or power to the input
amplitude or power, which is usually measured in decibels (dB). The power gain
of a PA in dB is given by:

⎛P ⎞
G(dB) = 10log ⎜ out ⎟ (5.1)
⎝ Pin ⎠
where Pin is the input power and Pout is the output power.

5.2.2 Bandwidth
The bandwidth of an amplifier is the range of frequencies for which the amplifier
delivers acceptable performance. A well-accepted metric for performance is the
half power points (i.e., frequencies where the power gain is half of its peak value).
Hence, the bandwidth can be defined as the difference between the lower and up-
per half power points and is also called a 3-dB bandwidth.

5.2.3 Noise Figure

The noise factor, F, is a metric that gives an indication of noise added by the cir-
cuit and is defined as the input signal-to-noise ratio (SNR) divided by the output
SNR, which is given by:
SNRin
F= (5.2)
SNRout
Since the amplifier always introduces noise, the SNR at the output is always less
than that at the input; therefore, the noise factor is always greater than unity.
If the noise factor is expressed in dB, it is called a noise figure (NF) for the
circuit, which is obtained as:
NF (dB) = 10log( F ) = SNRin, dB − SNRout , dB (5.3)

5.2.4 Power Efficiency

Power efficiency in PAs is a metric to quantify the ability of a system to transform the
given input power to useful output power. The most power-consuming component of
5.2 Transmitter Systems Parameters 79

a wireless transmitter is the PA; and, in fact, the PA is the dominant component in de-
termining the overall power efficiency of a wireless transmitter [5].
There are two forms of input power to the PA: direct current (DC) via the bias-
ing circuits to the gate and drain of the transistor, and RF power at the input of the
PA. The RF input power is amplified by transforming the DC power into RF pow-
er at the drain level. Part of the DC energy is dissipated as heat, which reduces the
PA power efficiency.
There are three different definitions of power efficiency in the literature, which
are the total efficiency, the drain efficiency, and the power-added efficiency [6].

5.2.4.1 Total Efficiency

The total efficiency can be obtained as the ratio of the RF power delivered at the
output of the PA to the summation of the RF and DC powers that entered the PA.
The total efficiency is obtained by:
Pout
ηt = (5.4)
Pdc + Pin

where η t is the total efficiency of the PA; and, Pdc , Pin , and Pout are the DC and
RF powers at the input and output of the PA, respectively.

5.2.4.2 Drain Efficiency

Drain efficiency, η D trans. , is defined as the ratio of the RF power to the DC power
at the drain level, which is specific to field-effect transistors (FETs). This efficien-
cy is called collector efficiency in bipolar junction transistors (BJTs). Drain effi-
ciency is given by:
PRF Drain
η D trans. = (5.5)
Pdc Drain

where Pdc Drain and PRF Drain are the DC and RF powers, respectively, at the drain
level of the transistor.
The DC power at the gate level of a transistor is usually very small compared to
the DC power at the drain level. Hence, the DC power at the drain level of the
transistor, Pdc Drain , can be approximated as the total DC power entering the PA,
Pdc . By considering losses in the matching networks, the drain efficiency, η D , is
given as the ratio of the RF output power, Pout , to the DC power that entered the
PA, Pdc , which is expressed as:

Pout
ηD = (5.6)
Pdc
80 Chapter 5 RF Power Amplifier and Linearization Techniques

5.2.4.3 Power-Added Efficiency

The power-added efficiency (PAE), η PAE , demonstrates the ability of the PA to
transform DC power to RF power. PAE can be obtained as the ratio of the added
power, which is the difference between the output RF power, Pout , and the input
RF power, Pin , to the DC power, Pdc . PAE is given by:

Pout − Pin
η PAE = (5.7)
Pdc

5.2.5 P1dB
The 1 dB compression point (P1dB) is a measure of amplitude linearity. The gain
of an amplifier compresses when the output signal level enters the compression
region before it reaches saturation. Higher output power corresponds to a higher
compression point. It is an input (or output) power for which the gain of the PA is
1 dB less than the small-signal gain (ideal linear gain).
The maximum saturation point corresponds to the point where the PA reaches its
output maximum power. This maximum power is called the saturation power
Psat[max] . The 3 dB saturation power Psat[ dB ] corresponds to the power for which the
gain of the PA is 3 dB less than the small-signal gain. Figure 5.1 illustrates the
P1dB, maximum saturation and 3 dB saturation points for a typical PA. To avoid in-
termodulation (IM) problems and distortion, the output power needs to be reduced
below the P1dB. Manufacturers usually back off about 10 dB from the P1dB point.

Fig. 5.1 P1dB and saturation power definition for PAs

5.2 Transmitter Systems Parameters 81

5.2.6 IP3
A third-order intercept point (IP3) is a widely used metric in PAs, which gives in-
formation about the linearity of an amplifier. A higher IP3 means better linearity
and lower distortion generation. It is a theoretical point at which the desired output
signal and the third-order IM (undesired) signal are equal in levels considering an
ideal linear gain for the PA. The theoretical input point is the input IP3 (IIP3) and
the output power is the output IP3 (OIP3).
The two-tone test, as shown in Figure 5.2, is commonly used to determine IP3,
which uses two closely spaced frequencies. When two signals at frequencies f1
and f 2 are incident on an amplifier, the output of the amplifier contains these two
signals, as well as IM products at frequencies mf1 ± nf 2 , where m + n is known as
the order of the IM product. The ratio of third-order IM products and the carrier is
known as third-order IM (IM3). IM3 products are important since their frequen-
cies, 2 f1 − f 2 and 2 f 2 − f1 , fall close to the desired signal, which makes filtering
of IM3 an issue. IM3 for 2 f1 − f 2 is given by:

⎛ P2 f − f ⎞
IM 3(dBc) = 10 log ⎜ 1 2 ⎟ (5.8)
⎜ Pf ⎟
⎝ 1 ⎠
and IM3 for 2 f 2 − f1 is expressed by:

⎛ P2 f − f ⎞
IM 3(dBc) = 10 log ⎜ 2 1 ⎟ (5.9)
⎜ Pf ⎟
⎝ 2 ⎠
where p f1 , p f2 , p2 f1 − f2 , and p2 f 2 − f1 are the power outputs at frequencies f1 , f 2 ,
2 f1 − f 2 , and 2 f 2 − f1 , respectively. IM3 is measured in units of dBc, because it is
calculated relative to the main tone power output.

Fig. 5.2 Third-order intermodulation distortion (IM3)

82 Chapter 5 RF Power Amplifier and Linearization Techniques

5.2.7 PAPR
A metric used to quantify required linearity is called the peak-to-average power
ratio (PAPR), which is the ratio between the peak power Ppeak (related to peak
amplitude) and the average power Pavg (related to mean amplitude) of a signal. It
is also called the crest factor and is given by:

⎛ max(| x(t ) |2 ) ⎞ ⎛ Ppeak ⎞

PAR(dB) = 10 log ⎜ 2 ⎟
= 10 log ⎜ ⎟⎟ (5.10)
⎝ mean(| x(t ) | ) ⎠ ⎜
⎝ Pavg ⎠

5.2.8 Power Back-Off

An amplifier appears linear for sufficiently small departures from its bias condi-
tions. The power back-off is defined as the ratio between the PA’s saturation pow-
er to the RF signal’s mean power. There are two types of power back-off: input
power back-off and output power back-off. The back-off at the input of the PA,
ξi , is obtained by:
ξi = Pi , sat − Pi , mean (dB) (5.11)

where Pi , sat and Pi , mean are the saturation power and mean signal power at the in-
put of the PA, respectively.
Similarly, the back-off at the output of the PA, ξ o , is given by:

ξo = Po, sat − Po, mean (dB) (5.12)

where Po, sat and Po, mean are the saturation power and mean signal power at the
output of the PA, respectively.
The input power back-off and output power back-off are illustrated in Figure 5.1.

5.2.9 ACPR
The adjacent channel power ratio (ACPR) is a critical figure of merit in the evalu-
ation of the IM distortion performance of RF PAs. It is a measure of spectral re-
growth and appears in the signal sidebands. ACPR is defined as the ratio of power
in a bandwidth adjacent to the main channel to the power within the main signal
bandwidth.
The ACPR for the right side of the power spectral density (PSD) can be defined as:
B
f c +Δf +
∫ f c +Δf −
B
2
PSD ( f )df
ACPR( Right ) = B
2
(5.13)
fc +
∫ fc −
B
2
2
PSD( f )df
5.2 Transmitter Systems Parameters 83

Similarly, the ACPR for the left side of the PSD can be obtained as:
B
f c −Δf +
∫ f c −Δf −
B
2
PSD ( f )df
ACPR ( Left ) = B
2
(5.14)
fc +
∫ fc −
B
2
2
PSD ( f )df

where f c is the carrier frequency, B is the bandwidth of the modulated signal,

and PSD ( f ) is the power spectral density at frequency f . Figure 5.3 illustrates
the way that ACPR is calculated.

Fig. 5.3 Adjacent channel power ratio (ACPR)

5.2.10 EVM

The effect of noise, distortion, and in-phase / quadrature (I/Q) imbalance on signal
quality can be analyzed based on its error vector magnitude (EVM) measurement.
The EVM is a common metric for the fidelity of the symbol constellation. Ideally,
a signal sent by a transmitter or received by a receiver would have all constellation
points precisely at the ideal locations. However, several deficiencies in implemen-
tation, such as I/Q imbalance, noise and carrier leakage, cause the deviation of
constellation points from their ideal locations. In other words, EVM is a measure
of how far the actual (measured) points are from the ideal locations. The discre-
pancy between actual and ideal symbol vectors is quantified through an error vec-
tor, as shown in Figure 5.4.
84 Chapter 5 RF Power Amplifier and Linearization Techniques

The EVM is the ratio of the power of the error vector to the power of the refer-
ence vector related to the ideal constellation. The EVM can be defined in decibels
(dB) or percentage (%) as:

⎛P ⎞
EVM (dB ) = 10 log10 ⎜ error ⎟ =
⎝ Pideal ⎠
(5.15)
⎛ ⎛ (I − I )2 + (Qactual − Qideal ) 2 ⎞ ⎞
10 log10 ⎜ mean ⎜ actual ideal2 ⎟ ⎟⎟
⎜ I ideal + Qideal
2
⎝ ⎝ ⎠⎠

Perror
EVM (%) = × 100 (5.16)
Pideal
where Perror and Pideal are the root mean square (RMS) power of the error vector
and the outermost (highest power) point in the reference signal constellation, I ideal
and Qideal are the ideal in-phase and quadrature signals, and I actual and Qactual are
the transmitted in-phase and quadrature signals.
EVM can be also represented as:

N
1
∑S
2
ideal , n − S actual , n
N
EVM = n =1
N
(5.17)
1
∑
2
Sideal , n
N n =1

where N is the number of unique symbols in the constellation.

EVM is useful in quantifying the amount of noise and distortion in a transmit-
ter. The random distribution of the error vectors is caused by noise, while patterns
in the error vector measurements correlated with symbol amplitude and phase may
be caused by distortion.
It is common to characterize AM/AM (amplitude modulation to amplitude
modulation) distortion and AM/PM (amplitude modulation to phase modulation)
distortion, which is shown in Figure 5.5. Variation in the transmitter gain causes
AM/AM distortion. Phase shifting of the carrier correlated to the signal amplitude
introduces AM/PM distortion [7].
Undesired expansion of the output spectrum is also caused by distortion. Non-
linearity of a PA that can be quantified as AM/AM and AM/PM distortion gene-
rates counterfeit energy near the fundamental and harmonics of the carrier fre-
quency. This broadening in the output spectrum is referred to as spectral regrowth
or out-of-band distortion, which deteriorates the ACPR. The transmitter output
spectrum showing spectral regrowth is illustrated in Figure 5.6. Spectral regrowth
is a potentially significant problem and demands special attention during the
transmitter design process. The nonlinearity in the device can be compensated for
using linearization techniques [4], which are covered in Section 5.4.
5.2 Transmitter Systems Parameters 85

Fig. 5.4 Error vector is the difference between the actual and ideal symbol vectors

Fig. 5.5 AM/AM distortion and AM/PM distortion in transmitters

86 Chapter 5 RF Power Amplifier and Linearization Techniques

Fig. 5.6 Spectral regrowth in transmitter output spectrum

5.2.11 Memory Effect

In a two- or multi-tone IM test, if the amplitude or phase of the IM signals is af-
fected by the tone frequency spacing, the PA exposes memory effects. Memory ef-
fects can be explained as time lags between the AM/AM and AM/PM responses of
the PA. Electrical and thermal are two types of memory effects.
The electrical memory effect is produced by poor gate and drain decoupling in
FET and base and collector decoupling in BJT at low frequencies. Electrical
memory effects cause the distortion of the envelope currents and result in IM
asymmetry. The memory effect is more significant for class AB PAs than class A,
with a reduced conduction angle.
Electrical memory effects are generated by non-constant node impedances
within frequency bands. Most of them are produced by frequency-dependent
envelope impedances. The thermal memory effect is generated by the junction
temperature that is modulated by the applied signal envelope.

5.3 RF Power Amplifiers

The wireless and satellite communications communities are always searching for
radios that are more power efficient. However, the power efficiency is greatly
dominated by the efficiency of the RF PA in the transmitting path. For a transistor,
the output power, gain and power efficiency depend on the biasing and matching
conditions, which define the PA’s class of operation. A diagram showing the con-
duction angles, as well as the input drive levels, is presented in Figure 5.7 for the
different classes of power amplifiers: (i) continuously driven class A, AB, B and C
PAs, (ii) saturated PAs and (iii) class D, E and F switching-modes PAs [1].
5.3 RF Power Amplifiers 87

By reducing the conduction angle, the efficiency of the PAs is increased. The effi-
ciency is even higher when operating the PA in a switching mode; however, efficiency
enhancement is accompanied with linearity degradation. The higher the efficiency
when passing from one class to another, the poorer is the linearity performance.

Fig. 5.7 Classes of power amplifiers [1]

Time-varying envelope signals are implemented in most communication sys-

tems that have recently been deployed or are to be deployed in the near future.
PAs are required to operate in their back-off region with such signals to meet the
required linearity. This linearity is defined either by the ACPR or EVM in the con-
text of wireless applications. The back-off level setting is generally a function of
the input signal’s PAPR. Unfavorably, the power efficiency of the PA decreases as
the back-off level increases. Therefore, this leads to the design of very low effi-
ciency amplifiers that need large DC power modules for fixed terminals or shorten
battery life for mobile terminals. In fact, linearization techniques are being used to
improve achievable power efficiency [1].

5.3.1 Linear PAs (Classes A, AB, B, C)

This section discusses several types of classic RF PAs, which are called classes A,
AB, B and C. Power amplifiers, depending on the class of operation, behave with
different levels of nonlinearity and provide different levels of efficiency. These
PAs, in terms of levels of linearity and power efficiency, are described; and, the
tradeoff between the linearity and efficiency is discussed.
A general block diagram of a PA is shown in Figure 5.8, which consists of a
transistor with its biasing circuits and the input and matching networks. The re-
quired current and voltage to bias the transistors are provided by the biasing
circuits. The input and output matching networks are designed to transfer the max-
imum power through the PA.
88 Chapter 5 RF Power Amplifier and Linearization Techniques

The class of operation of the PA relies upon the biasing level of the transistor.
For instance, the DC current consumption variation of the FET transistor is shown
in Figure 5.9, as a function of the gate voltage for a given drain voltage. The DC
current in this figure cuts into two regions. In the first region, the current is zero;
therefore, the transistor is off. In the second region, the current changes almost li-
nearly with the gate voltage. The gate voltage at the edge of these two regions is
described as the pinch-off voltage. The gate bias regions corresponding to class A,
AB, B and C PAs are also illustrated in Figure 5.9.

Fig. 5.8 A general block diagram of a PA

Fig. 5.9 The DC current consumption versus the gate voltage of the FET transistor
5.3 RF Power Amplifiers 89

In class B operation, the transistor starts conducting as soon as the input RF

signal is presented at the input of the PA; in fact, the PA consumes power as long
as the RF signal is present at the input. Operation of the PA in classes A, AB and
C can be achieved by biasing the transistor at the gate level, below or above the
pinch-off voltage. In class A, the transistor is on, even when there is no input RF
signal. Hence, the transistor conducts the input RF signal with no limitation on the
input power level up to the maximum level that the transistor can handle. In con-
trast, in class C, the transistor is off as long as the input RF signal amplitude does
not exceed a power threshold. Class A PAs have a linear power transfer function,
while it becomes nonlinear moving toward class C.
The conduction angle, α , is defined as the angular period of the input RF sig-
nal cycle in which the transistor conducts. It varies between 0 to 2π , and α = π
corresponds to class B operation. The conduction angle is calculated by:

⎛ VP ⎞
α = 2 cos −1 ⎜ 1 − ⎟⎟ (5.18)
⎜ Vgs
⎝ ⎠

where VP is the pinch-off voltage, and Vgs is the gate-source voltage.

It was theoretically proven [8] that the maximum drain efficiency of the PA can
be related to the conduction angle by the following equation:

α − sin(α )
ηmax = (5.19)
⎛α ⎞ ⎛α ⎞
4sin ⎜ ⎟ − 2α cos ⎜ ⎟
⎝2⎠ ⎝2⎠

Fig. 5.10 The maximum drain efficiency versus the conduction angle
90 Chapter 5 RF Power Amplifier and Linearization Techniques

Based on the previous equation, the maximum drain efficiency of the PA in-
creases from 50% for a class A PA with a conduction angle of 2π to 100% for a
class C PA with a conduction angle of 0. This is achieved by decreasing the con-
duction angle. The theoretical maximum drain efficiency of the PA for a conduc-
tion angle between 0 to 2π is demonstrated in Figure 5.10. It can be seen that the
maximum efficiency of the PA increases from class A to class C, where the class
A PA has a linear power transfer function and the class C PA has a nonlinear
power transfer function. Indeed, there should be a tradeoff between the efficiency
and linearity of the PA.
Considering the preceding discussion, the choice of a proper class of operation
for a PA in wireless transmitters is fairly challenging. Signal quality degradation can
be avoided by highly linear PAs, but it results in poor power efficiency. The overall
linearity and efficiency of wireless transmitters are highly affected by the PA, since
it is the dominant source of nonlinearities and power consumption. A class AB PA
supplies an acceptable tradeoff between linearity and power consumption.

5.3.1.1 Class A PA
The class A PA has the highest linearity over the other classes of operation. In the
class A amplifier, the conduction angle is set to α = 360D ; and, current flows con-
stantly and is not cut off during any part of the cycle, acting as a current source.
To achieve high linearity and gain, the amplifier’s base and drain DC voltage has
to be chosen properly so that the amplifier can operate in a linear region. There is
continuous loss of power in the amplifier, since it is constantly carrying current.
The class of operation A is distinguished owing to the fact that it is, in theory, per-
fectly linear, but inefficient.
In a Class A PA, the gate and drain of a FET transistor are biased according to:
VP + V f
Vgg = (5.20)
2
Vmax + Vknee
Vdd = (5.21)
2
where VP is the gate pinch-off voltage, V f is the gate diode forward voltage, Vmax
is the maximum tolerable drain voltage of the transistor, and Vknee is the drain knee
voltage. VP , V f , Vmax , and Vknee are inherent parameters of a FET transistor and
can be collected from the device data sheet.
To achieve the maximum drain current swing without clipping, the maximum
amplitude of the RF gate voltage should be chosen as V f − Vgg , which is illustrated
in Figure 5.11 (a). The input power corresponding to this gate voltage is
called the maximum linear input power. Consequently, the device conducts for the
entire RF cycle. The maximum drain current swing, and thus maximum linear
output power, can be obtained by proper selection of the load impedance. The gate
and drain voltages and drain current waveforms for a class A PA are illustrated in
Figure 5.11, as a function of the angular phase, θ .
5.3 RF Power Amplifiers 91

Fig. 5.11 Class A: (a) gate voltage, (b) drain voltage and (c) drain current waveforms

The drain current shown in Figure 5.11 (c) is the response of the transistor,
which is assumed to be an ideal voltage-dependent current source, to the applied
gate voltage shown in Figure 5.11 (a). The drain voltage is also represented in
Figure 5.11 (b) with the assumption of providing the optimal load for the maxi-
mum linear output power. The gate and drain voltages and current corresponding
to the waveforms in Figure 5.11 can be described as:

Vgs (θ ) = Vgg + vg ⋅ sin θ , where vg = V f − Vgg (5.22)

Vds (θ ) = Vdd − vd ⋅ sin θ , where vd = Vdd − Vknee (5.23)

I ds (θ ) = I dd + id ⋅ sin θ , where id = I dd = I max / 2 (5.24)

A
Thus, the optimal load, Ropt . , for the maximum drain current swing can be calcu-

lated as:

vd Vdd − Vknee V − Vknee

. = = = 2 dd
A
Ropt (5.25)
id I max / 2 I max
92 Chapter 5 RF Power Amplifier and Linearization Techniques

The maximum output power at peak drain efficiency is obtained if the transistor
A A
works on a load line of Ropt . . The required DC power, Pdc ,max , maximum RF out-

put power, PRFA ,max , and peak drain efficiency, η DA,max , for class A operation can be
obtained using the following equations:
Vdd ⋅ I max
PdcA,max = Vdd I dd = (5.26)
2
vd id (Vdd − Vknee ) ⋅ I max
PRFA ,max = = (5.27)
2 4
PRFA ,max (Vdd − Vknee )
η DA,max = A
= (5.28)
P dc ,max 2Vdd

1
if Vknee = 0 , then η DA,max =
.
2
The dissipated power in the transistor, which is the difference between the in-
jected DC power to the amplifier and the RF output power produced by the PA,
can be obtained as:
2π
1 Vmax ⋅ I max
= ∫V (θ ) ⋅ I ds (θ )dθ = Pdc − PRF =
A
Pdiss (5.29)
2π
ds
0 8

5.3.1.2 Class B PA

As shown in Figure 5.9, in the class B PA, the gate of the transistor is biased at the
pinch-off voltage, and the conduction angle is set to α = 180D . Hence, the transis-
tor remains in the active region during half of the RF cycle. A class B PA oper-
ates at zero quiescent current. This half-sine waveform of the drain current reduc-
es the corresponding DC power; therefore, the power efficiency is increased.
In the class B mode of operation, the gate and drain of a FET transistor are bi-
ased using:
Vgg = VP (5.30)

Vmax + Vknee
Vdd = (5.31)
2

The gate and drain voltages and drain current waveforms for a class B PA are de-
picted in Figure 5.12. As in the class A mode of operation, the load-line imped-
ance at the fundamental frequency is chosen to obtain the maximum drain voltage
and current swing. The harmonics of the fundamental frequency are ideally short-
ened to ground at the output of the PA.
5.3 RF Power Amplifiers 93

Fig. 5.12 Class B: (a) gate voltage, (b) drain voltage and (c) drain current waveforms

In the class B mode of operation, injecting a sinusoidal waveform to a transistor

with the gate biased at the pinch-off voltage results in a half-sine waveform of the
drain current that consists of fundamental and even harmonic frequencies. The
drain voltage only includes the fundamental frequency, since all higher harmonic
voltage components are suppressed by proper harmonic termination. The gate and
drain voltage equations are exactly the same as those in (5.22) and (5.23), respec-
tively, but with different Vgg . The drain current is calculated as:

⎧ I ⋅ sin θ 0 ≤θ ≤π
I ds (θ ) = ⎨ max (5.32)
⎩ 0 π < θ ≤ 2π
The amplitude of the DC and the fundamental frequency content of the drain cur-
rent of a class B PA can be calculated with the following equations:
2π
1 I max
I dd =
2π ∫I
0
ds (θ )dθ =
π
(5.33)

2π
1 I max
i d ,1 =
π ∫I
0
ds (θ ) ⋅ sin(θ )dθ =
2
(5.34)
94 Chapter 5 RF Power Amplifier and Linearization Techniques

Assuming the same maximum drain voltage and current as in the class A mode of
operation can be obtained, the peak output power of class B is equal to that of the
class A PA, but at the cost of 6 dB more input power requirement. This results in a
gain that is lower by 6 dB. The reason is that only half of the input RF signal is
amplified by the PA. However, class B mode of operation achieves higher peak
efficiency than class A, due to the lower average DC current, and, hence, the low-
er dissipated DC power.
A class B PA yields maximum power when offered the same optimal load line
as that of a class A PA, which is calculated in (5.25). The required DC power,
PdcB,max , maximum fundamental RF output power, PRFA ,max , and maximum power
efficiency, η DA,max , for the class B operation mode can be calculated based on the
following equations:

Vdd ⋅ I max
PdcB,max = Vdd I dd = (5.35)
π

vd ,1id ,1 (Vdd − Vknee ) ⋅ I max

PRFB , f0 ,max = = (5.36)
2 4

PRFB , f0 ,max π (Vdd − Vknee )

η DA,max = = (5.37)
PdcB,max 4Vdd

π
if Vknee = 0 , then η DB ,max =
= 78.5%
4
The theoretical optimal efficiency of the class B PA is η DB ,max = 78.5% higher
than that of the class A PA η DA,max = 50% , with the cost of reduction in the lineari-
ty of the device.
The dissipated power in the transistor for a class B mode of operation can be
calculated as:
2π
1 4 −π
= ∫V (θ ) ⋅ I ds (θ )dθ = Pdc − PRF = Vmax ⋅ I max
B
Pdiss (5.38)
2π 8π
ds
0

In comparison to (5.29), the dissipated power in the transistor in class B PA is

around 27% of the power dissipation in class A PA for the same maximum output
power.
In conclusion, a class B PA yields the same maximum output power as class A
PA, while producing higher efficiency, but 6 dB less gain. Furthermore, a more
complex load network is required, due to the requirement of proper harmonic fre-
quency termination.
5.3 RF Power Amplifiers 95

5.3.1.3 Class AB PA

A class AB PA compromises between class A and class B operation in terms of ef-

ficiency and linearity. The conduction angle is 180D < α < 360D , so that the transis-
tor remains in the active region for more than half of but less than the full cycle of
the input RF signal. Indeed, the efficiency in the class AB mode of operation is be-
tween 50% and 78.5%. Class AB PAs offer a wider dynamic range than either class
A or B PAs. The linearity of class AB is quite similar to a class A PA before its in-
put power becomes large enough to cause the drain current clipping from the bot-
tom, which is the time when nonlinearity starts and the gain decreases.

5.3.1.4 Class C PA

In the class C mode of operation, the gate bias point is chosen below the pinch-off
voltage of the device. In a class C PA, the conduction angle is α < 180D , and the
transistor remains in the active region for less than half of the RF cycle. Hence, in
order to achieve the maximum transistor drain current, I max , more input power
than in the class B is needed. The class C PA has higher efficiency than classes A,
AB and B, but it is highly nonlinear. It is biased so that the output current is zero
for more than one half of an input sinusoidal signal cycle.
In the class C operation mode, the gate and drain of a FET transistor are biased
according to:
Vgg < VP (5.39)

Vmax + Vknee
Vdd = (5.40)
2

Figure 5.13 shows the gate and drain voltages and drain current waveforms for a
class C PA. The gate and drain voltage equations are the same as those in (5.22)
and (5.23), respectively, but with a different Vgg . The conduction angle in Figure
5.13 (c) can be calculated by:
VP = Vgg + vg ⋅ sin γ (5.41)
π
where, γ = − θ 0 , then,
2

⎛ VP − Vgg ⎞
θ0 = cos −1 ⎜ ⎟⎟ (5.42)
⎜ v
⎝ g ⎠
Assuming an ideal linear relationship between the drain current, I ds , and the gate
voltage, Vgs , as demonstrated in Figure 5.9, the drain current can be specified as:

I max I ⋅V
I ds (θ ) = Vgs (θ ) − max P (5.43)
V f − VP V f − VP
96 Chapter 5 RF Power Amplifier and Linearization Techniques

Fig. 5.13 Class C: (a) gate voltage, (b) drain voltage and (c) drain current waveforms

Based on the class C conduction angle and applying (5.22) to (5.43), the drain cur-
rent can be calculated as:

⎧ I max vg I max (Vgg − VP ) π π

⎪ sin θ + − θ0 ≤ θ ≤ + θ0
I ds (θ ) = ⎨V f − VP V f − VP 2 2 (5.44)
⎪ otherwise
⎩ 0

This equation is valid for all the classes of operation, if vg , Vgg and θ 0 are suita-
bly chosen. In the class C PA, the drain current waveform includes both odd and
even harmonic frequency content. Hence, all higher frequency components need
to be short-circuited at the output to have zero drain voltage at all harmonics ex-
cept at the fundamental frequency.
Using (5.44), the drain current at DC and the fundamental frequency can be de-
rived in a few steps with the following equations:
1
2π
I max ⋅ v g
I dd =
2π ∫I
0
ds (θ )dθ =
π (V f − V p )
(sin θ 0 − θ 0 cos θ 0 ) (5.45)

1
2π
I max ⋅ vg
i d ,1 =
2π ∫I
0
ds (θ ) ⋅ sin(θ )dθ =
π (V f − V p )
(θ 0 − sin θ 0 cos θ 0 ) (5.46)
5.3 RF Power Amplifiers 97

To achieve peak drain current and, accordingly, maximum output power, the re-
quired RF gate voltage amplitude of a class C PA should be selected as:

vg = V f − Vgg (5.47)

Using (5.41) and (5.47), vg can be calculated as a function of θ 0 as:

V f − VP
vg = (5.48)
1 − cos θ 0
By substituting (5.48) in (5.45) and (5.46), the following equations can be
obtained:

I max (sin θ 0 − θ 0 cos θ 0 )

I = (5.49)
π 1 − cos θ 0
dd

I max (θ 0 − sin θ 0 cos θ 0 )

i = (5.50)
π 1 − cos θ 0
d ,1

C
The class C optimal load line, Ropt . , for the maximum drain voltage swing is

calculated as:
vdC,1 π (Vdd − Vknee ) 1 − cos θ 0
. = = ⋅
C
Ropt (5.51)
C
i
d ,1 I max (θ 0 − sin θ 0 ⋅ cos θ 0 )

Biasing the gate deeper in class C decreases the drain current; hence, a larger load
resistor is needed to achieve the maximum voltage swing.
The required DC power, PdcC,max , maximum fundamental RF output power,
,max , and maximum drain power efficiency, η D ,max , for the class C operation
C C
PRF
mode can be calculated based on the following equations:

Vdd ⋅ I max (sin θ 0 − θ 0 cos θ 0 )

PdcC,max = Vdd I dd = ⋅ (5.52)
π 1 − cos θ 0

vd ,1id ,1 I max ⋅ (Vdd − Vknee ) (θ 0 − sin θ 0 cos θ 0 )

, f 0 ,max = = ⋅
C
PRF (5.53)
2 2π 1 − cos θ0

(Vdd − Vknee ) (θ 0 − sin θ 0 cos θ 0 )

C
PRF
η DC,max = = ⋅
, f 0 ,max
(5.54)
P C
dc ,max 2Vdd (sin θ 0 − θ 0 cos θ 0 )

1 (θ 0 − sin θ 0 cos θ 0 )
if Vknee = 0 , then η DC,max =
2 (sin θ 0 − θ 0 cos θ 0 )
98 Chapter 5 RF Power Amplifier and Linearization Techniques

The dissipated power in the transistor for class C operation mode can be calcu-
lated as:
2π
1 Vmax ⋅ I max 2sin θ0 + sin θ0 cos θ0 − 2θ0 cos θ0 − θ0
= ∫V (θ ) ⋅ I ds (θ )dθ = Pdc − PRF = ⋅
B
Pdiss (5.55)
2π 4π 1 − cos θ0
ds
0

5.3.2 Switching-Mode PAs (Classes D, E, F)

In contrast to class A, AB, B and C PAs, where operation in the triode region
should be avoided, class D, E, and F PAs rely on operation in the triode region for
optimal efficiency and output power. In these amplifiers, which are also called
switching-mode amplifiers, the output device is driven by a large square wave
signal. For signals that have mainly phase and frequency modulation, such as qua-
drature phase-shift keying (QPSK) or Gaussian minimum-shift keying (GMSK),
the envelope is constant. This means that a nonlinear high-efficiency PA can be
used to amplify such signals.
The current and voltage waveforms at the drain level of the transistor can be
shaped to reduce the power dissipation and enhance the efficiency of PAs. This
reduction can be achieved by avoiding simultaneous presence of high current and
high voltage at the drain level of the transistor, since the power dissipation is equal
to the product of both current and voltage.
In the ideal case, no power dissipation occurs if the transistor behaves as a loss-
less switch, which makes the transistor operate in two states. In the first state,
when the transistor is on, the voltage across it is zero and the current is high, i.e.,
the transistor behaves as a closed switch during this part of a cycle. In the second
state, when the transistor is off, its current is zero and its voltage is high, i.e., the
transistor behaves as an open switch during this part of a cycle.
The gate and drain voltages and drain current waveforms for an ideal switch-
like transistor are shown in Figure 5.14. One way to achieve the switching
behavior is proper termination of the device at the fundamental and harmonic fre-
quencies, in order to shape the current and voltage waveforms to the switching-
mode operation [9].
The drain voltage and current corresponding to the waveforms in Figure 5.14
can be described as:

I max ∞ 2 I max
I ds (θ ) = +∑ sin(2n + 1)θ (5.56)
2 n = 0 (2n + 1)π

Vmax ∞ 2Vmax
Vds (θ ) = +∑ sin(2n + 1)θ (5.57)
2 n = 0 (2 n + 1)π

Vmax
where = Vdd .
2
5.3 RF Power Amplifiers 99

Fig. 5.14 Ideal switch-like transistor: (a) gate voltage, (b) drain voltage and (c) drain cur-
rent waveforms

The dissipated power in the transistor for ideal switching-mode operation is calcu-
lated as:
2π
1
= ∫V (θ ) ⋅ I ds (θ )dθ = Pdc − PRF = 0
B
Pdiss (5.58)
2π
ds
0

This is because the multiplication of the drain voltage and drain current is zero at
each angular phase. Hence, all the applied DC power is transfer to RF output pow-
er. The DC power is calculated as:
Vmax I max
Pdc = (5.59)
4
However, practically, the RF output power of the transistor consists of the funda-
mental frequency power, desired output power, and unwanted harmonic frequencies
power. The output RF power for different harmonic frequencies can be obtained as:
∞ ∞
1 2Vmax I max
PRF = ∑ Vds , nf0 I ds , nf0 = ∑ (5.60)
n = 0 (2n + 1) π
2 2
n =1 2
100 Chapter 5 RF Power Amplifier and Linearization Techniques

where Vds , nf0 and I ds , nf0 are the drain voltage and current amplitudes of the har-
monic signals. Simply, the following equation can be obtained:
Vmax I max
PRF = (5.60)
4
As a result, the load harmonic impedance network should be organized such that
unwanted harmonic power is rejected, while the drain waveforms are controlled to
keep the dissipated power in the transistor at zero. Consequently, all the DC power
is transferred to the fundamental frequency output power, and the output power ef-
ficiency theoretically becomes 100%.
The switch-like behavior of the PA is administered by the gate biasing and the RF
input signal. The fundamental and harmonic load impedances supplied for the drain
of the transistor shape the drain waveforms, which define the different classes of
switching-mode PAs. Switching-mode PAs can be categorized into two groups as:
• Single branch configurations, such as classes E and F, and inverse class F
PAs; and,
• Balanced or push-pull configurations, such as voltage mode class D and
current mode class D PAs.

5.3.2.1 Class D PA
The increased efficiency of class D PAs is a result of exploiting transistors as switch-
es. In most cases, a class D PA implements a pair of active devices operating in a
push-pull mode and a tuned output circuit. Switching-mode PAs with an output filter
tuned to the fundamental frequency ideally transform all the DC power to fundamen-
tal frequency power, which can be delivered to the load without power losses at the
harmonics. This results in a power output of (8 / π 2 )Vdd2 / R for the transformer-
coupled configuration. The output circuit is tuned to the switching frequency and
ideally removes its harmonic components, which results in a completely sinusoidal
signal that can be delivered to the load. Current is drained only through the transistor
that is on, resulting in a 100% efficiency for an ideal class D PA [6], [10].
Class D amplifiers can be categorized into two groups as:

• Class D voltage-switching PAs (also designated as voltage mode class D,

VMCD); and,
• Class D current-switching PAs (also designated as current mode class D,
CMCD).

Figure 5.15 shows a circuit schematic of a voltage-switching push-pull class D

PA, where Cb is the bypass capacitor, Co is the blocking capacitor, and R is the
load resistance. By using the input transformer, both transistors can be driven with
currents that are 180D out of phase. This can be done by reversing one secondary
winding on the transformer. The RF connection of the transistor output is in paral-
lel configuration as a result of the grounding effect of a bypass capacitor, Cb .
Hence, the equivalent load resistance is equal to 2R for each device.
5.3 RF Power Amplifiers 101

Fig. 5.15 Voltage-switching push-pull class D PA

The voltage and current waveforms are shown in Figure 5.16 for a voltage-
switching push-pull class D PA. The square voltage waveform enables good utili-
zation of the device’s breakdown capability. The maximum amplitude of the half
sinusoidal current through each transistor is I max = π ⋅ I dc .

Fig. 5.16 Voltage-switching class D voltage and current waveforms

102 Chapter 5 RF Power Amplifier and Linearization Techniques

The Fourier series of current waveforms, i1 (ω t ) and i2 (ωt ) , can be obtained as:

⎛ π ∞
cos nωt ⎞
i1 (ωt ) = I dc ⎜1 + sin(ωt ) − 2 ∑ ⎟ (5.61)
n = 2,4,6,... n − 1 ⎠
2
⎝ 2

⎛ π ∞
cos nωt ⎞
i2 (ω t ) = I dc ⎜ 1 − sin(ωt ) − 2 ∑ ⎟ (5.62)
n = 2,4,6,... n − 1 ⎠
2
⎝ 2

The Fourier series of the voltage at a connecting node can be obtained as:

⎛ 4 ⎞ ⎛ ∞ sin nωt ⎞
vtot (ω t ) = Vdd ⎜ ⎟ ⎜ ∑ ⎟ (5.63)
⎝ π ⎠ ⎝ n =1,3,5,... n ⎠

The current passing through the load is given by:

iR (ω t ) = i1 (ω t ) − i2 (ω t ) = π ⋅ I dc ⋅ sin(ω t ) (5.64)

Voltage-switching class D PAs are widely used in audio applications. However,

they have been rarely used at frequencies in the range of 1 GHz or higher. This is
because of the losses associated with the transition time, parasitic reactance and
turn-on resistance of the transistors. At high frequencies, the losses in the output
parasitic reactance, mainly the shunt capacitance, Cds , dominates the loss due to
the turn-on resistance. The energy, E , dissipated within Cds per switching cycle
can be obtained by:

1
E= Cds ⋅ Vdd2 (5.65)
2

where Vdd is the voltage across the transistor at switch closure [11].
To overcome this limitation, current-switching class D PAs have been pro-
posed. In this configuration, a shunt LCR (inductor, resistor, capacitor) output fil-
ter is implemented. In fact, the current waveform becomes a voltage waveform
and vice versa. A series connection transforms to a parallel connection and vice
versa. Furthermore, a voltage-source supply to transformer center-tap converts to a
RF-choke feed (current source) to transformer center-tap.
The use of the current-switching class D PA cancels out the losses in the shunt
capacitance by achieving a zero voltage switching. Therefore, current-switching
class D amplifiers are more appropriate for high-frequency applications resulting
in higher power efficiency.
Figure 5.17 shows a current-switching class D PA, in which the two devices are
connected in parallel. An input transformer is required to drive the devices out-of-
phase, and also an output transformer is required to extract the differential load
5.3 RF Power Amplifiers 103

voltage. As demonstrated in Figure 5.18, the currents through each transistor have
the shape of square waveforms. In order to have proper switching, large input
power is required. Voltages v1 and v2 have the shape of half-sinusoidal wave-
forms. The maximum amplitude of the half–sinusoidal voltage waveforms is
Vmax = π ⋅ Vdd , and high breakdown voltage is required.
The Fourier series of voltage waveforms, v1 (ω t ) and v2 (ωt ) , can be calculated as:

⎛ π ∞
cos nωt ⎞
v1 (ωt ) = Vdd ⎜1 + sin(ω t ) − 2 ∑ ⎟ (5.66)
n = 2,4,6,... n − 1 ⎠
2
⎝ 2

⎛ π ∞
cos nωt ⎞
v2 (ωt ) = Vdd ⎜ 1 − sin(ωt ) − 2 ∑ ⎟ (5.67)
n = 2,4,6,... n − 1 ⎠
2
⎝ 2

The Fourier series of the RLC tank current waveform is given by:

⎛ 4 ⎞ ⎛ ∞ sin nω t ⎞
I tot (ωt ) = I dc ⎜ ⎟ ⎜ ∑ ⎟ (5.68)
⎝ π ⎠ ⎝ n =1,3,5,... n ⎠

The voltage across the load is obtained as:

vload (ωt ) = v1 (ω t ) − v2 (ω t ) = π ⋅ V dd ⋅ sin(ωt ) (5.69)

Fig. 5.17 Current-switching class D PA

104 Chapter 5 RF Power Amplifier and Linearization Techniques

Fig. 5.18 Current-switching class D voltage and current waveforms

5.3.2.2 Class E PA

In a class E PA, a single transistor operates as a switch. The drain voltage wave-
form results from the sum of the DC and RF currents charging the drain current
capacitance. In an ideal class E PA, the drain voltage drops to zero and has a zero
slope when the transistor turns on (zero voltage switching). This eliminates the
losses associated with charging the drain capacitance in class D PAs and consider-
ably decreases the switching losses [12].
In order to optimize the efficiency, the transient response of the load network
has to be controlled even at the time the switching speed is a considerable fraction
of the RF cycle. Indeed, class E PAs can be designed to operate at frequencies
higher than those of class D PAs. Class E PAs can be designed at frequency bands
up to 3.5 GHz, while attaining high power efficiency.
In a switching-mode class E PA, it is possible to have simultaneous high vol-
tage and high current during the switching between on and off. Therefore, a
switching-mode PA with resistive load has power dissipation during switching
transitions. Class E can have good efficiency only if switching times are much
smaller than the waveform period. Figure 5.19 shows current and voltage wave-
forms in a square-wave switching-mode class E amplifier with resistive load.
Switching power dissipation at every instant of time can be obtained as the
multiplication of the voltage and current in the active part of the transistor. The
current and voltage must rise and fall, but not at the same time. In other words,
when the voltage rises, the current must fall and vice versa. Furthermore, time-
displace voltage and current transitions never have high current and high voltage
at the same time. In a class E PA, switching has to be performed as fast as possible
with a reasonable input drive. A class E PA has been utilized for high-efficiency
amplification at frequencies up to K-band [6].
5.3 RF Power Amplifiers 105

Fig. 5.19 Current and voltage waveforms in a square-wave switching-mode class E PA

5.3.2.3 Class F and Inverse Class F PAs

A class F PA is characterized by a load network that has resonances at one or more

harmonic frequencies as well as at the carrier frequency. Class F amplifiers are capable
of high efficiency (88.4% for traditionally defined class F or 100% if infinite harmonic
tuning is used), which results from a low DC voltage current product. In other words,
the drain voltage and current are shaped to minimize their overlap region.
Class F amplifier design is challenging mainly due to the complex design of the
output-matching network [6]. At microwave frequencies, it becomes difficult to
fabricate class F amplifiers, as capacitors and inductors function poorly; and, pla-
nar structures becomes hard to realize, especially when tuning multiple harmonics.
If the even harmonics are terminated to short circuits and the odd harmonics are
terminated to open circuits, the current and voltage waveforms have the shapes of
half-sinusoidal and square waveforms, respectively. The Fourier series of half-
sinusoidal current and square voltage waveforms, both at the same angular veloci-
ty, ω , are given by:
⎛ π ∞
cos nωt ⎞
i (ωt ) = I dc ⎜1 − sin(ωt ) − 2 ∑ ⎟ (5.70)
n = 2,4,6,... n − 1 ⎠
2
⎝ 2
106 Chapter 5 RF Power Amplifier and Linearization Techniques

⎛ 4 4 ∞ sin nωt ⎞
v(ωt ) = Vdc ⎜ 1 + sin(ωt ) + ∑ ⎟ (5.71)
⎝ π π n =3,5,7,... n ⎠

where I dc and Vdc are the DC current and voltage components, respectively [13].
This class of operation is called class F or voltage mode class F. The voltage
and current waveforms of the theoretical class F PA is demonstrated in Figure
5.20 (a). The power dissipation for this class of operation, which is equal to the
product of the current and voltage, is zero. Therefore, a class F PA is able to
achieve 100% efficiency in theory.
Inverse class F or current mode class F PA operation is obtained by terminating
the odd harmonics to short circuits and the even harmonics to open circuits. In
this case, the current and voltage waveforms have the shapes of square and
half-sinusoidal waveforms, respectively. The Fourier series of current and voltage
waveforms are in the form of:

⎛ 4 4 sin nω t ⎞
i (ωt ) = I dc ⎜ 1 + sin(ω t ) + ∑ ⎟ (5.72)
⎝ π π n =3,5,7,... n ⎠

⎛ π cos nω t ⎞
v(ω t ) = Vdc ⎜1 − sin(ωt ) − 2 ∑ ⎟ (5.73)
n = 2,4,6,... n − 1 ⎠
2
⎝ 2

The voltage and current waveforms of the theoretical inverse class F are shown in
Figure 5.20 (b). As in class F operation, the power dissipation of inverse class F
operation is zero; therefore, this class is also capable of achieving 100% efficiency
in theory [13].
However, due to some limitations, the theoretical 100% efficiency of class F and
inverse class F PAs is not possible in practice. In general, an infinite number of
serial odd-harmonic tank resonators can provide a half-sinusoidal current waveform
and a square voltage waveform. Figure 5.21 (a) shows such an output-matching
network with a multiple-resonator output filter that tunes the harmonics to the class
F operational conditions. All the even harmonics are tuned with one parallel tank
resonating at the fundamental frequency. Likewise, the harmonic matching for the
inverse class F operation needs an infinite number of serial even-harmonic tank re-
sonators and one parallel tank resonating at the fundamental frequency. The
harmonic matching network for the inverse class F is shown in Figure 5.21 (b). In-
deed, such a harmonic matching circuit is not feasible in practice.
Berini et al. [14] and Raab [15] showed that the significant effect on the wave-
form shaping and, in fact power efficiency, is only related to the first few harmon-
ics. Tuning more harmonics slightly enhances the efficiency of the PA at the cost
of increasing design complexity and output matching losses. In addition to the
matching networks’ complexity and losses, the intrinsic and extrinsic effects of the
transistor impact the maximum efficiency that can be achieved.
5.3 RF Power Amplifiers 107

Fig. 5.20 Ideal waveforms in switching mode PAs: a) class F and b) inverse class F

Fig. 5.21 Basic circuits of PAs: a) class F and b) inverse class F

108 Chapter 5 RF Power Amplifier and Linearization Techniques

5.3.3 Comparison of the Classes of Operation in RF PAs

Table 5.1 compares the performance of the different classes of operations de-
scribed in the preceding sections, in terms of output power, gain, efficiency, and
linearity. Table 5.1 shows that the efficiency of the linear amplifiers decreases
from class A to class C. However, the high linearity of the class A PA trends to
the high nonlinearity of the class C PA when moving from class A to class C. In
the case of switching-mode PAs, class D and E PAs have very high efficiencies,
but they are strongly nonlinear. The class F PA also has very high efficiency and
is highly nonlinear.

