UNIVERSITY OF CALIFORNIA, SAN DIEGO

Constant Envelope OFDM Phase Modulation

A dissertation submitted in partial satisfaction of the

requirements for the degree
Doctor of Philosophy

in

Electrical Engineering (Communications Theory and Systems)

by

Steve C. Thompson

Committee in charge:

Professor James R. Zeidler, Chair

Professor John G. Proakis, Co-Chair
Professor Robert R. Bitmead
Professor William S. Hodgkiss
Professor Laurence B. Milstein

2005
 Copyright

Steve C. Thompson, 2005

All rights reserved.

The dissertation of Steve C. Thompson is ap-
proved, and it is acceptable in quality and form
for publication on microfilm:

Co-Chair

Chair

University of California, San Diego

2005

iii
 “Before PhD,
I chopped wood and carried water;
After PhD,
I chopped wood and carried water.”
—[Slightly modified] Zen saying

“I wish I could be more moderate in my desires. But I can’t, so there is no rest.”

—John Muir, 1826

“I know this: a man got to do what he got to do. . . ”

—Casy, The Grapes of Wrath, John Steinbeck, 1939

iv
 TABLE OF CONTENTS

Signature Page . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii

Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v

List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii

List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii

Vita and Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiv

Abstract of the Dissertation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvi

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 An Introduction to OFDM . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.1.1 ISI-Free Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.1.2 A Multicarrier Modulation . . . . . . . . . . . . . . . . . . . . . . 5
1.1.3 Discrete-Time Signal Processing . . . . . . . . . . . . . . . . . . . 8
1.2 Problems with OFDM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 Constant Envelope Waveforms . . . . . . . . . . . . . . . . . . . . . . . . 12
1.4 Constant Envelope OFDM . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.5 Thesis Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2 OFDM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1 More OFDM Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.1.1 The Cyclic Prefix . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.1.2 Discrete-Time Model . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.1.3 Block Modulation with FDE . . . . . . . . . . . . . . . . . . . . . 20
2.1.4 System Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.2 PAPR Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3 Power Amplifier Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.4 Effects of Nonlinear Power Amplification . . . . . . . . . . . . . . . . . . . 30

v
 2.4.1 Spectral Leakage . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4.2 Performance Degradation . . . . . . . . . . . . . . . . . . . . . . . 32
2.4.3 System Range and PA Efficiency . . . . . . . . . . . . . . . . . . . 35
2.5 PAPR Mitigation Techniques . . . . . . . . . . . . . . . . . . . . . . . . . 37

3 Constant Envelope OFDM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

3.1 Signal Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.2 Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

4 Performance of Constant Envelope OFDM in AWGN . . . . . . . . . . . . . . . 58

4.1 The Phase Demodulator Receiver . . . . . . . . . . . . . . . . . . . . . . . 59
4.1.1 Performance Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.1.2 Effect of Channel Phase Offset . . . . . . . . . . . . . . . . . . . . 65
4.1.3 Carrier-to-Noise Ratio and Thresholding Effects . . . . . . . . . . 66
4.1.4 FIR Filter Design . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.2 The Optimum Receiver . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.2.1 Performance Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.2.2 Asymptotic Properties . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.3 Phase Demodulator Receiver versus Optimum . . . . . . . . . . . . . . . . 78
4.4 Spectral Efficiency versus Performance . . . . . . . . . . . . . . . . . . . . 80
4.5 CE-OFDM versus OFDM . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

5 Performance of CE-OFDM in Frequency-Nonselective Fading Channels . . . . . 86

6 Performance of CE-OFDM in Frequency-Selective Channels . . . . . . . . . . . 94

6.1 MMSE versus ZF Equalization . . . . . . . . . . . . . . . . . . . . . . . . 96
6.1.1 Channel Description . . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.1.2 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
6.1.3 Discussion and Observations . . . . . . . . . . . . . . . . . . . . . 103
6.2 Performance Over Frequency-Selective Fading Channels . . . . . . . . . . 108
6.2.1 Channel Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
6.2.2 Simulation Procedure and Preliminary Discussion . . . . . . . . . 112
6.2.3 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

vi
7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

A Generating Real-Valued OFDM Signals with the Discrete Fourier Transform . . 124
A.1 Signal Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
A.2 Spectral Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126

B More on the OFDM Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

C Sample Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134

C.1 GNU Octave Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
C.2 Gnuplot Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

Production Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194

vii
 LIST OF FIGURES

1.1 Representation of a wireless channel with multipath. . . . . . . . . . . . . 2

1.2 A wireless channel in time and frequency. . . . . . . . . . . . . . . . . . . 2
1.3 Intersymbol interference. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 OFDM with cyclic prefix (CP). . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5 Subcarrier and overall spectrum. (N = 16; |I 0,k | = 1, for all k) . . . . . . 7
1.6 OFDM converts wideband channel to N narrowband frequency bins. . . . 8
1.7 Frequency offset causes ICI. (fo = 0.25) . . . . . . . . . . . . . . . . . . . 9
1.8 A typical OFDM signal (N = 16). The PAPR is 9.5 dB. . . . . . . . . . . 10
1.9 Power amplifier transfer function. . . . . . . . . . . . . . . . . . . . . . . . 11
1.10 Comparison of OFDM and CE-OFDM signals. . . . . . . . . . . . . . . . 13

2.1 Sampling instances. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

2.2 Circular convolution with channel and the inverse channel. . . . . . . . . . 21
2.3 Block modulation with cyclic prefix and FDE. . . . . . . . . . . . . . . . . 21
2.4 OFDM is a special case. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.5 OFDM system diagram. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.6 Complementary cumulative distribution functions. (N = 64) . . . . . . . 25
2.7 PAPR CCDF lower bound (2.31) for N = 2 k , k = 5, 6, . . . , 10. . . . . . . . 26
2.8 AM/AM (solid) and AM/PM (dash) conversions (SSPA=thick, TWTA=thin)
for various backoff ratios K. . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.9 Fractional out-of-band power of OFDM with ideal PA and with TWTA
model at various input power backoff. (N = 64, IBO in dB) . . . . . . . . 31
2.10 Spectral growth versus IBO. (N = 64) . . . . . . . . . . . . . . . . . . . . 31
2.11 Performance of QPSK/OFDM with nonlinear power amplifier with various
input power backoff levels. (N = 64) . . . . . . . . . . . . . . . . . . . . . 33
2.12 Performance of M -PSK/OFDM with SSPA. (N = 64) . . . . . . . . . . . 34
2.13 The potential range of system is reduced with input backoff; the range is
reduced further from nonlinear amplifier distortion. . . . . . . . . . . . . . 36
2.14 Power amplifier efficiency. . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.15 Block diagram. The system is evaluated with and without PAPR reduction. 38
2.16 Unclipped OFDM signal (9.25 dB PAPR). The rings have radius A max
which correspond to various clipping ratios γ clip (dB). . . . . . . . . . . . 39

viii
2.17 PAPR CCDF of clipped OFDM signal for various γ clip (dB). [N = 64] . . 40
2.18 PAPR of clipped signal as a function of the clipping ratio. (N = 64) . . . 40
2.19 A comparison of the total degradation curves of clipped and unclipped
M -PSK/OFDM systems. (N = 64) . . . . . . . . . . . . . . . . . . . . . . 41

3.1 The CE-OFDM waveform mapping. . . . . . . . . . . . . . . . . . . . . . 43

3.2 Instantaneous signal power. . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.3 Basic concept of CE-OFDM. . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.4 Phase discontinuities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.5 Continuous phase CE-OFDM signal samples, over L blocks, on the com-
plex plane. (2πh = 0.7) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.6 Estimated fractional out-of-band power. (N = 64) . . . . . . . . . . . . . 52
3.7 Double-sided bandwidth as a function of modulation index. (N = 64) . . 53
3.8 Power density spectrum. (N = 64, 2πh = 0.6) . . . . . . . . . . . . . . . . 54
3.9 Fractional out-of-band power. (N = 64, 2πh = 0.6) . . . . . . . . . . . . . 55
3.10 CE-OFDM versus OFDM. (N = 64) . . . . . . . . . . . . . . . . . . . . . 56
3.11 CE-OFDM versus OFDM with nonlinear PA. (N = 64) . . . . . . . . . . 57

4.1 Phase demodulator receiver. . . . . . . . . . . . . . . . . . . . . . . . . . . 58

4.2 Bandpass to baseband conversion. . . . . . . . . . . . . . . . . . . . . . . 60
4.3 Discrete-time phase demodulator. . . . . . . . . . . . . . . . . . . . . . . . 62
4.4 Performance with and without phase offsets. System 1 (S1) has phase
offsets {(θi + φ0 ) ∈ [0, 2π)}, and System 2 (S2) doesn’t (θ i + φ0 = 0).
[M = 2, N = 64, J = 8] . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.5 Threshold effect at low CNR. (M = 8, N = 64, J = 8, 2πh = 0.5) . . . . . 68
4.6 Threshold effect at low CNR, various 2πh. (M = 8, N = 64, J = 8) . . . 68
4.7 Performance for various filter parameters L fir , fcut /W .
(M = 2, N = 64, J = 8, 2πh = 0.5 and Eb /N0 = 10 dB) . . . . . . . . . . 69
4.8 Magnitude response of various Hamming FIR filters. . . . . . . . . . . . . 70
4.9 CE-OFDM performance with and without FIR filter.
(M = 2, N = 64, J = 8) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.10 The optimum receiver. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.11 Correlation functions ρm,n (K). . . . . . . . . . . . . . . . . . . . . . . . . 76
4.12 CE-OFDM optimum receiver performance. (M = 2, N = 8) . . . . . . . . 77

ix
4.13 All unique ρm,n (K) for M = 2, N = 4 DCT modulation. . . . . . . . . . . 78
4.14 Phase demodulator receiver versus optimum. (N = 64) . . . . . . . . . . . 79
4.15 Noise samples PDF versus Gaussian PDF. (E b /N0 = 30 dB) . . . . . . . . 80
4.16 Performance of M -PAM CE-OFDM. (N = 64, †=leftmost curve, ‡=rightmost
curve) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.17 Spectral efficiency versus performance. . . . . . . . . . . . . . . . . . . . . 82
4.18 A comparison of CE-OFDM and conventional OFDM. (M = 2, N = 64) . 85

5.1 Performance of CE-OFDM in flat fading channels. (N = 64) . . . . . . . 88

5.2 A simplified two-region model. (M = 8, N = 64, 2πh = 0.6) . . . . . . . . 90
5.3 A (n + 1)-region model. (M = 8, N = 64, 2πh = 0.6) . . . . . . . . . . . . 91
5.4 Performance of CE-OFDM in flat fading channels. (Circle=Rayleigh;
square=Rice, K = 3 dB; triangle=Rice, K = 10 dB. Solid line=Semi-
analytical curve, (5.15); points=simulation. N = 64) . . . . . . . . . . . . 92
5.5 Comparison of semi-analytical technique (5.15) with (5.10) and (5.11).
(M = 4, N = 64, 2πh = 1.2) . . . . . . . . . . . . . . . . . . . . . . . . . . 93

6.1 CE-OFDM system with frequency-selective channel. . . . . . . . . . . . . 96

6.2 Channel D. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
6.3 Channel A results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
6.4 Channel B results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
6.5 Channel C results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.6 Channel D results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.7 Channel E results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
6.8 Channel F results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
6.9 Fundamental characteristic functions and quantities [(6.21)–(6.25)] of the
four channel models considered. . . . . . . . . . . . . . . . . . . . . . . . . 113
6.10 Performance results. (Multipath results are labeled with circle and trian-
gle points; the Rayleigh, L = 1 result is that of the frequency-nonselective
channel model. M = 4, N = 64, 2πh = 1.0) . . . . . . . . . . . . . . . . . 115
6.11 Single path versus multipath. (M = 4, N = 64, Channel C f , MMSE) . . . 119
6.12 CE-OFDM versus QPSK/OFDM. (SSPA model, Channel C f , N = 64,
MMSE) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

B.1 “OFDM” search on IEEE Xplore [222]. . . . . . . . . . . . . . . . . . . . . 130

B.2 Papers, filed and piled. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

x
B.3 Running average of papers read per day. . . . . . . . . . . . . . . . . . . . 132
B.4 Year histogram. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
B.5 Projected year histogram? . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

xi
 LIST OF TABLES

6.1 Channel samples of frequency-selective channels. . . . . . . . . . . . . . . 97

6.2 Channel model parameters. . . . . . . . . . . . . . . . . . . . . . . . . . . 111

6.3 Data symbol contribution per tone for m n (t), n =1, 2, and 3. . . . . . . . 118

xii
 ACKNOWLEDGEMENTS

I want to first thank my advisors, Professors Zeidler and Proakis, for giving me the
chance to do this work, for the encouragement, and for the guidance. I want to thank
Professor Milstein for the many helpful technical conversations and for his many sug-
gestions. Thanks to Professors Bitmead and Hodgkiss for taking the time to participate
as committee members. Also, thanks to Professor Proakis for carefully proofreading the
draft manuscripts of this thesis.

Thanks to UCSD’s Center for Wireless Communications for providing a good en-
vironment for conducting research; thanks to its industrial partners for the financial
support.

Thanks to my wife, Shannon, for the emotional and caloric support. Thanks to
Chaney the cat for waking me up in the morning. Thanks to my friends for fun support.
Thanks to my fellow graduate students in Professor Zeidler’s research group for the
camaraderie. Special thanks to Ahsen Ahmed for helpful collaboration over the past
couple years. Thanks to my family. Also, thanks to Karol Previte for her support early
in my graduate student existence.

Thanks to my teachers: Professors Duman, Masry, Milstein, Pheanis, and Wolf, to

name only a few.

Finally, I would like to thank the countless developers, documentation writers, bug
reporters, and users of the free software I’ve benefited from during the course of my PhD.

The text in this thesis, in part, was originally published in the following papers, of
which I was the primary researcher and author: S. C. Thompson, J. G. Proakis, and
J. R. Zeidler, “Constant Envelope Binary OFDM Phase Modulation,” in Proc. IEEE
Milcom, vol. 1, Boston, Oct. 2003, pp. 621–626; S. C. Thompson, A. U. Ahmed, J.
G. Proakis, and J. R. Zeidler, “Constant Envelope OFDM Phase Modulation: Spectral
Containment, Signal Space Properties and Performance,” in Proc. IEEE Milcom, vol. 2,
Monterey, Oct. 2004, pp. 1129–1135; S. C. Thompson, J. G. Proakis, and J. R. Zeidler,
“Noncoherent Reception of Constant Envelope OFDM in Flat Fading Channels,” in Proc.
IEEE PIMRC, Berlin, Sept. 2005; and S. C. Thompson, J. G. Proakis, and J. R. Zeidler,
“The Effectiveness of Signal Clipping for PAPR Reduction and Total Degradation in
OFDM Systems,” in Proc. IEEE Globecom, St. Louis, Dec. 2005.

xiii
 VITA

December 22, 1976 Born, Mesa, Arizona

1997–1998 Associate Engineer

Inter-Tel, Chandler, Arizona

Summer 1998 Summer Internship

Los Alamos National Laboratory
Los Alamos, New Mexico

1999 BSc in Electrical Engineering

Arizona State University, Tempe, Arizona

Summer 2001 Summer Internship

SPAWAR Systems Center, San Diego, California

2001 MSc in Electrical Engineering

University of California at San Diego, La Jolla, California

2001–2005 Research Assistant

Center for Wireless Communications
University of California at San Diego, La Jolla, California

Summer 2004 Summer Internship

SPAWAR Systems Center, San Diego, California

2005 PhD in Electrical Engineering

University of California at San Diego, La Jolla, California

PUBLICATIONS

S. C. Thompson, J. G. Proakis, and J. R. Zeidler, “Constant Envelope Binary OFDM

Phase Modulation,” in Proc. IEEE Milcom, vol. 1, Boston, Oct. 2003, pp. 621–626.

S. C. Thompson, A. U. Ahmed, J. G. Proakis, and J. R. Zeidler, “Constant Envelope

OFDM Phase Modulation: Spectral Containment, Signal Space Properties and Perfor-
mance,” in Proc. IEEE Milcom, vol. 2, Monterey, Oct. 2004, pp. 1129–1135.

S. C. Thompson, A. U. Ahmed, J. G. Proakis, and J. R. Zeidler, “Constant Envelope

OFDM Phase Modulation,” submitted to IEEE Transactions on Communications.

S. C. Thompson, J. G. Proakis, and J. R. Zeidler, “Noncoherent Reception of Constant

Envelope OFDM in Flat Fading Channels,” in Proc. IEEE PIMRC, Berlin, Sept. 2005.

S. C. Thompson, J. G. Proakis, and J. R. Zeidler, “The Effectiveness of Signal Clip-

ping for PAPR Reduction and Total Degradation in OFDM Systems,” in Proc. IEEE
Globecom, St. Louis, Dec. 2005.

xiv
S. C. Thompson, J. G. Proakis, and J. R. Zeidler, “The Effectiveness of Signal Clipping
for PAPR Reduction and Total Degradation in OFDM Systems,” in preparation.

S. C. Thompson, A. U. Ahmed, J. G. Proakis, and J. R. Zeidler, M -ary PAM Constant

Envelope OFDM,” in preparation.

S. C. Thompson, J. G. Proakis, and J. R. Zeidler, “Performance of CE-OFDM in

Frequency-Nonselective Fading Channels,” in preparation.

S. C. Thompson, J. G. Proakis, and J. R. Zeidler, “Performance of CE-OFDM in

Frequency-Selective Channels,” in preparation.

xv
 ABSTRACT OF THE DISSERTATION

Constant Envelope OFDM Phase Modulation

by

Steve C. Thompson

Doctor of Philosophy in Electrical Engineering (Communications Theory and

Systems)

University of California San Diego, 2005

Professor James R. Zeidler, Chair

Professor John G. Proakis, Co-Chair

Orthogonal frequency division multiplexing (OFDM) is a popular modulation technique

for wireless digital communications. It provides a relatively straightforward way to ac-
commodate high data rate links over harsh wireless channels characterized by severe
multipath fading. OFDM has two primary drawbacks, however. The first is a high sen-
sitivity to time variations in the channel caused by Doppler, carrier frequency offsets,
and phase noise. The second, and the focus of this thesis, is that the OFDM waveform
has high amplitude fluctuations, a drawback known as the peak-to-average power ratio
(PAPR) problem. The high PAPR makes OFDM sensitive to nonlinear distortion caused
by the transmitter’s power amplifier (PA). Without sufficient power backoff, the system
suffers from spectral broadening, intermodulation distortion, and, consequently, perfor-
mance degradation. High levels of backoff reduce the efficiency of the PA. For mobile
battery-powered devices this is a particularly detrimental problem due to limited power
resources.

A new PAPR mitigation technique is presented. In constant envelope OFDM (CE-

OFDM), the high PAPR OFDM signal is transformed to a constant envelope 0 dB PAPR
waveform by way of angle modulation. The constant envelope signal can be efficiently
amplified with nonlinear power amplifiers thus achieving greater power efficiency. In

xvi
this thesis, the fundamental aspects of the CE-OFDM modulation are studied, including
the signal spectrum, the signal space, optimum performance, and the performance of
a practical phase demodulator receiver. Performance is evaluated over a wide range of
multipath fading channel models. It is shown that CE-OFDM outperforms conventional
OFDM when taking into account the effects of the power amplifier.

This work was done at UCSD’s Center for Wireless Communication, under the “Mo-
bile OFDM Communications” project (CoRe research grant 00-10071).

xvii
Chapter 1

Introduction

Humans have always found ways to communicate, over space and over time. From
the messenger pigeon to the Pony Express, from the message in a bottle to cave drawings,
smoke signals and beacons, people have used inventive techniques, techniques derived
from their natural environment, to share information. A particularly good natural re-
source for communication is electricity for its speed and ability to be controlled with
devices like capacitors, microprocessors, electronic memory storage and batteries. Com-
munication was profoundly enhanced with Morse’s telegraph (1837), Bell’s telephone
(1876), Edison’s phonograph (1887), and Marconi’s radio (1896). From these early inven-
tions, communications technology has advanced with global telephone networks, satellite
communications, and magnetic storage systems; and with the rise of the internet and
digital computers, digital communications—the transfer of bits (1’s and 0’s) from one
point to another—has become important.

In particular, wireless digital communications is currently under intensive research,

development and deployment to provide high data rate access plus mobility. One chal-
lenge in designing a wireless system is to overcome the effects of the wireless channel,
which is characterized as having multiple transmission paths and as being time vary-
ing [421, 427]. Figure 1.1 illustrates a link with four reflecting paths between points A
and B. These reflections are caused by physical objects in the environment. Due to the
relative mobility between the points and the possibility that the reflecting objects are
mobile, the channel changes with time.

1
 2

point B
point A


Propagation paths

Figure 1.1: Representation of a wireless channel with multipath.

An example profile of the channel in Figure 1.1 is shown in Figure 1.2(a). Each path
has its own associated delay and power. The first path arrives at the receiver 0.5 µs after
the signal is transmitted; the last path arrives with a 14 µs delay. The Fourier transform
of the profile yields the frequency-domain representation shown in Figure 1.2(b). The
channel is viewed over a 2 MHz range centered at the center frequency f c . Notice that
the channel power fluctuates by 30 dB (a factor of 1000) over the frequency range. The
dispersion in the time domain leads to frequency-selectivity in the frequency domain.

1 5
0
Channel power (dB)

-5
Path power

-10
0.1
-15
-20
-25
0.01
-30−1 −0.5
0 2 4 6 8 10 12 14 0 0.5 1
Time (µs) Frequency, f − fc (MHz)

(a) Time domain. (b) Frequency domain.

Figure 1.2: A wireless channel in time and frequency.

In general, a digital communication system maps bits to k b -bit data symbols. In a

conventional single carrier system, the symbols are then transmitted serially. The signal
waveform of such a system is
X
s(t) = Ii g(t − iTs ), (1.1)
i

where t is the time variable, {Ii } are the data symbols, Ts is the symbol period, and g(t)
is a transmit pulse shape. For time-dispersive channels, such as the 4-path example in
 3

Figure 1.2, interference is caused from symbol to symbol. This intersymbol interference
(ISI) is illustrated in Figure 1.3. For simplicity, g(t) is rectangular. The channel is
represented by its time-variant impulse response h(τ, t), where τ is a propagation delay
variable. The received signal is expressed mathematically as [387, p. 97]

r(t) = s(t) ∗ h(τ, t) + n(t)

Z ∞ (1.2)
= h(τ, t)s(t − τ )dτ + n(t),
−∞
where ∗ represents the linear convolution operator and n(t) is additive noise. The effect of
the time-dispersive channel is shown to smear symbol 1 into symbol 2, therefore creating
intersymbol interference.
s(t) r(t)
Transmitter Channel Receiver

s(t) |h(τ, t)| r(t)

ISI
. .. symbol 1 symbol 2 ...
0 Ts 2Ts t τ 0 Ts 2Ts t

Figure 1.3: Intersymbol interference.

The severity of the ISI depends on the symbol period relative to the channel’s max-
imum propagation delay, τmax . Consider transmitting the signal in (1.1) over the 2
MHz channel in Figure 1.2. The signal bandwidth is roughly proportional to the sym-
bol rate 1/Ts Hz. Therefore making s(t) a 2 MHz signal, T s = (2 × 106 )−1 = 0.5 µs.
Since the maximum propagation delay of the channel is τ max = 14 µs, the ISI spans
τmax /Ts = (14 µs) / (0.5 µs) = 28 symbols. (For comparison, the ISI in Figure 1.3 spans
less than one symbol.) Such severe ISI must be corrected at the receiver in order to
provide reliable communication.

The traditional approach to combating intersymbol interference is with time-domain

equalizers [421]. There are many types, ranging in complexity and in effectiveness.
The optimum maximum-likelihood (ML) receiver is the most effective but is typically
impractical due to its high complexity, which grows exponentially with the ISI length.
Linear equalizers are much simpler, having a complexity which grows roughly linearly
with ISI length, but perform much worse than the optimum receiver. Nonlinear decision
feedback equalizers (DFEs) have similar complexity as the linear type and have better
performance.
 4

All of these techniques require knowledge of the channel, which is estimated by

transmitting a training sequence which is known at the receiver. Then by comparing
the received signal to what was transmitted, an estimate of h(τ, t) is made. There are
various algorithms available for the estimation process, each having its own complexity,
convergence rate, and stability. The least-mean-square (LMS) algorithm is the most
stable and the least complex, but suffers from a slow convergence rate. The recursive
least-square (RLS or Kalman) algorithm, on the other hand, converges quickly, but has
higher complexity and can be unstable.

For scenarios like the example above with an ISI spanning 28 symbols, conventional
equalization becomes difficult. Training times become long and convergence of the chan-
nel estimator is problematic, especially for time-varying channels. In the example, 2×10 6
symbols/s are transmitted. Using a QPSK (quadrature phase-shift keying) signal con-
stellation, which maps kb = 2 bits per symbol, the bit rate is 4 Mb/s. Such a bit rate is
desired in current wireless systems, and in many cases demand for many tens of Mb/s
is common.

1.1 An Introduction to OFDM

To meet the demanding data rate requirements, alternative techniques have been
considered. One approach, orthogonal frequency division multiplexing, has become ex-
ceedingly popular. OFDM has been implemented in wireline applications such as digital
subscriber lines (DSL) [95], in wireless broadcast applications such as digital audio and
video broadcasting (DAB and DVB) and in-band on-channel (IBOC) broadcasting [392].
It has been used in wireless local area networks (LANs) under the IEEE 802.11 and the
ETSI HYPERLAN/2 standards [552]. OFDM is being developed for ultra-wideband
(UWB) systems; cellular systems; wireless metropolitan area networks (MANs), under
the IEEE 802.16 (WiMax) standard; and for other wireline systems such as power line
communication (PLC) [119, 160, 264, 604].

1.1.1 ISI-Free Operation

OFDM’s main appeal is that it supports high data rate links without requiring
conventional equalization techniques. Instead of transmitting symbols serially, OFDM
 5

sends N symbols as a block. The OFDM block period, T B , is thus N times longer than
the symbol period. Continuing the example above, and choosing N = 300, the block
period is TB = N Ts = 300 × 0.5 µs = 150 µs, which is more than 10 times the duration
of the channel’s impulse response. ISI is avoided by inserting a guard interval between
successive blocks during which a cyclic prefix is transmitted. The interval duration, T g ,
is designed such that Tg ≥ τmax so that the channel is absorbed in the guard interval
and the OFDM block is uncorrupted. This is illustrated in the figure below. Selecting a
guard interval Tg = 15 µs for the channel in Figure 1.2 results in a transmission efficiency
ηt = TB /(TB + Tg ) = 150/165 ≈ 0.91. Therefore, with a small reduction in efficiency, ISI
is eliminated.
s(t)
Tg TB
CP OFDM block
t

|h(τ, t)|

r(t)

ISI-free block
t

Figure 1.4: OFDM with cyclic prefix (CP).

1.1.2 A Multicarrier Modulation

The OFDM signal can be expressed as1

X NX
" −1 #
s(t) = Ii,k ej2πfk t g(t − iTB ). (1.3)
i k=0

The pulse shape, g(t), is typically rectangular:


1, 0 ≤ t < TB ,

g(t) = (1.4)
0, otherwise.


Notice that the N data symbols {Ii,k }N −1

k=0 are transmitted during the ith block. The
set of complex sinusoids {exp (j2πf k t)}N −1
k=0 are referred to as subcarriers. The center
1
For simplicity, the guard interval is excluded from the signal definition in (1.3). The guard interval
and cyclic prefix is discussed in Chapter 2.
 6

frequency of the kth subcarrier is f k = k/TB and the subcarrier spacing, 1/TB Hz, makes
the subcarriers orthogonal over the block interval, expressed mathematically as
Z TB  Z TB
1 ∗   1
ej2πfk1 t ej2πfk2 t dt = ej2π(fk2 −fk1 )t dt
TB 0 TB 0

1, k1 = k2 ,
 (1.5)
=
0, k1 6= k2 ,


where (·)∗ represents the complex conjugate operation. The subcarrier orthogonality can
also be viewed in the frequency domain. Consider the 0th OFDM block:
N
X −1
s(t) = I0,k ej2πfk t , 0 ≤ t < TB . (1.6)
k=0

The frequency-domain representation is

N −1   
−j2πf TB /2
X k
S(f ) = F {s(t)} (f ) = TB e I0,k sinc f − TB , (1.7)
TB
k=0

where F{·}(f ) is the Fourier transform and


1,

x = 0,
sinc(x) = (1.8)
 sin πx ,

otherwise.
πx

Figure 1.5 plots |S(f )/TB | for N = 16 subcarriers and data symbols with normalized
amplitudes. The individual subcarrier spectra are also plotted. Notice that at the kth
subcarrier frequency, k/TB , the kth subcarrier has a peak and all the other subcarriers
have zero-crossings. Therefore, the subcarriers, while tightly packed (which improves
spectral efficiency), are non-interfering (i.e. orthogonal).

Figure 1.5 also demonstrates that OFDM is a multicarrier modulation, as opposed

to a single carrier modulation like the signal in (1.1). In general, a transmitted bandpass
signal is [421, p. 151]
n o
x(t) = < s(t)ej2πfc t , (1.9)

where fc is the carrier frequency. For single carrier,

X
xsc (t) = |Ii | cos [2πfc t + arg(Ii )] g(t − iTs ); (1.10)
i

while for multicarrier,

X NX
( −1    )
k
xmc (t) = |Ii,k | cos 2π fc + t + arg(Ii,k ) g(t − iTB ). (1.11)
TB
i k=0
 7

1.2 Subcarrier
Overall

1
Spectrum magnitude, |S(f )/TB |

0.8

0.6

0.4

0.2

0
-2 0 2 4 6 8 10 12 14 16 18
Normalized frequency, f TB

Figure 1.5: Subcarrier and overall spectrum. (N = 16; |I 0,k | = 1, for all k)

For single carrier each symbol occupies the entire signal bandwidth, while for multicarrier
the bandwidth is split into many frequency bands (also referred to as frequency bins).
Notice that the multicarrier signal transmits the N data symbols in parallel over multiple
carriers each centered at (fc + k/TB ) Hz, k = 0, 1, . . . , N − 1.

By properly designing the subcarrier spacing, each frequency bin is made frequency-
nonselective. The wideband frequency-selective channel is converted into N contigu-
ous narrowband frequency-nonselective bins. Figure 1.6 shows 18 bins in the range
[−0.9, −0.78] MHz for the N = 300 OFDM system over the channel in Figure 1.2(b).
Notice that the channel gain per bin varies over a 15 dB range. The OFDM modulation
can be optimized for the channel by sending more bits in frequency bins with high gain
and fewer bits in frequency bins with low gain. This technique, known as bit loading,
requires a fairly stable channel, one that can be accurately measured. For this reason,
bit loading is more common in wireline systems and stationary wireless systems than in
wireless systems with high mobility.

Frequency selectivity is the frequency-domain dual of intersymbol interference. Trans-

mitting the single carrier signal over the 2 MHz channel results in a frequency-selective
response. For OFDM, the overall channel is frequency-selective but for each bin the chan-
 8

Channel power (dB)

-5

-10

-15

-20
Frequency bins
-25

-30
−0.9 −0.85 −0.8
Frequency, f − fc (MHz)

Figure 1.6: OFDM converts wideband channel to N narrowband frequency bins.

nel is frequency non-selective and thus ISI is avoided. Therefore, Figure 1.6 illustrates a
frequency-domain interpretation of how OFDM avoids intersymbol interference.

1.1.3 Discrete-Time Signal Processing

Thus far, two of OFDM’s primary advantages have been discussed: the elimination
of ISI and the ability to optimize the modulation with bit loading. The third appeal of
OFDM is that the modulation and demodulation is done in the discrete-time domain with
the inverse fast Fourier transform (IFFT) and fast Fourier transform (FFT), respectively.
This is seen by sampling s(t) in (1.6) at N equally spaced time instances:
N
X −1
y[i] ≡ s(t)|t=iTB /N = I0,k ej2πki/N , i = 0, 1, . . . N − 1, (1.12)
k=0

which is the inverse discrete Fourier transform (IDFT) of the symbol vector I 0 =
[I0,0 , I0,1 , . . . , I0,N −1 ]. Therefore, s(t) is generated at the transmitter with an IDFT fol-
N −1
lowed by a digital-to-analog (D/A) converter. The frequency-domain symbols {I 0,k }k=0
can be expressed as
N −1
1 X
I0,k = y[i]e−j2πkn/N , k = 0, 1, . . . N − 1, (1.13)
N
i=0

which is the discrete Fourier transform (DFT) performed on the time-domain samples.
Consequently, the symbols are demodulated at the receiver with an analog-to-digital
(A/D) converter followed by a DFT.
 9

The IDFT/DFT is performed efficiently with IFFT/FFT algorithms. Doing so is

much simpler than performing the modulation/demodulation in the continuous-time
domain with N orthogonally tuned oscillators. Moreover, the signal processing can be
performed in software, making OFDM suitable for software defined radios (SDRs) [185].

1.2 Problems with OFDM

OFDM has two primary drawbacks. The first is sensitivity to imperfect frequency
synchronization which is common for mobile applications. This sensitivity arises from
the close subcarrier spacing. Figure 1.5 shows that the subcarriers are properly orthog-
onal at f = k/TB , k = 0, 1, . . . , N − 1. However, if the frequency synthesizer at the
receiver is misaligned by, say, fo /TB Hz, where −0.5 < fo < 0.5, the subcarriers are
not orthogonal and therefore interfering with one another. This intercarrier interference
(ICI) is illustrated in Figure 1.7: assuming that the receiver is tuned to (k +  fo )/TB
Hz rather than at the ideal k/TB Hz, the N − 1 neighboring subcarriers interfere with
the demodulation of the kth subcarrier. The intercarrier interference causes ISI—and
potentially high irreducible error floors.

The second problem with OFDM is that the signal has large amplitude fluctuations
caused by the summation of the complex sinusoids. The real and imaginary part of the
Spectrum magnitude, |S(f )/TB |

0.2

0.04

k−1 k k + fo k+1

Normalized frequency, f TB

Figure 1.7: Frequency offset causes ICI. ( fo = 0.25)

 10

OFDM signal is
N
X −1
< {s(t)} = < {I0,k } cos (2πkt/TB ) − = {I0,k } sin (2πkt/TB ) , (1.14)
k=0

and
N
X −1
= {s(t)} = < {I0,k } sin (2πkt/TB ) + = {I0,k } cos (2πkt/TB ) , (1.15)
k=0

respectively. Figure 1.8(a) shows the real and imaginary parts of an example OFDM
signal with N = 16 subcarriers. Also plotted are the individually modulated sinusoids.
Notice that each sinusoids has a constant amplitude, but when summing the sinusoids
the resulting OFDM signal fluctuates over a large range. The instantaneous signal power,
|s(t)|2 = <2 {s(t)} + =2 {s(t)}, is plotted in Figure 1.8(b). The ratio between the peak
power and the average power is 144/16 = 9 (or in decibels, 10 log 10 9 ≈ 9.5 dB).

160
12

10 140

8 Subcarriers |s(t)|2
<{s(t)} 120 Peak power
={s(t)} Average power
6
Power magnitude

100
Signal amplitude

2
80
0
60
-2

-4 40

-6
20
-8
0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Normalized time, t/TB Normalized time, t/TB

(a) Signal amplitude. (b) Signal power.

Figure 1.8: A typical OFDM signal (N = 16). The PAPR is 9.5 dB.

OFDM’s high peak-to-average power ratio (PAPR) requires system components with
a large linear range capable of accommodating the signal. Otherwise, the circuitry
 11

distorts the waveform nonlinearly, and nonlinear distortion results in a loss of subcarrier
orthogonality which degrades performance.

One such nonlinear device is the transmitter’s power amplifier (PA) which is respon-
sible for the system’s operational range [424]. Ideally the output of the PA is equal to
the input times a gain factor. In reality the PA has a limited linear region, beyond which
it saturates to a maximum output level. Figure 1.9 shows a representative input/output
curve, known as the AM/AM conversion. In the linear region the curve matches the
ideal, but as the input power increases the PA saturates. The most efficient operating
point is at the PA’s saturation point, but for signals with large PAPR the operating
point must shift to the left keeping the amplification linear. The average input power
is reduced and consequently this technique is called input power backoff (IBO). To keep
the peak power of the input signal less than or equal to the saturation input level, the
IBO must be at least equal to the PAPR. Thus the required IBO for the OFDM signal
in Figure 1.8 is 9.5 dB. At this backoff the efficiency of a Class A power amplifier is
less than 6%. Such an efficiency is detrimental to mobile battery-powered devices which
have limited power resources. Moreover, the operational range of the system is reduced
by a factor of nine2 .

Max output

Optimum

Linear region Saturation region

Output power

Actual

Backoff
AM/AM curve
Operating points
Ideal AM/AM

Input power

Figure 1.9: Power amplifier transfer function.

2
IBO of 9.5 dB corresponds to 109.5/10 ≈ 9 times less signal power transmitted in channel; the
(theoretical) efficiency of a Class A amplifier is 0.5/(109.5/10 ) ≈ 0.06 [374].
 12

Nonlinearities in the transmitter also cause the generation of new frequencies in

the transmitted signal. This intermodulation distortion causes interference among the
subcarriers, and a broadening of the overall signal spectrum. The later causes interference
between neighboring systems, an effect known as adjacent channel interference.

1.3 Constant Envelope Waveforms

Constant envelope (CE) waveforms are appealing since the optimum operating point
in Figure 1.9 is attainable. The baseband CE signal representation is

s(t) = Aejφ(t) , (1.16)

where A is the signal amplitude and φ(t) is the information bearing phase signal. The
advantage of the CE waveform is that the instantaneous power is constant: |s(t)| 2 = A2 .
Consequently, the PAPR is 0 dB and the required backoff is 0 dB. The PA can therefore
operate at the optimum (saturation) point, maximizing average transmit power (good
for range) and maximizing PA efficiency (good for battery life). Also, since the linearity
requirement is reduced, nonlinear PAs can be used which are generally more efficient
and less expensive than linear PAs. For example, the maximum theoretical efficiency of
a linear Class A power amplifier is 50%, while for a nonlinear Class E PA the maximum
theoretical efficiency is 100% [424].

Constant envelope signals are thus ideal in terms of the practical considerations of the
power amplifier. The question is how to embed digital information into φ(t) providing
good performance, spectral economy, and high data rates over the wireless channel.
Notice that the single carrier signal in (1.1) is constant envelope when |I i | = 1 and g(t)
is rectangular. This type of modulation, however, has large spectral sidelobes which
cause adjacent channel interference. In practice, non-rectangular pulse shapes are used
which result in a non-CE signal.

Continuous phase modulation (CPM) is a class of signaling that has very low sidelobe
power while maintaining the constant envelope property [14,421]. CPM uses memory to
smooth φ(t). The memory, however, increases the complexity of the receiver, which is a
key disadvantage of CPM. Also CPM systems have difficulty operating over frequency-
selective channels [118].
 13

1.4 Constant Envelope OFDM

Constant envelope OFDM (CE-OFDM) combines OFDM and constant envelope sig-
naling. The high peak-to-average power ratio OFDM signal is transformed into a CE
waveform. The CE-OFDM signal takes the form of (1.16) where the phase signal is an
OFDM waveform. For example, the phase signal can be the real part of the OFDM
signal:
N
X −1
φ(t) = < {sOFDM (t)} = < {I0,k } cos (2πkt/TB ) − = {I0,k } sin (2πkt/TB ) , (1.17)
k=0

where sOFDM (t) is the signal in (1.6). Figure 1.10 compares a conventional OFDM
bandpass signal with a bandpass CE-OFDM signal. Both are derived from the same
baseband OFDM message signal.


OFDM bandpass

OFDM message
R

CE-OFDM bandpass

Figure 1.10: Comparison of OFDM and CE-OFDM signals.

The motivation for CE-OFDM is to eliminate the PAPR problem of the conventional
OFDM system. Certainly, this is accomplished since the CE-OFDM signal has the 0 dB
PAPR property. The question is: at what cost? What is the performance of CE-
OFDM? What is its bandwidth? Can the guard interval be used in CE-OFDM as it is
in conventional OFDM? This thesis aims to answering these questions by analyzing the
various aspects of the CE-OFDM modulation.
 14

1.5 Thesis Overview

In Chapter 2 the basics of OFDM is further studied. The effect of the nonlinear
power amplification on OFDM is evaluated. In Chapter 3 the CE-OFDM modulation
format is defined and the spectral properties are studied. The performance aspects of
CE-OFDM in the presence of additive noise are analyzed in Chapter 4. Performance
analysis is extended to frequency-nonselective fading channels in Chapter 5, and multi-
path frequency-selective fading channels in Chapter 6.
Chapter 2

OFDM

In Sections 1.1 and 1.2 the basic properties of OFDM are identified. In this chapter,
OFDM is studied in more detail. Section 2.1 covers key properties of OFDM. In Section
2.1.1, the cyclic prefix is studied. In Section 2.1.2, the processing of the discrete-time
samples is described, and the equivalence of linear channel convolution and circular
channel convolution is explained. In light of this property, OFDM is considered a special
case of the more general block modulation with cyclic prefix scheme, as discussed in
Section 2.1.3. Finally, in Section 2.1.4 the main functional blocks of the OFDM system
are described.

The PAPR statistics are analyzed in Section 2.2 and power amplifier models used
to evaluated system performance are described in Section 2.3. Then in Section 2.4 the
effect of nonlinear power amplification on OFDM systems is studied in terms of spectral
leakage (Section 2.4.1), performance degradation (Section 2.4.2), and system range and
efficiency (Section 2.4.3). Lastly, the various PAPR mitigation techniques found in the
research literature are categorized in Section 2.5, and a technique called signal clipping
is evaluated in terms of its effectiveness to improve system performance.

15
 16

2.1 More OFDM Basics

2.1.1 The Cyclic Prefix

In Section 1.1.1 it is claimed that the use of the guard interval results in ISI-free
operation. This is true so long as a cyclic prefix is transmitted during the interval. This
is demonstrated below and it is shown that ISI results if anything but the cyclic prefix
is transmitted.

During the OFDM block interval, the waveform is

N
X −1
s(t) = Ik ej2πfk t , 0 ≤ t < TB , (2.1)
k=0

where {Ik }N −1 N −1
k=0 are the data symbols, {exp(j2πf k t)}k=0 are the subcarriers, N is the
total number of subcarriers, fk = k/TB is the center frequency of the kth subcarrier and
TB is the block period. The guard interval is defined during −T g ≤ t < 0, where Tg is
the guard period. To transmit a cyclic prefix, the last T g s of the block is transmitted
during the guard interval:
N
X −1 N
X −1 N
X −1
s(t) = Ik ej2πfk (t+TB ) = Ik ej2πfk t ej2πk = Ik ej2πfk t , (2.2)
k=0 k=0 k=0

−Tg ≤ t < 0. Notice that the above simplification is made due to the periodicity of the
signal. Thus the OFDM signal having a guard interval with cyclic prefix is simply
N
X −1
s(t) = Ik ej2πfk t , −Tg ≤ t < TB . (2.3)
k=0

The received signal is

r(t) = s(t) ∗ h(τ ) + n(t)

Z ∞
= h(τ )s(t − τ )dτ + n(t) (2.4)
Z−∞
τmax
= h(τ )s(t − τ )dτ + n(t),
0

where h(τ ) is the time-invariant channel impulse response 1 and n(t) is additive noise.
The bounds of integration are simplified since the channel is assumed causal [h(τ ) = 0
1
In (1.2), the received signal is expressed in terms of the time-variant channel impulse response h(τ, t).
If the channel is assumed to be time invariant, the impulse response is referred to as simply h(τ ).
 17

for τ < 0] and to have a maximum propagation delay τ max [h(τ ) = 0 for τ > τmax ]. The
received signal during the guard interval, which has interference from the previous block
(see Figure 1.4), is ignored and r(t) during 0 ≤ t < T B is processed. An estimate of the
k0 th data symbol is made by correlating r(t) with the k 0 th subcarrier:
Z TB
1 h i∗
Iˆk0 = r(t) ej2πfk0 t dt, (2.5)
TB 0
which expands to
TB
1
Z
Iˆk0 = r(t)e−j2πfk0 t dt
TB 0
Z TB "Z τmax N −1
#
1 X
j2πfk (t−τ )
= h(τ ) Ik e dτ e−j2πfk0 t dt + Nk0 (2.6)
TB 0 0 k=0
N −1 τ Z TB
1 X
Z max
= Ik h(τ )e−j2πfk τ dτ ej2πt(fk −fk0 ) dt + Nk0 ,
TB 0 0
k=0

where
TB
1
Z
N k0 = n(t)e−j2πfk0 t dt. (2.7)
TB 0
But since 
TB 1, k = k0 ,

1
Z
ej2πt(fk −fk0 ) dt = (2.8)
TB 0 0,

k 6= k0 ,
(2.6) simplifies to
Iˆk0 = Ik0 H[k0 ] + Nk0 , (2.9)

where Z τmax
H[k0 ] = h(τ )e−j2πfk0 τ dτ, (2.10)
0
which is the Fourier transform of h(τ ) evaluated at f = f k0 .

This shows that the N received data symbols { Iˆk }N −1

k=0 are equal to the transmitted
symbols {Ik }N −1 N −1
k=0 scaled by the complex-valued channel gains {H[k]} k=0 . ISI is avoided
since the kth symbol isn’t impacted by the N − 1 other symbols. Therefore, using the
guard interval with cyclic prefix provides ISI-free operation.

Now it is shown that by transmitting a signal other than the cyclic prefix during the
guard interval causes ISI. Suppose that the transmitted signal is

b(t),

−Tg ≤ t < 0,
s(t) = P (2.11)
 N −1 Ik ej2πfk t , 0 ≤ t < TB ,

k=0
 18

PN −1
where b(t) 6= k=0 Ik ej2πfk t . The estimate of the k0 th data symbols is
Z TB
ˆ 1
I k0 = r(t)e−j2πfk0 t dt
TB 0
Z TB Z τmax 
1 (2.12)
= h(τ )s(t − τ )dτ e−j2πfk0 t dt + Nk0
TB 0 0

= A k0 + B k0 + N k0 .