Table 5.1 Performance Comparison for Different Class of Operations of PAs

Class A Class Class Class Class Class Class

AB B C D E F
Out- ++++ +++ ++ + ++ ++ +++
put
power
Gain ++++ +++ ++ + ++ ++ +++
Effi- + ++ +++ ++++ ++++ ++++ +++
ciency
Li- ++++ +++ ++ + + + ++
neari-
ty

5.4 Linearization of RF Power Amplifiers

The nonlinear behavior of the RF front-end, especially RF transmitters, can signif-
icantly degrade the overall performance of wireless systems. The power efficiency
of an RF amplifier is optimal when it is operated near saturation [16]. An amplifi-
er operating in this nonlinear range generates IM distortion that interferes with
neighboring channels. Therefore, there should be compensation for the nonlineari-
ties and distortions of the RF transmitter.
An efficient PA design reduces the cost of power consumption and increases
the battery life of wireless mobile transmitters. However, the increase in efficiency
is usually accompanied by linearity deterioration, which requires a trade-off be-
tween efficiency and linearity.
The output power, gain and efficiency variations for a typical PA as functions
of the input power are shown in Figure 5.22. As can be observed from this figure,
there are two major operational regions for the amplifier. In the first region, the
amplifier has a linear gain, but the efficiency is low. In the second region, the am-
plifier has high efficiency, but the gain is nonlinear. Linearity deterioration is
caused by the distortions introduced to the signal by the compression of the PA in
the saturation region. As a result, the linearity/nonlinearity of the PA depends on
the input signal power [17].
5.4 Linearization of RF Power Amplifiers 109

Fig. 5.22 Output power, gain and efficiency of a PA versus input power

Linearization is a systematic approach to reduce an amplifier’s distortion and is

inevitable for enhancing the linearity of an amplifier to the high input power drive
levels and achieving linearity requirements when operating the device over its en-
tire power range [18]. Linearization allows a PA to generate more power and
operate with higher efficiency for a given level of distortion. There are different
methods for linearizing an RF amplifier. The three major linearization techniques
are feedback, feedforward, and predistortion [19].

5.4.1 Feedback Linearization

The feedback linearization technique is a closed-loop system that can provide high
levels of linearization. It is based on the concept of the feedback loop, which is
widely used in control theory. As shown in Figure 5.23, the amplifier’s output sig-
nal is fed back and subtracted from the amplifier’s input signal to force the output
signal to be a linear replica of the input signal. The use of feedback linearization
has been widely investigated, but has received little use at RF frequencies. The
most important reason is probably related to issues with amplifier stability and the
difficulty in production of networks with non-ideal components that function over
wide frequency ranges [19].
Indirect feedback techniques are reported to be more widely used. The correc-
tion in two alternative indirect feedback strategies, Cartesian feedback and polar
feedback, are processed in a baseband domain. Two loops are needed for the am-
plitude and phase. If only one loop is used for the amplitude at lower frequencies,
the technique is also called an automatic gain control (AGC).
110 Chapter 5 RF Power Amplifier and Linearization Techniques

There are some limitations with feedback linearization techniques that make
them unsuitable for wireless communication applications. The gain of the linea-
rized amplifier is reduced to the gain of feedback loop. Furthermore, the delays
associated with the feedback loop must be small enough to ensure stability, which
limits the use of feedback linearization techniques to narrowband signals [20].

Fig. 5.23 Feedback linearization block diagram

The frequency bandwidth of feedback linearization is limited by the inverse of

the propagation delay of the feedback loop. For third-order distortion correction,
the bandwidth of the loop must be three times the bandwidth of the baseband
envelope. Moreover, at small signals, the system becomes inefficient, and the
modulator/demodulator has a limited dynamic range. Operational amplifiers and
demodulators introduce noise in the system. Stability, noise and orthogonality of
the modulator/demodulator can be improved if the loop is digitized with a digital
signal processor (DSP). However, current DSPs cannot reach the speed of the
equivalent analog systems.
Different approaches have been proposed to address the limitations, including
envelope feedback, polar feedback, Cartesian feedback, adaptive double envelope
feedback, and digital Cartesian feedback. In envelope feedback linearization, in-
put and output samples are fed into two envelope detectors. As illustrated in
Figure 5.24, the error between the outputs of the envelope detectors is injected
to the main signal path via a vector modulator. The use of this error signal to con-
trol the amplifier’s output only allows the compensation for AM/AM nonlineari-
ties of the PA, not its AM/PM nonlinearities.
Hence, a second feedback loop is added in polar feedback linearization, which
is shown in Figure 5.25. In polar feedback linearization technique, the addition of
phase lock loop (PLL) to the envelope feedback linearizer allows the compensa-
tion for both AM/AM and AM/PM nonlinearities of the PA.
5.4 Linearization of RF Power Amplifiers 111

Fig. 5.24 The envelope feedback linearization technique

Fig. 5.25 The polar feedback linearization technique

An alternative approach is Cartesian feedback, which separates the signal into

in-phase and quadrature components. In this technique, the need for phase-shift
components is eliminated; and, correction of gain and phase is still possible by
tuning the amplitudes of two orthogonal components. The baseband in-phase and
quadrature components are compared to control the attenuators in the vector de-
modulator [19].

5.4.2 Feedforward Linearization

Feedforward linearization is a powerful technique that has the same fundamental

error-correcting operation as feedback, but removes the inherent bandwidth and
stability problems of a conventional feedback loop at RF frequencies. The feed-
forward technique is theoretically correct for all nonlinearity orders.
In feedforward linearization, a correction signal is injected at the output of
the PA to cancel its nonlinearity distortions. Figure 5.26 demonstrates the block
112 Chapter 5 RF Power Amplifier and Linearization Techniques

diagram of the feedforward linearization technique. As can be observed from this

figure, there are two loops in the feedforward system, which are the signal cancel-
lation loop and the distortion cancellation loop [6]. In the signal cancellation loop,
a portion of the PA’s output signal is compared to a time-, amplitude- and phase-
aligned replica of the input signal. The signal at the output of the comparator
represents the additive distortion products at the output of the PA. These products,
after being amplified to their original magnitudes, are then subtracted from the
time-aligned version of the PA output, which theoretically leads to perfect cancel-
lation of the distortions generated by the PA [2].

Fig. 5.26 Feedforward linearization block diagram

The feedforward linearization technique is naturally sensitive to changes in op-

erating conditions. However, the development of adaptation methods to compen-
sate for such changes has increased the interest in the technique. Due to the phase
shifts introduced in the carrier by the amplifiers, the feedforward linearizers are
very sensitive to the alignment in both loops [16]. Indeed, delay lines must be pre-
cisely tuned in order to achieve desirable performance. Furthermore, feedforward
linearization is intrinsically non-adaptive and requires adaptive control techniques
in both loops to retain the proper alignment. The gains of the attenuator and aux-
iliary PA must be very well matched to the gain of the PA. Furthermore, the delay
line and subtractor in the output path of the PA have to be low loss to achieve an
effective linearization.
Feedforward linearizers are commonly used in wireless communication base
stations. In these linearizers, very high linearity can be achieved (50 dBc or high-
er), and wide bandwidth can be supported (40-60 MHz or more). Power efficiency
is the major drawback of this technique. Although the main PA is operating at
high efficiency, the overall efficiency of the linearized amplifier is considerably
decreased by the error amplifier, which needs to be perfectly linear. The overall
efficiencies of third-generation (3G) multicarrier PAs linearized by the feedfor-
ward technique are in the range of 10%-15%.
5.4 Linearization of RF Po
ower Amplifiers 1113

5.4.3 Digital Predisstortion Linearization

The predistortion can be performed in the analog or digital domain and on basee-
band, intermediate frequeency (IF) or RF signals. Digital baseband predistortion,
which is an open-loop tecchnique, is currently the preferred linearization techniquue
and is widely used for ap pplications up to 20 MHz bandwidth. In a digital prediis-
tortion (DPD) technique, a complementary nonlinearity upstream of the PA is app-
plied, so that the cascade of the digital predistorter and the PA behaves as a linear
amplification system. Fig gure 5.27 presents a simplified block diagram of the DP PD
linearization technique [4 4]. DPD achieves high linearity performance with higgh
power efficiency. The effificiencies of 3G PAs linearized by digital predistorters arre
reported in the range of 300-45%.
Behavioral modeling of o RF PAs is an essential task in the design of highh-
performance wireless transmitters that combine high power efficiency and spectrum
compliant linearity performmance. This is even more important in modern communni-
cation systems that employ y envelope-varying signals with high PAPRs. Indeed, succh
signals set rough linearity requirements on the PA and usually lead to the forfeit oof
the system’s power efficieency, in order to meet the linearity requirements [1].

Fig. 5.27 Predistortion linearrization technique

The most important ad dvantage of behavioral modeling is that it does not requirre
extensive knowledge of th he RF circuit physics and functionality. As demonstrateed
in Figure 5.28, behavioraal modeling simplifies the modeling of the RF circuit tto
the identification of a matthematical formulation that relates the input and output oof
the device under test (DUUT) that can be considered as a black box. Consequently,
114 Chapter 5 RF Power Amplifier and Linearization Techniques

behavioral modeling appears as a time and resource efficient process for transmit-
ter performance evaluation and digital predistorter design [4].
Due to current broadband and highly varying signals, such as code-division
multiple-access (CDMA), wideband CDMA (WCDMA), andorthogonal frequen-
cy-division multiplexing (OFDM), PAs or wireless transmitters have to be consi-
dered as dynamic nonlinear systems. Therefore, an appropriate architecture has to
be chosen to extract and identify an accurate and robust forward or reverse model
for such systems [1].

Fig. 5.28 Black-box based behavioral modeling

The block diagram of a DPD-based linear amplification system is shown in

Figure 5.29. The signal at the input of the digital predistorer ( xin _ DPD (n) ) can be
derived from the signal at the output of the PA block according to:

xout (n)
xin _ DPD (n) = (5.74)
G

The behavioral modeling of the DUT identifies function f DUT that satisfies
f DUT ( xin (n)) = xout (n) . The synthesis of the DPD function is equivalent to the es-
timation of function f DPD , such that f DPD ( xout (n) / G ) = xin (n) .

Fig. 5.29 Block diagram of digital predistortion (DPD) based linear PA

5.4 Linearization of RF Power Amplifiers 115

Predistorters potentially can handle much wider modulation bandwidths, in-

cluding multicarrier signals. Unlike closed-loop systems, they do not have stability
problems. Predistorters can be integrated in monolithic microwave integrated cir-
cuits (MMICs). They provide good linearity improvement in traveling tube wave
amplifiers (TWTAs), but less significant improvement for metal semiconductor
field-effect transistor (MESFET) classes A and AB. Predistortion linearization
techniques have a smaller dynamic range than feedforward linearization tech-
niques. In addition, predistorters require temperature compensation or an adaptive
control mechanism.
Numerous formulations have been proposed for behavioral modeling and DPD
of RF PAs and transmitters. Some commonly used formulations are memoryless
look-up table, nested look-up table, Volterra, memory polynomial, forward, re-
verse and parallel twin nonlinear two-box, envelope memory polynomial, Wiener
and augmented Wiener, and Hammerstein and augmented Hammerstein models.
The look-up table and memory polynomial models are described in the following
sections. More information about the other models can be found in [4].

5.4.3.1 Look-Up-Table Model

The static look-up table (LUT) model is the basic behavioral model for memory-
less AM/AM and AM/PM nonlinearities. The complex gain of the DUT is stored
in two look-up tables. The output waveform can be shown as:

xout ( n) = G ( xin ( n) ) ⋅ xin ( n) (5.75)

where G ( xin (n) ) is the instantaneous complex gain of the DUT.

The AM/AM and AM/PM characteristics of the DUT can be obtained from the
raw measured data using averaging or polynomial fitting techniques.

5.4.3.2 Memory Polynomial Model

The memory polynomial model is widely used for behavioral modeling and DPD
of PAs and transmitters exhibiting memory effects. The output waveform of
memory polynomial model is given by:
M N
xout (n) = ∑∑ a ji ⋅ xin (n − j ) ⋅ xin (n − j )
i −1
(5.76)
j = 0 i =1

where N and M are the nonlinearity order and the memory depth of the DUT,
respectively; and, a ji are the model coefficients.
Figure 5.30 demonstrates typical spectrum results for a linearized Doherty am-
plifier using a memory polynomial based digital predistorter. This figure clearly
presents the perfect cancellation of the nonlinearities introduced by the DUT.
There are other variations of the memory polynomial model in the literature,
including the orthogonal memory polynomial model and the memory polynomial
model with cross-terms, which is also called as the generalized memory poly-
nomial model [4].
116 Chapter 5 RF Power Amplifier and Linearization Techniques

Input
Output (without DPD)
40 Output (with DPD)

20
PSD(dBm)

-20

-40

2100 2110 2120 2130 2140 2150 2160 2170 2180

Frequency (MHz)

Fig. 5.30 Performance of a DPD linearized 30-Watt GaN Doherty PA with four-carrier
WCDMA signal (PAPR = 11.25 dB) [4]

5.4.4 Analog Predistortion Linearization

RF linearizers based on predistortion have been used for several years in satellite
transponder PAs, either as solid-state power amplifiers (SSPAs) or TTWTAs, to
correct the nonlinear characteristics. The use of predistorters together with the PAs
on board a satellite allows for the operation of its transponders at reduced distor-
tion levels and higher power efficiency, while carrying multicarrier traffic [21].
Analog predistorters (PDs) are still in use for high power applications in the upper
GHz frequency bandwidths, where medium linearization performance is targeted.
PDs have less complexity, low implementation cost, wide bandwidth and the
capability to be added to existing PAs as a separate stand-alone component. Ap-
plications include mobile and handset PAs, which usually requires a simple PD
network, typically a diode gain amplifier [7].
A linearizer based on a series diode with a parallel capacitor is a very simple struc-
ture that utilizes the nonlinearity of the series resistance of the diode, which produces
a characteristic positive gain and negative phase with increasing input power. This is
a simple linearizer, but requires an additional isolation mechanism between the linea-
rization circuit and the PA. Furthermore, it has very limited control on the achieved
characteristics and, hence, has very limited practical applications [22].
The most basic linearization circuit with diodes is shown in Figure 5.31 (a) in a
head-tail configuration, which consists of semiconductor diodes as a nonlinear re-
sistor. This setup suffers from linear phase distortions in the AM/PM characteristic
5.4 Linearization of RF Po
ower Amplifiers 1117

of a PA. The linearizer caan be tuned to resonate at a center frequency to mitigaate

this effect, but the linearizzer is a nonlinear circuit. Therefore, the resonance cannoot
be achieved for the full range of input signals. Furthermore, at microwave free-
quencies, the problem is even
e more severe, as the capacitive effects of the junctioon
capacitance of the Schottk ky diodes limit the application of this linearizer to a verry
moderate range of input signals. This has been addressed in [22].
To overcome this prob blem, the resonance can be attained by utilizing a suitabble
inductance, Ls in shunt, as a illustrated in Figure 5.31 (b). By utilizing this shunt inn-
ductance, the problem of phase distortion can be reduced for any input power annd
any bias voltage, Vsch , of Schuttky diodes D1 and D2 . The actual bias voltage VPPK
across the Schottky diodess D1 and D2 is regulated by the bias feed resistance, Rb ,
T PIN diode, D3 , delivers a variable resistance in shunt
and the input RF power. The
to the biased Schottky diodes
d D1 and D2 . D3 provides a dynamic control to
achieve the required gain and phase characteristics and also provides an improveed
n and phase responses, Vpin . The actual bias of the PIN ddi-
dynamic range in the gain
ode is regulated by the biaas feed resistance, Rb1 , and the input RF power.

Fig. 5.31 (a) The basic linearrization circuit with diodes, (b) an analog linearization circuit

PAs may have single or o dual inflection points in their distortion characteristiccs
based on device technolog gy or bias conditions, as shown in Figure 5.32. A diode-
based RF PD linearizer that can be used to linearize PAs with dual inflectioon
points in their distortion characteristics
c is proposed in [23]. The block diagram oof
the suggested PD is show wn in Figure 5.33. Two linearizer branches connected tto
through and coupled portss of a branch line hybrid coupler. Each linearizer is com m-
posed of two Schottky diiodes connected in an anti-parallel configuration as weell
as a PIN diode in parallell with them. The key point in the design of the PD is thhe
cancellation of the imagin nary part of the admittance seen from connection poinnts
of the diodes in small-signnal operation.
118 Chapter 5 RF Power Amplifier and Linearization Techniques

Fig. 5.32 (a) Single and (b) dual inflection amplitude characteristics

Each linearizer generates a nonlinear reflection coefficient, Γ , that relates the input
signal, a1 , to the output signal, b4 , using the matrix of the hybrid coupler as follows:

⎡ b1 ⎤ ⎡0 1 − j 0 ⎤ ⎡ a1 ⎤
⎢b ⎥ ⎢1 0 0 − j ⎥⎥ ⎢⎢Γ ⋅ b2 ⎥⎥
⎢ 2⎥ = 1 ⎢ ⋅ (5.77)
⎢b3 ⎥ 2 ⎢− j 0 0 1 ⎥ ⎢ Γ ⋅ b3 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣b4 ⎦ ⎣ 0 −j 1 0⎦ ⎣ 0 ⎦

where a2 = Γ ⋅ b2 and a3 = Γ ⋅ b3 .
Further manipulation of the above matrix results in:
b4 = −Γ ⋅ a1 (5.78)

Considering the input power | a1 |2 and output power | b4 |2 , the following rela-
tionship can be derived:

Pout =| Γ |2 ⋅Pin (5.79)

Fig. 5.33 Analog predistorter architecture for linearization of the distortion with dual inflec-
tion point
5.4 Linearization of RF Power Amplifiers 119

This equation shows that the PD gain is equal to the square value of the reflection
coefficient seen from both linearizer branches.
Figure 5.34 shows the performance of the PD in suppressing the distortion of
the PA and flattening of the nonlinear part of the AM/AM characteristic of the PA.
It is obvious that the distortion is considerably compensated for by the PD.

44

43.5

43

42.5
Gain (dB)

PA AM/AM without PD
42 PA AM/AM with PD

41.5

41

40.5

40
-30 -25 -20 -15 -10 -5 0
Input Power (dBm)

Fig. 5.34 Measured AM/AM of AP603 with PD and without PD versus the input power [23]

5.4.5 Comparison of Linearization Techniques

Table 5.2 compares the performance of feedback, feedforward, and predistortion
linearization techniques, in terms of frequency bandwidth, linearity, complexity,
power efficiency, and adaptation.

Table 5.2 Performance Comparison of Different Linearization Methods

Feedback Feedforward Analog / Digital

Predistortion
Frequency Narrow Wide Ultra/Medium
Bandwidth
Linearity Good Ultra Medium/Ultra
Complexity Medium High Medium/Medium
Power High Low High/High
Efficiency
Adaptation Intrinsically Intrinsically non- Intrinsically non-
adaptive adaptive adaptive
120 Chapter 5 RF Power Amplifier and Linearization Techniques

5.5 RF Transmitter Architectures

Power efficient and linear and linear transmitters, such as polar, linear amplifica-
tion with nonlinear components (LINC), envelope elimination and restoration
(EER) and delta-sigma transmitters have received increased attention. These ad-
vanced transmitter architectures are theoretically highly efficient and linear, even
for sophisticated digital modulation signals with non-constant envelopes and large
PAPRs.

5.5.1 Polar Transmitter Architecture

In the polar transmitter, the in-phase (I) and quadrature (Q) baseband signals are
transformed from the Cartesian representation to the polar representation. The
Cartesian domain signal representation, S (t ) = I (t ) cos ωc t + Q(t ) sin ωc t , can be
transformed to the polar domain as S (t ) = A(t ) cos(ωc t + ϕ (t )) . The transformation
can be done with the following equations:

A(t ) = I 2 (t ) + Q 2 (t ) (5.80)

ϕ (t ) = arct tan(Q (t ) I (t )) (5.81)

A major difference between the Cartesian and polar representations is that the po-
lar basis vectors (amplitude and phase) have widely differing spectral properties
compared to the I/Q basis vectors and the modulated RF output. This is clear in
(5.80) and (5.81) as A(t ) and ϕ (t ) are derived from I (t ) and Q (t ) through non-
linear operations. The nonlinearities in (5.80) and (5.81) cause the polar basis vec-
tors to lose the bandlimited property of the Cartesian I/Q representation. To design
the system for high integrity, low EVM and high spectral fidelity, the wideband
nature of the amplitude and phase paths has to be considered in the architecture
and circuit-level design [24].
In a polar transmitter, the phase, ϕ (t ) , and amplitude, A(t ) , components of the
signal are processed separately. Figure 5.35 shows a general structure of the polar
transmitter. The phase of the carrier is modulated, and it then passes through an
amplifier, in which the value of the power supply varies proportionally to the
envelope. The PA is shown as a variable-gain amplifier (VGA) in Figure 5.35.
Generally, any signal that modulates the amplitude of the carrier can be em-
ployed in a polar transmitter. The phase information extracted from the original
signal with a constant or non-constant envelope is converted to a constant
envelope signal. This can be achieved by phase modulation with the help of a
phase lock loop (PLL) or digital phase lock loop (DPLL) to output the desired
transmitting frequencies [25]. The output signal can now be amplified without any
concern of distorting the amplitude information.
5.5 RF Transmitter Architectures 121

Fig. 5.35 General structure of the polar transmitter

5.5.2 LINC Transmitter Architecture

The principle of the linear amplification with nonlinear components (LINC) am-
plification system is based on converting the highly varying amplitude modulation
of the input signal, Sin (t ) , into phase modulations of two constant envelope sig-
nals, S1 (t ) and S2 (t ) , which can be obtained using the following equations:

S (t ) = r (t ) ⋅ e j ⋅φ ( t ) = S1 (t ) + S2 (t ) (5.82)

r (t ) = rmax ⋅ cos(θ (t )) (5.83)

hence,

⎧ rmax j (φ ( t ) +θ ( t ))
⎪⎪ S1 (t ) = 2 ⋅ e
⎨ (5.84)
⎪ S (t ) = rmax ⋅ e j (φ ( t ) −θ ( t ))
⎪⎩ 2 2
where rmax represents the maximum of r (t ) , φ (t ) is the phase of the baseband
signal, and θ (t ) is the additional phase modulation angle related to the amplitude
of signal, r (t ) , which is given by:

⎛ r (t ) ⎞
θ (t ) = arccos ⎜ ⎟ (5.85)
⎝ rmax ⎠
The resultant signal, Sout (t ) , is a linearly amplified version of the input signal,
Sin (t ) :

S out (t ) = 2 ⋅ G ⋅ S in (t ) (5.86)
where G is the gain of the branch amplifier.
122 Chapter 5 RF Power Amplifier and Linearization Techniques

Another way to compute S1 (t ) and S2 (t ) can illustrated as:

⎧ 1
⎪⎪ S1 (t ) = 2 S (t ) [1 + j ⋅ e(t )]
⎨ (5.87)
⎪ S (t ) = 1 S (t ) [1 − j ⋅ e(t )]
⎪⎩ 2 2
where e(t ) is obtained by:
2
rmax
e(t ) = −1 (5.88)
r 2 (t )
Considering that S1 (t ) and S2 (t ) have constant envelopes, they can be efficiently
amplified by means of power efficient or switching-mode nonlinear PAs. The two
amplified signals are then combined to retrieve a linearly amplified version of the
original amplitude modulated input signal. Therefore, RF PAs can be operated at
saturation, which results in maximum power efficiency.
The structure of a LINC amplification system is shown in Figure 5.36. A vector
representation of the separated baseband components is also given in Figure 5.37.
There are three main components in the LINC transmitter: the signal separator, the
nonlinear amplifiers, and the signal combiner [5].
Successful implementation of a LINC transmitter is highly dependent on the preci-
sion and control of signal separation, because the LINC architecture is sensitive to
amplitude and phase balance. Although there are some analog signal component se-
paration implementations, digital signal component separation implementation offers
the flexibility required to apply a control over the precision of the calculations. A dig-
ital signal separator also allows implementation of correction algorithms to compen-
sate for any residual imbalance in the analog sections. It is possible to implement the
signal separation by implementation of (5.85) or (5.88).

Fig. 5.36 LINC transmitter structure

5.5 RF Transmitter Architectures 123

Fig. 5.37 LINC vector decomposition

There are a number of issues that must be addressed in digital implementation

of the signal component separation block, such as the use of a look-up table tech-
nique, the bit resolution of the computations, the required memory resources, the
required computation speed, and the choice of digital-to-analog converters [5].
The PAs should be designed for the highest possible efficiency at saturation,
based on the selection of the PA’s biasing and impedance matching circuits. For
this reason, class F and inverse class F PA designs can be used, which have power
efficiencies of around 80%. While the load changes from matched to very high
impedance, the PAs should provide a rapid drop-off in DC power consumption in
an ideal case.
The bandwidth response of the PA needs to be carefully selected, as the LINC
separated signals have larger bandwidth than the original envelope-modulated sig-
nal, due to added phase modulation. Additionally, the power amplification block
should be made of two identical or quasi-identical amplifiers, in order to preserve
the amplitude and phase balance between the two branches [5].
The use of two highly efficient PAs operated with constant envelope signals in
the LINC system does not guarantee that the overall efficiency of the LINC
transmitter will be high, but the available power at the output of the combiner de-
termines the overall efficiency of the LINC transmitter. In [26], it is shown that
the average efficiency of the LINC system also depends on the power distribution
function (PDF) of the signal and the type of combiner used. Hence, the combining
structure is a key factor in the overall power efficiency and in the linearity of the
LINC transmitter. The linearity depends on how the combining structure affects
the impedances seen by the amplifiers and how it combines the output signals.
In this framework, the combining structures can be categorized into two
classes: 1) matched and isolated combiners, which are also referred to as hybrid or
resistive combiners; and, 2) non-matched and non-isolated combiners, which are
124 Chapter 5 RF Power Amplifier and Linearization Techniques

also referred to as out-phasing or Chireix combiners [5]. The overall system effi-
ciency, η LINC , of a LINC transmitter using an isolated combiner is given by:

η LINC = ηPA
max
⋅ηc (5.89)

where η PAmax
is the maximum amplifier efficiency, and ηc is the combiner efficien-
cy. This equation concludes that the LINC system efficiency drops rapidly for
low-level signals. The LINC system efficiency can drop as low as η PA max
/10 at
power back-off operation of 10 dB.
In order to make the LINC transmitter applicable for implementation in base
stations, some limitations in the LINC transmitter need to be addressed: the gain
and phase imbalance between the two branches; and, the bandwidth of the con-
stant envelope signals feeding the PAs. The bandwidths of the phase-modulated
signals at the output of the signal separator are generally five times wider than that
of the original input signal.

5.5.3 EER Transmitter Architecture

The unmatched efficiency of switching-mode PAs has motivated attempts to use
them for linear amplification. Envelope elimination and restoration (EER), pro-
posed by L. R. Kahn in 1952, is one such attempt and is presented in Figure 5.38.
A modulated RF signal is split into its polar components, the amplitude signal and
the constant amplitude phase-modulated signal by an envelope detector and a limi-
ter, respectively. The limiter output is a constant envelope signal that can be
amplified by a highly nonlinear but power efficient PA, usually operating
in switching mode, ideally without adding significant AM/AM and AM/PM
distortion.
In the envelope path, the amplitude information is extracted by utilizing an
envelope detector, which is used to modulate the supply voltage of the PA using
the amplitude amplifier / bias modulator. In the phase path, a limiter is utilized to
eliminate the amplitude signal variation, which generates the constant amplitude
phase-modulated signal to be applied to the PA [6]. The amplitude information is
restored to the envelope of the transmitted signal by modulating the supply voltage
of the PA, which leads to the EER distinction.
The most significant challenges the EER transmitter faces are mainly related to
the bandwidth of the signals to be amplified. The bandwidths of the phase and
envelope signals are five times wider than that of the input signal. In fact, both the
amplitude and RF power amplifiers should operate over a wide bandwidth, which
inevitably reduces their efficiency. The performance of the EER transmitter is sig-
nificantly affected by the synchronization between the envelope and the RF signal
paths [27]. The current state of the art is minimization of the worst-case differen-
tial delay between the envelope and RF signal paths.
5.5 RF Transmitter Architectures 125

Fig. 5.38 EER transmitter structure

5.5.4 Delta-Sigma Transmitter Architecture

A delta-sigma ( ΔΣ ) modulation based transmitter is another advanced amplifica-
tion system that transforms the envelope-varying signal to a constant envelope
signal and is used with high-efficiency PAs [28]. Figure 5.39 shows the block dia-
gram of an RF delta-sigma based transmitter. Two delta-sigma modulators
(DSMs), working at frequency, f s , are implemented to produce bi-level signals
for I and Q signals; and, two high-frequency multiplexers, working at frequen-
cy, Nf s , are used to up-convert the baseband signals to the carrier frequency,
f c = Nf s .
With the use of a third multiplexer, working at 2 f s , the modulated signals pro-
duced by the two multiplexers are combined to generate I and Q signals at carri-
er frequency f c . At the output of the third multiplexer, a switching-mode PA is
employed to amplify the bi-level I / Q signal. Prior to transmitting the signal
through the antenna, a bandpass filter is used to suppress all out-of-band distortion
and also to recover the modulated signal around the carrier frequency [1].
The general structure of a DSM is demonstrated in Figure 5.40, in which a quan-
tizer embedded in a loop with a digital-to-analog converter (DAC) in the feedback
path is shown. An integrator is in the forward path of the modulator. The input sig-
nal to the integrator is the difference between the input signal, x(t ) , and the quan-
tized output value, y (t ) , converted back to the predicted analog signal, yˆ(t ) .
Considering that the DAC is ideal and signal delays are negligible, the differ-
ence between the input signal, x(t ) , and the fed back signal, yˆ(t ) , at the integra-
tor input is equal to the quantization error. The quantization noise is denoted by
the additive term, E (t ) . This error is added up in the integrator and then quantized
by a 1-bit analog-to-digital converter (ADC).
126 Chapter 5 RF Power Amplifier and Linearization Techniques

The output of a DSM is illustrated in the z -domain by:

Y ( z ) = H ( z ) X ( z ) + G( z ) E ( z) (5.90)

where X ( z ) , Y ( z ) and E ( z ) denote the z -transforms of the input signal, the

output signal and the quantization error, respectively.
The signal transfer function, H ( z ) , transforms the signal at the desired fre-
quency band; and, the noise transform function, G ( z ) , suppresses the quantization
noise in this band [29].

Fig. 5.39 Block diagram of the all-digital delta-sigma transmitter

In the case of a first-order DSM, which is shown in Figure 5.40, the z -domains of
the signal and noise transfer functions can be obtained by:

H ( z ) = z −1 (5.91)

G( z ) = (1 − z −1 ) (5.92)

There are two DSM-based transmitter architectures in the literature to transform

baseband pulsed signals to the desired RF frequency band: the low-pass (LP)
DSM-based and the bandpass (BP) DSM-based transmitter architectures [30].

Fig. 5.40 A first-order low-pass DSM general structure

References 127

References
[1] Ghannouchi, F.M.: Power Amplifier and Transmitter Architectures for Software De-
fined Radio Systems. IEEE Circuits and Systems Magazine 10(4), 56–63 (2010)
[2] Larose, C.L., Ghannouchi, F.M.: Optimal Adaptation Methods and Class of Opera-
tion: Keys to Improving Feedforward Amplifier Power Efficiency. IEEE Transactions
on Vehicular Technology 54(2), 456–467 (2005)
[3] Taijun, L., Bumaiza, S., Ghannouchi, F.M.: Augmented Hammerstein Predistorter for
Linearization of Broad-Band Wireless Transmitters. IEEE Transactions on Micro-
wave Theory and Techniques 54(4), 1340–1349 (2006)
[4] Ghannouchi, F.M., Hammi, O.: Behavioural Modeling and Predistortion. IEEE Mi-
crowave Magazine 10(7), 52–64 (2009)
[5] Birafane, A., El-Asmar, M., Kouki, A.B., Helaoui, M., Ghannouchi, F.M.: Analyzing
LINC Systems. IEEE Microwave Magazine 11(5), 59–71 (2010)
[6] Raab, F.H., Asbeck, P., Cripps, S., Kenington, P.B., Popovic, Z.B., Pothecary, N.,
Sevic, J.F., Sokal, N.O.: Power Amplifiers and Transmitters for RF and Microwave.
IEEE Transactions on Microwave Theory and Techniques 50(3), 814–826 (2002)
[7] Cripps, S.C.: RF Power Amplifiers for Wireless Communications, 2nd edn. Artech
House, Norwood (2006)
[8] Kenington, P.B.: High-Linearity RF Amplifier Design. Artech House, Norwood
(2000)
[9] Ebrahimi, M.M., Helaoui, M., Ghannouchi, F.M.: Trading-off Stability for Efficiency
in Designing Switching-Mode GaN PAs for WiMAX Applications. In: Proc. IEEE
Microwave Conference 2009 (APMC 2009), Asia Pacific, pp. 2348–2351 (December
2009)
[10] Grebennikov, A.: RF and Microwave Transmitter Design. John Wiley & Sons (2011)
[11] Hung, T.P.: High Efficiency Switching-Mode Amplifiers for Wireless Communica-
tion Systems: ProQuest (2008)
[12] Raab, F.H.: Class-E, Class-C, and Class-F power amplifiers based upon a finite num-
ber of harmonics. IEEE Transactions on Microwave Theory and Techniques 49(8),
1462–1468 (2001)
[13] Grebennikov, A., Sokal, N.O.: Switchmode RF Power Amplifiers: Newnes (2007)
[14] Berini, P., Desgagne, M., Ghannouchi, F.M., Bosisio, R.G.: An Experimental Study
of the Effects of Harmonic Loading on Microwave MESFET Oscillators and Am-
plifiers. IEEE Transactions on Microwave Theory and Techniques 42(6), 943–950
(1994)
[15] Raab, F.H.: Maximum Efficiency and Output of Class-F Power Amplifiers. IEEE
Transactions on Microwave Theory and Techniques 49(6), 1162–1166 (2011)
[16] Larose, C.L., Ghannouchi, F.M.: Optimization of Feedforward Amplifier Power Effi-
ciency on the Basis of Drive Statistics. IEEE Transactions on Microwave Theory and
Techniques 51(1), 41–54 (2003)
[17] Bassam, S.A., Helaoui, M., Ghannouchi, F.M.: Crossover Digital Predistorter for the
Compensation of Crosstalk and Nonlinearity in MIMO Transmitters. IEEE Transac-
tions on Microwave Theory and Techniques 57(5), 1119–1128 (2009)
[18] Hammi, O., Ghannouchi, F.M.: Power Alignment of Digital Predistorters for Power
Amplifiers Linearity Optimization. IEEE Transactions on Broadcasting 55(1),
109–114 (2009)
128 Chapter 5 RF Power Amplifier and Linearization Techniques

[19] Katz, A.: Linearization: Reducing Distortion in Power Amplifiers. IEEE Microwave
Magazine 2(4), 37–49 (2001)
[20] Cardinal, J.S., Ghannouchi, F.M.: A New Adaptive Double Envelope Feedback
(ADEF) Linearizer for Solid State Power Amplifiers. IEEE Transactions on Micro-
wave Theory and Techniques 43(7), 1508–1515 (1995)
[21] Ghannouchi, F.M.: An S Band RF Digital Linearizer for TWTAs and SSPAs. In:
European Conference on Circuit Theory and Design (ECCTD 2009), Antalya, Tur-
key, pp. 735–738 (August 2009)
[22] Hashmi, M.S., Rogojan, Z.S., Ghannouchi, F.M.: A Flexible Dual-Inflection Point RF
Predistortion Linearizer for Microwave Power Amplifiers. Progress in Electromag-
netics Research C 13, 1–18 (2010)
[23] Rezaei, S., Hashmi, M.S., Dehlaghi, B., Ghannouchi, F.M.: A Systematic Methodol-
ogy to Design Analog Predistortion Linearizer for Dual Inflection Power Amplifiers.
In: International Microwave Symposium (IMS 2011), Baltimore, Maryland, USA
(August 2011)
[24] Nagle, P., Burton, P., Heaney, E., McGrath, F.: A wide-band linear amplitude mod-
ulator for polar transmitters based on the concept of interleaving delta modulation.
IEEE Journal of Solid-State Circuits 37(12), 1748–1756 (2002)
[25] Staszewski, R.B., Wallberg, J.L., Rezeq, S., Hung, C.M., Eliezer, O.E., Vemulapalli,
S.K., Fernando, C., Maggio, K., Staszewski, R., Barton, N., Lee, M.C., Cruise, P.,
Entezari, M., Muhammad, K., Leipold, D.: All-digital PLL and transmitter for mobile
phones. IEEE Journal of Solid-State Circuits 40(12), 2469–2482 (2005)
[26] Birafane, A., Kouki, A.: On the Linearity and Efficiency of Outphasing Microwave
Amplifiers. IEEE Transactions on Microwave Theory and Techniques 52(7),
1702–1708 (2004)
[27] Hammi, O., Helaoui, M., Ghannouchi, F.M.: Green Power Amplification Systems for
3G+ Wireless Communication Infrastructure. In: Proc. IEEE Vehicular Technology
Conference, pp. 1–5 (September 2010)
[28] Schreier, R., Temes, G.C.: Understanding Delta-Sigma Data Converters. Wiley-IEEE
Press (2004)
[29] Helaoui, M., Hatami, S., Negra, R., Ghannouchi, F.M.: A Novel Architecture of Del-
ta-Sigma Modulator Enabling All-Digital Multiband Multistandard RF Transmitters
Design. IEEE Transactions on Circuit and Systems II: Express Briefs 55(11),
1129–1133 (2008)
[30] Ghannouchi, F.M., Hatami, S., Aflaki, P., Helaoui, M., Negra, R.: Accurate Power
Efficiency Estimation of GHz Wireless Delta-Sigma Transmitters for Different
Classes of Switching Mode Power Amplifiers. IEEE Transactions on Microwave
Theory and Techniques 58(11), 2812–2819 (2010)
Chapter 6
Transmitter Design for MIMO Wireless
Communications

The high demand for broadband multimedia Internet access and wireless connec-
tions has increased the need for more advanced and sophisticated wireless com-
munication systems. However, wireless channels usually provide limited band-
width and lower quality links.
The next generation of wireless technologies is targeting two essential goals in
their design and development. One is the provision of high-speed data rates up to
100 megabits per second (Mb/s) for mobile users and 1 gigabit per second (Gb/s)
for stationary users. Achieving the goal of improving the data rate and increasing
the system capacity is feasible through the use of more advanced signal processing
and coding techniques, such as spectral efficient 64-QAM (quadrature amplitude
modulation), orthogonal frequency-division multiplexing (OFDM), and multiple
input multiple output (MIMO) topology.

6.1 Complexity and Cost in MIMO Systems

MIMO wireless communication systems are designed to deliver either maximal
diversity to enhance transmission reliability or to increase maximal multiplexing
gain to support high data rate [1]. The channel capacity and spectral efficiency of
a MIMO system is usually much higher than that of a single-input single-output
(SISO) system. This can be achieved by employing multiple antennas at the
transmitter and receiver.
This performance of a MIMO system can be quantified by the spatial multip-
lexing gain. Therefore, it is possible to send parallel independent data streams to
achieve overall system capacities scaled with min( N t , N r ) , where N t and N r
are the number of receiving and transmitting antennas [2]. On the other hand, if
the signal copies are transmitted from or received at multiple antennas, this mul-
tiantenna system can provide a gain that enhances the reliability of a wireless link.
This gain is known as diversity gain, which is achieved by space-time coding.
Both diversity and multiplexing gain can be supported by MIMO systems at the
same time, but there is a fundamental trade-off between them [1].
In multi-branch transceivers, multiple radio frequency (RF) front-ends are inte-
grated on the same platform. Multiple modulated signals are simultaneously

A. Mohammadi et al.: RF Transceiver Design for MIMO Wireless Communications, LNEE 145, pp. 129–149.
springerlink.com © Springer-Verlag Berlin Heidelberg 2012
130 Chapter 6 Transmitter Design for MIMO Wireless Communications

transmitted and received using these RF front-ends. The multi-branch MIMO tran-
sceiver is a simple case of MIMO transceiver architecture, which can improve the
power and diversity gains of the system. Transmitter and receiver nonlinearity,
modulator imbalance and other impairments should be considered in the design of
each branch of a multi-branch transceiver. In addition, when multiple transmis-
sion/reception paths are realized on the same chipset, new issues are generated,
such as RF crosstalk between the multiple paths, which are due to the proximity of
the different circuits [3].
A multi-branch transceiver with N antennas at the transmitter and N anten-
nas at the receiver is demonstrated in Figure 6.1. This system uses multiple paral-
lel RF front-ends, and the number of RF front-ends is equal to the number of an-
tennas. At the receiver, the baseband processing unit decodes N received
baseband paths to recover the signal and also to obtain the diversity gain.

Fig. 6.1 A conventional multiantenna receiver

The multiple antenna RF front-end architecture design usually leads to higher

complexity and hardware costs in the RF section. Furthermore, increasing the
number of antennas causes growth in RF circuit mismatches and coupling. In fact,
these issues limit the application of a high number of antennas at the transceiver.
One option to overcome such a problem is the adoption of a single RF front-end in
a MIMO system, where a single RF path is implemented instead of multiple paral-
lel RF paths [2], [4]. This reduces the complexity and cost of an RF section and al-
so enables a compact design with lower power consumption.
An orthogonal transmission of multiple RF streams over a single front-end
must be identified for the realization of a single RF front-end path. Common de-
signs that may be used for a single receiver front-end realization are the antenna
selection technique [5], frequency-division multiplexing, time-division multiplex-
ing [1], and code-division multiplexing [6].