The bounds of integration are separated into two segments, [0, T g ] and [Tg , TB ]:
Tg τmax
1
Z Z
A k0 = h(τ )s(t − τ )e−j2πfk0 t dτ dt, (2.13)
TB 0 0

and
TB τmax
1
Z Z
B k0 = h(τ )s(t − τ )e−j2πfk0 t dτ dt. (2.14)
TB Tg 0

Ak0 is a non-zero offset term which is a function of b(t). For the second term, t − τ > 0,
thus
N −1
"Z #
TB τmax
1
Z X
j2πfk (t−τ )
B k0 = h(τ ) Ik e dτ e−j2πfk0 t dt
TB Tg 0 k=0
(2.15)
N −1 τmax TB
1 X Z Z
= Ik h(τ )e−j2πfk τ dτ ej2πt(fk −fk0 ) dt.
TB 0 Tg
k=0

Due to the integration bounds for t, the orthogonality condition in (2.8) can’t be applied
to (2.15), and this results in ISI. The estimated data symbol is expressed as

Iˆk0 = Ik0 Hk0 C1 + Nk0 + ICI, (2.16)

where C1 = (TB − Tg )/TB , and the interference terms is

Z TB
1 X
ICI = Ak0 + H[k] Ik ej2πt(fk −fk0 ) dt. (2.17)
TB Tg k6=k0

The interference is denoted as ICI, intercarrier interference, since the subcarriers are no
longer orthogonal and interfere with one another. This phenomenon was described in
Section 1.2 in the context of imperfect frequency synchronization. Therefore, ICI can
manifest itself in more than one way, and when it does the data symbols interfere with
one another resulting in ISI.

In [358], cyclic prefixed OFDM is compared to zero-padded OFDM [b(t) = 0]. The
zero-padding causes ISI, but has the advantage of being able to recover data symbols
 19

located at channel zeros. This is in contrast with cyclic prefixed OFDM since, as shown
in (2.9), a channel zeros at the kth subcarrier, that is, H[k] = 0, results in an estimated
data symbol that consists entirely of noise. The zero-padded system avoids this problem
at the cost of increased receiver complexity due to equalization requirements.

2.1.2 Discrete-Time Model

It is convenient to describe OFDM by a discrete-time model. Consider sampling s(t),

h(τ ) and r(t) at the sampling rate fsa = JN/TB samp/s, where J ≥ 1 is the oversampling
factor. The sampling instances are shown in the figure below.
Signal sampling
−Tg TB

··· ··· t
−Ng Tsa −Tsa 0 Tsa (NB − 1)Tsa
−Ng Tsa ≥ −Tg

Channel sampling τmax

··· τ
0 Tsa (Nc − 1)Tsa
(Nc − 1)Tsa ≤ τmax

Figure 2.1: Sampling instances.

The number of guard samples, Ng , and channel samples, Nc , are defined as

 
Tg Tg
Ng ≡ ≤ , (2.18)
Tsa Tsa
and 
τmax τmax
Nc ≡ +1≤ + 1, (2.19)
Tsa Tsa
where Tsa = 1/fsa is the sampling period. The number of samples per block is N B = JN ;
and, by design, Ng ≥ Nc . The signal samples are

s[i] = s(t)|t=iTsa , i = −Ng , . . . , 0, . . . , NB − 1, (2.20)

and the channel samples are

h[i] = h(τ )|τ =iTsa , i = 0, . . . , Nc − 1. (2.21)

The received samples are expressed by the linear convolution sum

NX
c −1

r[i] = h[m]s[i − m] + n[i], i = −Ng , . . . , 0, . . . , NB − 1, (2.22)

m=0
 20

where {n[i]} are samples of the noise signal n(t). The guard interval samples are ignored
and the samples
NX
c −1

r[i] = h[m]s[i − m] + n[i], i = 0, . . . , NB − 1 (2.23)

m=0

are processed.

The linear convolution in (2.23) is equivalent to a circular convolution since, due

to the cyclic prefix, {s[i − m]} is periodic with period N B . The circular convolution
can be performed by taking the IDFT of the product of two DFTs [422, pp. 415–420].
Therefore, ignoring the noise samples, (2.23) can be expressed as

r[i] = IDFT {H[k]S[k]}

NDFT
X−1 (2.24)
1
= H[k]S[k]ej2πik/NDFT , i = 0, . . . , NB − 1,
NDFT
k=0

where IDFT{·} represents the inverse discrete Fourier transform;

NDFT
X−1
S[k] = s[i]e−j2πik/NDFT , k = 0, . . . , NDFT − 1 (2.25)
i=0

and
NDFT
X−1
H[k] = h[i]e−j2πik/NDFT , k = 0, . . . , NDFT − 1 (2.26)
i=0

are the NDFT -point DFTs of the signal and channel samples, respectively. The DFT
size is, in general, NDFT ≥ NB . If NDFT > NB , the signal vector is zero-padded. Since
NDFT > Ng , the channel samples are zero-padded: h[i] = 0 for i = N c , . . . , NDFT − 1.

Figure 2.2 shows a block diagram representing the calculation of (2.24). The effect of
the channel is simply a DFT followed by a multiplier bank (H[k]), which is then followed
by an IDFT. Also shown is the inverse channel which is a DFT followed by a multiplier
bank (1/H[k]) followed by an IDFT. Thus the transmit samples s[i] can be reconstructed
by passing the receive samples r[i] through the inverse channel.

2.1.3 Block Modulation with FDE

The inverse channel structure in Figure 2.2 corrects the distortion caused by the
channel in the frequency domain, and is therefore called a frequency-domain equalizer
 21

Channel

s[i] DFT H[k] IDFT r[i]

Inverse channel
1
r[i] DFT IDFT s[i]
H[k]

Figure 2.2: Circular convolution with channel and the inverse channel.

Frequency-domain equalizer

Data Multiplier Data

Modulator Channel DFT IDFT Demodulator
{Ik } bank {Iˆk }

Figure 2.3: Block modulation with cyclic prefix and FDE.

Frequency-domain equalizer

Data Multiplier Data

IDFT Channel DFT bank IDFT DFT {Iˆk }
{Ik }

Data Multiplier Data

IDFT Channel DFT
{Ik } bank {Iˆk }

Figure 2.4: OFDM is a special case.

 22

(FDE). Such an equalizer can be used only when the effect of the channel is a circular
convolution. This is the case for OFDM, but isn’t unique to OFDM since any modulation
can use a cyclic prefix. This observation was first identified by Sari et al. [462] and
suggests a more general modulation approach: block modulation with cyclic prefix and
frequency-domain equalization. Figure 2.3 shows a simplified block diagram of such
a system. (The insertion of the cyclic prefix at the transmitter and removal at the
receiver is implied but not included in the diagram for simplicity.) For the special case
of OFDM, the modulation is a IDFT and the demodulation is a DFT as shown in Figure
2.4. Notice that the DFT and IDFT cancel each other and the resulting diagram depicts
the conventional OFDM system.

The multiplier bank at the output of the DFT is often referred to as a one-tap
equalizer, one complex multiplication per frequency bin. This operation is required for
data symbols that rely on coherent demodulation, such as M -ary phase-shift keying
(M -PSK) and M -ary quadrature-amplitude modulation (M -QAM).

As Sari et al. pointed out, OFDM doesn’t eliminate the equalization problem (asso-
ciated with conventional single carrier modulation); rather, OFDM converts the problem
to the frequency domain. Since Sari’s original paper, there has been a considerable num-
ber of publications focused on the block modulation technique using conventional single
carrier modulations [8, 30, 54, 107, 116, 132, 142, 153, 154, 196, 197, 245, 388, 460, 461, 463,
481, 533, 565, 574].

2.1.4 System Diagram

The block diagram in Figure 2.4 conceptually illustrates the OFDM system. Figure
2.5 shows a more detailed description of OFDM’s functional blocks.

The encoder adds redundancy to the bit stream for error control. The encoded bits
are then mapped to the data symbols I k . In general, the data symbols are complex
numbers which result from mapping the bits to points on the complex plane. Next, the
symbols are serial-to-parallel (S/P) converted and processed by the IDFT. The cyclic
prefix is added and the signal samples, s[i], are passed through the digital-to-analog
(D/A) converter to obtain the continuous-time OFDM signal s(t). Finally, the signal is
amplified and transmitted.
 23

Transmitter

Bits 11101 Ik Add

Encoder Mapper S/P IDFT P/S
01101 CP

s[i] s(t) Power

D/A amplifier

Receiver

r(t) Remove r[i] Equalize

A/D CP S/P DFT C[k] P/S

Iˆk 11001 Bits

Detector Decoder
01101

Figure 2.5: OFDM system diagram.

At the receiver, the inverse operations are performed. First, the received signal,
r(t), is sampled to obtain the discrete-time sequence r[i]. The guard interval samples
are removed, the DFT is performed and each frequency bin is equalized by a complex
multiplication. The estimated data symbols, Iˆk , are processed by the detector which
outputs a stream of estimated receive bits, and the decoder attempts to correct any bit
errors that may have occurred.

As discussed in Section 1.2, one of OFDM’s key drawbacks is the high peak-to-
average power ratio. Nonlinearities in the power amplifier distort the transmitted signal
and large input power backoff is required which results in low amplifier efficiency. In the
next sections the impact of the PA is studied. But first, the statistical properties of the
PAPR are discussed.
 24

2.2 PAPR Statistics

The peak-to-average power ratio of the OFDM signal is best viewed statistically. For
any given block interval, the PAPR is a random quantity since it depends on the data
symbols {Ik }N −1
k=0 . Assuming that they’re selected randomly from a set of M complex
numbers, there are M N unique symbol sequences, and thus M N unique OFDM wave-
forms per block. Of these waveforms, some have a high PAPR, while others have a
relatively low PAPR. Therefore, it is desirable to understand the statistical distribution
of this quantity.

The OFDM signal is

N
X −1
s(t) = Ik ej2πfk t , 0 ≤ t < TB . (2.27)
k=0

The signal during the guard interval is ignored since it has no impact on the PAPR
distribution. M -PSK data symbols are assumed, therefore |I k | = 1 for all k. The
average power of s(t) is
TB
1
Z
Ps = |s(t)|2 dt = N. (2.28)
TB 0
The peak-to-average power ratio is defined as
.
PAPRs = max |s(t)|2 Ps . (2.29)
t∈[0,TB )

Notice that the absolute maximum signal power is N 2 , so the PAPR can be as high as
N . However, the likelihood that all the subcarriers align in phase is extremely low. For
example, as pointed out in [381], a N = 32 subcarrier system having 4-ary data symbols
and a block period of TB = 100 µs obtains the theoretical maximum PAPR once every
3.7 million years. Thus it is more meaningful to describe the PAPR statistically rather
than in absolute terms.

Since the average signal power is a constant, the randomness of the PAPR depends on
the randomness of the instantaneous power |s(t)| 2 , and more specifically, the maximum
instantaneous power over 0 ≤ t < TB . For large N , the real and imaginary parts of
s(t) are accurately modeled as Gaussian random processes (due to the application of
the central limit theorem [394, 421]). Consequently, the instantaneous signal power is
chi-squared distributed with two degrees of freedom [421, p. 41], and the complementary
 25

cumulative distribution function (CCDF) of the normalized instantaneous signal power

is approximated as
|s(t)|2
 
P >x ≈ e−x . (2.30)
Ps
A lower bound of the peak-to-average power ratio’s CCDF is [515]

P (PAPRs > x) ' 1 − (1 − e−x )N , (2.31)

where 1 − (1 − e−x )N is an approximation to the CCDF of the PAPR of the sequence

{s(t)|t=iTB /N ; i = 0, 1, . . . N −1} [173]. The PAPR of the discrete-time sequence provides
a lower bound to the continuous-time signal since peaks can occur between sampling
times.

100 100
Simulation Simulation
Approximation (2.30) Lower bound (2.31)

10−1 10−1
´
CCDF, P |s(t)|2 /Ps > x

CCDF, P (PAPRs > x)

10−2 10−2
`

10−3 10−3

10−4 0 2 4 6 8 10 10−4 4 6 8 10 12 14
x (dB) x (dB)

(a) Instantaneous power. (b) Peak-to-average power ratio.

Figure 2.6: Complementary cumulative distribution functions. (N = 64)

Figure 2.6(a) compares a simulated instantaneous power CCDF with the approxi-
mation in (2.30). This demonstrates the accuracy of the Gaussian approximation to the
real and imaginary part of s(t). Figure lower bound in (2.31). 2.6(b) compares PAPR
simulation results to the The bound is shown to be within 1 dB of the simulated result
for lower values of x. The 0.0001 PAPR is shown to be at around 11.25 dB, and at this
 26

level the bound is tight. Notice that essentially all OFDM blocks have a PAPR greater
than 6 dB, 10% have a PAPR greater than 8.5 dB, and 0.5% have a PAPR greater than
10 dB.

For the results in Figure 2.6, the number of subcarriers is N = 64 and QPSK data
symbols (4-ary PSK) are used, that is, I k ∈ {±1, ±j}. While the symbols constellation
has little impact on the PAPR statistics, the number of subcarriers does. Figure 2.7
shows the lower bound (2.31) over a range N = 32 to N = 1024. Notice that the 0.001
PAPR is 1 dB larger for N = 512 than for N = 32. For the N = 64 system, the PAPR
is greater than 8 dB for roughly 10% of the time. For the N = 1024 system, however,
the PAPR is greater than 8 dB nearly all of the time.
100

k 5 6 7 8 9 10

10−1
CCDF, P (PAPRs > x)

10−2

10−3

10−4 4 5 6 7 8 9 10 11 12
x (dB)

Figure 2.7: PAPR CCDF lower bound (2.31) for N = 2 k , k = 5, 6, . . . , 10.

2.3 Power Amplifier Models

To determine the impact of the PAPR on system performance, power amplifier mod-
els must be defined. Two models commonly used in the research literature are the
solid-state power amplifier (SSPA) model and the Saleh traveling-wave tube amplifier
(TWTA) model [454]. They are described here and then used in Section 2.4 for perfor-
mance evaluation.
 27

In general, modeling nonlinear power amplifiers is complicated (see [233, chap. 5]).
A common simplification is to assume that the PA is a memoryless nonlinearity, and
therefore has a frequency-nonselective response. For example, if the PA input is

sin (t) = A(t) exp[jφ(t)], (2.32)

the output is
 
sout (t) = G[A(t)] exp j{φ(t) + Φ[A(t)]} , (2.33)

where G(·) and Φ(·) are known as the AM/AM and AM/PM conversions, respectively.
The SSPA model is expressed as

g0 A
G(A) = h i1/2p , and Φ(A) = 0, (2.34)
2p
1 + (A/Asat )

where g0 is the amplifier gain, Asat is the input saturation level, and p controls the
AM/AM sharpness of the saturation region. For this model the AM/PM conversion is
assumed to be negligibly small.

Though widely known as the Rapp model [426], (2.34) should be credited to the
original work by A. J. Cann, published a decade earlier in the IEEE literature [71].
Cann’s formula is obtained with the simple manipulation:
A
G(A) =g0 h i1/2p
1 + (A/Asat )2p
A [(Asat /A)2p ]1/2p
=g0 h i1/2p × (2.35)
1 + (A/Asat )2p [(Asat /A)2p ]1/2p

Asat
=g0 h i1/2p ,
1 + (Asat /A)2p

which is precisely the nonlinearity presented in Cann’s paper.

Saleh’s TWTA model is expressed as [110]

g0 A αφ A2
G(A) = , and Φ(A) = . (2.36)
1 + (A/Asat )2 1 + β φ A2

Notice that the AM/PM conversion, determined by the constants α φ and βφ , is non-zero.
The TWTA model is therefore more nonlinear than the SSPA model.
 28

To reduce nonlinear distortion in the amplified OFDM signal, input power backoff
(IBO) is required. It is defined as [375]

A2sat
IBO = , (2.37)
Pin

where Pin = E{|sin (t)|2 } = E{A2 (t)} is the average power of the input signal. Equiva-
lently, (2.37) can be written as
A2sat
Pin = ; (2.38)
IBO
thus, given Asat and IBO, the input signal power can be scaled accordingly to satisfy
(2.38).

Assuming that the PAPR of the input signal is PAPR in , the peak power can be
written as
PAPRin 2 A2
Pmax = PAPRin · Pin = Asat = sat , (2.39)
IBO K
where
IBO
K= (2.40)
PAPRin
is defined as the backoff ratio. Notice that for K > 1 the backoff is greater than the input
signal’s PAPR; for K < 1 the backoff is less than the input PAPR. Now, the maximum
value of the input, Amax = max |A(t)|, can be written in terms of the backoff ratio and
the input saturation level:
p Asat
Amax = Pmax = √ . (2.41)
K

Figure 2.8 shows the AM/AM (solid lines) and AM/PM (dashed lines) conversions
for the SSPA (thick lines) and TWTA (thin lines) models for various backoff ratios K.
For the SSPA model, p = 2; for the TWTA model, α φ = π/12 and βφ = 1/4. The x-axis
is normalized to the maximum input level A max , and the y-axis is normalized to the
maximum output level g0 Asat . For K = −10 dB the IBO is one-tenth the input signal
PAPR, and thus the nonlinearity is severe. One the other hand, for K = 10 dB the
IBO is ten times the input signal PAPR and the PA response is nearly linear. As stated
above, the non-zero AM/PM conversion of the TWTA model makes it more nonlinear
than the SSPA model.

Insight can be gained by comparing Figure 2.6(b) and Figure 2.8. For example,
assuming that the backoff is IBO = 6 dB, the conversions are never as linear as the
K = 3 dB curves (the PAPR is a always greater than 3 dB) and are more nonlinear
 29

1 1

Normalized output value, G(A)/g0 Asat

Normalized output value, G(A)/g0 Asat

0.5 0.5

0 0

−0.5 −0.5

−1 −1
−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1
Normalized input value, A/Amax Normalized input value, A/Amax

(a) K = 10 dB (b) K = 3 dB

1 1
Normalized output value, G(A)/g0 Asat
Normalized output value, G(A)/g0 Asat

0.5 0.5

0 0

−0.5 −0.5

−1 −1
−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1
Normalized input value, A/Amax Normalized input value, A/Amax

(c) K = −3 dB (d) K = −10 dB

Figure 2.8: AM/AM (solid) and AM/PM (dash) conversions (SSPA=thick,

TWTA=thin) for various backoff ratios K.
 30

than the K = −3 dB curves for about 5% of the OFDM blocks (the 0.05 PAPR is 9 dB).
Therefore, even with a large IBO of 6 dB, the PA can impose high nonlinear distortion
on the transmitted signal. Also, the degree of distortion for a given OFDM block is
random (given a fixed IBO) since the PAPR for a given block is random.

2.4 Effects of Nonlinear Power Amplification

Power amplifier nonlinearities cause spectral leakage and performance degradation

to OFDM systems. These undesirable effects can be reduced with increase input backoff.
This is an unsatisfactory solution, however, since PA efficiency reduces with IBO. Also,
reducing the average transmit power reduces the operational range of the system. In
this section these various issues are studied.

2.4.1 Spectral Leakage

The first problem considered is spectral leakage. By using the Welch method [422, pp.
911–913], the power density spectrum at the output of the power amplifier can be quickly
estimated. The result is used to calculate estimated fractional out-of-band power curves,
defined as Rf
ˆ 0 Φ̂s (x)dx
FOBP(f )= , f > 0, (2.42)
0.5P̂s
R∞
where Φ̂s (f ) is the estimated power density spectrum of the signal and P̂s = −∞ Φ̂s (f )df
is the signal power. Figure 2.9 shows the curves for an N = 64 subcarrier OFDM signal
amplified by the TWTA power amplifier according to (2.36) at various backoff levels.
Also plotted is the FOBP curve for ideal linear amplification. These results show that
at least 6 dB backoff is required by the TWTA to avoid spectral broadening.

Figure 2.10 shows the 99.5% bandwidth as a function of IBO. The bandwidth of the
undistorted OFDM signal is f = 1.07W . For sufficient backoff, the bandwidth of the
nonlinearly amplified signal is the same. However, for IBO < 6 dB, the bandwidth is
shown to grow roughly linearly with IBO. For IBO = 1 dB, the 99.5% bandwidth is 73%
larger than the undistorted signal. Notice that the spectral leakage is roughly the same
for the two amplifier models.
 31

100
OFDM amplified with: TWTA PA
ideal PA

10−1
Fractional out-of-band power

IBO

0
10−2
2

10−3
0 0.25 0.5 0.75 1 1.25 1.5
Normalized frequency, f /W

Figure 2.9: Fractional out-of-band power of OFDM with ideal PA and with TWTA
model at various input power backoff. (N = 64, IBO in dB)

2.0 OFDM amplified with: TWTA PA

SSPA PA
ideal PA
1.8
99.5% bandwidth, f /W

1.6

1.4

1.2

1.0

0.8
0 2 4 6 8 10
Input power backoff, IBO (dB)

Figure 2.10: Spectral growth versus IBO. (N = 64)

 32

2.4.2 Performance Degradation

Next, the performance degradation caused by nonlinear amplification is considered.

The OFDM signal is passed through a PA and then it is corrupted by additive white
Gaussian noise (AWGN). The received signal is thus,

r(t) = sout (t) + n(t), (2.43)

where sout (t) is the output of the PA from (2.33) and n(t) is a complex-valued Gaussian
additive noise signal having a power density spectrum [421, p. 158]

N0 , |f | ≤ Bn /2,

Φn (f ) = (2.44)
0,

|f | > Bn /2,

where Bn is the bandwidth of the noise signal. The noise spectrum is assumed to be
constant over the effective bandwidth of the information bearing signal and is thus called
“white”. The transmitted data symbols are estimated by the correlation in (2.5) then
passed to the detector which makes the final decision. This decision is based on the
maximum-likelihood (ML) criterion assuming a linear PA; that is, the nearest point in
the symbol constellation [421, pp. 242–247].

The performance is estimated by way of computer simulation. Following the conven-

tion described in Section 2.1.2, the discrete-time signal representation is used and the
sampling rate fsa = JN/TB where J ≥ 1 is the oversampling factor. For the AWGN
channel, h(τ ) = δ(τ ), and therefore no guard interval is used. The noise samples {n[i]}
are Gaussian distributed and assumed independent:

σ 2 , i 1 = i 2 ,

n
E {n[i1 ]n[i2 ]} = (2.45)
0,

i1 6= i2 .

The autocorrelation function of n(t) [the inverse Fourier transform of (2.44)] has zero-
crossings at τ = 1/Bn . Thus assuming Bn = fs , (2.45) is satisfied and the noise sample
variance is σn2 = fsa N0 .

Figure 2.11 shows bit error rate (BER) performance as a function of E b /N0 , where
R TB
0|sout (t)|2 dt
Eb = (2.46)
Number of bits per block
 33

10−1
Nonlinear PA Nonlinear PA
Ideal PA Ideal PA
10−1

10−2

10−2
Bit error rate

Bit error rate

10−3
10−3

10−4
10−4

10−5 0 2 4 6 8 10 12 14 10−5 0 5 10 15 20 25 30
Signal-to-noise ratio per bit, Eb /N0 (dB) Signal-to-noise ratio per bit, Eb /N0 (dB)

(a) SSPA model, IBO = 0, 1, 2, 3, 4, 6, 8 dB; (b) TWTA model, IBO = 0, 1, . . . , 10, 16 dB;
0 = worst, 8 = best. 0 = worst, 16 = best.

Figure 2.11: Performance of QPSK/OFDM with nonlinear power amplifier with various
input power backoff levels. (N = 64)

is the energy per bit. The quantity E b /N0 is referred to as the signal-to-noise ratio (SNR)
per bit, or simply the SNR. QPSK data symbols are used, and the oversampling factor
is J = 4. For the SSPA results in Figure 2.11(a), the IBO ranges from 0 to 8 dB. At
the 0.0001 BER level, the IBO = 0 dB case suffers a 3 dB performance loss compared to
ideal AWGN performance, which is [421, pp. 268].
r !
Eb
BER = Q 2 , (2.47)
N0
R∞ 2 /2 √
where Q(x) = x e−y dy/ 2π is the Gaussian Q-function. To avoid degradation, 8
dB of backoff is required. The TWTA results in Figure 2.11(b) use IBO ranging from 0
to 16 dB. Notice the irreducible error floors for IBO ≤ 7 dB. To avoid degradation, 16
dB of backoff is required—8 dB more than for the SSPA case. The greater nonlinearity
of the TWTA model is evident from the results in this figure.

Figure 2.12 compares performance for higher-order PSK modulations. For M -PSK
 34

10
Ideal PA
10−1 SSPA: IBO = 3 dB
IBO = 6 dB M = 16
8
M = 16

Total degradation (dB)

10−2
M =8
Bit error rate

6
M =8

10−3
4 Ideal PA
SSPA
M = 2, 4

10−4 M = 2, 4
2

Target BER = 0.001

10−5 0 5 10 15 20 25 30
0
0 2 4 6 8 10
Signal-to-noise ratio per bit, Eb /N0 (dB) Input power backoff, IBO (dB)

(a) BER performance. (b) Total degradation.

Figure 2.12: Performance of M -PSK/OFDM with SSPA. (N = 64)

the data symbols are

Ik ∈ {exp(j2πm/M ); m = 0, 1, . . . , M − 1}. (2.48)

The number of bits per data symbols is log 2 M , therefore the bit energy is
R TB
|sout (t)|2 dt
Eb = 0 . (2.49)
N log2 M
Higher-order constellations are used for increased spectral efficiency at the price of BER
performance2 . In Figure 2.12(a) BER results for the SSPA model are shown. (The
results for M = 2 and M = 4 are very similar so only M = 2 is plotted.) The higher-
order modulations are shown to be more sensitive to the PA nonlinearity. For example,
the M = 16 result for IBO = 3 dB has an irreducible error floor at 5 × 10 −3 , while the
M = 2, 4 result at the same backoff shows only a 1 dB degradation. When increasing
the backoff to IBO = 6 dB, the error floor for M = 16 drops to 2 × 10 −5 and the 0.001
BER is about 2 dB worse than AWGN. Using IBO = 6 dB for M = 8 results in 2 dB
less degradation at the 0.001 bit error rate when compared to using IBO = 3 dB.
2
This is the case for linear modulation formats. This isn’t necessarily the case for nonlinear modulation
formats as discussed in Section 4.4.
 35

A more revealing way to view performance is in terms of total degradation, as shown

in Figure 2.12(b). The total degradation is defined as [121]

TD(IBO) = SNRPA (IBO) − SNRAWGN + IBO, [in dB] (2.50)

where SNRAWGN is the required signal-to-noise ratio per bit to achieve a target bit
error rate in AWGN; SNRPA (IBO) is the required SNR when taking into account the
distortion caused by the power amplifier at a given backoff. The “optimum” IBO, denote
as IBOopt , minimizes the total degradation, that is,

TD(IBOopt ) = TDmin = min TD(IBO). (2.51)

IBO≥0 dB

The target BER for the curves in Figure 2.12(b) is 0.001. Clearly the modulation order
influences the degradation. The minimum TD for M = 16 is 7.7 dB at IBO opt = 6.5
dB; for M = 8, TDmin = 5 dB at IBOopt = 3 dB. This can be interpreted as follows:
M = 8, while having lower spectral efficiency than M = 16 (3 b/s/Hz vs. 4 b/s/Hz),
suffers less degradation and can operate with less backoff, resulting in improved range
and higher PA efficiency. The M = 2 and M = 4 examples are shown to have the lowest
degradation and are thus the more robust against nonlinear distortion.

2.4.3 System Range and PA Efficiency

The total degradation is directly related to the system’s operational range. Consider
a transmitter operating at maximum transmit power. The range is represented by the
outermost ring in Figure 2.13. Now assume that the system requires a 3 dB backoff:
the range is reduced by one-half, as represented by the middle ring. Any degradation
caused by the PA further reduces range, as represented by the innermost circle. Thus
the actual range of the system is far less than the potential range of the transmitter.
The true capability of the power amplifier is greatly underutilized.

To quantify the relationship between the PA efficiency and the power backoff, the
theoretical efficiency of a Class A power amplifier is used [374]:

1 1
ηA = × 100%, IBO ≥ 1. (2.52)
2 IBO

The efficiency is thus inversely proportional to IBO and the maximum efficiency, 50%,
occurs at IBO = 1 (0 dB). The efficiency curve, shown in Figure 2.14, can be used
 36

Potential range
Potential range w/ IBO
Actual range

Figure 2.13: The potential range of system is reduced with input backoff; the range is
reduced further from nonlinear amplifier distortion.

in conjunction with Figures 2.10 and 2.12(b) to gain insight to the various tradeoffs
between PA efficiency, spectral containment, and performance/range. For example, the
optimum IBO in terms of total degradation for the 8-PSK SSPA example is IBO opt = 3
dB [Figure 2.12(b)]: however, the bandwidth expansion is 42% (Figure 2.10) and the
PA efficiency is ηA = 25% (Figure 2.14). The optimum IBO for the 16-PSK example,
6.5 dB, results in no bandwidth expansion but the PA efficiency is reduced to 11%. The
M = 2, 4 systems required minimal IBO for the SSPA, thus maximizing efficiency, but
the bandwidth expands by 87%.
50
45
Class-A PA efficiency, ηA (%)

40
35
30
25
20
15
10
5
00 1 2 3 4 5 6 7 8 9 10
Input power backoff, IBO (dB)

Figure 2.14: Power amplifier efficiency.

 37

2.5 PAPR Mitigation Techniques

There have been many schemes proposed in the research literature aimed at reducing
the impact of the PAPR problem. The goal of any scheme is to reduce the minimum
total degradation (for increased range) and the IBO opt (for increased PA efficiency). The
various schemes can be placed in one the following three categories:

1. transmitter enhancement techniques,

2. receiver enhancement techniques, or

3. signal transformation techniques.

Transmitter enhancement techniques include PAPR reduction schemes and PA lineariza-

tion schemes. The PAPR reduction schemes can be further divided into distortionless and
non-distortionless techniques. Distortionless techniques include coding (see [126,439,508]
and reference therein), constellation extension [269], tone reservation [169, 268, 512],
trellis-shaping [377], and multiple signal representation {aka selected mapping (SLM)
or partial transmit sequences (PTS), see [227] and its references}. Non-distortionless
schemes include signal clipping [27, 138, 290, 382], peak cancellation [330], and peak win-
dowing [403].

The PA linearization schemes attempt to predistort the OFDM signal such that the
overall response of the predistorter followed by the PA is linear—essentially equalizing
the amplifier. In [230], an LMS algorithm is applied for adaptive predistortion; in [395]
a neural network learning technique is used. Parametric techniques, which design a
predistorter based on a PA model, have been proposed. In [85, 122, 250, 567] nonlinear
polynomial models are used, and in [86] a Volterra-based model is suggested.

The second category, receiver enhancement techniques, have been suggested in [513],
[376] (maximum-likelihood decoding); in [259, 453] (signal reconstruction), and in [87]
(interference cancellation). Finally, the third category includes techniques that are based
on transforming the OFDM signal prior to the PA, and applying the inverse transform at
the receiver prior to demodulation. This category includes constant envelope OFDM (as
studied in the second half of this thesis) which uses a phase modulator as the transformer.
In [215, 329, 569–571] a companding transform is suggested.
 38

Signal Clipping

The remainder of this section focuses on the effectiveness of signal clipping, which has
been claimed to be the “simplest” and “most effective” PAPR reduction scheme [27, 87,
290, 375, 377, 380, 382, 391]. The impact of “clipping noise”—the intercarrier interference
caused by the clipping process—on system performance has been extensively analyzed
[39, 124, 382]. However, a common assumption is that the PA is linear [27, 39, 138,
290, 371, 380, 382, 391]. It is argued here that the effectiveness of a PAPR reduction
scheme must be measured not only by PAPR reduction, but by the more meaningful
measures of TDmin and IBOopt reduction. It is shown that clipping, while an effective
PAPR reduction scheme, does not reduce TD min nor does clipping reduce IBOopt for
an OFDM system. This result brings into question the usefulness of non-distortionless
PAPR reduction techniques in general.

The system under consideration is shown in Figure 2.15. When the switch is “on”
the PAPR reducing signal clipper is used. When “off” the system is identical to the one
studied in Section 2.4.2. Therefore, the earlier unclipped results serve as a performance
benchmark in which to compare the clipped results. The channel, as before, has an
impulse response h(τ ) = δ(τ ).

off sin (t)

OFDM s(t) sout (t) r(t) OFDM
PA h(τ )
modulator demodulator
PAPR on
reducing
sclip (t) n(t)
clipper

Figure 2.15: Block diagram. The system is evaluated with and without PAPR reduction.

The input to the clipping block is the OFDM signal s(t) from (2.27), the output is
the clipped OFDM signal:

s(t),

if |s(t)| ≤ Amax ,
sclip (t) = (2.53)
Amax ejψ(t) ,

if |s(t)| > Amax ,

where ψ(t) = arg[s(t)]. Therefore, the magnitude of the clipped signal does not exceed
Amax and the phase of s(t) is preserved. (This has been called “polar clipping” in the
literature [276].) The clipping severity is measured by the clipping ratio, defined as [375]
Amax
γclip = √ . (2.54)
Ps
 39

γclip OFDM signal

20 4 Clip radius

2
10
0
Imaginary axis

−10

−20

−20 −10 0 10 20
Real axis

Figure 2.16: Unclipped OFDM signal (9.25 dB PAPR). The rings have radius A max
which correspond to various clipping ratios γ clip (dB).

Figure 2.16 shows a typical OFDM signal on the complex plane. The dark rings have
radius Amax which correspond to clipping ratios γ clip = 0, 2, and 4 dB.

The PAPR of sclip (t) is

max |sclip (t)|2

t∈[0,T )
PAPRclip = R TB . (2.55)
1 2 dt
TB 0 |s clip (t)|

Clipping’s effectiveness at reducing PAPR is shown in Figure 2.17. For clipping ratio
γclip = 5 dB, the peak-to-average power ratio of the clipped signal is PAPR clip ≤ 10
dB; for γclip = 4 dB, PAPRclip ≤ 8 dB, and so forth. The 0.0001 PAPR improvement,
compared to the unclipped signal, is 1.2 dB for γ clip = 5 dB and by 3.2 dB for the
γclip = 4 dB.

Figure 2.18 shows PAPRclip as a function of the clipping ratio. The PAPR of
the unclipped signal is 13 dB3 . Notice that for large γclip , sclip (t) is unclipped, there-
3
This figure is made by generating 2 × 104 consecutive OFDM blocks. The PAPR of the overall block
is 13 dB.
 40

100

10−1 Clipped
Unclipped
P (PAPRclip > x)

10−2

10−3

γclip 3 4 5

10−4 0 2 4 6 8 10 12
x (dB)

Figure 2.17: PAPR CCDF of clipped OFDM signal for various γ clip (dB). [N = 64]

16

14
PAPRs
12
Peak-to-average power ratio (dB)

10

8
PAPRclip
6

0 PAPRclip as γclip → 0

−2
2
γclip
−4
−8 −6 −4 −2 0 2 4 6 8 10
Clipping ratio, γclip (dB)

Figure 2.18: PAPR of clipped signal as a function of the clipping ratio. (N = 64)
 41

fore PAPRclip = PAPRs . As γclip → 0, the peak and average powers converge, thus
PAPRclip → 0 dB. For the region 3 dB < γclip < 6.5 dB, sclip (t) is clipped so the
peak power is A2max = γclip
2 P . However, the clipping is mild so the average power is
s
2 P /P = γ 2 .
approximately the same as s(t); therefore, PAPR clip ≈ γclip s s clip

Clipping is clearly an effective technique at reducing the PAPR. The question is, does
the PAPR reduction translate into reduced total degradation? Figure 2.19 compares the
total degradation curves of the unclipped system [from Figure 2.12(b)] with the clipped
system. Interestingly, the unclipped results are shown to provide a lower bound for
the clipped, reduced PAPR, system results. The clipper is shown to increase both the
minimum total degradation and the optimum backoff. For example, using the clipping
ratio γclip = 3 dB for the M = 8 case increases the TD min by 0.2 dB; using γclip = 2 dB
increases TDmin by 1.2 dB. For M = 16, the γclip = 4 dB result is nearly identical to the
unclipped result; γclip = 3 dB increases the degradation by 1.2 dB, and the TD curve
associated with γclip = 2 dB is beyond the viewing range of the figure. For M = 2, 4 the
PAPR reducing clipping yields nearly identical results as the unclipped system.

10

M = 16
8
Total degradation (dB)

6
M =8

4 Ideal PA
Unclipped
M = 2, 4 Clipped: γclip = 4 dB
3 dB
2 2 dB

0
0 2 4 6 8 10
Input power backoff, IBO (dB)

Figure 2.19: A comparison of the total degradation curves of clipped and unclipped
M -PSK/OFDM systems. (N = 64)

Thus the effectiveness of a PAPR reduction scheme should be measured not only
by its PAPR reducing capabilities but by its effectiveness in reducing total degradation
(which increases range) and reducing the optimum IBO (which increases power amplifier
 42

efficiency). The distortion caused by non-distortionless schemes can outweigh the benefit
of the reduced PAPR. This is clearly shown to be the case for the clipped N = 64 M -
PSK/OFDM systems studied in this section. The clipping is shown to reduce the 0.0001
PAPR by > 1 dB, but this reduction does not translate into increased PA efficiency.

This result brings into question the validity of the claims that clipping is an effec-
tive scheme. In fact, the effectiveness of non-distortionless PAPR reduction schemes in
general is suspect. For these types of techniques it is important to take into account the
effect of the nonlinear power amplifier.

The effectiveness of distortionless PAPR reduction techniques are typically studied

in terms of PAPR reduction and complexity. It would be interesting to also study these
schemes in terms of total degradation. Does a 3 dB reduction in PAPR results in a 3
dB reducing in IBOopt ? What is the resulting minimum total degradation?
Chapter 3

Constant Envelope OFDM

Conventional OFDM systems, even with the use of effective PAPR reduction and/or
power amplifier linearization techniques, typically require more input power backoff than
convention single carrier systems. Therefore, OFDM is considered power inefficient,
which is undesirable particularly for battery-powered wireless systems.

The technique described in the remainder of the thesis takes a different approach to
the PAPR problem. CE-OFDM can be thought of as a mapping of the OFDM signal to
the unit circle, as depicted in Figure 3.1. The instantaneous power of the resulting signal
is constant. Figure 3.2 compares the instantaneous power of the OFDM signal and the
mapped CE-OFDM signal. For the CE-OFDM signal the peak and average powers are
the same, thus the PAPR is 0 dB.

Signal
Unit circle

⇒
CE-OFDM
OFDM

Figure 3.1: The CE-OFDM waveform mapping.

43
 44

5
OFDM
CE-OFDM

Instantaneous signal power

0
0 0.2 0.4 0.6 0.8 1
Normalized time

Figure 3.2: Instantaneous signal power.

The mapping is performed with an angle modulator, specifically, a phase modulator.

That is, the OFDM signal is used to phase modulate the carrier. This is in contrast to
conventional OFDM which amplitude modulates the carrier. To see this, consider the
baseband OFDM waveform
N
XX
m(t) = Ii,k qk (t − iTB ) (3.1)
i k=1

where {Ii,k } are the data symbols and {qk (t)} are the orthogonal subcarriers. For con-
ventional OFDM the baseband signal is up-converted to bandpass as
n o
y(t) = < m(t)ej2πfc t
(3.2)
= Am (t) cos [2πfc t + φm (t)] ,

where Am (t) = |m(t)| and φm (t) = arg[m(t)]. For real-valued m(t), φ m (t) = 0 and y(t)
is simply an amplitude modulated signal. (For complex-valued m(t), y(t) can be viewed
as an amplitude single-sideband modulation.) For CE-OFDM, m(t) is passed through a
phase modulator prior to up-conversion. The baseband signal is

s(t) = ejαm(t) , (3.3)

 45

where α is a constant. The bandpass signal is

n o
y(t) = < s(t)ej2πfc t
n o
= < ejαAm (t) exp[jφm (t)] ej2πfc t
n o (3.4)
= < e−αAm (t) sin φm (t) ej[2πfc t+αAm (t) cos φm (t)]

= e−αAm (t) sin φm (t) cos [2πfc t + αAm (t) cos φm (t)] .

For real-valued m(t),

y(t) = cos [2πfc t + αm(t)] . (3.5)

Therefore y(t) is a phase modulated signal.

CE-OFDM can also be thought of as a transformation technique, as shown in Fig-

ure 3.3. At the transmitter, the high PAPR OFDM signal is transformed into a low
PAPR signal prior to the power amplifier. At the receiver, the inverse transformation is
performed prior to demodulation.
Transmitter

Transform

OFDM m(t) Phase s(t) Power To

modulator modulator amplifier channel

Receiver

Inverse

From Phase OFDM

channel demodulator demodulator

transform

Figure 3.3: Basic concept of CE-OFDM.

As mentioned in Section 2.5, other approaches based on signal transformation have

been suggested. In particular, [215, 329, 569–571] suggest a companding transform. The
companded signal has an increased average power and thus a lower peak-to-average
power ratio than conventional OFDM. The PAPR is still large relative to single carrier
modulation, however. The advantage of the phase modulator transform is that the
resulting signal has the lowest achievable peak-to-average power ratio of 0 dB.
 46

The idea of transmitting OFDM by way of angle modulation isn’t entirely new.
In fact, Harmuth’s 1960 paper suggest transmitting information by orthogonal time
functions with “amplitude or frequency modulation, or any other type of modulation
suitable for the transmission of continuously varying [waveforms]” [202]. Using existing
FM infrastructure for OFDM transmission has been suggested in [76, 77, 575]. These
papers don’t consider the PAPR implications, however. Two conference papers, [101]
and [506], on the other hand, suggest using a phase modulator prior to the power amplifier
for PAPR mitigation—though intriguing, these papers lack a solid theoretical foundation
and ignore fundamental signal properties such as the signal’s power density spectrum.

The origin of this work, which is independent of the previous references, stems from
work done at the US Navy’s spawar Systems Center, (San Diego, CA). Mike Geile,
a principle engineer at Nova Engineering, (Cincinnati, OH), which is the contractor of
the OFDM component for JTRS (Joint Tactical Radio System), suggested a low PAPR
enhancement to OFDM by phase modulation. The motivation is to reduce the 6 dB
backoff used in the JTRS radio.

Transmitting OFDM with phase modulation raises several fundamental questions.

What is the power density spectrum of the modulation? How is the signal space af-
fected? What is the optimum AWGN performance? What is the performance of a phase
demodulator receiver (Figure 3.3)? How does the system perform in a frequency-selective
fading channel? These questions, and others, are addressed here. First, the CE-OFDM
modulation is defined.

3.1 Signal Definition

As indicated by (3.4), CE-OFDM requires a real-valued OFDM message signal, that

is, φm (t) = 0. Therefore the data symbols in (3.1) are real-valued:

Ii,k ∈ {±1, ±3, . . . , ±(M − 1)}. (3.6)

This one dimensional constellation is known as pulse-amplitude modulation (PAM). Thus

the data symbols are selected from an M -PAM set. The subcarriers {q k (t)} must also
 47

be real-valued. Three possibilities are considered: half-wave cosines,


cos πkt/TB , 0 ≤ t < TB ,

qk (t) = (3.7)
0,

otherwise,

for k = 1, 2, . . . , N ; half-wave sines,


sin πkt/TB ,

0 ≤ t < TB ,
qk (t) = (3.8)
0,

otherwise,

for k = 1, 2, . . . , N ; and full-wave cosines and sines,


N



 cos 2πkt/TB , 0 ≤ t < TB ; k ≤ 2,

qk (t) = sin 2π(k − N/2)t/TB , 0 ≤ t < TB ; k > N (3.9)
 2,


0,

otherwise.

For each case, the subcarrier orthogonality condition holds:


Z (i+1)TB Eq ,

k1 = k2 ,
qk1 (t − iTB )qk2 (t − iTB )dt = (3.10)
iTB 0,

k1 6= k2 ,

where Eq = TB /2.

In terms of implementation, (3.7) can be computed with a discrete cosine transform

(DCT); (3.8) with a discrete sine transform (DST); and (3.9) by taking the real part
of a discrete Fourier transform (DFT), or equivalently by taking a 2N -point DFT of a
conjugate symmetric data vector (see Appendix A.)

The baseband CE-OFDM signal is

s(t) = Aejφ(t) , (3.11)

where A is the signal amplitude. The phase signal during the ith block is written as
N
X
φ(t) = θi + 2πhCN Ii,k qk (t − iTB ), iTB ≤ t < (i + 1)TB , (3.12)
k=1

where h is referred to as the modulation index, and θ i is a memory term (to be described
below). The normalizing constant, C N , is set to
s
2
CN ≡ , (3.13)
N σI2
 48

where σI2 is the data symbol variance:

M
1 X
σI2 2
(2l − 1 − M )2

= E |Ii,k | =
M
l=1 (3.14)
M2 −1
= ,
3
assuming equally likely signal points, that is, P (I i,k = l) = 1/M , l = ±1, ±3, . . . , ±(M −
1), for all i and k. Consequently, the phase signal variance is
( Z (i+1)TB )
1
σφ2 = E [φ(t) − θi ]2 dt
TB iTB
Z (i+1)TB X N X N
(2πh)2 2
= E {Ik1 Ik2 } qk1 (t − iTB )qk2 (t − iTB )dt (3.15)
TB N σI2 iTB
k1 =1 k2 =1
N Z
(2πh)2 2 X TB 2 2
= σI qk (t)dt = (2πh)2 ,
TB N σI2 0
k=1

which is only a function of the modulation index. The signal energy is

Z (i+1)TB
Es = |s(t)|2 dt = A2 TB , (3.16)
iTB

and the bit energy is

Es A2 TB
Eb = = . (3.17)
N log2 M N log 2 M

The term θi is a memory component designed to make the modulation phase-continuous.