6.2 Transmitters Architectures

Several architectures have been proposed for transmitters in wireless communica-
tions, which are dependent on the performance requirements.
6.2 Transmitters Architectures 131

6.2.1 Direct Conversion Transmitter

The direct conversion transmitter architecture has received considerable attention
for multi-standard, multi-band applications, due to its smaller number of compo-
nents and usability. This architecture consists of a digital baseband processing
unit, digital-to-analog converters (DACs), a direct up-converter, a PA, and an an-
tenna. The digital baseband processing unit has the responsibilities of digital mod-
ulation, coding and filtering, in order to develop and synthesize an appropriate
transmission signal from the digital input data.
The DACs convert the digital in-phase (I) and quadrature (Q) baseband signals
to analog signals, which feed the quadrature modulator for direct up-conversion.
Analog signals are directly up-converted to the RF with I and Q carriers. The
block diagram of a direct conversion transmitter is shown in Figure 6.2. The
bandpass filter after the signal summation is used to suppress the out-of-band sig-
nals produced by the harmonic distortion of the carrier. The RF signal power is in-
creased by the PA for transmission through the antenna.

Fig. 6.2 Block diagram of a direct conversion transmitter

The direct conversion transmitter is theoretically simple, and there are no in-
termediate frequency (IF) components. The drawbacks of this architecture are lo-
cal oscillator (LO) leakage at RF frequencies, voltage controlled oscillator (VCO)
pulling, and the requirement of an I/Q mixer at RF frequencies [7]. The direct
conversion transmitter suffers from unequal complex gains of the I and Q paths.
This depends on the frequency and operating temperature. The quality of the out-
put RF signal is strongly affected by the I/Q imbalance caused by the modulators.
In addition, the PA stage has nonlinear behavior, which deteriorates the output
signal quality and produces out-of-band power emission. Digital predistortion
(DPD) techniques are suitable candidates to enhance the signal quality and com-
pensate for the out-of-band power. However, the I/Q imbalance at the transmitter
can considerably reduce the linearization capacity of the digital predistorter [8].
132 Chapter 6 Transmitter Design for MIMO Wireless Communications

6.2.2 Superhetrodyne Transmitter

The most common double conversion approach is the superheterodyne architec-
ture. A schematic of a superheterodyne transmitter is shown in Figure 6.3. In the
first stage, the baseband input is up-converted to IF by an I/Q mixer. The first
bandpass filter, BP1, shown in this figure reduces any spurious out-of-band power
prior to the second up-conversion. In the second stage, the IF signal goes through
a second up-conversion to achieve the desired transmit frequency. A second band-
pass filter, BP2, can be utilized to filter additional spurious out-of-band power
caused by the second mixing operation [9]. The IF frequency is determined by the
first LO frequency, f1 . Two frequencies are generated by the second mixer stage,
f 2 + f1 and f 2 − f1 . One of the frequencies is selected by the second bandpass
filter and the unwanted frequency (image frequency) is rejected.
The superheterodyne architecture was developed to overcome the inherent
problems in the direct conversion or homodyne architecture. In this architecture,
some of the unwanted frequency components originating from nonlinearities (e.g.,
in the mixers) can be removed by filters. Figure 6.4 shows the frequency plan of
the superheterodyne transmitter with one IF frequency.

Fig. 6.3 Block diagram of a superheterodyne transmitter

Fig. 6.4 Frequency plan of the superheterodyne transmitter

6.3 Brief Overview of MIMO Transmission Schemes 133

6.3 Brief Overview of MIMO Transmission Schemes

In telecommunications, multiplexing is a technique by which multiple analog sig-
nals or digital data streams are combined into a single signal to be transmitted over
a shared medium. The basis of the multiplexing technique is the division of the en-
tire signaling dimensions into parts or channels and the allocation of these parts or
channels to different users. The most common techniques of dividing signal space
are along the frequency, time and code axes, which are called frequency-division
multiplexing (FDM), time-division multiplexing (TDM), and code-division mul-
tiplexing (CDM).

6.3.1 FDM Technique

FDM is a technique that combines several signals into one medium by sending sig-
nals in several distinct frequency ranges over that medium, as shown in Figure 6.5
(a). This is a familiar technique in many industries, such as telephone and commer-
cial radio and television broadcasting. Although the FDM technique can be applied
to both analog and digital systems, it has been widely used in the analog communi-
cation systems.
Guard bands are used in the FDM technique to avoid interference between
channels. The FDM produces a spectral efficiency, which is due to the transmis-
sion rate that is quite close to the maximum rate needed by the user. FDM needs
frequencies that can provide different carriers with different channels, because the
transmission in communication systems is continuous.

6.3.2 TDM Technique

TDM is a multiplexing technique that allows more than one user to access RF chan-
nels without any interference between them. Figure 6.5 (b) shows how the frequency
channels are shared at different times. It involves the sequencing of groups of sig-
nals from each individual input, so that they can be associated with the appropriate
receiver. TDM has been used in digital communication technologies, including sig-
nificant application in cellular phone technologies. The guard times are used in the
TDM technique to reduce the transmission impairments, such as delay propagation,
the transient response of the pulse signal, and other impairments.

6.3.3 CDM Technique

CDM is a multiplexing technique where several channels share the same frequen-
cy spectrum at the same time, as shown in Figure 6.5 (c). CDM employs spread-
spectrum technology and a special coding method in which each transmitter is
134 Chapter 6 Transmitter Design for MIMO Wireless Communications

assigned a code, allowing multiple users to be multiplexed over a shared physical

medium. Direct sequence spread spectrum (DSSS) and frequency hopping are two
forms of the CDM technique. In CDM, the modulated coded signal has a much
higher bandwidth than the data being communicated. One of the early applications
for CDM is in Global Positioning Systems (GPSs). CDM techniques are also used
in wireless networks due to the advantage of reducing the interference among dif-
ferent users.

Fig. 6.5 Time and/or frequency sharing in (a) TDM, (b) FDM and (c) CDM techniques

6.4 MIMO Transceiver Architectures

MIMO refers to a system where multiple inputs have interaction with multiple
outputs. In the context of wireless communications, MIMO refers to topologies in
which multiple modulated signals, separated in the space, time or frequency do-
main are simultaneously transmitted through an RF front-end. A MIMO system
with modulated signals separated in the space domain is are already known as a
MIMO system in wireless communication theory. This topology has multiple
branches of RF front-ends that are involved in simultaneous transmission of sig-
nals. A MIMO system with modulated signals separated in the frequency domain
is a system in which multiple signals modulated in different carrier frequencies are
transmitted all together through a single-branch RF front-end.

6.4.1 Antenna Selection Architecture

Antenna selection techniques in spatial multiplexing at the transmitter and/or re-
ceiver can lead to simpler RF front-end MIMO systems. Based on antenna selec-
tion techniques, some of the available antennas are chosen; therefore, fewer RF
chains than the number of transmitter and/or receivers antennas are used. Conse-
quently, this reduces cost and complexity; and, at the same time, the system per-
formance is maintained. The antenna selection technique, both in the transmitter
6.4 MIMO Transceiver Architectures 135

and receiver of a MIMO system, is demonstrated in Figure 6.6. Although the im-
plementation of the antenna selection technique in the receiver is straightforward,
the implementation in the transmitter needs a feedback path from the transmitter to
the receiver [2].

Fig. 6.6 A MIMO system with antenna selection technique

The performance of the antenna selection algorithm is largely dependent on the

objective function or the selection criterion utilized to construct the subset of an-
tennas. Several criteria have been introduced. A number of common selection cri-
teria are briefly presented in the following subsections.

6.4.1.1 Maximum Capacity Criterion

Channel capacity maximization is a typical criterion for antenna selection. An ana-

lytical bound for the channel capacity of MIMO systems with antenna selection
[1], [10] can be given as:

Kr
⎛ ρ ⎞
C≤ ∑ log
i =1
⎜
2 ⎜1 +
⎝ Nt
γ i ⎟⎟
⎠
(6.1)

where ρ is the mean signal-to-noise ration (SNR), K r is the number of selected

antennas in the receiver, and γ i represents the squared norm of the i th row of the
MIMO channel matrix, H , which is ordered from the smallest to the largest.
Based on the capacity formula in (6.1), the capacities for all possible subsets of a
receiver’s antennas, p ∈ P , are calculated; and, the subset with the largest ca-
pacity C p is selected.
136 Chapter 6 Transmitter Design for MIMO Wireless Communications

6.4.1.2 Maximum Minimum Singular Value Criterion

For each subset of receiver antennas, p ∈ P , the minimum singular value ( λ min )
corresponding to different H p is calculated. Then, the subset with the largest
λ min is selected.

6.4.1.3 Norm-Based Selection (NBS) Criterion

In the norm-based selection (NBS) criterion, the subset of receiver antennas re-
lated to the rows of H with the largest Euclidean norm is chosen. Although this
technique is not optimal when the number of RF front-ends is greater than one
[11], it remains a popular criterion because of its simplicity.

6.4.2 Frequency-Division Multiplexing (FDM) Architecture

In FDM architecture, the signals of different antennas are shifted in frequency
with mixing by different LOs, added together, and transmitted through a single-
branch RF frond-end. Figure 6.7 shows the general block diagram of FDM
architecture. The multiple signals from different antennas are separated in the fre-
quency domain. A single RF front-end is used to down-convert the signals, and
the frequency shifts of the multiple streams are removed in the baseband block.
The diversity gain is extracted by further processing [4]. Examples of this type of
system are concurrent dual-band transmitters and multicarrier transmitters.
A system-level performance of a scenario with two antennas at the receiving
side was studied in [4]. Considering s(t ) as the passband signal with a center fre-
quency, f c , and a bandwidth, ω , the received signal of the first antenna is
h1 ⋅ s (t ) + n1 (t ) ; and, the received signal from the second antenna is
h2 ⋅ s (t ) + n2 (t ) , where n1 (t ) and n2 (t ) are the passband noise, and h1 and h2
are the Rayleigh flat fading channel coefficients. The signal of second antenna is
combined with a low-frequency oscillator at the same frequency ( ω ), then filtered
to remove the IF term.
The deficiency of this technique is the requirement of a narrowband filter in the
RF frequency. The simulation results for binary phase shift keying (BPSK) mod-
ulation using two antennas are demonstrated in Figure 6.8. The variation between
the ideal and simulated curves for diversities 1 and 2 is related to the filter design
and its quality factor.

6.4.3 Time-Division Multiplexing (TDM) Architecture

The general block diagram of TDM architecture is shown in Figure 6.9. In this ar-
chitecture, the same number of antennas as in the conventional topology is used,
6.4 MIMO Transceiver Architectures 137

Fig. 6.7 Realization of single-branch multiantenna receiver based on FDM

0
10
Diversity = 1(Ideal)
Diversity = 1(FDM)
Diversity = 2(Ideal)
Diversity = 2(FDM)

-1
10
BER

-2
10

-3
10
0 1 2 3 4 5 6 7 8 9 10
Eb/N0(dB)

Fig. 6.8 Bit error rate (BER) versus Eb / N0 for a single frond-end receiver based on FDM
using BPSK modulation [4]

but instead of having multiple RF front-ends, a single pole, multiple-throw RF

switch along with a single RF front-end are used to down-convert the RF signals
to baseband. Furthermore, the signals are carried to the baseband processing unit
using a de-multiplexer. The switch is used to capture the signals of all the anten-
nas for every symbol time interval of the modulated signal. This offers some con-
straints on the switching speed. In addition, careful alignment of the data before
138 Chap
pter 6 Transmitter Design for MIMO Wireless Communicationns

and after the multiplexin ng and de-multiplexing of the multiple signals in thhe
MIMO receivers is desired.
A single front-end MIMMO receiver using the TDM technique can be realized aas
a single RF architecture by
b time-multiplexing the antennas’ signals using a singlee-
pole multiple-throw RF switch.
s The received signals are de-multiplexed after R RF
processing according to a diversity algorithm. Although this architecture implee-
ments the same number of o antennas as in the conventional architecture, it reducees
the number of RF front-eends from N to 1 . Consequently, the overall cost annd
size of the multiantenna receiver and also RF design mismatch are significantlly
reduced [2].

Fig. 6.9 Realization of singlle-branch multiantenna receiver based on TDM

0
10
diversity = 1(Ideal)
diversity = 1(TDM)
diversity = 2(Ideal)
diversity = 2(TDM)

-1
10
BER

-2
10

-3
10
0 1 2 3 4 5 6 7 8 9 10
Eb/N0(dB)

Fig. 6.10 BER versus Eb / N0 for a single frond-end receiver based on TDM using BPS
SK
modulation [4]
6.4 MIMO Transceiver Architectures 139

Figure 6.10 shows the results of a time-multiplexed receiver using a raised co-
sine pulse shape filter with a roll-off factor of 0.5 for single antenna and two-
antenna cases using BPSK modulation. Zero forcing (ZF) receivers are imple-
mented. To compare the TDM architecture with a conventional architecture, the
bit error rate (BER) of the multiple frond-end systems is demonstrated as well.
The diversity gain using a single time-multiplexed RF front-end is equal to the di-
versity gain of the multiple RF front-end, as shown in Figure 6.10. This architec-
ture demands a wider bandwidth than the one used in the conventional MIMO re-
ceiving front-end.

6.4.4 Code-Division Multiplexing (CDM) Architecture

In code-division multiplexing (CDM), the signals from different antennas are mul-
tiplied by orthogonal codes and added together. At the receiver, a single RF front-
end is used to down-convert the resultant RF signals to baseband, where the sig-
nals are again multiplied by the orthogonal codes, integrated, and de-multiplexed.
This technique is demonstrated in Figure 6.11. In the case of two receiver anten-
nas, the first and second signals are multiplied by C1 (t ) and C2 (t ) codes, respec-
tively, in the symbol duration, and then summed together. These codes should be
orthogonal, i.e., C1 (t ) is equal to 1 between 0 and Ts , and C2 (t ) is equal to 1 be-
tween 0 and Ts / 2 and equal to -1 between Ts / 2 and Ts . Ts is the symbol dura-
tion. After summing the signals, the total signal,
(h1 ⋅ s (t ) + n1 (t )) ⋅ c1 (t ) + (h2 ⋅ s (t ) + n2 (t )) ⋅ c2 (t ) , is down-converted using a single
RF front-end. The baseband signal is expressed as
(h1 ⋅ s (t ) + n1 (t )) ⋅ c1 (t ) + (h2 ⋅ s (t ) + n2 (t )) ⋅ c2 (t ) .
Followed by down-conversion in the baseband processing unit block, the signal
is multiplied by C1 (t ) and C2 (t ) . This separates the signal into different paths.
An integrator removes the effect of the other signals in each path. However, the
effect of the noise of both antennas is preserved in each path. The output of the in-
tegrator in the first path can be obtained as:
h1 ⋅ s + n =
Ts
1
Ts ∫ [(h s(t ) + n (t )) ⋅ c (t ) ⋅ c (t ) + (h
0
1 1 1 1 2 ⋅ s (t ) + n2 (t )) ⋅ c2 (t ) ⋅ c1 (t )] dt (6.2)

Ts Ts Ts
1 1 1
=
Ts ∫ h1 s (t )dt +
0 Ts ∫ n1 (t )dt + 0 +
0 Ts ∫ n (t ) ⋅ c (t ) ⋅ c (t )dt
0
2 2 1

where ~s (t ) is invariable in the symbol duration, Ts , and the codes have unit
energy. The third term is zero, which is due to the orthogonality of C1 (t ) and
C2 (t ) ; and, ~
s (t ) is constant in the symbol duration, Ts . The fourth term is non-
zero, because n2 (t ) is stochastic and variable in the symbol duration.
140 Chapter 6 Transmitter Design for MIMO Wireless Communications

The second and fourth terms of the integral have similar power; therefore, the
output noise power using CDM is twice that of conventional design. The noise
level is increased by 10 log( N ) dB in each branch, when N antennas are uti-
lized and the signals are down-converted with the CDM technique [12]. Figure
6.12 demonstrates the simulation results for BPSK modulation using two receiv-
ing antennas. A separation of 3 dB can be noticed between the CDM technique
and the conventional diversity system [4].

Fig. 6.11 Realization of single-branch multiantenna receiver based on CDM

0
10
Diversity = 1(Ideal)
Diversity = 2(Ideal)
Diversity = 2(CDM)

-1
10
BER

-2
10

-3
10
0 1 2 3 4 5 6 7 8 9 10
Eb/N0(dB)

Fig. 6.12 BER versus Eb / N0 for a single frond-end receiver based on CDM using BPSK
modulation [4]
6.5 Distortion and Impairment Compensation in MIMO Transmitters 141

6.5 Distortion and Impairment Compensation in MIMO

Transmitters
MIMO topology is a highly efficient solution for improving the spectral efficiency
of wireless systems. Indeed, moving from SISO transceivers to MIMO transceiv-
ers could theoretically multiply the system capacity or data rate by the number of
outputs integrated in the MIMO transceiver. However, MIMO topology faces var-
ious implementations issues that can be classified into two major categories. The
first category consists of issues related to the general transceiver design, such as
transmitter linearity, receiver dynamic range, and imbalance and leakage in mixers
[3]. This group is not unique to MIMO systems.
The second category is specific to MIMO transceivers. In multi-branch MIMO
transceivers, crosstalk results from the coupling and interference between the sig-
nals of different branches. Crosstalk is more likely to take place between the
branches, because the signals in different branches use the same operating fre-
quency and have equal transmission power [13]. This is more significant in an in-
tegrated circuit (IC) design, especially when the actual footprint of the circuit is
small [14].
RF crosstalk can be categorized as linear and nonlinear. Linear crosstalk occurs
beyond the output of the transmitter, i.e., at the antenna, and can be modeled as a
linear function of the interference and desired signals. This means the signal af-
fected by linear crosstalk does not pass through nonlinear components. Converse-
ly, nonlinear crosstalk affects the signal before it passes through nonlinear compo-
nents. This crosstalk that takes place in the transmitter circuit prior the PA is the
main source of nonlinear crosstalk, since the PA is the main source of nonlinearity
[3]. The origin and nature of both types of crosstalk are illustrated in Figure 6.13.

Fig. 6.13 Linear and nonlinear crosstalk in dual-branch MIMO transmitters

142 Chapter 6 Transmitter Design for MIMO Wireless Communications

6.5.1 Antenna Crosstalk

Antenna crosstalk is the effect of the transmitted signal from one antenna element
to another antenna element, which can be modeled as:
G G
Y = AX (6.3)
G G
where X = [ x1 x2 ]T and Y = [ y1 y2 ]T are the signals at the input and output of
⎡1 α⎤
the antennas, respectively; and, A = ⎢ ⎥ is the antenna crosstalk matrix.
⎣β 1 ⎦
For symmetric antenna crosstalk, α and β are equal [3].
With the expansion of (6.3), the signals at the output of the antennas after
crosstalk can be shown as:
⎧ y1 = x1 + αx 2
⎨ (6.4)
⎩ y 2 = β x1 + x 2
From this equation, it can be observed that y1 and y 2 are linear functions of the
signals at the antennas’ input.
Linear crosstalk has been addressed in several research studies [15], [16]; and,
several techniques have been proposed to compensate for it. The compensation is
simultaneously performed mostly at the receiver side for the composite linear cross-
talk generated at the transmitter and receiver antennas and also by the channel.
A 2 × 2 MIMO system with three types of linear crosstalk is shown in Figure
6.14. The total linear crosstalk can be modeled as:
G G G
Y = (BHA) X + BN (6.5)

where A and B are crosstalk matrices for the transmitting and receiving anten-
⎡ h11 h12 ⎤
nas, respectively; and, H = ⎢ ⎥ is a crosstalk matrix for the channel; and,
⎣ h21 h22 ⎦
G
N is an additive white Gaussian channel noise (AWGN) vector.
The uncorrelated received signal using the matrix inversion algorithm is given by:
G G
Yuncorr . = (BHA) −1 Y (6.6)

Substituting (6.5) into (6.6):

G G G
Yuncorr . = X + (HA) −1 N (6.7)
G
where Yuncorr. is the uncorrelated component of the received signal without the
noise component [3].
6.5 Distortion and Impairment Compensation in MIMO Transmitters 143

From (6.7), it can be observed that the matrix inversion method is practical if
the (BHA) matrix is invertible. Furthermore, the crosstalk in the transmitter an-
tenna and MIMO channel may deteriorate the performance of the MIMO system
due to the noise enhancement, as the last term in (6.7). The performance of the
MIMO system is not sensitive to the receiver antenna crosstalk as long as matrix
B is invertible.

Fig. 6.14 A 2× 2 MIMO channel and antenna crosstalk

6.5.2 Nonlinear RF Crosstalk

MIMO transmitters need to maintain certain levels of efficiency and linearity in
the system, similar to those of single-branch transmitters. Actually, the power effi-
ciency and linearity requirements of MIMO transmitters are more restrained than
those of single-branch transmitters. As demonstrated in Figure 6.13, crosstalk that
occurs before the PA is recognized as nonlinear. Some sources of this type of
crosstalk can be the leakage of the RF signal through the common LO [14] and in-
terference in the chipset.
Considering RF nonlinear crosstalk and transmitters’ nonlinearities, the trans-
mitters’ output in a dual-branch MIMO transmitter can be modeled as:

y1 = f1 ( x1 + αx2 ) (6.8)

y 2 = f 2 ( βx1 + x2 ) (6.9)

where f1 (⋅) and f 2 (⋅) are nonlinear functions that represent the transmitter res-
ponses of each branch; xi and yi are the input and output signals, respectively;
and, α and β are the RF nonlinear crosstalk. Since f1 (⋅) and f 2 (⋅) are nonli-
near, a simple matrix inversion is not sufficient to compensate for the effect of
nonlinear crosstalk.
144 Chapter 6 Transmitter Design for MIMO Wireless Communications

Due to unavoidable RF crosstalk between branches of a MIMO transmitter, in-

dependent modeling of each branch is not adequate to include the effects of RF
nonlinear crosstalk. Hence, the traditional DPD techniques developed for SISO
systems are not suitable for MIMO systems. In fact, a new DPD architecture with
the capability of having more than one input signal is required [3].

6.5.3 Effects of Nonlinear Crosstalk on DPD Extraction

In Figure 6.15 (a), a MIMO transmitter is shown as a black box with two inputs
and two outputs. One way to model the dual-branch MIMO transmitter is inde-
pendent modeling of each branch of the MIMO transmitter, which is illustrated in
Figure 6.15 (b). The independent modeling is reliable if there are no interactions
between the two branches. However, as discussed previously, nonlinear crosstalk
is an inevitable incident in dual-branch MIMO transmitters. As a matter of the
fact, the transmitters’ output with the effect of crosstalk can be modeled as in (6.8)
and (6.9). It can be observed from the expressions in (6.8) and (6.9) that the signal
at the output of each branch is a function of both input signals, in the presence of
RF nonlinear crosstalk. Hence, independent modeling of each branch, according to
the input and output signals of that branch, leads in an inefficient model for the
dual-branch transmitter. Similarly, the use of DPD linearization techniques leads
to a similar scenario when linearization blocks are realized separately for each
branch of the dual-branch transmitter.

Fig. 6.15 A MIMO transmitter as (a) a black box, (b) two independent parallel models

The cascade of DPD and the RF frond-end complex transfer function in an

ideal transmitter without nonlinear crosstalk results in a linear complex transfer
function. However, in MIMO transmitters, the quality of the DPD extraction de-
grades when nonlinear crosstalk modifies the complex envelope of the RF signal
at the input of the PA of a transmitter [3]. As shown in Figure 6.16, the DPD coef-
ficients of the upper branch transmitter are extracted using the digital baseband
signal to be transmitted ( z1 ) and the equivalent complex envelope of the PA RF
output ( y1 ). The coupled signal from the second path ( z 2 ) is also added to signal
6.5 Distortion and Impairment Compensation in MIMO Transmitters 145

z1 , and the combined signal ( z1 + αz 2 ) is amplified by the PA. Hence, the output
PA signal can be expressed as:

y1 = f1 ( z1 + αz 2 ) (6.10)

In the DPD extraction process, the DPD function, g1 (⋅) , is the inverse function of
f1 (⋅) , which depends on both input signals, x1 and x 2 , through z1 and z 2
terms. Considering that g1 (⋅) is only a function of the x1 input signal, the PA
output signal is:
y1 = f1 ( g1 ( x1 ) + αz 2 ) ≠ G0 ⋅ x1 (6.11)

where G0 is the linear or small-signal gain of the PA.

Hence, the nonlinear crosstalk influences the extraction and effectiveness of the
DPD function.

Fig. 6.16 DPD and PA with nonlinear crosstalk in a MIMO system

6.5.4 Impairment and Distortion Compensation

Most of the techniques proposed so far have addressed either nonlinear distortion,
the I/Q imbalance of the modulator, or the coupling effects in MIMO transmitters.
However, there are not many proposed solutions that jointly address all the issues
at the same time. In order to characterize the static and dynamic (memory effect)
nonlinear behavior of the transmitter, the multi-branch polynomial model can be
146 Chapter 6 Transmitter Design for MIMO Wireless Communications

used [17]. The output complex envelope signal at the output of the nonlinear
transmitter can be expressed as:
M N

∑∑ b
i −1
y ( n) = i, j ⋅ x( n − j ) ⋅ x( n − j ) (6.12)
j =0 i =1

where x(n) and y (n) are the input and output complex signal envelopes, respec-
tively; bi , j are the model coefficients of the j th filter tap; and, N and M are
the maximum polynomial order and memory depth, respectively.
In a MIMO transmitter, for the modeling of the coupling effects in addition to
the PA nonlinearity, each output of the transmitter should contain cross terms be-
tween the input signals [3]. The memory polynomial model can be extended for a
dual-input dual-output transmitter with crosstalk effects. The signal at each output
of the transmitter is given by:
M N
yk (n) = ∑∑ bk ,1,i , j ⋅ x1 (n − j ) ⋅ x1 (n − j )
i −1
+
j = 0 i =1
M N
(6.13)
∑∑ bk ,2,i, j ⋅ x2 (n − j ) ⋅ x2 (n − j )
i −1

j = 0 i =1

where x1 (n) and x2 (n) are the input complex signals envelopes, and yk (n) is
the output complex signal envelop at the k -branch of the MIMO transmitter. The
modeling of the nonlinear crosstalk (memoryless case) in a MIMO transmitter is
illustrated in Figure 6.17.

Fig. 6.17 Modeling of a 2× 2 MIMO transmitter with nonlinear crosstalk

6.5 Distortion and Impairment Compensation in MIMO Transmitters 147

In addition to crosstalk, the I/Q imbalance of the modulator affects the perfor-
mance of the MIMO transmitter. The modeling of the I/Q imbalance of the mod-
ulator is based on the modeling of the cross coupling channels between the I and
Q components of the modulator signal input. The cross coupling terms can be pre-
sented in the model by utilizing the conjugate of the input signal [18]. I/Q imper-
fections are generally gain and phase imbalance and DC offset. The signal at the
output of a SISO transmitter suffering from such impairments can be expressed as:
M N
y (n) = ∑∑ bi , j ⋅ x(n − j ) ⋅ x(n − j )
i −1

j = 0 i =1
M N
(6.14)
+ ∑∑ bi′, j ⋅ x (n − j ) ⋅ x(n − j )
i −1
*
+ bdc
j = 0 i =1

*
where x(n) , x (n) and y (n) are the input, input conjugate and output complex
signals envelopes, respectively; and, bdc is a DC term used to estimate the DC
offset of the modulator.
Figure 6.18 demonstrates the modeling of the modulator’s I/Q imbalance for a
SISO transmitter for the memoryless case.

Fig. 6.18 Modeling of modulator I/Q imbalance for a SISO transmitter

Based on the prior models, a dual-input dual-output MIMO transmitter that suf-
fers from PA nonlinearity, modulator I/Q imbalance and coupling effects can be
* *
counted as a nonlinear system with four inputs ( x1 , x 2 , x1 , x 2 ) and two outputs
( y1 , y 2 ). The outputs can be related to the inputs as:

[y1 y 2 ] = W x1
T
[ x2 x1* x 2* ]T
(6.15)

where W is a nonlinear matrix function that characterizes the behavior of the

transmitter. The signal at the output of the MIMO transmitter suffering from PAs
nonlinearity, nonlinear crosstalk, modulator I/Q imbalance and DC offset can be
expressed as:
148 Chapter 6 Transmitter Design for MIMO Wireless Communications

M N
yk (n) = ∑∑ bk ,1,i , j ⋅ x1 (n − j ) ⋅ x1 (n − j )
i −1
+
j = 0 i =1
M N

∑∑ b′
i −1
k ,1, i , j ⋅ x1* (n − j ) ⋅ x1 (n − j ) +
j = 0 i =1
M N

∑∑ b
i −1
k ,2, i , j ⋅ x2 (n − j ) ⋅ x2 (n − j ) +
j = 0 i =1
M N

∑∑ b′
i −1
k ,2, i , j ⋅ x2* (n − j ) ⋅ x2 (n − j ) + bdc
j = 0 i =1

Digital baseband preprocessing techniques can be used to compensate for all the
effects previously mentioned. Hence, identification needs to be performed to ex-
tract the coefficients of the inverse model. The inverse model is used to generate
the MIMO transmitter input when fed with the MIMO transmitter output.

References
[1] Duman, T.M., Ghrayeb, A.: Coding for MIMO Communication Systems. Wiley
(2008)
[2] Mohammadi, A., Ghannouchi, F.M.: Single RF Front-End MIMO Transceivers. IEEE
Communications Magazine (December 2011)
[3] Bassam, S.A., Helaoui, M., Ghannouchi, F.M.: Crossover Digital Predis-torter for the
Compensation of Crosstalk and Nonlinearity in MIMO Transmitters. IEEE Transac-
tions on Microwave Theory and Techniques 57(5), 1119–1128 (2009)
[4] Lari, M., Bassam, S.A., Mohammadi, A., Ghannouchi, F.M.: Time-Multiplexed Sin-
gle Front-End Multiple-Input Multiple-Output Receivers with Preserved Diversity
Gain. IET Communications 5(6), 789–796 (2011)
[5] Molisch, A., Win, M., Winters, J.: Reduced-Complexity Trans-mit/Receive-Diversity
Systems. IEEE Transactions on Signal Processing 51(11), 2729–2738 (2003)
[6] Tzeng, F., Jahanian, A., Pi, D., Heydari, P.: A CMOS code-modulated path-sharing
multi-antenna receiver front-end. IEEE Journal of Solid-State Circuits 44(5),
1321–1335 (2009)
[7] Vankka, J.: Digital Synthesizers and Transmitters for Software Radio. Springer,
Heidelberg (2005)
[8] Bassam, S.A., Boumaiza, S., Ghannouchi, F.M.: Block-Wise Estimation of and Com-
pensation for I/Q Imbalance in Direct-Conversion Transmitters. IEEE Transactions
on Signal Processing 57(12), 4970–4973 (2009)
[9] Wiser, R.: Tunable Bandpass RF Filters for CMOS Wireless Transmitters. ProQuest
(2008)
[10] Gesbert, D., Shafi, M., Shiu, D.-S., Smith, P.J., Naguib, A.: From Theory to Practice:
An Overview of MIMO Space-Time Coded Wireless Systems. IEEE Journal on Se-
lected Areas in Communications 21(3) (April 2003)
[11] Molisch, A.F., Win, M.Z., Winters, J.H.: Reduced-Complexity Transmit/Receive Di-
versity Systems. In: Proc. IEEE Vehicular Technology Conference, pp. 1996–2000
(May 2001)
References 149

[12] Jahanian, A., Tzeng, F., Heydari, P.: Code-Modulated Path-Sharing Multi-Antenna
Receivers: Theory and Analysis. IEEE Transactions on Wireless Communica-
tions 8(5), 2193–2201 (2009)
[13] Bassam, S.A., Helaoui, M., Boumaiza, S., Ghannouchi, F.M.: Experimental Study of
the Effects of RF Front-End Imperfection on the MIMO Transmitter Performance. In:
Proc. IEEE MTT-S International Symposium Digest, pp. 1187–1190 (June 2008)
[14] Palaskas, Y., Ravi, A., Pellerano, S., Carlton, B.R., Elmala, M.A., Bishop, R., Baner-
jee, G., Nicholls, R.B., Ling, S.K., Dinur, N., Taylor, S.S., Soumyanath, K.: A 5-GHz
108-Mb/s 2 ×2 MIMO Transceiver RFIC with Fully In-tegrated 20.5-dBm P1dB Power
Amplifiers in 90-nm CMOS. IEEE Journal of Solid-State Circuits 41(12), 2746–2756
(2006)
[15] Tse, D., Viswanath, P.: Fundamentals of Wireless Communication. Cambridge Uni-
versity Press, Cambridge (2005)
[16] Medvedev, I., Bjerke, B.A., Walton, R., Ketchum, J., Wallace, M., Howard, S.: A
Comparison of MIMO Receiver Structures for 802.11n WLAN—Performance and
Complexity. In: The 17th Annual IEEE International Symposium on Personal, In-
door, and Mobile Radio Communications, Helsinki, Finland (September 2006)
[17] Ghannouchi, F.M., Hammi, O.: Behavioural Modeling and Predistortion. IEEE Mi-
crowave Magazine 10(7), 52–64 (2009)
[18] Anttila, L., Handel, P., Valkama, M.: Joint Mitigation of Power Amplifier and I/Q
Modulator Impairments in Broadband Direct-Conversion Transmitters. IEEE Trans-
actions on Microwave Theory and Techniques 58(4) (April 2010)
Chapter 7
Receiver Design for MIMO Wireless
Communications

In the previous chapters, it has been shown that multiple input multiple out (MIMO)
systems can provide either diversity order improvement or spatial multiplexing ad-
vantages. In this chapter, the receiver front-end architectures in MIMO systems are
discussed. Starting with the traditional single input single output (SISO) receiver
front-ends, their extensions are presented for MIMO applications. Moreover, the ca-
pacity reduction due to a noisy front-end and radio frequency interference are ex-
amined. In addition, a MIMO testbed realization technique is presented using a
MIMO testbed; and, the more advanced MIMO transceiver implementation tech-
niques are described under the commercial MIMO transceivers section.

7.1 Receiver Architectures

7.1.1 Superheterodyne Receiver

The superheterodyne receiver architecture, which is widely used in wireless com-
munications, includes two mixers. The receiver is shown in Figure 7.1. The first
mixer down-converts a radio frequency (RF) modulated signal to an intermediate
frequency (IF) using a local oscillator (LO). The IF signal is generated either
above or below the LO frequency. The undesired signal is called an image signal.
The frequency interval between the desired and image signals is 2IF.

90o

Fig. 7.1 The architecture of a superheterodyne receiver

A. Mohammadi et al.: RF Transceiver Design for MIMO Wireless Communications, LNEE 145, pp. 151–180.
springerlink.com © Springer-Verlag Berlin Heidelberg 2012
152 Chapter 7 Receiver Design for MIMO Wireless Communications

It is likely that an unwanted signal is allocated in the 2IF distance of an RF sig-

nal. In this case, the unwanted signal interferes with the desired IF signal. An image
rejection filter before the down-converter is used to prevent from this problem.
Moreover, to remove the unwanted components, a highly selective bandpass IF fil-
ter is used at receiver. The next stage is the automatic gain control (AGC) section,
which should keep the input level of the analog-to-digital converter (ADC) constant
by adjusting the gain. The IF signal is then down-converted to the baseband signal
using an IF oscillator. Finally, the baseband signal is processed in the digital signal
processor (DSP) after analog-to-digital conversion. The DSP can demodulate any
modulation format within its processing and data conversion bandwidth [1], [2].

7.1.2 Direct Conversion Receiver

The direct conversion, or zero IF, receiver is proposed to overcome the need to use
of an IF section at the receiver. This architecture down-converts the RF modulated
digital signal directly to a baseband signal. A typical architecture of this receiver is
shown in Figure 7.2. The oscillator frequency in this receiver is the same as the
carrier frequency of input signal. The channels are selected by tuning the LO and
removing unwanted signals using a low-pass filter [2].
The architecture of a direct conversion receiver is simpler than that of a super-
heterodyne receiver. This results in a smaller sized and lower cost receiver. How-
ever, there are more design challenges in these receivers. The main design issues
are due to the RF impairments, such as DC offset, I/Q imbalance, and second-
order distortion. On the other hand, this architecture provides less spurious power,
making it very attractive for the receiver applications.

90o
LO (@ RF)

Fig. 7.2 The architecture of a direct conversion receiver

7.1.3 Low IF Receiver

The architecture of a low IF receiver is shown in Figure 7.3. This architecture has
been proposed in order to take advantage of both direct conversion and superhete-
rodyne receivers [3]. The lack of DC offset is the main advantage of a low IF re-
ceiver over the direct conversion structure. However, the image rejection poses
greater difficulty compared to the superheterodyne receiver, because the image
frequency is closer to the RF frequency.
7.1 Receiver Architectures 153

The IF frequency in this architecture is usually selected from half the channel
bandwidth to several times the channel bandwidth. This creates more challenges in
removing the image signal. If the low IF frequency is selected to be equal to at
least half the channel bandwidth, the adjacent channel interference and image in-
terference can be filtered out by a bandpass filter known as a polyphase filter [3].

7.1.4 Image Rejection Receiver

The image reject receiver was developed to overcome the image problem in superhete-
rodyne receivers. The most common image reject receiver architectures are Hartley
and Weaver. The architecture of image reject receivers is shown in Figure 7.4.

Cos (ωo − ωi )t Sinωit

Sin(ωo − ωi )t

Fig. 7.3 The architecture of a low IF receiver

Cosωot

Sinωot

(ω RF − ωo )
ωRF Cosωot

(ω RF − ωo )
Sinωot

Fig. 7.4 The architecture of image rejection receivers: (top) Hartley architecture and (bot-
tom) Weaver architecture
154 Chapter 7 Receiver Design for MIMO Wireless Communications

7.1.5 Six-Port Receiver

A six-port receiver consists of two inputs and four outputs. In direct conversion
receiver applications, the RF and LO signals are applied to the two input ports, as
shown in Figure 7.5. The six-port structure produces four various combinations of
RF and LO signals and injects them to four power detectors, so that the outputs
can be obtained [4]. If the six-port junction is a linear network, the four outputs of
the multi-port structure are as follows:

2
Pi = Si1aLO + Si 2 aRF i = 1, 2,..., 4 (7.3a)

Si1 = Si1 e jφi1 (7.3b)

Si 2 = Si 2 e jφi 2 (7.3c)

Fig. 7.5 Block diagram of a direct conversion based six-port receiver

where Sij is the scattering parameter of the multi-port structure.

Assuming the RF signal with digital modulation has the same frequency as the LO
signal, the RF and LO signals can be represented by:

1
aLO = aLO e jϕLO (7.4)
2

1
aRF = aRF I 2 + Q 2 e jϕRF (7.5)
2

where I and Q are in-phase and quadrature-phase modulating signals.

7.1 Receiver Architectures 155

By substituting (7.4) and (7.5) into (7.3) and using the relations of:

I = I 2 + Q2 Cos (ϕ RF ) (7.6a)

Q = − I 2 + Q2 Sin (ϕRF ) (7.6b)

the expressions of the output powers can be derived as:

Pi =
1
2
2 2 1 2
Si1 aLO + Si 2 a RF I 2 + Q 2
2
2
( ) (7.7)
+ Si1 Si 2 a RF a LO ( ICos (ϕi ) + QSin (ϕi ) )

where ϕ = ϕ RF − ϕ LO and ϕi = ϕi1 − ϕi 2 .

Assuming that the LO power is known, the unknown parameters are (I2+Q2), I,
and Q. Thus, (7.7) can be written as:

( )
Pi = M i + Li I 2 + Q 2 + N i Cos ( ϕi ) I + N i Sin ( ϕi ) Q
1 2 2 1 2 2 (7.8)
Mi = Si1 aLO , Li = Si 2 aRF , Ni = Si1 Si 2 aRF aLO
2 2
ϕi = ϕi1 − ϕi 2 , for i = 1, 2,..., N

where Mi, Li and Ni are constant governing parameters for given operating condi-
tions, which are functions of scattering parameters and the power of LO and re-
ceived RF signals.
Equation (7.8) presents N equations with three unknown parameters. From
(7.8), the six-port equations can be written in matrix form as:

⎡ 1 ⎤ ⎡ P1 ⎤
⎢ 2 ⎥ ⎢ ⎥
(
⎢ I +Q
2
)
⎥ = D −1 ⋅ ⎢ P2 ⎥ (7.9)
⎢ I ⎥ T
⎢ P3 ⎥
⎢ ⎥ ⎢ ⎥
⎢⎣ Q ⎥⎦ ⎣ P4 ⎦

where DT is a matrix defined as:

⎡ M1 L1 N1Cos(φ1 ) N1Sin(φ1 ) ⎤
⎢M L2 N 2 Cos(φ2 ) N 2 Sin(φ2 ) ⎥
DT = ⎢ ⎥
2
(7.10)
⎢M3 L3 N 3Cos(φ3 ) N 3Sin(φ 3 ) ⎥
⎢ ⎥
⎣M 4 L4 N 4 Cos(φ4 ) N 4 Sin(φ 4 ) ⎦
156 Chapter 7 Receiver Design for MIMO Wireless Communications

If the determinant of the DT matrix is non-zero, (7.9) can be solved with respect to
unknown parameters:

I = α1′P1 + α 2′ P2 + α 3′ P3 + α 4′ P4 (7.11.1)

Q = β1′P1 + β 2′ P2 + β3′P3 + β 4′ P4 (7.11.2)

Equation (7.9) shows the general demodulation equations in six-port structures.

7.1.6 Five-Port Receiver

As seen in (7.9), there is a dependent equation that leads to a fixed value in the
first element of the calculated matrix [4], [5]. This dependent equation is the result
of the known LO power. This assumption is reasonable, because the LO power is
known to the receiver designer and can be removed from the unknown parameters.

Fig. 7.6 A five-port receiver

Nevertheless, if the LO power is so unstable that it cannot be treated as a con-

stant, the DC terms of each output can be easily removed by DC cancellation
methods (in a simple case, by a capacitor), which are always used in direct con-
version receivers for the DC offset cancellation. Therefore, the slow variation of
the LO signal with the other DC terms is removed, and it is not necessary to know
7.2 Smart Antenna Receiver Architectures 157

or calibrate the LO power exactly. Note that blindly removing LO power cannot
cause any problems, since the information needed for I/Q detection exists in terms
of the alternating current (AC) outputs. Note that the (I2+Q2) parameter is variable
and cannot be removed from demodulation calculations.
This is the reason for reducing the number of outputs from four to three; there-
fore, (7.9) may be changed to:

(
⎡ I 2 + Q2
⎢
)⎤⎥ ⎡ P1 − M 1 ⎤
−1 ⎢ ⎥
⎢ I ⎥ = D . ⎢ P2 − M 2 ⎥ (7.12)
⎢ Q ⎥ ⎢⎣ P3 − M 3 ⎥⎦
⎣ ⎦

where the matrix D is defined as:

⎡ L1 N1Cos (φ1 ) N1Sin (φ1 )⎤

⎢ ⎥
D = ⎢ L2 N 2 Cos (φ2 ) N 2 Sin (φ2 )⎥ (7.13)
⎢⎣ L3 N 3Cos (φ 3 ) N 3 Sin (φ 3 )⎥⎦

where the determinant of matrix D is non-zero. Therefore, the demodulation equa-

tions will be:

I = α1 P1 + α 2 P2 + α 3 P3 − c1 (7.14.1)

Q = β1 P1 + β 2 P2 + β 3 P3 − c2 (7.14.2)

It should be mentioned that the additional equation in the common six-port struc-
ture is only used to simplify I/Q calculations. In this case, it is sometimes better to
reduce the number of ports from six to five, if the complexity of the calculation
remains the same. In regards to (7.9) and (7.12), the additional equation in the six-
port architecture leads to more calibration parameters, which may increase the er-
rors of the I/Q calculations due to errors in the calibration procedure. A low-cost
generation of an I/Q signal, as shown in Figure 7.6, uses conventional Op-Amp
(operational amplifier) circuits [5], although the accuracy may be slightly reduced.