At the ith signaling interval boundary, the phase discontinuity is

ci = φ(iTB − ) − φ(iTB + ),  → 0. (3.18)

Since qk (t) = 0 for t ∈

/ [0, TB ), it follows that
N
X
φ(iTB − ) = K Ii−1,k Ae (k), (3.19)
k=1

and
N
X
φ(iT + ) = K Ii,k Ab (k), (3.20)
k=1

where K ≡ 2πhCN , Ab (k) = qk (0) and Ae (k) = qk (TB − ),  → 0. Therefore,

N
X
ci = θi−1 − θi + K [Ii−1,k Ae (k) − Ii,k Ab (k)] . (3.21)
k=1
 49

To guarantee continuous phase, that is, c i = 0, the memory term is set to

N
X
θi ≡ θi−1 + K [Ii,k Ab (k) − Ii−1,k Ae (k)] . (3.22)
k=1

Notice that θi depends on θi−1 ; the OFDM signal at the beginning of the ith block,
PN PN
k=1 Ii,k Ab (k); and the OFDM signal at the end of the (i−1)th block, k=1 Ii−1,k Ae (k).
The recursive relationship can be written as
∞ X
X N
θi = K [Ii−l,k Ab (k) − Ii−1−l,k Ae (k)] . (3.23)
l=0 k=1

Thus, the memory term is a function of all data symbols during and prior to the ith
block.

Figure 3.4 plots the phase discontinuities {c i } at the boundary times t = iTB , i =
0, 1, . . . , 49. In Figure 3.4(a), ci is plotted for memoryless modulation, that is, θ i = 0,
for all i; therefore, ci = K N
P
k=1 [Ii−1,k Ae (k) − Ii,k Ab (k)]. Figure 3.4(b) shows that the
phase discontinuities are eliminated with the use of memory as defined in (3.22).

1.5 1.5

1 1
Phase discontinuity, ci
Phase discontinuity, ci

0.5 0.5

0 0

−0.5 −0.5

−1 −1

−1.5 −1.5
0 10 20 30 40 50 0 10 20 30 40 50
Normalized time, t/TB Normalized time, t/TB

(a) Without memory. (b) With memory.

Figure 3.4: Phase discontinuities.

The benefit of continuous phase CE-OFDM is a more compact signal spectrum. This
property is studied further in Section 3.2. A second consequence of the memory terms
is the entire unit circle is used for the CE-OFDM phase modulation. This is illustrated
in Figure 3.5 which plots continuous phase CE-OFDM signal samples on the complex
 50

Unit circle

Starting point

(a) L = 1 (b) L = 100

Figure 3.5: Continuous phase CE-OFDM signal samples, over L blocks, on the complex
plane. (2πh = 0.7)

plane. The modulation index is 2πh = 0.7. Figure 3.5(a) shows signal samples over
L = 1 block, where the phase signal occupies about one-half the unit circle. Viewing
samples over L = 100 blocks, Figure 3.5(b) shows that the phase signal occupies the
entire unit circle.

3.2 Spectrum

CE-OFDM is a complicated nonlinear modulation and a general closed-form expres-

sion for the power density spectrum is not available. The approach taken in [34], [421, pp.
207–217] to calculate the power spectrum of conventional CPM signals can be applied
to CE-OFDM. The Fourier transform of the average autocorrelation function results in
a two-dimensional definite integral. The problem is there are N sinusoidal phase pulses
in CE-OFDM, versus a single phase pulse as in CPM. This makes the integrand very
jagged for all but trivial values of N , and numerical integration algorithms (for example,
those in [328, 419]) fail to converge in a timely manner. Insight can be gained by taking
this approach, however. It can be shown that memoryless modulation (θ i = 0) results
in spectral lines at the frequencies f k = k/TB , k = 0, ±1, ±3, . . . [17]. Using memory as
defined by (3.22) eliminates these lines.

Since the Fourier transform approach isn’t computationally feasible, other techniques
are required to understand the CE-OFDM spectrum. The simplest is with the Taylor
 51

P∞
expansion ex = n=0 x
n /n!. The CE-OFDM signal, with θi = 0, can be written as

s(t) = Aejσφ m(t)

∞  (3.24)
(jσφ )n
X 
=A mn (t),
n=0
n!

where
N
XX
m(t) = CN Ii,k qk (t − iTB ) (3.25)
i k=1
is the normalized OFDM message signal. The effective double-sided bandwidth, defined
as the twice the highest frequency subcarrier, of m(t) is
N N
W =2× = . (3.26)
2TB TB
The bandwidth of s(t) is at least W : in (3.24), the n = 0 term contains no information
and thus has zero bandwidth; the n = 1 term is information bearing and has bandwidth
W ; the n = 2 term has a bandwidth 2W ; and so on. Thus, due to the n = 1 term, the
bandwidth of s(t) is at least W , and depending on the modulation index the effective
bandwidth can be greater than W .

The power density spectrum, Φs (f ), can be easily estimated by the Welch method
of periodogram averaging [526]. The result, Φ̂s (f ) ≈ Φs (f ), is used to calculate the
fractional out-of-band power,
Rf Rf
0 Φs (x)dx 0 Φ̂s (x)dx ˆ
FOBP(f ) = ≈ = FOBP(f ), (3.27)
0.5Ps 0.5Ps
R∞
where Ps = −∞ Φs (f )df = Es /TB = A2 is the signal power. Figure 3.6 shows estimated
fractional out-of-band power curves for N = 64 and various 2πh. Due to the normalizing
constant CN these curves are valid for any M . The dashed lines represent the RMS
bandwidth,
Brms = σφ W = 2πhN/TB . (3.28)

The RMS (root-mean-square) bandwidth is obtained by borrowing a result from analog

angle modulation [423, pp. 340–343] [437], which assumes a Gaussian message signal;
for large N , the OFDM waveform is well modeled as such (see Section 2.2). The results
in Figure 3.6 shows that Brms accounts for at least 90% of the signal power.

As defined in (3.28), the RMS bandwidth can be less than W , but, as shown by the
Taylor expansion in (3.24), the CE-OFDM bandwidth is at least W . A more suitable
 52

100
ˆ
FOBP(f )
Brms

10−1

2πh
10−2
2.0
Fractional out-of-band power

1.8

1.6
10−3
1.4

1.2
10−4
1.0

0.8
10−5
0.6

0.4
10−6

0.2

10−7 0 0.5 1 1.5 2

Normalized frequency, f /W

Figure 3.6: Estimated fractional out-of-band power. (N = 64)

bandwidth is thus
Bs = max(2πh, 1)W. (3.29)

Figure 3.7 plots Bs versus 2πh, and compares it with the 90–99% bandwidths as de-
termined by the Welch method. Notice that (3.29) is an accurate 90–92% bandwidth
measure for 2πh ≥ 1.0. For small modulation index, B s is a conservative bandwidth.
With 2πh = 0.4, for example, (3.29) accounts for 99.8% of the signal power (from Figure
3.6).

Figure 3.8 compares spectral estimates for CE-OFDM signals with the three sub-
carrier modulations from (3.7), (3.8) and (3.9). The modulation index is 2πh = 0.6.
Memoryless, non-continuous phase CE-OFDM is compared to continuous phase CE-
OFDM (the continuous phase examples are prefixed with “CP”). The estimates are also
 53

2.8
Bs
Welch: 90%
2.6 92%
95%
99%
Normalized double-sided bandwidth, B/W 2.4

2.2

1.8

1.6

1.4

1.2

0.8

0.6

0.4

0.2

0
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Modulation index, 2πh

Figure 3.7: Double-sided bandwidth as a function of modulation index. (N = 64)

compared to the Abramson spectrum [1]:

∞
X
2
ΦAb (f ) = A an Un (f ), (3.30)
n=0

where 2
e−σφ σφ2n
an = , (3.31)
n!
and 



 δ(f ), n = 0,

Un (f ) = Φm (f ), n = 1, (3.32)


Φm (f ) ∗n Φm (f ),

n > 1.


P∞ n
The weighting factors {an } are Poisson distributed, and n=0 an = 1; ∗ denotes the
3
n-fold convolution, for example x(t) ∗ x(t) = x(t) ∗ x(t) ∗ x(t); and Φm (f ) is the power
 54

Welch estimate
0 terms from (3.30)
Φ̂Ab (f )

−10
n=2

−20
n=3
Power spectrum (dB)

−30 n=1

n=4
DCT
DFT

−40

−50 DST
CP-DFT

−60

CP-DCT
−70

−80−3 −2 −1 0 1 2 3
Normalized frequency, f /W

Figure 3.8: Power density spectrum. (N = 64, 2πh = 0.6)

density spectrum of the message signal m(t) according to (3.25):

N      
TB X 2 k 2 k
Φm (f ) = sinc f− TB + sinc f+ TB , (3.33)
2N 2TB 2TB
k=1

where 
1,

x = 0,
sinc(x) = (3.34)
 sin πx ,

otherwise.
πx
R∞
The functions {Un (f )} have the property: −∞ Un (f )df = 1, for all n [1]. Therefore
the nth term in (3.30) has an an × 100% contribution to the overall spectrum. For
example, the carrier component, represented as δ(f ), has a fractional contribution of
2 2 2
e−σφ ; Φm (f ) ∗ Φm (f ) has a fractional contribution (e −σφ σφ4 )/2; and so on. Notice that
2
for 2πh = 0.2, the carrier component accounts for e −0.2 ×100 ≈ 96% of the signal power.
(This explains why the 90–92% curves at 2πh = 0.2 in Figure 3.7 are equal zero.)
 55

Figure 3.8 plots the n = 1, 2, 3, 4 terms in (3.30), and the resulting sum
4
X
2
Φ̂Ab (f ) = A an Un (f ) ≈ ΦAb (f ). (3.35)
n=0

The Abramson spectrum is shown to match all estimates over the range |f /W | ≤ 1. For
|f /W | > 1, the spectral height depends on the overall smoothness of the phase signal. For
example, DST has a continuous phase [with or without memory since A b (k) = Ae (k) = 0,
for all k] and has a lower out-of-band power than memoryless DFT, which isn’t phase-
continuous. Memoryless DFT results in a slightly smoother phase than memoryless DCT
since one-half of the subcarriers have zero-crossings at the signal boundaries [A b (k) =
Ae (k) = 0, for k = N/2 + 1, . . . , N , and Ab (k) = Ae (k) = 1, otherwise] while DCT
doesn’t [Ab (k) = Ae (k) = 1, for all k]. The smoothest phase results from CP-DCT
which, unlike DST and CP-DFT, has a first derivative equal to zero at the boundary
times t = iTB . Consequently, the CP-DCT is the most spectrally contained.

Figure 3.9 shows estimated fractional out-of-band power curves that correspond to
the signals in Figure 3.8. For reference, conventional OFDM is also plotted. Notice that
the 99% spectral containment at f /W = 0.5 is the same for each signal. The continu-
ous phase CE-OFDM signals are the most spectrally contained and are shown to have
better than 99.99% containment at f /W = 1.25. Over the range 0.5 ≤ f /W ≤ 0.8, This

100
CE-OFDM
OFDM
10−1 Brms

10−2
Fractional out-of-band power

10−3

10−4
DCT
DFT

10−5
DST

10−6 CP-DFT

10−7
CP-DCT

10−8 0 0.5 1 1.5 2 2.5 3

Normalized frequency, f /W

Figure 3.9: Fractional out-of-band power. (N = 64, 2πh = 0.6)

 56

figure shows that the CE-OFDM spectrum has more out-of-band power than conven-
tional OFDM. Since the modulation index controls the CE-OFDM spectral containment,
smaller h can be used if a tighter spectrum is required. The tradeoff is that smaller h
results in worse performance, as will be discussed in the next chapter. Therefore, the
system designer can trade performance for spectral containment, and visa versa.

Figure 3.10 compares CE-OFDM, with CP-DFT modulation over a large range of
modulation index, to conventional OFDM. For 2πh ≤ 0.4 the fractional out-of-band
power of CE-OFDM is always better than OFDM; otherwise CE-OFDM has more out-
of-band power for at least some frequencies f /W > 0.5. The 2πh = 2.0 example has a
broad spectrum, greater than OFDM over all frequencies. Notice that the shape of the
spectrum appears Gaussian shaped. This is due to the fact that for a large modulation
index, the higher-order terms in (3.32) dominate. They are Gaussian shaped due to the
multiple convolutions of (3.33). The shape of “wideband FM” signals is well covered in
the classical works of [1, 341, 437, 472].

100
CE-OFDM
OFDM

10−1

2πh
10−2
2.0
Fractional out-of-band power

1.8
1.6
10−3
1.4

1.2
10−4
1.0

0.8
10−5
0.6

0.4
10−6
0.2

10−7 0 0.5 1 1.5 2

Normalized frequency, f /W

Figure 3.10: CE-OFDM versus OFDM. (N = 64)

 57

Finally, Figure 3.11 compares CE-OFDM and OFDM with nonlinear power ampli-
fication. The OFDM curves (from Figure 2.9) require > 6 dB backoff to avoid spectral
broadening. The CE-OFDM signals have a bandwidth that depends only on the modu-
lation index and are not effected by the PA nonlinearity.
100
OFDM, TWTA
OFDM, Ideal
CE-OFDM

10−1
IBO (dB)
0
2
Fractional out-of-band power

4
6
10−2

10−3

10−4
2πh 0.4 0.5 0.6 0.7

10−5 0 0.5 1 1.5 2

Normalized frequency, f /W

Figure 3.11: CE-OFDM versus OFDM with nonlinear PA. (N = 64)

Chapter 4

Performance of Constant
Envelope OFDM in AWGN

In this chapter the basic performance properties of CE-OFDM are studied. The
baseband signal, represented by (3.11) and (3.12), is up-converted and transmitted as
the bandpass signal
n o
sbp (t) = < s(t)ej2πfc t = A cos [2πfc t + φ(t)] , (4.1)

where fc is the carrier frequency. The received signal is

rbp (t) = sbp (t) + nw (t), (4.2)

where nw (t) denotes a sample function of the additive white Gaussian noise (AWGN)
process with power density spectrum Φ nw (f ) = N0 /2 W/Hz. The primary focus of the
chapter is to analyze the phase demodulator receiver, depicted by the block diagram
below. An expression for the bit error rate (BER) is derived by making certain high
carrier-to-noise ratio (CNR) approximations. The analytical result is then compared
against computer simulation and it is shown to be accurate for BER < 0.01. It is also

rbp (t) Bandpass Phase OFDM To

filter demodulator demodulator detector

Figure 4.1: Phase demodulator receiver.

58
 59

demonstrated that with the use of a phase unwrapper, the receiver is insensitive to phase
offsets caused by the channel and/or by the memory terms {θ i }.

The phase demodulator receiver is a practical implementation of the CE-OFDM

receiver and is therefore of practical interest. However, it isn’t necessarily optimum,
since the optimum receiver is a bank of M N matched filters [421, p. 244], one for each
potentially transmitted signal. In Section 4.2 a performance bound and approximation
for the optimum receiver is derived; and then in Section 4.3, the performance of the
phase demodulator receiver is compared to the optimum result. It is shown that under
certain conditions the phase demodulator receiver has near-optimum performance.

In Section 4.4 CE-OFDM’s spectral efficiency versus performance is compared to

channel capacity. Finally, the chapter is concluded in Section 4.5 with a comparison
between CE-OFDM and conventional OFDM in terms of power amplifier efficiency, total
degradation, and spectral containment.

4.1 The Phase Demodulator Receiver

The phase demodulator receiver essentially consists of a phase demodulator followed

by a conventional OFDM demodulator. Figure 4.2 shows the model used in this analy-
sis. The received signal is first passed through a front-end bandpass filter, centered at
the carrier frequency fc , which limits the bandwidth of the additive noise. Then the
bandpass signal is down-converted to r(t), sampled, and processed in the discrete-time
domain. The conversion from rbp (t) to r(t) is described first1 , making use of the following
trigonometric identities:

cos(x − y) − cos(x + y)
sin(x) sin(y) = , (4.3)
2
sin(x + y) + sin(x − y)
sin(x) cos(y) = , (4.4)
2
cos(x + y) + cos(x − y)
cos(x) cos(y) = , (4.5)
2
sin(x + y) − sin(x − y)
cos(x) sin(y) = . (4.6)
2

1
This is the standard model used for representing received baseband signals, and more discussion of
the model can be found in [421, sec. 4.1], [624, sec. 5.5], among other places.
 60

Lowpass
filter

2 cos(2πfc t)

rbp (t) Bandpass u(t) r(t) r[i] Phase OFDM

filter t = iTsa demodulator demodulator
−2 sin(2πfc t)
j
Lowpass
filter

Figure 4.2: Bandpass to baseband conversion.

The output of the bandpass filter is

u(t) = sbp (t) + nbp (t), (4.7)

where
nbp (t) = nc (t) cos(2πfc t) − ns (t) sin(2πfc t) (4.8)

is the result of passing nw (t) through the bandpass filter. The terms n c (t) and ns (t)
are referred to as the in-phase and quadrature components of the narrowband noise,
respectively, and have the power density spectrum

N0 , |f | ≤ Bbpf /2,

Φnc (f ) = Φns (f ) = (4.9)
0,

|f | > Bbpf /2,

where Bbpf is the bandwidth of the bandpass filter. Note that B bpf is assumed to be
sufficiently large so sbp (t) is passed through the front-end filter with negligible distortion
[421, pp. 157–158]. Writing sbp (t) in the form

sbp (t) = sc (t) cos(2πfc t) − ss (t) sin(2πfc t), (4.10)

where sc (t) = A cos[φ(t)] and ss (t) = A sin[φ(t)], the filter output can then be written as

u(t) = [sc (t) + nc (t)] cos(2πfc t) − [ss (t) + ns (t)] sin(2πfc t). (4.11)
 61

The output of the top (in-phase) branch of the down-converter is 2

rc (t) = LP {u(t) × 2 cos(2πfc t)}

= LP{[sc (t) + nc (t)] + [sc (t) + nc (t)] cos(4πfc t)
(4.12)
− [ss (t) + ns (t)] sin(4πfc t)}
= sc (t) + nc (t),

where LP{·} denotes the lowpass component of its argument (i.e., double-frequency terms
are rejected) [624, p. 364]. Likewise, the output of the bottom (quadrature) branch is

rs (t) = LP {u(t) × −2 sin(2πfc t)}

= LP{−[sc (t) + nc (t)] sin(4πfc t) + [ss (t) + ns (t)]
(4.13)
− [ss (t) + ns (t)] cos(4πfc t)}
= ss (t) + ns (t).

The two are combined to obtain

r(t) = s(t) + n(t), (4.14)

where s(t) is the lowpass equivalent CE-OFDM signal from (3.11), and

n(t) = nc (t) + jns (t) (4.15)

is the lowpass equivalent representation of the bandpass white noise, n bp (t) [421, p. 158].
The power density spectrum of n(t) is [421, p. 158]

N0 , |f | ≤ Bn /2,

Φn (f ) = (4.16)
0,

|f | > Bn /2,

where Bn = Bbpf is the noise bandwidth. The corresponding autocorrelation of n(t)

is [421, p. 158]
sin πBn τ
φn (τ ) = N0 . (4.17)
πτ

The continuous-time receive signal is then sampled at the rate f sa = 1/Tsa samp/s
to obtain the discrete-time signal 3

r[i] = s[i] + n[i], i = 0, 1, . . . , (4.18)

2
Here, ideal phase coherence and frequency synchronization is assumed. In Section 4.1.2 the effect of
channel phase offsets is considered.
3
Perfect timing synchronization is assumed.
 62

Phase demodulator

r[i] FIR Phase To OFDM

arg(·) unwrapper demodulator
filter

Figure 4.3: Discrete-time phase demodulator.

where s[i] = s(t)|t=iTsa and n[i] = n(t)|t=iTsa . As discussed in Section 2.4.2, the noise
samples {n[i]} are assumed independent:

σ 2 ,

i1 = i2 ,
n
E {n[i1 ]n[i2 ]} = (4.19)
0,

i1 6= i2 ;

and therefore the sampling rate is f sa = Bn , and σn2 = fsa N0 .

The discrete-time phase demodulator studied in this thesis is shown in Figure 4.3.
The finite impulse response (FIR) filter is optional, but has been found effective at
improving performance; arg(·) simply calculates the arctangent of its argument; and the
phase unwrapper is used to minimize the effect of phase ambiguities. As will be shown,
the unwrapper makes the receiver insensitive to phase offsets caused by the channel
and/or by the memory terms.

The output of the phase demodulator is processed by the OFDM demodulator which
consists of the N correlators, one corresponding to each subcarrier. This correlator bank
is implemented in practice with the fast Fourier transform.

4.1.1 Performance Analysis

In this section a bit error rate approximation is derived for the phase demodulator
receiver. Although the receiver operates in the discrete-time domain, it is convenient to
analyze it in the continuous-time domain. The angle of the received signal is
N
X
arg[r(t)] = θi + 2πhCN Ii,k qk (t − iTB ) + ξ(t), (4.20)
k=1

iTB ≤ t < (i + 1)TB , where

 
N (t) sin [Θ(t) − φ(t)]
ξ(t) = arctan (4.21)
A + N (t) cos [Θ(t) − φ(t)]
is the corrupting noise [624, p. 416]. The terms N (t) and Θ(t) in (4.21) are the envelope
and phase of n(t).
 63

The kth correlator in the OFDM demodulator computes

Z (i+1)TB
1
arg[r(t)]qk (t − iTB )dt = Si,k + Ni,k + Ψi,k . (4.22)
TB iTB
The signal term is
(i+1)TB
1
Z
Si,k = [φ(t) − θi ]qk (t − iTB )dt
TB iTB
N
(i+1)TB X
2πhCN
Z
= Ii,n qn (t − iTB )qk (t − iTB )dt (4.23)
TB iTB n=1
s
2πhCN 1
= Ii,k Eq = 2πh Ii,k .
TB 2N σI2

The noise term is

(i+1)TB
1
Z
Ni,k = ξ(t)qk (t − iTB )dt. (4.24)
TB iTB
For example, with DST subcarrier modulation (3.8),
Z (i+1)TB
1
Ni,k = ξ(t) sin [πk(t − iTB )/TB ] dt, (4.25)
TB iTB
which can be viewed as a Fourier coefficient of ξ(t) at f = k/2T B Hz. As TB → ∞,
the variance of the coefficient is proportional to the power density spectrum function
evaluated at f = k/2TB [442, pp. 41–43]. It is well known that, given a high CNR, the
noise at the output of a phase demodulator has a power density spectrum [423, p. 410]
N0
Φξ (f ) ≈ , |f | ≤ W/2, (4.26)
A2
where, from (3.26), W = N/TB is the effective bandwidth of φ(t). Moreover, for high
CNR, ξ(t) is well modeled as a sample function of a zero mean Gaussian process. There-
fore, Ni,k is approximated as a zero mean Gaussian random variable with variance [442,
pp. 41–43]
1 1 N0
var{Ni,k } ≈ Φξ (f )|f =k/2TB ≈ . (4.27)
2TB 2TB A2
This result is the same for DCT and DFT subcarrier modulation.

The third term in (4.22), Ψi,k , is expressed as

Z (i+1)TB
1
Ψi,k = θi qk (t − iTB )dt. (4.28)
TB iTB
Since Z TB
qk (t)dt = 0, k = 1, 2, . . . , N, (4.29)
0
 64

for DCT and DFT modulations [(3.7), (3.9)], Ψ i,k = 0 and therefore has no effect on sys-
tem performance. This highlights an important observation: DST subcarrier modulation
(3.8) is inferior to DCT and DFT since Ψ i,k = 0 isn’t guaranteed.

The symbol error rate is computed by determining the probability of error for each
signal point in the M -PAM constellation. For the M − 2 inner points, the probability of
error is
Pinner = P (|Ni,k | > d) = 2P (Ni,k > d), (4.30)

where s
1
d = 2πh . (4.31)
2N σI2
[Notice that (4.30) is not averaged over i nor k since var{N i,k }, as approximated by (4.27),
is a constant.] Due to the Gaussian approximation applied to the random variable N i,k ,
Z ∞
1
exp −x2 / 2N0 /(2A2 TB ) dx
 

Pinner ≈ 2 p
d 2πN0 /(2A2 TB )
Z ∞
1
√ exp −x2 /2 dx


=2 (4.32)
d[N0 /(2A2 TB )]−0.5 2π
s ! r !
A2 TB 6 log 2 M Eb
= 2Q 2πh = 2Q 2πh .
N0 N σI2 M 2 − 1 N0

For the two outer points, the probability of error is

1
Pouter = P (Ni,k > d) = Pinner . (4.33)
2

Therefore, the overall symbol error rate is

M −2 2
SER = Pinner + Pouter
M M
(4.34)
  r !
M −1 6 log 2 M Eb
≈2 Q 2πh .
M M 2 − 1 N0

Notice that for 2πh = 1, (4.34) is equivalent to the SER for conventional M -PAM [483,
pp. 194–195]. For high SNR, the only significant symbol errors are those that occur in
adjacent signal levels, in which case the bit error rate is approximated as [483, p. 195]
  r !
SER M −1 6 log 2 M Eb
BER ≈ ≈2 Q 2πh . (4.35)
log2 M M log2 M M 2 − 1 N0
 65

4.1.2 Effect of Channel Phase Offset

Suppose the channel imposes a phase offset of φ 0 . The received signal is then

r(t) = s(t)ejφ0 + n(t). (4.36)

The angle of r(t) is

N
X
arg[r(t)] = θi + 2πhCN Ii,k qk (t − iTB ) + φ0 + ξ(t), (4.37)
k=1

iTB ≤ t < (i + 1)TB . Which is identical to (4.20) with the addition of the channel offset
term. The kth correlator is the same as (4.22), except the third term is
(i+1)TB
1
Z
Ψi,k = [θi + φ0 ]qk (t − iTB )dt = 0. (4.38)
TB iTB

Therefore, the phase offset due to the channel has no impact on performance, and the
analytical approximation in (4.35) is applicable.

Figure 4.4 compares the performance of N = 64, M = 2 CE-OFDM with phase offset
{(θi + φ0 ) ∈ [0, 2π)}, and without (θi + φ0 = 0). The former is referred to as System
1 (S1), the later as System 2 (S2). The system is computer simulated with a sampling
rate fsa = JN/TB , where J = 8 is the oversampling factor 4 . For Eb /N0 ≥ 10 dB and
2πh ≤ 0.5, S1 and S2 are shown to have identical performance. For these cases the
analytical approximation (4.35) closely matches the simulation results for BER < 0.01.
With the 2πh = 0.7 example, S1 is shown to have a 1 dB performance loss compared
to S2. In this case, the analytical approximation is shown to be overly optimistic. This
demonstrates a limitation of the phase demodulator receiver: for a large modulation
index and low signal-to-noise ratio, the phase demodulator has difficulty demodulating
the noisy samples. The performance of S1 is slightly worse than S2 since the output
of the phase demodulator, the arg(·) block in Figure 4.3, has more phase jumps since
the received phase crosses the π boundary more frequently. Proper phase unwrapping
is therefore required. However, phase unwrapping a noisy signal is a difficult problem
and the unwrapper makes mistakes. As a result the performance degrades slightly. For
a smaller modulation index, the unwrapper works perfectly and the performance of S1
isn’t degraded.
4
Also, the FIR filter (see Figure 4.3) has length Lfir = 11 and normalized cutoff frequency fcut /W =
0.2. See Section 4.1.4 for more on the filter design.
 66

100
System 1
System 2
Approx (4.35)

10−1

Bit error rate 10−2

10−3 0.7 0.5 0.3 0.2 0.1

2πh

10−4

10−5 0 5 10 15 20 25 30
Signal-to-noise ratio per bit, Eb /N0 (dB)

Figure 4.4: Performance with and without phase offsets. System 1 (S1) has phase offsets
{(θi + φ0 ) ∈ [0, 2π)}, and System 2 (S2) doesn’t (θ i + φ0 = 0). [M = 2, N = 64, J = 8]

4.1.3 Carrier-to-Noise Ratio and Thresholding Effects

The high-CNR approximation made in (4.26), which leads to the BER approximation
(4.35), is a standard technique for analyzing phase demodulator receivers [423, 624]. A
well-known characteristic of such receivers is: at low CNR, below a threshold value, the
approximation is invalid and system performance degrades drastically. In this section, the
CNR is defined and the threshold effect for CE-OFDM is observed by way of computer
simulation.

The CNR at the output of the analog front end, r(t), is

A2
CNR = , (4.39)
Pn
where A2 is the carrier power, and
Z ∞
Pn = Φn (f )df = Bn N0 (4.40)
−∞

is the noise power. From (3.17), the carrier power can be written in the form
Eb N log 2 M
A2 = ; (4.41)
TB
 67

thus
(Eb /N0 )N log2 M
CNR = . (4.42)
TB Bn
Since the noise samples are assumed independent [see (4.19)],

Bn = fsa = JN/TB , (4.43)

and (4.42) reduces to

(Eb /N0 ) log 2 M
CNR = . (4.44)
J
Therefore, the carrier-to-noise ratio is proportional to E b /N0 and M , and inversely pro-
portional to the oversampling factor.

A commonly accepted threshold CNR for analog FM systems is 10 dB [472, pp.

120–138], [501, pp. 87–91]. This threshold level is studied in the following two figures.
In Figure 4.5, simulation results for an M = 8, N = 64, J = 8, 2πh = 0.5 system are
compared to (4.35). In subfigures (a) and (b) the system is below and above the 10
dB threshold, respectively. Clearly, above CNR = 10 dB, the system is observed to be
above threshold, with simulation results closely matching the analytical approximation.
Below 10 dB, the performance begins to deviate from (4.35); and for CNR < 5 dB, the
performance quickly degrades to a bit error rate of 1/2. Figure 4.6 shows results for more
values of 2πh. For each case, 10 dB can be considered an appropriate threshold level.
There is, however, a transition region—that is, a region where the system is useless,
with a BER of 1/2, to where the system is above threshold. This transition region is
difficult to study analytically. Gaining more insight into this issue is a subject for future
investigation.
 68

Bit error rate

Bit error rate

10−2

Simulation
Approx (4.35)
10−1

Simulation
Approx (4.35)

−2 0 2 4 6 8 10 10−310 11 12 13 14 15 16
Carrier-to-noise ratio (dB) Carrier-to-noise ratio (dB)

(a) Below 10 dB threshold. (b) Above 10 dB threshold.

Figure 4.5: Threshold effect at low CNR. (M = 8, N = 64, J = 8, 2πh = 0.5)

Simulation
2πh Approx (4.35)
10−1

0.2

0.4
Bit error rate

Bit error rate

10−1

0.6 10−2

Simulation
Approx (4.35) 0.8 0.6 0.4 2πh 0.2
0.8

10−2
−2 0 2 4 6 8 10 10−310 12 14 16 18 20 22 24
Carrier-to-noise ratio (dB) Carrier-to-noise ratio (dB)

(a) Below 10 dB threshold. (b) Above 10 dB threshold.

Figure 4.6: Threshold effect at low CNR, various 2πh. (M = 8, N = 64, J = 8)

 69

4.1.4 FIR Filter Design

The FIR filter preceding the phase demodulator (see Figure 4.3) can improve per-
formance. Figure 4.7 shows BER simulation results of an M = 2, N = 64, J = 8,
2πh = 0.5 system. The SNR is held constant at E b /N0 = 10 dB. The filter, designed
using the window technique described in [422, pp. 623–630], has a length 3 ≤ L fir ≤ 101
and a normalized cutoff frequency 0 < f cut /W ≤ 1. Hamming windows are used5 . The
performance without a filter is shown to be BER = 0.05, while the analytical approxi-
mation (4.35) is BER = 0.012. For fcut /W ≥ 0.4 all the filtered results are shown to be
better than the unfiltered result. The filters with L fir > 5 and fcut /W > 0.5 are shown
to have roughly the same performance. The higher-order filters, which have a narrower
transition bands, require fcut /W > 0.5 to yield good performance. This is explained
by noting that the (single-sided) signal bandwidth is at least W/2 Hz. Therefore, the
higher-order filters with fcut /W < 0.5 distort the signal. Notice that the L fir = 11 filter
has equally good performance so long as f cut /W ≥ 0.1. This is due to the wide transition
band of the lower-order filter.

Lfir = 3
5
7
9
11
21
31
61
101
10−1
Bit error rate

No filter

Approx (4.35)

10−2 0 0.2 0.4 0.6 0.8 1

Normalized cutoff frequency, fcut /W

Figure 4.7: Performance for various filter parameters L fir , fcut /W .

(M = 2, N = 64, J = 8, 2πh = 0.5 and Eb /N0 = 10 dB)

5
It has been observed that the window type has negligible impact on performance.
 70

Lfir , fcut /W

−20
9, 0.7

3, 0.1
Magnitude response (dB)

−40
101, 0.7 9, 0.1

−60

31, 0.1

−80

−100
0 0.5 1 1.5 2 2.5 3
Normalized frequency, f /W

Figure 4.8: Magnitude response of various Hamming FIR filters.

The figure above shows the magnitude response of the various Hamming FIR filters.
The filters with relatively flat response over |f /W | ≤ 0.5 result in good performance.
The Lfir = 31, fcut /W = 0.1 example is shown to not have this property, and, as shown
in Figure 4.7, has worse BER performance than the other filters.

Figure 4.9 compares the performance of binary (M = 2) CE-OFDM with and without
the FIR filter. The Lfir = 11, fcut /W = 0.2 filter is used. These results show that the
filter becomes important for larger modulation index: for 2πh = 0.1 the filtered and
unfiltered results are the same; for 2πh = 0.3 the filtered performance is a fraction of
a dB better than the unfiltered; for 2πh = 0.7 there is a 2 dB improvement in the
range 10−3 < BER < 10−5 . Notice the error floor developing below 10 −5 . This is a
consequence of imperfect phase demodulation. The filter lowers the error floor resulting
in a 9 dB improvement at BER = 10−6 .
 71

100
Without FIR filter
With FIR filter
Approx (4.35)

10−1

10−2
Bit error rate

10−3 2πh 0.7 0.3 0.1

10−4

10−5

10−6 0 5 10 15 20 25 30
Signal-to-noise ratio per bit, Eb /N0 (dB)

Figure 4.9: CE-OFDM performance with and without FIR filter.

(M = 2, N = 64, J = 8)

4.2 The Optimum Receiver

As mentioned in the introduction to this chapter, the phase demodulator receiver is

a practical implementation, but not necessarily optimum. In this section, the optimum,
yet impractical, CE-OFDM receiver is studied. Results obtained here are used in the
following section to compare the phase demodulator receiver to optimum performance.

During each block one of M N CE-OFDM signals is transmitted. Consider the mth
bandpass signal
N
" #
(m)
X
sm (t) = A cos 2πfc t + θ0 + K Ik qk (t) , 0 ≤ t < TB , (4.45)
k=1

N
where K = 2πhCN . The set of all possible signals, {s m (t)}M
m=1 , is determined by the
(m) (m) (m) N
set of all possible data symbol vectors {I (m) = [I1 , I2 , . . . , IN ]}M
m=1 . The optimum
 72

receiver, as shown in Figure 4.10, correlates the received signal, r bp (t) = sm (t) + nw (t),
with each potentially transmitted signal. The detector then selects the largest result [421,
pp. 242–247].

R TB
0
(·)dt

s1 (t)
R TB
0 (·)dt

s2 (t)
Select
Received the Output
signal rbp (t) largest decision
.. ..
. .

R TB
0
(·)dt

sM N (t) Sample
at t = TB

Figure 4.10: The optimum receiver.

4.2.1 Performance Analysis

It is desired to obtain an analytical expression for the bit error probability 6 , P (bit error).
However, there are two other probabilities to consider:

P (signal error) and P (symbol error) .

The first is the probability that the output of the optimum receiver is in error—that is,
the receiver selects a different signal than the one transmitted. The second is the data
symbol error probability. Determining exact expressions for the above probabilities is
intractable for large N . However, upperbounds and approximations can be derived in a
straightforward way, as described below.
6
The bit error probability is used interchangeably with the bit error rate. Likewise for the symbol
error probability and symbol error rate.
 73

An upperbound for P (signal error) is [373]:

Z ∞h
1 M N −1
i
P (signal error) ≤ √ 1 − [1 − Q(y)] ×
2π −∞
2 
(4.46)
  s
 1 2E (1 − λ) 
s
exp − y −  dy.
 2 N0 

The above expression is the probability of detection error for M N signals with equal
correlation −1 ≤ λ ≤ 1. Therefore, it provides an upperbound given that

λ = ρmax = max ρm,n , (4.47)

m,n; m6=n

where ρm,n is the normalized correlation between s m (t) and sn (t):

TB
1
Z
ρm,n = sm (t)sn (t)dt. (4.48)
Es 0

An approximation for P (signal error) is [421, p. 288]

s 
d2min 
P (signal error) ≈ Kd2 Q  , (4.49)
min 2N0

where Kd2 is the number of neighboring signal points having the minimum squared
min

Euclidean distance
d2min = min d2m,n , (4.50)
m,n; m6=n

where Z TB
d2m,n = [sm (t) − sn (t)]2 dt (4.51)
0
is the squared Euclidean distance between s m (t) and sn (t). This quantity is related to
the signal correlation as
d2m,n = 2Es (1 − ρm,n ), (4.52)

thus
d2min = 2Es (1 − ρmax ). (4.53)

Therefore to obtain the performance bound (4.46) and the approximation (4.49) the
signal correlation properties must be studied, and in particular ρ max must be determined.
The normalized correlation between the mth and nth signal, as a function of the phase
 74

constant K = 2πhCN , is
TB
1
Z
ρm,n (K) = sm (t)sn (t)dt
Es 0
N
" #
TB
A2
Z
(m)
X
= cos 2πfc t + θ0 + K Ik qk (t) ×
Es 0 k=1
" N
# (4.54)
(n)
X
cos 2πfc t + θ0 + K Ik qk (t) dt
k=1
N
" #
TB
A2
Z X
= cos 2K ∆m,n (k)qk (t) dt,
2Es 0 k=1

(m) (n)
where ∆m,n (k) = 0.5[Ik − Ik ]. The double frequency term is ignored since f c  1/TB
is assumed. Notice that for k where ∆ m,n (k) = 0, the data symbols are the same, and
these indices don’t contribute to the correlation. Therefore
D
" #
A 2 TB
Z X
ρm,n (K) = cos 2K ∆m,n (kd )qk (t) dt, (4.55)
2Es 0
d=1

where {kd }D
d=1 are the indices where the data symbols differ, that is, ∆ m,n (kd ) 6= 0, and
D is the total number of differences. Writing (4.55) in exponential form yields
D
( " #)
A 2 TB
Z X
ρm,n (K) = < exp j2K ∆m,n (kd )qk (t) dt
2Es 0
d=1
(D ) (4.56)
A 2 TB
Z Y
= < exp [j2K∆m,n (kd )qk (t)] dt.
2Es 0
d=1

To proceed, the DCT modulation (3.7) is assumed. Making use of the Jacobi-Anger
expansion [580],
∞
X
ja cos b
e = Ji (a)eji(b+π/2) , (4.57)
i=−∞

where Ji (a) is the ith-order Bessel function of the first kind, (4.56) is written as
" ∞ ∞
A 2 TB
Z X X
ρm,n (K) = < ···
2Es 0
i1 =−∞ iD =−∞
#
Ji1 [2K∆m,n (k1 )] × · · · × JiD [2K∆m,n (kD )]ejσ(i) dt
(4.58)
TB ∞ ∞
A2
Z X X
= ··· Ji1 [2K∆m,n (k1 )]×
2Es 0 i1 =−∞ iD =−∞

· · · × JiD [2K∆m,n (kD )] cos[ω(i) + ψ(i)]dt,

 75

πt PD π PD
where σ(i) = ω(i) + ψ(i), ω(i) ≡ TB d=1 id kd and ψ(i) ≡ 2 d=1 id . Index values that
result in ω(i) 6= 0 have no contribution, so (4.58) simplifies to
D
XY
ρi,j (K) = Ji0i,d [2K∆m,n (kd )] cos[ψ(i0i )], (4.59)
i d=1

where i0i ≡ [i0i,1 , . . . , i0i,D ], i = 1, 2, . . ., represent the vectors whereby ω(i 0i ) = 0. This
result is the same for DST modulation except ψ(i 0i ) = 0. For DFT modulation, (4.59) is
slightly different since both sinusoids and cosinusoids are used as subcarriers.

For D = 1,
ρm,n (K) = J0 [2K∆m,n (k1 )]. (4.60)

Therefore the correlation is simply the 0th-order Bessel function. Figure 4.11(a) plots
(4.60) for |∆m,n (k1 ) = 1|. Also plotted is the envelope of the 0th-order Bessel func-
tion [580, p. 121]. Note that ρm,n (K) doesn’t depend on the subcarrier frequency
fk1 = k1 /TB , k1 ∈ {1, 2, . . . , N }, just on the magnitude of the difference |∆ m,n (k1 )| ∈
{1, 2, . . . , (M − 1)}.

For CE-OFDM signals of interest,

ρmax = J0 (2K). (4.61)

Figure 4.11(b) plots all unique ρm,n (K) for M = 2, N = 8 DCT subcarrier modulation.
Notice that the largest correlation function is associated with D = 1. For any given
signal, there are N other signals with D = 1: therefore, K d2 = N , and from (4.49),
min

the probability of signal error is approximated as

s 
d2min 
P (signal error) ≈ Kd2 Q 
min 2N0
p  (4.62)
= NQ Es [1 − ρmax ]/N0
p 
≈ NQ Es [1 − J0 (2K)]/N0 .

A minimum distance signal error results in one data symbols error. Therefore, the symbol
error probability is approximated as
P (signal error) p 
P (symbol error) ≈ ≈Q Es [1 − J0 (2K)]/N0 . (4.63)
N
For M = 2, one symbol error corresponds to one bit error. For M > 2, a symbol error
can result in 1 to log 2 M bit errors. Assuming each outcome is equally likely, a symbol
 76

0.5
p
1/πK
ρm,n (K)

−0.5
0 1 2 3 4 5
K

(a) D = 1.

J0 (2K)

0.8
ρm,n (K)

0.6

0.40 0.1 0.2 0.3 0.4 0.5

K

(b) All unique ρm,n (K) for M = 2, N = 8 DCT modulation.

Figure 4.11: Correlation functions ρ m,n (K).

 77

1 Plog2 M
error results in log2 M i i = 0.5(log 2 M + 1) bit errors. Thus

0.5(log 2 M + 1)
P (bit error) ≈ P (symbol error)
log2 M
(4.64)
0.5(log 2 M + 1) p 
≈ Q Es [1 − J0 (2K)]/N0 .
log2 M

The bit error probability is bounded by noting that P (bit error) ≤ P (signal error),
and using (4.46) with λ = ρmax = J0 (2K):
Z ∞h
1 N
i
P (bit error) ≤ √ 1 − [1 − Q(y)]M −1 ×
2π −∞
2 
(4.65)
  s
 1 2Es [1 − J0 (2K)]  
exp − y − dy.
 2 N0 

Figure 4.12 shows simulation results of the optimum receiver for M = 2 and N = 8.
The number of correlators at the receiver is therefore 2 8 = 256. Two values of modulation
index are plotted: 2πh = 0.3 and 2πh = 0.7 which corresponds to K = 0.15 and
K = 0.35. The upperbound (4.65) is shown to be within 3 dB of the simulated results
for high SNR. The analytical approximation (4.64) is shown to be very accurate.

100

10−1

10−2
Bit error rate

10−3
Approx (4.64)
Bound (4.65)
Simulation
10−4
0.7 2πh 0.3

10−5

10−6 0 3 6 9 12 15 18 21
Signal-to-noise ratio per bit, Eb /N0 (dB)

Figure 4.12: CE-OFDM optimum receiver performance. (M = 2, N = 8)

 78

4.2.2 Asymptotic Properties

In Figure (4.13) each correlation function is plotted for M = 2, N = 4 DCT modu-

lation. The functions are shown to be bounded by
r
1
ρm,n (K) ≤ ρmax (K) ≤, (4.66)
πK
the envelope of the 0th-order Bessel function. Therefore,
r !
1
d2m,n (K) ≥ d2min (K) ≥ 2Es 1− . (4.67)
πK
Notice that as K → ∞ the CE-OFDM signals become orthogonal. The phase modulator
thus drastically alters the signal space. Prior to the phase demodulator, the OFDM
signal space is described by 2N dimensions (2 per subcarrier). At the output of the
phase modulator, the space is transformed into a M N -dimensional space (due to the
linear independence of the signal set [421, p. 164]); and as the modulation index becomes
very large, a M N -dimensional orthogonal space. However, from (3.29), the bandwidth
tends to infinity as 2πh → ∞.
1

0.5
ρm,n (K)

p
1/πK

−0.5
0 1 2 3 4 5
K

Figure 4.13: All unique ρm,n (K) for M = 2, N = 4 DCT modulation.

4.3 Phase Demodulator Receiver versus Optimum

Figure 4.14 shows simulation results for the phase demodulator receiver with N = 64
and for various modulation index values 2πh and modulation order M . The simulation
 79

100
Approx (4.35)
Approx (4.64)
Simulation
10−1

Bit error rate

10−2

10−3 M , 2πh 8, 1.2 16, 0.8 16, 0.2

10−4 2, 0.3 4, 0.2

10−5 5 10 15 20 25 30 35 40
Signal-to-noise ratio per bit, Eb /N0 (dB)

Figure 4.14: Phase demodulator receiver versus optimum. (N = 64)

results are compared to the analytical approximation (4.35) and the optimum receiver
approximation (4.64). All curves are shown to be essentially identical for BER < 0.01.
This implies that the phase demodulator receiver is nearly optimum. For this to be
true, the phase demodulator must perfectly invert the phase modulation done at the
transmitter, and the noise at the output of the phase demodulator must be “white” and
Gaussian. That is, the OFDM demodulator is optimum given that the input, φ(t) + ξ(t),
is comprised of the transmitted message signal plus an AWGN corrupting signal. As
shown by (4.26), ξ(t) is approximately “white”. The probability density function of ξ(t)
samples is represented by the well-known form [421, p. 268]
Z ∞  2
y + A2 − 2yA cos x

y
pξ (x) = exp − dy, (4.68)
0 2πσn2 2σn2

where σn2 = Bn N0 is the power of the noise signal n(t). Figure 4.15 compares (4.68) to
the Gaussian probability density function. The SNR per bit is E b /N0 = 30 dB. This
shows that ξ(t) is well approximated as Gaussian, and near optimum performance of the
phase demodulator receiver is expected.
 80

pξ (x)
100 Gaussian

Probability density function, p(x)

10−5

10−10

10−15

10−20
−1.5 −1 −0.5 0 0.5 1 1.5
x

Figure 4.15: Noise samples PDF versus Gaussian PDF. (E b /N0 = 30 dB)

4.4 Spectral Efficiency versus Performance

In the previous sections, it is shown that the performance of CE-OFDM is deter-

mined by the modulation index, which, as shown in Section 3.2, also controls the signal
bandwidth. In this section, the spectral efficiency (b/s/Hz) versus performance (E b /N0
to achieve a target bit error rate) is plotted for a variety of CE-OFDM signals. The
results are compared to channel capacity.