7.2 Smart Antenna Receiver Architectures

The principle of operation of a smart antenna is presented in Figure 7.7 [6]. In this
model, a set of antenna elements is arranged in space; and, the output of each ele-
ment is multiplied by a complex weight and summed. New patterns result from
combining the radiation patterns of the individual elements with phases and
158 Chapter 7 Receiver Design for MIMO Wireless Communications

amplitudes, depending on the weights applied. The weights may be allowed to

vary in time, and the adaptive array is then realized. This array can be used to im-
prove the performance of a mobile communication system by choosing the
weights so as to optimize some measure of the system performance [1], [6].

Element − 1

W1
Element − 2

Output
W2 Σ
Element − n

Wn

Fig. 7.7 Principle of operation of smart antennas

■ Example 7.1:

This example shows how adaptive antennas at a receiver can be used to minimize
the impact of interference [6]. Figure 7.8 shows a desired radiator and an interfer-
ing antenna transmitting co-channel signals s1 and s2, respectively. The signals are
received at a receiver with two independent antenna elements. The signals re-
ceived by the antenna elements are multiplied by complex weights, w, and
summed, yielding the output, y, which would then be demodulated.

S1
h11 Element − 1
h12
W1
h21 Element − 2 y
S2 Σ
h22 W2

Fig. 7.8 Two elements smart antenna

7.2 Smart Antenna Receiver Architectures 159

It is assumed that there is complete channel state information (CSI) at the re-
ceiver, where the channel coefficients, hij, are known. One can set the weights, wj,
in such way the output y minimizes the output due to the interferer, while leaving
the desired signal unaffected. The signals received at the antenna elements are:

x1 = h11 s1 + h21 s2
x2 = h12 s1 + h22 s2

where x1 and x2 are the received signals at element 1 and element 2. The output of
the combiner is:

y = x1 w1 + x2 w2

The output can be written as:

y = ( h11 s1 + h21 s2 ) w1 + ( h12 s1 + h22 s2 ) w2

y = s1 ( h11 w1 + h12 w2 ) + s2 ( h21 w1 + h22 w2 )

If the output must be the same as the signal from the desired mobile, the term mul-
tiplying s1 should be set to 1 and the term multiplying s2 must be set to 0. Accor-
dingly, the weighting functions are obtained as [6]:

h22
w1 =
h11h22 − h12 h21
−h21
w2 =
h11h22 − h12 h21

As may be seen from this example, the selection of the weighting functions is the
main task in the realization of smart antenna architecture. There are different me-
thods to select and adaptively adjust the weighting functions using DSP. The
common techniques to implement the weighting functions are shown in Figure 7.9.
Figure 7.9 (a) shows the implementation of the weighting functions in RF. The
weighting functions are usually realized using the variable attenuator and the vari-
able phase shifter. This technique has also found some applications for uncom-
pressed high-definition (HD) video transmission in a 60 GHz link and is being
supported by IEEE 802.15 [7].
Figure 7.9 (b) shows the implementation of the weighting functions in the IF
section. Figure 7.9 (c) shows the weighting functions implementation in LOs: this
technique is more suitable for superheterydyne receivers. In this design, the varia-
ble gain modules and variable phase shifter are implemented in IF. The weighting
function implementation in baseband is illustrated in Figure 7.9 (d). This tech-
nique is very common, due to its ease of implementation.
160 Chapter 7 Receiver Design for MIMO Wireless Communications

Element − 1
Baseband
W1
Element − 2

Output
W2 DSP
Element − n

Wn

(a)

Element − 1
Baseband
W1
Element − 2
LORF LOIF
Output
W2 DSP
Element − n
LORF LOIF
Wn

LORF LOIF

(b)

Element − 1
Baseband

Element − 2
W1
Output
DSP
Element − n
W2

Wn

(c)

Fig. 7.9 The realization of the smart antenna receiver: (a) weighting function in RF, (b) IF
weighting function in IF, (c) weighting function in LO, and d) weighting function in baseband
7.3 MIMO Receiver Architectures 161

Element − 1
Baseband
W1
Element − 2

Output
W2 DSP
Element − n

Wn

(d)
Fig. 7.9 (continued)

Figure 7.10 illustrates a simple implementation of the smart antenna and beam-
forming structure using weighting functions in baseband and having only an RF
front-end [8]. This structure uses parasitic antenna concepts and is very promising
for the provision of low-cost smart antenna and beamforming systems. The design
technique for this receiver is discussed in Chapter 10.

Active Ref.
Element
RF , frontend
Parasitic
Element #1 Adaptive
Tunable
Beamforming Output
Load − 1
Parasitic Algorithm
Element # n Tunable
Load − n

Baseband

Fig. 7.10 Adaptive smart antenna receiver using parasitic elements

7.3 MIMO Receiver Architectures

The MIMO receiver front-end is traditionally realized using parallel SISO receiv-
ers. In baseband, it is usually assumed that the spatial and temporal properties of
the MIMO impulse response, H(t), can be separated as:

H (t ) = u (t )H (7.15)
where u(t) is the pulse shape.
162 Chapter 7 Receiver Design for MIMO Wireless Communications

The matched filter, H (−t ) , in the MIMO baseband receiver can, therefore, be
*

decomposed into a cascade of a space-only column matrix, H*, and followed by a

bank of time-only matched filters, u (−t ) . This is a common assumption in nar-
rowband models [9]. It is also usually considered that the channel is known on the
receiver side.
In this section, three architectures are discussed to implement an RF front-end
for MIMO receivers. All these architectures use parallel SISO front-ends. There
are also alternative designs that use a single RF front-end for MIMO receivers
[10]. This technique is discussed in details in the last chapter.

7.3.1 Superheterodyne MIMO Front-End

A MIMO superheterodyne receiver is illustrated in Figure 7.11. As can be seen,
the architecture is basically designed based on N parallel SISO receivers. Howev-
er, common LOs, both in RF and IF, are used. The common LO minimizes the
voltage controlled oscillator (VCO) / synthesizer power dissipation and chip area
and prevents pulling between multiple VCOs on the same die [39]. The parallel
down-converted streams are delivered to the baseband section using analog-to-
digital sections. The signals are then detected by decomposing the spatial and
temporal properties of received signals using (7.15). Furthermore, the detected
signals are combined depending on their encoding format. The encoding can be
used to increase the reliability, known as space-time coding, or can be used to in-
crease the rate, referred to as spatial multiplexing.

90 o

90o

90o

Fig. 7.11 The architecture of a MIMO superheterodyne receiver

7.3 MIMO Receiver Architectures 163

Having N antenna elements in this architecture, N SISO front-ends are required.

This makes the receiver expensive and increases the power consumption of the re-
ceiver. The RF impairments, such as phase noise and I/Q imbalance, also impact
on the performance of this receiver. On the positive side, this design has the ad-
vantages of the SISO superheterodyne front-end.

7.3.2 Direct Conversion MIMO Front-End

The architecture of a MIMO direct conversion receiver is shown in Figure 7.12.
This design uses an LO in the same frequency as the input RF signal. Accordingly,
it removes the IF section and uses fewer components compared to superhetero-
dyne architecture.
Moreover, the size and cost of this architecture are usually less than those of the
superheterodyne design. However, the direct conversion design suffers from DC
offset impairment [2]. The modeling and compensation of the DC offset impair-
ment in direct conversion receiver is discussed in the future chapters.

90o

Fig. 7.12 The architecture of a MIMO direct conversion receiver

164 Chapter 7 Receiver Design for MIMO Wireless Communications

7.3.3 Six-Port MIMO Front-End

The architecture of a MIMO six-port receiver is presented in Figure 7.13. This ar-
chitecture essentially belongs to the direct conversion class, where only an LO is
used to convert the RF signal to baseband signals. However, due to the use of
passive circuits and detector diodes instead of mixers, it promises simple and low-
cost implementation, particularly at high microwave and millimeter-wave fre-
quencies. Moreover, this technique provides lower power consumption compared
to the other designs. Meanwhile, the RF impairments in this technique are similar
to those in the direct conversion receivers.

Element #1
LPF

RF Six-Port LPF
BPF LNA
Fabric LPF

LPF

Element #2

LPF A/D
Six-Port LPF
& Output Data
RF LNA DSP
BPF Fabric LPF

LPF

Element #N
LPF

RF Six-Port LPF
BPF LNA
Fabric LPF

LPF

Fig. 7.13 The architecture of a MIMO six-port receiver

7.4 Capacity Reduction of MIMO System due to the Front-End

As seen in (4.32), an antenna can be modeled by a Thevenin equivalent circuit
with the open circuit voltage as u and the impedance, z A = rA + jx A . This model
can be readily extended to an antenna array [11]. The open circuit array voltage is
obtained as:
u = ∑ h i xi + n o (7.16)
i
7.4 Capacity Reduction of MIMO System due to the Front-End 165

where xi is the transmitted signal, hi is the spatial channel vector, and no is the
thermal noise, and nο ~ CN (0, 4kTo B.Re[Z A ]) , where B is bandwidth in Hertz
and ZA is the impedance matrix of the antenna array.
The front-end may include parallel sets of amplifiers, mixers and filters [13-
15]. The impedance matrix of the frond-end is represented as Z . Figure 7.14 al-
so shows that the front-end contributes noise to the delivered signal to the load
[12], [16]. Accordingly, the output signal can be related to input signal, u, as:
N
r = ∑ h i xi + n = Hx + n (7.17)
i =1

where n=no+z, z is the noise contribution from the front-end, n ~ CN (0, R nn )

and, hi ~ CN (0, R hh ) where R nn is the covariance matrix of noise, and R hh is
the covariance matrix of the channel.
For the correlated antenna, the Ergodic capacity with complete channel state in-
∑ E xi
2
formation (CSI) at the receiver and power constraint < P can be ob-
i
tained as [15]:
C = E[log2 det(1 + HHW RHW )] (7.18)

where

P 1/ 2 -1 1/ 2
R= Rh Rn Rh (7.19)
N

This matrix (7.19) is called the signal-to-noise ratio (SNR) matrix as it is analog-
ous to the SNR of a SISO system [16]. By extending relation (4.33) of the SISO
case to the MIMO system, the noise factor matrix for the MIMO front-end can be
defined as:

F = (R n z =0
)−1/2 R n (R n z=0
)1/ 2
(7.20)

The SNR matrix may be also presented, in terms of noise factor matrix, as [16]:

P −1/2 −1
R= R1/ 2
h Re{Z A } F Re{Z A }−1/ 2 R1/
h
2
(7.21)
4kT0 BN

It is derived that two otherwise identical MIMO systems, having front-ends with
noise factor matrices of F1 ≤ F2 , result in corresponding capacities of C1 ≥ C2 ,
with equality if and only if F1 = F2 [16].
166 Chapter 7 Receiver Design for MIMO Wireless Communications

u1 n11 n21
ZL
⎛ z11 z12 ⎞
ZA ⎜ ⎟
uM n1M ⎝ z 21 z 22 ⎠ n2M
ZL

Fig. 7.14 A circuit model for a receiver front-end

7.5 Radio Frequency Interference on MIMO Receivers

Radio frequency interference (RFI) is a combination of independent radiation
events that usually have non-Gaussian statistics. As an example, a computer has
emissions of varying intensities that interfere with wireless transceivers. Some in-
terfering sources from computer subsystems are shown in Table 7. 1.

Table 7.1 Example of Computer Subsystems Interfering with Wireless Standards [18]

Carrier Example Interfering

Standard Wireless Networking
(GHz) Computer Subsystems
Gigabit Ethernet, PCI Express
Bluetooth 2.4 Personal Area Network
Bus, Memory, Processor, LCD
IEEE Gigabit Ethernet, PCI Express
2.4 Wireless LAN (Wi-Fi)
802.11b/g Bus, Memory, Processor, LCD
High-Speed Gigabit Ethernet, PCI Express
IEEE 802.11n 2.4
Wireless LAN Bus, Memory, Processor, LCD
2.5–2.69
Mobile Broadband PCI Express Bus, Memory,
IEEE 802.16e 3.3–3.8
(Wi-Max) Processor, LCD
5.725–5.85
PCI Express Bus, Memory,
IEEE 802.11a 5.2 Wireless LAN (Wi-Fi)
Processor, LCD

The impact of computational platform generated RFI on wireless local area

network (WLAN, IEEE 802.11b/g) data communication transceivers has been
studied [19]. It was shown that computational platform generated RFI may in-
crease the receiver noise floor ten-fold, causing a more than 50% loss in range and
a significant impact on the throughput performance. As an illustration, the impact
of LCD (liquid crystal display) noise on the throughput of an IEEE802.11g em-
bedded receiver is presented in Figure 7.15 [18]. As may be seen, the LCD interfe-
rence may highly reduce the throughput of the wireless system.
7.5 Radio Frequency Interfe
ference on MIMO Receivers 1667

The common RFI mod del for single antenna systems are Middleton class A, B
and C models [17]. RFI may have an even larger impact on MIMO transceiveer
performance. The extensiion of the single antenna RFI models to MIMO system is
not a very straightforward
d task. Figure 7.16 shows the performance degradation oof
a typical MIMO receiver in the presence of RFI.The RFI model is an additive bbi-
variate Middleton class A noise model [20]. In this model, the energy of the Midd-
dleton Class A model is a sum of the Gaussian and impulsive components of thhe
noise. At high SNRs, thee receivers become sensitive to the impulsive componennt
of RFI, thereby causing seevere degradation of performance due to RFI.

Fig. 7.15 The impact of LCD

D interference on throughput of an IEEE 802.11g receiver: (topp)
channel 1 (bottom) channel 7 © IEEE [19]
168 Chapter 7 Receiver Design for MIMO Wireless Communications

0
10

-1
10
Vector Symbol Error Rate

-2
10

-3
10
SM with ML (Gaussian noise)
SM with ZF (Gaussian noise)
-4
10 Alamouti coding (Gaussian noise)
SM with ML (Middleton noise)
SM with ZF (Middleton noise)
-5
Alamouti coding (Middleton noise)
10
-10 -5 0 5 10 15 20
SNR [in dB]

Fig. 7.16 The performance of spatial multiplexing (SM) and Alamouti coding MIMO re-
ceivers in presence of Gaussian and bivariate Middleton class A noise © Springer [20]

7.6 MIMO Testbed Design

Wireless MIMO communications have necessitated a variety of performance ex-
aminations; therefore, a MIMO testbed, which shows the system performance and
examines the analytical results, is required in research laboratories in both acade-
mia and industry. Moreover, the performance degradation of the system over
realistic and imperfect channels needs to be accurately characterized. The first
narrowband MIMO prototype was vertical BLAST (Bell Laboratories Layered
Space-Time) or V-BLAST (Vertical BLAST) architecture [21]. This was followed
by several MIMO testbeds that have been reported in the literature, e.g. [22]-[32].
A modem MIMO testbed design needs flexibility in software, since firmware im-
plementation of different transmission schemes, such as space-time processing, is
typically tedious work. In this section, a typical MIMO test-bed is described.

7.6.1 Testbed Architecture

Figure 7.17 shows a functional block diagram of a testbed. For a given data
stream, a program in a PC builds up frames by using a desired transmission
scheme. These frames are then transmitted to the field-programmable gate array
7.6 MIMO Testbed Design 169

(FPGA) board and are buffered in its internal RAM (random access memory). Af-
ter some processing, these signals are converted to analog signals by four digital-
to-analog converters (DACs). The analog converted signals are fed to the RF
transmitters.
On the receiver side, the received signals are down-converted by means of four
synchronous superheterodyne receivers. The digitized signals are fed to the FPGA.
An energy detector is implemented in the FPGA to detect the frame position [33].
The received frames are sent to the PC for processing.
The testbed specifications are presented in Table 7.2, and the various sections
of the testbed are described in the following subsections.

FPGA FPGA

PC & & PC

DAC ADC

RF Transmitters RF Receivers

Fig. 7.17 Functional block diagram of the testbed

Table 7.2 Testbed specifications

Parameter Description

Radio Frequency 2.4~2.5 GHz

Bandwidth <4 MHz
Modulation Any
Channel Up to 4×4
TX DAC Resolution 12 bit
RX ADC Resolution 12 bit
TX Output power <18 dBm
TX IF 21.4 MHz
RX IF1 268 MHz
RX IF2 10.7 MHz
Connection type USB
170 Chapter 7 Receiver Design for MIMO Wireless Communications

7.6.1.1 RF Transmitter

The RF transmitter module for a path is shown in Figure 7.18. The RF transmitter
consists of an I/Q mixer, a common oscillator for all modulators and a power am-
plifier with 18 dBm maximum output power. The input signal to the transmitter
block is fed by a DAC. A common LO with a power level of -7 dBm is used to
provide the oscillator signals for all paths. It is shown in Figure 7.19.

Fig. 7.18 The transmitter module

Fig. 7.19 Transmitter oscillator in 2358.6 MHz

7.6 MIMO Testbed Design 171

7.6.1.2 RF Receiver

The RF receiver diagram is shown in Figure 7.20 The RF receiver consists of a band-
pass filter (BPF), a low-noise amplifier (LNA), an image rejection (IMR) filter, an RF
to IF down-converter, two common oscillators for all down-converters and an auto-
matic gain control (AGC) amplifier. The first stage of the receiver is a micro strip
BPF filter to protect the following LNA from blocking by other radio applications.
An IMR filter is used to additionally suppress image frequency signals.
The down-conversion to both the first and second IF frequencies is performed by a
double-balanced mixer. The LO signals with a power level of +12 and +7 dBm are
applied for the first and second mixers, respectively. A bandpass SAW (surface
acoustic wave) filter is applied at the first IF stage to limit the noise bandwidth of the
RF system to 4 MHz and to filter undesired spectral components. An AGC amplifier
with 68 dB dynamic range is in the last stage of the receiver. The 10.7 MHz IF signal
feeds the ADC. The RF receiver module is shown in Figure 7.21.

IMR SAW
BPF LNA AGC
Filter Filter

LO1 LO2

Fig. 7.20 Receiver functional diagram

Fig. 7.21 Receiver module

172 Chapter 7 Receiver Design for MIMO Wireless Communications

The developed testbed is shown in Figure 7.22 In addition, a channel simulator

can be implemented in MATLAB software and is placed after MIMO transmitter.
The overall system is shown in Figure 7.23. This testbed is used in the future
chapters to investigate the performance of MIMO transceivers.

(a)

(b)

Fig. 7.22 Block diagrams of (a) the RF receiver, and (b) the RF transmitter
7.7 Commercial MIMO Transceivers 173

Transmitters

Attenuators

Receivers
FPGA FPGA
MIMO MIMO
H 0 −1

RF
RF
H & &
Transmitter Receiver
DACs ADCs
MATLAB

Feedback

Fig. 7.23 Block diagram of the overall testbed

7.7 Commercial MIMO Transceivers

Some manufacturers have recently introduced a few MIMO transceivers to the
wireless market. These transceivers usually target commercial applications, in-
cluding WiMAX (Worldwide Interoperability for Microwave Access) and LTE
(Long Term Evolution).

7.7.1 UXA234 MIMO Transceiver from NXP-Philips

The functional block diagram of the UXA234 MIMO transceiver from NXP-
Philips is shown in Figure 7.24. The UXA234 covers the frequencies from 2.3
GHz to 3.8 GHz using conversion transceiver structures. A direct conversion ar-
chitecture is used to achieve the lowest power consumption and the least number
of external components [34]. On the receiver side, the signal enters at the LNA, is
mixed down in an I/Q mixer to a zero-IF. It is then low-pass filtered and buffered
before being fed to the ADC (located in the baseband chip). On the transmitter
side, I/Q signals are created by the current mode DAC of the baseband integrated
circuit (IC) and converted on to the transceiver’s IC to a voltage. The signal is
then passed through reconstruction filters before being mixed up to the RF and
amplified. The LO signal is shared between the receiver and transmitter.
On the receiver side, the high- and mid-gain modes of the LNA are imple-
mented with active circuitry, while the low-gain mode is passive. Over the whole
receiver band, the RF front-end requires a very low noise figure of 3 dB. The cir-
cuit has been fabricated with SiGe (silicon-germanium) BiCMOS technology [3].
Without calibration, the receiver’s error vector magnitude (EVM) is reported at
about -32dB and, after calibration, is about - 38dB [35].
174 Chapter 7 Receiver Design for MIMO Wireless Communications

Fig. 7.24 Functional block diagram for the UXA234 MIMO transceiver © NXP-Philips
[35]

7.7.2 MAX2842 MIMO Transceiver from MAXIM

The functional block diagram of the MAX2842 MIMO transceiver is shown in
Figure 7.25 [36]. The MAX2842 is a direct conversion zero-IF RF transceiver that
operates from 3.3 GHz to 3.9 GHz. It has two transmitters and two receivers and
may be used to realize a 2x2 MIMO system. The receiver noise figure is 3.8 dB. This
transceiver supports channel bandwidths of 3.5 MHz, 5 MHz, 7 MHz, and 10 MHz.
7.7 Commercial MIMO Transceivers 175

Fig. 7.25 Functional block diagram for the MAX2842 MIMO transceiver © MAXIM [36]

The IC integrates an on-chip AM (amplitude modulation) detector for measur-

ing the transmitter’s I/Q imbalance and LO leakage. An internal transmit-to-
receive loopback mode allows for receiver I/Q imbalance calibration. The
MAX2842 completely eliminates the need for external SAW filters by implement-
ing on-chip programmable monolithic filters for both the receiver and transmitter
for channel bandwidths from 3.5MHz to 10MHz [36].

7.7.3 PM8800 MIMO Transceiver from PMC-Seirra

The functional block diagram of the PM8800 MIMO transceiver is shown in Fig-
ure 7.26 [37]. The PM8800 2x2 is a zero-IF MIMO transceiver. It has two inde-
pendent radio paths (2 transmitters and 2 receivers) and supports operation in the
2.3 GHz to 2.9 GHz band or the 3.3 GHz to 3.8 GHz band.
176 Chapter 7 Receiver Design for MIMO Wireless Communications

Fig. 7.26 Functional block diagram for the PM8800 MIMO transceiver © PMC-Seirra [37]

The EVM in the transmitting path is about -36 dB at output power of 0 dBm
(2.5 GHz) and -34 dB at output power of 0 dBm (3.5 GHz). Moreover, the EVM
in the receiver path is about -36 dB at 2.5 GHz and -34 dB at 3.5 GHz. The noise
figure of the receiver is about 3 dB. Moreover, the adjacent channel power ratio
(ACPR) for 64QAM (quadrature amplitude modultion) is about 26 dB [37].

7.7.4 AD9356 MIMO Transceiver from Analog Devices

A functional block diagram of the ADS9356 MIMO transceiver is shown in
Figure 7.27 [38]. The AD9356 is an RF transceiver with dual receivers and trans-
mitters and operates in the 2.3 GHz to 2.7 GHz range. This transceiver supports
channel bandwidths of 3.5 MHz, 4.375 MHz, 5 MHz, 7 MHz, 8.75 MHz, and 10
MHz. The receiver uses direct conversion architecture. The transmitting path has
excellent spectral purity with sideband noise of less than -130 dBc/Hz at an 8 MHz
frequency offset and offers an EVM of -40 dB. Two high dynamic range ADCs,
followed by decimation and channel filters, digitize the received signals and pro-
duce 12-bit output signals at a sample rate determined by the bandwidth mode.
The transmitting path takes 12-bit input data and interpolates before converting
to the analog domain and up-converting to the carrier frequency [38]. This chip
can be used in MIMO base station design by adding suitable power amplifiers and
duplexers. This is presented in Figure 7.28.
7.7 Commercial MIMO Transceivers 177

Fig. 7.27 Functional block diagram for the AD9356 MIMO transceiver © Analog Devices
[38]

Fig. 7.28 2×2 MIMO base station transceiver architecture using the AD9356 MIMO tran-
sceiver ©Analog Device [40].
178 Chapter 7 Receiver Design for MIMO Wireless Communications

References

[1] Ellinger, F.: Radio Frequency Integrated Circuits Technology, 2nd edn. Springer,
Heidelberg (2008)
[2] Razavi, B.: RF Microelectronics. Prentice Hall (1998)
[3] Crols, J., Steyaert, M.S.J.: Low-IF Topologies for High-Performance Ana-log Front
Ends of Fully Integrated Receivers. IEEE Transactions on Circuits and Systems II 45,
269–282 (1998)
[4] Ghannouchi, F.M., Mohammadi, A.: Six-port Technique with Microwave and Wire-
less Applications. Artech House (2009)
[5] Mirzavand, R., Mohammadi, A., Ghannouchi, F.M.: Five-port receivers, architectures
and applications. IEEE Communications Magazine 48(6), 30–36 (2010)
[6] Saunders, S.R., Aragon-Zavala, A.: Antenna and Propagation for Wireless Communi-
cations, 2nd edn. John Wiley & Sons (2007)
[7] Kraemer, R., Katz, M.D.: Short Range Wireless Communications, Emerging Tech-
nologies and Applications. Wiley (2009)
[8] Chandran, S.: Advances in Direction-of-Arrival Estimation. Artech House (2006)
[9] Barry, J., Lee, E.A., Messerschmitt, D.G.: Digital Communications, 3rd edn. Kluwer
Academic (2004)
[10] Lari, M., Bassam, A., Mohammadi, A., Ghannouchi, F.M.: Time-Multiplexed Single
Front-End MIMO Receivers with Preserved Diversity Gain. IET Communications
(2010)
[11] Balanis, C.A.: Antenna Theory: Analysis and Design, 3rd edn. John Wiley (2005)
[12] Rothe, H., Dahlke, W.: Theory of noisy four poles. In: Proceeding of IRE, vol. 44,
pp. 811–818 (June 1956)
[13] Gans, M.J.: Channel capacity between arrays – Part I: Sky noise dominates. IEEE
Transactions on Communications 54, 1586–1592 (2006)
[14] Gans, M.J.: Channel capacity between arrays – Part II: Amplifier noise dominates.
IEEE Transactions on Communications 54, 1983–1992 (2006)
[15] Shiu, D., et al.: Fading correlation and its effect on the capacity of multielement an-
tenna systems. IEEE Transactions on Communications 48, 502–513 (2000)
[16] Domizioli, C.P., Hughes, B.L., Gard, K.G., Lazzi, C.: Optimal front-end design for
MIMO receivers. In: IEEE Global Communications Conference, Globcom 2008,
New Orleans (2008)
[17] Middleton, D.: Non-Gaussian noise models in signal processing for telecommunica-
tions: New methods and results for Class A and Class B noise models. IEEE Transac-
tions on Information Theory 45(4), 1129–1149 (1999)
[18] Nassar, M., Gulati, K., DeYoung, M.R., Evans, B.L., Tinsley, K.R.: Mitigating near-
field interference in laptop embedded wireless transceivers. Journal of Signal
Processing Systems (March 2009)
[19] Shi, J., Bettner, A., Chinn, G., Slattery, K., Dong, X.: A study of platform EMI from
LCD panels - impact on wireless, root causes and mitigation methods. In: IEEE Inter-
national Symposium on Electromagnetic Compatibility, August 14-18, vol. 3,
pp. 626–631 (2006)
References 179

[20] Gulati, K., Chopra, A., Heath Jr., R.W., Evans, B.L., Tinsley, K.R., Lin, X.E.: MIMO
Receiver Design in the Presence of Radio Frequency Interference. In: Proc. IEEE
Global Communications Conference, November 30 - December 4, New Orleans
(2008)
[21] Wolniansky, P.W., Foschini, G.L., Golden, G.D., Valenzuela, R.A.: V-Blast: an ar-
chitecture for realizing very high data rates over the richscattering wireless channel.
In: Proceedings of International Symposium on Signals, Systems, and Electronics
(ISSSE 1998), Pisa, Italy, pp. 295–300 (September- October 1998)
[22] Sampath, H., Talwar, S., Tellado, L., Erceg, V., Paulraj, A.: A fourth-generation
MIMO-OFDM broadband wireless system: Design, performance, and field trial re-
sults. IEEE Communications Magazine 40(9), 143–149 (2002)
[23] Murphy, P., Lou, F., Sabharwal, A., Frantz, L.P.: An FPGA based rapid prototyping
platform for MIMO systems. In: Conference Record of the 37th Asilomar Conference
on Signals, Systems, and Computers, ASILOMAR 2003 (November 2003)
[24] Adjoudani, A., Beck, E.C., Burg, A.P., et al.: Prototype experience for MIMO
BLAST over third-generation wireless system. IEEE Journal on Selected Areas in
Communications 21(3), 440–451 (2003)
[25] Aschbacher, E., Langwieser, R., Caban, S., Mehlfuehrer, S., Caban, G., Keirn, W.,
Scholtz, A.L., Rupp, M.: Design and subsystem verification of a flexible and scalable
4X4 MIMO testbed. In: IEEE Radio and Wireless Conference (RAWCON 2004),
Workshop on MIMO Implementation Aspects (Septembeer 2004)
[26] Garrett, D., Woodward, G., Davis, L., Knagge, G., Nicol, C.: A 28.8 Mb/s 4x4
MIMO 3G highspeed downlink packet access receiver with normalized least mean
square equalization. In: Proceedings of IEEE International Solid-State Circuits Con-
ference (ISSCC 2004), San Francisco, vol. 1, pp. 420–536 (February 2004)
[27] Borkowski, D., Bruhl, L.: Hardware implementation for real-time multi-user MIMO
systems. In: Proceedings of IEEE Radio and Wireless Conference (RAWCON 2004)
Workshop 2 on MIMO Implementation Aspects, Atlanta (September 2004)
[28] Bialkowski, K.S., Zagriatski, S., Postula, A., Bialkowski, M.E.: MIMO Test-bed with
an insight into signal strength distribution around transmitter/receiver sites. In: Pro-
ceedings of EuMW 2005, Paris, pp. 117–120 (October 2005)
[29] Liberti, J.C., Koshy, J.C., Hoerning, T.R., Martin, C.C., Dixon, J.L., Triolo, A.A.,
Murray, R.R., McGiffen, T.G., Military, T.: Experimental results using a MIMO test
bed for wideband, high spectral efficiency tactical communications. In: IEEE Com-
munications Conference, MILCOM, October 17-20, vol. 3, pp. 1340–1345 (2005)
[30] Bialkowski, K.S., Postula, A., Abbosh, A., Bialkowski, M.E.: 2x2 MIMO Testbed for
Dual 2.4GHz/5GHz Band. In: International Conference on Electromagnetics in Ad-
vanced Applications, ICEAA 2007, September 17-21, pp. 1–4 (2007)
[31] Chiu, C.Y., Cheng, C.H., Wan, Y.S., Rowell, C.R., Murch, R.D.: Design of a Flat
Fading 4 x 4 MIMO Testbed for Antenna Characterization using a Modular Ap-
proach. In: IEEE Conference on Wireless Communications and Networking, Kow-
loon, pp. 2913–2918. Dept. of Electron. Comput. Eng, Hong Kong Univ. of Sci.
Technol. (March 2007)
[32] Spring, R., Zhou, L., Gogate, N., Daryoush, A.S.: 4x4 MIMO Experimental Test-bed
using COTS at ISM Band. In: IEEE Radio and Wireless Symposium, pp. 173–176.
Dept. of ECE, Drexel Univ., Philadelphia, PA (2007)
[33] Azami, F., Ghorssi, A., Hemesi, H., Mohammadi, A., Abdipour, A.: Design and Im-
plementation of a Flexible 4x4 MIMO Testbed. In: IEEE International Symposium on
Telecommunications, IST 2008, Isfehan (2008)
180 Chapter 7 Receiver Design for MIMO Wireless Communications

[34] Locher, M., Tomesen, M., Kuenen, J., Daanen, A., Visser, H., Essink, B., Vervoort,
P.P., Nijrolder, M., Kopmeiners, R., Redman-White, W., Balmford, R., El Waffaoui,
R.: A Low Power, High Performance BiCMOS MIMO/Diversity Direct Conversion
Transceiver IC for WiBro/WiMAX (802.16e). In: IEEE 2007 Custom Intergrated
Circuits Conference, CICC (2007)
[35] NXP Philips Inc., WiMAX 802.16e MIMO transceivers UXA234xx (2007)
[36] Maxim Integrated Product, 3.3GHz to 3.9GHz MIMO Wireless Broadband RF Tran-
sceiver, 2842 data sheet (2010)
[37] PMC Seirra Inc., PM8800 WiZIRD 2x2, Released Product Brief (2008)
[38] Analog Devices Inc., WiMAX/BWA/WiBRO/LTE RF MxFE 2 × 2 MIMO Tran-
sceiver, AD9356 data sheet (2010)
[39] Palaskas, Y., Ravi, A., Pellerano, S., Carlton, B.R., Elmala, M.A., Bishop, R., Baner-
jee, G., Nicholls, R.B., Ling, S.K., Dinur, N., Taylor, S.S., Soumyanath, K.: A 5-GHz
108-Mb/s 2x2 MIMO Transceiver RFIC With Fully Integrated 20.5-dBm P1dB Pow-
er Amplifiers in 90-nm CMOS. IEEE Journal of Solid-State Circuits 41(12),
2746–2756 (2006)
[40] Wier, P.: Synchronizing NxN MIMO Basestations to an External Timing Reference.
Analog Devices Inc. (2010)
Chapter 8
RF Impairments in OFDM Transceivers

Orthogonal frequency-division multiplexing (OFDM) is a very attractive transmis-

sion technique for future high data rate wireless multimedia communication sys-
tems. OFDM has been developed for wideband wireless digital communication
and is used in applications such as Digital Video and Audio Broadcasting (DVB,
DAB, respectively), wireless networking and broadband Internet access [1]-[3].
OFDM has advantages, such as high spectral efficiency, easy adaptation to se-
vere channel conditions without complex equalization, and efficient implementa-
tion using fast Fourier transform (FFT). However, it also has some disadvantages,
such as sensitivity to frequency offset, a high peak-to-average-power ratio (PAPR)
and sensitivity to noises [1], [4], [5].
As discussed in Chapter 3, OFDM is successfully used in MIMO systems. This
chapter is devoted to investigate the various RF impairments, specifically noise
and nonlinearity, and their mutual impacts on OFDM transceivers.

8.1 OFDM Transceivers

Orthogonal frequency-division multiplexing is a frequency-division multiplexing
(FDM) scheme utilized as a digital multicarrier modulation method. A large num-
ber of closely spaced orthogonal subcarriers are used to carry data, which are di-
vided into several parallel data streams or channels, one for each subcarrier. Each
sub-carrier is modulated with a conventional modulation scheme, such as quadra-
ture amplitude modulation (QAM) or phase-shift keying (PSK) at lower symbol
rates, maintaining total data rates similar to conventional single-carrier modulation
schemes in the same bandwidth [2], [20]. Figure 8.1 shows a block diagram of an
OFDM transceiver system.
The input data is first de-multiplexed into N parallel streams, and each one is
mapped to a (possibly complex) symbol stream using some modulation constella-
tion (QAM, PSK, etc.). An inverse FFT is computed on each set of symbols,
giving a set of complex time-domain samples. These samples are then quadrature-
mixed to passband in the standard way. The real and imaginary components are
first converted to the analog domain using digital-to-analog converters (DACs);
and, the analog signals are then used to modulate cosine (real) and sine (imagi-
nary) waves at the carrier frequency, fc. These signals are then summed to give the
transmission signal, S (t ) .

A. Mohammadi et al.: RF Transceiver Design for MIMO Wireless Communications, LNEE 145, pp. 181–226.
springerlink.com © Springer-Verlag Berlin Heidelberg 2012
182 Chapter 8 RF Impairments in OFDM Transceiveers

Fig. 8.1 Block diagram of OF

FDM communication system

t signal, r (t ) , which is then quadrature-mixed down tto

The receiver picks up the
baseband using cosine an nd sine waves at the carrier frequency. This also createes
signals centered on 2fc, so
o low-pass filters are used to reject these. The basebannd
signals are then sampled and digitized using analog-to-digital converters (ADCss),
and a forward FFT is used d to convert back to the frequency domain. This returns N
parallel streams, each of which
w is converted to a binary stream using an approprri-
ate symbol detector. Theese streams are then re-combined into a serial stream m,
which is an estimate of th
he original binary stream at the transmitter [1], [20].

8.2 Noise in OFDM

M Transceivers

It is very important to exaactly predict and analyze the oscillator noise influences iin
OFDM communication systemss and quantify the tolerable level of noise. Phasse
noise effects in OFDM haave been analyzed in many papers by several authors [44],
[5], [7]-[19]. However th he effect of amplitude noise was not considered and aal-
ways ignored relative to phase
p noise. In this section, the theoretical analysis of thhe
impact of the oscillator phase and amplitude noise on the signal-to-noise ratiio
(SNR) and, hence, the bit error rate (BER) performance of MQAM-OFDM signaals
over an additive white Gaussian
G noise (AWGN) channel are presented; and, thhe
tolerable levels of phase and
a amplitude noise are extracted.
8.2 Noise in OFDM Transceivers 183

8.2.1 Phase and Amplitude Noise

An oscillator is a system that generates a periodic signal with a specified or con-
trollable frequency. There are the two kinds of noise in radio frequency (RF) oscil-
lator output, namely amplitude noise and phase noise. The output voltage of an
oscillator can be represented by [6]:
Vout (t ) = ( A0 + aN (t )).Cos(2π f 0 t + ϕ N (t )) (8.1)

Vout (t ) = A(t ).Cos(2π f 0 t + ϕ (t )) (8.2)

where A0 is the average amplitude of the output signal, f 0 is nominal frequency

of oscillation, and aN (t ) and ϕ N (t ) represent the time-varying components of the
amplitude and phase, respectively. These components are considered noise, be-
cause an ideal oscillator would have constant amplitude A0 and phase varying at a
constant rate of 2π f 0 .
It can be shown that the white noise power of the amplifier electronics is equal-
ly partitioned into amplitude and phase noise components with a zero mean and
variances of σ A2 and σ ϕ2 , respectively [13]. The SNR performances that can be
achieved in an OFDM system with a noisy oscillator can be approximated by inte-
grating its phase and amplitude noise power density functions (PDFs) over the
channel bandwidth. The phase noise PDF of an oscillator is approximated by a Lo-
renzian function with uniform phase distribution. It is parameterized by its total
integrated phase noise, K, and -3dB bandwidth, B, to which we superimpose a
noise floor (set at 35dB above the thermal noise), as shown in (8.3) [5]:

1 K .B
L( f )2 = . + L0 (8.3)
π f 2 + B2

1 K .B
L( f )2 = . + L0 (8.3)
π f 2 + B2
Figure 8.2 shows phase noise PDFs for different K and a -3dB bandwidth of 100
Hz. The amplitude noise PDF of an oscillator is similar to the phase noise PDF,
but is, on average, n dB less than it is in spectrum bandwidth, where n for a prac-
tical parameter of an oscillator is normally between 10 to 20 dB relative to the os-
cillator’s quality [21]. Therefore, the phase and amplitude noise variances are:
Bs Bs
2 2
1 KB 2K B 1
σ =2 ∫ L( f ) df = 2 ∫ + L0 ) df = arctg ( s ) + L0 Bs
2 2
ϕ ( (8.4a)
0
π 0 f +B
2 2
π 2B π

n
−
σ A2 = σ ϕ2 ×10 10 (8.4b)
184 Chapter 8 RF Impairments in OFDM Transceiveers

Fig. 8.2 Phase noise PDFs fo

or different K, B = 100 Hz.

Table 8.1 shows the phase

p and amplitude noise variances of an oscillator foor
different K and n. In the next
n section, we use these values for calculating MQAMM-
OFDM degradation caused by phase and amplitude noise in simulation. Noisse
includes components at many
m frequencies, so its phase with respect to the maiin
carrier is random, and its amplitude is also random.
Figure 8.3 shows noisse added to the carrier phasor. Random amplitude annd
phase variation due to nooise changes the constellation of M-QAM modulated sigg-
nal and degrade the perforrmance of the systems.

Fig. 8.3 Noise added to the carrier

c phasor [22]
8.2 Noise in OFDM Transceivers 185

Table 8.1 Phase and Amplitude Noise Variances for Different K , n

Bs(Hz) 107
K -30 -20 -10 -5 0
σϕ2
0.0010 0.0100 0.1 0.3158 1
n = 100 1e-13 1e-12 3.158e-11 1e-10 1e-10
n = 20 0.00001 0.0001 0.0032 0.0100 0.0100
0.00003
σ A2 n = 15 0.00032 0.0100 0.0316 0.0316
2
n = 10 0.0001 0.0010 0.0316 0.1000 0.1000
n=5 0.00032 0.0032 0.1000 0.3162 0.3162

8.2.2 Analysis of Phase and Amplitude Noise Impacts in OFDM

As shown in Figure 8.1, if N subcarriers are used, and each subcarrier is mod-
ulated using M alternative symbols, the OFDM symbol alphabet consists of MN
combined symbols. The low-pass equivalent OFDM signal is expressed as [1]:

N −1
S (t ) = ∑ X k e j 2π kt / T 0≤t <T (8.5)
k =0

where j = −1 , X k is the data symbol for the kth subcarrier, N is the number of
subcarriers, and T is the OFDM symbol time.
The subcarrier spacing of 1/T makes them orthogonal over each symbol period.
S (t ) is corrupted by the phase and amplitude noise of the transmitter’s local os-
cillator (LO). The received signal is also influenced by the phase and amplitude
noise of the receiver’s LO. It can be assumed that the phase and amplitude noises
of the LOs at the transmitter and the receiver are independent and identical. There-
fore, it is expressed as follows:

r (t ) = {S (t ). ATX (t ).e jϕTX (t ) + n(t )}. ARX (t ).e jϕRX (t ) (8.6)

where n(t) is the complex Gaussian noise, and ϕTX (t ), ATX (t ) & ϕ RX (t ), ARX (t ) are
the time-varying phase and amplitude noise processes generated by the RF LO at
the transmitter (TX) and the receiver (RX), respectively.
186 Chapter 8 RF Impairments in OFDM Transceivers

The sampled signal for the kth subcarrier after the FFT processing stage in the
receiver can be written as:

N −1 2π
1 −j
∑ r ( m) e
km
Yk = N
N m =0
N −1 2π
1 −j
∑ S (m) ATX (m) ARX (m)e j (ϕTX (m )+ϕRX (m )) e
km
= N
N m =0
N −1 2π
1 −j
∑ n(m) ARX (m).e jϕRX (m) e
km
+ N
N m =0
(8.7)
N −1 N −1 2π 2π
1 −j
∑ (∑ Xle
j lm km
= N ). A(m).e jϕ ( m ) e N + NK
N m=0 l =0
N −1 N −1 2π
1 −j (l − k ) m
=
N
∑ X l ∑ A(m)e jϕ (m) e N + NK
l =0 m=0
N −1
= ∑ X l Ql −k + N K
l =0

where N K is a sampled FFT version of the AWGN noise multiplied by the phase
and amplitude noise of the RX LO with a variance of σ n2 . A(m) and ϕ (m) are
defined as :
⎧ A( m)  ATX (m) ARX (m)
⎪
⎨ (8.8)
⎪ϕ (m)  ϕ (m) + ϕ (m)
⎩ TX RX

and Qk is given by :

N −1 2π
1
∑ A(m)e
j km
jϕ ( m )
Qk = e N (8.9)
N m=0

The received signal is composed of three contributions: Yk1 is the kth desired sub-
carrier, Yk2 are the other subcarriers, and N k is the AWGN [7], [9], [10], [13]-[15]:

Yk = YK1 + Yk2 + N k (8.10)

or:

N −1
Yk = X k .Q0 + ∑ X l .Ql − k + N k (8.11)
l = 0,l ≠ k
8.2 Noise in OFDM Transceivers 187

The expected value of X k . Q0 is the desired signal component; therefore, the de-
sired signal power can be found as:

Ps − desired = [ E[ X k Q0 ]]2 = Ps [ E[Q0 ]]2 (8.12)

where:

N −1 σ ϕ2
1 1 −
E[Q0 ] =
N
∑ EA(m)e jϕ ( m )
= .N .EA(m)e jϕ ( m ) = EA(m) Ee jϕ ( m ) = e
N
2 (8.13)
m=0

+∞
ϕ2 +∞
ϕ2 σ ϕ2
− + jϕ −
1 2σ ϕ2 1 2σ ϕ2
∫e ∫
jϕ ( m ) jϕ
Ee = . .e dϕ = .e dϕ = e 2 (8.14)
−∞ 2π .σ ϕ −∞ 2π .σ ϕ

Therefore:

−σ ϕ2
Ps − desired = Ps .e (8.15)

The variance of X k . Q0 is the phase and amplitude noise power caused by com-
mon phase error (CPE); and, PCPE is calculated as follows:

PCPE = Var[ X k Q0 ] = Ps .Var[Q0 ] (8.16)

where:

N −1 N −1
1 1
Var[Q0 ] = Var[
N
∑ A(m)e jϕ ( m) ] = N 2 ∑ Var ( A(m)e jϕ (m ) )
m=0 m=0
1 −σ 2
= ( E | A(m) |2 . | e jϕ |2 − A02 .e ϕ ) (8.17)
N
1 −σ 2
= ((1 + σ A2 ) − e ϕ )
N

Therefore:

1 −σ 2
PCPE = Ps × ((1 + σ A2 ) − e ϕ ) (8.18)
N
188 Chapter 8 RF Impairments in OFDM Transceivers

N −1
Next, ∑ X l .Ql − k is the interference component caused by ICI (inter-subcarrier
l = 0,l ≠ k

interference). So, PICI can be calculated as:

N −1
PICI = Ps .{Var[ ∑ Ql ] − Var[Q0 ]} (8.19)
l =0

where:

N −1 N −1 N −1 2π
1
Var[ ∑ Ql ] = Var[ ∑ { ∑ A(m)e jϕ ( m) e
j lm
N }]
l =0 l =0 N m =0
N −1 N −1 2π
1
Var[ ∑ { ∑ A(m)e jϕ ( m ) e N }]
j lm
= 2
(8.20)
N l = 0 m =0
−σ ϕ2
= (1 + σ A2 − e )

Therefore:

N −1 −σ 2
.(1 + σ A2 − e ϕ )
PICI = Ps . (8.21)
N
N k is a sampled FFT version of the AWGN noise multiplied by the phase and
amplitude noise of the RX LO, as seen in (7). Hence, for the sake of simple analy-
sis, we consider Var[ N k ] as that of the AWGN [38]:

Pwith{ϕ , A} = Var [ N k ] = Pn (8.22)

From (8.12) to (8.22), the SNR, including the phase and amplitude noise, can be
arranged as :

−σ 2 −σ 2
S PS .e ϕ PS .e ϕ
( )With{ϕ , A} = =
( PCPE + PICI ) + Pn P (1 + σ 2 − e−σ ϕ ) + P
2
N
S A n
S −σ 2 (8.23)
( )Without{ϕ , A} e ϕ
= N
S −σ 2
1+ ( )Without{ϕ , A} (1 + σ A2 − e ϕ )
N
8.2 Noise in OFDM Transceivers 189

The above-mentioned powers can be used to evaluate the performance degradation

of the OFDM system in the presence of phase and amplitude noise of LOs. There-
fore, the degradation factor (DF) is calculated as follows:

S
(
)Without{φ , A}
DF = 10 log N
S (8.24)
( )With{φ , A}
N
−σ φ2 −σ φ2
= 10 log{(1 + (1 + σ A2 − e ).SNRWithout {φ , A} ) / e }

In (8.24), σ ϕ2 is the variance of random variables ϕTX and ϕ RX , and σ A2 is the va-
riance of random variables ATX and ARX . These are calculated by integrating the
power spectrum density of the total output phase and amplitude noise.