It is first demonstrated that CE-OFDM with modulation index 2πh > 1 can out-
perform the underlying M -PAM subcarrier modulation. Figure 4.16 shows simulation
results7 for M = 2, 4, 8 and 16. The bit error rate is plotted against the SNR per bit on
the bottom x-axis and the carrier-to-noise ratio on the top x-axis. The viewable range
is such that CNR ≥ 5 dB. Notice that for M ≥ 4 and 2πh > 1, CE-OFDM outperforms
M -PAM. This is predicted by (4.35), since for 2πh = 1.0, the expression is equal to the
performance of M -PAM, and for 2πh > 1.0, it is better than M -PAM. For CE-OFDM
to operate in the region 2πh > 1, the carrier-to-noise ratio must be above threshold.
7
The oversampling factor is J = 8 for M = 2, 4 and 8, and J = 16 for M = 16. The FIR filter has
length Lfir = 11 and a normalized cutoff frequency 0.2 cycles per sample for M = 2, 4 and 16, and 0.3
cycles per sample for M = 8.
 81

Carrier-to-noise ratio (dB) Carrier-to-noise ratio (dB)

5 10 15 20 5 10 15 20 25
10−1 10−1
Simulation Simulation
(4.35) (4.35)
4-PAM

10−2 10−2
Bit error rate

Bit error rate

10−3 10−3

10−4 2πh ∈ {0.5† , 0.4, . . . , 0.1‡ } 10−4 2πh ∈ {1.0† , 0.9, . . . , 0.1‡ }

10−5 16 18 20 22 24 26 28 30 10−5 15 20 25 30 35
Signal-to-noise ratio per bit, Eb /N0 (dB) Signal-to-noise ratio per bit, Eb /N0 (dB)

(a) M = 2. (b) M = 4.

Carrier-to-noise ratio (dB) Carrier-to-noise ratio (dB)

5 10 15 20 25 30 35 5 10 15 20 25 30 35 40
10−1 10−1
Simulation Simulation
(4.35) (4.35)
8-PAM 16-PAM

10−2 10−2
Bit error rate

Bit error rate

10−3 10−3

10−4 2πh ∈ {1.5† , 1.2, 1.0, 0.9, . . . , 0.1‡ } 10−4 2πh ∈ {2.0† , 1.5, 1.2, 1.1, . . . , 0.1‡ }

10−5 10 15 20 25 30 35 40 10−5 15 20 25 30 35 40 45
Signal-to-noise ratio per bit, Eb /N0 (dB) Signal-to-noise ratio per bit, Eb /N0 (dB)

(c) M = 8. (d) M = 16.

Figure 4.16: Performance of M -PAM CE-OFDM. (N = 64, †=leftmost curve,

‡=rightmost curve)
 82

To plot the spectral efficiency versus performance, the data rate must be defined,
which for uncoded CE-OFDM is
N log 2 M
R= b/s. (4.69)
TB
Using (3.29) as the effective signal bandwidth, the spectral efficiency is
R log2 M
R/B = = b/s/Hz. (4.70)
Bs max(2πh, 1)
Figure 4.17 shows result for M = 2, 4, 8 and 16. The target bit error rate is 0.0001. For
reference the channel capacity is also plotted, which is expressed as [421, p. 387]
 
C Eb
C = B log 2 1 + , (4.71)
B N0
or equivalently,
Eb 2C/B − 1
= . (4.72)
N0 C/B

10
M =2
M =4 M = 16: 2πh = 2.0, 1.8, . . . , 0.6
7 M =8
6 M = 16
Capacity M = 8: 2πh = 1.4, 1.2, . . . , 0.4
5

4
Spectral efficiency (b/s/Hz)

M = 4: 2πh = 1.0, 0.8, . . . , 0.2

M = 2: 2πh = 0.5, 0.4, 0.3, 0.2

0.5-1.6 0 5 10 15 20 25
Performance: Eb /N0 (dB) to achieve 0.0001 bit error rate

Figure 4.17: Spectral efficiency versus performance.

There are two main observations to be made. First, for a fixed modulation index,
CE-OFDM has improved spectral efficiency with increase modulation order M at the cost
of performance degradation. For example consider 2πh = 0.4. The spectral efficiency
 83

is 1, 2 and 3 b/s/Hz for M = 2, 4 and 8, respectively. However, M = 4 requires 4 dB

more power than M = 2, and M = 8 requires nearly 5 dB more power than M = 4.
This type of spectral efficiency/performance tradeoff is the same for conventional linear
modulations such as M -PAM, M -PSK and M -QAM [421, p. 282].

The second observation is that CE-OFDM can have both improvements in spectral
efficiency and in performance. Compare M = 2, 2πh = 0.5 with M = 4, 2πh = 1.0,
for example. The spectral efficiency doubles in the later case while also having a 2 dB
performance gain. Conventional CPM systems also have the property of increase spectral
efficiency and performance [14]. However, with CPM the receiver complexity increases
drastically with M (due to phase trellis decoding), which isn’t the case for CE-OFDM.

4.5 CE-OFDM versus OFDM

The total degradation, as defined in Section 2.4.2, is

TD(IBO) = SNRPA (IBO) − SNRAWGN + IBO, [in dB]

where SNRAWGN is the required signal-to-noise ratio required to achieve a target bit error
rate, SNRPA (IBO) is the required SNR when taking into account the nonlinear power
amplifier at a given backoff. Applying the PA model from Section 2.3 to CE-OFDM, the
input signal is
sin (t) = A exp[jφ(t)], (4.73)

and the output is

sout (t) = G(A) exp j[φ(t) + Φ(A)] . (4.74)

The instantaneous nonlinearity results in a constant amplitude and a constant phase

shift. Therefore the PA has no impact on the CE-OFDM performance and no backoff is
needed. The total degradation for CE-OFDM is defined as

TD = SNRPM − SNRsub , (4.75)

where SNRsub is the required SNR for the underlying subcarrier modulation and SNR PM
is the required SNR for the phase modulated CE-OFDM system. By this definition,
the total degradation can be negative since, as observed in Figure 4.16, CE-OFDM can
outperform the underlying subcarrier modulation at the price of lower spectral efficiency.
 84

Figure 4.18 compares CE-OFDM with conventional OFDM in terms of PA efficiency,

total degradation and spectral containment. Binary modulation is used in both systems.
The target BER is 10−5 and the number of subcarriers is N = 64. Both the SSPA and
TWTA models are considered. The lowest TD for the TWTA system is 10.5 dB at 8 dB
backoff, which corresponds to an 8% efficiency as shown in Figure 4.18(a). At this backoff
level, the 99.5% bandwidth occupancy is roughly the same as undistorted ideal OFDM
as shown in Figure 4.18(c). For the SSPA model, the lowest TD is 3.8 dB at IBO = 1 dB.
In this case, the PA efficiency is improved to 40% but the bandwidth requirement is 73%
more than ideal OFDM. Since CE-OFDM has a constant envelope, the PA can operate
at IBO = 0 dB thus maximizing amplifier efficiency. The total degradation is 5 dB for
2πh = 0.6 and the corresponding bandwidth requirement is 26% more than ideal OFDM.
For 2πh = 0.4, the total degradation is 8 dB but the bandwidth reduces to f /W = 0.98
which is 8% less than ideal OFDM. This shows that the modulation index for CE-OFDM
can be chosen accordingly to balance performance and bandwidth. Also, since the PA
imposes no additional distortion on the CE-OFDM signal, the resulting spectrum can be
well contained with no power backoff and at the same time have optimal PA efficiency.
 85

50

Class-A PA efficiency, ηA (%)

45
40
35
30
25
20
15
10
5
00 1 2 3 4 5 6 7 8 9 10
Input power backoff, IBO (dB)

(a) PA efficiency.

16
OFDM, TWTA
14 OFDM, SSPA
Total degradetion (dB)

OFDM, ideal
12 CE-OFDM: 2πh = 0.4
10 0.5
0.6
8
6
4
2
00 2 4 6 8 10
Input power backoff, IBO (dB)

(b) Total degradation for target BER 10−5 .

2 OFDM, TWTA
OFDM, SSPA
99.5% bandwidth, f /W

1.8 OFDM, ideal

CE-OFDM: 2πh = 0.4
1.6 0.5
0.6
1.4

1.2

0.8
0 2 4 6 8 10
Input power backoff, IBO (dB)

(c) Spectral containment.

Figure 4.18: A comparison of CE-OFDM and conventional OFDM. (M = 2, N = 64)

Chapter 5

Performance of CE-OFDM in
Frequency-Nonselective Fading
Channels

In this chapter, performance analysis of the phase demodulator receiver is extended

to fading channels. The lowpass equivalent representation of the received signal is

r(t) = αejφ0 s(t) + n(t) (5.1)

where s(t) is the CE-OFDM signal according to (3.11), α and φ 0 is the channel amplitude
and phase, respectively, and n(t) is the complex Gaussian noise term represented in
R∞
(4.15). The received signal can be written as r(t) = −∞ h(τ )s(t − τ )dτ + n(t) [see (1.2),
(2.4)], where the channel impulse response is h(τ ) = αe jφ0 δ(τ ). In the frequency domain,
the channel is H(f ) = F{h(τ )}(f ) = αe jφ0 , and is thus constant at all frequencies—that
is, the channel is frequency nonselective.

In the previous chapter only the simple case of α = 1 (i.e. no fading) was considered.
In this chapter the channel amplitude is treated as a random quantity. Such a channel
model, since it’s frequency nonselective, is commonly referred to as flat fading. The
signal-to-noise ratio per bit for a given α is

Eb
γ = α2 , (5.2)
N0

86
 87

and the average SNR per bit is [421, p. 817]

 Eb
γ̄ = E{γ} = E α2 . (5.3)
N0
It is desired to calculate the bit error rate at a given γ̄, denoted here as BER(γ̄). This
quantity depends on the statistical distribution of γ. For channels with a line-of-sight
(LOS) component, the probability density function of γ is [483, p. 102]

−K
  " s #
(1 + KR )e R (1 + KR )x KR (1 + KR )x
pγ (x) = exp − I0 2 , x ≥ 0, (5.4)
γ̄ γ̄ γ̄

where I0 (·) is the 0th-order modified Bessel function of the first kind, and

ρ2
KR = (5.5)
2σ02

is the Rice factor: ρ2 and 2σ02 represent the power of the LOS and scatter component,
respectively [401, p. 40]. For channels without a line-of-sight, ρ → 0 and γ is Rayleigh
distributed [483, p. 101]:
 
1 x
pγ (x) = exp − , x ≥ 0. (5.6)
γ̄ γ̄
To obtain BER(γ̄), the conditional BER is averaged over the distribution of γ [421, p.
817]: Z ∞
BER(γ̄) = BER(x)pγ (x)dx. (5.7)
0

In Section 4.1.1 it is shown that

√

BER(x) ≈ c1 Q c2 x , (5.8)
p
where c1 = 2(M − 1)/(M log 2 M ) and c2 = 2πh 6 log 2 M/(M 2 − 1), so long as the
system is above threshold. For the moment, assume
√
BER(x) = c1 Q(c2 x), for all x ≥ 0. (5.9)

If this were true, the bit error rate for the Ricean channel, described by (5.4), is [483, p.
102]
π/2
c1 (1 + KR ) sin2 θ
Z
BERRice (γ̄) = ×
π 0 (1 + KR ) sin2 θ + c22 γ̄/2
(5.10)
KR c22 γ̄/2
 
exp − dθ,
(1 + KR ) sin2 θ + c22 γ̄/2
 88

and for the Rayleigh channel, as described by (5.6), [483, p. 101]

s !
c1 c22 γ̄/2
BERRay (γ̄) = 1− . (5.11)
2 1 + c22 γ̄/2

However, as discussed in Section 4.1.3, the bit error rate of CE-OFDM, as a result of
the threshold effect, isn’t simply expressed by the Q-function for all values of SNR.
Consequently (5.10) and (5.11) are not generally accurate.

Figure 5.1(a) compares simulation results 1 to (5.10) for an M = 8, N = 64 system in

the Ricean channel with KR = 10 dB. For 2πh = 0.6 the simulation result closely matches
(5.10) for γ̄ > 15 dB. For lower values of γ̄, (5.10) is overly optimistic since the system is
more likely to experience channel fades which take the system below threshold—in which
case the bit error rate isn’t accurately represented by the Q-function, that is, (5.9) is
false. For the 2πh = 1.8 example, (5.10) is overly optimistic by at least 3 dB for all
values of γ̄. This is due to the inaccuracy of the Q-function for large modulation index
cases (see Figure 4.4, for example).

100 100
Simulation Simulation
Approx (5.10) Approx (5.11)

10−1 10−1
Bit error rate

Bit error rate

10−2 10−2

2πh 1.8 0.6 2πh 1.2 0.4

10−3 10−3

10−4 10−4

10−5 0 5 10 15 20 25 30 10−5 0 10 20 30 40 50
Average signal-to-noise ratio per bit, γ̄ (dB) Average signal-to-noise ratio per bit, γ̄ (dB)

(a) M = 8, Ricean KR = 10 dB. (b) M = 4, Rayleigh.

Figure 5.1: Performance of CE-OFDM in flat fading channels. (N = 64)

1
Unless otherwise stated, the simulation parameters—J, Lfir , normalized cutoff frequency, and so
forth—are the same as those used for the result shown in Figure 4.16 (see the footnote in on page 80).
 89

Figure 5.1(b) further illustrates the inaccuracy of assuming (5.9). An M = 4, N = 64

system is simulated in the Rayleigh channel. For the low modulation index case of
2πh = 0.4, (5.11) is somewhat accurate. However, for the large modulation index case
of 2πh = 1.2, (5.11) is shown to be off by 5–7 dB.

A Semi-Analytical Approach

The problem with (5.10) and (5.11) is that the conditional bit error rate, BER(x),
is not accurately described by the Q-function at low SNR and/or for large modulation
index. For a limited range of 2πh (for example, the values shown in Figure 4.16) the
following observation can be made: above a certain SNR, say x 0 , the conditional bit
error rate closely matches the Q-function, that is, (5.8) holds. Therefore (5.7) can be
approximated as
Z x0 Z ∞
BER(γ̄) = BER(x)pγ (x)dx + BER(x)pγ (x)dx
0
x0 Zx0∞ (5.12)
√
Z
≈ BER(x)pγ (x)dx + c1 Q(c2 x)pγ (x)dx.
0 x0
R x0
Determining x0 for a given M and 2πh, and dealing with 0 BER(x)pγ (x)dx in (5.12) are
the problems that remain to obtain an accurate approximation of BER(γ̄). As observed
in Section 4.1.3 [see Figure 4.6(a)], at low SNR the bit error rate is roughly 1/2. Assume
for the moment that BER(x) = 1/2 for x ≤ x 0 ; then

1 x0
Z ∞
√
Z
BER(γ̄) ≈ pγ (x)dx + c1 Q(c2 x)pγ (x)dx. (5.13)
2 0 x0

This simplified model, referred to as a two-region model since the conditional BER is split
into two regions, is illustrated in Figure 5.2: below x 0 the BER is 1/2, otherwise the BER
is equal to the Q-function. Also shown is the observed simulation result. Notice that
the two-region model doesn’t account for the transition region in which BER(x) ≈ 1/2
√
to where BER(x) ≈ c1 Q(c2 x). [For more examples of the transition region, see Figure
4.6.] Consequently, (5.13) is not generally accurate, and a more elaborate approach is
required which accounts for the transition region.
 90

Transition region
1

0.5
Conditional bit error rate, BER(x)

0.1

Two-region model
Observed (simulation)
Q-function (4.35)

0.01

x0
Signal-to-noise ratio per bit, x (dB)

Figure 5.2: A simplified two-region model. (M = 8, N = 64, 2πh = 0.6)

This is done by splitting the SNR region 0 ≤ x ≤ x 0 into n sub-regions:

Z x0 Z γ1
BER(x)pγ (x)dx = BER(x)pγ (x)dx+
0 γ0
Z γ2 Z γn (5.14)
BER(x)pγ (x)dx + . . . + BER(x)pγ (x)dx,
γ1 γn−1

where γi > γi−1 , i = 1, 2, . . . , n, γ0 = 0 and γn = x0 . Due to the analytical difficulty

of describing BER(x) over 0 ≤ x ≤ x0 , computer simulation is used. The system is
simulated at SNR values γi , i = 1, 2, . . . , n − 1, to get the result BER i , i = 1, 2, . . . , n − 1.
It is assumed that BER(x) ≈ BERi for γi ≤ x ≤ γi+1 to obtain the approximation
n−1
X Z γi+1 Z ∞
√
BER(γ̄) ≈ BERi pγ (x)dx + c1 Q(c2 x)pγ (x)dx. (5.15)
i=0 γi γn

For SNR in the range 0 ≤ x ≤ γ1 the bit error rate is assumed to be BER 0 = 1/2. Figure
5.3 illustrates the n + 1 regions of (5.15). Notice that for n = 1, (5.15) is equivalent to
(5.13). In other words, (5.15), a (n+1)-region model, is a generalization of the two-region
model (5.13).

CE-OFDM systems are simulated in Rayleigh and Ricean (K R = 3 dB and KR = 10

dB) channels. The values of modulation index are as follows: for M = 2, 2πh ≤ 0.6;
 91

1
BER0 = 1/2

BER1
BER3
BER2
Conditional bit error rate, BER(x)
BER4

..
.

BERn−2
(n + 1)-region model
Observed (simulation)
Q-function (4.35)
BERn−1

0.01

← γ0 = −∞ γ1 γ2 γ3 γ4 ... γn−2 γn−1 γn

Signal-to-noise ratio per bit, x (dB)

Figure 5.3: A (n + 1)-region model. (M = 8, N = 64, 2πh = 0.6)

for M = 4, 2πh ≤ 1.2; for M = 8, 2πh ≤ 1.8; and for M = 16, 2πh ≤ 2.4. The
results are shown in Figure 5.4: the circles represent Rayleigh results; the squares and
triangles represent the Ricean results for K R = 3 dB and KR = 10 dB, respectively.
The solid lines are the results of the semi-analytical approach, (5.15). The transition
region is sampled every 0.5 dB, that is, γ i+1 − γi = 0.5 dB, i = 1, 2, . . . , n − 1; the
starting point is γ1 = −5 dB. Therefore γi = 0.5(i − 1) − 5 dB, i = 1, 2, . . . , n. The
sampling continues until BERn < 0.01. For SNR x ≥ γn the conditional bit error rate
is approximated with the Q-function (5.8). This criteria used for γ n is based on the
observation that, for the modulation index values under consideration, the Q-function is
accurate for BER < 0.01. As shown in the figure, this semi-analytical approach yields
curves for BER(γ̄) that closely match simulation.

Figure 5.5 shows the improvement of (5.15) over (5.10) and (5.11). The semi-
analytical approach closely matches the simulation results, even at low SNR, while (5.10)
and (5.11) are overly optimistic by several dB.

The advantage of the technique described in this section is it gives an accurate

result in a small fraction of the time required for direct simulation. For example, the
 92

(a) M = 2, 2πh = 0.2 (b) M = 2, 2πh = 0.6

100 100

10−1 10−1
Bit error rate

Bit error rate

10−2 10−2

10−3 10−3

10−4 10−4

10−5 0 10 20 30 40 50 10−5 0 10 20 30 40 50
Average SNR per bit, γ̄ (dB) Average SNR per bit, γ̄ (dB)

(c) M = 4, 2πh = 0.4 (d) M = 4, 2πh = 1.2

100 100

10−1 10−1
Bit error rate

Bit error rate

10−2 10−2

10−3 10−3

10−4 10−4

10−5 0 10 20 30 40 50 10−5 0 10 20 30 40 50
Average SNR per bit, γ̄ (dB) Average SNR per bit, γ̄ (dB)

(e) M = 8, 2πh = 0.6 (f) M = 8, 2πh = 1.8

100 100

10−1 10−1
Bit error rate

Bit error rate

10−2 10−2

10−3 10−3

10−4 10−4

10−5 0 10 20 30 40 50 10−5 0 10 20 30 40 50
Average SNR per bit, γ̄ (dB) Average SNR per bit, γ̄ (dB)

(g) M = 16, 2πh = 0.8 (h) M = 16, 2πh = 2.4

100 100

10−1 10−1
Bit error rate

Bit error rate

10−2 10−2

10−3 10−3

10−4 10−4

10−5 0 10 20 30 40 50 10−5 0 10 20 30 40 50
Average SNR per bit, γ̄ (dB) Average SNR per bit, γ̄ (dB)

Figure 5.4: Performance of CE-OFDM in flat fading channels. (Circle=Rayleigh;

square=Rice, K = 3 dB; triangle=Rice, K = 10 dB. Solid line=Semi-analytical curve,
(5.15); points=simulation. N = 64)
 93

100
Rayleigh simulation
Rayleigh approximation (5.11)
Ricean (KR = 3 dB) simulation
10−1 Ricean (KR = 3 dB) approximation (5.10)
Ricean (KR = 10 dB) simulation
Ricean (KR = 10 dB) approximation (5.10)
Bit error rate Semi-analytical technique (5.15)

10−2

10−3

10−4

10−5 0 10 20 30 40 50
Average signal-to-noise ratio per bit, γ̄ (dB)

Figure 5.5: Comparison of semi-analytical technique (5.15) with (5.10) and (5.11).
(M = 4, N = 64, 2πh = 1.2)

simulated Rayleigh result in Figure 5.5 requires about 6 hours of computer time (on a
workstation with 1 gigabytes of memory and a single 3 gigahertz microprocessor). The
semi-analytical result, on the other hand, requires less than 7 s (to obtain {BER i }, and
perform numerical integration): a speed improvement of 4 orders of magnitude.

The disadvantage, however, is that this technique doesn’t yield a closed-form expres-
sion. As of the time of this writing, such a solution, that is general and accurate, doesn’t
seem possible.
Chapter 6

Performance of CE-OFDM in
Frequency-Selective Channels

In this chapter the performance of CE-OFDM in frequency-selective channels is stud-

ied. The channel is time dispersive having an impulse response h(τ ) that can be non-zero
over 0 ≤ τ ≤ τmax , where τmax is the channel’s maximum propagation delay. The received
signal is
Z ∞
r(t) = h(τ )s(t − τ )dτ + n(t)
Z−∞
τmax (6.1)
= h(τ )s(t − τ )dτ + n(t),
0
where s(t) is the CE-OFDM signal according to (3.11) and n(t) is the complex Gaussian
noise term represented by (4.15). The lower bound of integration in (6.1) is due to the
law of causality [401, p. 245]: h(τ ) = 0 for τ < 0. The upperbound is τ max since, by
definition of the maximum propagation delay, h(τ ) = 0 for τ > τ max .

CE-OFDM has the same block structure as conventional OFDM, with a block period,
TB , designed to be much longer than τmax . A guard interval of duration Tg ≥ τmax is
inserted between successive CE-OFDM blocks to avoid interblock interference. At the
receiver, r(t) is sampled at the rate f sa = 1/Tsa samp/s, the guard time samples are
discarded and the block time samples are processed. Using the discrete-time model
outlined in Section 2.1.2, the processed samples are
NX
c −1

rp [i] = r[i] = h[m]s[i − m] + n[i], i = 0, . . . , NB − 1. (6.2)

m=0

94
 95

Note that the discarded samples are {r[i]} −1

i=−Ng . Transmitting a cyclic prefix during
the guard interval makes the linear convolution with the channel equivalent to circular
convolution. Thus
NDFT
X−1
1
rp [i] = H[k]S[k]ej2πik/NDFT , i = 0, . . . , NB − 1, (6.3)
NDFT
k=0

where {H[k]} is the DFT of {h[i]} and {S[k]} is the DFT of {s[i]}. The effect of the
channel can be reversed with the frequency-domain equalizer: a DFT followed by a
multiplier bank, followed by an IDFT. The FDE output is
NDFT
X−1
1
ŝ[i] = Rp [k]C[k]ej2πik/NDFT , i = 0, . . . , NB − 1, (6.4)
NDFT
k=0

where {Rp [k]} is the DFT of the processed samples and {C[k]} are the equalizer correc-
tion terms, which are computed as [463]

1
C[k] = (6.5)
H[k]

for the zero-forcing (ZF) criterion, and

H ∗ [k]
C[k] = (6.6)
|H[k]|2 + (Eb /N0 )−1

for the minimum mean-square error (MMSE) criterion.

Ignoring noise (n[i] = 0), the output of the frequency-domain equalizer using (6.5) is
NDFT
X−1
1
ŝ[i] = H[k]S[k]C[k]ej2πik/NDFT
NDFT
k=0
NDFT
X−1
1 1 j2πik/NDFT
= H[k]S[k] e
NDFT H[k] (6.7)
k=0
NDFT
X−1
1
= S[k]ej2πik/NDFT
NDFT
k=0

= s[i], i = 0, . . . , NB − 1.

Therefore, the ZF frequency-domain equalizer perfectly reverses the effect of the channel.
When noise can’t be ignored, the ZF suffers from noise enhancement. For example, a
fade of −30 dB results in a correction term with gain +30 dB, which corrects the channel
but amplifies the noise by a factor of 1000. The MMSE criterion (6.6) takes into account
 96

the signal-to-noise ratio, making an optimum trade between channel inversion and noise
enhancement. Notice that the MMSE and ZF are equivalent at high SNR:

H ∗ [k] 1
lim C[k]|MMSE = 2
= = C[k]|ZF . (6.8)
Eb /N0 →∞ |H[k]| H[k]

The system under consideration is shown in Figure 6.1. System performance is

estimated by way of computer simulation. The samples {h[i]}, {s[i]} and {n[i]} are
generated then used to calculate the received samples (6.2) which are then processed by
the FDE and the demodulator.

CE-OFDM s(t) r(t) r[i] Remove rp [i] CE-OFDM

h(τ ) FDE
Modulator CP Demodulator

n(t)

Figure 6.1: CE-OFDM system with frequency-selective channel.

The study is separated into two parts. In Section 6.1, the performance of the MMSE
and ZF equalizers are compared over various frequency-selective channels. In Section 6.2,
performance is evaluated for frequency-selective fading channels, in which case {h[i]} is
described statistically. In both sections an N = 64 CE-OFDM system is considered,
with a block period of TB = 128 µs. The subcarrier spacing is 1/T B = 7812.5 Hz and
the mainlobe bandwidth is W = N/TB = 500 kHz. The guard period is Tg = 10 µs,
resulting in a transmission efficiency η t = 128/138 ≈ 0.93. The simulation uses an
oversampling factor J = 8; therefore the sampling rate is f sa = JN/TB = 4 Msamp/s,
and the sampling period is Tsa = 1/fsa = 0.25 µs.

6.1 MMSE versus ZF Equalization

In this section, the performance of CE-OFDM using the MMSE and ZF frequency-
domain equalizers is compared over six frequency-selective channels.

6.1.1 Channel Description

The channel samples {h[i]}, over the corresponding guard interval [0, 10 µs], are
shown in Table 6.1. For Channels A–C the maximum propagation delay is τ max = 0.75
 97

Table 6.1: Channel samples of frequency-selective channels.

Delay (µs) Channel A Channel B Channel C Channel D Channel E Channel F

i τi = iTsa h[i] h[i] h[i] h[i] h[i] h[i]
0 0.00 0.59e+j3.04 0.93e−j1.11 0.71e−j0.77 0.14e+j1.99 0.56e−j0.40 0.62e+j0.67
1 0.25 0.80e−j2.22 0.30e−j2.90 0.70e+j2.00 0.47e−j1.01 0.24e+j0.98 0.47e−j0.95
2 0.50 0 0 0 0.61e+j0.26 0.51e−j0.06 0.33e+j2.58
3 0.75 0.10e−j0.37 0.20e+j2.97 0.07e+j0.98 0.42e−j0.01 0.21e−j2.12 0.22e+j0.10
4 1.00 – – – 0.23e+j1.09 0.24e+j1.14 0.25e−j1.92
5 1.25 – – – 0.10e+j1.00 0.11e+j1.64 0.16e−j0.20
6 1.50 – – – 0.18e+j1.82 0.25e−j1.28 0.14e−j2.30
7 1.75 – – – 0.13e+j2.36 0.12e−j0.93 0.21e−j1.14
8 2.00 – – – 0.13e−j0.60 0.27e+j1.82 0.13e+j0.34
9 2.25 – – – 0.12e+j1.00 0.12e+j1.49 0.16e−j2.43
10 2.50 – – – 0.08e−j2.30 0.15e+j0.15 0.17e+j0.36
11 2.75 – – – 0.09e−j1.91 0.19e+j0.23 0.08e−j0.93
12 3.00 – – – 0.13e+j2.99 0.05e+j2.57 0.06e−j1.08
13 3.25 – – – 0.04e−j1.97 0.07e−j0.17 0.05e+j0.13
14 3.50 – – – 0.08e+j1.05 0.04e+j3.00 0.02e+j3.11
15 3.75 – – – 0.08e+j1.01 0.05e−j1.20 0.07e−j2.81
16 4.00 – – – 0.05e+j1.42 0.09e+j0.54 0.05e−j2.87
17 4.25 – – – 0.06e−j0.18 0.12e−j0.10 0.04e−j1.39
18 4.50 – – – 0.09e+j0.56 0.03e+j0.05 0.01e−j0.89
19 4.75 – – – 0.05e+j0.72 0.03e+j0.96 0.02e−j2.00
20 5.00 – – – 0.01e+j3.13 0.02e−j0.33 0.02e+j2.22
21 5.25 – – – 0.05e+j1.11 0.03e−j1.53 0.03e+j0.92
22 5.50 – – – 0.01e+j2.42 0.01e+j0.29 0.01e−j1.56
23 5.75 – – – 0.02e−j1.92 0.02e+j2.58 0.02e+j0.55
24 6.00 – – – 0.02e−j1.20 0.01e−j1.33 0.01e+j2.83
25 6.25 – – – 0.03e+j2.07 0.02e−j1.96 0.02e+j0.48
26 6.50 – – – 0.01e+j0.17 0.02e+j2.29 0.01e+j2.68
27 6.75 – – – 0.01e−j0.93 0.03e+j2.86 0.01e+j2.03
28 7.00 – – – 0.01e+j2.93 0.01e−j0.14 0.01e−j1.76
29 7.25 – – – 0.02e−j2.91 0.01e−j0.36 0.01e−j2.42
30 7.50 – – – 0.01e−j0.76 0.02e+j1.98 0.01e+j1.11
31 7.75 – – – 0.01e−j1.88 0.01e−j2.38 0.01e+j0.01
32 8.00 – – – 0.01e−j2.96 0.01e+j0.19 0.01e+j0.40
33 8.25 – – – 0.01e−j0.89 0.02e+j2.18 0.01e+j1.69
34 8.50 – – – 0.01e−j1.54 0.01e−j2.41 0.01e−j0.49
35 8.75 – – – 0.01e−j3.01 0.01e−j3.11 0.01e+j2.67
36 9.00 – – – – – –
37 9.25 – – – – – –
38 9.50 – – – – – –
39 9.75 – – – – – –
40 10.0 – – – – – –
 98

µs, which results in Nc = bτmax /Tsa c + 1 = b0.75/0.25c + 1 = 4 samples [see (2.19)].

For Channels D–F, τmax = 8.75 µs, thus Nc = b8.75/0.25c + 1 = 36. The channels are
normalized such that
NX
c −1

|h[i]|2 = 1. (6.9)
i=0

Channels A–C are single realizations of an approximation to the maritime channel

model in [350]. Channels D–F are single realizations of a stochastic model which has an
exponential delay power density spectrum 1 .

Figure 6.2 shows Channel D in the time and frequency domains. In subfigure (a),
|h[i]|2 , that is, the power of the time samples, is plotted. In subfigure (b), |H(f 0 )|2 is
plotted, where [422, p. 256]
NX
c −1
0 0
H(f ) = h[i]e−j2πf i , (6.10)
i=0

is the Fourier transform of h[i]. The x-axis is scaled as [422, p. 24]

f = f 0 fsa Hz, (6.11)

where f 0 is the normalized frequency variable having units cycles/samp [422, p. 16].
Notice that over the signal’s mainlobe frequency range, −250 kHz ≤ f ≤ 250 kHz, the
channel is frequency selective. The magnitude response fluctuates over a 8.5 dB range,
−2.5 dB ≤ |H(f 0 )|2 ≤ 6 dB.

The Fourier transform (6.10) is related to the discrete Fourier transform,

NX
c −1

H[k] = h[i]e−j2πik/NDFT , k = 0, . . . , NDFT − 1, (6.12)

i=0

as
H[k] = H(fk0 ), k = 0, 1, . . . , NDFT − 1, (6.13)

where the discrete set of frequencies {f k0 } are defined as


 k , k = 0, 1, . . . , NDFT
2 ,

0 NDFT
fk ≡ (6.14)
 k − 1, k = NDFT + 1, . . . , NDFT − 1.

NDFT 2

1
Stochastic models are discussed in the next section.
 99

0.4

0.35

0.3

0.25
|h[i]|2

0.2

0.15

0.1

0.05

0 0 1 2 3 4 5 6 7 8 9
Propagation delay, iTsa (µs)

(a) Time domain.

Channel D response, |H(f 0 )|2

10
Equalizer response, ZF
Equalizer response, MMSE: Eb /N0 = 0 dB
10 dB
20 dB
Magnitude response (dB)

−5

−200 −100 0 100 200

Frequency, f = f 0 fsa (kHz)

(b) Frequency domain.

Figure 6.2: Channel D.

 100

Using a DFT size NDFT = JN = NB and noting (6.11), the frequency samples {H[k]}
correspond to the frequencies

k, k = 0, 1, . . . , NDFT
2 ,

0 0 JN TB
fk = fk fsa = fk = (6.15)
TB  k − fsa ,

k= NDFT
+ 1, . . . , . . . , NDFT − 1.
TB 2

Included in Figure 6.2(b) is the response of the MMSE and ZF equalizers. The ZF
response, (6.5), is simply the inverse of the channel. The MMSE response, (6.6), is shown
for Eb /N0 = 0, 10, and 20 dB. Notice that at high SNR the MMSE approaches the ZF
equalizer, which is to be expected from (6.8). For this particular channel the MMSE and
ZF are shown to be equivalent for Eb /N0 ≥ 20 dB.

6.1.2 Simulation Results

The N = 64 CE-OFDM system is simulated over Channels A–F. The modulation

order is M = 2, and different values of the modulation index, h, are selected. Due to
the channel normalization (6.9), the simulation results are compared against the simple
AWGN channel. The results are shown in Figures 6.3–6.8. For each case, |h[i]| 2 is plotted
in subfigure (a); the channel and equalizer frequency-domain responses are plotted in
subfigure (b); and the bit error rate performance results are shown in subfigure (c).

The results for Channel A are shown in Figure 6.3. Of the six test channels, Channel
A is the most mild in terms of its frequency-domain response. The magnitude response
|H(f 0 )|2 spans a 3 dB region in a nearly linearly manner. The equalizers are shown to
effectively correct the channel: the BER curves in Figure 6.3(c) are nearly indistinguish-
able from the simple AWGN curves. Results are plotted for 2πh = 0.1, 0.3 and 0.6. For
the 2πh = 0.6 example at the lower SNR values E b /N0 < 10 dB, the ZF result is shown
to be slightly worse than the MMSE result; for higher values of SNR the performance of
the two equalizers becomes nearly identical. This is to be expected since, as illustrated
in Figure 6.3(b), their frequency response become the same at high E b /N0 .

Results for 2πh = 0.1, 0.2, 0.4 and 0.6 over Channel B are shown in Figure 6.4.
The frequency response of this channel is more severely varying than Channel A. Over
the signal bandwidth, |H(f 0 )|2 spans a 6 dB range. As with the previous example, the
MMSE is shown to slightly outperform the ZF at low SNR (i.e., the 2πh = 0.6 example
for Eb /N0 < 10 dB), but the two equalizers have essentially the same performance at the
 101

0.7

2
0.6

Magnitude response (dB)

0.5 0

0.4
|h[i]|2

−2
0.3
Channel A
0.2 −4 ZF
MMSE: Eb /N0 = 0 dB
10 dB
0.1 20 dB
−6

0 −200 −100
0 1 2 3 4 5 6 7 8 9 0 100 200
Propagation delay, iTsa (µs) Frequency, f = f 0 fsa (kHz)

(a) Time domain. (b) Frequency domain.

10−1
ZF
MMSE
AWGN sim
AWGN approx (4.35)

10−2
Bit error rate

10−3 2πh 0.6 0.3 0.1

10−4 5 10 15 20 25 30
Signal-to-noise ratio per bit, Eb /N0 (dB)

(c) Performance for MMSE and ZF compared to AWGN and (4.35).

Figure 6.3: Channel A results.

 102

0.9
4
0.8

0.7

Magnitude response (dB)

2
0.6
0
0.5
|h[i]|2

0.4
−2
Channel B
0.3 ZF
MMSE: Eb /N0 = 0 dB
−4 10 dB
0.2
20 dB
30 dB
0.1
−6
0 −200 −100
0 1 2 3 4 5 6 7 8 9 0 100 200
Propagation delay, iTsa (µs) Frequency, f = f 0 fsa (kHz)

(a) Time domain. (b) Frequency domain.

10−1
ZF
MMSE
AWGN sim
AWGN approx (4.35)

10−2
Bit error rate

10−3 2πh 0.6 0.4 0.2 0.1

10−4 5 10 15 20 25 30
Signal-to-noise ratio per bit, Eb /N0 (dB)

(c) Performance for MMSE and ZF compared to AWGN and (4.35).

Figure 6.4: Channel B results.

 103

higher SNR values. For BER ≤ 0.001 the degradation caused by the frequency selective
channel, when compared to the simple AWGN result, is slightly less than 1 dB.

Channel C has the most frequency-selective response of the three maritime channel
realizations. As shown in Figure 6.5(b), the magnitude response varies over a 20 dB
range. It is also shown that very high SNR is required for the MMSE response to
approach the ZF response. Over the frequency range −250 kHz ≤ f 0 fsa ≤ −200 kHz,
for example, the two are equivalent only for E b /N0 > 35 dB. This equivalence is also
demonstrated in Figure 6.5(c): for the 2πh = 0.1 example, the ZF performance gradually
approaches the MMSE performance at these high SNR values. Clearly, the large amount
of frequency selectivity of this channel results in a large performance degradation when
compared to the AWGN results. At the bit error rate 0.001, the degradation is 10 dB
for the 2πh = 0.1 case. The improvement of the MMSE is pronounced for 2πh = 0.5.
At the bit error rate 0.001, the MMSE outperforms the ZF by 7 dB, and is only 2 dB
worse than the performance over the simple AWGN channel.

Figures 6.6–6.8 show the results for Channels D–F. As stated earlier, the three chan-
nels are three different realizations of a stochastic model with an exponential delay power
density spectrum. The degree that the each channel varies over the signal bandwidth
progresses from Channel D to Channel F. Channel F, having a 50 dB attenuation at
185 kHz, is the most harsh of the test channels. The results in Figure 6.8(c) show the
dramatic performance degradation as a consequence of the severe frequency selectivity.
An 18 dB loss, compared to the AWGN performance, is experienced for the 2πh = 0.6,
MMSE example at the bit error rate 0.001; the ZF case degrades more than 20 dB further.
A 40 dB loss is suffered for the 2πh = 0.1 and 0.3 cases. These results show that fre-
quency selective channels having deep fades in the signal bandwidth impact performance
greatly.

6.1.3 Discussion and Observations

At this point, several observations can be made. First, the performance of the
equalized CE-OFDM systems studied depends on the amount of frequency selectivity
over the signal bandwidth. For channels with a relatively mild frequency response—
Channels A, B and D, for example—the performance degradation is minor. The noise
enhancement that results from equalizing channels with severe frequency responses—
 104

0.6
Channel C
ZF
20 MMSE: Eb /N0 = 0 dB
0.5 10 dB
20 dB

Magnitude response (dB)

30 dB
0.4 10 35 dB
|h[i]|2

0.3 0

0.2 −10

0.1
−20

0 −200 −100
0 1 2 3 4 5 6 7 8 9 0 100 200
Propagation delay, iTsa (µs) Frequency, f = f 0 fsa (kHz)

(a) Time domain. (b) Frequency domain.

10−1
ZF
MMSE
AWGN sim
AWGN approx (4.35)

10−2
Bit error rate

10−3 2πh 0.5 0.1

10−4 5 10 15 20 25 30 35 40
Signal-to-noise ratio per bit, Eb /N0 (dB)

(c) Performance for MMSE and ZF compared to AWGN and (4.35).

Figure 6.5: Channel C results.

 105

0.4 6

0.35 4

Magnitude response (dB)

0.3
2
0.25 Channel D
0 ZF
MMSE: Eb /N0 = 0 dB
|h[i]|2

0.2 10 dB
−2 20 dB
0.15
−4
0.1

0.05 −6

0 −8 −200 −100
0 1 2 3 4 5 6 7 8 9 0 100 200
Propagation delay, iTsa (µs) Frequency, f = f 0 fsa (kHz)

(a) Time domain. (b) Frequency domain.

10−1
ZF
MMSE
AWGN sim
AWGN approx (4.35)

10−2
Bit error rate

10−3 2πh 0.6 0.2 0.1

10−4 5 10 15 20 25 30 35
Signal-to-noise ratio per bit, Eb /N0 (dB)

(c) Performance for MMSE and ZF compared to AWGN and (4.35).

Figure 6.6: Channel D results.

 106

0.35 10
Channel E
ZF
0.3 MMSE: Eb /N0 = 0 dB
10 dB
5

Magnitude response (dB)

20 dB
0.25 30 dB

0.2
|h[i]|2

0.15

0.1 −5

0.05
−10
0 −200 −100
0 1 2 3 4 5 6 7 8 9 0 100 200
Propagation delay, iTsa (µs) Frequency, f = f 0 fsa (kHz)

(a) Time domain. (b) Frequency domain.

10−1
ZF
MMSE
AWGN sim
AWGN approx (4.35)

10−2
Bit error rate

10−3 2πh 0.6 0.1

10−4 5 10 15 20 25 30 35 40
Signal-to-noise ratio per bit, Eb /N0 (dB)

(c) Performance for MMSE and ZF compared to AWGN and (4.35).

Figure 6.7: Channel E results.

 107

0.4

40 Channel F
0.35 ZF
MMSE: Eb /N0 = 0 dB

Magnitude response (dB)

0.3 10 dB
20 20 dB
30 dB
0.25
|h[i]|2

0.2 0

0.15
−20
0.1

0.05 −40

0 −200 −100
0 1 2 3 4 5 6 7 8 9 0 100 200
Propagation delay, iTsa (µs) Frequency, f = f 0 fsa (kHz)

(a) Time domain. (b) Frequency domain.

10−1
ZF
MMSE
AWGN sim
AWGN approx (4.35)

10−2
Bit error rate

10−3 0.6 0.3 0.1

0.6 0.3 0.1 2πh

10−4 5 10 15 20 25 30 35 40 45 50 55 60 65 70
Signal-to-noise ratio per bit, Eb /N0 (dB)

(c) Performance for MMSE and ZF compared to AWGN and (4.35).

Figure 6.8: Channel F results.

 108

Channels C, E and F—degrades performance dramatically. Second, the complexity of

the frequency-domain equalizers is determined by the DFT size, not by the number of
non-zero channel terms h[i]. This is in contrast to conventional time-domain equalizers
which have a complexity that depends on the number of paths in the multipath channel.

Last, the MMSE equalizer is more complicated than the ZF equalizer since the SNR
per bit, Eb /N0 , must be estimated at the receiver. The results of this study show that
this added complexity doesn’t always translate into improved performance. That is, the
ZF performance is the same as the MMSE performance for many cases—the 2πh ≤ 0.4
cases in Channel B, for example. In other cases, the MMSE performs much better, and
thus estimating Eb /N0 pays substantial dividends—the 2πh = 0.5 case for Channel C
illustrates this point.

As demonstrated in the following section, the MMSE equalizer offers significant im-
provement over the ZF equalizer when averaging performance over many channel real-
izations of a stochastic channel model.

6.2 Performance Over Frequency-Selective Fading Chan-

nels

In contrast to the test channels used in the previous section, which were deterministic
as defined in Table 6.1, the channels used in this section are described statistically.
The mathematical foundation for stochastic time-variant linear channels was pioneered
by Bello [50]; more recently Pätzold’s text, Mobile Fading Channels [401], provides a
excellent treatment of the topic, with a focus on the various aspects of simulation. In the
study here, the widely used assumption of WSSUS (wide-sense stationary uncorrelated
scattering) is applied. Also, it is assumed that the channel is composed of discrete paths,
each having an associated gain and discrete propagation delay. This assumption is based
on the Parsons and Bajwa ellipse model for describing multipath channel geometry [401,
p. 244]. The channel’s impulse response is
L−1
X
h(τ ) = al δ(τ − τl ), (6.16)
l=0
 109

where al is the complex channel gain and τl is discrete propagation delay of the lth path;
the total number of paths is represented by L. The propagation delay differences are

∆τl = τl − τl−1 ≡ Tsa , l = 1, 2, . . . , L − 1. (6.17)

That is, they are set equal to the sampling period of the simulation [401, p. 269]. The
delay of the 0th path is defined as τ0 ≡ 0, thus

τl = lTsa , l = 0, 1, . . . , L − 1. (6.18)

For each simulation trial, the set of path gains {a l }L−1

l=0 are generated randomly. Each
gain is complex valued, has a zero mean and a variance

σa2l = E |al |2 ,

l = 0, 1, . . . , L − 1. (6.19)

Both the real and imaginary parts of the path gains are Gaussian distributed [401, p.
267]; thus the envelope |al |2 is Rayleigh distributed. Also, the channels are normalized
such that
L−1
X
σa2l = 1. (6.20)
l=0

As outlined in Pätzold’s text (pp. 276–279) the parameters σ a2l , τl and L determine
the fundamental characteristic functions and quantities of the channel models, such as the
delay power spectral density and the delay spread 2 . The relevant formulas are expressed
below.