8.2.3 Impacts of Phase and Amplitude Noise on OFDM Systems

This section presents a performance analysis of an OFDM communication system

influenced by the phase and amplitude noise of the RF LO in an AWGN channel.
It is difficult to compensate for the phase and amplitude noise in the OFDM sys-
tem. Therefore, it is important to quantify the permissible levels of phase and am-
plitude noise in the LO to maintain system performance and to design the LO with
limited maximum phase and amplitude noise.
The BER formula for a system performance analysis can be found by the SNR
with phase and amplitude noise, instead of the SNR without phase and amplitude
noise. Equations (8.25) and (8.26) shows the probability of the bit error per carrier
and the probability of the bit error for QAM modulation, respectively.

2 1 3k E
Pbc = (1 − )Q ( . b) (8.25)
k M ( M − 1) N 0

where
1 x
Q( x ) = erfc( )
2 2

Pb = 1 − (1 − Pbc ) 2 = 2 Pbc − Pbc 2 (8.26)

190 Chapter 8 RF Impairments in OFDM Transceivers

where k is the number of bits per symbol, M is the number of symbols in the mod-
ulation constellation, Eb is the energy per bit, and N0 is the noise power spectral
density. Equations (8.27) to (8.30) are the fundamental BER formulas that can be
used in the system using the Gray-encoded 4,16,64,256QAM modulation method
and the coherent detection technique in the AWGN channel [23].

1 1
P4−QAM = erfc( SNRwith(ϕ , A) ) − erfc 2 ( SNRwith(ϕ , A) ) (8.27)
2 16

3 2 9 2
P16 −QAM = erfc( SNRwith (ϕ , A) ) − erfc 2 ( SNRwith (ϕ , A) ) (8.28)
8 5 256 5

7 4 49 4
P64 −QAM = erfc( SNRwith (ϕ , A) ) − erfc 2 ( SNRwith (ϕ , A) ) (8.29)
24 21 4096 21

15 4 225 4
P256 −QAM = erfc( SNRwith (ϕ , A) ) − erfc 2 ( SNRwith (ϕ , A) ) (8.30)
64 85 16384 85

where SNRwith{φ , A} is the SNR with phase and amplitude noise in the OFDM sys-
tem as in (8.23).
The phase and amplitude noise variances, σ φ2 , σ A2 , achieved in the previous
section can be used to analyze the BER performance in OFDM system with phase
and amplitude noise. Figures 8.4 to 8.7 show the theoretical and simulation BER
performances of an OFDM communication system employing 4,16,64,256QAM
modulations in the AWGN channel [38]. IER stands for irreducible error rate.
Detailed results for a BER of 10-6, every phase noise variance and different am-
plitude noise order relative to it are shown in Table 8.2. As shown in Figures 8.4
to 8.7, the relative SNR penalty (degradation) to attain a BER of 10-6 for different
amplitude noise orders, n, are presented in Table 8.3.
The results from Tables 8.2 and 8.3 shows that, if the amplitude noise order rela-
tive to phase noise, n, is higher than 15 dB (15-20 dB), we can ignore the amplitude
noise effect in the performance analysis. However, if it is less than 15 dB (10-15
dB), we must consider the amplitude noise for an exact analysis and determination
of the tolerable level of amplitude noise.
Figure 8.8 shows the combination of two effects (CPE, ICI) on a 4QAM-
OFDM signal after demodulation caused by phase and amplitude noise. The first
scatter plot (a) corresponds to the demodulated signal when the variance of the
phase noise that affects the OFDM signal is zero. The second scatter plot (b) cor-
responds to the demodulated signal when the variance of the phase noise is 0.0316
and the amplitude noise order relative to the phase noise is n = 100 dB (ideal). The
third scatter plot (c) corresponds to the demodulated signal when the variance of
8.2 Noise in OFDM Transcceivers 1991

Fig. 8.4 4QAM-OFDM systeem with ( σ ϕ = 0.0316 )

2

Fig. 8.5 16QAM-OFDM sysstem with amplitude and phase noise ( σ ϕ = 0.01 )
2
192 Chapter 8 RF Impairments in OFDM Transceiveers

Fig. 8.6 64QAM-OFDM sysstem with amplitude and phase noise( σ ϕ = 0.0032 )
2

ystem with amplitude and phase noise ( σ ϕ2 = 0.001 )

Fig. 8.7 256QAM-OFDM sy
8.2 Noise in OFDM Transceivers 193

Table 8.2 Phase and Amplitude Noise Degredation in 4,16,64,256 QAM-OFDM for
BER=10-6.

σ ϕ2 0.0316 0.01 0.0032 0.001

Pe Modulation AM-Noise Degradation (dB)

n=100
6.083 1.218 0.351 0.104
(ideal)
n=20 6.211 1.232 0.354 0.105
4QAM n=15 6.500 1.260 0.362 0.107
n=10 7.568 1.348 0.386 0.113
n=5 17.01 1.648 0.464 0.135
n=100 IER 3.619 0.872 0.254
n=20 IER 3.674 0.881 0.256
16QAM n=15 IER 3.796 0.901 0.262
n=10 IER 4.208 0.967 0.279
n=5 IER 5.849 1.179 0.334
n=100 IER IER 2.904 0.720
n=20 IER IER 2.945 0.727
64QAM n=15 IER IER 3.035 0.744
n=10 IER IER 3.334 0.797
n=5 IER IER 4.442 0.971
10-6 n=100 IER IER IER 2.599
n=20 IER IER IER 2.635
256QAM n=15 IER IER IER 2.713
n=10 IER IER IER 2.969
n=5 IER IER IER 3.897
194 Chapter 8 RF Impairments in OFDM Transceivers

the phase noise is 0.0316 and the amplitude noise order relative to the phase noise
is n = 10 dB. As shown in Figure 8.8, clouds in the constellation of (c) have the
higher spreading tendency and cause higher performance degradation.

Table 8.3 Relative Degradation for Amplitude Noise Order in 4,16,64,256 QAM-OFDM
for BER=10-6 and Specified Phase Noise

σ ϕ2 0.0316 0.01 0.0032 0.001

Modulation 4QAM 16QAM 64QAM 256QAM
Pe AM-Noise Relative Degradation (dB)
n=100
0 0 0 0
(ideal)
10-6 n=20 0.1272 0.0554 0.0412 0.0356
n=15 0.4168 0.1778 0.1314 0.1136
n=10 1.4854 0.5892 0.4300 0.3700
n=5 10.9285 2.2301 1.5376 1.2977

2 2
1
1.5 1.5

1 1
0.5
0.5 0.5
Quadrature

Quadrature
Quadrature

0 0
0

-0.5 -0.5

-0.5 -1 -1

-1.5 -1.5
-1
-2
-2

-2 -1 0 1 2 -2 -1 0 1 2
-1 -0.5 0 0.5 1 In-Phase In-Phase
In-Phase

(a) (b) (c)

SNR=20dB, SNR=20dB, SNR=20dB,

σϕ = 0 ,
2
σ ϕ = 0.0316 ,
2 σ ϕ2 = 0.0316 ,
n=100 dB n=100 dB n=10dB
(ideal) (ideal)

Fig. 8.8 Constellation point of 4QAM signal

8.3 Nonlinearity in OFDM Transceivers

According to the theory of smoothly nonlinear systems, we can represent a com-
ponent’s transfer function by a truncated power series around the DC operating
point. The power series law can be used to describe the transfer characteristics of
any two-port component, as long as a frequency independent model can represent
it [24]. All the internal nonlinearities, such as parasitic capacitances and
8.3 Nonlinearity in OFDM Transceivers 195

inductances, are assumed to have insignificant frequency dependence within the

frequency band of operation. Moreover, it is assumed that the circuit is smoothly
nonlinear and the circuit output, Vout (t ) = f NL [Vin (t )] , can be approximated by the
first five terms of its Taylor series expansion:
5
f [ x(t )] ≅ ∑ an x n = a0 + a1 x + a2 x 2 + a3 x 3 + a5 x 5 (8.31)
n =0

The above approximation is realistic for a wide range of RF devices, such as am-
plifiers and mixers [25]-[30]. It is important to note that even-order distortion (n
even) is mapped to bands that are far from the original OFDM signal passband. It is
assumed that these components are filtered out and do not influence the BER.
Therefore, only odd-order distortion is of importance. Without loss of generality,
we may neglect the DC term, a0 , and express the output of the nonlinear circuit as:

vout (t ) = a1vin (t ) + a3vin3 (t ) + a5vin5 (t ) (8.32)

where a1 , a3 and a5 are the linear gain, third-order and fifth-order nonlinearity
coefficients, respectively, while vin is the input OFDM voltage.
In the case of OFDM and other multi-tone signals, it can be inferred from the
above description that a large number of intermodulation distortion (IMD) prod-
ucts are generated. We calculate the number of IMD products using the generating
function technique [17]-[19]. As discussed in Chapter 4, in order to assess nonli-
near characteristics, circuit designers make use of either the 1-dB compression
point or the input/output inferred third-order intercept points (IIP3/OIP3) [31].
Three coefficients, a1 , a3 and a5 , can be derived from the electrical amplifier
parameters, such as the gain of amplifier ( G0 ), the third-order output intercept
point ( OIP 3 ) and the 1dB compression point ( P1dB ), as below [32]:
G0
a1 = 10 20 (8.33)

3G OIP3
2 0
−
a3 = − 10 20 10 (8.34)
3

8a1 (1 − 100.05 ) − 6a3α100.05

a5 = (8.35)
5α 2 100.05
P1dB +1− G 0
where α = 2 × 10 10
196 Chapter 8 RF Impairments in OFDM Transceiveers

Fig. 8.9 Illustration of 1-dB compression point and IIP3/OIP3

8.3.1 Analysis of No
onlinear Circuit Impact in OFDM
If we consider N modullated data symbols from a particular signaling constella-
tion, X k = ( X 0 , X1 ,..., X N −1 ) , over a time interval [0, T ] , the OFDM symbol caan
be written as:
N −1 2π kt
x(t ) = ∑ X k e
j
T 0≤t <T (8.366)
k =0

where j = −1 , X k is th he data symbol for the kth subcarrier, N is the number oof
subcarriers, and T is the OFDM
O symbol time.
The subcarrier spacing g of 1/T makes them orthogonal over each symbol period.
The output voltage can bee expressed as in (8.31), obtained by passing a low-pass
equivalent OFDM signall through nonlinear circuit and substitution of (8.36) intto
(8.31) [32].
N −1 2π 2π
3a3 j [( k1 + k2 − k3 )
f [ x(t )] = vin (t ) = a1 ∑ X k e ∑
j[ k t] t]
T + X k1 X k2 X k3 e T

k =0 4 k1 , k2 , k3
(8.377)
2π
5a j[( k1 + k2 + k3 − k4 − k5 )
∑
t]
+ 5 X k1 X k2 X k3 X k4 X k5 e T
8 k1 , k2 , k3 , k4 , k5

The received signal is ex

xpressed as (8.38), where n(t) is the complex Gaussiaan
noise:

vout (t ) = vin (t ) + a1n(t ) (8.388)

8.3 Nonlinearity in OFDM Transceivers 197

The sampled signal for the kth subcarrier after the FFT processing stage in the re-
ceiver can be written as:

N −1 2π N −1 N −1 2π 2π
1 −j 1 m] − j
∑ vout (m).e ∑ a1 ∑ X l e
km j[l km
Yk = N = N e N
N m=0 N m =0 l =0
N −1 2π 2π
1 3a3 j [( k1 + k2 − k3 ) m] − j
∑ ∑
km
+ X k1 X k2 X k3 e N e N
N m=0 4 k1 , k2 , k3

N −1 2π 2π
1 5a5 j[( k1 + k2 + k3 − k4 − k5 ) m] − j
∑ ∑
km
+ X k1 X k2 X k3 X k4 X k5 e N e N (8.39)
N m=0 8 k1 , k2 , k3 , k4 , k5

N −1 2π
a1 −j
∑ n(m).e
km
+ N
N m=0

or:

N −1 N −1
Yk = ∑ X l Ql − k + ∑ X k1 X k2 X k3 Qk′1 + k
2
− k3 − k
l =0 k1 , k2 , k3
(8.40)
N −1
+ ∑ X k1 X k2 X k3 X k4 X k5 Qk′′1 + k
2
+ k3 − k4 − k5 − k + Nk
k1 , k2 , k3 , k4 , k5

where N k is a sampled FFT version of the AWGN noise, Qk , Qk′ and Qk′′ are
given by :

N −1 2π
a1
∑e
j km
Qk = N (8.41-a)
N m=0

N −1 2π
3a3
∑e
j km
Qk′ = N (8.41-b)
4N m=0

N −1 2π
5a5
∑e
j km
Qk′′ = N (8.41-c)
8N m=0

The received signal is composed of four contributions: Yk1 is the kth desired sub-
carrier, Yk2 is the interference component caused by third-order intermodulation
(IM) of the subcarriers, Yk3 is the interference component caused by fifth-order
IM of the subcarriers, and N k is the AWGN:
Yk = Yk1 + Yk2 + Yk3 + N k (8.42)
198 Chapter 8 RF Impairments in OFDM Transceivers

or:

Yk = X k Q0 + U 3 ( N , k ). X k1 X k2 X k3 k1 + k2 − k3 = k . Q0′

+ U 5 ( N , k ). X k1 X k2 X k3 X k4 X k5 k1 + k2 + k3 − k4 − k5 = k . Q0′′
N −1
+ ∑ X k1 X k2 X k3 Qk′1 + k2 − k3 − k (8.43)
k1 , k2 , k3
, k1 + k2 − k3 ≠ k
N −1
+ ∑ X k1 X k2 X k3 X k4 X k5 .Qk′′1 + k2 + k3 − k4 − k5 − k
k1 , k2 , k3
, k1 + k2 + k3 − k4 − k5 ≠ k

+ Nk

where U 3 ( N , k ) and U 5 ( N , k ) are the number of integer solutions of the equations:

k1 + k2 − k3 = k & k1 + k2 + k3 − k4 − k5 = k (8.44)

with condition:

0 ≤ k1 , k2 , k3 , k4 , k5 ≤ N − 1 , k ∈ {0,..., N − 1} .

The number | U 3 | and | U 5 | are derived as below [30].

3
U 3 ( N , b) = (2bN − 2b 2 + 2b + N 2 − N ) (8.45)
2

5
U 5 ( N , b) = [6(2b − b 2 − 2b3 + b 4 ) + 6( −1 + b + 3b 2 − 2b3 ) N
12 (8.46)
+ (1 + 6b − 6b 2 ) N 2 + 6(2b − 1) N 3 + 11N 4 ]

The desired signal component can be found as:

3a3
Ps − desired = [ E[ a1 X k ]]2 + [ E[ U 3 ( N , k ) X k1 X k2 X k3 ]]2
4 (8.47)
5a5
+ [ E[ U 5 ( N , k ) X k1 X k2 X k3 X k4 X k5 ]]2
8

Therefore:

U 32 ( N , k ) × 9 × a32 × Ps 3 U 52 ( N , k ) × 25 × a52 × Ps 5
Ps − desired = a12 Ps + + (8.48)
16 64
8.3 Nonlinearity in OFDM Transceivers 199

If we consider a normalized OFDM signal as in (8.49), the amount of Ps can be

calculated as in (8.50):

2
N −1 2π kt
+φTX ( t )]
∑ Xke
j[ 2
POFDM − Signal = E T = N .E X k = N .Ps = 1 (8.49)
k =0
φTX ( t ) = 0

1
Ps = (8.50)
N

Therefore:

U 32 ( N , k ) × 9 × a32 × Ps U 52 ( N , k ) × 25 × a5 2 × Ps
Ps − desired = a12 Ps + + (8.51)
16 N 2 64 N 4

The power of interference component caused by third- and fifth-order IM of the

subcarriers, PIM 3 and PIM 5 can be calculated respectively as :

N −1
9 × a32
PIM 3 = [ E[ ∑ X k1 X k2 X k3 Qk′1 + k2 − k3 − k ]]2 =
16 N 2
.Ps3 .( ∑ U 3 ( N , b))2 (8.52)
k1 , k2 , k3 b≠k
, k1 + k2 − k3 ≠ k

Therefore:

9 × a32
PIM 3 = .Ps .( ∑ U 3 ( N , b))2 (8.53)
16 N 4 b≠k

and

N −1
PIM 5 = [ E[ ∑X
k1 , k2 ,k3 ,k 4 ,k5
k1 X k2 X k3 X k4 X k5 Qk′1 + k2 +k3 −k4 −k5 −k ]]2
,k1 + k 2 + k3 −k 4 − k5 ≠ k

25 × a52 5
= .Ps .(∑U 5 ( N , b)) 2
64 N 2 b≠ k (8.54)

Therefore:

25 × a52
PIM 5 = 6
.Ps .(∑ U 5 ( N , b))2 (8.55)
64 N b≠ k
200 Chapter 8 RF Impairments in OFDM Transceivers

The power of noise, Pn1 , can be calculated as:

Pn1 = Pn With{ Nonlinearity} = Var[ Nk ] = a12 .Pn (8.56)

From (8.48), (8.53), (8.55) and (8.56), the SNR including the nonlinearity can be
arranged as:

S PS ,desired
( )With{ Nonlinearity} = (8.57)
N ( PIM 3 + PIM 5 ) + Pn1
S
( ) Without { Nonlinearity } (64a12 N 6 + 36a 3 N 4 × U 32 ( N , k ) + 25a5 N 2 × U 52 ( N , k ))
2 2
S N
( )With { Nonlinearity } = (8.58)
S
( ) Without { Nonlinearity } × (36a32 N 2 .(∑ U 3 ( N , b)) 2 + 25a 52 × (∑ U 5 ( N , b)) 2 ) + 64a12 N 6
N
N b≠k b≠k

The expected value of (8.58) for 1 < k < N has been used in simulation as the
SNR. Next, the above-mentioned powers are used to evaluate the performance de-
gradation of the OFDM system in the presence of the nonlinearity factor. There-
fore, the degradation factor (DF) is calculated as follows:

S
(
)Without{Nonlinearity}
DF = 10 log N
S (8.59)
( )With{Nonlinearity}
N

DF = 10 log
S
(36a32 N 2 .( ∑ U 3 ( N , b)) 2 + 25a52 × ( ∑ U 5 ( N , b)) 2 ) × ( )Without{ Nonlinearity} + 64a12 N 6
b≠k b≠k N (8.60)
{ }
64a12 N 6 + 36a32 × N 4 × U 32 ( N , k ) + 25a5 2 × N 2 × U 52 ( N , k )

■ Example 8.1: Impact of Nonlinearity in OFDM Systems

OFDM modulation has a high peak-to-average power ratio (PAPR), necessitating

high linearity active devices to be able to process the signal with low distortion.
For validation of the analytical results, we shall consider one linear high power
amplifier with nonlinear coefficients. The power amplifier product is WPS-
495922-02. It operates over a 4.9 GHz to 5.9 GHz band with 11 dB gain and 32
dBm output power at the 1dB compression point.
Thia product uses high linearity device technology, which allows us to address the
important 802.11 Wi-Fi and the emerging 802.16 WiMAX (Worldwide Interopera-
bility for Microwave Access) infrastructure markets [34]. The electrical parameters
and corresponding polynomial coefficients of the amplifier are shown in Tables 8.4
and 8.5 by considering the datasheet of the amplifier and the above equations.
8.3 Nonlinearity in OFDM Transceivers 201

Table 8.4 Electrical Parameters of Amplifier WPS-495922-02

Electrical WPS- Application

Parameter 495922-02
G0 11.0 dB 802.16
OIP3 (dBm) 47 dBm WiMAX
P1dB(dBm) 32 dBm 256 carriers
64 QAM

Table 8.5 Polynomial Coefficients of Amplifier WPS-495922-02

Polynomial WPS-495922-02
Coefficients
a1 3.55
a3 -5.94*10-4
a5 -8.39*10-6

Figure 8.10 shows the theoretical and simulation BER performances of an

OFDM communication system employing 64QAM modulation and 256 subcar-
riers with the electrical parameters of the WPS-495922-02 power amplifier. Figure
8.11 shows the theoretical and simulation performance degradation of an OFDM
communication system with the electrical parameters of the WPS-495922-02
power amplifier and different numbers of subcarriers. It is shown that increasing
the number of subcarriers causes higher performance degradation.

Fig. 8.10 64QAM-OFDM systems with specified electrical parameter and 256 subcarriers
202 Chapter 8 RF Impairments in OFDM Transceivers

Fig. 8.11 Performance degradation of OFDM system with specified electrical parameters
and different subcarriers

8.4 Concurrent Analysis of Nonlinearity and Phase Noise

in OFDM Transceivers
8.4 Concu rre nt Analysi s of No nlinearity and Phase Noise

This section investigates the concurrent circuit nonlinearity and phase noise ef-
fects on the OFDM signal. Their effects are investigated analytically, and the
closed form SNR expressions are derived. By using a truncated power series to
represent the nonlinear behavior of an RF circuit and a Lorentzian function to
model phase noise, the effects of nonlinearity and phase noise on OFDM signals
are examined. A closed form expression is obtained for the SNR degradation due
to third-order nonlinearity and phase noise effects.
Considering linear and cubic terms alone, OIP3 is defined as [35]:

2 a13
OIP3 = (8.61)
3 a3

Also, the OIP3 of N stages in cascade is as:

1 1 1 1 1
= + ... + + (8.62)
OIP3 (OIP3 )1 G2 ... GN (OIP3 ) N − 2 GN −1GN (OIP3 ) N −1 GN (OIPN ) N
8.4 Concurrent Analysis of Nonlinearity and Phase Noise 203

In the following studies, we consider a1 = 1 and the typical range of OIP3 to be 30

dBM to 50 dBm. On the other hand, According to (8.4), the phase noise variance is:
Bs
2 (8.63)
2K Bs 1
σϕ = 2 ∫ L( f ) 2 df = ) + L0 Bs
2
arctg (
0
π 2B π

where Bs is the OFDM bandwidth.

8.4.1 Theoretical Analysis

As shown, the low-pass equivalent OFDM signal corrupted by transmitter oscilla-
tor phase noise is expressed as:

N −1 2π kt
+φTX ( t )]
vin (t ) = ∑ X k e
j[
T 0≤t <T (8.64)
k =0

where X k is the data symbol for the kth subcarrier, N is the number of subcarriers,
φTX (t ) is the transmitter LO phase noise, and T is the OFDM symbol time. The
subcarrier spacing of 1/T makes them orthogonal over each symbol period.
The output voltage can be obtained by passing a low-pass equivalent OFDM
signal through a nonlinear circuit and the substitution of (8.64) into (8.65). The de-
rivation calculus can be found in [38].

s(t ) = a1vin (t ) + a3vin (t )3 (8.65)

N −1 2π 2π
t +φTX ( t )] 3a3 j[( k1 + k2 − k3 ) t +φTX ( t )]
s(t ) = a1 ∑ X k e ∑
j[ k
T + X k1 X k2 X k3 e T (8.66)
k =0 4 k1 , k2 , k3

The received signal is:

r (t ) = s(t ) + a1n(t ) (8.67)

where n(t) is the complex Gaussian noise. Moreover, the received signal is influ-
enced by the phase noise of RX LO ( φRX (t ) ) as:

vout (t ) = r (t )e jφRX (t ) = s (t )e jφRX (t ) + a1n(t )e jφRX (t )

(8.68)
204 Chapter 8 RF Impairments in OFDM Transceivers

N −1 2π 2π
t +φ ( t )] 3a3 j[( k1 + k2 − k3 ) t +φ ( t )]
vout (t ) = a1 ∑ X k e ∑
j[ k
T + X k1 X k2 X k3 e T + a1n(t )e jφRX (t ) (8.69)
k =0 4 k1 , k2 , k3

where φ (t ) = φTX (t ) + φRX (t ) .

It can be assumed that the phase noise of the local oscillators at the transmitter
and receiver are independent and identically distributed (i.i.d.) with a zero mean
and variance of σ φ2 ; hence, Var (φ (t )) = 4σ φ2 [38].
The sampled signal for the kth subcarrier after the FFT processing stage in the
receiver can be written as:

N −1 2π N −1 N −1 2π 2π
1 −j 1 m +φ ( m )] − j
∑ vout (m)e ∑ a1 ∑ X l e
km j[l km
Yk = N = N e N
N m =0 N m =0 l =0
N −1 2π 2π
1 3a3 j [( k1 + k2 − k3 ) m +φ ( m )] − j
∑ ∑
km
+ X k1 X k2 X k3 e N e N
N m =0 4 k1 , k2 , k3
(8.70)
N −1 2π
1 −j
∑ n(m)e jφ
km
+ RX (m)
.e N
N m =0
N −1 N −1
= ∑ X l Ql − k + ∑ X k1 X k2 X k3 Qk′1 + k − k3 − k + Nk
2
l =0 k1 , k2 , k3

where N k is a sampled FFT of the AWGN noise multiplied by the phase noise of
the receiver LO with a variance of σ n2 , φ (m) = φTX (m) + φRX (m) , and Qk and
Qk′ are given by:
N −1 2π
a1
∑ e jφ ( m) e
j km
Qk = N (8.71)
N m=0

N −1 2π
3a3
∑ e jφ (m) e
j km
Qk′ = N (8.72)
4N m =0

The received signal is composed of four contributions: Yk1 is the kth desired sub-
carrier, Yk2 is the interference component caused by inter-subcarrier interference,
Yk3 is the interference component caused by intermodulation of the subcarriers,
and N k is the AWGN, i.e.:

Yk = Yk1 + Yk2 + Yk3 + N k

8.4 Concurrent Analysis of Nonlinearity and Phase Noise 205

or:
N −1
Yk = X k Q0 + U 3 ( N , k ) X k1 X k2 X k3 k1 + k2 − k3 = k Q0′ + ∑ X l Ql − k
l = 0,l ≠ k
N −1 (8.73)
+ ∑ X k1 X k2 X k3 Qk′1 + k2 − k3 − k + N k
k1 , k2 , k3
, k1 + k2 − k3 ≠ k

where U 3 ( N , k ) is the number of integer solutions of the equation

k1 + k2 − k3 = k with condition 0 ≤ k1 , k2 , k3 ≤ N − 1 , k ∈ {0,..., N − 1} [38].
The expected value of X k Q0 + U3 ( N , k ) X k1 X k2 X k3 k1 + k2 − k3 = k Q0′ is the desired
signal component. Hence, the desired signal power can be obtained as:

Ps − desired = [ E[ X k Q0 ]]2 + [ E[U 3 ( N , k ) X k1 X k2 X k3 Q0′ ]]2 (8.74)

−2σ φ2 3a3 −2σ φ2

where E[Q0 ] = a1e and E[Q0′ ] = e . Accordingly:
4

−4σ φ2
−4σ 2 9U 2 ( N , k )a32 Ps3e
Ps − desired = a12 Ps e φ + 3 (8.75)
16

Using of a normalized OFDM signal, the amount of Ps can be calculated as:

2
N −1 2π kt
+φTX ( t )]
∑ Xke
j[ 2
POFDM − Signal = E T = NE X k = NPs = 1 (8.76)
k =0
φTX ( t ) = 0

Therefore:
−4σ φ2
−4σ 2 9U 32 ( N , k )a32 Ps e
Ps − desired = a12 Ps e φ + (8.77)
16 N 2

The variance of X k Q0 + U3 ( N , k ) X k1 X k2 X k3 k1 + k2 − k3 = k Q0′ is the phase noise power

caused by the common phase error (CPE) and PCPE is calculated as:

PCPE = Var[ X k Q0 ] + Var[U3 ( N , k ) X k1 X k2 X k3 Q0′ ] = PVar

s [Q0 ] + Ps3U32 ( N , k )Var[Q0′ ] (8.78)
206 Chapter 8 RF Impairments in OFDM Transceivers

a12 −4σ 2 9a 2 −4σ 2

where Var[Q0 ] = (1 − e φ ) and Var[Q0′ ] = 3 (1 − e φ ) . Hence, one may
N 16 N
obtain PCPE as:
a12 −4σ 2 9a32 −4σ 2
PCPE = Ps (1 − e φ ) + PU
s 3
2
( N , k ) 3
(1 − e φ ) (8.79)
N 16 N

N −1
On the other hand, in (8.78), ∑ X l Ql − k is the interference component caused
l = 0,l ≠ k

by the inter-subcarrier interference (ICI). So, PICI can be calculated as:

N −1
PICI = Ps {Var[ ∑ Ql ] − Var[Q0 ]} (8.80)
l =0

N −1 N −1
9a32
where Var[ ∑ Ql ] = a12 (1 − e ) and Var[ ∑ Ql′] =
−4σ φ2 −4σ 2
(1 − e φ ) . The derivation
l =0 l =0 16
N −1 N −1
calculus of Var[ ∑ Ql ] and Var[ ∑ Ql′] is described in detail [38],[38]. Accordingly:
l =0 l =0

N −1 −4σ 2
PICI = a12 Ps (1 − e φ ) (8.81)
N

N −1
Moreover, in (8.71), the term ∑ X k1 X k2 X k3 Qk′1 + k2 − k3 − k is the interference
k1 , k2 , k3
, k1 + k2 − k3 ≠ k

component caused by intermodulation of the subcarriers. Hence, PIM can be cal-

culated as:

N −1
9a32
∑ ( ∑ U 3 ( N , b))2 (8.82)
−4σ φ2
PIM = [ E[ X k1 X k2 X k3 Qk′1 + k2 − k3 − k ]]2 = 2
Ps3e
k1 , k2 , k3 16 N b≠k
, k1 + k2 − k3 ≠ k

9a32
( ∑ U 3 ( N , b))2
−4σ φ2
PIM = 4
Ps e (8.83)
16 N b≠k

As mentioned, N k is a sampled FFT of the AWGN noise multiplied by the phase

noise of the receiver LO. Therefore, for the sake of simple analysis, we consider
Var ( N k ) as that of the AWGN, since it has a zero mean, the variance is
8.4 Concurrent Analysis of Nonlinearity and Phase Noise 207

unchanged on the basis that the FFT process is linear, and the phase noise corrupt-
ing the AWGN works as the phase rotation factor.

Pn1 = Pn With{φ ,Nonlinearity} = Var[ Nk ] = a12 Pn (8.84)

From the above relations, the SNR, including the nonlinearity and phase noise,
can be arranged as:

S PS , desired
( )With{φ , Nonlinearity} = (8.85)
N ( PCPE + PICI + PIM ) + Pn1

S −4σ 2 −4σ 2
)Without{φ , Nonlinearity} (16 N 4 a12 e φ + 9a32 N 2U 32 ( N , k )e φ )
(
S
( )With{φ , Nonlinearity} = N (8.86)
S
A( )Without{φ , Nonlinearity} + 9a32 ( ∑ U 3 ( N , b)) 2 e φ + 16a12 N 4
N −4σ 2

N b≠k

−4σ φ2
where A  (1 − e )(16 N 3 a12 + 9a32 NU 32 ( N , k ) + 16 N 3a12 ( N − 1)) .
The expected value of (8.84) for 1 < k < N has been used in simulations as the
SNR. In addition, the above-mentioned powers are used to evaluate the perfor-
mance degradation of the OFDM system in the presence of nonlinearity and phase
noise factors. Therefore, the degradation factor (DF) is calculated as follows:

S
(
)Without{φ , Nonlinearity}
DF = 10 log N (8.87)
S
( )With{φ , Nonlinearity}
N

S
( A + 9a32 ( ∑ U 3 ( N , b)) 2 )e
−4σ φ2
( ) + 16a12 N 4
b≠ k N Without{φ , Nonlinearity}
DF = 10 log −4σ φ2 −4σ φ2
(8.88)
16 N 4 a12 e + 9a32 N 2U 32 ( N , k )e

If we ignore from nonlinearity and suppose that OIP3 = ∞ (a3 = 0), the SNR be-
comes equal to equation (8.87), which is consistent with the literature [38], [39]:

S −4σ 2
)Without {φ } e φ
(
S N
( )With{φ } = (8.89)
N S −4σ 2
1 + ( )Without {φ } (1 − e φ )
N

Figure 8.12 shows simulation results of (8.87) for the BER of a 4-16-64-
256QAM-OFDM system with different phase noise variances, which has been
compared with available results.
208 Chapter 8 RF Impairments in OFDM Transceivers

Fig. 8.12 QAM-OFDM systems with phase noise variances

8.4.2 Performance Analysis

The OFDM system performance is degraded by the nonlinear power amplifier and
noisy local oscillator. The simulation model to investigate this degradation is pre-
sented in Figure 8.13. In this simulation, the behavioral model for RF circuits has
been used. As mentioned, the phase noise power density function of the oscillator
is modeled with a Lorentzian function with specified parameters and the nonlinear
transfer function is modeled with a third-order memoryless polynomial model.

Fig. 8.13 OFDM simulation model

8.4 Concurrent Analysis of Nonlinearity and Phase Noise 209

It is generally difficult to compensate for the circuit nonlinearity and phase

noise degradation in the OFDM system. Therefore, it is important to quantify their
permissible levels in RF tuners, inorder to maintain the system performance. This
leads to design RF tuners with limited maximum nonlinearity and phase noise.
The BER formula for system performance with phase noise and nonlinearity can
be found based on SNR. By use of the Gray-encoded 4,16,64,256QAM modula-
tion method and the coherent detection technique in the AWGN channel, the BER
in QAM can be obtained as [38].

1 1
P4 −QAM = erfc ( SNRwith{φ , Nonlinearity} ) − erfc 2 ( SNRwith{φ , Nonlinearity} ) (8.90-a)
2 16

3 2 9 2
P16 −QAM = erfc ( SNRwith{φ , Nonlinearity} ) − erfc 2 ( SNRwith{φ , Nonlinearity} ) (8.90-b)
8 5 256 5

7 4 49 4
P64 −QAM = erfc( SNRwith{φ , Nonlinearity} ) − erfc 2 ( SNRwith{φ , Nonlinearity} ) (8.90-c)
24 21 4096 21

15 4 225 4
P256 −QAM = erfc( SNRwith{φ , Nonlinearity} ) − erfc 2 ( SNRwith{φ , Nonlinearity} ) (8.90-d)
64 85 16384 85

where SNRwith (ϕ , Nonlinearity ) is the SNR with phase noise and nonlinearity in an
OFDM system. This can be used to analyze the BER performance in the OFDM
system with nonlinearity and phase noise.
Figure 8.14 shows the BER performance of the OFDM system employing
4,16,64,256QAM modulations in the AWGN channel. For each modulation
scheme, we only considered the nonlinearity with OIP3 in the range of 37 dBm to
41dBm for 256 and 1024 subcarriers in an OFDM system. Figure 8.15 shows the
BER performance of the OFDM system employing 4,16,64,256QAM modulations
with different subcarriers and a specified OIP3; and, Figure 8.16 shows the BER
performance of the same system with different OIP3 and specified subcarriers.
Figure 8.17 shows the combination effect of phase noise and nonlinearity for
different phase noise variances and OIP3 on the BER performance employing
4QAM modulation with different subcarriers. In addition, Figure 8.18 shows the
constellation of a 16QAM-OFDM system with OIP3 = 40dBm and σ φ2 = 0.001
for 64,256,1024,2048 subcarriers in OFDM system.
Using these results, engineers are given a trade-off for choosing the best values
for the parameters of an OFDM transceiver system, such as the constellation size
of modulation, number of subcarriers, phase noise variance of an oscillator and
third-order intercept point of a power amplifier, to meet the acceptable BER in a
specified SNR.
For example, if RF devices with a specified phase noise and nonlinearity exist,
designers can choose a suitable scheme for modulation and an appropriate number
of subcarriers; or, if an OFDM system with a specified scheme of modulation and
210 Chapter 8 RF Impairments in OFDM Transceivers

Fig. 8.14 QAM-OFDM systems with different OIP3 and 256 subcarriers

Fig. 8.15 QAM-OFDM systems with different subcarriers and specified OIP3 (SNR = 14dB)

.
8.4 Concurrent Analysis of Nonlinearity and Phase Noise 211

Fig. 8.16 QAM-OFDM systems with different OIP3 and specified subcarriers (SNR =
14dB)

Fig. 8.17 QAM-OFDM system with different OIP3 and specified subcarriers and phase
noise variances (SNR = 10dB)
212 Chapter 8 RF Impairments in OFDM Transceivers

number of subcarriers exists, engineers can choose RF devices with acceptable

phase noise and nonlinearity to meet the acceptable BER in a specified SNR

8.4.3 Fifth-Order Nonlinear Circuit and Phase Noise Modeling

Without loss of generality, we may neglect the DC term, a0 , and express the out-
put of the fifth0order nonlinear circuit as:

vout (t ) = a1vin (t ) + a3vin3 (t ) + a5vin5 (t ) (8.91)

where a1 , a3 and a5 are the linear gain and third- and fifth-order nonlinearity
coefficients, respectively, while vin is the input OFDM voltage. The coefficients
are as determined in (8.33), (8.34), and (8.35).

Fig. 8.18 Constellation points of 16QAM signal for different numbers of subcarriers (SNR
= 14dB, OIP3 = 40dBm, )
8.4 Concurrent Analysis of Nonlinearity and Phase Noise 213

Here, the AM/AM effects on the performance of the OFDM system are studied.
The AM/PM effects are excluded because the AM/AM effects on the OFDM per-
formance have been found to be much more significant than those of AM/PM.
Similar to the third-order nonlinearity, the SNR with the nonlinearity and phase
noise for the fifth-order nonlinearity can be arranged as:

S PS , desired
( )With{φ , Nonlinearity} = (8.92)
N ( PCPE + PICI + PIM 3 + PIM 5 ) + Pn1

S
( )Without{ϕ , Nonlinearity}
S N
( )With{ϕ , Nonlinearity} = ×
N S
( )Without{ϕ , Nonlinearity}
N (8.93)
−4σ ϕ2
e (64a12 × N + 36a3 × N
6 2 4
× U 32 ( N , k ) + 25a52 ×N 2
× U 52 ( N , k ))
× N × ( ∑ U 3 ( N , b)) × ( ∑ U 5 ( N , b)) 2 )) + 64a12 N 6
−4σ ϕ2
(B + e × (36a32 2 2
+ 25a52
b≠ k b≠ k

where:

−4σ ϕ2
B  (1 − e ) × (64a12 × N 5 + 36a32 × N 3 × U 32 ( N , k ) + 25a52 × N × U 52 ( N , k ))

The expected value for 1 < k < N has been used in simulations as the SNR. In ad-
dition, the above-mentioned powers are used to evaluate the performance degrada-
tion of the OFDM system in the presence of nonlinearity and phase noise factors.
Therefore, the degradation factor (DF) is calculated as follows:

S
(
)Without{φ , Nonlinearity}
DF = 10 log N (8.94)
S
( )With{φ , Nonlinearity}
N

1
DF = 10 log{ ×
−4σ ϕ2
e
× (36a32 × N 2 × ( ∑ U 3 ( N , b)) 2 + 25a52 × ( ∑ U 5 ( N , b))2 ))
−4σ ϕ2
(B + e
b≠k b≠ k (8.95)
S
( )Without{ϕ , Nonlinearity} + 64a12 N 6
N }
(64a12 × N 6 + 36a32 × N 4 × U 32 ( N , k ) + 25a52 × N 2 × U 52 ( N , k ))

In this analysis, if we ignore the nonlinearity and suppose that a3 = a5 = 0, the SNR
becomes consistent with previous works [23], [26]:
214 Chapter 8 RF Impairments in OFDM Transceivers

S −4σ 2
( )Without {φ } .e φ
S N
( )With{φ } = (8.96)
N S −4σ 2
1 + ( )Without {φ } × (1 − e φ )
N

Figure 8.19 shows simulation results for the BER of a 4-16-64-256QAM-OFDM

system with different phase noise variances that have been compared with availa-
ble results.

Fig. 8.19 QAM-OFDM systems with phase noise variances

■ Example 8.2: Performance Analysis

We consider the parameters of one actual and applicable high power amplifier
with fifth-order nonlinear coefficients. This power amplifier product is WPS-
495922-02. It operates over a 4.9 to 5.9 GHz band with an 11 dB gain and a 32
dBm output power at the 1dB compression point. The electrical parameters, cor-
responding polynomial coefficients and application of amplifier are shown in
Table 8.6, by considering the datasheet of the amplifier and the above equations.
Figure 8.20 shows the combined effect of phase noise and nonlinearity with a
specified parameter and different output back-offs (OBOs) on the BER perfor-
mance of an OFDM communication system employing 4QAM with different sub-
carriers. Figure 8.21 shows the constellation of a 4QAM-OFDM system with an
8.4 Concurrent Analysis of Nonlinearity and Phase Noise 215

OBO = 10dB and σ φ2 = 0.001 for 64,256,1024,2048 subcarriers in an OFDM sys-

tem. The SNR is 14 dB. Figure 8.22 shows the BER performance of an OFDM
communication system employing 64QAM and 256 subcarriers with specified pa-
rameters for different OBOs.

Table 8.6 Electrical Parameters of WPS-495922-02

Polynomial
Electrical Parameter Application
Coefficients
G0 11.0 dB a1 3.55 802.16 Wi-
OIP3 (dBm) 47 dBm a3 -5.94*10-4 MAX
256 carriers
P1dB (dBm) 32 dBm a5 -3.89*10-6
64 QAM

Fig. 8.20 QAM-OFDM system with specified parameters and different OBOs and subcar-
riers (SNR = 10dB)

Figure 8.23 shows the performance degradation of an OFDM communication

system with specified parameters for different subcarriers and OBO = 10 dB. Us-
ing these results, one has a metric to select the suitable values of parameters in
OFDM transceivers. These parameters can be the constellation size, number of
216 Chapter 8 RF Impairments in OFDM Transceivers

subcarriers, phase noise variance of an oscillator, and the OBO of a power am-
plifier. These are used to meet the acceptable BER in a specified SNR. For in-
stance, if RF devices with specified phase noise and nonlinearity are assumed, one
can choose a suitable scheme of modulation and an appropriate number of subcar-
riers. On the other hand, if an OFDM system with a specified scheme of modula-
tion and number of subcarriers is considered, a device with acceptable phase noise
and nonlinearity must be selected.

Fig. 8.21 The constellations of 4QAM-OFDM system with different numbers of subcar-
riers, OBO = 10dB and σ φ2 = 0.001
8.4 Concurrent Analysis of Nonlinearity and Phase Noise 217

Fig. 8.22 64QAM-OFDM system with specified parameters and different OBOs

Fig. 8.23 Performance degradation of an OFDM system with specified parameters for dif-
ferent subcarriers and OBO = 10 dB
218 Chapter 8 RF Impairments in OFDM Transceivers

8.5 OFDM Signal with Phase Noise Passing through Dynamic

Nonlinear Circuits
In this section, the SNR degradation due to the effects of dynamic nonlinearity and
phase noise is investigated.

8.5.1 Dynamic Nonlinear Circuits

In reality, the output of a nonlinear dynamic circuit, such as a high power amplifi-
er, depends on the previous inputs as well as the current input. This phenomenon
is called memory effects or simply temporal dynamics. These memory effects are
due to thermal effects and long-time constants in DC bias circuits [30].
The Volterra series is a general nonlinear model with memory, but its practical
application is strongly restricted due to its complexity. A simpler and particular
case of the Volterra series is the memory polynomial model (MPM), which is less
complicated and more analytically tractable [40]-[45]. The MPM consists of sev-
eral delay taps and nonlinear static functions. This model is a truncation of the
general Volterra series, consisting of only the diagonal terms in the Volterra ker-
nels. The number of parameters is, therefore, significantly reduced compared to
the general Volterra series. The model is shown in Figure 8.24.

Fig. 8.24 Memory polynomial model [28]

The baseband MPM is now widely used to describe nonlinear effects in a pow-
er amplifier with memory effects. As can be seen in Figure 8.24, the general form
of a baseband MPM can be written as [40]:
Q K
VOut ( x) = ∑∑ akq .Vin ( x − q )2 k −1 (8.97)
q = 0 k =1
8.5 OFDM Signal with Phase Noise Passing through Dynamic Nonlinear Circuits 219

where a kq are complex memory polynomial coefficients; k is an integer number;

Vin ( x) and Vout ( x) are the discrete input and output complex envelope signals of
x th sample, respectively; q = 0,1,..., Q is the memory interval and equal to the
sampling interval, T ; and, Q and K are the maximum memory and polynomial
order, respectively. Equation (8.97) only contains odd-order terms, because the
signals obtained from the even-order terms are far from the carrier frequency. Eq-
uation (8.97) can also be rewritten in a compact form if Fkq ( x − q ) = Vin ( x − q ) :

Q K Q
Vout ( x) = ∑∑ akq . Fkq ( x − q) = ∑ Fq ( x − q ) = F0 + F1 + ... + FQ (8.98)
q = 0 k =1 q =0

where:
K
Fq = ∑ akq .Vin ( x − q )2 k −1 (8.99)
k =1

The least mean square error method can be used to find the complex polynomial
coefficients, a kq . The least mean square solution for (8.98) is
A = ( F ∗T .F )−1.F ∗T .Vout .
where:

Vout = [Vout (0),...,Vout ( X − 1)]T , A = [a10 ,..., a K 0 , a11 ,..., a K 1 ,..., a1Q ,..., a KQ ]T ,
F = [ F10 ,..., FK 0 , F11 ,..., FK 1 ,..., F1Q ,..., FKQ ] .

In a memoryless polynomial model, the coefficients can be calculated relative to

the electrical parameters as in (8.35), (8.36), and (8.37). Moreover, the phase noise
power density function of a noisy oscillator is approximated by a Lorentzian func-
tion with a uniform phase distribution. It is parameterized by its total integrated
phase noise, K, and the -3dB bandwidth, B.

8.5.2 Modeling and Analysis

When the low-pass equivalent OFDM signal corrupted by transmitter oscillator phase
noise passes through the third-order memory nonlinear circuit that is modeled as:
Q 2
s(t ) = ∑∑ akq vin (t − qts )2 k −1 (8.100)
q = 0 k =1
220 Chapter 8 RF Impairments in OFDM Transceivers

The output voltage, s (t ) , by using the third-order nonlinearity model can be ob-
tained as:
Q N −1 2π
( t − qts ) +φTX ( t − qts )]
s(t ) = ∑ a1q ∑ X k e
j[ k
T

q=0 k =0
Q 2π
(8.101)
3a3q j[( k1 + k2 − k3 ) ( t − qts ) +φTX ( t − qts )]
+∑ ∑ X k1 X k2 X k3 e T

q =0 4 k1 , k2 , k3

The approximated received signal is expressed as (8.101), where n(t) is the com-
plex Gaussian noise.
Q
r (t )  s(t ) + ∑ a1q . n(t ) (8.102)
q=0

Moreover, the received signal is influenced by the phase noise of the receiver
(RX) LO ( φRX (t ) ) as:

Q
vout (t ) = r (t )e jφRX (t ) = s (t )e jφRX (t ) + ∑ a1q . n(t )e jφRX (t ) (8.103)
q =0

or, alternatively:

Q N −1 2π 2π kqts
t− +φTX ( t − qt s ) +φRX ( t )]
vout (t ) = ∑ a1q ∑ X k e
j[ k
T T

q =0 k =0 (8.104)
Q 2π 2π kqt s Q
3a3q j[( k1 + k2 − k3 ) t− +φTX ( t − qt s ) +φRX (t )]
+∑ ∑ X k1 X k2 X k3 e T T + ∑ a1q . n(t )e jφRX (t )
q =0 4 k1 , k2 , k3 q=0

Similar to the previous calculation [39], the performance degradation of the

OFDM system is determined in the presence of dynamic nonlinearity and phase
noise factors; therefore, the degradation factor (DF) is calculated as follows:

S
(
)Without{φ , Nonlinearity}
DF = 10 log N (8.105)
S
( )With{φ , Nonlinearity}
N
8.5 OFDM Signal with Phase Noise Passing through Dynamic Nonlinear Circuits 221

Q
).∑ a12q + Ω
− (Q + 2) 2 σ φ2
( Ψ .(1 − e
q=0
DF = 10 log{ Q Q
}
(16 N × ( ∑ a1q ) + 9 N × ( ∑ a3q )
− (Q + 2) σ φ
2 2
e 4 2 2 2
× U 32 ( N , k ))
q=0 q=0
(8.106)
where :
Q Q
Ψ  16 N 3 ∑ a12q + 9 N .∑ a32q × U 32 ( N , k )) + 16 N 3 ( N − 1).(1 − e
− (Q + 2)2 σ φ2
)
q=0 q =0
Q Q
S
Ω  9 × ( ∑ a3q )2 .e ( ∑ U 3 ( N , b) ) 2 ) × ( )Without{φ , Nonlinearity} + 16 N 4 ∑ a12q
− (Q + 2)2 σ φ2

q=0 b≠k N q =0

In this analysis, if we ignore the memory and suppose that Q = 0 , the SNR is
consistent with the memoryless models previously extracted. Figure 8.25 shows
the simulation results of (8.106) for the BER of a 4QAM-OFDM system with dif-
ferent OIP3 and specified subcarriers and phase noise variances.
If we ignore the memory and nonlinearity simultaneously, the SNR is:

S −4σ 2
( )Without {φ } .e φ
S N
( )With{φ } = (8.107)
N S −4σ 2
1 + ( )Without {φ } × (1 − e φ )
N

Figure 8.26 shows the simulation result of (8.107) for the BER of a 4-16-64-
256QAM-OFDM system with different phase noise variances that has been com-
pared with available results. Figure 8.27 shows the theoretical and simulation BER
performances of an OFDM communication system with phase noise and memory
and without nonlinearity, employing 4-16-64-256QAM modulations. For each
modulation, we only considered σ ϕ2 = 0.001 and Q = 0,1,2,3, 4,5 .
Figure 8.28 shows the theoretical and simulation BER performances of an
OFDM communication system employing 4QAM modulation with different sub-
carriers and memory length and specified OIP3 and phase noise. Figure 8.29
shows the BER performance of the same system with different OIP3 and memory
length and specified subcarriers and phase noise.
222 Chapter 8 RF Impairments in OFDM Transceivers

Fig. 8.25 4QAM-OFDM system with different OIP3 and specified subcarriers and phase
noise variances (SNR=10dB)

Fig. 8.26 QAM-OFDM systems with phase noise variances

8.5 OFDM Signal with Phase Noise Passing through Dynamic Nonlinear Circuits 223

Fig. 8.27 QAM-OFDM systems with different memory lengths and specified phase noise
variances

Fig. 8.28 4QAM-OFDM system with different subcarriers and memory length and speci-
fied OIP3 and phase noise variances (SNR = 12dB)
224 Chapter 8 RF Impairments in OFDM Transceivers

Fig. 8.29 4QAM-OFDM system with different OIP3 and memory lengths and specified
subcarriers and phase noise variances (SNR = 12dB).

References
[1] Nasser, C.R., Natarajan, B., Wu, Z., Weigandt, D., Zekavat, S.A., Shattil, S.: Multi
Carrier Technologies for Wireless Communication. Kluwer Academic Publishers
(2002)
[2] Riaz Ahamed, S.S.: Performance Analysis of OFDM. Journal of Theoretical and Ap-
plied Information Technology (2008)
[3] Armada, A.G., Calvo, M.: Rapid Prototyping of a Test Modem for Terrestrial Broad-
casting of Digital Television. IEEE Transactions on Consumer Electronics 43(4),
1100–1109 (1997)
[4] Costa, E., Pupolin, S.: M-QAM-OFDM System Performance in the Presence of a
Nonlinear Amplifier and Phase Noise. IEEE Transactions on Communications 50(3)
(March 2002)
[5] Come, B., Ness, R., Donnay, S., Van der Perre, L.: Impact of Front-End Non-
Idealities on Bit Error Rate Performance of WLAN-OFDM Transceivers. In: IEEE
Radio and Wireless Conference, September 10-13, pp. 91–94 (2000)
[6] Lee, T.H.: The Design of CMOS Radio Frequency Integrated Circuits, 2nd edn.,
pp. 664–667. Cambridge University Press, New York (2004)
[7] Liu, P., Bar-Ness, Y., Zhu, J.: Effects of Phase Noise at Both Transmitter and Re-
ceiver on the Performance of OFDM Systems. In: IEEE 40th Annual Conference on
Information Sciences and Systems (March 2006)
[8] Wu, S., Bar-Ness, Y.: OFDM Systems in The Presence of Phase Noise: Conse-
quences and Solutions. IEEE Transactions on Communications 52(11) (November
2004)
References 225

[9] Xu, L., Chen, Y., Feng, Z.: Effects of RF Oscillator Phase Noise on Performance of
Communication System. In: 4th International Conference on Microwave and Milli-
meter Wave Technology, pp. 429–432 (August 2004)
[10] Ryu, H.G., Li, Y.S., Park, J.S.: Nonlinear Analysis of the Phase Noise in the OFDM
Communication System. IEEE Transactions on Consumer Electronics 50(1) (Febru-
ary 2004)
[11] Petrovic, D., Rave, W., Fettweis, G.: Performance Degradation of Coded-OFDM Due
to Phase Noise. In: 57th IEEE Vehicular Technology Conference, vol. 2(22-25),
pp. 1168–1172 (April 2003)
[12] Shentu, J., Panta, K., Armstrong, J.: Effects of Phase Noise on Performance of
OFDM Systems Using an ICI Cancellation Scheme. IEEE Transactions on Broadcast-
ing 49(2), 221–224 (2003)
[13] Ryu, H.G., Lee, Y.S.: Phase Noise Analysis of the OFDM Communication System
By the Standard Frequency Deviation. IEEE Transactions on Consumer Electron-
ics 49(1), 41–47 (2003)
[14] Wu, S., Bar-Ness, Y.: Performance Analysis on Effect of Phase Noise in OFDM Sys-
tems. In: IEEE 7th International Symposium On Spread-Spectrum Techniques and
Applications, Prague (September 2002)
[15] Armada, A.G.: Understanding The Effects of Phase Noise in OFDM. IEEE Transac-
tions on Broadcasting 47(2), 153–159 (2001)
[16] Sathananthan, K., Tellambura, C.: Performance Analysis of an OFDM System With
Carrier Frequency Offset and Phase Noise. In: 54th IEEE Vehicular Technical Confe-
rence, pp. 2329–2332 (2001)
[17] Tomba, L.: On The Effect of Wiener Phase Noise in OFDM Systems. IEEE Transac-
tions on Communications 46(5), 580–583 (1998)
[18] Muschallik, C., Multimedia, T.: Influence of RF Oscillators On An OFDM Signal.
IEEE Transactions on Consumer Electronics 41(3) (August 1995)
[19] Pollet, T., Blade, M.V., Moeneclaey, M.: BER Sensitivity of OFDM Systems to Car-
rier Frequency Offset and Wiener Phase Noise. IEEE Transactions on Communica-
tions 43(2/3/4) (February, March, April 1995)
[20] http://en.wikipedia.org/wiki/OFDM
[21] Schiek, B., Rolfes, I., Siweris, H.J.: Noise in High-Frequency Circuits and Oscilla-
tors. John Wiley & Sons (2006)
[22] http://www.zen118213.zen.co.uk/Systems_And_Devices_Files/
PhaseNoise.pdf
[23] Sampei, S.: Applications of Digital Wireless Technologies to Global Wireless Com-
munications. Prentice-Hall (1997)
[24] Deutsch, R.: Non-linear Transformations of Random Processes. Prentice-Hall Inter-
national (1962)
[25] Cripps, S.C.: RF Power Amplifiers for Wireless Communications. Artech House,
Norwood (1999)
[26] Chien, C.: Digital Radio on a Chip. Kluwer Academic Publishers (2001)
[27] Fager, C., et al.: A comprehensive analysis of IMD behavior in RF CMOS power
amplifiers. IEEE Journal of Solid-State Circuits 39(1), 24–34 (2004)
[28] Bradley, J.: Applied Combinatorics With Problem Solving. Addison-Wesley, Read-
ing (1990)
[29] Madani, M.H., Abdipour, A., Mohammadi, A.: Analytical Performance Evaluation of
the OFDM Systems Passing Through Nonlinear Circuits. IEICE Electronics Ex-
press 7(3) (2010)
226 Chapter 8 RF Impairments in OFDM Transceivers

[30] Boulejfen, N., Harguem, A., Ghannouchi, F.M.: New Closed-Form Expressions for
the Prediction of Multitone Inter-Modulation Distortion in Fifth-Order Nonlinear RF
Circuits/Systems. IEEE Transactions on Microwave Theory and Techniques 52(1)
(January 2004)
[31] Gregorio, F.H.: Analysis and Compensation of Nonlinear Power Amplifier Effects in
Multi Antenna OFDM Systems, PhD dissertation, Helsinki University of Technology,
Electrical and Communications Engineering Department (2007)
[32] Hieu, N.T., Jianxue, J., Ryu, H.G.: Polynomial Modeling for Efficient Predistortion
of HPA. Journal of the Research Institute for Computer and Information Communica-
tion 12(3) (December 2004)
[33] Sakran, H.Y., Shokair, M., Elazm, A.A.: An Efficient Technique for Reducing PAPR
of OFDM Sysyem in the Presence of Nonlinear High Power Amplifier. Progress in
Electromagnetics Research C 2, 233–241 (2008)
[34] http://www.mwtinc.com/cat/wireless/pdf/WPS-495922-02.pdf
[35] Rodhe, U.: RF/Microwave Circuit Design for Wireless Applications. Wiley Inters-
cience (2000)
[36] Lee, T.H.: The Design of CMOS Radio Frequency Integrated Circuits, 2nd edn.
Cambridge University Press (2004)
[37] Schiek, B., Rolfes, I., Siweris, H.J.: Noise in High-Frequency Circuits and Oscilla-
tors. John Wiley & Sons (2006)
[38] Madani, M.H., Abdipour, A., Mohammadi, A.: Analysis of performance degradation
due to non-linearity and phase noise in orthogonal frequency division multiplexing
systems. IET Communications Journal 4(10), 1226–1237 (2010)
[39] Madani, M.H., Abdipour, A., Mohammadi, A.: Analytical Performance Evaluation of
the OFDM Systems in the Presence of Jointly Fifth Order Nonlinearity and Phase
Noise. Analog Integrated Circuits Journal (January 2011)
[40] Salman Ahmed, A.M.: Analysis, Modeling and linearization of Nonlinearity and
memory effects in power amplifiers used for microwave and mobile communications,
PhD dissertation, Kassel University (2005)
[41] Sakran, H.Y., Shokair, M., Elazm, A.A.: An Efficient Technique for Reducing PAPR
of OFDM Sysyem in the Presence of Nonlinear High Power Amplifier. Progress in
Electromagnetics Research C 2, 233–241 (2008)
[42] Ding, L., Zhou, G.T., Morgan, D.R., Ma, Z., Kenney, J.S., Kim, J., Giardina, C.R.: A
robust digital baseband predistorter constructed using memory polynomials. IEEE
Transactions on Wireless Communications 52(1), 159–165 (2004)
[43] Kim, J., Konstantinou, K.: Digital predistortion of wideband signals based on power
amplifier model with memory. Electronic Letters 37(23), 1417–1418 (2001)
[44] Ku, H., Kenney, J.S.: Behavioral modeling of nonlinear RF power amplifiers consi-
dering memory effects. IEEE Transactions on Microwave Theory and Tech-
niques 51(12), 2495–2504 (2003)
[45] Bohara, V.A., Ting, S.H.: Analysis of OFDM signals in nonlinear high power am-
plifier with memory. In: IEEE International Conference On Communications, Bei-
jing, China, pp. 3653–3657 (May 2008)
Chapter 9
RF Impairments in MIMO Transceivers

Direct conversion transceivers are very attractive in wireless communications, due

to their simplicity and ease of integration. However, their performances are de-
graded with radio frequency (RF) impairments. On the other hand, multiple input
multiple output (MIMO) wireless communication systems promise higher spectral
efficiency than what is possible with single input single output (SISO) wireless
communications [1], [2]. However, the significant improvement in wireless sys-
tem performance using MIMO is achieved by increasing the system complexity,
which may result in higher sensitivity and lower performance in practical systems.
Therefore, the degradations due to system impairments must be studied to evaluate
the realistic behavior of the MIMO systems. Moreover, compensation techniques
can be applied by understanding the effects of these impairment factors. These
impairments may be noticed as phase noise, DC offset, in-phase / quadrature (I/Q)
imbalance, and power amplifier nonlinearity [3]. In this chapter, the impairments
due to phase noise, DC offset and I/Q imbalance are analyzed and the different
techniques for their prevention and compensation are discussed. The power am-
plifier nonlinearity effects are investigated in the following chapter.

9.1 Phase Noise in MIMO Transceivers

Among various RF impairment factors, phase noise is one of the major sources of
performance reduction in wireless communications systems. This section is main-
ly focused on the characterization and evaluation of phase noise effects on MIMO
systems. By modeling the performance degradation due to the phase noise, a clear
understanding of the system and compensation techniques are provided. On the
other hand, by providing analytical relations to model the phase noise effects, the
performance can be evaluated without the requirement of conducting very time-
consuming simulation studies.

9.1.1 Phase Noise Model

Two types of phase noise models exist. When the system is phase locked, the result-
ing phase noise is low; and, it is modeled as a zero-mean, stationary, finite-power

A. Mohammadi et al.: RF Transceiver Design for MIMO Wireless Communications, LNEE 145, pp. 227–263.
springerlink.com © Springer-Verlag Berlin Heidelberg 2012
228 Chapter 9 RF Impairments in MIMO Transceivers

random process [4]. On the other hand, when the system is only frequency locked,
the resulting phase noise slowly varies, but is not limited; and, it is modeled as a ze-
ro-mean, non-stationary, infinite-power Wiener process [5]. We assume the system
employs a phase-locked loop (PLL) for its local oscillators (LO); hence, we use the
first model. For simplicity, it is assumed for simplicity that the phase noise ( θ n ), is a
zero-mean Gaussian distributed random variable [6], [7]:
−θ n2
1 2σ θ2n
f Θn (θ n ) = e (9.1)
2π σ θn

where σ θ2n is the phase noise power.

In the first step, we extract the bit error rate (BER) of a SISO system with phase
noise. If a SISO system has phase noise in the receiver, the received signal is ex-
pressed as:
r = hak e jθn + w (9.2)

where r is the received signal, h denotes the channel gain, ak is the transmitted
symbol, θ n is the random phase caused by the phase noise, and w is the additive
white Gaussian noise (AWGN), which is a statistically independent and identical-
ly distributed (i.i.d.) complex-valued Gaussian random variable with a variance of
N0. It is assumed that h is a complex random variable.
The BER relation for MQAM (M-ary quadrature amplitude modulation) using
the signal-space concept is extracted as [8]:
M /2 ⎛
4 1 3Eb log 2 M ⎞
Pe, MQAM ≅ (1 − ) ∑ Q ⎜ (2i − 1) ⎟ (9.3)
log 2 M M i =1 ⎜⎝ ( M − 1) N o ⎟⎠

It is shown that, using only the first term in (9.3), the BER can be closely approx-
imated as [8]:

4 1 ⎛ 3Eb log 2 M ⎞
Pe, MQAM ≅ (1 − )Q ⎜ ⎟⎟ (9.4)
log 2 M ⎜ ( M − 1) N o
M ⎝ ⎠
It is a straightforward task to extract the BER of MQAM due to a sample function
of θ n using same procedure as in [8], [9]:

M
−1
4( M − 1) 2 ⎛ 3γ s ⎞
p pn ( M , θ n ) = ∑ Q ⎜ (1 − (2i + 1) sin θ n ) ⎟ (9.5)
M log 2 ( M ) ⎜ M − 1 ⎟⎠
i=
− M ⎝
2

where γs denotes the received signal-to-noise ratio (SNR)

(i.e. γ s = received signal power / σ 2
w) and M is the constellation size. Hence, the
9.1 Phase Noise in MIMO Transceivers 229

SER (θ n )
BER can be approximated as pb (θ n ) ≅ . This relation provides the same
log 2 ( M )
results as in [8], [9], if θ n = 0 . The average BER is obtained by taking the ex-
pected value of (9.5) with respect to the probability density function (pdf) of θ n ;
therefore:
∞
Pe ( M ) = ∫ p pn ( M , θ n ) f Θ n (θ n ) dθn . (9.6)
−∞

SISO Case Study

In a SISO scenario, one transmitter and one receiver are connected in an AWGN
channel. The analytic, simulation and measurement results of the SISO system
over the AWGN channel using a very low phase noise oscillator are presented in
Figure 9.1. The measurement setup is discussed in the appendix. As can be seen,
the simulation agrees very well with the analytic results. However, there is a gap
of about 1 dB between the analytical and measurement results. This is due to im-
plementation loss and the phase noiseless assumption of the signal generators that
operate as LOs. Nevertheless, there is good agreement between analytical and
measurement results. Figure 9.2 shows the results of the SISO system over the
AWGN channel with phase noise. As shown, the measurement and analytical re-
sults are well matched.

0
10
SISO AWGN BER
AWGN
Measurement

Measurement
-1 fitted curve
10

256QAM
-2
10
BER

fine AWGN
BER relation 4QAM
-3
10
16QAM
AWGN 64QAM
Simulation

-4
10
0 5 10 15 20 25 30
SNR (dB)

Fig. 9.1 BER of AWGN SISO without phase noise

230 Chapter 9 RF Impairments in MIMO Transceivers

SISO BER (With PN)

-1 Measurement
10

256QAM
-2
10
BER

64QAM

-3
10
Analytic
Sigma2=0.004

16QAM
4QAM
-4
10

0 5 10 15 20 25 30
SNR (dB)

Fig. 9.2 BER of AWGN SISO with phase noise (PN)

9.1.2 Impact of Phase Noise on MIMO Systems

A flat-fading MIMO channel with NT transmitting antennas and NR receiving an-
tennas is considered. Assuming perfect synchronization, the input-output relation-
ship is given by:

y=Hx+w (9.7)

where y is an NR×1 vector of the received signal, H denotes the NR×NT channel
matrix, x is an NT×1 vector of transmitted symbols, and w is an NR×1 additive
{ }μ
N ,N
R T
white Gaussian noise vector. The channel coefficients hμυ are statistically
=1,υ =1
i.i.d. complex-valued Gaussian random variables with a variance of 1. Compo-
nents of the noise vector have the same distribution with a variance of N0.
If singular value decomposition (SVD) is applied to H , it can be expressed as

H=UDV H (9.8)

where (.) H denotes the conjugate transpose; D is an NR×NT matrix with singular

{ λ}
m Δ
values of H, and i and m = min( N t , N r ), are its main diagonal elements;
i =1
9.1 Phase Noise in MIMO Transceivers 231

U and V are NR×NR and NT×NT unitary matrices with left and right singular vectors
of H as their columns, respectively. Substituting (9.8) into (9.7) we have:

y ′ = Dx′ + w′ (9.9)

where:
Δ Δ Δ
y ′ = U H y , x ′ = V H x , w′ = U H w (9.10)

Since U and V are unitary matrices, the powers of x and x ′ are the same, as well
y and y′ , w and w ′ . The equivalent model of the system is depicted in Figure
9.3, which shows that the MIMO channel is converted into m parallel SISO sub-
channels by SVD.

w
x′ x y′
V H UH

w1′
λ1
x1′ y1′

λ2 w2′
x2′ y2′

wm′
λm
xm′ ym′

Fig. 9.3 MIMO system and equivalent model

Due to the randomness of the entries of H , the subchannel power gain ( λ ) is a

random process. The marginal pdf of the unordered eigenvalues is [10]

1 m −1 k !
∑ ⎡ Ldk ( λ1 ) ⎤ λ1d e − λ1
2
Pdf λ1 (λ1 ) = ⎣ ⎦
(9.11)
m k = 0 ( k + d )!
232 Chapter 9 RF Impairments in MIMO Transceivers

where Ldk ( λ1 ) is the associated Laguerre polynomial of order k , i.e.:

k
( k + d )!
Ldk ( λ1 ) = ∑ ( −1) λl
l
(9.12)
l =0 ( k − l )!( d + l )!l ! 1

w e j θn
x′ x y y′
V H UH

λ1 w1′ e j θn
x1′ y1′

λ2
w2′ e j θn
x2′ y2′

wm′
λm e j θn
xm′ ym′

Fig. 9.4 MIMO system and equivalent model with phase noise

The MIMO system model with phase noise is presented in Figure 9.4. The input-
output relationship is given by:

y=ΘHx+Θw (9.13)

where Θ is an NR× NR matrix of phase noise and is defined as follows:

⎛ e jθ1 ⎞
⎜ 0 ⎟
Θ=⎜ ⎟ (9.14)
⎜⎜ jθ1 ⎟
⎝ 0 e ⎟⎠ N × N
R R
9.1 Phase Noise in MIMO Transceivers 233

Equation (9.13) can be rewritten as:

y ′ = U H ΘHVx′ + UH Θw (9.15)

Since the Θ is diagonal, (9.13) can be written as (9.15); and, since the Θ is di-
agonal and its entries have unit norm and random phase, the entries of Θw′ are
complex Gaussian random variables with a zero mean and a variance of N0. It is
easy to see that the received signal can be again converted into m parallel SISO
subchannels by SVD. The difference from the previous situation is that each sub-
channel is affected by similar phase noise.

y ′ = ΘUH HVx′ + Θw′ = ΘDx′ + Θw′ (9.16)

MIMO Case Study

The results of the MIMO system with adaptive modulation using NT = NR = 4 are
presented. The Rayleigh fading channel is used in the channel simulator. Moreo-
ver, we used zero forcing receivers for MIMO detection. Figure 9.5 shows the re-
sults. As shown, there is good agreement between the analytic and measurement
results.

9.1.3 Adaptive Modulation MIMO System

In the adaptive modulation scheme, the channel is estimated in the receiver and
fed back to the transmitter. The transmitter adapts the transmitting signal, consi-
dering the feedback information, in order to maximize the spectral efficiency. We
assume that the feedback path does not introduce any errors and delay. The avail-
ability of channel information at the transmitter allows it to adapt its transmission
power and rate relative to the channel variation [9].
A variable-rate variable-power (VRVP) adaptation scheme is considered here.
According to the above decomposition, the adaptive modulation MIMO problem
can be considered as an adaptive modulation SISO problem using the unordered
eigenvalue distribution [12]. Hence, the following problem should be solved by
maximizing the average spectral efficiency (ASE) [13]:
maximize ASE = mε λ1 [ k1 (λ1 )]
P1 , k1

subject to ε λ1 [ P1 (λ1 ) ] = P (9.17)

m
BER1 (λ1 ) = BERtgt

where λ1 is the first unordered subchannel power gain, and P1 , k1 and BER1 are
the power, rate and instantaneous BER in the first unordered subchannel, respec-
tively. We use only square MQAMs; therefore, the available rates are:
234 Chapter 9 RF Impairments in MIMO Transceivers

MIMO BER (without PN)

MIMO Nt =4, Nr =4
Rayleigh Channel
Without PN
-1
10
BER

256QAM
Measurement

analytic result
sigma2 = 0
-2
10

16QAM
4QAM 64QAM

0 5 10 15 20 25 30 35 40
. SNR (dB)

(a)

MIMO BER (with PN)

-1
10
256QAM
BER

Measurement
-2
10
MIMO Nt =4, Nr =4
Rayleigh Channel
With PN
64QAM
analytic result 16QAM
sigma2=0.004
-3
10
4QAM

0 5 10 15 20 25 30 35 40 45
SNR (dB)

(b)

Fig. 9.5 Rayleigh MIMO: (a) without phase noise, (b) with phase noise
9.1 Phase Noise in MIMO Transceivers 235

T1 R1

det ection
adaptive T2 R2 and
mod ulation estimation

TNt RN r

feedback

Fig. 9.6 An adaptive modulation MIMO structure

k1 (λ1 ) ∈ {r0 , r1 ,… , rN }
(9.18)
ri = 2i i = 0,1,… , N
The solution to the problem in (9.17) has been given in [3]. Therefore, the optimal
power and rate policy is:

⎧ k1 (λ1 ) = ri , υi ≤ λ1 < υi +1
⎪
⎨ M i − 1 σ w2 (9.19)
P
⎪ 1 1( λ ) = , υi ≤ λ1 < υi +1
⎩ λ1 K

−1.5
where M i = 2ri , K = and υi denotes the power gain boundary and
ln(5 BERtgt )
equals to

M i − M i −1 μ
υi = ri − ri−1
(9.20)
K
where μ > 0 and is the Lagrangian multiplier that is determined from the average
power constraint equation:

N
M i − 1 ⎡ ∞ Pdf λ1 (λ1 ) ∞ Pdf λ (λ1 ) ⎤ P
∑ ⎢ ∫υ
λ1
d λ1 − ∫
λ1
1
d λ1 ⎥ = (9.21)
⎦ mσ w
K ⎣ i υi+1 2
i =1

After obtaining μ , the average spectral efficiency can be achieved from:

N
R = m∑ ri ⎡ ∫ Pdf λ1 (λ1 ) d λ1 − ∫ Pdf λ1 (λ1 ) d λ1 ⎤
∞ ∞
(9.22)
⎢
i =1 ⎣ i
υ υ i +1 ⎥⎦
236 Chapter 9 RF Impairments in MIMO Transceivers

9.1.4 BER of Adaptive Modulation MIMO System with Phase Noise

The average BER of the system can be expressed as:
N
∑ Pe (M j ) × log 2 (M j ) P(υ j ≤ λ1 < υ j +1 )
j =1
Average BER = N
(9.23)
∑ log 2 (M j ) P(υ j ≤ λ1 < υ j +1 )
j =1

where Pe ( M j ) is the BER corresponding to M j , and P (υ j ≤ λ1 < υ j +1 ) is the

probability that λ1 falls in the jth region, so that:

υ j +1
p(υ j ≤ λ1 < υ j +1 ) = ∫ Pdf λ1 (λ1 )d λ1 (9.24)
υj

From (9.17), the received signal power ( Prec ) due to adaptive modulation can be
expressed as:

−2 ln(5 BERtgt )σ w2
Prec = λ1 P1 (λ1 ) = ( M i − 1) (9.25)
3
From (9.5) and (9.25), for a sample function of θn , Pe ( M j ) can be written:

Mj
−1
4( M j − 1)
( )
2
p pn ( M j ,θ n ) = ∑ Q (1 − (2i + 1) sin θ n ) −2 ln(5 BERtgt ) (9.26)
M j log 2 ( M j ) − Mj
i=
2

The BER of the system for a constellation size of M j is obtained by taking the
expected value of (9.26) with respect to θn , so that:

−θ n2
∞ 1 2σ θ2n
Pe ( M j ) = ∫ p pn ( M j ,θ n )e dθn (9.27)
−∞ 2π σ θn

Therefore, the total BER of the adaptive modulation MIMO system under the im-
pact of phase noise can be expressed in a closed form as:

∑ ⎡⎢⎣ Pe (M j ) × log 2 (M j ) × ∫υ Pdf λ1 (λ1 )d λ1 ⎤⎥

N υ j +1

j =1 j ⎦
BERPN = (9.28)
∑ ⎡⎢⎣log 2 (M j ) × ∫υ j Pdfλ1 (λ1 )d λ1 ⎤⎥⎦
N υ j +1

j =1
9.1 Phase Noise in MIMO Transceivers 237

Case Study: Adaptive Modulation MIMO

The performance of the adaptive modulation MIMO system under the impact of
phase noise is presented in this section. The measurement results of the adaptive
modulation MIMO system are presented for a set of average SNRs, NT = NR = 4
and k1 (λ1 ) ∈ {0, 2, 4,6,8} . The Rayleigh fading channel is used in the channel si-
mulator. A BERtgt of 10-2 is selected for our measurements. Perfect channel state
information (P-CSI) is assumed at both the transmitter and receiver [14], [15].
However, the phase noise affects the channel estimation. The channel is estimated
in the receiver and fed back to the transmitter. To determine the training sequence
length, we use the available algorithm [16]. Due to P-CSI, the total average spec-
tral efficiency, which is only related to channel model and average SNR, is the
same as the case without phase noise. This is validated in Figure 9.7 using simula-
tion and measurement results.

30

AM MIMO
25 (With PN)
ASE of MIMO adaptive modulation, R(bps/Hz)

20

Nt=4 Nr=4
15 BERtgt=0.01

10
Analytical result
Simulation result
5 Measurement with PN
Measurement without PN

0
0 5 10 15 20 25 30
SNR (dB)

Fig. 9.7 ASE of adaptive modulation MIMO system under impact of phase noise
−2 P
( BERtgt = 10 , SNR = )
σ w2

In Figure 9.8-a, the measurement results of an adaptive modulation MIMO sys-

tem in a Rayleigh channel without phase noise are depicted. As shown, the mea-
surement results are close to the analytical results.The measurement results of an
adaptive modulation MIMO system over a Rayleigh channel with phase noise are
shown in Figure 9.8-b. This figure shows an unusual behavior, where by increasing
the average SNR, the BER is more degraded. This is due to the use of higher con-
stellation sizes in larger SNRs by the adaptive modulator. The higher constellation
238 Chapter 9 RF Impairments in MIMO Transceivers

sizes are more sensitive to the phase noise. Hence, the total impact of higher con-
stellation sizes and greater sensitivity to phase noise results in BER degradation. As
can be seen in this figure, there is good agreement between the measurement and
analytical results.

AM MIMO BER (Without PN)

Measurement
-2 fitted curve
10

Measurement
BER

analytic result
sigma2=0
AM MIMO
BERtgt=0.01, k (λ )ε [0 2 4 6 8]
1 1
Nt=4, Nr=4
Rayleigh Channel
Without PN
-3
10
0 5 10 15 20 25 30 35
SNR (dB)
(a)
-1
10
AM MIMO BER (With PN)

Measurement
fitted curve

-2
10
Measurement
BER

analytic result
sigma2=0.004 AM MIMO
BERtgt=0.01, k (λ )ε [0 2 4 6 8]
1 1
Nt=4, Nr=4
Rayleigh Channel
With PN
-3
10
0 5 10 15 20 25 30 35
SNR (dB)
(b)

Fig. 9.8 BER of an adaptive MIMO system: (a) without phase noise, (b) with phase noise
9.2 DC Offset in MIMO Transceivers 239

The average spectral efficiency (ASE) of an adaptive modulation (AM) MIMO

system for six various antennas configurations are shown in Figure 9.9. The re-
sults are obtained with BERtgt = 10−3 and k1 (λ1 ) ∈ {0, 2, 4,6,8,10} ,and agree with
the available literature [12].
The average BER rate versus the variance of phase noise, σ θ2n , for different
SNR values are shown in Figure 9.10. The average BERs are presented for a
BERtgt equal to .01 and .001. As expected, by increasing the phase noise power,
the BER is more degraded.

40

35 (16,4)
ASE of MIMO adaptive modulation, R(bps/Hz)

analytic result
30 simulation result (8,4)

25

20 (8,2)
(4,4)
15 (4,2)

10
(2,2)
5

0
0 5 10 15 20 25 30
SNR (dB)

, k1 (λ1 ) ∈ {0, 2, 4,6,8,10} ,

−3
Fig. 9.9 ASE of an AM MIMO system ( BERtgt = 10
P
SNR = )
σ n2

9.2 DC Offset in MIMO Transceivers

The DC offset is the other impairment that reduces the performance of MIMO
transceivers. In this section, the DC offset impairment is investigated. The com-
plete removal of the DC offset in the receiver can be very difficult; therefore, the
baseband compensations are usually required [19].
240 Chapter 9 RF Impairments in MIMO Transceivers

-1
10 AM MIMO (With PN)

Nt=4 Nr=4 SNR=20dB

Nt=2 Nr=2 SNR=20dB
Nt=4 Nr=4 SNR=10dB
Nt=2 Nr=2 SNR=10dB

-2
10
BER

BERtgt=0.01

10
-3 BERtgt=0.001

-5 -4 -3 -2 -1
10 10 10 σ2 (PN) 10 10

2
Fig. 9.10 Average BER of an AM MIMO system versus σθ (analytic results)
n

9.2.1 DC Offset
The DC offset sources can be broken down into static DC error and dynamic DC
[20]. Static DC errors are generally caused by self-mixing of an LO within the re-
ceiver itself. On the other hand, dynamic DC errors are caused by time-varying ef-
fects within the receiver environment. Some examples of the dynamic DC offset
errors are [21], [22]:

• Reflections of the receiver LO, which is radiated from the receiver anten-
na and are picked up by receiver and down-converted to DC.
• Rapid changes in signal strength (such as those caused by fading), which
are not tracked quickly enough by the receiver automatic gain control
(AGC). The receiver is overloaded for a short period of time, and second-
order nonlinearity causes the DC offset.

DC offset degrades the BER of the receiver. It may also saturate the baseband ana-
log-to-digital converters (ADCs), which dramatically reduces their dynamic range.
Therefore, the DC offset must be removed by means of a calibration method.
9.2 DC Offset in MIMO Transceivers 241

Sometimes a capacitive coupling can be used, although it removes some of the

wanted signal energy. When the signal has significant energy at or close to DC,
capacitive coupling is not an option; however, it is possible to perform DC calibra-
tion, which is done by injecting an appropriate DC level to cancel the DC offset.
The measurement process is typically performed in the digital domain by long-
term averaging. The calculated DC offset is subtracted from the received signal
(typically in the analog domain). DC calibration has a disadvantage in that it is not
capable of completely compensating for dynamic DC offsets, whereas static DC
offsets can just be removed. In the remainder of this section, wherever DC offset
mentioned, it means the uncompensated part of the DC offset. Since the nature of
a dynamic DC offset is random and is best modeled by a zero-mean complex
Gaussian distributed random variable [23], we assume the same model for the un-
compensated part.

9.2.2 BER OF MQAM Modulation under Impact of DC Offset

In this section, a closed form expression for the MQAM BER under the impact of
DC offset is extracted. The relation between the input and output of a digital
communication system can be written as:

s' = s + c + w (9.29)

where s = sI + jsQ is the transmitted signal, c = cI + jcQ is the DC offset, s’ is the

received signal, and w is the AWGN.
It is assumed that static DC offsets have been removed entirely and that the re-
maining DC offset is modeled as a zero-mean, complex Gaussian distributed ran-
dom variable. It is considered that the channel varies very slowly, as well as the
DC offset; and, the channel is constant in the transmission interval of a frame.
(Since the sources of dynamic DC offset are related to the time-varying effects
within the receiver environment, the same flat model as the channel model is ac-
ceptable for DC offset.)
The constellation diagram of a square QAM signal is depicted in Figure 9.9.
This figure shows the s and s ′ points with Gray-encoded bit mapping. If we use
the method of signal-space concepts or the method of [25], the BER for a specific
realization of c is:

(1 − 1
)× M
⎧⎪ ⎛ (2i − 1)d + c ⎞
∑ ⎨Q ⎜⎜ σ w
2

⎟+
M
PbMQAM (d , c) ≅ I
log 2 M i =1 ⎪ ⎝
⎟
⎩ 2 ⎠
(9.30)
⎛ (2i − 1)d + cQ ⎞ ⎛ (2i − 1)d − c ⎞ ⎛ (2i − 1)d − cQ ⎞ ⎪⎫
Q⎜ ⎟ + Q⎜ I
⎟ + Q⎜ ⎟⎬
⎜ σw ⎟ ⎜ σw ⎟ ⎜ σw ⎟⎪
⎝ 2 ⎠ ⎝ 2 ⎠ ⎝ 2 ⎠⎭
242 Chapter 9 RF Impairments in MIMO Transceivers

s0 s4 s12
• • • s8′∗ •s8
∗s′ s12′∗
∗ s0′ 4

s1′
•∗
s s s
•5 d 2 • 13 • 9
s∗13′ ∗ s9′
s1
s5′ ∗

d 2
s15′
•
s3
• s7 ∗ ∗s11′
s3′∗ s15• s11•
s7′
∗

s2
∗s2′ • • s6 ∗s14′ •s14 s10′
∗•s10
s6′
∗

Fig. 9.9 The constellation diagram of a 16-QAM signal

where d is the Euclidean distance of two adjacent points. The Q function is very
low for a large argument, so we can approximate relation (9.30) to:

PbMQAM (d , c) ≅
(1 − 1
M ) × ⎧⎪Q ⎛⎜ d + c I
⎞
⎟+
⎨ ⎜ σw ⎟
log 2 M
⎩⎪ ⎝ 2 ⎠
(9.31)
⎛ d + cQ ⎞ ⎛ d −c ⎞ ⎛ d −c ⎞ ⎫⎪
Q⎜ σ ⎟ + Q⎜ σ I ⎟ + Q⎜ σ Q ⎟⎬
⎜ w ⎟ ⎜ w ⎟ ⎜ w ⎟⎪
⎝ 2 ⎠ ⎝ 2 ⎠ ⎝ 2 ⎠⎭

It is known that cI and cQ are Gaussian with a zero-mean value and a variance of
σ c2 2 and that:

d 3γ s
= ,
σw 2( M − 1)

where γ s = is the received symbol SNR, and σ w2 is the AWGN

received signal power

σ w
2

variance.
9.2 DC Offset in MIMO Transceivers 243

After some calculations, the MQAM BER under the impact of a DC offset is
approximated as:

PDC ( M ) = PbMQAM (γ s ) ≅
(
4 1− 1
M ) ×
2π P
log 2 M
mσ w2 SDCR (9.32)
− mσ w2 β 2
∞⎡ ⎛ ⎞ ⎛ 3γ s ⎞⎤
3γ s 2P
∫ ⎢⎢Q ⎜⎜
M −1
+ β ⎟ + Q⎜
⎟ ⎜ M −1
− β ⎟⎥
⎟⎥
e SDCR dβ
0⎣ ⎝ ⎠ ⎝ ⎠⎦
Δ

where SDCR = mP σ c2 (signal-to-DC ratio).