• Delay power spectral density:

L−1
X
Sτ τ (τ ) = σa2l δ(τ − τl ). (6.21)
l=0

• Average delay:
L−1
X
Bτ(1)
τ = σa2l τl . (6.22)
l=0
2
The phrase “delay power spectral density” is also commonly referred to as “power delay profile”
(PDP) or “multipath intensity profile” (MIP). For the sake of being consistent with [401], “delay power
spectral density” is used here. In Pätzold’s text, a clear distinction is made between stochastic channel
models, which provide the theoretical and mathematical foundations, and “deterministic” channel models
which are generated in software or hardware for simulation purposes. For the sake of simplicity, this
distinction isn’t stressed here (which results in a slightly different notation for the expressed formulas
in his text). Also, since only time-invariant channels are considered in this thesis, the Doppler power
spectral density, time correlation function and coherence time (see [401, pp. 277–279]) are not discussed.
 110

• Delay spread: v
uL−1
(1) 2
uX  
Bτ(2)
τ =t (σal τl )2 − Bτ τ (6.23)
l=0

• Frequency correlation function:

L−1
0
X
0
rτ τ (v ) = σa2l e−j2πv τl (6.24)
l=0

The variable v 0 is referred to as the frequency separation variable [401, p. 278].

• Coherence bandwidth: The coherence bandwidth is the smallest positive value

BC which fulfils |rτ τ (BC )| = 0.5|rτ τ (0)|; which, due to (6.20) and (6.24), is equiv-
alent to L−1 
X  1
2 −j2πBC τl 
σ e  − = 0. (6.25)

al
 2


l=0

Notice that BC is the 3 dB bandwidth of rτ τ (v 0 ).

6.2.1 Channel Models

CE-OFDM is simulated over four frequency-selective fading channel models. Table

6.2 defines the parameters {σa2l } and {τl }. Channel Af and Bf are similar to the maritime
channel models in [350]3 . Both have a secondary path with a 5 µs propagation delay.
Channel Af has a weak secondary path (one-tenth, i.e., −10 dB, the power of the primary
path); Channel Bf has a stronger secondary path (one-half, i.e., −3 dB, the power of the
primary path).

Channel Cf has an exponential delay power spectral density:


CC e−τl /2µs , 0 ≤ τl ≤ 8.75 µs,

2 f
σal ,C = (6.26)
0,

otherwise,

where
35
X
C Cf = 1 exp(−τl /2e-6) = 0.1188 . . . (6.27)
l=0

is the normalizing constant used to guarantee (6.20). Note that the maximum propaga-
tion delay is 8.75 µs.
3
To avoid notational ambiguities, the channel model labels in this section have the subscript “f”
(“fading”).
 111

Table 6.2: Channel model parameters.

Path no. Delay (µs) Channel Af Channel Bf Channel Cf Channel Df

l τl = lTsa σa2l ,A σa2l ,B σa2l ,C σa2l ,D
0 0.00 10/11 2/3 1.18e-1 1/36
1 0.25 0 0 1.04e-1 1/36
2 0.50 0 0 9.25e-2 1/36
3 0.75 0 0 8.16e-2 1/36
4 1.00 0 0 7.20e-2 1/36
5 1.25 0 0 6.36e-2 1/36
6 1.50 0 0 5.61e-2 1/36
7 1.75 0 0 4.95e-2 1/36
8 2.00 0 0 4.37e-2 1/36
9 2.25 0 0 3.85e-2 1/36
10 2.50 0 0 3.40e-2 1/36
11 2.75 0 0 3.00e-2 1/36
12 3.00 0 0 2.65e-2 1/36
13 3.25 0 0 2.33e-2 1/36
14 3.50 0 0 2.06e-2 1/36
15 3.75 0 0 1.82e-2 1/36
16 4.00 0 0 1.60e-2 1/36
17 4.25 0 0 1.41e-2 1/36
18 4.50 0 0 1.25e-2 1/36
19 4.75 0 0 1.10e-2 1/36
20 5.00 1/11 1/3 9.75e-3 1/36
21 5.25 0 0 8.60e-3 1/36
22 5.50 0 0 7.59e-3 1/36
23 5.75 0 0 6.70e-3 1/36
24 6.00 0 0 5.91e-3 1/36
25 6.25 0 0 5.22e-3 1/36
26 6.50 0 0 4.60e-3 1/36
27 6.75 0 0 4.06e-3 1/36
28 7.00 0 0 3.58e-3 1/36
29 7.25 0 0 3.16e-3 1/36
30 7.50 0 0 2.79e-3 1/36
31 7.75 0 0 2.46e-3 1/36
32 8.00 0 0 2.17e-3 1/36
33 8.25 0 0 1.92e-3 1/36
34 8.50 0 0 1.69e-3 1/36
35 8.75 0 0 1.49e-3 1/36
36 9.00 0 0 0 0
37 9.25 0 0 0 0
38 9.50 0 0 0 0
39 9.75 0 0 0 0
40 10.0 0 0 0 0
 112

The last model, Channel Df , has a uniform delay power density spectrum:

CD , 0 ≤ τl ≤ 8.75 µs,

2 f
σal ,D = (6.28)
0,

otherwise,

where the normalizing constant is

CDf = 1/36. (6.29)

In Figure 6.9 the delay power density spectrum (6.21) and the frequency correlation
function (6.24) are plotted for each of the four models. The corresponding average
delay (6.22), delay spread (6.23) and coherence bandwidth (6.25) for each model is
labeled. Notice that Channel Df has the smallest coherence bandwidth, B C = 67 kHz.
For Channel Af the coherence bandwidth isn’t finite since, as shown in subfigure (b),
|rτ τ (v 0 )| > −3 dB for all frequency separation values 4 .

6.2.2 Simulation Procedure and Preliminary Discussion

The average performance of various CE-OFDM systems is evaluated over the four
stochastic channel models. This is done by randomly generating {a l }—which, as stated
above, are complex-valued quantities, drawn from the Gaussian distribution, with zero
mean and variance {σa2l }—computing the received samples (6.2), then processing the
samples with the frequency-domain equalizer and the CE-OFDM demodulator. At each
average Eb /N0 considered, the simulation runs for at least 20,000 bit errors, or until
100,000,000 bits are transmitted, whichever happens first. This corresponds to many
thousands of channel realizations5 . Some channel realizations result in very poor per-
formance (for example, see Figure 6.8), while others result in a bit error rates not much
worse than that of the simple AWGN channel. This performance difference is attributed
to the severity of the channel’s frequency response, as observed with the several examples
in Section 6.1.

The performance also depends on the gain of the channel realization. Due to (6.20)
the channel gain, on average, is normalized to unity; however, for a given trial, the
channel may be fading such that the gain is less than unity, resulting in degraded per-
formance. The likelihood of a deep channel fade depends on the number of independent
4
˛ 10 1 9 1
For Channel Af , min |rτ τ (v 0 )| = min ˛ 11 exp(−j2πv 0 5 µs)˛ =
˛
+ 11 11
> 2
≈ −3 dB.
5
Example simulation code can be found in Appendix C.
 113

(a) Delay power spectral density, Channel Af (b) Frequency correlation function, Channel Af
0 0
(1)
−5 Bτ τ = 0.45 µs −3

10 log 10 [rτ τ (v 0 )]
10 log10 [Sτ τ (τ )]

(2) BC → ∞
Bτ τ = 1.44 µs
−10
−6
−15
−9
−20

−25 −12

−30 −15
−450 −300 −150
0 1 2 3 4 5 6 7 8 9 10 0 150 300 450
Propagation delay, τ (µs) Frequency separation, v 0 (kHz)
(c) Delay power spectral density, Channel Bf (d) Frequency correlation function, Channel Bf
0 0
(1)
−5 Bτ τ = 1.67 µs −3

10 log 10 [rτ τ (v 0 )]
10 log10 [Sτ τ (τ )]

(2) BC
Bτ τ = 2.36 µs
−10
−6
−15
−9
−20

−25 −12

−30 −15
−450 −300 −150
0 1 2 3 4 5 6 7 8 9 10 0 74 150 300 450
Propagation delay, τ (µs) Frequency separation, v 0 (kHz)
(e) Delay power spectral density, Channel Cf (f) Frequency correlation function, Channel Cf
0 0
(1)
−5 Bτ τ = 1.78 µs −3
10 log 10 [rτ τ (v 0 )]
10 log10 [Sτ τ (τ )]

(2) BC
Bτ τ = 1.75 µs
−10
−6
−15
−9
−20

−25 −12

−30 −15
−450 −300 −150
0 1 2 3 4 5 6 7 8 9 10 0 140 300 450
Propagation delay, τ (µs) Frequency separation, v 0 (kHz)
(g) Delay power spectral density, Channel Df (h) Frequency correlation function, Channel Df
0 0
(1)
−5 Bτ τ = 4.38 µs −3
10 log 10 [rτ τ (v 0 )]
10 log10 [Sτ τ (τ )]

(2) BC
Bτ τ = 2.60 µs
−10
−6
−15
−9
−20

−25 −12

−30 −15
0 1 2 3 4 5 6 7 8 9 10 −450 −300 −150 0 67 150 300 450
Propagation delay, τ (µs) Frequency separation, v 0 (kHz)

Figure 6.9: Fundamental characteristic functions and quantities [(6.21)–(6.25)] of the

four channel models considered.
 114

propagation paths [the WSSUS assumption makes each path in (6.16) independent]. It is
unlikely that multiple paths fade simultaneously. For this reason, channels characterized
by multiple propagation paths possess a type of diversity known at multipath diversity—
which can be exploited by the receiver. Of the four models considered in this study,
Channel Df can be said to have the most multipath diversity: the gain of a given realiza-
tion depends on 36 independent paths, each having, on average, an equal contribution.
Channel Af can be said to have the least amount of multipath diversity: over 90% of the
channel gain depends on a single path. Channel B f has more multipath diversity than
Channel Af since the gain is distributed more equally between the two paths. That is,
the multipath diversity depends not only on the number of independent paths but also
on the way in which the power is distributed over the paths, as determined by {σ a2l }. [It
is worth noting that the frequency-nonselective channel models considered in Chapter 5
have L = 1 path of which 100% of the channel gain depends (σ a21 = 1), and thus these
channels have no multipath diversity.] In the results that follow, the impact of multipath
diversity—and its frequency-domain dual frequency diversity—on CE-OFDM systems is
studied.

6.2.3 Simulation Results

The simulation results of this study are presented over three figures: Figure 6.10
compares the performance of a CE-OFDM system, with fixed modulation order M and
modulation index h, over the four channel models; Figure 6.11 compares the performance
of a CE-OFDM system with fixed M but varying h over Channel C f ; and Figure 6.12
compares the performance of constant envelope and conventional OFDM systems, in the
presence of power amplifier nonlinearities, over Channel C f . For each case, the number
of subcarriers is N = 64.

In Figure 6.10, performance results of an M = 4, N = 64, 2πh = 1.0 CE-OFDM

system are plotted. The simulation results over the multipath channel models A f –Df are
labeled with circles and triangles; the MMSE equalized results have solid lines connecting
the points, while the ZF equalized results use dashed lines. For reference, the performance
of the system over the simple AWGN channel is plotted (with dash-dot lines) along with
the performance over the Rayleigh frequency-nonselective fading channel (represented
by the thick solid line). These results show the significant performance improvement
 115

MMSE: Channel Af
Bf
10−1 Cf
Df
ZF: Channel Af
Bf
Cf
Df
Rayleigh, L = 1
AWGN
AWGN approx (4.35)

10−2
Bit error rate

10−3

10−4 5 10 15 20 25 30 35 40
Average signal-to-noise ratio per bit, Eb /N0 (dB)

Figure 6.10: Performance results. (Multipath results are labeled with circle and triangle
points; the Rayleigh, L = 1 result is that of the frequency-nonselective channel model.
M = 4, N = 64, 2πh = 1.0)

that is to be had by using the MMSE equalizer. At the bit error rate 0.001, for example,
MMSE outperforms ZF by 10 dB for Channel D f . These results also show the impact of
multipath diversity. Consider the MMSE results. For E b /N0 > 15 dB, the performance
over Channels Af –Df is better than the performance over the frequency-nonselective
Rayleigh (L = 1 path) channel. For BER ≤ 0.001, the performance over the multipath
channels is at least 5 dB better than the performance over the single path channel.
Notice that Channel Df , which has the most multipath diversity, results in a better
performance that all the other channels. The performance over Channel B f , which has
more multipath diversity than Channel A f , is in fact better than the performance over
Channel Af . These results indicate that the CE-OFDM receiver exploits the multipath
diversity of the channel.

The fact that constant envelope OFDM exploits multipath diversity is an interesting
result since conventional OFDM doesn’t. This was shown in Section 2.1.1; specifically,
 116

by (2.9). So long as the duration of the guard interval is greater than or equal to
the channel’s maximum propagation delay, that is, T g ≥ τmax , and a cyclic prefix is
transmitted during the guard interval, the performance of OFDM in a time-dispersive
channel is equivalent to flat fading performance. In other words, the multipath fading
performance is the same as single path fading performance. In the context of Section
2.1.1, this property was considered beneficial since ISI is avoided. In the context here,
however, this property is considered a weakness since the multipath diversity of the
channel isn’t leveraged6 .

To understand why CE-OFDM has improved performance over multipath fading

channels (compared to single path fading channels) while OFDM doesn’t, it is best to
view the problem in the frequency domain. The frequency domain dual to multipath
diversity is frequency diversity. It can be said that OFDM lacks frequency diversity
as well. As identified in Section 1.1.2, the wideband frequency-selective fading channel
is converted into N contiguous frequency-nonselective fading channels. Therefore any
frequency diversity inherent to the channel—that is, over the signal bandwidth the fre-
quency response of the channel varies, which can be taken advantage of by the receiver
to obtain performance better than flat fading—is not exploited by the OFDM receiver.

CE-OFDM, in contrast, has the ability to exploit the frequency diversity of the
channel since the phase modulator, in effect, spreads the data symbol energy in the
frequency domain. This can be seen by viewing the CE-OFDM waveform by the Taylor
series expansion [see Section 3.2, (3.24)]:
" #
σφ2 2
σφ3 3
s(t) = A 1 + jσφ m(t) − m (t) − j m (t) + . . . , (6.30)
2 6

0 ≤ t < TB , where A is the signal amplitude, σφ2 = (2πh)2 is the phase signal variance,
and m(t) = CN N
p
6/N (M 2 − 1), is the normalized
P
k=1 Ik qk (t), 0 ≤ t < TB , CN =
OFDM message signal. The higher-order terms m n (t), n ≥ 2, results in a frequency
spreading of the data symbols. This property is best demonstrated by way of a simple
example.

Example 6.2.1
Consider a CE-OFDM waveform with an OFDM message signal composed of N = 2 orthogonal
6
Note that OFDM systems typically employ channel coding and frequency-domain interleaving, which
offers diversity. However, since this thesis only deals with uncoded systems, these topics are beyond its
scope—and are topics for further research.
 117

cosine subcarriers modulated with binary data symbols (M = 2):

2
X
m(t) = Ik cos 2πkt/TB , 0 ≤ t < TB , (6.31)
k=1

where Ik ∈ {±1}, k = 1, 2. Assume that the modulation index, h, is such that the higher-order
terms m2 (t) and m3 (t) contribute to the make up of s(t) according to (6.30). It is desired to
write m2 (t) and m3 (t) in terms of I1 , I2 and {cos 2πkt/TB }. This task requires some algebra,
but is simply done. For notational simplicity, let’s define

ck ≡ cos 2πkt/TB . (6.32)

Thus, (6.31) is written as

m(t) = I1 c1 + I2 c2 . (6.33)

The second-order term is calculated as

m2 (t) = (I1 c1 + I2 c2 )(I1 c1 + I2 c2 )

(6.34)
= 0.5I12 + 0.5I22 c0 + (I1 I2 ) c1 + 0.5I12 c2 + (I1 I2 ) c3 + 0.5I22 c4 ,

and the third-order term as

m3 (t) = 0.5I12 + 0.5I22 c0 + (I1 I2 ) c1 + 0.5I12 c2



+ (I1 I2 ) c3 + 0.5I22 c4 (I1 c1 + I2 c2 )



= 0.75I12 I2 c0 + 0.75I13 + 1.5I1 I22 c1 + 1.25I12I2 + 0.5I23 c2

(6.35)

+ 0.25I13 + 0.75I1 I22 c3 + 0.75I12 I2 c4

+ 0.75I1 I22 c5 + 0.25I23 c6 .

The expansions above are represented in Table 6.3. The data symbol contribution at each
tone cos 2πkt/TB , k = 0, 1, . . . , 6, for m(t), m2 (t) and m3 (t) is shown. Referring to the tones
as frequency bins, it can be said that for m(t) the two data symbols are simply contained in
the k = 1 and k = 2 frequency bins. For the second-order term, m2 (t), the data symbols mix
across the k = 0, 1, 2, 3, and 4 frequency bins. For m3 (t), the data symbols mix across the
k = 0, 1, . . . , 6 frequency bins.

The simple example above shows how the data symbols spread across multiple fre-
quency bins. In general, it can be said that the N data symbols that constitute the
constant envelope OFDM signal are not simply confined to N frequency bins—as is the
case with conventional OFDM. The phase modulator mixes and spreads—albeit in a
nonlinear and exceedingly complicated manner—the data symbols in frequency, which
gives the CE-OFDM system the potential to exploit the frequency diversity in the chan-
nel. This isn’t necessarily the case, however. For small values of modulation index,
 118

Table 6.3: Data symbol contribution per tone for m n (t), n =1, 2, and 3.
kth tone, cos 2πkt/TB
0 1 2 3 4 5 6
m(t) – I1 I2 – – – –
0.5I12 ,
m2 (t) I1 I2 0.5I12 I1 I2 0.5I22 – –
0.5I22
0.75I13 , 1.25I12 I2 , 0.25I13 ,
m3 (t) 0.75I12 I2 0.75I12 I2 0.75I1 I22 0.25I23
1.5I1 I22 0.5I23 0.75I1 I22

where only the first two terms in (6.30) contribute, that is,

s(t) ≈ A [1 + jσφ m(t)] , (6.36)

the CE-OFDM signal doesn’t have the frequency spreading given by the higher-order
terms. In this case, the CE-OFDM signal is essentially equivalent to a conventional
OFDM signal, jσφ m(t), (plus a relatively large DC term, A) and therefore doesn’t have
the ability to exploit the frequency diversity of the channel. Simply put, CE-OFDM
has frequency diversity when the modulation index is large and doesn’t have frequency
diversity when the modulation index is small.

This property is demonstrated in Figure 6.11. Simulation results of an M = 4, N =

64 CE-OFDM system are shown. The system is simulated over the single path Rayleigh
flat fading channel and over the multipath fading model Channel C f . To demonstrate that
CE-OFDM with a small modulation index lacks frequency diversity, results for 2πh = 0.1
are shown. Notice that the single path and multipath performance is essentially the
same. By contrast, for the large modulation index example 2πh = 1.1, the multipath
performance is significantly better than the single path performance. For example, at
the bit error rate 0.001 the multipath performance is over 10 dB better than the single
path performance.

In the final figure, Figure 6.12, the performance of constant envelope OFDM is
compared to conventional OFDM in the presence of power amplifier nonlinearities. The
SSPA model (see Section 2.3) is used at various input backoff levels. The x-axis is
adjusted to account for the negative impact of input power backoff. The systems are
simulated over Channel Cf . For the OFDM system, QPSK data symbols are used. Three
different CE-OFDM systems are tested: M = 4, 2πh = 0.9; M = 8, 2πh = 2.0; and
M = 16, 2πh = 3.0. The advantage of the CE-OFDM systems is twofold. First, the
CE-OFDM systems operate with IBO = 0 dB. Second, the CE-OFDM systems exploit
 119

100
Multipath
Single path

10−1
Bit error rate

10−2

10−3 2πh 1.1 0.1

10−4 5 10 15 20 25 30 35 40 45 50
Average signal-to-noise ratio per bit, Eb /N0 (dB)

Figure 6.11: Single path versus multipath. (M = 4, N = 64, Channel C f , MMSE)

the frequency diversity inherent to the channel.

At the bit error rate 0.001 the CE-OFDM systems outperform the OFDM system
by at least 10 dB. At this bit error rate, the OFDM system has essentially the same
performance with backoff levels of 6 and 10 dB; therefore, IBO = 6 dB is preferred since
the performance is the same but the power efficiency is higher (see Figure 2.14). Even
so, the 6 dB backoff required by the OFDM system is still far less desirable as the 0 dB
backoff used by the CE-OFDM system. Notice that the OFDM system with IBO = 0
dB results in an irreducible error floor just below the bit error rate 0.1.

The results in Figure 6.12 also highlight the poor performance of CE-OFDM at
low SNR due to the threshold effect (as studied in Section 4.1.3). Over the region 0 dB
≤ Eb /N0 ≤ 10 dB, the OFDM system performs better than the CE-OFDM system. Also,
it should be noted that the M = 8 and M = 16 CE-OFDM systems shown have large
modulation index values (2πh = 2.0 and 2πh = 3.0 respectively) which results in spectral
broadening. Roughly speaking, the spectral efficiency of the QPSK/OFDM system is 2
b/s/Hz, which, according to (4.70), is about the same as the M = 4, 2πh = 0.9 CE-
OFDM system. The M = 8 and M = 16 systems have spectral efficiencies of 1.5 and
1.3 b/s/Hz, respectively.

Making a direct comparison between CE-OFDM and conventional OFDM is difficult

 120

10−1

Bit error rate

10−2

OFDM: IBO = 0 dB
10−3 3 dB
6 dB
10 dB
CE-OFDM: M = 4, 2πh = 0.9
M = 8, 2πh = 2.0
M = 16, 2πh = 3.0

10−4 0 5 10 15 20 25 30 35 40
Eb /N0 + IBO (dB)

Figure 6.12: CE-OFDM versus QPSK/OFDM. (SSPA model, Channel C f , N = 64,

MMSE)

due to the various parameters involved (M , 2πh, IBO, etc.), and due to the fact that
system requirements vary from system to system. For example, if power amplifier effi-
ciency is the most important requirement, then the input power backoff of 0 dB should
be chosen. At this backoff level, the OFDM system has a very high irreducible error
floor due to the power amplifier distortion, while the CE-OFDM system is relatively
unaffected. Alternatively, if operation at low SNR is important, then CE-OFDM may
not be well suited due to the threshold effect.

The results in this chapter show that CE-OFDM can perform quite well in multipath
fading channels—so long as the channel information (i.e., {H[k]}) is known at the receiver
and so long as the added complexity of the frequency-domain equalizer (i.e., two extra
FFTs) is acceptable. Further work is needed to study the effects of channel coding,
time-varying channels, phase noise, and so forth. Also, a thorough study comparing CE-
OFDM, OFDM and single carrier frequency-domain equalizer (SC-FDE) systems could
provide for interesting results.
Chapter 7

Conclusions

In this thesis the peak-to-average power ratio problem associated with orthogonal
frequency division multiplexing is evaluated. The PAPR statistics are studied and the
effect of power amplifier nonlinearities as a function of power backoff is evaluated by
computer simulation. It is shown that the amount of backoff required to reduce spectral
growth and performance degradation is significant: 6–10 dB depending on the subcarrier
modulation used. Large backoff is an unsatisfactory solution for battery-powered systems
since PA efficiency is low.

A signal transformation method for solving the PAPR problem is presented and
analyzed. The high PAPR OFDM signal is transformed to a 0 dB PAPR constant enve-
lope waveform. At the receiver, the inverse transform is performed prior to the OFDM
demodulator. For the CE-OFDM technique described, phase modulation is used. The
effect of the phase modulator on the transmitted signal’s spectrum is studied. It is shown
that the modulation index controls the spectral containment. The modulation index also
controls the system performance. The optimum receiver is analyzed and a performance
bound and approximation is derived. For a large modulation index, the CE-OFDM sig-
nals become less correlated which improves detection performance. The approximation
of the optimum receiver closely matches simulation results. It also closely matches a
derived bit error rate approximation for a practical phase demodulator receiver. For a
small modulation index and high signal-to-noise ratio, the phase demodulator receiver is
nearly optimum. For a larger modulation index the phase demodulator receiver becomes
sub-optimum due to the limitations of the phase demodulator and phase unwrapper.

121
 122

This problem can be suppressed with the use of a properly designed finite impulse re-
sponse lowpass filter which precedes the phase demodulator.

The simulation results of the CE-OFDM performance curves use an oversampling

factor of J = 8. Future work includes experimenting with lower sampling rates for
reduced receiver complexity. The performance of the phase demodulator is a crucial
element to the overall CE-OFDM performance. Therefore, further research is needed
to evaluate more advanced phase demodulation techniques such as digital phase-locked
loops.

Phase modulation is used exclusively in this work. It would be interesting to evaluate

CE-OFDM frequency modulation systems and compare them to the results in this thesis.

In terms of performance over frequency-selective fading channels, the frequency-

domain equalizer requires knowledge of the channel. Many conventional OFDM systems
(those that don’t use differentially encoded modulations) also require channel state in-
formation. Thus techniques for channel estimation in OFDM has been extensively re-
searched [105, 144, 204, 213, 251, 257, 259, 273, 469, 547]. Applying the known techniques,
such as linear minimum mean-squared error (LMMSE) estimation and reduced complex-
ity singular value decomposition (SVD) approaches, to CE-OFDM is a subject for future
investigation. The impact of imperfect channel state information on the performance of
the frequency-domain equalizer is of interest.

CE-OFDM might be used as a stand-alone modulation technique or as a supplement

to an existing OFDM system. For example, a conventional OFDM system is designed for
severe multipath channels. However, at times the channel might be relatively benign so
the OFDM systems is an overkill and, due to power backoff, inefficient. An adaptive radio
might sense times where power efficient CE-OFDM, which requires minimal backoff, is
more applicable. Such a system can adaptively switch between conventional and constant
envelope modes.

For systems, such as power-limited satellite communications, where a constant en-

velope is very desirable, if not required, CE-OFDM might be a viable alternative to
convention continuous phase modulation systems which are complex due to phase trellis
decoding and sensitive to multipath. CE-OFDM is relatively robust in multipath fading
channels with the use of the frequency-domain equalizer. Depending on the channel
condition, equalization might not be required, therefore reducing receiver complexity.
 123

For example, a channel characterized by a two-path model with a weak secondary path,
CE-OFDM might provide acceptable performance without equalization. CPM systems
in the other hand require high quality coherent channels.

In the near term a CE-OFDM prototype is being developed by Nova Engineering

(Cincinnati, OH). This work is being funded by the United States Office of Naval Re-
search under an STTR (small business technology transfer) initiative with UCSD being
the university partner. The goal of the prototype is to offer a second low-power mode
for the existing JTRS (Joint Tactical Radio System) wideband component which uses
OFDM. Research challenges that remain include evaluating CE-OFDM with many sub-
carriers (in this thesis, only 64 subcarriers are used), considering different equalization
techniques, developing synchronization schemes and studying the impact of channel cod-
ing and the effects of time-varying channels.

Additional future work includes comparing CE-OFDM with other block modulation
technique in terms of PAPR, spectral efficiency, power amplifier efficiency, performance
and complexity. There has been an increasing amount of attention given to conventional
single carrier modulation with the addition of a cyclic prefix which allows for frequency-
domain equalization [107, 154, 460, 463, 574]. However, most single carrier modulations
have a non-constant envelope due to pulse shaping and multilevel QAM symbol con-
stellations. A study is needed to compare these modulation techniques to CE-OFDM
taking into account the effects of the PA at various backoff levels. Also, using CPM with
a cyclic prefix is an interesting idea. Comparing the complexity and spectral efficiency of
such a technique with CE-OFDM would be interesting. Such research will help provide
insight into good designs for future wireless digital communication systems that require
power efficiency and high data rates.
Appendix A

Generating Real-Valued OFDM

Signals with the Discrete Fourier
Transform

For some applications, a real-valued OFDM signal is required. This can be done by
taking a DFT of a conjugate symmetric vector. The spectral efficiency of the real-valued
OFDM signal is the same as the spectral efficiency of the complex-valued OFDM signal.

A.1 Signal Description

The baseband OFDM signal is typically written as

N
X −1
x(t) = Xk ej2πkt/TB , 0 ≤ t < TB , (A.1)
k=0

where N is the number of subcarriers, {X k }N −1

k=0 are the data symbols and TB is the
block period. Sampling x(t) at N equally spaced intervals over 0 ≤ t < T B yields the
sequence,
N
X −1
x[i] = x(t)|t=iTB /N = Xk ej2πki/N , i = 0, 1, . . . , N − 1, (A.2)
k=0

which is the inverse discrete Fourier transform (IDFT) of the vector

X = [X0 , X1 , . . . , XN −1 ]. (A.3)

124
 125

The sequence is complex-valued in general. However it can be made real-valued by

making X conjugate symmetric:

∗
XN/2+k = XN/2−k (A.4)

and
X0 = XN/2 = 0. (A.5)

The IDFT is then

N
X −1
x[i] = Xk ej2πki/N
k=1
N/2−1
X
= XN/2−k ej2π(N/2−k)i/N + XN/2+k ej2π(N/2+k)i/N (A.6)
k=1
N/2−1
X
= XN/2−k ej2π(N/2−k)i/N + XN/2−k
∗
ej2π(N/2+k)i/N ,
k=1

i = 0, 1, . . . , N − 1. But since

ej2π(N/2+k)i/N = ej2π(N/2+k)i/N e−j2πN i/N

= ej2π(−N/2+k)i/N (A.7)

= e−j2π(N/2−k)i/N ,

(A.6) can be written as

N/2−1
X
x[i] = XN/2−k ej2π(N/2−k)i/N + XN/2−k
∗
e−j2π(N/2−k)i/N , (A.8)
k=1

i = 0, 1, . . . , N − 1. Using the identity A + A ∗ = 2<{A},

 
N/2−1
X 
x[i] = 2< XN/2−k ej2π(N/2−k)i/N
 
k=1
  (A.9)
N/2−1
X 
= 2< Xk ej2πki/N , i = 0, 1, . . . , N − 1.
 
k=1

And since <{AB} = <{A}<{B} − ={A}={B},

N/2−1
X
x[i] = 2 <{Xk } cos(2πki/N ) − ={Xk } sin(2πki/N ), (A.10)
k=1
 126

i = 0, 1, . . . , N − 1. Thus, x[i] is real. Passing the sequence through a D/A converter

yields the continuous-time real-valued OFDM signal:
N/2−1
X
x(t) = 2 <{Xk } cos(2πkt/TB ) − ={Xk } sin(2πkt/TB ). (A.11)
k=1

Now, suppose the data symbols are derived from a M 2 -QAM (quadrature-amplitude
modulation) constellation; that is,

Xk = <{Xk } + j={Xk }, (A.12)

where
<{Xk }, ={Xk } ∈ {±1, ±3, ±(M − 1)}, for all k. (A.13)

In other words, the real and imaginary components are derived from M -PAM (pulse-
amplitude modulation) constellations. Therefore, processing M 2 -QAM data with the
IDFT, (A.11) is a real-valued M -PAM OFDM signal.

A.2 Spectral Efficiency

Complex-valued baseband signals are transmitted as bandpass signals, centered at

a carrier frequency fc Hz. This is the case for the complex-valued signal in (A.1). The
transmitted signal is represented as
n o
s1 (t) = < x(t)ej2πfc t . (A.14)

In the frequency domain, x(t) is shifted to the right by f c Hz, and the subcarriers are
centered at fc , fc + 1/TB , fc + 2/TB , . . . , fc + (N − 1)/TB Hz. The effective bandwidth of
the signal is therefore N/TB Hz. Each data symbol represents log 2 M bits (i.e., they are
assumed to be selected from a M -ary constellation), therefore the spectral efficiency is
Bits per second (b/s) N log2 M/TB
S1 = = = log2 M b/s/Hz. (A.15)
Bandwidth (Hz) N/TB

The real-valued OFDM signal in (A.11) has the same spectral efficiency as the
complex-valued signal, so long as it is transmitted at baseband. Transmitting the
signal as-is, <{Xk }, k = 1, 2, . . . , (N/2) − 1, modulate cosine subcarriers centered at
1/TB , 2/TB , . . . , [(N/2) − 1]/TB Hz; and likewise, ={Xk }, k = 1, 2, . . . , (N/2) − 1, mod-
ulate sine subcarriers at the same frequencies. The effective bandwidth of the signal is
 127

(N/2)/TB Hz1 , and since the real and imaginary parts of X k represent 0.5 log 2 M bits,
the spectral efficiency of the real-valued OFDM signal is

Bits per second (b/s) 2 × 0.5(N/2) log 2 M/TB

S2 = = = log 2 M b/s/Hz. (A.16)
Bandwidth (Hz) (N/2)/TB

Therefore the spectral efficiency is the same as for the complex case.

However, the spectral efficiency of the real-valued signal is 1/2 that of the complex-
valued signal if the real-valued signal is translated up to a carrier frequency. This is due to
the fact that the cosine and sine subcarriers in (A.11) have a double sideband spectrum:
that is, cos(2πkt/TB ) [or sin(2πkt/TB )] has a spectral components at ±k/T B Hz. [This
isn’t the case for the complex-valued signal, which has complex sinusoids: exp(j2πkt/T B )
has a spectral component only at k/T B Hz and is thus considered single sideband.] The
carrier frequency is typically much larger than the signal bandwidth, so the frequency
translation brings all the negative frequencies to the positive side: −(N/2)/T B + fc  0.
Consequently, the passband transmission of (A.11) results in a signal with double the
bandwidth and 1/2 the spectral efficiency.

1
Only the positive frequencies, f ≥ 0, count.
Appendix B

More on the OFDM Literature

The first OFDM-like radio to be found in the research literature is the Kineplex sys-
tem presented by Mosier et. al in 1958 [354]. Developed at the Collins Radio Company,
Burbank, CA, the radio used 20 tones separated by 110 Hz, each differentially phase
modulated. This paper caused some interest and some controversy as indicated by E.
D. Sunde’s (Bell Laboratories) comments found at the end of the journal paper.

In his 1960 paper [202], H. F. Harmuth, a researcher at General Dynamics, Rochester,

NY, suggested multiplexing orthogonal waveforms. Then, in 1967 M. S. Zimmerman et.
al described a 34 subcarrier military radio named Kathryn. The first paper to identify
the Doppler sensitivity of such a radio was by P. A. Bello [51]. Significant theoretical
contributions were made by B. R. Saltzberg and R. W. Chang of Bell Laboratories
[83, 84, 455]. In 1970 Chang was issued US patent 3,488,445 on OFDM [82].

Weinstein and Ebert, in 1971, where to first to suggest using a DFT for OFDM
modulation [579]. This observation was made six years after Cooley and Tukey published
details of the fast Fourier transform; these developments were significant since all modern
OFDM systems are based on the FFT.

A decade passed with little mention of OFDM in the literature. Then, in the early
80’s researchers from IBM’s Watson Research Center suggests OFDM for a wireline DSL-
type application [408]. They were the first to suggest bit loading. Around this time,
Japanese researcher suggest OFDM for wireless communications [207–209] (also see [6]).
L. J. Cimini’s 1985 paper [102] generated interest when he suggested applying OFDM

128
 129

to mobile systems.

In the late 80’s and early 90’s OFDM received wide interest for the applications
of DSL and for wireless digital broadcasting. Kalet and Zervos compare OFDM to
single carrier with decision feedback equalization [248, 614]. The acceptance of OFDM
into xDSL standards was lead primarily by Stanford University’s J. M. Cioffi et al.
[9, 61, 95–97, 105, 446]. Now, OFDM is widely deployed for this consumer electronics
application. In terms of digital broadcasting, OFDM has been accepted for the European
DAB and DVB standards [162, 477, 552]. In the US, OFDM is being used for IBOC
broadcasting [221, 392].

OFDM is being applied to indoor wireless local area networks under the IEEE 802.11
and the ETSI HYPERLAN/2 standards [552]. And as mentioned in Chapter 1, OFDM
is being developed for ultra-wideband systems; cellular systems; wireless metropolitan
area networks; and for power line communication [119, 160, 264, 604].

Active OFDM research continues. The major focus in the OFDM literature includes
OFDM’s sensitivity to Doppler, phase noise, carrier frequency offsets, and nonlinearities.
Channel estimation and synchronization techniques are of interest, along with techniques
to address the PAPR problem.

Literature Survey Statistics

The OFDM literature is immense, so a detail discussion of it here would be overly

ambitious. The bibliography of this thesis does provide a somewhat current snapshot of
the OFDM literature. Also included in the bibliography are papers dealing with gen-
eral digital communications, continuous phase modulation, FM analog communications,
power amplifiers, computer simulation techniques, and other miscellaneous papers that
have, in some way, contributed to this work.

Conducting a 100% thorough literature review in this field, over the course of a PhD,
is a formidable, if impossible, task. Some statistics of the current author’s attempt are
displayed below.
 130

First, to get an idea of the size of the literature, Figure B.1 shows the result of
searching for “OFDM” in the IEEE online literature database. As of the year 2004,
there are over 800 OFDM-specific IEEE journal papers and over 4300 papers when
including papers presented at IEEE conferences.

1400 4500
Journal papers Journal papers
Journal plus conference papers Journal plus conference papers
4000
1200

3500

1000
3000

800
2500
Papers

Papers

2000
600

1500
400

1000

200
500

0 0
1988 1992 1996 2000 2004 1988 1992 1996 2000 2004

(a) Papers each year. (b) Cumulative paper count.

Figure B.1: “OFDM” search on IEEE Xplore [222].

So, there are many papers to read and to learn from. Besides the OFDM-specific
papers, there are many interesting and fundamental papers dealing with the general
area of digital communications and information theory. Being familiar with the relevant
literature, which may include several thousands of papers published over many decades,
is the goal, however long-term it may be.
 131

600

550

500

450
Papers

400

350

300

250

200

150
Oct 2004 Jan 2005 Apr 2005 Jul 2005 Oct 2005
Filed Piled

Figure B.2: Papers, filed and piled.

This figure shows the number of filed and the number of piled papers as a function
of time, spanning my final year as a PhD student. A filed paper has been printed out,
read, added to a citation list (using BibTeX), and briefly summarized in one or two
paragraphs. A piled paper is in queue waiting to be filed. As the figure shows, the pile
is in good health. In late Spring 2005, a concerted effort was made to “kill the pile”. It
briefly dipped below 150 papers, but the literature is too large—and the battle continues.
 132

8
Papers read per day (log scale)

Oct 2004 Jan 2005 Apr 2005 Jul 2005 Oct 2005
Running average Daily points

Figure B.3: Running average of papers read per day.

Figure B.3 shows the running average of papers read per day, and Figure B.4 shows
a histogram of the filed papers’ publication year. One unknown is the true papers-of-
interest count. A simple model might be: 20 papers per year from 1920–1960; 50 papers
per year from 1960–1980; and 100 papers per year from 1980 to present. A histogram of
this projected goal in relation to the current progress is shown in Figure B.5. According
to the model, 4300 papers are of interest, of which roughly 3700 have yet to be filed. Say
350 papers are read per year (which, according to Figure B.3, isn’t entirely unreasonable).
Of these 350 papers, assume that 100 are current-year, leaving the remaining 250 papers
to be from the past. It would therefore take 3700/250 = 14.8 years to “kill the pile”.
 133

70

60

50

40
Papers

30

20

10

1930 1940 1950 1960 1970 1980 1990 2000

Figure B.4: Year histogram.

Desired?
100

80
Papers

60

40

20 Current

1920 1940 1960 1980 2000 2020

Figure B.5: Projected year histogram?

Appendix C

Sample Code

The simulations were performed using GNU Octave [188] and the figures were gen-
erated with Gnuplot [189]. In this appendix sample code is provided.

C.1 GNU Octave Code

Below is GNU Octave code used to obtain the results for the Channel C f , MMSE
curve in Figure 6.10. The code can easily be adapted to obtain other results, as outlined
below.

% GNU Octave code for M=4, N=64, 2pih=1, Channel Cf result.

% Written by: Steve Thompson
% ------- Simulation parameters ------------------------------------
% for a good time, max min sqrt
shortrun=0; % equals 0 or 1
if shortrun % (use for speed/testing)
Trans_max=1e5; % max bits sent per SNR
Trans_min=2e4; % min bits sent per SNR
Error_min=2e1; % min errors per SNR
else % long run (use for accuracy/final result)
Trans_max=100e6; % max bits sent per SNR
Trans_min=1e6; % min bits sent per SNR
Error_min=2e5; % min errors per SNR
end

targetBER=1e-5; % target BER

SNRmax=50; % max SNR (dB)

134
 135

io=1; % index offset

A=1; % signal amplitude
M=4; % modulation order
modh=1.0/(2*pi); % modulation index
N=64; % number of subcarriers
TB=128e-6; % block time
J=8; % oversampling factor
Fsa=J*N/TB; % sampling rate
Tsa=1/Fsa; % sampling period
Tg=10e-6; % guard time
TF=Tg+TB; % frame time
Ng=Tg*Fsa; % samples per guard interval
NB=TB*Fsa; % samples per symbol
NF=TF*Fsa; % samples per frame
ip=[Ng:NF-1]+io; % processing indices
Ndft=512; % DFT size (for equalizer)
taumax=9e-6; % maximum delay spread of channel (sec)
Nc=taumax*Fsa; % number of channel taps
Nr=Nc+NF-1; % number of received samples
L=8; % blocks/channel realization (vectorize)

%% Bit and symbol mappings (depends on modulation order)

if M==2
SymMap=[-1;1]; % data symbol mapping
BitMap=[0; 1]; % bit mapping
end
if M==4
SymMap=[-3;-1;1;3]; % data symbol mapping
BitMap=[... % bit mapping
0 0; 0 1; 1 1; 1 0];
end
if M==8
SymMap=[-7:2:7]’; % data symbol mapping
BitMap=[... % bit mapping
0 0 0; 0 0 1; 0 1 1; 0 1 0; 1 1 0; 1 1 1; 1 0 1; 1 0 0];
end
if M==16
SymMap=(-15:2:15)’; % data symbol mapping
BitMap=[... % bit mapping
0 0 0 0; 0 0 0 1; 0 0 1 1; 0 0 1 0; 0 1 1 0; 0 1 1 1; ...
0 1 0 1; 0 1 0 0; 1 1 0 0; 1 1 0 1; 1 1 1 1; 1 1 1 0; ...
1 0 1 0; 1 0 1 1; 1 0 0 1; 1 0 0 0];
end

varI=sum(SymMap.^2)/M; % variance of data

 136

CN=sqrt(2/(N*varI)); % normalizing constant

%% Subcarrier matrix
t=0:Tsa:(TB-Tsa); % time vector
W=zeros(NB,N); % initialize unitary matrix
for k=1:N/2 % W is a set of orth. sines and cosines
W(:,k)=cos(2*pi*k*t/TB)’;
end
for k=(N/2+1):N
W(:,k)=sin(2*pi*(k-N/2)*t/TB)’;
end

%% Design FIR filter: improves performance of phase demodulator

%% See Proakis’s DSP text for design details
Mf=11; % filter length
n1=0:(Mf-1); % filter sample index
d=(Mf-1)/2; % delay
n2=(d+1):(d+NB); % desired, delayed indices
fc=0.2; % normalized cutoff frequency (cyc/samp)
wc=2*pi*fc; % normalized cutoff frequency (rad/samp)
h1=zeros(1,Mf); % initialize
for i=1:Mf % compute coefficients
if n1(i)==((Mf-1)/2)
h1(i)=wc/pi;
else
h1(i)=sin(wc*(n1(i)-(Mf-1)/2))/(pi*(n1(i)-(Mf-1)/2));
end
end
w1=0.54-0.46*cos(2*pi*n1/(Mf-1)); % Hamming window
hf=h1.*w1; % windowed filter coefficients

%% Channel delay power spectral density (exponential)

t=[0:Nc-1]’*Tsa; % time vector
p=1/tauRms*exp(-t/2e-6); % delay PDS

% ------- Simulation -----------------------------------------------

BER=0; % initialize BER vector
EbN0_dB=0; % initialize SNR vector
dx=2.5; % SNR step size
iSNR=1; % SNR counter
go=1; % initialize loop
while go % run until max SNR condition
Error_num=0; Trans_num=0; % initialize
while Trans_num<=Trans_min | ...
(Error_num<=Error_min & Trans_num<=Trans_max)
 137

%% Generate L blocks
in=ceil(M*rand(N,L)); % random symbol index
I=SymMap(in); % data symbols
m=CN*W*I; % OFDM message signal
theta0=2*pi*rand(1,L)-pi; % memory terms (assume uniform)
phi=zeros(NF,L); % initialize CE-OFDM phase signal
for i=1:L % cyclic prefix
phi(:,i)=[2*pi*modh*m(NB-Ng+1:NB,i)+theta0(i);...
2*pi*modh*m(:,i)+theta0(i)];
end
s=A*exp(j*phi); % CE-OFDM signal

%% Determine noise power

Es=sum(sum(abs(s).^2))*Tsa; % signal energy
Eb=Es/(L*N*log2(M)); % bit energy
EbN0=10^(EbN0_dB(iSNR)/10); % SNR
N0=Eb./EbN0; % noise spectral height

%% Channel
tmp=sqrt(1/2)*(randn(Nc,1)+j*randn(Nc,1)); % Gaussian vector
Ch=sqrt(p/sum(p)).*tmp; % channel (normalize average power)

%% Received signal plus noise (to be processed by FDE)

rp=zeros(NB,L); % initialize
for i=1:L
tmp1=(conv(Ch,s(:,i))).’; % received samples
tmp1=tmp1(ip); % discard cyclic prefix
tmp2=sqrt(1/2)*(randn(NB,1)+j*randn(NB,1)); % complex Gaussian
noise=sqrt(N0*Fsa)*tmp2; % Gaussian noise
rp(:,i)=tmp1+noise; % received samples plus noise
end

%% Frequency-domain equalizer
H=fft(Ch,Ndft); % channel gains
C=conj(H)./(abs(H).^2+EbN0^(-1)); % correction term (MMSE)
X=fft(rp,Ndft); % to frequency domain
hatS=X.*(C*ones(1,L)); % equalize
x=ifft(hatS,Ndft); % to time domain

%% Filter signal
hats=zeros(NB,L); % initialize
for i=1:L
tmp=(conv(hf,x(:,i))).’; % filtered signal
hats(:,i)=tmp(n2); % filtered signal, desired indices
 138

end

%% Demodulate and detect

hatphi=unwrap(angle(hats)); % phase demodulate
Ihat=W’*hatphi/((2*pi*modh*CN)*NB*1/2); % matched-filter output
inHat=min(round((Ihat+(M-1))/2)+io,M); % index estimate, (<=M)
inHat=max(inHat,1); % (>=1)
Errors=sum(sum(BitMap(in,:)~=BitMap(inHat,:))); % bit errors
Error_num=Error_num+Errors; % cumulative bit errors
Trans_num=Trans_num+L*N*log2(M); % cumulative bits

%% Display (optional)
if rem(Trans_num,10*L*N*log2(M))==0 % print-frequency
clc
printf([’MMSE, fading ChC, EQ, M=%d, 2pih=%1.1f, J=%d, ’...
’fc=%1.1f, EbN0=%2.1f, Trans_num=%d, ’...
’Error_num=%d, BER=%1.1e’], M, 2*pi*modh, J, fc,...
EbN0_dB(end), Trans_num, Error_num, Error_num/Trans_num)
end
end % end this SNR

BER(iSNR)=Error_num/Trans_num; % bit error rate for current SNR

%% Test for max SNR condition

if BER(iSNR)<targetBER | EbN0_dB(iSNR)>=SNRmax
go=0;
else % keep going
iSNR=iSNR+1;
EbN0_dB(iSNR)=EbN0_dB(iSNR-1)+dx;
end
end % end simulation

%% Plot
semilogy(EbN0_dB,BER)

%% Save
tmp=[EbN0_dB’ BER’];
save -ascii data tmp

To get other results, the above code is used with different values of M , 2πh, equalizer
settings, and/or channel definitions. The ZF equalizer is simulated by changing the
equalizer to

C=1./H; % correction term (ZF)

The other fading channels are generated by changing the code that defines the channel.
For Channel Af :
 139

%% Channel delay power spectral density (two-path)

tau=[0 5e-6]; % path delays
power_dB=[0 -10]; % path power (dB)
power=10.^(power_dB/10); % path power
for n=1:length(tau)
i=tau(n)*Fs; % path index
p(i+io,1)=power(n); % delay PSD
end
p=[p; zeros(Nc-length(p),1)]; % zero-pad

For Channel Bf :

%% Channel delay power spectral density (two-path)

tau=[0 5e-6]; % path delays
power_dB=[0 -3]; % path power (dB)
power=10.^(power_dB/10); % path power
for n=1:length(tau)
i=tau(n)*Fs; % path index
p(i+io,1)=power(n); % delay PSD
end
p=[p; zeros(Nc-length(p),1)]; % zero-pad

For Channel Df :

%% Channel delay power spectral density (uniform)

tau=[0:Nc-1]’*Ts; % discrete propagation delays
p=ones(size(t)); % delay PSD

Additionally, the above template can be used for conventional OFDM with some
minor alterations.