9.2.3 MIMO System Model

Each receiving path is impaired by a DC offset, which causes BER degradation in
comparison with the ideal case. The input signal, x ( Nt × 1 ), and the output sig-
nal, y ( N r × 1 ), are related by:

y = Hx + w + c (9.33)

where w ( N r × 1 ) is the AWGN and its entries are independent and

wi ~ CN (0, σ w2 ) ; and, c ( N r × 1 ) is the DC offset.
The following definitions are supposed:
Δ Δ Δ
m = min( N t , N r ), n = max( N t , N r ), d = n − m (9.34)

If singular value decomposition (SVD) is applied to H , (9.33) is converted to:

y′ = Dx ′ + w ′ + c′ (9.35)

where:
Δ Δ Δ Δ
y′ = U H y , x ′ = V H x , w ′ = U H w , c ′ = U H c (9.36)

and U , V and D are the decomposing elements of H , which means H=UDV H .

U and V are unitary matrices, so the powers of x and x ′ , y and y′ , w and
w ′ , c and c′ are the same, respectively. D is a diagonal matrix with the singular

{λ}
m
values of H , with i as its main diagonal elements. As discussed in the
i =1
previous section, a MIMO channel by SVD is converted into m parallel SISO
subchannels.
244 Chapter 9 RF Impairments in MIMO Transceivers

9.2.4 BER of Adaptive Modulation MIMO under the Impact

of DC Offset
In this section we investigate the impact of DC offset on adaptive modulation in
MIMO systems. It is assumed that all subchannels have been impaired by the DC off-
set with identical variances, so the total bit error probability is equal to the average
BER in each subchannel. The average BER in each subchannel can be expressed as:
Average BER =
N
∑ PDC (M j ) × log 2 (M j ) P(υ j ≤ λ1 < υ j +1 ) (9.37)
j =1
N
∑ log 2 ( M j ) P(υ j ≤ λ1 < υ j +1 )
j =1

where PDC ( M j ) is the BER that corresponds to M j , and P (υ j ≤ λ1 < υ j +1 ) is the

probability that λ1 falls in the jth region, so that:

υ j +1
p(υ j ≤ λ1 < υ j +1 ) = ∫ Pdf λ1 (λ1 )d λ1 (9.38)
υj

The adaptation scheme is

received signal power =

−2 ln(5BERtgt )σ w2 (9.39)
λ1 P1 (λ1 ) = ( M i − 1)
3
From (9.32), the PDC ( M j ) , under the impact of DC offset in the adaptation
scheme, can be written as:

PDC ( M j ) ≅
(
4 1− 1
Mj ) ∞
× ∫ ⎡⎢Q ( −2 ln(5 BERtgt ) + β )
2π P ⎣
0
log 2 M j
mσ w2 SDCR (9.40)
− mσ β
2
w
2

( ) ⎤ e SDCR
2P
+Q −2 ln(5BERtgt ) − β ⎥ dβ
⎦
Therefore, the total BER of MIMO adaptive modulation under the impact of DC
offset can be achieved as:

BERT = Average BER (9.41)

9.2 DC Offset in MIMO Transceivers 245

where BERT is the total BER of the adaptive MIMO system, including the effect
of the DC offset.

9.2.5 BER Upper Bound of Adaptive Modulation under the

Impact of DC Offset
In [26], the following upper bound relation is used for computation of the AM
power gain boundaries:

⎛ −1.5γ s ⎞
BER ≤ 0.2 exp ⎜ ⎟ (9.42)
⎝ M −1 ⎠

where γ s is the symbol SNR.

To improve the DC-affected AM system, adaptation should be done with re-
spect to the DC offset, because a DC offset degrades the AM MIMO BER, allow-
ing it to become greater than BERtgt . If we change the mechanism of adaptation
by considering DC offset, we may have a BER better than BERtgt .
In order to obtain the optimal power and rate adaptation for different modula-
tion schemes, we need an expression for each modulation technique’s BER in
AWGN that is easily inverted, with respect to rate and power.
Unfortunately, the approximated DC affected BERis neither easily invertible
nor easily differentiable in its argument, but these properties are needed for
adaptive modulation design. Therefore, we now introduce a new tight BER ap-
proximation for MQAM affected by DC offset in AWGN, which can be easily dif-
ferentiated and inverted. The following upper bound for (9.32) is an acceptable
approximation for optimal adaptation:
⎛ ⎞
⎜ ⎟
⎜ −1.5γ s ⎟ (9.43)
PDC ( M ) ≤ 0.2 exp ⎜ ⎟
⎛
⎜ ( M − 1) ⎜ 1 + P ⎞ ⎟
⎜ ⎜ mσ 2 SDCR ⎟⎟ ⎟
⎝ ⎝ w ⎠⎠

As can be seen, the variance of noise ( σ w2 ) has been added to the variance of the
DC offset ( σ c2 ); and, an upper bound relation for the BER in the presence of DC
offset has been calculated from (9.42) (replacing σ w2 by σ w2 + σ c2 ). It is a logical
approximation due to the nature of the dynamic DC offset, as explained in Section
4. The accuracy of this approximation has been verified by simulation and is shown
in Figure 9.12. Based on this upper bound, a new adaptation mechanism is used:
⎧k1 (λ1 ) = ri , υi′ ≤ λ1 < υi′+1
⎪
⎨ M i − 1 σ w2 ⎛ P ⎞ (9.44)
⎪ P1 (λ1 ) = λ K
⎜⎜1 +
σ 2 ⎟⎟ , υi′ ≤ λ1 < υi′+1
⎩ 1 ⎝ m w SDCR ⎠
246 Chapter 9 RF Impairments in MIMO Transceivers

where

M i − M i−1 μ ⎛ P ⎞
υi′ = r − r ⎜⎜1 + ⎟⎟ (9.45)
⎝ mσ w SDCR ⎠
2
i i −1
K

and μ is calculated from:

N
M i − 1 ⎡ ∞ Pdf λ1 (λ1 ) ∞ Pdf λ (λ1 ) ⎤ P
∑ K ⎣ i
⎢ ∫υ ′
λ1
d λ1 − ∫
υi′+1 λ1
1
d λ1 ⎥ =
⎛ ⎞
(9.46)
i =1 ⎦ mσ w ⎜⎜ 1 +
2 P
⎟⎟
⎝ mσ w SDCR ⎠
2

0
DC offset effect on MQAM, σ 2c =1
10

BER Upper Bound

-1
10 Approximation

accurate BER

-2
10
BER

-3
10

256QAM

σ 2c =1
With DC offset
-4
10
64QAM

σ 2c =0 4QAM
without DC offset
-5
10 16QAM

-6
10
0 5 10 15 20 25 30 35
SNR (dB)

Fig. 9.12 The accurate and approximate BER under the impact of DC offset
9.2 DC Offset in MIMO Transceivers 247

In this case, the BER is calculated as:

PDC ( M j ) ≅
(
4 1− 1
Mj ) ×
2π P
log 2 M j
mσ w2 SDCR
∞⎡ ⎛ ⎞
⎛ P ⎞
∫ ⎢⎢Q ⎜⎜
−2 ln(5BERtgt ) ⎜1 + +β⎟ (9.47)
⎜ mσ 2 SDCR ⎟⎟ ⎟
0
⎣ ⎝ ⎝ w ⎠ ⎠
− mσ β
2 2
⎛ ⎛ P ⎞ ⎞⎤ 2P w
+ Q ⎜ −2 ln(5 BERtgt ) ⎜1 + − β ⎟ ⎥ SDCR d β
⎜ mσ 2 SDCR ⎟⎟
e
⎜ ⎝ ⎠ ⎟⎥
⎝ w
⎠⎦

The ASE is calculated from (9.22) with υi′ instead of υi .

9.2.6 Throughput Analysis

The throughput is defined as the ratio of the average number of bits in packets that
are successfully transmitted in any given time interval, divided by the number of
attempted transmissions in that interval [29]. It is a key measure of quality of ser-
vice (QoS) for wireless data transmission systems. The transmitter constructs the
packet and transmits it through the air. The receiver processes the received packet.
Upon detecting the packet, the receiver sends an acknowledgment, either positive
or negative, back to the transmitter. For ease of analysis we assume this feedback
packet goes through a separate control channel and arrives at the transmitter in-
stantaneously and without error. If the receiver detects any error and issues a nega-
tive acknowledgement, the transmitter uses a selective repeat protocol to resend
the packet. It repeats the process until the packet is successfully delivered.
We define the throughput of a system as the number of payload bits received
correctly per second:
L−C
T= Rb Psp (9.48)
L
where T is the throughput, L is the packet length, C is the non-information bits
in a packet, Rb is the bit rate, and Psp is the packet success probability. It is as-
L −C
sumed that L >> C , so is ignored; therefore, T = Rb Psp .
L
Rb is the bit rate, so it is equal to Rb = Rs ri1 + Rs ri2 + ...
+ Rs rim = Rs (ri1 + … + rim ) , where ri1 is the number of bits per each symbol in the
first subchannel, ri2 is the number of bits per each symbol in the second subchan-
nel … to rim , which is the number of bits per each symbol in the last subchannel.
248 Chapter 9 RF Impairments in MIMO Transceivers

Rs is the symbol rate in each subchannel. Each packet contains N s symbols in the
first subchannel, N s symbols in the second subchannel ... to N s symbols in the
last subchannel. The BER in the first subchannel is Pb (ri1 ) and Pb (ri2 ) in the
second subchannel ... to Pb (rim ) in the last subchannel. The probability that all the

( )
N s ri1
bits are transmitted successfully in the first subchannel is 1 − Pb (ri1 ) and … to

(1 − P (r ) )
N s rim
b i1 , which is the probability that all the bits are transmitted success-
fully in the last subchannel .
Since any bit error in the packet results in a loss of the packet, the Psp in our
VRVP AM system is calculated as:

( ) ( )
N s ri1 N s rim
Psp = 1 − Pb ( ri1 ) …… 1 − Pb ( rim ) (9.49)

where N s is the number of symbols in each subchannel, ri j , j = 1,..., m is the

number of bits per each symbol of the jth subchannel, and Pb (ri j ) is the error
probability of a bit. Now the throughput can be calculated as:

T (ri1 ,… , rim ) = Rs (ri1 + … + rim )

(9.50)
( ) ( )
N s ri1 N s rim
× 1 − Pb (ri1 ) …… 1 − Pb (rim )

where Rs is the symbol rate of the system and is the same for all subchannels. Fi-
nally, the average throughput is obtained as:
N N
AT = ∑… ∑ T (ri1 ,… , rim ) ×
i1 =1 im =1 (9.51)
p(υi1 ≤ λ1 < υi1 +1 ,… ,υim ≤ λm < υim +1 )

where p(υi1 ≤ λ1 < υi1 +1 ,…,υim ≤ λm < υim +1 ) is the probability that the constella-
r r
tion size in the first subchannel is 2 i1 and in the second subchannel is 2 i2 and …
r
to 2 im in the last subchannel. The probability is calculated using the joint proba-
bility density function (pdf) of the unordered eigenvalues [28].
Relation (9.51) is very complicated, so we may approximate it with a simpler
relation. For an approximation of (9.51), we may replace the Pb ( ri j ) , j = 1,..., m ,
with the average BER ( BERT ) by using the average value approximation. On the
other hand, the average of (ri1 + … + rim ) is equal to ASE, defined as R. Equation
(9.51) can be approximated as:
9.2 DC Offset in MIMO Transceivers 249

( )
N N
AT ≈ ASE × Rs × ∑ … ∑ 1 − Pb ( ri1 )
N s ri1
……
i1 =1 im =1

(1 − P (r ) )
N s rim
b im × p (υi1 ≤ λ1 < υi1 +1 ,… ,υim ≤ λm < υim +1 )

N N
⇒ AT ≈ ASE × Rs × ∑… ∑ (1 − BERT )
N s ( ri1 + ri2 +....+ rim )
×
i1 =1 im =1

p(υi1 ≤ λ1 < υi1 +1 ,… ,υim ≤ λm < υim +1 )

AT ≈ ASE × Rs (1 − BERT )
ASE × N s
(9.52)

where BERT is calculated from (9.41).

Equation (9.52) is very simple and easy to detect, but it causes approximation
when calculating the average throughput (AT). Accordingly, the BER degradation
of AM MIMO systems under the impact of DC offset can be evaluated. The com-
ponent modulation schemes are uncoded quadrature amplitude modulations
(QAMs) with k1 (λ1 ) ∈ {0, 2, 4, 6, 8, 10} .
We set BERtgt = 10−3 . It is assumed that the channel is known perfectly. At
first, the effect of the DC offset on rgw adaptation design is not considered; hence,
the ASE remains unchanged (similar to corresponding curves in [28], but later, in
the new design of AM, ASE alters). Six transmit-receive antenna configurations
are achieved and depicted in Figure 9.13. The simulation results are also depicted
in Figure 9.13. For all receiving branches, the DC offset is set to SDCR = 20 dB .
The BERT curve can be achieved from (9.41).
Figure 9.14 shows the BERT curve for 6 different configurations of transmit-
receive antenna in the non-ideal case ( SDCR = 20 dB ) and the ideal case
( SDCR = Inf ). As can be seen in Figure 9.14, by increasing the number of anten-
nas, the MIMO adaptive modulation BER is less degraded. The SISO system is
more degraded in comparison with the MIMO systems. The (2,2) configuration
has the worst BER performance, and the (16,4) has the best BER performance
among the MIMO systems. This is expected, since according to the definition,
σ c2 is proportionate to the reverse of m .
The BER of MIMO adaptive modulation under various SDCR values for (2,2)
and (4,4) structures is depicted in Figure 9.15. Moreover, Monte Carlo simulation
results are shown in Figure 9.15. As expected, by decreasing the SDCR , the
MIMO adaptive modulation BER is more degraded. Due to the adaptation me-
chanism in the ideal case, the actual instantaneous BER lies below the BERtgt .
250 Chapter 9 RF Impairments in MIMO Transceivers

40

ASE of MIMO adaptive modulation, R (bps/Hz)

35
analytic result
simulation result
30

25

20 (8,2)
(16,4) (4,4)

15

10 (4,2) (2,2)

5
SISO
(1,1)
0
0 5 10 15 20 25 30
SNR (dB)

Fig. 9.13 ASE of AM MIMO in 6 various system configurations by supposing

BERtgt = 10−3 , versus SNR = P
σ w2

When the SNR is low or the DC offset is not critical, the actual BER is as good as
the ideal case; however, when DC offset gets more severe, the BER becomes
greater than BERtgt .
Figures 9.14 and 9.15 also show that the impact of DC offset on the BER is
higher for higher SNRs, meaning that the BER is more degraded in the high SNR
region. This is because the AM system uses high order component modulations in
this region, which are more sensitive to the DC offset. This result can also be seen
γs
easily with considering that is constant in VRVP AM systems. It can al-
( M − 1)
so be seen that, in low SNRs (low P ), the effect of the DC offset is very small,
which is also shown in Figures 9.14 and 9.15. The BER of a new design for an
AM MIMO system for a (2,2) configuration is depicted in Figure 9.16
( SDCR = 15 dB , 25 dB ). As shown in this figure, in the new design of AM, the
BER remains below BERtgt = 10−3 .
In Figure 9.17, the ASE of the new design of an AM MIMO system for a (2,2)
configuration is depicted ( SDCR = 15 dB, 25 dB ) and compared to the ASE of the
traditional design of an AM MIMO system. It is shown that the ASE of the new
design of an AM MIMO system is decreased. Thus, improvement of the BER
leads to a lower ASE. The proposed AM mechanism helps to avoid BER degrada-
tion by adjusting the spectral efficiency.
9.2 DC Offset in MIMO Transceivers 251

-1
10

Nonideal (SDCR=20 dB):

BER of MIMO adaptive modulation

SISO (1,1)
(2,2)
-2
(4,2)
10 (8,2)
(4,4)
(16,4)

-3 BERtgt
10

Ideal
(without DC offset)
-4
10
0 5 10 15 20 25 30
SNR (dB)

Fig. 9.14 Average BER of AM MIMO systems under the impact of DC offsets (analytic re-
−3
sult), BERtgt = 10 , SDCR = 20 dB SNR = P
σ w2

-1
10

analytic result
simulation result
BER of MIMO adaptive modulation

(2,2) , SDCR=20

-2
10
(4,4) , SDCR=20

(2,2) , SDCR=30

(4,4) , SDCR=30

-3 BERtgt
10

(2,2) , SDCR=Inf
(4,4) , SDCR=Inf
-4
10
0 5 10 15 20 25 30
SNR (dB)

Fig. 9.15 Average BER of AM MIMO systems under the impact of DC offsets under vari-
−3
ous SDCR values, BERtgt = 10 , SNR = P
σ w2
252 Chapter 9 RF Impairments in MIMO Transceivers

10
-1 New Design of AM MIMO
SDCR= 15dB New Design
SDCR= 25dB New Design
BERtgt
SDCR= 15dB Traditional Design
-2
SDCR= 25dB Traditional Design
10
BER

-3
10

-4
10
0 5 10 15 20 25 30
SNR (dB)

Fig. 9.16 Average BER of a new design of AM MIMO system, N t = 2 , N r = 2

BERtgt = 10−3 , SNR = P (2, 2)

σ w2

14
New Design of AM MIMO
SDCR=25dB Traditional Design
12
SDCR=15dB Traditional Design
10
SDCR=15dB New Design
SDCR=25dB New Design
R (bps / Hz)

0
0 5 10 15 20 25 30
SNR (dB)
Fig. 9.17 ASE of the new design of an AM MIMO system, N t = 2 , N r = 2 ,

BERtgt = 10−3 , SNR = P (2, 2)

σ w2
9.2 DC Offset in MIMO Transceivers 253

The throughput results for the (2,2) configuration are shown in Fig. 9.18. It has
been supposed that Rs = 1 M Symbols / S and N s = 50 . In this figure, the
throughput of an AM MIMO system under the impact of a DC offset
( SDCR = 25 dB ) is compared to the throughput of ideal case (i.e. without DC off-
set) and the throughput of the new design of an AM MIMO system. As can be
seen in this figure, the DC offset degrades the throughput dramatically, and the
new design of AM improves it. Due to the BER degradation, the throughput is de-
creased in the high SNR region. There is an acceptable agreement between simu-
lation results and analytic results in most situations.As may be seen, the difference
between the analytical and simulation results is higher with larger SNRs. Howev-
er, this is predictable, due to the use of larger constellation sizes by the system in
higher SNRs. As discussed, the analytical relation only predicts the performance
degradation due to the DC offset imperfection.
The imperfections due to the other parameters, such as phase noise and I/Q im-
balance, also have impacts that are more noticeable in larger constellation sizes,
which correspond to larger SNRs. Accordingly, the difference between the simula-
tion and analytical results will be greater with high SNRs than those with small
SNRs.

12
Without DC Offset, Anal.
Without DC Offset, Simul.
SDCR=25dB, New Design,Anal.
10 SDCR=25dB, New Design,Simul.
SDCR=25dB, Traditional Design,Anal.
SDCR=25dB, Traditional Design,Simul.
8
AT (M bps)

0
0 5 10 15 20 25 30
SNR (dB)

Fig. 9.18 Average throughput of AM MIMO in the presence of DC offset. N s = 50

Rs = 1 M Symbol / S N t = 2 , N r = 2 , BERtgt = 10−3 , SNR = P (2, 2)

σ w2
254 Chapter 9 RF Impairments in MIMO Transceivers

The calculated closed form expressions for BER and throughput give us a use-
ful tool to determine whether an ideally designed VRVP AM MIMO system can
work in the presence of DC offset.

9.3 I/Q Imbalance in MIMO Transceivers

The I/Q (in-phase/quadrature) imbalance in MIMO systems is also a limiting fac-
tor. In this section, the impact of I/Q imbalance on adaptive modulation MQAM
MIMO systems is investigated.

9.3.1 I/Q Imbalance Model

The distortion parameters, μi and υi , are related to the amplitude and phase im-
balances between the I and Q branches of each receiving path in the RF/analog
demodulation process.

(1 + α )cos(2π f Lo t + )
θ
2 ziI
LPF ADC
yi (t )
LNA DSP
ziQ
LPF ADC

−(1 − α )sin(2π f Lo t − )
θ
2
Fig. 9.19 I/Q imbalance in a direct conversion receiver

A simplified model for this distortion is the assumption that there is an ampli-
tude imbalance, α, and a phase imbalance, θ, between the I and Q paths of the
mixer, as depicted in Figure 9.19. For such a case, the μi and υi parameters can
be written as [31]:

θ θ
μ = cos( ) + jα sin( )
2 2
(9.53)
θ θ
υ = α cos( ) − j sin( )
2 2
9.3 I/Q Imbalance in MIMO Transceivers 255

9.3.2 MIMO System Model

A flat-fading MIMO channel with Nt transmitting antennas and Nr receiving an-
tennas is considered. Assuming perfect synchronization and perfect channel state
information (CSI) at the receiver, the input-output relationship is given by:

y=Hx+w (9.54)

where y is an Nr×1 vector of the received signal; H denotes the Nr× Nt channel ma-
trix; x is an Nt×1 vector of the transmitted symbols; and, w is an Nr×1 additive
white Gaussian noise (AWGN) vector. The entries of w are also assumed to be

{hμυ }μ = υ = , are
N ,N
independent and wi ∼ CN (0, σ 2 ) . The channel coefficients, r t
1, 1
statistically independent and identically distributed (i.i.d.) complex-valued Gaus-
sian random variables with a variance of 1.
Components of the noise vector have the same distribution with a variance of
N0. The received signal, y, after distortion by I/Q imbalances becomes [32]:

z=μy + υy* (9.55)

⎛ μ1 0 … 0 ⎞ ⎛υ1 0 … 0 ⎞
⎜ ⎟ ⎜ ⎟
⎜0 μ2 … ⎟ and υ= ⎜ 0 υ2 … ⎟.
where μ =
⎜ 0 ⎟ ⎜ 0 ⎟
⎜ ⎟ ⎜ ⎟
⎜0 ⎟
… μ Nr ⎠ ⎜ ⎟
⎝ 0 ⎝ 0 0 … υ Nr ⎠

The distortion parameters, μi and υi , are related to the amplitude and phase im-
balances between the I and Q branches of each receiving path in the RF/analog
demodulation process. A simplified model for this distortion is the assumption that
there is an amplitude imbalance, α, and a phase imbalance, θ, between the I and Q
paths of the mixer, as depicted in Figure 9.20.
For such a case, the μi and υi parameters can be written as [31]:

θ θ
μi = cos( i ) + jα i sin( i )
2 2

θ θ
υi = α i cos( i ) − j sin( i ) (9.56)
2 2

Δ Δ Δ
We define m = min( N t , N r ) , n = max( N t , N r ), d = n − m . If the singular value de-
composition (SVD) is applied to H , it can be expressed as:

H=UDV H (9.57)
256 Chapter 9 RF Impairments in MIMO Transceivers

x1 y1 Z1
quadrature
x2 y2 demodulator Z 2 detection
(IQ Distortion) and
estimation

xNt yNr Z Nr

Fig. 9.20 An adaptive modulation MIMO structure

where (.) H denotes the conjugate transpose; D is an Nr× Nt matrix with singular

{ λ}
m
values of H, and i are its main diagonal elements; and, U and V are Nr× Nr
i =1
and Nt× Nt unitary matrices with left and right singular vectors of H as their col-
umns, respectively. For the detection and estimation process, the MIMO system
should be converted to an m parallel SISO system. Hence, z should be multip-
lied by U H [33]:

z′=UH μUDx′ + UH υU*Dx′* + UH μw + U H υw* (9.58)

Δ Δ
where z ′ = U H z and x′ = V H x . Since U and V are unitary matrices, the powers of
x and x ′ are the same, as well z and z′ . If we assume that μ ′ = I − μ , where I
is the identity matrix; then, (9.58) is written as:

z′=Dx′ - U H μ′UDx′ + UH υU* Dx′* + UH μw + U H υw* (9.59)

Because D is a diagonal matrix, the elements of z′ can be written as:

zi′ = λi xi′ + n1i′ + n2 i′ + n3i′ (9.60)

where n1′ = -ui H μ′UDx′ + ui H υU* Dx′* , n′2 =ui H μw and n′3 =u i H υw* .
From (9.60), we can see that the I/Q imbalance introduces n1′i as the cross
channel interference; therefore, the subchannels are no longer parallel. Although
in reality the elements of x′ usually have values chosen from a set of finite sym-
bols, we assume the entries of x′ are i.i.d. Gaussian variables, i.e. [33]:
9.3 I/Q Imbalance in MIMO Transceivers 257

P
xi′ ∼ CN (0, ) (9.61)
m
P
where is the average transmitting power constraint of each subchannel. Under
m
this assumption, n′ is a zero-mean complex Gaussian random variable with the
variance of σ n2i′ . We can calculate σ n2′ as:

σ n′2 = E[n1′ n1′ H ] + E[n′2 n′2 H ] + E[n′3 n′3 H ]

i

(9.62)
=
P H
m
( ) (
u i μ ′UDD H U H μ ′* + υU* DD H U T υ* ui + σ w2 u i H μμ* + υυ* ui )

9.3.3 Impact of I/Q Imbalance on BER of Adaptive Modulation

MIMO
The BER of an MQAM signal in an AWGN channel can be approximated as:

4(1 − 1 Mi ) ⎛ 3γ s ⎞
Q⎜
⎜ M i − 1 ⎟⎟
BER ( M i ) (9.63)
log 2 ( M i ) ⎝ ⎠

where γ s is the symbol of the received SNR.

As seen, the received signal power ( Prec ) due to adaptive modulation can be
expressed as:
−2 ln(5 BERtgt )σ w2
Prec = λ1 P1 (λ1 ) = ( M i − 1) (9.64)
3

One can approximate the BER of a subchannel by considering (9.63) as:

4(1 − 1 Mi ) ⎛ −2 ln(5 BER )σ 2 ⎞

Q⎜ ⎟
tgt w
BERIQ ( M i , H ) (9.65)
log 2 ( M i ) ⎜ σ n2′ ⎟
⎝ ⎠

BER( M i , H ) has the joint pdf of unordered eigenvalues [35]. We can approx-
imate the subchannel BER by averaging the σ n2′ using:

EH [σ n2′ ] = λ1 ×
P H
m
( ) ( )
ui μ′μ′* + υυ * ui + σ w2 ui H μμ* + υυ * ui (9.66)

∞
where λ1 = ∫ λ1 Pdf λ1 (λ1 ) d λ1 .
0
258 Chapter 9 RF Impairments in MIMO Transceivers

If we consider:
⎛ P θ ⎞
Δ = max ⎜ λ1 (2 + α 2j − 2 cos( 2j )) + σ w2 (1 + α 2j ) ⎟
⎝ m ⎠ (9.67)
j = 1,..., N r

An approximated upper bound for the BER can then be achieved:

4(1 − 1 ⎛ −2 ln(5 BER )σ 2

Mi ) ⎞
BERIQ ( M i ) ≤ Q⎜ ⎟
tgt w
(9.68)
log 2 ( M i ) ⎜ Δ ⎟
⎝ ⎠
The average BER of the adaptive modulation MIMO system impaired by I/Q im-
balance can be expressed as:
N
∑ BERIQ (M j ) × log 2 (M j ) P(η j ≤ λ1 < η j +1 )
j =1
BERave = N
(9.69)
∑ log 2 (M j ) P(η j ≤ λ1 < η j +1 )
j =1

where P (η j ≤ λ1 < η j +1 ) is the probability that λ1 falls in the jth region as:
η j +1
p(η j ≤ λ1 < η j +1 ) = ∫ Pdf λ1 (λ1 )d λ1 (9.70)
ηj

9.3.4 I/Q Imbalance Compensation in Adaptive Modulation

MIMO Systems
To compensate for a distorted adaptive modulation MIMO system, adaptation
should be done with respect to the I/Q imbalance. In order to obtain the optimal
power and rate adaptation for different modulation schemes, we need an expres-
sion for the BER in AWGN that is easily inverted, with respect to rate and power.
Accordingly, we now introduce a new BER approximation for the MQAM dis-
torted by an I/Q imbalance in AWGN.
From the results of the last section, the following upper bound can be intro-
duced for optimal adaptation as:

⎛ −1.5γ ⎞
BERIQ ( M ) ≤ 0.2 exp ⎜ s ⎟ (9.71)
⎜ ( M − 1) Δ σ 2 ⎟
⎝ w ⎠
The values of α and θ are not known at the receiver; therefore, for the implemen-
tation of a new adaptation scheme, we use the average of Δ , in terms of α and θ
statistics:
9.3 I/Q Imbalance in MIMO Transceivers 259

Δ P
E[ ] = λ1 (2 + E[α 2 − 2 cos( θ2 )]) + (1 + E[α 2 ]) (9.72)
σ w2 mσ w2

Based on the definitions of α and θ , E[α ] = E[θ ] = 0 . If we assume a small θ :

1
E[cos( θ2 )] 1 − E[θ 2 ] (9.73)
8

Accordingly,
Δ P 1
E[ ] = λ1 (σ α2 + σ θ2 ) + 1 + σ α2 (9.74)
σ 2
w mσ w
2
4

Now the BER relation is approximated as:

⎛ ⎞
⎜ −1.5 γ ⎟
BERIQ ( M ) ≤ 0.2exp ⎜ s ⎟ (9.75)
⎜ P 1 2 2 ⎟
⎜ ( M − 1) (λ1 mσ 2 (σ α + 4 σ θ ) + 1 + σ α ) ⎟
2

⎝ w ⎠

Based on this approximation, we introduce a new adaptation mechanism as:

⎧ k1 (λ1 ) = ri , ηi′ ≤ λ1 < ηi′+1

⎪
⎨ M i − 1 σ w2 (9.76)
⎪ P1 (λ1 ) = λ CF , ηi′ ≤ λ1 < ηi′+1
⎩ 1 K

where the compensation factor ( CF ) is defined as:

P 1
CF = λ1 (σ α2 + σ θ2 ) + 1 + σ α2 (9.77)
mσ w2 4

M −M μ
and ηi′ = ri − r i−1 CF .
i i −1
K
The Lagrangian multiplier, β , is calculated as:

N
M i − 1 ⎡ ∞ Pdf λ1 (λ1 ) ∞ Pdf λ (λ1 ) ⎤ P
∑ ⎢ ∫η ′
λ1
d λ1 − ∫ 1

λ1
d λ1 ⎥ = (9.78)
⎦ mσ w CF
K ⎣ i η ′
i +1
2
i =1

and the ASE is calculated as:

N
R = m∑ ri ⎡ ∫ Pdf λ1 (λ1 )d λ1 − ∫ Pdf λ1 (λ1 )d λ1 ⎤
∞ ∞
⎢ (9.79)
i =1 ⎣ ηi′ ηi′+1 ⎦⎥
260 Chapter 9 RF Impairments in MIMO Transceivers

9.3.5 BER Analysis

In this section, an experimental study is presented; and, the analytical relations are
compared with measurement results. The effect of an I/Q imbalance (for two sets
of phase and amplitude errors) on the BER of adaptive modulation MIMO is de-
picted in Figure 9.21. The BER of an ideal adaptive modulation MIMO is also il-
lustrated in this figure. The comparison between the ideal and impaired systems
shows the extreme degradation of the adaptive MIMO system due to the I/Q im-
balance. Good agreement between the measurement results and the simulation re-
sults can be observed in this figure. It can be seen that the approximated analytic
BER upper bound follows the simulation and measurement results very well. The
impact of different parameters on BER is determined from this analytic expres-
sion. Figure 9.21, however, shows an unusual behavior: by increasing the average
SNR, the BER became more degraded. This response is expected, because the
adaptive modulator uses high-order components, which are more sensitive to I/Q
imbalance, in the modulation of high SNR regions.

-1
10 α =[.374 .682 .338 .306] (dB)
θ =[3.84 7.68 4.2 6.7] (degree)

Analytic (Up.Bou.)
Measurement
10
-2 Simulation
BER

10
-3 α =[.17 .086 .128 .253](dB)
θ =[5 2 7 1] (degree)

without IQ imbalance

0 5 10 15 20 25 30
SNR (dB)

Fig. 9.21 BER of an adaptive modulation MIMO system with I/Q imbalance

Figure 9.22 (top) shows the BER of an adaptive modulation MIMO system im-
paired by an I/Q imbalance before and after the compensation operation by
supposing σ α = 0.3 dB and σ θ = 5 .
9.3 I/Q Imbalance in MIMO Transceivers 261

Measurement
Analitic(Up.Bou.)
Simulation

α =[0.274 0.245 0.191 0.334] (dB)

-2 θ =[5.2 3.8 3.5 4.2] (degree)
10
BER

Compensation by:
σ =0.3 dB
α Non-Compensated
σ =5 degree
θ

-3
10 Compensated

0 5 10 15 20 25 30
SNR (dB)

Analytic
25 Simulation
Measurement

Non-Compensated
20

α =[0.274 0.245 0.191 0.334] (dB)

15 θ =[5.2 3.8 3.5 4.2] (degree)
ASE

10 Compensated

5 Compensation by:
σ =0.3 dB
α
σ =5 degree
θ

0
0 5 10 15 20 25 30
SNR (dB)

Fig. 9.22 (Top) BER of an I/Q imbalance compensated adaptive modulation MIMO system,
and (bottom) ASE of an I/Q imbalance compensated adaptive modulation MIMO system
262 Chapter 9 RF Impairments in MIMO Transceivers

References
[1] Paulraj, A., Nabar, R., Gore, D.: Introduction to Space-Time Wireless Communica-
tions. Cambridge University Press (2003)
[2] Bolcskei, H., Gesbert, D., Papadias, C.B., Veen, A.: Space-Time Wireless System
From Array Processing to MIMO Communications. Cambridge University Press
(2006)
[3] Schenk, T., Linnartz, J.: RF Imperfections in High-Rate Wireless Systems, Impact
and Digital Compensation. Springer, Heidelberg (2008)
[4] Piazzo, L., Mandarini, P.: Analysis of phase noise effects in OFDM modems. IEEE
Transactions on Communications 50(10), 1696–1705 (2002)
[5] Tomba, L.: On the effect of wiener phase noise in OFDM systems. IEEE Transac-
tions on Communications 46(5), 580–583 (1998)
[6] Howald, R., Kesler, S., Kam, M.: BER performance of MQAM using OFDM with RF
carrier phase noise. In: Proceedings of Southeastern Symposium on System Theory,
pp. 419–423 (1998)
[7] Liu, P., Bar-Ness, Y.: Closed-form expressions for BER performance in OFDM sys-
tems with phase noise. In: Proceedings of International Conference on Communica-
tions (ICC), Turkey, pp. 5366–5370 (2006)
[8] Lu, J., Letaief, K.B., Chuang, J., Liou, M.L.: M-PSK and M-QAM BER computation
using signal-space concepts. IEEE Transactions on Communications 47(2), 181–184
(1999)
[9] Proakis, J.G., Salehi, M.: Digital Communications, 5th edn. McGraw-Hill (2007)
[10] Dohler, M.: Virtual antenna arrays, Ph.D. dissertation, King’s College, London, U.K
(2003)
[11] Chung, S.T., Goldsmith, A.J.: Degrees of freedom in adaptive modulation: A unified
view. IEEE Transactions on Communications 49(9), 1561–1571 (2001)
[12] Keshavarzi, M.R., Mohammadi, A., Abdipour, A.: Characterization of Adaptive
Modulation MIMO Systems in the Presence of Phase Noise. In: International Confe-
rence on Wireless Communications & Mobile Computing, IWCMC, Leipzig, Germa-
ny, pp. 1243–1247 (2009)
[13] Muschallik, C.: Influence of RF oscillators on an OFDM signal. IEEE Transactions
on Consumer Electronics 41, 592–603 (1995)
[14] Gershman, A.B., Sidiropoulos, N.D.: Space-Time Processing for MIMO Communica-
tions. Wiley (2005)
[15] Tsoulos, G.: MIMO System Technology for Wireless Communications. CRC Press
(2006)
[16] Zhou, S., Giannakis, G.B.: How accurate channel prediction needs to be for transmit-
beamforming with adaptive modulation over Rayleigh MIMO channels? IEEE Trans-
actions on Wireless Communications 3(4), 1285–1294 (2004)
[17] Leung, B.: VLSI for Wireless Communications. Prentice Hall (2002)
[18] Hemesi, H., Azami, F., Ghorssi, A., Abdipour, A., Mohammadi, A.: Low-complexity
channel and frequency offset estimation in MIMO systems. Electronic Letters 44(9),
599–600 (2008)
[19] Larson, E.: RF and Microwave Circuit Design for Wireless Communications. Artech
House, Norwood (1996)
References 263

[20] Bateman, A., Haines, D.M.: Direct conversion transceiver design for compact low-
cost portable mobile radio terminals. In: Proceedings of the 43th IEEE Vehicular
Technology Conference, pp. 55–62 (1989)
[21] Keshavarzi, M.R., Mohammadi, A., Abdipour, A., Ghannouchi, F.M.: Characteriza-
tion and Compensation of DC Offset on Adaptive MIMO Direct Conversion Tran-
sceivers. IEICE Transactions on Communications (1), 253–261 (2011)
[22] Kenington, P.B.: RF and Baseband Techniques for Software Defined Radio. Artech
House, Norwood (2005)
[23] Lindoff, B., Malm, P.: BER performance analysis of a direct conversion receiver.
IEEE Transactions on Communications 50(5), 856–865 (2002)
[24] Lu, J., Letaief, K.B., Chuang, J.C.-I., Liou, M.L.: M-PSK and M-QAM BER Compu-
tation Using Signal-Space Concepts. IEEE Transactions on Communications 47,
181–184 (1999)
[25] Proakis, J.G., Salehi, M.: Digital Communications, 5th edn. McGraw-Hill (2007)
[26] Mohammadi, A., Kumar, S.: Characterization of adaptive modulators in fixed wire-
less ATM networks. KICS/IEEE Journal of Communications and Networks 6(2),
123–132 (2004)
[27] Dohler, M., Aghvami, H.: On the approximation of MIMO capacity. IEEE Transac-
tions on Wireless Communications 4(1), 30–34 (2005)
[28] Zhou, Z., Vucetic, B., Dohler, M., Li, Y.: MIMO systems with adaptive modulation.
IEEE Transactions on Vehicular Technology 54(5), 1828–1842 (2005)
[29] Goldsmith, A.: Wireless Communications. Cambridge University Press (2005)
[30] Pursley, M.B., Shea, J.M.: Adaptive nonuniform phase-shift-key modulation for mul-
timedia traffic in wireless networks. IEEE Journal on Selected Areas in Communica-
tions 18, 1394–1407 (2000)
[31] Liu, C.L.: Impacts of I/Q imbalance on QPSK-OFDM-QAM detection. IEEE Trans-
actions on Consumer Electronics 44(3), 984–989 (1998)
[32] Tarighat, A., Sayed, A.H.: MIMO OFDM receivers for systems with IQ imbalances.
IEEE Transactions on Signal Processing 53(9), 3583–3596 (2005)
[33] Zhou, Z., Vucetic, B., Dohler, M., Li, Y.: MIMO systems with adaptive modulation.
IEEE Transactions on Vehicular Technology 54(5), 1828–1842 (2005)
[34] Dohler, M.: Virtual antenna arrays, Ph.D. dissertation, King’s College, London
(2003)
[35] Telatar, I.E.: Capacity of multi-antenna Gaussian channels. European Transactions on
Telecommunications 10, 585–595 (1999)
[36] Hemesi, H., Azami, F., Ghorssi, A., Mohammadi, A., Abdipour, A.: Design and Im-
plementation of a Flexible 4x4 MIMO Testbed. In: Proceedings of the International
Symposium on Telecommunications (IST 2008), pp. 266–272 (2008)
Chapter 10
Single RF Front-End MIMO Transceivers

The main limitation of using a multiple antenna architecture is the complexity and
high cost of the hardware in the radio frequency (RF) section, which rise with an
increase in the number of antennas. In addition, RF circuit mismatches [1-3] and
coupling [4] grow with an increase in the number of antennas; consequently, these
factors also limit the use of high numbers of antennas at the transceiver.
One solution to compensate for the extra hardware cost and RF circuit imper-
fection is the utilization of a single RF front-end in a multiple input multiple out-
put (MIMO) system, where a single RF path is used instead of multiple parallel
RF paths. This results in an RF section that has lower complexity and cost, a simp-
ler RF design, a compact size, and lower power consumption.
A conventional MIMO receiver with N antennas at the receiver is shown in
Figure 10.1. The receiver uses multiple parallel RF front-ends, where the number
of RF front-ends is equal to the number of antennas. In this architecture, the base-
band processing section decodes N received baseband paths to recover the signal
and obtain the diversity or spatial multiplexing gain. To realize a single RF front-
end path, orthogonal transmission of multiple RF streams over a single front-end
must be recognized. In the following subsections, the different techniques to real-
ize a single RF front-end in a MIMO system are discussed.

RFChain RFChain
Baseband

Baseband

Input MIMO Output

Channel
RFChain RFChain

Fig. 10.1 A conventional MIMO system

A. Mohammadi et al.: RF Transceiver Design for MIMO Wireless Communications, LNEE 145, pp. 265–288.
springerlink.com © Springer-Verlag Berlin Heidelberg 2012
266 Chapter 10 Single RF Front-End MIMO Transceivers

10.1 RF Front-End MIMO Using Antenna Selection

MIMO systems increase spatial diversity; and, in proper channel conditions, the
data rate may be increased using spatial multiplexing implementation. As dis-
cussed in Chapter 2, diversity methods improve the robustness of the communica-
tion system, in terms of BER, by exploiting the multiple paths between the
transmitting and receiving antennas. This is achieved by space-time coding. The
space-time codes are capable of delivering a diversity order of NrNt, where Nr and
Nt are the number of receiving and transmitting antennas, respectively. A MIMO
system can be also designed to provide spatial multiplexing. In this implementa-
tion, the receiver descrambles signals that are transmitted simultaneously from
multiple antennas. Accordingly, it is possible to send parallel independent data
streams and achieve overall system capacities that scale with min(Nr, Nt).
The performance enhancement of MIMO systems comes with an increased RF
hardware complexity and cost. This problem can be mitigated using antenna selec-
tion technique at the transmitter and/or receiver to realize a simpler RF front-end
MIMO system [5]. By using such a technique, some of the available antennas are
selected, and the MIMO system uses fewer RF chains than the number of trans-
mitter and/or receiver antennas. Therefore, the complexity and cost are reduced
while maintaining the system performance.
The antenna selection system, both in the receiver and the transmitter, is pre-
sented in Figure 10.2. In this technique, the received signal-to-noise ratio (SNR)
values must be known on both the receiver and transmitter sides. Although imple-
mentation of antenna selection in the receiver is very straightforward, antenna selec-
tion in the transmitter requires a feedback path from the transmitter to the receiver.

Nr
Lt Nt
Lr

RFChain RFChain
RFswitch
RFswitch
Baseband

Baseband

Input MIMO Output

Channel
RFChain RFChain

Rx − Selection
Tx − Selection

Fig. 10.2 A single antenna selection system

A common criterion for antenna selection is channel capacity maximization. An

analytical bound for the channel capacity of MIMO systems with antenna selec-
tion is [6]:
Lr
ρ
Cselect ≤ ∑ log 2 (1 + γi ) (10.1)
i =1 Nt
10.1 RF Front-End MIMO Using Antenna Selection 267

where ρ is the mean SNR, Lr is the number of selected antennas in the receiver,
and γ(i) represents the squared norm of the ith row of H after ordering from the
smallest to the largest. They are obtained by ordering a set of Nr independent and
identically distributed (i.i.d.) chi-square random variables with 2Nt degrees of free-
dom (DOF). Without loss of generality, when the receiver selects the best Lr anten-
nas that maximize capacity, the resulting channel capacity can be upper bounded [6].
It has been shown that most of the capacity of the MIMO system is retained with an-
tenna selection, provided that the number of selected antennas on one end should be
at least as many as the number of available antennas on the other end.
Some suboptimal antenna selection algorithms have also developed [8], [9]. The
fastest algorithm starts with zero antennas and adds one antenna per step, where the
added antenna is the greatest contribution to the channel capacity [9], [7].
The other criterion in the selection of antennas is outage probability [7], [10].
When the information transmission rate is greater than the instantaneous mutual
information, an outage event occurs. In quasi-static fading, since the fading coeffi-
cients are constant over the whole frame, we cannot average them with an ergodic
measure. In such an event, channel capacity does not exist in the ergodic sense.
The probability of such an event is normally referred to as outage probability.
The outage probability, when the best Lr receive antennas are selected, can be
upper bounded as [7]:
2 R − 1 ⎛ Nt ⎞ Nt N r
Pselect ≤ P[ ⎜ ⎟ , Nt N r ] → ko γ − Nt Nr (10.2)
R ⎝ Lr ⎠ γ

where R is the information transmission rate in bits per second per hertz, P(x,a) is
the normalized in complete gamma function, and:

2 R − 1 ⎛ Nt ⎞
ko = ( ⎜ ⎟ Nt N r )
Nt N r
/ ( Nt N r )! (10.3)
R ⎝ Lr ⎠

Expression (10.2) suggests that the diversity order is maintained with antenna se-
lection for any N t , N r and Lr. If one selects Lt = Lr = 1 , a single RF transceiver
can be realized for a MIMO system. The single antenna selection diagram is
shown in Figure 10.3.