C.2 Gnuplot Code

The majority of the figures in this thesis were generated with Gnuplot. Below is
sample code which generates Figure 6.10.

# Tell Gnuplot what kind of plot to generate and give it

# some parameters.
set term pslatex monochrome dashed rotate 8
set format "$%g$"
set logscale y 10
set format y "$10^{%T}$"
 140

set ticscale 0.5

set border 31 linewidth 0.5
set grid
set size 1.0,1.4
set key width -23.5 height 1 box lw 0.1 41.4,1.5e-1
set output "p_ber"

# Define line styles.

set style line 1 lt 1 lw 1 pt 9 ps 1.0
set style line 11 lt 3 lw 1 pt 9 ps 1.0
set style line 2 lt 1 lw 1 pt 6 ps 1.0
set style line 22 lt 3 lw 1 pt 6 ps 1.0
set style line 3 lt 1 lw 1 pt 7 ps 1.0
set style line 33 lt 3 lw 1 pt 7 ps 1.0
set style line 4 lt 1 lw 1 pt 8 ps 1.0
set style line 44 lt 3 lw 1 pt 8 ps 1.0
set style line 5 lt 1 lw 3
set style line 6 lt 5 lw 3
set style line 7 lt 5 lw 1

# Define labels.
set xlabel ’[t]{Average signal-to-noise ratio per bit,\
$\mathcal{E}_\text{b}/N_0$ (dB)}’
set ylabel ’Bit error rate’

# Now, plot. (The data files are in a make-believe

# directory called ‘results’
plot [5:44][1e-4:2e-1]\
"results/MMSE/ChA" t ’MMSE: Channel A’ w lp ls 1,\
"results/MMSE/ChB" t ’B’ w lp ls 2,\
"results/MMSE/ChC" t ’C’ w lp ls 3,\
"results/MMSE/ChD" t ’D’ w lp ls 4,\
"results/ZF/ChA/" t ’ZF: Channel A’ w lp ls 11,\
"results/ZF/ChB/" t ’B’ w lp ls 22,\
"results/ZF/ChC/" t ’C’ w lp ls 33,\
"results/ZF/ChD/" t ’D’ w lp ls 44,\
"results/flat" t ’Rayleigh, $\mathcal{L}=1$’ w l ls 5,\
"results/AWGN" t ’AWGN’ w l ls 6,\
"results/approx" t ’AWGN approx \eqref{eqn:approx}’ w l ls 7
Abbreviations

A/D analog-to-digital converter

AM/AM amplitude/amplitude conversion of power amplifier
AM/PM amplitude/phase conversion of power amplifier
AWGN additive white Gaussian noise
b bit
BER bit error rate
CCDF complementary cumulative distribution function
CE constant envelope
CE-OFDM constant envelope OFDM
CNR carrier-to-noise ratio
CP cyclic prefix
CPM continuous phase modulation
dB decibels, 10 log 10 (·)
D/A digital-to-analog converter
DAB digital audio broadcasting
DC direct current
DFE decision feedback equalizer
DFT discrete Fourier transform
DSL digital subscriber line
DVB digital video broadcasting
ETSI European Telecommunications Standards Institute
FDE frequency-domain equalizer
FFT fast Fourier transform
FIR finite impulse response
FOBP fractional out-of-band power
Hz Hertz (1 cycle/s)
IBO input power backoff
IBOC in-band on-channel
ICI intercarrier interference
IDFT inverse discrete Fourier transform
IEEE Institute of Electrical and Electronic Engineers
IFFT inverse fast Fourier transform
ISI intersymbol interference
JTRS Joint Tactical Radio System

141
 142

kHz kilohertz (1 thousand cycles/s)

LAN local area network
LMMSE linear minimum mean-squared error
LMS least-mean-square
LOS line-of-signal
M -PSK M -ary phase-shift keying
M -PAM M -ary pulse-amplitude modulation
M -QAM M -ary quadrature-amplitude modulation
MAN metropolitan area network
Mb/s megabits per second (1 million b/s)
Msamp megasample (1 million samples)
MHz megahertz (1 million cycles/s)
ML maximum-likelihood
OFDM orthogonal frequency division multiplexing
P/S parallel-to-serial conversion
PA power amplifier
PAM pulse-amplitude modulation
PAPR peak-to-average power ratio
PLC power line communication
PSK phase-shift keying
QAM quadrature-amplitude modulation
QPSK quadrature phase-shift keying
RLS recursive least-square
RMS root-mean-square
s second
S/P serial-to-parallel conversion
samp sample
SC-FDE single carrier frequency-domain equalizer
SER symbol error rate
SDR software defined radio
SNR signal-to-noise ratio
SSPA solid-state power amplifier
STTR small business technology transfer
SVD singular value decomposition
TWTA traveling-wave tube amplifier
UWB ultra-wideband
W Watts
WSSUS wide-sense stationary uncorrelated scattering
µs microsecond (1/1,000,000 s)
Symbols

Set Theory
∈ is an element of
∈
/ is not an element of
[·] closed interval
[·) open interval
{xn }N
n=1 set of elements x1 , x2 , . . . , xN

Operators and Miscellaneous Symbols

arg(·) argument
cos(·) cosine
DFT{·} discrete Fourier transform
e 2.71828182845905. . .
e(·) exponential function
exp(·) exponential function
E{·} expected value
F{·}(f ) Fourier transform
I0 (·) 0th-order modified Bessel function of the first kind
IDFT{·} inverse discrete Fourier transform
={·} imaginary
√ part
j −1
Ji (·) ith-order Bessel function of the first kind
max maximum
min minimum
lim limit
ln(·) natural log
logx (·) log base x
LP{·} lowpass component
P (·) probability
Q(·) Gaussian Q-function
<{·} real part
sin(·) sine
sinc(·) sinc function
var{·} variance

143
 144

x(t) x as a function of t
x[i] discrete-time samples of x at the ith index
δ(·) delta function
π 3.14159265358979. . .
∞ infinity
Rb
Ra (·)dx definite integral
(·)dx indefinite integral
QN
multiple product
Pn=1
N
n=1 multiple sum
n! factorial
x→a x approaches a
x∗y x convolved with y
|·| absolute value
(·)∗ complex conjugate
d·e ceiling function
b·c floor function
= equal
≡ equal by definition
6= not equal
≈ approximately equal
≤ less than or equal to
≥ greater than or equal to
< strictly less than
> strictly greater than
 much less than
 much greater than

Power Amplifier
Amax maximum input level
Asat input saturation level
g0 gain
G(·) AM/AM conversions
p sharpness parameter for the SSPA model
αφ , β φ AM/PM parameters for the TWTA model
ηA efficiency of Class-A power amplifier
K backoff ratio
Φ(·) AM/PM conversions
 145

Channel
2σ02 scatter component power of frequency-nonselective channel
al complex-valued gain of the lth path
BC coherence bandwidth
(1)
Bτ τ average delay
(2)
Bτ τ delay spread
C channel capacity
h(τ, t) time-variant channel impulse response
h(τ ) time-invariant channel impulse response
h[i] samples of the channel impulse response
H[k] discrete Fourier transform of h[i]
KR Rice factor
L number of discrete paths
rτ τ (v 0 ) frequency correlation function
Sτ τ (τ ) delay power spectral density
v0 frequency separation variable
∆τl propagation delay difference between τ l and τl−1 , that is, ∆τl = τl − τl−1
ρ line-of-sight component power of frequency-nonselective channel
σa2l average power of the lth path
τ continuous propagation delay
τl discrete propagation delay of the lth path
τmax maximum propagation delay

Signal
A signal amplitude
Ab (k) the value of the kth subcarrier at the beginning of the block interval
Ae (k) the value of the kth subcarrier at the end of the block interval
Amax clip level
Bbpf bandwidth of bandpass filter
Bn noise bandwidth
Brms root-mean-square bandwidth
Bs effective bandwidth of CE-OFDM signal
C[k] frequency-domain equalizer terms
CN normalizing constant
d2m,n squared Euclidean distance between mth and nth signal
d2m,n (K) squared Euclidean distance between mth and nth signal as a function of the
phase constant
d2min minimum squared Euclidean distance
D total number of data symbol differences
Eb energy per bit
Eb /N0 signal-to-noise ratio per bit
Eq subcarrier energy
Ex energy of signal x
f frequency variable (cycles/s)
 146

f0 normalized frequency variable (cycles/samp)

fc carrier (or center) frequency (cycles/s)
fsa sampling rate (samp/s)
FOBP(f ) fractional out-of-band power
ˆ
FOBP(f ) estimated fractional out-of-band power
g(t) pulse shape
h modulation index
I data symbol
Iˆ estimated data symbol
J oversampling factor
kb bits per symbol
K phase signal constant, K = 2πhCN
K d2 number of neighboring signal points having minimum squared Euclidean
min
distance d2min
Lfir filter length
m(t) message signal
M modulation order of data symbol constellation
n(t) lowpass complex-valued zero mean additive Gaussian noise
n[i] samples of n(t)
nbp (t) bandpass representation of n(t) [bandpass Gaussian noise]
nc (t) in-phase component of nbp (t)
ns (t) quadrature component of nbp (t)
nw (t) white Gaussian noise
N number of subcarriers
N0 /2 spectral height of additive white Gaussian noise
Nc number of channel samples
NB number of block samples
Ng number of guard samples
pγ (x) probability density function of signal-to-noise ratio per bit
pξ (x) probability density function of ξ(t) samples
Px average power of signal x
PAPRx the peak-to-average power ratio of signal x
qk (t) kth subcarrier
r(t) lowpass equivalent representation of received signal
rbp (t) bandpass representation of r(t)
R rate, b/s
R/B spectral efficiency, b/s/Hz
s(t) lowpass equivalent representation of transmitted signal
s[i] samples of s(t)
sbp (t) bandpass representation of s(t)
sc (t) in-phase component of sbp (t)
ss (t) quadrature component of sbp (t)
S(f ) frequency domain representation of s(t)
S[k] discrete Fourier transform of s[i]
t time variable
 147

TB block period
Tg guard period
Ts symbol period
Tsa sampling period
W effective bandwidth of OFDM signal, W = N/T B
∆m,n (k) data symbol difference between mth and nth signal at the kth subcarrier
γ signal-to-noise ratio per bit (used interchangeably with E b /N0 )
γ̄ average signal-to-noise ratio per bit
γclip clipping ratio
fo normalized carrier frequency offset
ηt transmission efficiency
θi memory term during ith CE-OFDM block interval
ξ(t) noise at the output of phase demodulator
σI2 data symbol variance
σn2 variance of noise samples, n[i]
σφ2 phase signal variance
ρm,n correlation between mth and nth signal
ρm,n (K) correlation between mth and nth signal as a function of the phase constant
ρmax maximum correlation among signals
φ(t) phase signal
φn (t) noise autocorrelation function
ΦAb (f ) Abramson spectrum
Φ̂Ab (f ) estimated Abramson spectrum
Φx (f ) power density spectrum of signal x.
Φ̂x (f ) estimated power density spectrum of signal x.
Bibliography

[1] N. Abramson, “Bandwidth and Spectra of Phase-and-Frequency-Modulated

Waves,” IEEE Trans. Commun., pp. 407–414, Dec. 1963.

[2] F. Abrishamkar, “Class of Compact Spectrum Coded Signals Over Multipath

Fading Channel,” in Proc. IEEE Milcom, vol. 1, Oct. 1988, pp. 265–270.

[3] M. T. Abuelma’atti, “Frequency-Dependent Nonlinear Quadrature Model for

TWT Amplifiers,” IEEE Trans. Commun., vol. 32, no. 8, pp. 982–986, Aug. 1984.

[4] F. Adachi and J. D. Parsons, “Error Rate Performance of Digital FM Mobile

Radio with Postdetection Diversity,” IEEE Trans. Commun., vol. 37, no. 3, pp.
200–210, Mar. 1989.

[5] J. Ahn and H. S. Lee, “Frequency Domain Equalisation of OFDM Signals Over
Frequency Nonselective Rayleigh Fading Channels,” IEE Electr. Lett., vol. 29,
no. 16, pp. 1476–1477, Aug. 1993.

[6] S. Aikawa, Y. Nakamura, and H. Takanashi, “Performance of Trellis Coded 256

QAM Super-Multicarrier Modem using VLSI’s for SDH Interface Outage-Free
Digital Microwave Radio,” IEEE Trans. Commun., vol. 42, no. 2/3/4, pp.
1415–1421, Feb./Mar./Apr. 1994.

[7] A. N. Akansu, P. Duhamel, X. Lin, and M. de Courville, “Orthogonal

Transmultiplexers in Communication: a Review,” IEEE Trans. Acoust., Speech,
Signal Processing, vol. 46, no. 4, pp. 979–995, Apr. 1998.

[8] N. Al-Dhahir, “Single-Carrier Frequency-Domain Equalization for Space–Time

Block-Coded Transmissions Over Frequency-Selective Fading Channels,” IEEE
Commun. Lett., vol. 5, no. 7, pp. 304–306, July 2001.

[9] N. Al-Dhahir and J. M. Cioffi, “Optimum Finite-Length Equalization for

Multicarrier Transceivers,” IEEE Trans. Commun., vol. 44, no. 1, pp. 56–64, Jan.
1996.

[10] ——, “A Bandwidth-Optimized Reduced-Complexity Equalized Multicarrier

Transceiver,” IEEE Trans. Commun., vol. 45, no. 8, pp. 948–956, Aug. 1997.

148
 149

[11] S. M. Alamouti, “A Simple Transmit Diversity Technique for Wireless

Communications,” IEEE J. Select. Areas Commun., vol. 16, no. 8, pp.
1451–1458, Oct. 1998.

[12] G. Almási and D. Padua, “MaJIC: Compiling MATLAB for Speed and
Responsiveness,” in Proc. AMC SIGPLAN Conf., vol. 37, no. 5, Berlin, May
2002, pp. 294–303.

[13] F. Amoroso, “Pulse and Spectrum Manipulation in the Minimum (Frequency)

Shift Keying (MSK) Format,” IEEE Trans. Commun., vol. com-24, no. 3, pp.
381–384, Mar. 1976.

[14] J. B. Anderson, T. Aulin, and C.-E. Sundberg, Digital Phase Modulation. New
York: Plenum Press, 1986.

[15] J. B. Anderson and C.-E. W. Sundberg, “Advances in Constant Envelope Coded

Modulation,” IEEE Commun. Mag., pp. 36–45, Dec. 1991.

[16] J. R. Anderson, C.-E. W. Sundberg, T. Aulin, and N. Rydbeck,

“Power-Bandwidth Performance of Smoothed Phase Modulation Codes,” IEEE
Trans. Commun., vol. 29, no. 3, pp. 187–195, Mar. 1981.

[17] R. R. Anderson and J. Salz, “Spectra of Digital FM,” Bell Syst. Tech. J., vol. 44,
pp. 1165–1189, July–Aug. 1965.

[18] S. Andreoli, H. G. McClure, and P. Banelli, “Digital Linearizer for RF

Amplifiers,” IEEE Trans. Broadcast., vol. 43, no. 1, pp. 12–19, Mar. 1997.

[19] O. Andrisano, G. Corazza, and G. Immovilli, “On the Availability of Multilevel

CPFSK Systems with Modulation Pulse Shaping During Multipath Propagation,”
IEEE Trans. Commun., vol. com-33, no. 9, pp. 975–985, Sept. 1985.

[20] ——, “Adjacent Channel and Cochannel Interferences in CPFSK Systems with
Nonlinear Transmitters and Limiter-Discriminator Detection,” IEEE Trans.
Commun., vol. 36, no. 5, pp. 544–552, May 1988.

[21] O. Andrisano and N. Ladisa, “On the Spectral Efficiency of CPM Systems over
Real Channel in the Presence of Adjacent Channel and Cochannel Interference:
A Comparison Between Partial and Full Response Systems,” IEEE Trans. Veh.
Technol., vol. 39, no. 2, pp. 89–100, May 1990.

[22] A. Annamalai, C. Tellambura, and V. K. Bhargava, “A General Method for

Calculating Error Probabilities Over Fading Channels,” IEEE Trans. Commun.,
vol. 53, no. 5, pp. 841–852, May 2005.

[23] M. Ardakani, T. Esmailian, and F. R. Kschischang, “Near-Capacity Coding in

Multicarrier Modulation Systems,” IEEE Trans. Commun., vol. 52, no. 11, pp.
1880–1889, Nov. 2004.
 150

[24] A. G. Armada, “Phase Noise and Sub-Carrier Spacing Effects on the

Performance of an OFDM Communication System,” IEEE Commun. Lett.,
vol. 2, no. 1, pp. 11–13, Jan. 1998.

[25] ——, “Understanding the Effects of Phase Noise in Orthogonal Frequency

Division Multiplexing (OFDM),” IEEE Trans. Broadcast., vol. 47, no. 2, pp.
153–159, June 2001.

[26] S. Armour, A. Nix, and D. Bull, “The Impact of FFT Size on the Performance of
a Combined OFDM-Equalization Radio Modem,” in Proc. IEEE VTC, vol. 3,
Sept. 1999, pp. 1506–1510.

[27] J. Armstrong, “Peak-to-average power reduction for OFDM by repeated clipping

and frequency domain filtering,” IEE Electr. Lett., vol. 38, no. 5, pp. 246–247,
Feb. 2002.

[28] ——, “Analysis of New and Existing Methods of Reducing Intercarrier

Interference Due to Carrier Frequency Offset in OFDM,” IEEE Trans. Commun.,
vol. 47, no. 3, pp. 365–369, Mar. 1999.

[29] ——, “New OFDM Peak-to-Average Power Reduction Scheme,” in Proc. IEEE
VTC, vol. 1, Rhodes, Greece, May 2001, pp. 756–760.

[30] V. Aue and G. P. Fettweis, “A Comparison of the Performance of Linearly

Equalized Single Carrier and Coded OFDM Over Frequency Selective Fading
Channels Using The Random Coding Technique,” in Proc. IEEE ICC, vol. 2,
Atlanta, June 1998, pp. 753–757.

[31] T. Aulin, N. Rydbeck, and C.-E. W. Sundberg, “Continuous Phase

Modulation—Part II: Partial Response Signaling,” IEEE Trans. Commun.,
vol. 29, no. 3, pp. 210–224, Mar. 1981.

[32] T. Aulin and C.-E. Sundberg, “Exact Asymptotic Behavior of Digital FM

Spectra,” IEEE Trans. Commun., vol. 30, no. 11, pp. 2438–2449, Nov. 1982.

[33] ——, “Minimum Euclidean Distance and Power Spectrum for a Class of
Smoothed Phase Modulation Codes with Constant Envelope,” IEEE Trans.
Commun., vol. 30, no. 7, pp. 1721–1729, July 1982.

[34] ——, “An easy way to calculate power spectra of digital FM,” IEE Proc., vol.
130, no. 6, pp. 519–526, Oct. 1983.

[35] T. Aulin and C.-E. W. Sundberg, “Continuous Phase Modulation—Part I: Full

Response Signaling,” IEEE Trans. Commun., vol. 29, no. 3, pp. 196–209, Mar.
1981.

[36] N. J. Baas and D. P. Taylor, “Bandwidth Expansion for Robust, Low-Complexity

Communication Over Fading Dispersive Channels,” IEEE Trans. Commun.,
vol. 49, no. 10, pp. 1748–1754, Oct. 2001.
 151

[37] ——, “Pulse Shaping for Wireless Communication Over Time- or

Frequency-Selective Channels,” IEEE Trans. Commun., vol. 52, no. 9, pp.
1477–1479, Sept. 2004.

[38] A. R. S. Bahai and B. R. Saltzberg, Multi-Carrier Digital Communications:

Theory and Applications of OFDM. New York: Kluwer Academic / Plenum
Publishers, 1999.

[39] A. R. S. Bahai, M. Singh, A. J. Goldsmith, and B. R. Saltzberg, “A New

Approach for Evaluating Clipping Distortion in Multicarrier Systems,” IEEE J.
Select. Areas Commun., vol. 20, no. 5, pp. 1037–1046, June 2002.

[40] P. Banelli, “Theoretical Analysis and Performance of OFDM Signals in Nonlinear

Fading Channels,” IEEE Trans. Wireless Commun., vol. 2, no. 2, pp. 284–293,
Mar. 2003.

[41] P. Banelli, G. Baruffa, and S. Cacopardi, “Effects of HPA Non Linearity on

Frequency Multiplexed OFDM Signals,” IEEE Trans. Broadcast., vol. 47, no. 2,
pp. 123–136, June 2001.

[42] P. Banelli and S. Cacopardi, “Theoretical Analysis and Performance of OFDM

Signals in Nonlinear AWGN Channels,” IEEE Trans. Commun., vol. 48, no. 3,
pp. 430–441, Mar. 2000.

[43] I. Bar-David and S. S. (Shitz), “On Information Transfer by

Envelope-Constrained Signals over the AWGN Channel,” IEEE Trans. Inform.
Theory, vol. 34, no. 3, pp. 371–379, May 1988.

[44] ——, “On the Rice Model of Noise in FM Receiver,” IEEE Trans. Inform.
Theory, vol. 34, no. 6, pp. 1406–1419, Nov. 1988.

[45] S. Barbarossa and R. Torti, “Chirped-OFDM For Transmissions over

Time-Varying Channels with Linear Delay/Doppler Spreading,” in Proc. IEEE
ICASSP, vol. 4, Salt Lake City, May 2001, pp. 2377–2380.

[46] A. Barbieri, G. Cariolaro, and L. Vangelista, “Nonlinear Models of TWT

Revisited for OFDM Systems,” in Proc. Midwest Symp. Circ. Sys., vol. 1, Aug.
1995, pp. 522–525.

[47] C. W. Baum and K. P. Conner, “A Multicarrier Transmission Scheme for

Wireless Local Communications,” IEEE J. Select. Areas Commun., vol. 14, no. 3,
pp. 521–529, Apr. 1996.

[48] R. W. Bäuml, R. F. H. Fischer, and J. B. Huber, “Reducing the peak-to-average

power ratio of multicarrier modulation by selected mapping,” IEE Electr. Lett.,
vol. 32, no. 22, pp. 2056–2057, Oct. 1996.

[49] N. C. Beaulieu, C. C. Tan, and M. O. Damen, “A “Better Than” Nyquist Pulse,”

IEEE Commun. Lett., vol. 5, no. 9, pp. 367–368, Sept. 2001.
 152

[50] P. A. Bello, “Characterization of Randomly Time-Variant Linear Channels,”

IEEE Trans. Commun., pp. 360–393, Dec. 1963.

[51] ——, “Selective Fading Limitations of the Kathryn Modem and Some System
Design Considerations,” IEEE Trans. Commun., pp. 320–333, Sept. 1965.
[52] P. A. Bello and B. D. Nelin, “The Effect of Frequency Selective Fading on
Intermodulation Distortion and Subcarrier Phase Stability in Frequency
Modulation Systems,” IEEE Trans. Commun., pp. 87–101, Mar. 1964.
[53] M.-G. D. Benedetto and P. Mandarini, “An Application of MMSE Predistortion
to OFDM System,” IEEE Trans. Commun., vol. 44, no. 11, pp. 1417–1420, Nov.
1996.

[54] N. Benvenuto and S. Tomasin, “On the Comparison Between OFDM and Single
Carrier Modulation With a DFE Using a Frequency-Domain Feedforward Filter,”
IEEE Trans. Commun., vol. 50, no. 6, pp. 947–955, June 2002.

[55] K. Berberidis and J. Palicot, “A Frequency-Domain Decision Feedback Equalizer

for Multipath Echo Cancellation,” in Proc. IEEE Globecom, vol. 1, Singapore,
Nov. 1995, pp. 98–102.
[56] P. Bianchi, P. Loubaton, and F. Sirven, “On the Blind Estimation of the
Parameters of Continuous Phase Modulated Signals,” IEEE J. Select. Areas
Commun., vol. 23, no. 5, pp. 944–962, May 2005.
[57] E. Biglieri, S. Barberis, and M. Catena, “Analysis and Compensation of
Nonlinearities in Digital Transmission Systems,” IEEE J. Select. Areas
Commun., vol. 6, no. 1, pp. 42–51, Jan. 1988.

[58] E. Biglieri, M. Elia, and L. Lopresti, “The Optimal Linear Receiving Filter for
Digital Transmission Over Nonlinear Channels,” IEEE Trans. Inform. Theory,
vol. 35, no. 3, pp. 620–625, May 1989.

[59] E. Biglieri and M. Visintin, “A Simple Derivation of the Power Spectrum of

Full-Response CPM and Some of its Properties,” IEEE Trans. Commun., vol. 38,
no. 3, pp. 267–269, Mar. 1990.
[60] H. Bilgekul and H. Amea, “Multicarrier transmission using quadrature
frequency/phase modulation,” IEE Electr. Lett., vol. 32, no. 22, pp. 2051–2053,
Oct. 1996.
[61] J. A. C. Bingham, “Multicarrier Modulation for Data Transmission: An Idea
Whose Time Has Come,” IEEE Commun. Mag., pp. 5–14, May 1990.

[62] S. Boyd, “Multitone Signals with Low Crest Factor,” IEEE Trans. Circuits Syst.,
vol. 33, no. 10, pp. 1018–1022, Oct. 1986.
[63] A. Brajal and A. Chouly, “Compensation of nonlinear distortions for orthogonal
multicarrier schemes using predistortion,” in Proc. IEEE Globecom, vol. 3, San
Francisco, Nov. 1994, pp. 1909–1914.
 153

[64] J. Brown and R. Churchill, Complex Variables and Applications, 6th ed.
McGraw-Hill, 1996.

[65] K. Brüninghaus and H. Rohling, “Multi-Carrier Spread Spectrum and its

Relationship to Single-Carrier Transmission,” in Proc. IEEE VTC, vol. 3,
Ottawa, May 1998, pp. 2329–2332.

[66] S. B. Bulumulla, S. A. Kassam, and S. S. Venkatesh, “A Systematic Approach to

Detecting OFDM Signals in a Fading Channel,” IEEE Trans. Commun., vol. 48,
no. 5, pp. 725–728, May 2000.

[67] J. J. Bussgang, L. Ehrman, and J. W. Graham, “Analysis of Nonlinear Systems

with Multiple Inputs,” Proc. IEEE, vol. 62, no. 8, pp. 1088–1119, Aug. 1974.

[68] C. R. Cahn, “Performance of Digital Phase-Modulation Communication

Systems,” IRE Trans. Commun., pp. 3–6, May 1959.

[69] M. Campanella, U. L. Faso, and G. Mamola, “Optimum Bandwidth-Distance

Performance in Full Response CPM Systems,” IEEE Trans. Commun., vol. 36,
no. 10, pp. 1110–1118, Oct. 1988.

[70] ——, “Some Results on the Spectral Efficiency Improvement in Partial Response
CPM Systems,” in Proceedings of the IEEE National Aerospace and Electronics
Conference, vol. 1, Dayton, OH, USA, May 1989, pp. 53–58.

[71] A. J. Cann, “Nonlinearity Model With Variable Knee Sharpness,” IEEE Trans.
Aerosp. Electron. Syst., vol. 16, no. 6, pp. 874–877, Nov. 1980.

[72] G. Cariolaro, G. Michieletto, G. Stivanello, and L. Vangelista, “Spectral Analysis

at the Output of a TWT Driven by an OFDM Signal,” in Proc. ICCS, vol. 2,
Singapore, Nov. 1994, pp. 653–657.

[73] G. Cariolaro and F. C. Vagliani, “An OFDM Scheme with a Half Complexity,”
IEEE J. Select. Areas Commun., vol. 13, no. 9, pp. 1586–1599, Dec. 1995.

[74] G. Carron, R. Ness, L. Deneire, L. V. der Perre, and M. Engles, “Comparison of

Two Modulation Techniques using Frequency Domain Processing for In-House
Networks,” IEEE Trans. Consumer Electron., vol. 47, no. 1, pp. 63–72, Feb. 2001.

[75] N. Carson and T. A. Gulliver, “Performance of OFDM With Modified RA Codes

and Clipping,” IEEE Commun. Lett., vol. 7, no. 9, pp. 440–442, Sept. 2003.

[76] E. F. Casas and C. Leung, “OFDM for Data Communication Over Mobile Radio
FM Channels–Part I: Analysis and Experimental Results,” IEEE Trans.
Commun., vol. 39, no. 5, pp. 783–793, May 1991.

[77] ——, “OFDM for Data Communication Over Mobile Radio FM Channels–Part
II: Performance Improvement,” IEEE Trans. Commun., vol. 40, no. 4, pp.
680–683, Apr. 1992.
 154

[78] R. Castle and A. Jones, “A 20 Mbits/s OFDM Demodulator at 5 GHz: System

Design, Implementation and Experimental Results,” in Proc. IEEE VTC, vol. 2,
Ottawa, May 1998, pp. 1295–1299.

[79] A. C. Caswell, “Multicarrier transmission in a mobile radio channel,” IEE Electr.

Lett., vol. 32, no. 21, pp. 1962–1963, Oct. 1996.

[80] R. Cendrillion and M. Moonen, “Efficient Equalizers for Single and Multi-Carrier
Environments with Known Symbol Padding,” in Proc. IEEE ISSPA, vol. 2,
Kuala Lumpur, Aug. 2001, pp. 607–610.

[81] M.-X. Chang and Y. T. Su, “Performance Analysis of Equalized OFDM Systems
in Rayleigh Fading,” IEEE Trans. Wireless Commun., vol. 1, no. 4, pp. 721–732,
Oct. 2002.

[82] R. W. Chang, “Orthogonal frequency division multiplexing,” US patent

3,488,445, filed Nov. 14, 1966, issued Jan. 6, 1970.

[83] ——, “Synthesis of band-limited orthogonal signals for multichannel data

transmission,” Bell Syst. Tech. J., vol. 45, pp. 1775–1796, Dec. 1966.

[84] R. W. Chang and R. A. Gibby, “A Theoretical Study of Performance of an

Orthogonal Multiplexing Data Transmission Scheme,” IEEE Trans. Commun.
Technol., vol. com-16, no. 4, pp. 529–540, Aug. 1968.

[85] S. Chang and E. J. Powers, “A Simplified Predistorter for Compensation of

Nonlinear Distortion in OFDM Systems,” in Proc. IEEE Globecom, vol. 5, San
Antonio, Nov. 2001, pp. 3080–3084.

[86] S. Chang, E. J. Powers, and J. Chung, “A Compensation Scheme for Nonlinear

Distortion in OFDM Systems,” in Proc. IEEE Globecom, vol. 2, San Francisco,
Nov. 2000, pp. 736–740.

[87] H. Chen and A. M. Haimovich, “Iterative Estimation and Cancellation of

Clipping Noise for OFDM Signals,” IEEE Commun. Lett., vol. 7, no. 7, pp.
305–307, July 2003.

[88] H. Chen and G. J. Pottie, “An Orthogonal Projection-Based Approach for PAR
Reduction in OFDM,” IEEE Commun. Lett., vol. 6, no. 5, pp. 169–171, May
2002.

[89] W. cheol Lee, H.-M. Park, K. jin Kang, and K. bae Kim, “Performance Analysis
of Viterbi Decoder Using Channel State Information in COFDM System,” IEEE
Trans. Broadcast., vol. 44, no. 4, pp. 488–496, Dec. 1998.

[90] C. Cherubini, E. Eleftheriou, and S. O. Ölçer, “Filtered Multitone Modulation

for Very High-Speed Digital Subscriber Lines,” IEEE J. Select. Areas Commun.,
vol. 20, no. 5, pp. 1016–1028, June 2002.
 155

[91] E. Chiavaccini and G. M. Vitetta, “Maximum-Likelihood Frequency Recovery for

OFDM Signals Transmitted Over Multipath Fading Channels,” IEEE Trans.
Commun., vol. 52, no. 2, pp. 244–251, Feb. 2004.
[92] A. Chini, Y. Wu, M. El-Tanany, and S. Mahmoud, “Filtered Decision Feedback
Channel Estimation for OFDM-Based DTV Terrestrial Broadcasting Systems,”
IEEE Trans. Broadcast., vol. 44, no. 1, pp. 2–11, Mar. 1998.
[93] ——, “Hardware Nonlinearities in Digital TV Broadcasting Using OFDM
Modulation,” IEEE Trans. Broadcast., vol. 44, no. 1, pp. 12–21, Mar. 1998.
[94] Y.-S. Choi, P. J. Voltz, and F. A. Cassara, “On Channel Estimation and
Detection for Multicarrier Signals in Fast and Selective Rayleigh Fading
Channels,” IEEE Trans. Commun., vol. 49, no. 8, pp. 1375–1387, Aug. 2001.
[95] J. S. Chow, J. C. Tu, and J. M. Cioffi, “A Discrete Multitone Transceiver System
for HDSL Applications,” IEEE J. Select. Areas Commun., vol. 9, pp. 895–908,
Aug. 1991.
[96] P. S. Chow, J. M. Cioffi, and J. A. C. Bingham, “A Practical Discrete Multitone
Transceiver Loading Algorithm for Data Transmission over Spectrally Shaped
Channels,” IEEE Trans. Commun., vol. 43, no. 2/3/4, pp. 773–775,
Feb./Mar./Apr. 1995.
[97] P. S. Chow, J. C. Tu, and J. M. Cioffi, “Performance Evaluation of a
Multichannel Transceiver system for ADSL and VHDSL Services,” IEEE J.
Select. Areas Commun., vol. 9, pp. 909–919, Aug. 1991.
[98] J. Chuang, L. J. Cimini, Jr., G. Y. Li, B. McNair, N. Sollenberger, H. Zhao,
L. Lin, and M. Suzuki, “High-Speed Wireless Data Access Based on Combining
EDGE with Wideband OFDM,” IEEE Commun. Mag., pp. 92–98, Nov. 1999.
[99] J. Chuang and N. Sollenberger, “Beyond 3G: Wideband Wireless Data Access
Based on OFDM and Dynamic Packet Assignment,” IEEE Commun. Mag., pp.
78–87, July 2000.
[100] J. C.-I. Chuang, “The Effects of Time Delay Spread on Portable Radio
Communications Channels with Digital Modulation,” IEEE J. Select. Areas
Commun., vol. SAC-5, no. 5, pp. 879–889, June 1987.
[101] C.-D. Chung and S.-M. Cho, “Constant-Envelope Orthogonal Frequency Division
Multiplexing Modulation,” in Proc. APCC/OECC, vol. 1, Beijing, Oct. 1999, pp.
629–632.
[102] L. J. Cimini, Jr., “Analysis and Simulation of a Digital Mobile Channel Using
Orthogonal Frequency Division Multiplexing,” IEEE Trans. Commun., vol. 33,
no. 7, pp. 665–675, July 1985.
[103] L. J. Cimini, Jr., B. Daneshrad, and N. R. Sollenberger, “Clustered OFDM with
Transmitter Diversity and Coding,” in Proc. IEEE Globecom, vol. 1, London,
Nov. 1996, pp. 703–707.
 156

[104] L. J. Cimini, Jr. and N. R. Sollenberger, “Peak-to-Average Power Ratio

Reduction of an OFDM Signal Using Partial Transmit Sequences,” IEEE
Commun. Lett., vol. 4, no. 5, pp. 86–88, Mar. 2000.

[105] J. M. Cioffi and J. A. C. Bingham, “A Data-Driven Multitone Echo Canceller,”

IEEE Trans. Commun., vol. 42, no. 10, pp. 2853–2869, Oct. 1994.

[106] S. Clanton, “Speeding Up the Scientific Process,” Linux Journal, no. 110, pp.
56–60, June 2003.

[107] M. V. Clark, “Adaptive Frequency-Domain Equalization and Diversity

Combining for Broadband Wireless Communications,” IEEE J. Select. Areas
Commun., vol. 16, no. 8, pp. 1385–1395, Oct. 1998.

[108] I. B. Collings and I. V. L. Clarkson, “A Low-Complexity Lattice-Based Low-PAR

Transmission Scheme for DSL Channels,” IEEE Trans. Commun., vol. 52, no. 5,
pp. 755–764, May 2004.

[109] M. S. Corrington, “Frequency-Modulation Distortion Caused by Multipath

Transmission,” Proc. IRE, pp. 878–891, Dec. 1945.

[110] E. Costa, M. Midrio, and S. Pupolin, “Impact of Amplifier Nonlinearities on

OFDM Transmission System Performance,” IEEE Commun. Lett., vol. 3, no. 2,
pp. 37–39, Feb. 1999.

[111] E. Costa and S. Pupolin, “M -QAM-OFDM System Performance in the Presence

of Nonlinear Amplifier and Phase Noise,” IEEE Trans. Commun., vol. 50, no. 3,
pp. 462–472, Mar. 2002.

[112] A. J. Coulson, “Maximum Likelihood Synchronization for OFDM Using a Pilot

Symbol: Algorithms,” IEEE J. Select. Areas Commun., vol. 19, no. 12, pp.
2486–2494, Dec. 2001.

[113] ——, “Maximum Likelihood Synchronization for OFDM Using a Pilot Symbol:
Analysis,” IEEE J. Select. Areas Commun., vol. 19, no. 12, pp. 2495–2503, Dec.
2001.

[114] S. C. Cripps, Advanced Techniques in RF Power Amplifier Design. Boston:

Artech House, 2002.

[115] M. G. Crosby, “Frequency Modulation Propagation Characteristics,” Proc. IRE,

1936.

[116] A. Czylwik, “Comparison Between Adaptive OFDM and Single Carrier

Modulation with Frequency Domain Equalization,” in Proc. IEEE VTC, vol. 2,
Phoenix, May 1997, pp. 865–869.

[117] ——, “Low Overhead Pilot-Aided Synchronization for Single Carrier Modulation
With Frequency Domain Equalization,” in Proc. IEEE Globecom, vol. 4, Sydney,
Nov. 1998, pp. 2068–2073.
 157

[118] R. Dabora, J. Goldberg, and H. Messer, “Inherent Limitations in Data-Aided

Time Synchronization of Continuous Phase-Modulation Signals Over
Time-Selective Fading Channels,” IEEE Trans. Signal Processing, vol. 50, no. 6,
pp. 1470–1482, June 2002.

[119] H. Dai and H. V. Poor, “Advanced Signal Processing for Power Line
Communications,” IEEE Commun. Mag., vol. 41, no. 5, pp. 100–107, May 2003.

[120] A. N. D’Andrea, A. Ginesi, and U. Mengali, “Frequency Detectors for CPM

Signals,” IEEE Trans. Commun., vol. 43, no. 2/3/4, pp. 1828–1837,
Feb./Mar./Apr. 1995.

[121] A. N. D’Andrea, V. Lottici, and R. Reggiannini, “RF Power Amplifier

Linearization Through Amplitude and Phase Distortion,” IEEE Trans.
Commun., vol. 44, no. 11, pp. 1477–1484, Nov. 1996.

[122] ——, “Nonlinear Predistortion of OFDM Signals over Frequency-Selective Fading

Channels,” IEEE Trans. Commun., vol. 49, no. 5, pp. 837–843, May 2001.

[123] D. Dardari and V. Tralli, “High-Speed Indoor Wireless Communications at 60

GHz with Coded OFDM,” IEEE J. Select. Areas Commun., vol. 47, no. 11, pp.
1709–1721, Nov. 1999.

[124] D. Dardari, V. Tralli, and A. Vaccari, “A Theoretical Characterization of

Nonlinear Distortion Effects in OFDM Systems,” IEEE Trans. Commun., vol. 48,
no. 10, pp. 1755–1764, Oct. 2000.

[125] J. A. Davis and J. Jedwab, “Peak-to-mean power control and error correction for
OFDM transmission using Golay sequences and Reed-Muller codes,” IEE Electr.
Lett., vol. 33, no. 4, pp. 267–268, Feb. 1997.

[126] ——, “Peak-to-mean Power Control in OFDM, Golay Complementary Sequences,

and Reed-Muller Codes,” IEEE Trans. Inform. Theory, vol. 45, no. 7, pp.
2397–2417, Nov. 1999.

[127] F. de Jager and C. B. Dekker, “Tamed Frequency Modulation, A Novel Method

to Achieve Spectrum Economy in Digital Transmission,” IEEE Trans. Commun.,
vol. 26, no. 5, pp. 534–542, May 1978.

[128] L. De Rose and D. Padua, “Techniques for the Translation of MATLAB

Programs into Fortran 90,” ACM Trans. Prog. Lang. and Syst., vol. 21, no. 2, pp.
286–323, Mar. 1999.

[129] G. de Veciana and A. Zokhor, “Neural Net-Based Continuous Phase Modulation

Receivers,” IEEE Trans. Commun., vol. 40, no. 8, pp. 1396–1408, Aug. 1992.

[130] M. Debbah, P. Loubaton, and M. de Courville, “Asymptotic Performance of

Successive Interference Cancellation in the Context of Linear Precoded OFDM
Systems,” IEEE Trans. Commun., vol. 52, no. 9, pp. 1444–1448, Sept. 2004.
 158

[131] Debian—The Universal Operating System. [Online]. Available:

http://www.debian.org/

[132] L. Deneire, B. Gyselinckx, and M. Engels, “Training Sequence versus Cyclic

Prefix—A New Look on Single Carrier Communication,” IEEE Commun. Lett.,
vol. 5, no. 7, pp. 292–294, July 2001.

[133] P. Dent, G. E. Bottomley, and T. Croft, “Jakes Fading Model Revisited,” IEE
Electr. Lett., vol. 29, no. 13, pp. 1162–1163, June 1993.

[134] L. Ding, G. T. Zhou, D. R. Morgan, Z. Ma, J. S. Kenny, J. Kim, and C. R.

Giardina, “A Robust Digital Baseband Predistorter Constructed Using Memory
Polynomials,” IEEE Trans. Commun., vol. 52, no. 1, pp. 159–165, Jan. 2004.

[135] R. Dinis and A. Gusmão, “Carrier synchronization with CEPB-OFDM,” in Proc.

IEEE VTC, vol. 3, May 1997, pp. 1370–1374.

[136] ——, “Performance evaluation of a multicarrier modulation technique allowing

strongly nonlinear amplification,” Proc. IEEE ICC, vol. 2, pp. 791–796, 1998.

[137] ——, “On the Performance Evaluation of OFDM Transmission Using Clipping
Techniques,” in Proc. IEEE VTC, vol. 5, Amsterdam, Sept. 1999, pp. 2923–2928.