Input Output
MIMO
Baseband RFChain Channel RFChain Baseband

Selection − Feedback

Fig. 10.3 A single antenna selection system

268 Chapter 10 Single RF Front-End MIMO Transceivers

10.2 Single RF Front-End MIMO Using FDM

The frequency-division multiplexing (FDM) technique is presented in Figure 10.4.
In this method, the signals of different antennas are shifted in frequency with mix-
ing by different local oscillators (LOs) and added together; therefore, the multiple
signals of different antennas are separated in the frequency domain. After down-
conversion using a single RF front-end, the frequency shifts of the multiple
streams are removed in the baseband section.
By further processing, the diversity gain is extracted. The method is also applica-
ble when different signals are transmitted, e.g., Vertical Bell Laboratories Layered
Space-Time (V-BLAST) or other spatial multiplexing MIMO transmission.
The system-level performance of a scenario with two antennas at the receiving
side is studied. Assuming s (t ) as the passband signal with a center frequency, f c ,
and a bandwidth, W . The received signal of the first antenna is h1 s(t ) + n1 (t ) and
that of the second antenna is h2 s (t ) + n2 (t ) , where n1 (t ) and n2 (t ) are the pass-
band noise and h1 , h2 are the Rayleigh flat-fading channel coefficients. The signal
of the second antenna is mixed with a low-frequency oscillator at the same fre-
quency ( W ) and filtered to removed the unwanted terms. The drawback of this
technique is that a narrowband RF filter and extra oscillators are required.

BPF LPF
Baseband
Output

BPF

Cos 2π Wt Cos 2π Wt

Fig. 10.4 Realization of single-branch multiantenna receiver based on FDM

The performance of this structure has been investigated using binary phase-
shift keying (BPSK) modulation over two receiver antennas and is depicted in
Figure 10.5. The difference between the ideal and simulated curves for diversities
1 and 2 is related to filter design and its quality factor.
10.3 Single RF Front-End MIMO Using TDM 269

0
10
Diversity = 1(Ideal)
Diversity = 1(FDM)
Diversity = 2(Ideal)
Diversity = 2(FDM)

-1
10
BER

-2
10

-3
10
0 1 2 3 4 5 6 7 8 9 10
Eb/N0(dB)

Fig. 10.5 The BER versus Eb/N0 for single front-end receiver FDM

10.3 Single RF Front-End MIMO Using TDM

The idea of the time-division multiplexing (TDM) technique in MIMO receivers is
the use of an RF switch and a single receiver, where signals from different antennas
were selected, captured and digitally processed separately. In this regard, the pur-
pose of the switch is the capture of the signals of all the antennas for every symbol
time interval of the modulated signal. This introduces constraints on the switching
speed and also requires a careful alignment of the data before and after the multip-
lexing and de-multiplexing of the multiple signals in the MIMO receivers.
In this section, we investigate the possibility of using the TDM technique to
simultaneously transfer and extract the data of multiple antennas through a single
RF front-end. This receiver can be realized as a single RF architecture by time-
multiplexing the antennas’ signals using a single-pole, multiple-throw RF switch.
The received signals are de-multiplexed after RF processing according to a diver-
sity algorithm. Figure 10.6 shows a general block diagram of this architecture.
This architecture uses the same number of antennas as the conventional topology;
however, instead of having multiple parallel RF front-ends, a single-pole, mul-
tiple-throw RF switch along with a single RF front-end is used to down-convert
the RF signals to baseband. Finally, the signals are conveyed to the baseband sec-
tion using a de-multiplexer. The proposed architecture reduces the number of RF
front-ends from N to 1, which significantly reduces the overall cost and size of the
multiantenna receiver and also decreases the RF design mismatch.
270 Chapter 10 Single RF Front-End MIMO Transceivers

X1 Y1
BPF LPF
Baseband
Receive Processing
Front -End
BPF LPF
XN YN

Fig. 10.6 Realization of single-branch multiantenna receiver based on TDM

The results of the time-multiplexed receiver using a raised cosine pulse shape
filter with a roll-off factor of 0.5 for single antenna and two-antenna cases using
BPSK modulation are shown in Figure 10.7. The zero-forcing (ZF) receivers are
used [11], [16], [21]. To compare the proposed architecture with a conventional
receiver, the bit error rate (BER) of the multiple front-end receivers is plotted as
well. As can be seen, the diversity gain using a single time-multiplexed RF front-
end was equal to the diversity gain of the multiple RF front-end.
In the following subsections, the performance of a single front-end MIMO re-
ceriver is investigated both theoretically and experimentally.

0
10
diversity = 1(Ideal)
diversity = 1(TDM)
diversity = 2(Ideal)
diversity = 2(TDM)

-1
10
BER

-2
10

-3
10
0 1 2 3 4 5 6 7 8 9 10
Eb/N0(dB)

Fig. 10.7 The BER versus Eb/N0 for single front-end receiver using TDM
10.3 Single RF Front-End MIMO Using TDM 271

10.3.1 Input Signal

The received signal considered is an MQAM (multiple or M-ary quadrature ampli-
tude modulation) signal with a carrier frequency of fc. The MQAM modulated sig-
nal, sm(t), can be represented as:

sm ( t ) = Amc u ( t ) cos 2π f c t − Ams u ( t ) sin 2π fc t m = 1, 2, … , L (10.4)

where u(t) is a signal pulse shape, fc is the carrier frequency, Amc and Ams are the
in-phase and quadrature signal amplitudes, M is the constellation size, and
L= M .
The nature of the received signal can be assumed to be deterministic, in order to
simplify the representation of the incoming signal in the frequency domain. The
spectrum of sm(t) can, therefore, be represented as:

Amc A
Sm ( f ) = ⎡⎣U ( f − fc ) + U ( f + f c ) ⎤⎦ − ms ⎡⎣ − jU ( f − fc ) + jU ( f + fc ) ⎤⎦ (10.5)
2 2

where Sm ( f ) and U ( f ) are the Fourier transforms of sm(t) and u(t), respectively.
Meanwhile, it can be easily shown that the power spectral density of an
MQAM modulated signal is [12]:

σ2 ⎡
φss ( f ) = ⎤
2 2

T ⎣⎢ U ( f − f c ) + U (− f − fc ) ⎥ ⎦
(10.6)

where σ 2 is the variance of the information sequence in the in-phase and quadra-
ture paths. The spectral efficiency of the MQAM signal is controlled by the base-
band pulse shape, u(t).

10.3.2 Switch Model

According to the block diagram of the architecture in Figure 10.6, the received
MQAM signals are sampled using a single-pole, N-throw (SPNT) RF switch. It is
assumed herein that the switch takes K samples from each antenna signal during
the symbol period, T. The switch function, which obtains a sample from the first
antenna, can be shown as:

⎛t ⎞
z ( t ) = repT / K [ rect ⎜ ⎟] (10.7)
⎝τ ⎠
272 Chapter 10 Single RF Front-End MIMO Transceivers

The switch waveform and its spectrum are shown in Figure 10.8. The Fourier
transform of this switch waveform can be shown as:

K
Z(f )= rep K ⎡⎣δ ( f ) ⎤⎦ τ Sinc ( f τ ) (10.8)
T T

The number of antennas is considered to be equal to N; therefore, the switch

waveform can be written as:
⎡ ⎛ T/K ⎞ ⎛ ( N − 1) T / K ⎞ ⎤
⎢
⎛ ⎞ ⎜t− N ⎟ ⎜t− ⎟⎥
⎟ ⎥ (10.9)
t
z N ( t ) = repT / K ⎢rect ⎜ ⎟ + rect ⎜ ⎟ +…+ rect ⎜
N
⎢ ⎝τ ⎠ ⎜⎜ τ ⎟⎟ ⎜ τ ⎟⎥
⎢ ⎝ ⎠ ⎜ ⎟⎥
⎣ ⎝ ⎠⎦
The sampling of different signals must use the concept of orthogonal signals in the
time domain to prevent overlapping among the different samples. According to
(10.9), in an ideal condition, the relation between the duty factor and the period is
T /K
obtained as τ ≤ .
N

time
T/ K
(a)

0
−1/τ ... -K/T K/ T ...1/τ frequency

(b)

Fig. 10.8 (a) the time domain switch waveform and (b) its spectrum

10.3.3 Time Sampling of Received Signals

According to Figure 10.6, the switch takes the sample from a particular antenna
output, where the received signal arrives through a Rayleigh fading channel and is
corrupted by AWGN. Let us concentrate on a sample from the first antenna. This
10.3 Single RF Front-End MIMO Using TDM 273

antenna signal is called X1 in Figure 10.6. The signal, after passing through the
filter, can be represented as:

x1 ( t ) = [h1I ( t ) . Amc u ( t ) + n1I (t )]cos(2πf c t ) − [h1Q ( t ) . Ams u ( t ) + n1Q (t )]sin(2πf c t ) (10.10)

where h1I ( t ) and h1Q ( t ) are the real and imaginary parts of the channel coeffi-
cient corresponding to the first antenna, respectively, and are assumed constant
over the symbol duration, T; and, n1I (t ) and n1Q (t ) are the real and imaginary parts
of the AWGN at the output of band-pass filter (BPF), respectively. This signal is
sampled by the RF switch and is represented as:

{ }
x1s ( t ) = ⎡⎣ h1I ( t ) . Amc u ( t ) + n1I ( t ) ⎤⎦ cos(2π f c t ) z (t )
(10.11)
−[h1Q ( t ) . Ams u ( t ) + n1Q (t )]sin(2 π f c t ) z (t )

where the superscript ‘s’ represents the sampled signal.

These samples then pass through the single front-end receiver. It is assumed
herein that the receiver has enough bandwidth that an RF pulse can pass through it
without any distortion. The other RF samples also pass through the receiver at the
different time intervals, and the samples from different antennas are orthogonal in
each symbol duration. Eventually, the sample of the first antenna is delivered to
the corresponding port in the baseband processing section using a low-pass filter
(LPF). This port is called Y1 in Figure 10.6.
Complete synchronization between the RF switch and the de-multiplexer is as-
sumed. If a coherent LO is used at the receiver (i.e., xLO(t)), the output signal, y1(t),
is expressed as:

y1 ( t ) = x1s ( t ) xLO ( t ) (10.13)

where the LO signals in the receiver in the in-phase and quadrature paths are
I
xLO ( t ) = cos ( 2π fC t ) and xLO
Q
( t ) = sin ( 2π fC t ) , respectively. The signal y1 ( t )
also encompasses the second harmonic frequency component, 2 f C , which is
eventually eliminated using a LPF in the baseband section. Accordingly, the out-
put signals can be shown as:

1 I
yDI ( t ) = [h1 Amc u ( t ) z (t ) + n1I ( t ) z (t )] (10.14-a)
2
1 Q
yDQ ( t ) = [h1 Ams u ( t ) z (t ) + n1Q ( t ) z (t )] (10.14-b)
2

where yDI ( t ) and yDQ ( t ) are the in-phase and quadrature components, respective-
ly, of the received signals.
274 Chapter 10 Single RF Front-End MIMO Transceivers

The switch effect is also considered by z ( t ) . If z ( t ) = 1 in (10.14-a) and

(10.14-b), these two equations are true for in-phase and quadrature components of
the first antenna’s baseband received signals. In baseband, the receiver samples
each antenna’s signals and performs matched filtering and maximum ratio com-
bining. If the sampler takes K samples in each symbol duration, T, (10.14-a) and
(10.14-b) are accurate for the output of the sampler when z ( t ) is the same as
(10.7) and τ << T . Therefore, these equations are the same for the common re-
ceiver and the time-multiplexed receiver.

10.3.4 Baseband Processing

According to Figure 10.6, the received signals are delivered using a de-
multiplexer to the baseband processing section. The LPFs are used to extract the
sampled signals, which must be processed in the baseband section to extract the
spatial diversity or spatial multiplexing gain. This is a common assumption in nar-
rowband models [13], where the spatial and temporal properties of the MIMO im-
pulse response, H(t), can be separated as:

H (t ) = u (t ) H (10.15)

where u(t) is the pulse shape. The matched filter, H * (−t ) , in the MIMO base-
band receiver can, therefore, be decomposed into a cascade of a space-only col-
umn matrix, H*, and followed by a bank of time-only matched filters, u ( −t ) . It is
also assumed that the channel is known on the receiver side.
In a classic multiantenna receiver with N antennas and N RF front-ends, K
samples are taken from the baseband signal of each antenna after down-
conversion. The matched filter then combines the samples and improves the SNR
from the samples of each antenna. Consequently, the transmitted symbol is esti-
mated with the information of each fading channel coefficient. The same proce-
dure can be followed in this system, where there is a corresponding baseband sig-
nal for each antenna signal. A zero-forcing receiver to recover the transmitted
symbols [20,21] can be used. When the channel coefficients are known, the sym-
bols are estimated as:

1 I
yDI (t ) = [h1 Amc u (t ) z ( t ) + n1I (t ) z ( t )] (10.16a)
2

1 Q
yDQ (t ) = [h1 Ams u (t ) z ( t ) + n1Q (t ) z ( t )] (10.17a)
2

where Amc and Ams are the estimated in-phase and quadrature signal amplitudes,
respectively.
10.3 Single RF Front-End MIMO Using TDM 275

10.3.5 Experimental Studies

Experimental studies can evaluate the performance of the proposed time-
multiplexed single RF front-end receiver and validate the concept of the topology
in practical scenarios. The experimental studies were conducted for a 16-QAM
modulation scheme, where the number of receiving antennas was selected as
N = 2. The 16-QAM signaling scheme was eight times oversampled and passed
through a raised cosine filter with a roll-off factor of 0.3 [14]. The channel model
was a Rayleigh fading channel with f d Ts = 0.3 . The maximum ratio combining
method was used at the receiver, in order to achieve the maximum diversity.
The objective was a demonstration of the potential of using the proposed time-
multiplexed single-branch receiver instead of conventional multiantenna receivers.
The measurement setup is shown in Figure 10.9. It consisted of two identical signal
generators, which were used to emulate the signal received from the two antennas
through uncorrelated fading channels. These two signal generators were connected
through a general purpose interface bus (GPIB) connection and controlled with a
computer. An evaluation board (MAX2830) was used as the receiver.

Computer running
ADS 2006U & MATLAB
IB
GP
Fi
re
w
ire
Function Gen.
Parallel
33250A
port
ESG 4438C
switch LNA Down VSA
fsw=2fs VGA Converter 89600
ESG 4438C
MAX 2830 Evaluation board

Fig. 10.9 Block diagram of the measurement setup

The MAX2830 is a completely integrated solution for the implementation of

RF transceivers: it contains two RF inputs, an antenna diversity switch, a low
noise amplifier (LNA), a programmable voltage-gain amplifier (VGA), an RF-to-
baseband down-converter and a programmable LPF. Finally, the output of the re-
ceiver board was connected to a baseband vector signal analyzer (VSA), which
was controlled with the computer [14].
276 Chapter 10 Single RF Front-End MIMO Transceivers

Fig. 10.10 Block diagram of the measurement setup

Two signal sources transmitted the RF signal with the same carrier frequency.
The faded 16-QAM modulated signals were downloaded to each source. The out-
puts of the two sources were connected to the two RF inputs of the receiver board.
The received signals through the RF inputs were time-multiplexed using the inte-
grated antenna diversity switch on the receiver board. The switch was controlled
by a function generator, where the switch speed was adjusted based on the sam-
pling rate of the modulated signal. In this case, the sampling rate was
f s = 1.6 MSPS ; therefore, the switch speed was set to f = 3.2 MHz . The re-
sw
ceiver board was programmed to down-convert the RF signal to a low intereme-
diate frequency (IF) signal with f = 3.2 MHz . The IF signal was then captured
IF
with the VSA.
Finally, an off-line process was carried out for down-conversion, de-
multiplexing, maximal ratio combining and low-pass filtering of the captured sig-
nal, in order to evaluate the performance of the proposed topology based on the
measured BER. A picture of the measurement setup is shown in Figure 10.10.
Figure 10.11 shows the measured power spectra of the received signal after time-
multiplexing and RF down-conversion, but before de-multiplexing. The signal
power spectra is split into three lobes: the main one is at f IF = 3.2 MHz , and the
± sw = 3.2 ± 1.6 MHz . This spectrum was exactly as ex-
f
two side lobes are at f
IF 2
pected from the analytical explanation in the previous section.
10.3 Single RF Front-End MIMO
M Using TDM 2777

Fig. 10.11 Measured powerr spectra of the time-multiplexed signal before baseband dde-
multiplexing

It should be noted thaat the proposed topology requires wider bandwidth thaan
conventional multiantenna receivers. This results in an extra cost; however, thhe
overall cost is reduced, as the number of RF front-ends is decreased from N in conn-
ventional multiantenna recceivers to 1 in the proposed single RF front-end receiver.
Figures 10.12(a) and 10.12(b) compare the envelopes of the RF signals and thhe
extracted baseband signaals related to each of the antennas after passing througgh
the proposed receiver top pology for SNR = 18 dB. The results show very goood
agreement between the tw wo signals, thereby proving the functionality of the proo-
posed architecture [14].
The measurement BER R of the proposed time-multiplexed single RF front-ennd
receiver is shown in Figurre 10.13. The measured BER results follow the analyticcal
results of an ideal SIMO (single
( input multiple out) system with two antennas. Thhe
small deviation between thet simulation and measurement results can be attributeed
to the unavoidable errorss between practical measurement and theoretical simulaa-
tion. For comparison, thee simulation and measurement BER results of a singlee-
input single-output (SISO O) case is also plotted in Figure 10.11. As can be seen,
having multiple antennass (in this case, two antennas) improved the BER perfoor-
mance of the communicaation systems due to the diversity gain. Moreover, Figurre
10.13 shows that the div versity gain using the proposed time-multiplexed singgle
branch receiver was equall to the diversity gain of the multiple RF front-ends. Thhis
validates the idea of usinng a time-multiplexed RF front-end in multiantenna syys-
tems. It should be noted that
t this topology requires wider bandwidth than convenn-
tional receivers. Accordinngly, it is most suitable for the single user applications oor
multi-user applications wh hen the adjacent channel interference can be neglected.
278 Chapter 10 Single RF Front-End MIMO Transceivers

Fig. 10.12 Comparing the time domain signals before and after the proposed time-
multiplexed single branch receiver: (a) antenna 1 and (b) antenna 2
10.4 Single RF Front-End MIMO Using CDM 279

16-QAM fd*T=0.03 T=1/1.6e6

-1
10

-2
10
BER [%]

-3 Simulated BER (l=1)

10
Simulated BER (l=2)
Measured BER (l=1)
Measured BER (l=2)

-4
10
2 4 6 8 10 12 14 16 18 20
SNR [in dB]

Fig. 10.13 BER comparison of the simulation and measurement for 16-QAM modulated
signal over a Rayleigh fading channel.

10.4 Single RF Front-End MIMO Using CDM

In code-division multiplexing (CDM), e.g. [15], the signals of different antennas
must be multiplied by orthogonal codes and then added together, where a single
RF front-end is used to down-convert the summed RF signals to baseband. In the
baseband section, the signals are multiplied by the orthogonal codes, integrated
and de-multiplexed. This technique is depicted in Figure 10.14.
In this method, the first signal is multiplied by the c1 (t ) code and the second
signal by the c2 (t ) code in the symbol duration and then added together. These
two codes are orthogonal, i.e. c1 (t ) is equal to 1 between 0 and Ts, and c2 (t ) is
equal to 1 between 0 and Ts/2 and to -1 between Ts/2 and Ts, where Ts is the sym-
bol duration. The signal ( h1s(t ) + n1 (t ) ) c1 (t ) + ( h2 s(t ) + n2 (t ) ) c2 (t ) is then down-
converted using a single RF front-end. The baseband signal is represented as
( h1s(t ) + n1 (t ) ) c1 (t ) + ( h2 s(t ) + n2 (t ) ) c2 (t ) .
280 Chapter 10 Single RF Front-End MIMO Transceivers

BPF ∫

Baseband
C1 (t ) Re ceiver Output
Front − End
C1 (t )

BPF ∫
C2 (t ) C2 (t )
Fig. 10.14 Realization of single-branch multiantenna receiver based on CDM

In the baseband section, the signal is multiplied by c1 (t ) and c2 (t ) and sepa-

rated into different paths. In each path, an integrator can remove the effect of the
other signal; however, the effect of the noise of both antennas is maintained in
each path.

Ts
h1 s + n = 1 / Ts ∫ [(h1 s (t ) + n1 (t ))c1 (t )c1 (t ) + (h2 s (t ) + n2 (t ))c2 (t )c1 (t )]dt
0
(10.17)
Ts Ts Ts
= 1 / Ts ∫ h1 s (t )dt +1 / Ts ∫ n1 (t )dt + 0 + 1 / Ts ∫ n2 (t )c1 (t )c2 (t )dt
0 0 0

This is the output of the integrator in the first path, where s (t ) is constant in the
symbol duration, Ts, and the codes have unit energy. The first term of the integral
is h1s and the second part of integral is the integration of the first antenna noise.
The third term is zero, because the two codes are orthogonal and s (t ) is invariable
in the symbol duration, Ts.
The fourth term of the integral is not zero, although c1 (t ) and c2 (t ) are ortho-
gonal, because n2 (t ) is stochastic and is not constant in the symbol duration. This
term has the same power as the second part. Hence, in each subchannel in base-
band, the output noise power using CDM is twice of that of a traditional design.
Indeed, the noise level is increased by 10log(N)dB in each path when N antennas
are used and the signals are down-converted with CDM method.
The simulation result for BPSK modulation using two receiving antennas is il-
lustrated in Figure 10.15. A distance of 3 dB can be observed between the conven-
tional diversity system and this technique.
An integrated circuit is designed and fabricated using CDM technique to realize
a single front-end receiver for two antennas [16]. Figure 10.16(a) illustrates the
different sections in the integrated circuit (IC) and Figure 10.16(b) shows the die
photo of the receiver in 5 GHz.
10.4 Single RF Front-End MIMO
M Using CDM 2881

0
10
Diversity = 1(Ideal)
Diversity = 2(Ideal)
Diversity = 2(CDM)

-1
10
BER

-2
10

-3
10
0 1 2 3 4 5 6 7 8 9 10
Eb/N0(dB)

Fig. 10.15 The BER versus Eb/N0 for single front-end receiver using CDM

(a)
Fig. 10.16 The 5 GHz IC for two-antenna
t single front-end receiver: a) circuits b) die photo [16]]
282 Chapter 10 Single RF Front-End MIMO Transceivers

(b)

Fig. 10.16 (continued)

10.5 Single RF Front-End MIMO Using a Parasitic Antenna

The concept of digital beamforming (DBF) antenna arrays is based on the classic
phase array method. This technique may be used in different wireless communica-
tions systems. According to Figure 10.17, signals received by individual antenna
elements are down-converted into baseband signals. These signals are digitized
and fed into the digital signal processing (DSP) chip in which the algorithms re-
side. However, RF circuit branches connected to the array elements, analog-to-
digital converters (ADCs) and the baseband DSP chip consume a considerable
amount of DC power. Furthermore, each channel connected to the array sensor has
the same structure, so the cost of fabrication increases with the number of array
elements [22], [23]. These factors limit DBF applications in the wireless commu-
nications systems. The parasitic array has been proposed to provide the solutions
for the disadvantages of DBF technique [24].
The functional block diagram of a parasitic antenna digital beamformer is
shown in Figure 10.18. In this structure, one central element is connected to the
single RF front-end, and the surrounding parasitic elements realize the array.
Beamsteering is achieved by tuning the load reactances at the parasitic elements
surrounding the central active element. The loads may be realized with the use of
varactor diodes. The power consumption of this architecture is very small. Fur-
thermore, the system has only one RF front-end, and the small inter-element spac-
ing can be as small as 0.05λ [24].
10.5 Single RF Front-End MIMO Using a Parasitic Antenna 283

Element − 1
Baseband
RF , frontend #1 W1
Element − 2

Output
RF , frontend # 2 W2 DSP
Element − n

RF , frontend # n Wn

Fig. 10.17 Block diagram of digital beam forming (DBF) antenna array

Active Ref.
Element
RF , frontend
Parasitic
Element #1 Adaptive
Tunable
Beamforming Output
Load − 1
Parasitic Algorithm
Element # n Tunable
Load − n

Baseband

Fig. 10.18 Block diagram of steerable parasitic array radiator (ESPAR) antennas

To analyze the operation of the architecture, we assume that the parasitic array
operates in the transmitting mode. The antenna generates a directional beam based
on tuning load reactances ( x1 , x2 ,", xn ) on the parasitic monopoles. Signals
transmitted from the central RF front-end excite the parasitic monopoles with sub-
stantial induced mutual currents. Vectors I and V represent the currents and the
voltages on the monopoles, respectively:

I = (i0 , i1 ,", in )T (10.18)

V = (v0 , v1 ,", vn )T (10.19)

where i0 , v0 represent the current and voltage on the central element, respectively.
284 Chapter 10 Single RF Front-End MIMO Transceivers

One may show the mutual admittances by using a Y matrix, where yij is the mu-
tual admittance between the ith element and the jth element. The induced currents
are represented with mutual admittances as follows:

⎡ y00 y01 … y0 n ⎤ ⎡ v0 ⎤
⎢y y1n ⎥⎥ ⎢⎢ # ⎥⎥ (10.20)
I = YV = ⎢ 10
⎢ # # ⎥⎢ ⎥
⎢ ⎥⎢ ⎥
⎣ yn 0 yn1 … ynn ⎦ ⎣vn ⎦

The parasitic array has a symmetrical structure; therefore:

y11 = y22 = ... = ynn

y01 = y02 = ... = y0 n

y12 = y23 = ... = y( n −1) n = yn1

y13 = y24 = ... = y( n − 2)1 = yn 2

y1 j = y2( j +1) = ...

The admittance matrix can thus be determined only by n parameters. The voltages
on the active central monopole and the mth parasitic monopole are obtained as:

v0 = vs − z0 i0 (10.21)

vm = − jxm im (10.22)

where m = 1, 2,..., n. z0 and is the characteristic impedance of 50 at the RF port,

and vs represents the transmitted voltage signal source with the amplitude and the
phase from the driving RF port at the central element. These equations can be
represented in a matrix form as:

⎡v0 ⎤
⎢0⎥
V = ⎢ ⎥ − XI = vsU − XI (10.23)
⎢#⎥
⎢ ⎥
⎣0⎦

where
⎡ z0 0 … 0 ⎤
⎢0 0 ⎥⎥
jx1 (10.24)
X =⎢
⎢# # ⎥
⎢ ⎥
⎣0 0 … jxn ⎦

U = [1,0,", 0]T
10.5 Single RF Front-End MIMO Using a Parasitic Antenna 285

The matrix induced current is then obtained as:

I = YV = Y (vsU − XI )
(10.25)
Or, similarly, we can have:

W  (Y −1 + X )−1U
(10.26)
This is called an equivalent weight vector [22].
The far-field radiation pattern is the superposition of all monopoles’ radiation
patterns. Therefore, the far-field current signal in the azimuth direction with its
amplitude, and the phase is represented as:

y (θ ) far − field = I T α (θ ) = W T α (θ )vs

(10.27)

The steering vector, α (θ ) , is defined based on the array geometry:

⎡ 1 ⎤
⎢ π ⎥
⎢ e j 2 cos(θ ) ⎥
⎢ ⎥
⎢ π 2π ⎥ (10.28)
α (θ ) = ⎢ e j 2 cos(θ − n ) ⎥
⎢ ⎥
⎢ ⎥
⎢ j π cos(θ − 2( n −1)π ) ⎥
⎢⎣e 2 n ⎥⎦

According to the reciprocity theory for radiation patterns, if the antenna is work-
ing in the receiving mode, the voltage signals, u(t), at the RF port are:

u(t ) = W T α (θ )s(t ) (10.29)

where s (t ) represents the far-field incident current waves with the amplitude and
phase in the azimuthal direction ( θ ), W is dependent on the reactance at each pa-
rasitic monopole.
The desired beam patterns can be formed by tuning the reactances
( x1 , x2 ,", xn ) . This is achieved by changing the control voltages of the diodes,
which are connected to the parasitic elements. This relation is the same form as
the beamforming equation in the array processing literature [23]. Signals received
at the antenna receiver are the sum of the weighted samples at individual antenna
sensor elements. This technique is also known as steerable parasitic array radiator
(ESPAR) antennas.
The ESPAR technique is used on the transmitter side for implementing
MIMO/MISO transmitters for BPSK and QPSK (quadrature phase-shift keying)
modulation [25]. Moreover, it has been implemented to realize a single RF front-
end MIMO system for MPSK (multiple phase-shift keying) modulation [26].
286 Chapter 10 Single RF Front-End MIMO Transceiveers

A simplified schemattic diagram of a three-element system is shown iin

Figure 10.19. As can be seen, an antenna element is connected to the RF fronnt-
end, and the other antennaas are controlled using two reactances. The idea is to maap
the spatially multiplexed signals to the spatial basis function where the desireed
beam pattern in the far fieeld becomes a combination of the individual signals. Thhe
reactive loads are used to
o create linear combination of spatial basis function. Thhe
probability of bit error based
b on Eb/No is shown in Figure 10.20 for a three-
element antenna using BP PSK modulation [27],[28].

jX L jX L

Data
Mapper
RF FrontEnd

Fig. 10.19 Block diagram off a three-element parasitic antenna MIMO transmitter [26]

Fig. 10.20 Bit error probabillity of BPSK parasitic antenna MIMO transmitter [27]
References 287

References
[1] Janaswamy, R.: Radiowave Propagation and Smart Antennas for Wireless Communi-
cations. Kluwer Academic Publishers (2001)
[2] Oestges, C., Clerckx, B.: MIMO Wireless Communications: From Real World Prop-
agation to Space Time Code Design. Academic Press (2007)
[3] Liu, J., Khaled, N., Petre, F., Bourdoux, A., Barel, A.: Impact and mitigation of mul-
tiantenna analog front-end mismatch in transmit maximum ration combining.
EURASIP Journal on Applied Signal Processing 2006, 1–14 (2006)
[4] Morris, M.L., Jensen, M.A.: Impact of receive amplifier signal coupling on MIMO
system performance. IEEE Transactions on Vehicular Technology 54, 1678–1683
(2005)
[5] Sanayei, S., Nosratinia, A.: Antenna selection in MIMO systems. IEEE Communica-
tions Magazine 42(10) (October 2004)
[6] Molisch, A., Win, M., Winters, J.: Reduced-complexity transmit/receive-diversity
systems. IEEE Transactions on Signal Processing 51(11), 2729–2738 (2003)
[7] Duman, T.M., Ghrayeb, A.: Coding for MIMO Communication Systems. Wiley
(2008)
[8] Gorokhov, A., Gore, D., Paulraj, A.: Receiver Antenna Selection for Spatial Multip-
lexing: Theory and Algorithem. IEEE Transactions on Signal Processing 51,
2796–2807 (2003)
[9] Gharavi-Alkhansari, M., Greshman, A.: Fast antenna selection in MIMO systems.
IEEE Transactions on Signal Processing 52(2), 339–347 (2004)
[10] Gucluoglu, T., Duman, T.M.: Performance of Space-Time Coded Systems with
Transmit Antenna Selection. In: Proceedings of the Conference on Information
Sciences and Systems (CISS), Baltimore, MD, pp. 863–868 (March 2007)
[11] Greshman, A., Sidiropoulos, N.D.: Space Time Processing for MIMO Communica-
tions. John Wiley & Sons (2005)
[12] Proakis, J.G., Salehi, M.: Digital Communications, 5th edn. McGraw-Hill, New York
(2008)
[13] Barry, J., Lee, E.A., Messerschmitt, D.G.: Digital Communications, 3rd edn. Kluwer
Academic Publishers (2004)
[14] Lari, M., Bassam, A., Mohammadi, A., Ghannouchi, F.M.: Time-Multiplexed Single
Front-End MIMO Receivers with Preserved Diversity Gain. IET Communica-
tions 5(6), 789–796 (2011)
[15] Jahanian, A., Tzeng, F., Heydari, P.: Code-modulated path-sharing multi-antenna re-
ceivers: theory and analysis. IEEE Transactions on Wireless Communications 8(5),
2193–2201 (2009)
[16] Tzeng, F., Jahanian, A., Pi, D., Heydari, P.: A CMOS Code-Modulated Path-Sharing
Multi-Antenna Receiver Front-End. IEEE Journal on Solid-State Circuits 44(5),
1321–1335 (2009)
[17] Golden, G.D., Foschini, G.J., Valenzuela, R.A., Wolniansky, P.W.: Detection algo-
rithm and initial laboratory results using the V-BLAST space time communication ar-
chitecture. Electronics Letters 35(1), 14–16 (1999)
[18] Xiong, F.: Digital Modulation Techniques, 2nd edn. Artech House (2006)
[19] Paulraj, A., Nabar, R., Gore, D.: Introduction to Space-Time Wireless Communica-
tions. Cambridge University Press (2003)
288 Chapter 10 Single RF Front-End MIMO Transceivers

[20] Kwan, M., Kok, C.: FIR zero-forcing equalizer for ISI MIMO channels using space-
time modulated codes. IEEE Transactions on Wireless Communications 5, 123–132
(2006)
[21] Wang, C., Au, E.K.S., Murch, R.D., Mow, W.H., Cheng, R.S., Lau, V.: On the Per-
formance of the MIMO Zero-Forcing Receiver in the Presence of Channel Estimation
Error. IEEE Transactions on Wireless Communications 6, 805–810 (2007)
[22] Sun, C., Hirata, A., Ohira, T., Karmakar, N.C.: Fast Beamforming of Elec-tronically
Steerable Parasitic Array Radiator Antennas: Theory and Experiment. IEEE Transac-
tions on Antennas & Propagation 52(7), 1819–1832 (2004)
[23] Johnson, D.H., Dudgeon, D.E.: Array Signal Processing: Concepts and Techniques,
pp. 91–365. Prentice-Hall, Englewood Cliffs (1993)
[24] Bains, R., Muller, R.: Using parasitic elements for implementing the rotat-ing antenna
for MIMO receivers. IEEE Transactions on Wireless Communications 7(11),
4522–4533 (2008)
[25] Kalis, A., Kanatas, A.G., Papadias, C.B.: A novel approach to MIMO transmission
using a single RF front end. IEEE Journal on Selected Areas in Communica-
tions 26(6) (August 2008)
[26] Alrabadi, O.N., Papadias, C.B., Kalis, A., Prasad, R.: A Universal Encoding Scheme
for MIMO Transmission Using a Single Active Element for PSK Modulation
Schemes. IEEE Transactions on Wireless Communications 8(10), 5133–5143 (2009)
[27] Alrabadi, O.N., Divarathne, C., Tragas, P., Kalis, A., Marchetti, N., Kalis, A.,
Papadias, C.B., Prasad, R.: Spatial Multiplexing with a Single Radio: Proof-of-
Concept Experiments in an Indoor Environment with a 2.6-GHz Proto-type. IEEE
Communications Letters (2011)
[28] Mohammadi, A., Ghannouchi, F.M.: Single RF Front-End MIMO Transceivers. IEEE
Communications Magazine (December 2011)
Index

1 Class A PA 90 coupling 265

adaptive array 158 covariance matrix 165
adaptive modulation 233 cross Modulation 60
additive white Gaussian noise crosstalk 141
(AWGN) 229 cyclic Prefix 48
adjacent channel power ratio cyclic prefix (CP) 47
(ACPR) 62
alamouti 18 DC offset 7, 147
AM/AM conversion 61 degradation factor (DF) 207
AM/PM conversion 61 degrees of freedom (DOF) 267
analog Predistortion Linearization 116 delta-sigma () modulation 129
analog-to-digital conwverters delta-sigma modulators (DSMs) 125
(ADCs) 69 digital beamforming (DBF) 282
antenna selection 267 digital phase lock loop (DPLL) 120
application specific integrated circuits digital predistortion (DPD) 114
(ASICs) 3 digital signal processors (DSPs) 3
automatic gain control (AGC) 109, 152 digital Video and Audio
average error probability 31 Broadcasting 181
average spectral efficiency (ASE) 233 digital-to-analog converters (DACs) 69
direct Conversion Transmitter 131
back-off 82 distortion. 273
bandwidth efficiency 6 diversity 9
beamforming 161 diversity order 4
bell layered space-time (BLAST) 21 diversity-multiplexing tradeoff
(DMT) 21
capacity reduction factor 30 down conversion 3
channel matrix 11 drain efficiency 79
channel state information (CSI) 15, 159 DUT 115
class AB 95 dynamic DC errors 240
class B PA 92 dynamic Range 68
class D current-switching 100
class D PA 100 eigenvalues 248
class D voltage-switching 102 envelope elimination and restoration
class E PA 105 (EER) 124
class F 108 equal gain combining 1
coherence bandwidth 45 equalization 181
commercial 173 ergodic 267
compensation 141 error vector magnitude (EVM) 55
compensation techniques 7, 227
constellation sizes 238 feedback Linearization 110
COST 259 23 feedforward Linearization 115
COST 273 23 FFT 47
290 Index

field-effect transistors (FETs) 79 matched filter 274

field-programmable gate arrays maximal ratio combining 1
(FPGAs) 3 maximum likelihood (ML) 17
five-Port Receiver 156 memory Effect 86
flat-fading 255 memory polynomial 115
FPGA 169 MFSK (multiple frequency-shift keying)
frequency selective channel 49 30
frequency synthesizing 3 middleton 169
frequency-division multiplexing millimeter-wave 164
(FDM) 45 MIMO testbed 168
MIMO-OFDM 6
guassian minimum-shift keying modulation constellation 190
(GMSK) 30 monte Carlo 249
MPSK (M-ary phase-shift keying) 30
hartley 153 MPSK-MIMO 43
heterodyne technique 3 MQAM (M-ary quadrature amplitude
high-definition (HD) 159 modulation) 30
MQAM-OFDM 182
I/Q imbalance 7 multicarrier modulation 45
I/Q mixer 132 multipath fading 11
IEEE 802.11a/g 4 multiple antenna 265
IEEE 802.11n 4 multiplexing 9, 269
IEEE 802.15 159 mutual information 267
IEEE 802.16e 4
IEEE 802.16m 4 noise factor matrices 165
IFFT 45 noise figure 64
image Rejection Receiver 153 normalized in complete gamma
image signal 153 function 267
impairment 163
interference 166 OBOs 214
intermediate frequency (IF) 131 OFDM 7
inter-subchannel interference (ICI) 52 one-dB Compression Point 57
intersymbol interference (ISI) 47 -order intermodulation distortion
inverse Class F PAs 106 (IMD3) 73
ISI 47 orthogonal frequency-division
multiplexing (OFDM) 6
laguerre polynomial 32, 232 outage probability 267
layer space-time (LAST) 19
layered space-time 19 parallel streams 181
LCD interference 166 parasitic antenna 283
linear amplification with nonlinear path loss 11
components (LINC) 120 peak-to-average power ratio
linear time-invariant (LTI) 55 (PAPR) 82
linearization 111 performance degradation 207
link budget 67 phase lock loop (PLL) 110
local oscillator (LO) 131 phase noise 7
Long Term Evolution (LTE) 5 phase-shift keying (PSK) 181
lorentzian function 207 phase-shift keying (QPSK) 40
lorenzian 50 polar Transmitter Architecture 120
low IF Receiver 152 polynomial coefficients 214
low noise amplifiers (LNAs) 3 power efficiency 6
low-pass equivalent 219 power-added efficiency (PAE) 80
Index 291

probability density function (pdf) 248 steerable parasitic array radiator

prototype 168 (ESPAR) 286
subcarrier 185
quadrature amplitude modulation superheterodyne method 39
(QAM) 181 superhetrodyne Transmitter 131
quality factor 70 switching-Mode PAs 98
quality of service (QoS) 247 symbol error rate (SER) 40
synthesizer 162
raised cosine filter 275
rayleigh fading 12, 31 tap-delay lines 23
receiver diversity 1 temporal dynamics 218
relative constellation error (RCE) 74 Terrestrial Trunked Radio (TETRA) 63
reliability 9 thermal effects 218
RF impairments 7, 152 thermal noise 66
RF Impairments, 227 thevenin equivalent 66
RF switch 269 third-Order Intercept Point 58
RFI 166 throughput 248
root mean square (RMS) 84 -to-average power ratio (PAPR) 50
traveling wave tube amplifiers
SAW filters 175 (TWTA) 61
second-Order Intercept Point 60 truncated power series 202
selection combining 1
self-mixing 240 unitary matrices 231
sensitivity 67
shadowing 11 variable-rate variable-power
shannon capacity 30 (VRVP) 233
singular value decomposition V-BLAST 168
(SVD) 230 voltage controlled oscillator
six-port junction 154 (VCO) 131
six-Port MIMO 164 voltage-gain amplifier (VGA) 275
six-Port Receiver 154 volterra series 218
smart Antenna 157
SNR degradation 218 WCDMA 71
solid-state power amplifiers weaver 153
(SSPAs) 116 wideband CDMA 114
space-time block code (STBC) 21 wide-sense stationary uncorrelated
space-time block codes 18 scattering homogeneous
space-time codes 17 (WSSUSH) 12
space-time trellis codes 18 Wi-Fi 200
spatial multiplexing 4, 9 wireless local area network
spectral efficiency 27 (WLAN) 166
spectral Regrowth 62 world Interoperability for Microwave
spurious 152 Access (WiMAX) 4
spurious free dynamic range
(SFDR) 68 zero-forcing 274
Stanford University Interim (SUI) 23 zero-IF 175
About the Authors

Abbas Mohammadi is a professor at Electrical Engineering Department of

Amirkabir University of Technology where he is head of Radio Communications
center of excellence and Director of Microwave/Millimeter-wave and Wireless
Communications Research Lab. He has held an invited professor position at the
University of Calgary. Prof. Mohammadi authored two books, awarded four pat-
ents, and published more than 150 peer-reviewed papers. His research interests are
in the areas of microwave and wireless communications, software defined radio,
adaptive transceivers and MIMO systems. He is a senior member of Institute of
Electrical and Electronics Engineers (IEEE).
Fadhel M. Ghannouchi, a Senior Canada Research Chair in Intelligent RF Radio
Technology, iCORE Professor, founder and Director of iRadio Laboratory, Uni-
versity of Calgary is regarded as one of the foremost authorities in the fields of RF
Electronics and Radio technology. His current research interests are in their appli-
cations in broadband wireless, satellite and biomedical sciences. Other recogni-
tions received by Prof. Ghannouchi include: The Alberta Professional Engineers
Geologists and Geophysicist (APEGGA) Alberta Ingenuity Fund Research Excel-
lence Award, The Research Excellence Award of Schulich school of Engineering,
University of Calgary and, being elected Distinguished Lecturer of the IEEE Mi-
crowave Theory and Technique. For his scientific and technical accomplishments,
Dr. Ghannouchi’s peers have elected him to Fellow in The Canadian Academy of
Engineering, Fellow of The Engineering Institute of Canada, Fellow of The Insti-
tution of Engineering and Technology (IET formally IEE), and Fellow of Institute
of Electrical and Electronics Engineers (IEEE). Prof. Ghannouchi held several in-
vited positions at several academic and research institutions in Europe, North
America and Japan. He has provided consulting services to a number of micro-
wave and wireless communications companies worldwide. His research interests
are in the areas of microwave instrumentation and measurements, nonlinear mod-
eling of microwave devices and communications systems, design of power and
spectrum efficient microwave amplification systems, design of intelligent RF
transceivers for wireless and satellite communications and software defined and
cognitive Radio systems. Prof. Ghannouchi has published extensively. He has au-
thored two books, was awarded 7 patents, and published more that 400 peer-
reviewed articles, given over 50 invited/keynote/plenary conference presentations,
trained over 60 Master and PhD graduates and founded three Spin-off companies.