[138] ——, “A Class of Nonlinear Signal-Processing Schemes for Bandwidth-Efficient

OFDM Transmission With Low Envelope Fluctuation,” IEEE Trans. Commun.,
vol. 52, no. 11, pp. 2009–2018, Nov. 2004.

[139] N. Dinur and D. Wulich, “Peak-to-Average Power Ratio in High-Order OFDM,”

IEEE Trans. Commun., vol. 49, no. 6, pp. 1063–1072, June 2001.

[140] D. Divsalar and M. K. Simon, “Multiple-Symbol Differential Detection of

MPSK,” IEEE Trans. Commun., vol. 38, no. 3, pp. 300–308, Mar. 1990.

[141] B. J. Dixon, R. D. Pollard, and S. Iezekiel, “A discussion of the effects of

amplifier back-off on OFDM,” in High Frequency Postgraduate Student
Colloquium, Leeds, Sept. 1999, pp. 14–19.

[142] P. A. Dmochowski and P. J. McLane, “Frequency Domain Equalization for High

Data Rate Multipath Channels,” in Proc. IEEE PACRIM, vol. 2, Victoria, BC,
Aug. 2001, pp. 534–537.

[143] O. Dural and J. G. Proakis, “Signal Space Representation of Binary Continuous

Phase Modulated Signals,” in Proc. IEEE VTC, vol. 3, Atlantic City, NJ, Oct.
2001, pp. 1663–1667.

[144] O. Edfors, M. Sandell, J.-J. van de. Beek, S. K. Wilson, and P. O. Börjesson,
“OFDM Channel Estimation by Singular Value Decomposition,” IEEE Trans.
Commun., vol. 46, no. 7, pp. 931–939, July 1998.
 159

[145] P. V. Eetvelt, G. Wade, and M. Tomlinson, “Peak to average power reduction for
OFDM schemes by selective scrambling,” IEE Electr. Lett., vol. 32, no. 21, pp.
1963–1964, Oct. 1996.

[146] H. M. Eissa, K. Sharaf, and H. Ragaie, “ARCTAN Differentiated Digital

Demodulator for FM/FSK Digital Receivers,” in The 2002 45th Midwest
Symposium on Circuits and Systems, vol. 2, Aug. 2002, pp. II–200—II–203.

[147] M. S. El-Tanany, Y. Wu, and L. Házy, “Analytical Modeling and Simulation of

Phase Noise Interference in OFDM-Based Digital Television Terrestrial
Broadcasting Systems,” IEEE Trans. Broadcast., vol. 47, no. 1, pp. 20–31, Mar.
2001.

[148] ——, “OFDM Uplink for Interactive Broadband Wireless: Analysis and
Simulation in the Presence of Carrier, Clock and Timing Errors,” IEEE Trans.
Broadcast., vol. 47, no. 1, pp. 3–19, Mar. 2001.

[149] E. Eleftheriou, S. Ölçer, and H. Sadjadpour, “Application of

Capacity-Approaching Coding techniques to Digital Subscriber Lines,” IEEE
Commun. Mag., pp. 88–94, Apr. 2004.

[150] S. M. Elnoubi, “Predetection Filtering Effect on the Probability of Error of

GMSK with Discriminator Detection in Mobile Radio Channels,” IEEE Trans.
Veh. Technol., vol. 37, no. 2, pp. 104–107, May 1988.

[151] D. Elphick, M. Leuschel, and S. Cox, “Partial evaluation of matlab,” in GPCE

’03: Proceedings of the second international conference on Generative
programming and component engineering. Springer-Verlag New York, Inc., 2003,
pp. 344–363.

[152] T. Erseghe, N. Laurenti, and V. Cellini, “A Multicarrier Architecture Based

Upon the Affine Fourier Transform,” IEEE Trans. Commun., vol. 53, no. 5, pp.
853–862, May 2005.

[153] D. Falconer, S. L. Ariyavisitakul, A. Benyamin-Seeyar, and B. Eidson,

“Frequency Domain Equalization for Single-Carrier Broadband Wireless
Systems,” IEEE Commun. Mag., pp. 58–66, Apr. 2002.

[154] D. D. Falconer and S. L. Ariyavisitakul, “Broadband Wireless Using Single

Carrier and Frequency Domain Equalization,” in Proc. IEEE Wireless Pers.
Multi. Commun., vol. 1, Oct. 2002, pp. 27–36.

[155] B. Farhang-Boroujeny and M. Ding, “Design Methods for Time-Domain

Equalizers in DMT Transceivers,” IEEE Trans. Commun., vol. 49, no. 3, pp.
554–562, Mar. 2001.

[156] C. W. Farrow, “A Continuously Variable Digital Delay Element,” in Proc. IEEE

Symp. Circuits. Syst., vol. 3, Espoo, Finland, June 1988, pp. 2641–2645.
 160

[157] O. Feely, “Nonlinear Dynamics of First-order DPLL with FM Input,” in Proc.

IEEE Symp. Circuits. Syst., vol. 4, Geneva, May 2000, pp. 477–480.

[158] J. Fernándex, M. Anguita, S. Mota, A. Cañas, E. Ortigosa, and F. J. Rojas,

“MPI Toolbox for Octave,” in Proc. VECPAR, Velencia, Spain, June 2004.

[159] R. F. H. Fischer, L. H.-J. Lampe, and S. H. Müller-Weinfurtner, “Coded

Modulation for Noncoherent Reception with Application to OFDM,” IEEE
Trans. Veh. Technol., vol. 2001, no. 4, pp. 910–919, July 2001.

[160] “OFDM for Mobile Data Communications,” Flarion Technologies. [Online].

Available: http://www.flarion.com/

[161] B. L. Floch, M. Alard, and C. Berrou, “Coded Orthogonal Frequency Division

Multiplex,” Proc. IEEE, vol. 83, no. 6, pp. 982–996, June 1995.

[162] B. L. Floch, R. Halbert-Lassalie, and D. Castelain, “Digital Sound Broadcasting

to Mobile Receivers,” IEEE Trans. Consumer Electron., vol. 35, no. 3, pp.
493–503, Aug. 1989.

[163] P. Fortier, A. Ruiz, and J. M. Cioffi, “Multidimensional Signal Sets Through the
Shell Construction for Parallel Channels,” IEEE Trans. Commun., vol. 40, no. 3,
pp. 500–512, Mar. 1992.

[164] M. E. Fox and M. W. Marcellin, “Shaped BPSK and the 5 kHz UHF Satcom
Channels,” in Proc. IEEE Milcom, vol. 1, Nov. 1991, pp. 326–332.

[165] S. Fragiacomo, C. Matrakidis, and J. J. O’Reilly, “Multicarrier transmission

peak-to-average power reduction using simple block code,” IEE Electr. Lett.,
vol. 34, no. 10, pp. 953–954, May 1998.

[166] M. Freidhof and P. Havezov, “A Numerical Method for Spectral Shaping of CPM
Signals,” in Proc. IEEE Globecom, vol. 5, Sydney, Australia, Nov. 1998, pp.
2859–2863.

[167] G. Freiman, S. Litsyn, and A. Yudin, “A Method to Suppress High Peaks in

BPSK-Modulated OFDM Signal,” IEEE Trans. Commun., vol. 52, no. 9, pp.
1440–1443, Sept. 2004.

[168] P. K. Frenger and N. A. B. Svensson, “Parallel Combinatory OFDM Signaling,”

IEEE Trans. Commun., vol. 47, no. 4, pp. 558–567, Apr. 1999.

[169] M. Friese, “Multicarrier modulation with low peak-to-average power ratio,” IEE
Electr. Lett., vol. 32, no. 8, pp. 713–714, Apr. 1996.

[170] ——, “Multitone Signals with Low Crest Factor,” IEEE Trans. Commun.,
vol. 45, no. 10, pp. 1338–1344, Oct. 1997.

[171] ——, “OFDM Signals with Low Crest-Factor,” in Proc. IEEE Globecom,
Phoenix, Nov. 1997, pp. 290–294.
 161

[172] ——, “On the degradation of OFDM-signals due to peak-clipping in optimally

predistorted power amplifiers,” in Proc. IEEE Globecom, vol. 2, Sydney, Nov.
1998, pp. 939–944.

[173] ——, “On the Achievable Information Rate with Peak-Power-Limited Orthogonal
Frequency-Division Multiplexing,” IEEE Trans. Inform. Theory, vol. 46, no. 7,
pp. 2579–2587, Nov. 2000.

[174] N. J. Frigo, “A Model of Intermodulation Distortion in Non-Linear Multicarrier

Systems,” IEEE Trans. Commun., vol. 42, no. 3, pp. 1216–1222, Apr. 1994.

[175] I. Galton, “Analog-Input Digital Phase-Locked Loops for Precise Frequency and
Phase Demodulation,” IEEE Trans. Circuits Syst. II, vol. 42, no. 10, pp.
621–630, Oct. 1995.

[176] M. J. F.-G. Garcia, S. Zazo, and J. M. Páez-Barrello, “Pilot patters for channel
estimation in OFDM,” IEE Electr. Lett., vol. 36, no. 12, pp. 1049–1050, June
2000.

[177] F. M. Gardner and J. F. Heck, “Phaselock Loop Cycle Slipping Caused by

Excessive Angle Modulation,” IEEE Trans. Commun., vol. com-26, no. 8, pp.
1307–1309, Aug. 1978.

[178] W. A. Gardner, Ed., Cyclostationarity in Communications and Signal Processing.

New York: IEEE, 1994.

[179] W. A. Gardner, “On “The Optimal Linear Receiving Filter for Digital
Transmission Over Nonlinear Channels”,” IEEE Trans. Inform. Theory, vol. 37,
no. 1, p. 219, Jan. 1991.

[180] I. Ghareeb, “Bit Error Rate Performance and Power Spectral Density of a
Noncoherent Hybrid Frequency-Phase Modulation System,” IEEE J. Select.
Areas Commun., vol. 13, no. 2, pp. 276–284, Feb. 1995.

[181] M. Ghogho and A. Swami, “Blind Frequency-Offset Estimator for OFDM

Systems Transmitting Constant-Modulus Symbols,” IEEE Commun. Lett., vol. 6,
no. 8, pp. 343–345, Aug. 2002.

[182] M. Ghosh, “Analysis of the Effect of Impulse Noise on Multicarrier and Single
Carrier QAM Systems,” IEEE Trans. Commun., vol. 44, no. 2, pp. 145–147, Feb.
1996.

[183] J. D. Gibson, Principles of Digital and Analog Communications. New York:

Macmillan, 1989.

[184] S. Gifford, J. Kleider, and S. Chuprun, “OFDM Transmitter Power Amplifier and
PAR Reduction Performance: Measurement and Simulation,” in Proc. IEEE
Milcom, vol. 1, Oct. 2002, pp. 591–595.
 162

[185] S. Gifford, J. E. Kleider, and S. Chuprun, “Broadband OFDM Using 16-bit

Precision on a SDR Platform,” in Proc. IEEE Milcom, vol. 1, Oct. 2001, pp.
180–184.

[186] D. R. Gimlin and C. R. Patisaul, “On Minimizing the Peak-to-Average Power

Ratio of the Sum of N Sinusoids,” IEEE Trans. Commun., vol. 41, no. 2, pp.
631–635, Apr. 1993.

[187] A. Glavieux, P. Y. Cochet, and A. Picart, “Orthogonal Frequency Division

Multiplexing with BFSK Modulation in Frequency Selective Rayleigh and Rician
Fading Channels,” IEEE Trans. Commun., vol. 42, no. 2/3/4, pp. 1919–1928,
Feb./Mar./Apr. 1994.

[188] GNU Octave. [Online]. Available: http://www.octave.org/

[189] Gnuplot. [Online]. Available: http://www.gnuplot.info/

[190] D. L. Goeckel, “Coded Modulation with Non-Standard Signal Sets for Wireless
OFDM Systems,” in Proc. IEEE ICC, vol. 2, Vancouver, BC Canada, June 1999,
pp. 791–795.

[191] L. Goldfeld, V. Lyandres, and D. Wulich, “Minimum BER Power Loading for
OFDM in Fading Channel,” IEEE Trans. Commun., vol. 50, no. 11, pp.
1729–1733, Nov. 2002.

[192] M. Goossens, F. Mittelbach, and A. Samarin, The LATEXCompanion. Boston:

Addison-Wesley, 1994.

[193] A. J. Grant and R. D. van Nee, “Efficient Maximum-Likelihood Decoding of

Q-ary Modulated Reed–Muller Codes,” IEEE Commun. Lett., vol. 2, no. 5, pp.
134–136, May 1998.

[194] R. Gross and D. Veeneman, “SNR and Spectral Properties for a Clipped DMT
ADSL Signal,” in Proc. IEEE ICC, vol. 2, New Orleans, May 1994, pp. 843–847.

[195] N. Guo and L. B. Milstein, “The Impact of Nonlinear Amplification of

Multi-Code CDMA Systems,” in Proc. IEEE ICC, vol. 2, New Orleans, June
2000, pp. 1034–1038.

[196] A. Gusmão, R. Dinis, J. Conceição, and N. Esteves, “Comparison of Two

Modulation Choices for Broadband Wireless Communications,” in Proc. IEEE
VTC, vol. 2, Tokyo, May 2000, pp. 1300–1305.

[197] A. Gusmão, R. Dinis, and N. Esteves, “On Frequency Domain Equalization and
Diversity Combining for Broadband Wireless Communications,” IEEE Trans.
Commun., vol. 51, no. 7, pp. 1029–1033, July 2003.

[198] E. Haas and S. Kaiser, “Two-Dimensional Differential Demodulation for OFDM,”

IEEE Trans. Commun., vol. 51, no. 4, pp. 580–586, Apr. 2003.
 163

[199] L. Haendel, “The Function Pointer Tutorials,” 2003. [Online]. Available:

http://www.newty.de/zip/e− fpt.pdf

[200] J. Hamkins, “Accurate Computation of the Performance of M -ary Orthogonal

Signaling on a Discrete Memoryless Channel,” IEEE Trans. Commun., vol. 52,
no. 11, pp. 1844–1845, Nov. 2004.

[201] L. Hanzo, W. Webb, and T. Keller, Single- and Multi-carrier Quadrature

Amplitude Modulation. Chichester: John Wiley & Sons, 2000.

[202] H. F. Harmuth, “On the Transmission of Information by Orthogonal Time

Functions,” AIEE. Trans., vol. 79, pp. 248–255, July 1960.

[203] P. Hasan, “PLL FM Demodulator Performance Under Gaussian Modulation,”

IEEE Trans. Commun., vol. 46, no. 4, pp. 437–440, Apr. 1998.

[204] R. W. Heath, Jr. and G. B. Giannakis, “Exploiting Input Cyclostationarity for

Blind Channel Identification in OFDM Systems,” IEEE Trans. Acoust., Speech,
Signal Processing, vol. 47, no. 3, pp. 848–856, Mar. 1999.

[205] W. Henkel and T. Kessler, “Maximizing the Channel Capacity of Multicarrier

Transmission by Suitable Adaptation of the Time-Domain Equalizer,” IEEE
Trans. Commun., vol. 48, no. 12, pp. 2000–2004, Dec. 2001.

[206] W. Henkel and B. Wagner, “Another Application for Trellis Shaping: PAR
Reduction for DMT (OFDM),” IEEE Trans. Commun., vol. 48, no. 9, pp.
1471–1476, Sept. 2000.

[207] B. Hirosaki, “An Analysis of Automatic Equalizers for Orthogonally Multiplexed

QAM Systems,” IEEE Trans. Commun., vol. com-28, no. 1, pp. 73–83, Jan. 1980.

[208] ——, “An Orthogonally Multiplexed QAM System using the Discrete Fourier
Transform,” IEEE Trans. Commun., vol. 29, no. 7, pp. 982–989, July 1981.

[209] B. Hirosaki, S. Hasegawa, and A. Sabato, “Advanced Groupband Data Modem

Using Orthogonally Multiplexed QAM Technique,” IEEE Trans. Commun., vol.
com-34, no. 6, pp. 587–592, June 1986.

[210] P. Höeher, “A Statistical Discrete-Time Model for the WSSUS Multipath

Channel,” IEEE Trans. Veh. Technol., vol. 41, no. 4, pp. 461–468, Nov. 1992.

[211] J. hong Jong and W. E. Stark, “Performance Analysis of Coded Multicarrier

Spread-Spectrum Systems in the Presence of Multipath Fading and
Nonlinearities,” IEEE Trans. Commun., vol. 49, no. 1, pp. 168–179, Jan. 2001.

[212] J. hong Jong, K. Yang, W. E. Stark, and G. I. Haddad, “Power Optimization of

OFDM Systems with dc Bias Controlled Nonlinear Amplifiers,” in Proc. IEEE
VTC, vol. 1, Amsterdam, Sept. 1999, pp. 268–272.
 164

[213] H.-H. Hsieh and C.-H. Wei, “Channel Estimation for OFDM Systems Based on
Comb-Type Pilot Arrangement in Frequency Selective Fading Channels,” IEEE
Trans. Consumer Electron., vol. 44, no. 1, pp. 217–225, Feb. 1998.

[214] X. Huang, J. Lu, and J. Zheng, “Tradeoffs Between BER Performance and PAPR
Reduction for Broadband OFDM Transmission with Signal Processing,” in
Proceedings of the IEEE International Symposium on Wireless Personal
Multimedia Communications, vol. 1, Oct. 2002, pp. 154–157.

[215] X. Huang, J. Lu, J. Zheng, J. Chuang, and J. Gu, “Reduction of peak-to-average

power ratio of OFDM signals with companding transform,” IEE Electr. Lett.,
vol. 37, no. 8, pp. 506–507, Apr. 2001.

[216] X. Huang and Y. Li, “The PAM Decomposition of CPM Signals with Integer
Modulation Index,” IEEE Trans. Commun., vol. 51, no. 4, pp. 543–546, Apr.
2003.

[217] L. W. Hughes, “A Simple Upper Bound on the Error Probability for Orthogonal
Signals in White Noise,” IEEE Trans. Commun., vol. 40, no. 4, p. 670, Apr. 1992.

[218] C.-S. Hwang, “Peak power reduction method for multicarrier transmission,” IEE
Electr. Lett., vol. 37, no. 17, pp. 1075–1077, Aug. 2001.

[219] T. Hwang and Y. G. Li, “Iterative Cyclic Prefix Reconstruction for Coded
Single-Carrier Systems with Frequency-Domain Equalization (SC-FDE),” in
Proc. IEEE VTC, vol. 3, Apr. 2003, pp. 1841–1845.

[220] ——, “A Bandwidth Efficient Block Transmission with Frequency-Domain

Equalization,” in Proc. IEEE Circuits Syst. Symp., vol. 2, June 2004, pp.
433–436.

[221] Ibiquity—Digital Radio. [Online]. Available: http://www.ibiquity.com/

[222] IEEE Xplore. [Online]. Available: http://ieeexplore.ieee.org/

[223] M. I. Irshid and I. S. Solous, “Bit Error Probability for Coherent M -ary PSK
Systems,” IEEE Trans. Commun., vol. 39, no. 3, pp. 349–352, Mar. 1991.

[224] Y. Iwanami and T. Ikeda, “A Numerical Method for Evaluating the Distortion of
Angle-Modulated Signals in a Time Domain,” IEEE Trans. Commun., vol. 34,
no. 11, pp. 1151–1156, Nov. 1986.

[225] P. C. Jain, “Error Probabilities in Binary Angle Modulation,” IEEE Trans.

Inform. Theory, vol. it-20, no. 1, pp. 36–42, Jan. 1974.

[226] S. Jaugegui, “Complementary Sequences of Length 26,” IEEE Trans. Inform.

Theory, vol. it-8, no. 4, p. 323, July 1962.

[227] A. D. S. Jayalath and C. R. N. Athaudage, “On the PAR Reduction of OFDM

Signals Using Multiple Signal Representation,” IEEE Commun. Lett., vol. 8,
no. 7, pp. 425–427, July 2004.
 165

[228] A. D. S. Jayalath and C. Tellambura, “Reducing the peak-to-average power ratio

of orthogonal frequency division multiplexing signal through bit or symbol
interleaving,” IEE Electr. Lett., vol. 36, no. 13, pp. 1161–1163, June 2000.
[229] J. Jedwab, “Comment: M-sequences for OFDM peak-to-average power ratio
reduction and error correction,” IEE Electr. Lett., vol. 33, no. 15, pp. 1293–1294,
July 1997.
[230] W. G. Jeon, K. H. Chang, and Y. S. Cho, “An Adaptive Data Predistorter for
Compensation of Nonlinear Distortion in OFDM Systems,” IEEE Trans.
Commun., vol. 45, no. 10, pp. 1167–1171, Oct. 1997.
[231] ——, “An Equalization Technique for Orthogonal Frequency-Division
Multiplexing Systems in Time-Variant Multipath Channels,” IEEE Trans.
Commun., vol. 47, no. 1, pp. 27–32, Jan. 1999.
[232] W. Jeong, H. Park, H. Lee, and S. Hwang, “Performance Improvement
Techniques for CCK-OFDM WLAN Modem,” IEEE Trans. Consumer Electron.,
vol. 49, no. 3, pp. 602–605, Aug. 2003.
[233] M. C. Jeruchim, P. Balaban, and K. S. Shanmugan, Simulation of
Communication Systems—Modeling, Methodology, and Techniques, 2nd ed. New
York: Kluwer Academic/Plenum Publishers, 2000.
[234] T. Jiang and G. Zhu, “Complement Block Coding for Reduction of
Peak-to-Average Power Ratio of OFDM Signals,” IEEE Commun. Mag., pp.
S17–S22, Sept. 2005.
[235] D. A. Johnston and S. K. Jones, “Spectrally Efficient Communication via Fading
Channels Using Coded Multilevel DPSK,” IEEE Trans. Commun., vol. com-29,
no. 3, pp. 276–284, Mar. 1981.
[236] P. G. Joisha, A. Kanhere, P. Banerjee, U. N. Shenoy, and A. Choudhary,
“Handling Context-Sensitive Syntactic Issues in the Design of a Front-end for a
MATLAB Compiler,” APL Quote Quad, vol. 31, no. 3, pp. 27–40, Mar. 2001.
[237] A. E. Jones and T. A. Wilkinson, “Combined Coding for Error Control and
Increased Robustness to System Nonlinearities in OFDM,” in Proc. IEEE VTC,
vol. 2, Atlanta, GA, Apr. 1996, pp. 904–908.
[238] ——, “Performance of Reed–Muller Codes and a Maximum-Likelihood Decoding
Algorithm for OFDM,” IEEE Trans. Commun., vol. 47, no. 7, pp. 949–952, July
1999.
[239] A. E. Jones, T. A. Wilkinson, and S. K. Barton, “Block coding scheme for
reduction of peak to mean envelope power ratio of multicarrier transmission
schemes,” IEE Electr. Lett., vol. 30, no. 25, pp. 2098–2099, Dec. 1994.
[240] D. L. Jones, “Peak Power Reduction in OFDM and DMT via Active Channel
Modification,” in Conference Record of the Asilomar Conference on Signals,
Systems and Computers, vol. 2, Pacific Grove, Oct. 1999, pp. 1076–1079.
 166

[241] S. M. Ju and S. H. Leung, “Clipping on COFDM With Phase On Demand,”

IEEE Commun. Lett., vol. 7, no. 2, pp. 49–51, Feb. 2003.

[242] P. Jung, “Laurent’s Representation of Binary Digital Continuous Phase

Modulated Signals with Modulation Index 1/2 Revisited,” IEEE Trans.
Commun., vol. 42, no. 2/3/4, pp. 221–224, Feb./Mar./Apr. 1994.

[243] P. Jung and P. W. Baier, “On the Representation of CPM Signals by Linear
Superposition of Impulses in the Bandpass Domain,” IEEE J. Select. Areas
Commun., vol. 10, no. 8, pp. 1236–1242, Oct. 1992.

[244] P. Kabal and S. Pasupathy, “Partial-Response Signaling,” IEEE Trans.

Commun., vol. com-23, no. 9, pp. 921–934, Sept. 1975.

[245] G. Kadel, “Diversity and Equalization in Frequency Domain—A Robust and

Flexible Receiver Technology for Broadband Mobile Communication Systems,” in
Proc. IEEE VTC, vol. 2, Phoenix, May 1997, pp. 894–898.

[246] R. E. Kahn and J. B. Thomas, “Bandwidth Properties and Optimum

Demodulation of Single-Sideband FM,” IEEE Trans. Commun. Technol., vol. 14,
no. 2, pp. 113–117, Apr. 1966.

[247] G. K. Kaleh, “Simple Coherent Receiver for Partial Response Continuous Phase
Modulation,” IEEE J. Select. Areas Commun., vol. 7, no. 9, pp. 1427–1436, Dec.
1989.

[248] I. Kalet, “The Multitone Channel,” IEEE Trans. Commun., vol. 37, no. 2, pp.
119–124, Feb. 1989.

[249] I. Kalet, S. Shitz, Z. Haddad, A. Trachtman, and Y. Baruch, “Examples of

Continuous Phase Modulation Using Discriminator Detection,” IEEE Trans.
Commun., vol. 34, no. 11, pp. 1148–1150, Nov. 1986.

[250] H. W. Kang, Y. S. Cho, and D. H. Youn, “On Compensating Nonlinear

Distortions of an OFDM System Using an Efficient Adaptive Predistorter,” IEEE
Trans. Commun., vol. 47, no. 4, pp. 522–526, Apr. 1999.

[251] M.-S. Kang and W.-J. Song, “A Robust Channel Equalizer for OFDM TV
Receivers,” IEEE Trans. Consumer Electron., vol. 44, no. 3, pp. 1129–1133, Aug.
1998.

[252] S. G. Kang and E. K. Joo, “OFDM system with linear-phase transmultiplexer,”

IEE Electr. Lett., vol. 34, no. 13, pp. 1292–1293, June 1998.

[253] G. Karam and H. Sari, “Analysis of Predistortion, Equalization, and ISI

Cancellation Techniques in Digital Radio Systems with Nonlinear Transmit
Amplifiers,” IEEE Trans. Commun., vol. 37, no. 12, pp. 1245–1253, Dec. 1989.

[254] ——, “A Data Predistortion Technique with Memory for QAM Radio Systems,”
IEEE Trans. Commun., vol. 39, no. 2, pp. 336–344, Feb. 1991.
 167

[255] A. R. Kaye, D. A. George, and M. J. Eric, “Analysis and Compensation of

Bandpass Nonlinearities for Communications,” IEEE Trans. Commun. Technol.,
vol. 20, no. 5, pp. 965–972, Oct. 1972.

[256] T. Keller and L. Hanzo, “Adaptive Multicarrier Modulation: A Convenient

Framework for Time-Frequency Processing in Wireless Communications,” Proc.
IEEE, vol. 88, no. 5, pp. 611–640, May 2000.

[257] T. Keller, L. Piazzo, P. Mandarini, and L. Hanzo, “Orthogonal Frequency

Division Multiplex Synchronization Techniques for Frequency-Selective Fading
Channels,” IEEE J. Select. Areas Commun., vol. 19, no. 6, pp. 999–1008, July
2001.

[258] D. Kim and G. L. Stüber, “Residual ISI Cancellation for OFDM with
Applications to HDTV Broadcasting,” IEEE J. Select. Areas Commun., vol. 16,
no. 8, pp. 1590–1599, Oct. 1998.

[259] ——, “Clipping Noise Mitigation for OFDM by Decision-Aided Reconstruction,”

IEEE Commun. Lett., vol. 3, no. 1, pp. 4–6, Jan. 1999.

[260] J. Kim, J. Kang, and E. J. Powers, “A Bandwidth Efficient OFDM Transmission

Scheme,” in Proc. IEEE ICASSP, vol. 3, May 2002, pp. iii–2329–iii–2332.

[261] Y. H. Kim, I. Song, H. G. Kim, T. Chang, and H. M. Kim, “Performance

Analysis of a Coded OFDM System in Time-Varying Multipath Rayleigh Fading
Channel,” IEEE Trans. Veh. Technol., vol. 48, no. 5, pp. 1610–1615, Sept. 1999.

[262] J. E. Kleider, G. Maalouli, S. Gifford, and S. Chuprun, “Preamble and

Embedded Synchronization for RF Carrier Frequency-Hopped OFDM,” IEEE J.
Select. Areas Commun., vol. 23, no. 5, pp. 920–931, May 2005.

[263] H. Kobayashi, T. Fukuhara, H. Yuan, and Y. Takeuchi, “Proposal of Single

Carrier OFDM Technique with Adaptive Modulation Method,” in Proc. IEEE
VTC, vol. 3, Apr. 2003, pp. 1915–1919.

[264] I. Koffman and V. Roman, “Broadband Wireless Access Solutions Based on

OFDM Access in IEEE 802.16,” IEEE Commun. Mag., pp. 96–103, Apr. 2002.

[265] S. Kondo and L. B. Milstein, “Performance of Multicarrier DS CDMA Systems,”

IEEE Trans. Commun., vol. 44, no. 2, pp. 238–246, Feb. 1996.

[266] H. Kopka and P. W. Daly, A Guide to LATEX. London: Addison-Wesley, 1999.

[267] W. Kozek and A. F. Molisch, “Nonorthogonal Pulseshapes for Multicarrier

Communications in Doubly Dispersive Channels,” IEEE J. Select. Areas
Commun., vol. 16, no. 8, pp. 1579–1589, Oct. 1998.

[268] B. S. Krongold and D. L. Jones, “An Active-Set Approach for OFDM PAR
Reduction via Tone Reservation,” IEEE Trans. Signal Processing, vol. 52, no. 2,
pp. 495–509, Feb. 2004.
 168

[269] ——, “PAR Reduction in OFDM via Active Constellation Extension,” IEEE
Trans. Broadcast., vol. 49, no. 3, pp. 258–268, Sept. 2003.

[270] M. Kuipers and R. Prasad, “Pre-Distorted Amplifiers for OFDM In Wireless

Indoor Multimedia Communications,” in Proc. IEEE VTC, vol. 1, Tokyo, May
2000, pp. 395–399.

[271] H. M. Kwon and K. B. E. Lee, “A Novel Digital FM Receiver for Mobile and
Personal Communications,” IEEE Trans. Commun., vol. 44, no. 11, pp.
1466–1476, Nov. 1996.

[272] L. Lamb and A. Robbins, Learning the vi Editor, 6th ed. Beijing: O’Reilly, 1998.

[273] U. Lambrette, M. Speth, and H. Meyr, “OFDM Burst Frequency Synchronization

by Single Carrier Training Data,” IEEE Commun. Lett., vol. 1, no. 2, pp. 46–48,
Mar. 1997.

[274] M. Lampe and H. Rohling, “Reducing Out-of-Band Emissions Due to

Nonlinearities in OFDM Systems,” in Proc. IEEE VTC, vol. 3, Houston, May
1999, pp. 2255–2259.

[275] R. Landauer, “The physical nature of information,” Physics Letters A, vol. 217,
pp. 188–193, July 1996.

[276] N. Lashkarian, H. Tarn, and C. Dick, “Crest Factor Reduction in Multi-carrier

WCDMA Transmitters,” in Proc. IEEE PIMRC, Berlin, Sept. 2005, in press.

[277] J. Lassing, E. G. Ström, E. Agrell, and T. Ottoson, “Computation of the Exact

Bit-Error Rate of Coherent M -ary PSK With Gray Code Bit Mapping,” IEEE
Trans. Commun., vol. 51, no. 11, pp. 1758–1760, Nov. 2003.

[278] V. K. N. Lau, “Peak-to-average ratio (PAR) reduction by Walsh-code selection

for IS-95 and CDMA2000 systems,” IEE Proc.-Commun, vol. 147, no. 6, pp.
361–364, Dec. 2000.

[279] P. A. Laurent, “Exact and Approximate Construction of Digital Phase

Modulations by Superposition of Amplitude Modulated Pulses (AMP),” IEEE
Trans. Commun., vol. 34, no. 2, pp. 150–160, Feb. 1986.

[280] G. Lazzarin, S. Pupolin, and A. Sarti, “Nonlinearity Compensation in Digital

Radio Systems,” IEEE Trans. Commun., vol. 42, no. 2/3/4, pp. 988–999,
Feb./Mar./Apr. 1994.

[281] M. T. Le and L. Thibault, “Performance Evaluation of CEOFDM for Digital

Audio Broadcasting Part II: Effects of HPA Nonlinearities,” IEEE Trans.
Broadcast., vol. 44, no. 2, pp. 165–171, June 1998.

[282] J. C. Lee and C. K. Un, “Performance Analysis of Digital Tanlock Loop,” IEEE
Trans. Commun., vol. com-30, no. 10, pp. 2398–2411, Oct. 1982.
 169

[283] J. S. Lee and D. G. Woodring, “Considerations of Nonlinear Effects in

Phase-Modulation Systems,” IEEE Trans. Commun., vol. com-20, no. 6, pp.
1063–1073, Dec. 1972.

[284] R. M. Lerner, “Communication Theory Vs Communications,” IEEE Trans.

Inform. Theory, vol. it-8, no. 4, pp. 323–325, July 1962.

[285] S. H. Leung, S. M. Ju, and G. G. Bi, “Algorithm for repeated clipping and
filtering in peak-to-average power reduction for OFDM,” IEE Electr. Lett.,
vol. 38, no. 25, pp. 1726–1727, Dec. 2002.

[286] J. Li, G. Liu, and G. B. Giannakis, “Carrier Frequency Offset Estimation for
OFDM-Based WLANs,” IEEE Signal Processing Lett., vol. 8, no. 3, pp. 80–82,
Mar. 2001.

[287] J. Li and M. Kavehrad, “Effects of Time Selective Multipath Fading on OFDM

Systems for Broadband Mobile Applications,” IEEE Commun. Lett., vol. 3,
no. 12, pp. 332–334, Dec. 1999.

[288] R. Li and G. Stette, “Time-Limited Orthogonal Multicarrier Modulation

Schemes,” IEEE Trans. Commun., vol. 43, no. 2/3/4, pp. 1269–1272,
Feb./Mar./Apr. 1995.

[289] X. Li and L. J. Cimini, Jr., “Effects of Clipping and Filtering on the Performance
of OFDM,” in Proc. IEEE VTC, vol. 3, May 1997, pp. 1634–1638.

[290] ——, “Effects of Clipping and Filtering on the Performance of OFDM,” IEEE
Commun. Lett., vol. 2, no. 5, pp. 131–133, May 1998.

[291] X. Li and J. A. Ritcey, “M-sequences for OFDM peak-to-average power ratio

reduction and error correction,” IEE Electr. Lett., vol. 33, no. 7, pp. 554–555,
Mar. 1997.

[292] Y. Li, S. McLaughlin, and D. G. M. Cruickshank, “Bandwidth efficient single

carrier systems with frequency domain equalisation,” IEE Electr. Lett., vol. 41,
no. 15, pp. 857–858, July 2005.

[293] Y. G. Li, “Pilot-Symbol-Aided Channel Estimation for OFDM in Wireless

Systems,” IEEE Trans. Veh. Technol., vol. 49, no. 4, pp. 1207–1215, July 2000.

[294] ——, “Simplified Channel Estimation for OFDM Systems with Multiple Transmit
Antennas,” IEEE Trans. Wireless Commun., vol. 1, no. 1, pp. 67–75, 2002.

[295] Y. G. Li, J. C. Chuang, and N. R. Sollenberger, “Transmitter Diversity for

OFDM Systems and Its Impact on High-Rate Data Wireless Networks,” IEEE J.
Select. Areas Commun., vol. 17, no. 7, pp. 1233–1243, July 1999.

[296] Y. G. Li and L. J. Cimini, Jr., “Bounds on the Interchannel Interference of

OFDM in Time-Varying Impairments,” IEEE Trans. Commun., vol. 49, no. 3,
pp. 401–404, Mar. 2001.
 170

[297] Y. G. Li, L. J. Cimini, Jr., and N. R. Sollenberger, “Robust Channel Estimation

for OFDM Systems with Rapid Dispersive Fading Channels,” IEEE Trans.
Commun., vol. 46, no. 7, pp. 902–915, July 1998.

[298] Y. G. Li, N. Seshadri, and S. Ariyavisitakul, “Channel Estimation for OFDM

Systems with Transmitter Diversity in Mobile Wireless Channels,” IEEE J.
Select. Areas Commun., vol. 17, no. 3, pp. 461–471, Mar. 1999.

[299] Y. G. Li and N. R. Sollenberger, “Adaptive Antenna Arrays for OFDM Systems

with Cochannel Interference,” IEEE Trans. Commun., vol. 47, no. 2, pp.
217–229, Feb. 1999.

[300] C.-P. Liang, J.-H. Jong, W. E. Shark, and J. R. East, “Nonlinear Amplifier
Effects in Communications Systems,” IEEE Trans. Microwave Theory Tech.,
vol. 47, no. 8, pp. 1461–1466, Aug. 1999.

[301] D. W. Lin, “Wide-Band Digital Subscriber Access with Multidimensional Block

Modulation and Decision-Feedback Equalization,” IEEE J. Select. Areas
Commun., vol. 7, no. 6, pp. 996–1005, Aug. 1989.

[302] R. Lin and A. P. Petropulu, “Linear Precoding Assisted Blind Channel

Estimation for OFDM Systems,” IEEE Trans. Veh. Technol., vol. 54, no. 3, pp.
983–995, May 2005.

[303] Y.-P. Lin and S.-M. Phoong, “Perfect Discrete Multitone Modulation with
Optimal Transceivers,” IEEE Trans. Signal Processing, vol. 48, no. 6, pp.
1702–1711, June 2000.

[304] ——, “MMSE OFDM and Prefixed Single Carrier Systems: BER Analysis,” in
Proc. IEEE ICASSP, vol. 4, Apr. 2003, pp. iv–229–iv–232.

[305] W. C. Lindsey and M. K. Simon, “L-Orthogonal Signal Transmission and

Detection,” IEEE Trans. Commun., vol. 20, no. 5, pp. 953–960, Oct. 1972.

[306] L. Litwin, “An introduction to multicarrier modulation,” IEEE Potentials, pp.

36–38, Apr./May 2000.

[307] C.-L. Liu, “The Effect of Nonlinearity on a QPSK-OFDM-QAM Signal,” IEEE

Trans. Consumer Electron., vol. 43, no. 3, pp. 443–447, Aug. 1997.

[308] H. Liu and U. Tureli, “A High-Efficiency Carrier Estimator for OFDM

Communications,” IEEE Commun. Lett., vol. 2, no. 4, pp. 104–106, Apr. 1998.

[309] J. Liu, D. M. Nicol, L. G. Perrone, and M. Lilgenstarn, “Towards High

Performance Modeling of the 802.11 Wireless Protocol,” in Proc. IEEE Winter
Simulation Conf., vol. 2, Arlington, VA, Dec. 2001, pp. 1315–1320.

[310] W. Liu, J. Lau, and R. S. Cheng, “Considerations on Applying OFDM in a

Highly Efficient Power Amplifier,” IEEE Trans. Circuits Syst. II, vol. 46, no. 11,
pp. 1329–1336, 1999.
 171

[311] Z. Liu, “Maximum Diversity in Single-Carrier Frequency-Domain Equalization,”

IEEE Trans. Inform. Theory, vol. 51, no. 8, pp. 2937–2940, Aug. 2005.

[312] J. Louveaux, L. Vandendorpe, and T. Sartenaer, “Cyclic Prefixed Single Carrier

and Multicarrier Transmission: Bit Rate Comparison,” IEEE Commun. Lett.,
vol. 7, no. 4, pp. 180–182, Apr. 2003.

[313] B. Lu, X. Wang, and Y. G. Li, “Iterative Receivers for Space–Time Block-Coded
OFDM Systems in Dispersive Fading Channels,” IEEE Trans. Wireless
Commun., vol. 1, no. 2, pp. 213–225, Apr. 2002.

[314] J. Lu, T. Tjhung, F. Adachi, and C. L. Huang, “BER Performance of

OFDM-MDPSK System in Frequency-Selective Rician Fading with Diversity
Reception,” IEEE Trans. Veh. Technol., vol. 49, no. 4, pp. 1216–1225, July 2000.

[315] J. E. Ludman, “Gray Code Generation for MPSK Signals,” IEEE Trans.
Commun., vol. com-29, no. 19, pp. 1519–1522, Oct. 1981.

[316] M. Luise and R. Reggiannini, “Carrier Frequency Acquisition and Tracking for
OFDM Systems,” IEEE Trans. Commun., vol. 44, no. 11, pp. 1590–1598, Nov.
1996.

[317] X. Ma, H. Kobayashi, and S. C. Schwartz, “Effect of Frequency Offset on BER of

OFDM and Single Carrier Systems,” in Proc. IEEE PIMRC, vol. 3, Sept. 2003,
pp. 2239–2243.

[318] F. R. Magee and J. G. Proakis, “Maximum-Likelihood Sequence Estimation for

Digital Signaling in the Presence of Intersymbol Interference,” IEEE Trans.
Inform. Theory, pp. 120–124, Jan. 1973.

[319] D. Makrakis and K. Feher, “Optimal Noncoherent Detection of PSK Signals,”

IEE Electr. Lett., vol. 26, no. 6, pp. 398–400, Mar. 1990.

[320] G. D. Mandyam, “Sinusoidal Transforms in OFDM Systems,” IEEE Trans.

Broadcast., vol. 50, no. 2, pp. 172–184, June 2004.

[321] D. Marelli and M. Fu, “Subband Methods for OFDM Equalization,” in Proc.
IEEE ICC, Anchorage, May 2003, pp. 2350–2354.

[322] G. Marsaglia, “Seeds for Random Number Generators,” Commun. ACM, vol. 46,
no. 5, pp. 90–93, 2003.

[323] G. Marsaglia and W. W. Tsang, “The Ziggurat Method for Generating Random
Variables,” Journal of Statistic Software, vol. 5, no. 8, pp. 1–7, 2000.

[324] G. Marsaglia, A. Zaman, and W. W. Tsang, “Toward a Universal Random

Number Generator,” Statistics & and Probability Letters, vol. 9, no. 1, pp. 35–39,
1990.
 172

[325] M. M. Martone, “A Multicarrier System Based on the Fractional Fourier

Transform for Time–Frequency-Selective Channels,” IEEE Trans. Commun.,
vol. 49, no. 6, pp. 1011–1020, June 2001.

[326] T. Maseng, “Digitally Phase Modulated (DPM) Signal,” IEEE Trans. Commun.,
vol. 33, no. 9, pp. 911–918, Sept. 1985.
[327] E. Masry, “Distortionless Demodulation of Narrow-Band Single-Sideband Angle
Modulated Signals,” IEEE Trans. Inform. Theory, vol. 23, no. 5, pp. 582–591,
Sept. 1977.
[328] Mathematica. [Online]. Available: http://www.wolfram.com/

[329] A. Mattsson, G. Mendenhall, and T. Dittmer, “Comments on: “Reduction of

peak-to-average power ratio of OFDM system using a companding technique”,”
IEEE Trans. Broadcast., vol. 45, no. 4, pp. 418–419, Dec. 1999.

[330] T. May and H. Rohling, “Reducing the Peak-to-Average Power Ratio in OFDM
Radio Transmission Systems,” in Proc. IEEE VTC, vol. 3, Ottawa, May 1998,
pp. 2774–2478.
[331] T. May, H. Rohling, and V. Engels, “Performance Analysis of Viterbi Decoding
for 64-DAPSK and 64-QAM Modulated OFDM Signals,” IEEE Trans.
Commun., vol. 46, no. 2, pp. 182–190, Feb. 1998.
[332] J. E. Mazo and J. Salz, “Spectral Properties of Single-Sideband Angle
Modulation,” IEEE Trans. Commun. Technol., vol. 16, no. 1, pp. 52–62, Feb.
1968.

[333] J. T. E. McDonnell and T. A. Wilkinson, “Comparison of Computational

Complexity of Adaptive Equalization and OFDM for Indoor Wireless Networks,”
in Proc. IEEE PIMRC, vol. 3, Oct. 1996, pp. 1088–1091.
[334] P. J. W. Melsa, R. C. Younce, and C. E. Rohrs, “Impulse Response Shortening
for Discrete Multitone Transceivers,” IEEE Trans. Commun., vol. 44, no. 12, pp.
1662–1672, Dec. 1996.
[335] U. Mengali and M. Morelli, “Decomposition of M -ary CPM Signals into PAM
Waveforms,” IEEE Trans. Inform. Theory, vol. 41, no. 5, pp. 1265–1275, Sept.
1995.

[336] V. Menon and K. Pingali, “A Case for Source-Level Transformations in

MATLAB,” in Proceedings of the 2nd Conference on Domain-Specific Languages,
Austin, 1999, pp. 53–65.

[337] S. Merchán, A. G. Armada, and J. L. Garcı́a, “OFDM Performance in Amplifier

Nonlinearity,” IEEE Trans. Broadcast., vol. 44, no. 1, pp. 106–114, Mar. 1998.
[338] D. J. G. Mestdagh, P. M. P. Spruyt, and B. Biran, “Effect of Amplitude Clipping
in DMT-ADSL Transceivers,” IEE Electr. Lett., vol. 29, no. 15, pp. 1354–1355,
July 1993.
 173

[339] D. J. G. Mestdagh and P. M. P. Spruyt, “A Method to Reduce the Probability of

Clipping in DMT-Based Transceivers,” IEEE Trans. Commun., vol. 44, no. 10,
pp. 1234–1238, Oct. 1996.

[340] S. T. Meyers, “Nonlinearity in Frequency-Modulation Radio Systems Due to

Multipath Propagation,” Proceedings of the I.R.E. and Waves and Electrons, pp.
256–265, May 1946.

[341] D. Middleton, An Introduction to Statistical Communication Theory.

McGraw-Hill Book Company, 1960.

[342] V. Mignone and A. Morello, “CD3-OFDM: A Novel Demodulation Scheme for

Fixed and Mobile Receivers,” IEEE Trans. Commun., vol. 44, no. 9, pp.
1144–1151, Sept. 1996.

[343] S. L. Miller and R. J. O’Dea, “Peak Power and Bandwidth Efficient Linear
Modulation,” IEEE Trans. Commun., vol. 46, no. 12, pp. 1639–1648, Dec. 1998.

[344] L. B. Milstein and M. C. Austin, “Performance of Noncoherent FSK and

AM-FSK Systems with Postdetection Filtering,” IEEE Trans. Commun., pp.
1300–1306, Nov. 1975.

[345] J. Minkoff, “The Role of AM-to-PM Conversion in Memoryless Nonlinear

Systems,” IEEE Trans. Commun., vol. 33, no. 2, pp. 139–144, Feb. 1985.

[346] H. Minn, C. Tellambura, and V. K. Bhargava, “On the Peak Factors of Sampled
and Continuous Signals,” IEEE Commun. Lett., vol. 5, no. 4, pp. 129–131, Apr.
2001.

[347] H. Minn, M. Zeng, and V. K. Bhargava, “On Timing Offset Estimation for
OFDM Systems,” IEEE Commun. Lett., vol. 4, no. 7, pp. 242–244, July 2000.

[348] F. Mittelbach and M. Goossens, The LATEXCompanion, 2nd ed. Boston:

Addison-Wesley, 2004.

[349] P. H. Moose, “A Technique for Orthogonal Frequency Division Multiplexing

Frequency Offset Correction,” IEEE Trans. Commun., vol. 42, no. 10, pp.
2908–2914, Oct. 1994.

[350] P. H. Moose, D. Roderick, R. North, and M. Geile, “A COFDM-based Radio for

HDR LOS Networked Communications,” in Proc. IEEE ICC, vol. 1, June 1999,
pp. 187–192.

[351] M. Morelli, A. N. D’Andrea, and U. Mengali, “Frequency Ambiguity Resolution

in OFDM Systems,” IEEE Commun. Lett., vol. 4, no. 4, pp. 134–136, Apr. 2000.

[352] M. Morelli and U. Mengali, “An Improved Frequency Offset Estimator for
OFDM Applications,” IEEE Commun. Lett., vol. 3, no. 3, pp. 75–77, Mar. 1999.
 174

[353] R. Morrison, L. J. Cimini, Jr., and S. K. Wilson, “On the Use of a Cyclic
Extension in OFDM,” in Proc. IEEE VTC, vol. 2, Atlantic City, Oct. 2001, pp.
664–668.

[354] R. R. Mosier and R. G. Clabaugh, “Kineplex, a Bandwidth-Efficient Binary

Transmission System,” AIEE Trans., vol. 76, pp. 723–728, Jan. 1958.

[355] S. H. Müller and J. B. Huber, “OFDM with reduced peak-to-average power ratio
by optimum combination of partial transmit sequences,” IEE Electr. Lett.,
vol. 33, no. 5, pp. 368–369, Feb. 1997.

[356] S. H. Müller-Weinfurtner, “Optimum Nyquist Windowing in OFDM Receivers,”

IEEE Trans. Commun., vol. 49, no. 3, pp. 417–420, Mar. 2001.

[357] B. Muquet, M. de Courville, P. Duhamel, G. B. Giannakis, and P. Magniez,

“Turbo Demodulation of Zero-Padded OFDM Transmissions,” IEEE Trans.
Commun., vol. 50, no. 11, pp. 1725–1728, Nov. 2002.

[358] B. Muquet, Z. Wang, G. B. Giannakis, M. de Courville, and P. Duhamel, “Cyclic

Prefixing or Zero Padding for Wireless Multicarrier Transmissions?” IEEE
Trans. Commun., vol. 50, no. 12, pp. 2136–2148, Dec. 2002.

[359] K. Murota and K. Hirade, “GMSK Modulation for Digital Mobile Radio
Telephony,” IEEE Trans. Commun., vol. com-29, no. 7, pp. 1044–1050, July
1981.

[360] C. Muschallik, “Improving an OFDM Reception Using an Adaptive Nyquist

Windowing,” IEEE Trans. Consumer Electron., vol. 42, no. 3, pp. 259–269, Aug.
1996.

[361] ——, “Influence of RF Oscillators on an OFDM Signal,” IEEE Trans. Consumer

Electron., vol. 41, no. 3, pp. 592–603, Aug. 1995.

[362] ——, “Improving an OFDM Reception using an Adaptive Nyquist Windowing,”

in Proceedings of the International Conference on Consumer Electronics, June
1996, pp. 6–7.

[363] Y. Nakamura, Y. Saito, and S. Aikawa, “256 QAM Modem for Multicarrier 400
Mbit/s Digital Radio,” IEEE J. Select. Areas Commun., vol. sac-5, no. 3, pp.
329–335, Apr. 1987.

[364] A. Napolitano and C. M. Spooner, “Cyclic Spectral Analysis of

Continuous-Phase Modulated Signals,” IEEE Trans. Signal Processing, vol. 49,
no. 1, pp. 30–44, Jan. 2001.

[365] S. Narahashi and T. Nojima, “New phasing scheme of N -multiple carriers for
reducing peak-to-average power ratio,” IEE Electr. Lett., vol. 30, no. 17, pp.
1382–1383, Aug. 1994.
 175

[366] R. Negi and J. Cioffi, “Pilot Tone Selection for Channel Estimation in a Mobile
OFDM System,” IEEE Trans. Consumer Electron., vol. 44, no. 3, pp. 1122–1128,
Aug. 1998.
[367] S. Neugebauer, G. Ford, and M. Ready, “Characterization of Multipath-Distorted
and Frequency-Discriminated CPM Signals,” in Conference Record of the
Asilomar Conference on Signals, Systems and Computers, vol. 2, Pacific Grove,
CA, Oct. 2000, pp. 819–822.
[368] W. T. Ng and V. K. Dubey, “Analysis of PCC-OFDM Systems for General
Time-Varying Channel,” IEEE Commun. Lett., vol. 9, no. 5, pp. 394–396, May
2005.
[369] Y. H. Ng, P. M. Grant, and R. A. Stirling-Gallacher, “Carrier tracking technique
for OFDM signal transmission,” IEE Electr. Lett., vol. 32, no. 22, pp. 2047–2048,
1996.
[370] H. Nikookar and K. S. Lidsheim, “Random Phase Updating Algorithm for
OFDM Transmission With Low PAPR,” IEEE Trans. Broadcast., vol. 48, no. 2,
pp. 123–128, June 2002.
[371] H. Nikopour and S. H. Jamali, “On the Performance of OFDM Systems Over a
Cartesian Clipping Channel: A Theoretical Approach,” IEEE Trans. Wireless
Commun., vol. 3, no. 6, pp. 2083–2096, Nov. 2004.
[372] A. J. Noga and T. K. Sarkar, “A Discrete-Time Method of Demodulating Large
Deviation FM Signals,” IEEE Trans. Commun., vol. 47, no. 8, pp. 1194–1200,
Aug. 1999.
[373] A. H. Nuttall, “Error Probabilities for Equicorrelated M -ary Signals Under
Phase-Coherent and Phase-Incoherent Reception,” IRE Trans. Inform. Theory,
vol. 8, pp. 305–314, July 1962.
[374] H. Ochiai, “Power Efficiency Comparison of OFDM and Single-Carrier Signals,”
in Proc. IEEE VTC, vol. 2, Sept. 2002, pp. 899–903.
[375] ——, “Performance Analysis of Peak Power and Band-Limited OFDM System
With Linear Scaling,” IEEE Trans. Wireless Commun., vol. 2, no. 5, pp.
1055–1065, Sept. 2003.
[376] ——, “Performance of Optimal and Suboptimal Detection for Uncoded OFDM
System with Deliberate Clipping and Filtering,” in Proc. IEEE Globecom, vol. 3,
Dec. 2003, pp. 1618–1622.
[377] ——, “A Novel Trellis-Shaping Design With Both Peak and Average Power
Reduction for OFDM Systems,” IEEE Trans. Commun., vol. 52, no. 11, pp.
1916–1926, Nov. 2004.
[378] H. Ochiai, M. P. C. Fossorier, and H. Imai, “On Decoding of Block Codes with
Peak-Power Reduction in OFDM Systems,” IEEE Commun. Lett., vol. 4, no. 7,
pp. 226–228, July 2000.
 176

[379] H. Ochiai and H. Imai, “Channel Capacity of Clipped OFDM Systems,” in Proc.
IEEE Intern. Sympo. Inform. Theory, Sorrento, June 2000, p. 219.

[380] ——, “Performance of the Deliberate Clipping with Adaptive Symbol Selection
for Strictly Band-Limited OFDM Systems,” IEEE J. Select. Areas Commun.,
vol. 18, no. 11, pp. 2270–2277, Nov. 2000.

[381] ——, “On the Distribution of the Peak-to-Average Power Ratio in OFDM
Signals,” IEEE Trans. Commun., vol. 49, no. 2, pp. 282–289, Feb. 2001.

[382] ——, “Performance Analysis of Deliberately Clipped OFDM Signals,” IEEE

Trans. Commun., vol. 50, no. 1, pp. 89–101, Jan. 2002.

[383] M. S. O’Droma and N. Mgebrishvili, “Signal Modeling Classes for Linearized

OFDM SSPA Behavioral Analysis,” IEEE Commun. Lett., vol. 9, no. 2, pp.
127–129, Feb. 2005.

[384] R. O’Neill and L. B. Lopes, “Performance of Amplitude Limited Multitone

Signals,” in Proc. IEEE VTC, June 1994, pp. 1675–1679.

[385] ——, “Envelope Variations and Spectral Splatter in Clipped Multicarrier

Signals,” in Proc. IEEE PIMRC, vol. 1, 1995, pp. 71–75.

[386] A. V. Oppenheim, R. W. Schafer, and J. R. Buck, Discrete-Time Signal

Processing, 3rd ed. New Jersey: Prentice Hall, 1998.

[387] A. V. Oppenheim and A. S. Willsky, Signals and Systems. Upper Saddle River,
New Jersey: Prentice Hall, 1997.

[388] F. Pancaldi and G. M. Vitetta, “Block Equalization Techniques in the Frequency

Domain,” in Proc. IEEE WCNC, vol. 4, Mar. 2004, pp. 2295–2300.

[389] R. Pandey, H. Leib, and S. Pasupathy, “Orthogonal Phase Functions for Full
Response Non-Coherent CPM Systems,” in Proceedings of the IEEE Pacific Rim
Conference on Communications, Computers and Signal Processing, Victoria, BC,
Canada, June 1989, pp. 320–323.

[390] ——, “Noncoherently Orthogonal Digital Phase Modulation,” IEEE Trans.

Commun., vol. 40, no. 6, pp. 992–995, June 1992.

[391] K. R. Panta and J. Armstrong, “Effects of Clipping on the Error Performance of

OFDM in Frequency Selective Fading Channels,” IEEE Trans. Wireless
Commun., vol. 3, no. 2, pp. 668–671, Mar. 2004.

[392] H. C. Papadopoulos and C.-E. W. Sundberg, “Simultaneous Broadcasting of

Analog FM and Digital Audio Signals by Means of Adaptive Precanceling
Techniques,” IEEE Trans. Commun., vol. 46, no. 9, pp. 1233–1242, Sept. 1998.

[393] P. Papantoni-Kazakos and I. M. Paz, “The Performance of a Digital FM System

with Discriminator: Intersymbol Interference Effects,” IEEE Trans. Commun.,
vol. com-23, no. 9, pp. 867–877, Sept. 1975.
 177

[394] A. Papoulis, Probability, Random Variables, and Stochastic Processes, 3rd ed.
Boston, MA: McGraw-Hill, 1991.

[395] I.-S. Park and E. J. Powers, “Compensation of Nonlinear Distortion in OFDM

Systems Using a New Predistorter,” in Proc. IEEE PIMRC, vol. 2, Boston, Sept.
1998, pp. 811–815.

[396] J. Park, E. Joe, M.-J. Choe, and B.-S. Song, “A 5-MHz IF Digital FM
Demodulator,” IEEE J. Solid-State Circuits, vol. 34, no. 1, pp. 3–11, Jan. 1999.

[397] J. S. Park, S. R. Park, H. J. Roh, and K. H. Koo, “Power Amplifier Back-off

Analysis with AM-to-PM for Millimeter-wave OFDM Wireless LAN,” in Proc.
IEEE RAWCOM, Waltham, Aug. 2001, pp. 189–192.

[398] S. Pasupathy, “Minimum Shift Keying: A Spectrally Efficient Modulation,”

IEEE Commun. Mag., pp. 14–22, July 1979.

[399] K. G. Paterson, “Generalized Reed-Muller Codes and Power Control in OFDM

Modulation,” IEEE Trans. Inform. Theory, vol. 46, no. 1, pp. 104–120, Jan. 2000.

[400] K. G. Paterson and V. Tarokh, “On the Existence and Construction of Good
Codes with Low Peak-to-Average Power Ratios,” IEEE Trans. Inform. Theory,
vol. 46, no. 6, pp. 1974–1987, Sept. 2000.

[401] M. Pätzold, Mobile Fading Channels. West Sussex, England: John Wiley &
Sons, 2002.

[402] M. Pätzold, U. Killat, F. Laue, and Y. Li, “On the Statistical Properties of
Deterministic Simulation Models for Mobile Fading Channels,” IEEE Trans. Veh.
Technol., vol. 47, no. 1, pp. 254–269, Feb. 1998.

[403] M. Pauli and H.-P. Kuchenbecker, “Minimization of the Intermodulation

Distortion of a Nonlinearly Amplified OFDM Signal,” Wireless Personal
Communications, vol. 4, no. 1, pp. 93–101, Jan. 1996.

[404] D. R. Pauluzzi and N. C. Beaulieu, “A Comparison for SNR Estimation

Techniques for the AWGN Channel,” IEEE Trans. Commun., vol. 48, no. 10, pp.
1681–1691, Oct. 2000.

[405] R. F. Pawula, “The Effects of Quadratic AM-PM Conversion in

Frequency-Division Multiplexed Multiple-Access Communication Satellite
Systems,” IEEE Trans. Commun. Technol., vol. 19, no. 3, pp. 345–349, June
1971.

[406] ——, “On the Theory of Error Rates for Narrow-Band Digital FM,” IEEE Trans.
Commun., vol. com-29, no. 11, pp. 1634–1643, Nov. 1981.

[407] ——, “Refinements to the Theory of Error Rates for Narrow-Band Digital FM,”
IEEE Trans. Commun., vol. 36, no. 4, pp. 509–513, Apr. 1988.
 178

[408] A. Peled and A. Ruiz, “Frequency Domain Data Transmission Using Reduced
Computational Complexity Algorithms,” in Proc. IEEE ICASSP, vol. 5, Apr.
1980, pp. 964–967.

[409] N. Petersson, A. Johansson, P. Ödling, and P. O. Börjesson, “A Performance

Bound on PSD-constrained PAR Reduction,” in Proc. IEEE ICC, vol. 5, May
2003, pp. 3498–3502.

[410] R. Petrovic, W. Roehr, and D. W. Cameron, “Multicarrier Modulation for

Narrowband PCS,” IEEE Trans. Veh. Technol., vol. 43, no. 4, pp. 856–862, Nov.
1994.

[411] R. H. Pettit, B. E. Wahlen, C. Y. Mai, and R. C. North, “A

Maximum-Likelihood-Based Frequency Synchronizer for Dual-h Continuous
Phase Modulation (CPM),” in Proc. IEEE Milcom, vol. 2, Oct. 2001, pp.
1219–1223.

[412] L. Piazzo and P. Mandarini, “Analysis of Phase Noise Effects in OFDM

Modems,” IEEE Trans. Commun., vol. 50, no. 10, pp. 1696–1705, Oct. 2002.

[413] G. L. Pierobon, S. G. Pupolin, and G. P. Tronca, “Power Spectrum of Angle

Modulated Correlated Digital Signals,” IEEE Trans. Commun., vol. 30, no. 2,
pp. 389–396, Feb. 1982.

[414] T. Pollet and M. Peeters, “Synchronization with DMT Modulation,” IEEE

Commun. Mag., pp. 80–86, Apr. 1999.

[415] T. Pollet, M. Van Bladel, and M. Moeneclaey, “BER Sensitivity of OFDM

Systems to Carrier Frequency Offset and Wiener Phase Noise,” IEEE Trans.
Commun., vol. 43, no. 2/3/4, pp. 191–193, Feb./Mar./Apr. 1995.

[416] M. O. Polley, W. F. Schreiber, and S. J. Wee, “Comments on “Transmission

Techniques for Digital Terrestrial TV Broadcasting”,” IEEE Commun. Mag., pp.
22–26, Nov. 1995.

[417] M. F. Pop and N. C. Beaulieu, “Limitations of the Sum-of-Sinusoids Fading

Channel Simulators,” IEEE Trans. Commun., vol. 49, no. 4, pp. 699–708, Apr.
2001.

[418] B. M. Popović, “Synthesis of Power Efficient Multitone Signals with Flat

Amplitude Spectrum,” IEEE Trans. Commun., vol. 39, no. 7, pp. 1031–1033,
July 1991.

[419] W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, Numerical

Recipes in C, 2nd ed. Cambridge University Press, 1992.

[420] J. G. Proakis, “Probabilities of Error for Adaptive Reception of M -Phase

Signals,” IEEE Trans. Commun., vol. 16, no. 1, pp. 71–81, Feb. 1968.

[421] ——, Digital Communications, 4th ed. New York: McGraw-Hill, 2001.
 179

[422] J. G. Proakis and D. G. Manolakis, Digital Signal Processing: Principles,

Algorithms, and Applications, 3rd ed. Upper Saddle River, NJ: Prentice Hall,
1996.

[423] J. G. Proakis and M. Salehi, Communication Systems Engineering. New Jersey:

Prentice Hall, 1994.

[424] F. H. Raab, P. Asbeck, S. Cripps, P. B. Kenington, Z. B. Popovic, N. Pothecary,

J. F. Sevic, and N. O. Sokal, “Power Amplifiers and Transmitters for RF and
Microwave,” IEEE Trans. Microwave Theory Tech., vol. 50, no. 3, pp. 814–826,
Mar. 2002.

[425] M. Rabzel and S. Pasupathy, “Spectral Shaping in Minimum Shift Keying

(MSK)-Type Signals,” IEEE Trans. Commun., vol. com-26, no. 1, pp. 189–195,
Jan. 1978.

[426] C. Rapp, “Effects of HPA-Nonlinearity on a 4–DPSK/OFDM–Signal for a Digital

Sound Broadcasting System,” in Second European Conference on Satellite
Communications, Liége, Belgium, Oct. 1991, pp. 179–184. [Online]. Available:
http://www.e-technik.fh-muenchen.de/fb/prof/rapp/bib − Rapp.html

[427] T. S. Rappaport, Wireless Communications: Principles and Practice, 2nd ed.

Upper Saddle River, New Jersey: Prentice Hall, 2002.

[428] K. R. Raveendra and R. Srinivasan, “Threshold detection of continuous

phase-modulated signals,” IEE Proceedings, vol. 136, no. 6, pp. 414–423, Dec.
1989.

[429] A. J. Redfern, “Receiver Window Design for Multicarrier Communication

Systems,” IEEE J. Select. Areas Commun., vol. 20, no. 5, pp. 1029–1036, June
2002.

[430] G. Ren, H. Zhang, and Y. Chang, “A Complementary Clipping Transform

Technique for the Reduction for Peak-to-Average Power Ratio of OFDM System,”
IEEE Trans. Consumer Electron., vol. 49, no. 4, pp. 922–926, Nov. 2003.

[431] R. B. Rhodes, Jr., S. G. Wilson, and A. B. Svensson, “MSK-Type Reception of

Continuous Phase Modulation: Cochannel and Adjacent Channel Interference,”
IEEE Trans. Commun., vol. com-35, no. 2, pp. 185–193, Feb. 1987.

[432] S. O. Rice, “Distributions for Noise Plus Several Sine Waves—The Problem of
Computation,” IEEE Trans. Commun., pp. 851–853, June 1974.

[433] B. E. Rimoldi, “A Decomposition Approach to CPM,” IEEE Trans. Inform.

Theory, vol. 34, no. 2, pp. 260–270, Mar. 1988.

[434] J. Rinne and M. Renfors, “The Behavior of Orthogonal Frequency Division

Multiplexing Signals in an Amplitude Limiting Channel,” in Proc. IEEE ICC,
New Orleans, May 1994, pp. 381–385.
 180

[435] ——, “Pilot Spacing in Orthogonal Frequency Division Multiplexing Systems on

Practical Channels,” IEEE Trans. Consumer Electron., vol. 42, no. 4, pp.
959–962, Nov. 1996.

[436] A. D. Rizos, J. G. Proakis, and T. Q. Nguyen, “Comparison of DFT and Cosine

Modulated Filter Banks in Multicarrier Modulation,” in Proc. IEEE Globecom,
vol. 2, Nov. 1994, pp. 687–691.

[437] J. H. Roberts, Angle Modulation. London: Peter Peregrinus Ltd., 1977.

[438] M. R. D. Rodrigues, I. Darwazeh, and J. J. O’Reilly, “Error probability of
nonlinearly distorted OFDM signals,” IEE Electr. Lett., vol. 39, no. 22, pp.
1620–1622, Oct. 2003.

[439] M. R. D. Rodrigues and I. J. Wassell, “A Novel Coding Strategy to Improve the

Error Probability Performance of Non-linearly Distorted OFDM Signals,” in
Proc. IEEE VTC, vol. 1, Oct. 2003, pp. 294–298.

[440] H. Rohling and T. May, “Comparison of PSK and DPSK Modulation in a Coded
OFDM System,” in Proc. IEEE VTC, vol. 2, Phoenix, May 1997, pp. 870–874.
[441] C. Rößing and V. Tarokh, “A Construction of OFDM 16-QAM Sequences Having
Low Peak Powers,” IEEE Trans. Inform. Theory, vol. 47, no. 5, pp. 2091–2094,
July 2001.
[442] H. E. Rowe, Signals and Noise in Communication Systems. Princeton, N. J.: D.
Van Nostrand Company, Inc., 1965.

[443] H. E. Rowe and V. K. Prabhu, “Power Spectrum of a Digital,

Frequency-Modulation Signal,” Bell Syst. Tech. J., vol. 54, pp. 1095–1125,
July–Aug. 1975.
[444] D. N. Rowitch and L. B. Milstein, “Convolutionally Coded Multicarrier
DS-CDMA Systems in a Multipath Fading Channel—Part I: Performance
Analysis,” IEEE Trans. Commun., vol. 47, no. 10, pp. 1570–1582, Oct. 1999.
[445] ——, “Convolutionally Coded Multicarrier DS-CDMA Systems in a Multipath
Fading Channel—Part II: Narrow-Band Interference Suppression,” IEEE Trans.
Commun., vol. 47, no. 11, pp. 1729–1736, Nov. 1999.

[446] A. Ruiz, J. M. Cioffi, and S. Kasturia, “Discrete Multiple Tone Modulation with
Coset Coding for the Spectrally Shaped Channel,” IEEE Trans. Commun.,
vol. 40, no. 6, pp. 1012–1029, June 1992.

[447] M. Russell and G. L. Stüber, “Interchannel Interference Analysis of OFDM in a

Mobile Environment,” in Proc. IEEE VTC, vol. 2, Chicago, July 1995, pp.
820–824.
[448] H.-G. Ryu, T. P. Hoa, K. M. Lee, S.-W. Kim, and J.-S. Park, “Improvement of
Power Efficiency of HPA by the PAPR Reduction and Predistortion,” IEEE
Trans. Consumer Electron., vol. 50, no. 1, pp. 119–124, Feb. 2004.
 181

[449] H.-G. Ryu, B.-I. Jin, and I.-B. Kim, “PAPR Reduction Using Soft Clipping and
ACI Rejection in OFDM System,” IEEE Trans. Consumer Electron., vol. 48,
no. 1, pp. 17–22, Feb. 2002.
[450] H.-G. Ryu, J.-E. Lee, and J.-S. Park, “Dummy Sequence Insertion (DSI) for
PAPR Reduction in the OFDM communication System,” IEEE Trans. Consumer
Electron., vol. 50, no. 1, pp. 89–94, Feb. 2004.
[451] H.-G. Ryu and K.-J. Youn, “A New PAPR Reduction Scheme: SPW (Subblock
Phase Weighting),” IEEE Trans. Consumer Electron., vol. 48, no. 1, pp. 81–89,
Feb. 2002.
[452] M. Sablatash, “Transmission of All-Digital Advanced Television: State of the Art
and Future Directions,” IEEE Trans. Broadcast., vol. 40, no. 2, pp. 102–121,
June 1994.
[453] H. Saeedi, M. Sharif, and F. Marvasti, “Clipping Noise Cancellation in OFDM
Systems Using Oversampled Signal Reconstruction,” IEEE Commun. Lett.,
vol. 6, no. 2, pp. 73–75, Feb. 2002.
[454] A. A. M. Saleh, “Frequency-Independent and Frequency-Dependent Nonlinear
Models of TWT Amplifiers,” IEEE Trans. Commun., vol. 29, no. 11, pp.
1715–1720, Nov. 1981.
[455] B. R. Saltzberg, “Performance of an Efficient Parallel Data Transmission
System,” IEEE Trans. Commun. Technol., vol. com-15, no. 6, pp. 805–811, Dec.
1967.
[456] A. A. Salvekar, C. Aldana, J. Tellado, and J. Cioffi, “Peak-to-average power ratio
reduction for block transmission systems in the presence of transmit filtering,” in
Proc. IEEE ICC, Helsinki, June 2001, pp. 175–178.
[457] S. D. Sandberg and M. A. Tzannes, “Overlapped Discrete Multitone Modulation
for High Speed Copper Wire Communications,” IEEE J. Select. Areas Commun.,
vol. 13, no. 9, pp. 1571–1585, Dec. 1995.
[458] G. Santella, “A Frequency and Symbol Synchronization System for OFDM
Signals: Architecture and Simulation Results,” IEEE Trans. Veh. Technol.,
vol. 49, no. 1, pp. 254–275, Jan. 2000.
[459] G. Santella and F. Mazzenga, “A Hybrid Analytical-Simulation Procedure for
Performance Evaluation in M-QAM-OFDM Schemes in Presence of Nonlinear
Distortions,” IEEE Trans. Veh. Technol., vol. 47, no. 1, pp. 142–151, Feb. 1998.
[460] H. Sari, G. Karam, and I. Jeanclaude, “An Analysis of Orthogonal
Frequency-Division Multiplexing for Mobile Radio Applications,” in Proc. IEEE
VTC, vol. 3, Stockholm, June 1994, pp. 1635–1639.
[461] ——, “Frequency-Domain Equalization of Mobile Radio and Terrestrial
Broadcast Channels,” in Proc. IEEE Globecom, vol. 1, San Francisco, Nov. 1994,
pp. 1–5.
 182

[462] ——, “Channel equalization and carrier synchronization in OFDM systems,” in

Proc. Tirrenia Int’l Workshop on Dig. Commun., Tirrenia, Italy, Sept. 1993, pp.
191–202.

[463] ——, “Transmission Techniques for Digital Terrestrial TV Broadcasting,” IEEE

Commun. Mag., pp. 100–109, Feb. 1995.

[464] K. Sathananthan and C. Tellambura, “Forward Error Correction Codes to

Reduce Intercarrier Interference in OFDM,” in Proc. IEEE ICASSP, vol. 4,
Sydney, May 2001, pp. iv–566–iv–569.

[465] ——, “Probability of Error Calculation of OFDM Systems With Frequency

Offset,” IEEE Trans. Commun., vol. 49, no. 11, pp. 1884–1888, Nov. 2001.

[466] ——, “Coding to Reduce Both PAR and PICR of an OFDM Signal,” IEEE
Commun. Lett., vol. 6, no. 8, pp. 316–318, Aug. 2002.

[467] ——, “Partial Transmit Sequence and Selected Mapping Schemes to Reduce ICI
in OFDM Systems,” IEEE Commun. Lett., vol. 6, no. 8, pp. 313–315, Aug. 2002.

[468] T. M. Schmidl and D. C. Cox, “Blind synchronisation for OFDM,” IEE Electr.
Lett., vol. 33, no. 2, pp. 113–114, Jan. 1997.

[469] ——, “Robust Frequency and Timing Synchronization for OFDM,” IEEE Trans.
Commun., vol. 45, no. 12, pp. 1613–1621, Dec. 1997.

[470] K. W. Schneider and W. H. Tranter, “Efficient Simulation of Multicarrier Digital

Communication Systems in Nonlinear Channel Environments,” IEEE J. Select.
Areas Commun., vol. 11, no. 3, pp. 328–339, Apr. 1993.

[471] M. R. Schroeder, “Synthesis of Low-Peak-Factor Signals and Binary Sequences

With Low Autocorrelation,” IEEE Trans. Inform. Theory, pp. 85–89, Jan. 1970.

[472] M. Schwartz, W. R. Bennett, and S. Stein, Communication Systems and

Techniques. McGraw-Hill, 1966.

[473] M. Sharif, C. Florens, M. Fazel, and B. Hassibi, “Amplitude and Sign

Adjustment for Peak-to-Average-Power Reduction,” IEEE Trans. Commun.,
vol. 53, no. 8, pp. 1243–1247, Aug. 2005.

[474] M. Sharif, M. Gharavi-Alkhansari, and B. H. Khalaj, “On the Peak-to-Average

Power of OFDM Signals Based on Oversampling,” IEEE Trans. Commun.,
vol. 31, no. 1, pp. 72–78, Jan. 2003.

[475] M. Sharif and B. Hassibi, “On Multicarrier Signals Where the PMEPR of a
Random Codeword is Asymptotically log n,” IEEE Trans. Inform. Theory,
vol. 50, no. 5, pp. 895–903, May 2004.

[476] M. Shell, “How to Use the IEEEtran LATEXClass,” Journal of LATEXFiles, vol. 1,
no. 11, pp. 1–22, Nov. 2002.
 183

[477] P. Shelswell, “The COFDM modulation system: the heart of digital audio
broadcasting,” IEEE Electron. Commun. Eng. J., pp. 127–136, June 1995.

[478] S. J. Shepherd, P. W. J. V. Eevelt, C. W. Wyatt-Millington, and S. K. Barton,

“Simple coding scheme to reduce peak factor in QPSK multicarrier modulation,”
IEE Electr. Lett., vol. 31, no. 14, pp. 1131–1132, July 1995.

[479] S. Shepherd, J. Orriss, and S. Barton, “Asymptotic Limits in Peak Envelope

Power Reduction by Redundant Coding in Orthogonal Frequency-Division
Multiplex Modulation,” IEEE Trans. Commun., vol. 46, no. 1, pp. 5–10, Jan.
1998.

[480] Q. Shi, “OFDM in Bandpass Nonlinearity,” IEEE Trans. Consumer Electron.,

vol. 42, no. 3, pp. 253–258, Aug. 1996.

[481] T. Shi, S. Zhou, and T. Tao, “Capacity of Single Carrier Systems With
Frequency-Domain Equalization,” in Proc. IEEE CAS Symp., vol. 2, Shanghai,
June 2004, pp. 429–432.

[482] S. S. (Shitz) and I. Bar-David, “On the Capacity Penalty Due to

Input-Bandwidth Restrictions with an Application to Rate-Limited Binary
Signaling,” IEEE Trans. Inform. Theory, vol. 36, no. 3, pp. 623–627, May 1990.

[483] M. K. Simon and M.-S. Alouini, Digital Communication over Fading Channels.
New York: John Wiley & Sons, Inc., 2000.

[484] M. K. Simon and D. Divalar, “Maximum-Likelihood Block Detection of

Noncoherent Continuous Phase Modulation,” IEEE Trans. Commun., vol. 41,
no. 1, pp. 90–98, Jan. 1993.

[485] R. Sinha and R. D. Yates, “Performance of Multicarrier MFSK in Fading

Channels,” in Proc. IEEE VTC, vol. 3, Atlantic City, Oct. 2001, pp. 1848–1851.

[486] C. V. Sinn, J. Götze, and M. Haardi, “Efficient Data Detection Algorithms in

Signal- and Multi-Carrier Systems without the Necessity of a Guard Period,” in
Proc. IEEE ICASSP, vol. 3, May 2002, pp. iii–2737–iii–2740.

[487] B. Sklar, “Rayleigh Fading Channels in Mobile Digital Communication System,

Part I: Characterization,” IEEE Commun. Mag., pp. 90–100, July 1997.

[488] ——, “Rayleigh Fading Channels in Mobile Digital Communication System, Part
II: Mitigation,” IEEE Commun. Mag., pp. 102–109, July 1997.

[489] H.-K. Song, Y.-H. You, J.-H. Paik, and Y.-S. Cho, “Frequency-Offset
Synchronization and Channel Estimation for OFDM-Based Transmission,” IEEE
Commun. Lett., vol. 4, no. 3, pp. 95–97, Mar. 2000.

[490] R. Song and S.-H. Leung, “A Novel OFDM Receiver with Second Order
Polynomial Nyquist Window Function,” IEEE Commun. Lett., vol. 9, no. 5, pp.
391–393, May 2005.
 184

[491] E. A. Sourour and M. Nakagawa, “Performance of Orthogonal Multicarrier

CDMA in a Multipath Fading Channel,” IEEE Trans. Commun., vol. 44, no. 3,
pp. 356–367, Mar. 1996.

[492] M. Speth, S. A. Fechtel, G. Fock, and H. Meyr, “Optimum Receiver Design for
Wireless Broad-Band Systems Using OFDM–Part I,” IEEE Trans. Commun.,
vol. 47, no. 11, pp. 1668–1677, Nov. 1999.

[493] A. Stamoulis, S. N. Diggavi, and N. Al-Dhahir, “Intercarrier Interference in

MIMO OFDM,” IEEE Trans. Signal Processing, vol. 50, no. 10, pp. 2451–2464,
Oct. 2002.
[494] B. Stantchev and G. Fettweis, “Time-Variant Distortions in OFDM,” IEEE
Commun. Lett., vol. 4, no. 9, pp. 312–314, Sept. 2000.
[495] W. E. Stark, E. B. Felstead, D. L. Nonaeker, and B. M. Sadler, “Guest Editorial:
Advances in Military Wireless Communications,” IEEE J. Select. Areas
Commun., vol. 23, no. 5, pp. 881–883, May 2005.

[496] H. Steendam and M. Moeneclaey, “Analysis and Optimization of the

Performance of OFDM on Frequency-Selective Time-Selective Fading Channels,”
IEEE Trans. Commun., vol. 47, no. 12, pp. 1811–1819, Dec. 1999.
[497] ——, “Sensitivity of orthogonal frequency-division multiplexed systems to carrier
and clock synchronization errors,” Signal Processing, vol. 80, no. 7, pp.
1217–1229, July 2000.
[498] H. Steendam, M. Moeneclaey, and H. Sari, “The Effect of Carrier Phase Jitter on
the Performance of Orthogonal Frequency-Division Multiple-Access Systems,”
IEEE Trans. Commun., vol. 46, no. 4, pp. 456–459, Apr. 1998.

[499] Y. Sun, “Bandwidth-Efficient Wireless OFDM,” IEEE J. Select. Areas Commun.,

vol. 19, no. 11, pp. 2267–2278, Nov. 2001.

[500] C.-E. Sundberg, “Continuous Phase Modulation,” IEEE Commun. Mag., vol. 24,
no. 4, pp. 25–37, Apr. 1986.
[501] E. D. Sunde, Communication Systems Engineering Theory. New York: John
Wiley & Sons, Inc., 1969.

[502] T. Svensson and A. Svensson, “Constrained Envelope Continuous Phase

Modulation,” in Proc. IEEE VTC, vol. 4, Apr. 2003, pp. 2623–2627.
[503] ——, “Multi-Pulse Continuous Phase Modulation,” in Proceedings of The 5th
International Symposium on Wireless Personal Multimedia Communications,
vol. 3, Oct. 2003, pp. 981–985.
[504] G. S. Takhar and S. C. Gupta, “Discrete Estimation of Continuous
Angle-Modulated Signals Over Multipath Channels for Aeronautical
Communication,” IEEE Trans. Commun. Technol., vol. 24, no. 3, pp. 365–374,
Mar. 1976.
 185

[505] B. T. Tan and Y. K. Some, “Crest Factor Minimisation in FDM PSK Systems,”
IEE Electr. Lett., vol. 26, no. 13, pp. 859–861, 1990.

[506] J. Tan and G. L. Stüber, “Constant Envelope Multi-Carrier Modulation,” in

Proc. IEEE Milcom, vol. 1, Anaheim, Oct. 2002, pp. 607–611.

[507] B. Tarokh and H. R. Sadjadpour, “Construction of OFDM M -QAM Sequences

with Low Peak-to-Average Power Ratio,” IEEE Trans. Commun., vol. 51, no. 1,
pp. 25–28, Jan. 2003.

[508] V. Tarokh and H. Jafarkhani, “On the Computation and Reduction of the
Peak-to-Average Power Ratio in Multicarrier Communications,” IEEE Trans.
Commun., vol. 48, no. 1, pp. 37–44, Jan. 2000.

[509] I. A. Tasadduq and R. K. Rao, “OFDM-CPM Signals,” IEE Electr. Lett., vol. 38,
no. 2, pp. 80–81, Jan. 2002.

[510] ——, “PAPR Reduction of OFDM Signals using Multiamplitude CPM,” IEE
Electr. Lett., vol. 38, no. 16, pp. 915–917, Aug. 2002.

[511] ——, “Detection of OFDM-CPM Signals over Multipath Channels,” in Proc.

IEEE ICC, vol. 3, May 2002, pp. 1651–1655.

[512] J. Tellado and J. M. Cioffi, “Efficient Algorithms for Reducing PAR in

Multicarrier Systems,” in Proc. IEEE Intern. Sympo. Inform. Theory,
Cambridge, Aug. 1998, p. 191.

[513] J. Tellado, L. M. C. Hoo, and J. M. Cioffi, “Maximum-Likelihood Detection of

Nonlinearly Distorted Multicarrier Symbols by Iterative Decoding,” IEEE Trans.
Commun., vol. 51, no. 2, pp. 218–228, Feb. 2003.

[514] C. Tellambura, “Upper bound on peak factor of n-multiple carriers,” IEE Electr.
Lett., vol. 33, no. 19, pp. 1608–1609, Sept. 1997.

[515] ——, “Use of m-sequences for OFDM peak-to-average power ratio reduction,”
IEE Electr. Lett., vol. 33, no. 15, pp. 1300–1301, July 1997.

[516] ——, “Comment: Multicarrier transmission peak-to-average power reduction

using simple block code,” IEE Electr. Lett., vol. 34, no. 17, p. 1646, Aug. 1998.

[517] ——, “Phase optimisation criterion for reducing peak-to-average power ratio in
OFDM,” IEE Electr. Lett., vol. 34, no. 2, pp. 169–170, Jan. 1998.

[518] ——, “Computation of the Continuous-Time PAR of an OFDM Signal with

BPSK Subcarriers,” IEEE Commun. Lett., vol. 5, pp. 185–187, May 2001.

[519] C. Tellambura, I. R. Johnson, Y. J. Guo, and S. K. Barton, “Frequency-Offset

Estimation for HIPERLAN,” IEEE Trans. Commun., vol. 47, no. 8, pp.
1137–1139, Aug. 1999.
 186

[520] C. Tellambura and M. G. Parker, “Relationship between Hamming weight and

Peak-to-mean envelope power ratio of orthogonal frequency division
multiplexing,” in Proc. IEEE Intern. Sympo. Inform. Theory, Lausanne,
Switzerland, June 2002, p. 245.

[521] C. Tellambura, M. G. Parker, Y. J. Guo, S. J. Shepherd, and S. K. Barton,

“Optimal Sequences for Channel Estimation Using Discrete Fourier Transform
Techniques,” IEEE Trans. Commun., vol. 47, no. 2, pp. 230–238, Feb. 1999.

[522] C. Tellambura, “Improved Phase Factor Computation for the PAR Reduction of
an OFDM Signal Using PTS,” IEEE Commun. Lett., vol. 5, no. 4, pp. 135–137,
Apr. 2001.

[523] The VIM Text Editor. [Online]. Available: http://www.vim.org/

[524] L. Thibault and M. T. Le, “Performance Evaluation of COFDM for Digital

Audio Broadcasting, Part I: Parametric Study,” IEEE Trans. Broadcast., vol. 43,
no. 1, pp. 64–75, Mar. 1997.

[525] S. C. Thompson, “Baseband Equivalent Simulation of the Optimum Receiver.”

[Online]. Available: http://zeidler.ucsd.edu/∼sct/holdings/baseband − model/

[526] ——, “Spectral Estimation of Digital Signaling Using The Welch Method.”
[Online]. Available: http://zeidler.ucsd.edu/∼sct/holdings/welch/

[527] S. C. Thompson, A. U. Ahmed, J. G. Proakis, and J. R. Zeidler, “Constant

Envelope OFDM Phase Modulation: Spectral Containment, Signal Space
Properties and Performance,” in Proc. IEEE Milcom, vol. 2, Monterey, Oct. 2004,
pp. 1129–1135.

[528] S. C. Thompson, J. G. Proakis, and J. R. Zeidler, “Constant Envelope Binary

OFDM Phase Modulation,” in Proc. IEEE Milcom, vol. 1, Boston, Oct. 2003, pp.
621–626.

[529] ——, “Noncoherent Reception of Constant Envelope OFDM in Flat Fading

Channels,” in Proc. IEEE PIMRC, Berlin, Sept. 2005.

[530] ——, “The Effectiveness of Signal Clipping for PAPR and Total Degradation
Reduction in OFDM Systems,” in Proc. IEEE Globecom, St. Louis, Dec. 2005.

[531] T. T. Tjhung, “Power Spectra and Power Distributions of Random Binary F.M.
Signals with Premodulation Shaping,” IEE Electr. Lett., vol. 1, no. 6, pp.
176–178, Aug. 1965.

[532] T. T. Tjhung, K. K. Teo, and P. H. Wittke, “Effects of Pulse Shaping and Soft
Decisions on the Performance of Digital FM with Discriminator Detection,”
IEEE Trans. Commun., vol. 34, no. 11, pp. 1116–1122, Nov. 1986.

[533] S. Tomasin, “Overlap and Save Frequency Domain DFE for Throughput Efficient
Single Carrier Transmission,” in Proc. IEEE PIMRC, Berlin, Sept. 2005, in press.
 187

[534] L. Tomba, “On the Effect of Wiener Phase Noise in OFDM Systems,” IEEE
Trans. Commun., vol. 46, no. 5, pp. 580–583, May 1998.

[535] A. Trachtman, I. Kalet, and S. S. (Shitz), “Limiter-Discriminator Detection of

Continuous Phase Modulation (CPM) Tomlinson Filtering,” IEEE Trans.
Commun., vol. 42, no. 2/3/4, pp. 819–825, Feb./Mar./Apr. 1994.
[536] W. H. Tranter, K. S. Shanmugan, T. S. Rappaport, and K. L. Kosbar,
Communication Systems Simulation. Upper Saddle River, New Jersey: Prentice
Hall, 2004.
[537] S. Trautmann and N. J. Fliege, “Perfect Equalization for DMT Systems Without
Guard Interval,” IEEE J. Select. Areas Commun., vol. 20, no. 5, pp. 987–996,
June 2002.
[538] J. R. Treichler and B. G. Agee, “A New Approach to Multipath Correction of
Constant Modulus Signals,” IEEE Trans. Acoust., Speech, Signal Processing, vol.
assp-31, no. 2, pp. 459–472, Apr. 1983.

[539] J. Tubbax, B. Come, L. V. der Perre, L. Deneire, S. Donnay, and M. Engels,

“OFDM verses Single Carrier with Cyclic Prefix: A System-Based Comparison,”
in Some Conference, CHECK, May 2001, pp. 1115–1119.
[540] J. Tubbax, B. Come, L. V. der Perre, S. Donnay, M. Engels, H. D. Man, and
M. Moonen, “Compensation of IQ Imbalance and Phase Noise in OFDM
Systems,” IEEE Trans. Wireless Commun., vol. 4, no. 3, pp. 872–877, May 2005.
[541] U. Tureli, D. Kivanc, and H. Liu, “Experimental and Analytical Studies on a
High-Resolution OFDM Carrier Frequency Offset Estimator,” IEEE Trans. Veh.
Technol., vol. 50, no. 2, pp. 629–643, Mar. 2001.

[542] U. Tureli, H. Liu, and M. D. Zoltowski, “OFDM Blind Carrier Offset Estimation:
ESPRIT,” IEEE Trans. Commun., vol. 48, no. 9, pp. 1459–1461, Sept. 2000.

[543] V. Vadde, “PAPR Reduction by Envelope Stabilization Using Partial Response

Signaling in OFDM Systems,” in Proc. IEEE RAWCOM, Waltham, MA USA,
Aug. 2001, pp. 197–201.
[544] V. Vadde and S. Gray, “Partial Response Signaling for Enhanced Spectral
Efficiency and RF Performance in OFDM Systems,” in Proc. IEEE Globecom,
vol. 5, San Antonio, TX, Nov. 2001, pp. 3120–3124.
[545] A. Vahlin and N. Holte, “Use of a guard interval in OFDM on multipath
channels,” IEE Electr. Lett., vol. 30, no. 24, pp. 2015–2016, Nov. 1994.

[546] ——, “Optimal Finite Duration Pulses for OFDM,” IEEE Trans. Commun.,
vol. 44, no. 1, pp. 10–14, Jan. 1996.
[547] J.-J. van de Beek, M. Sandell, and P. O. Börjesson, “ML Estimation of Time and
Frequency Offset in OFDM Systems,” IEEE Trans. Signal Processing, vol. 45,
no. 7, pp. 1800–1805, July 1997.
 188

[548] C. van den Bos, M. H. L. Kouwenhoven, and W. A. Serdijn, “Effect of Smooth

Nonlinear Distortion on OFDM Symbol Error Rate,” IEEE Trans. Commun.,
vol. 49, no. 9, pp. 1510–1514, Sept. 2001.

[549] E. Van Der Ouderaa, J. Schoukens, and J. Renneboog, “Comments on

“Multitone Signals with Low Crest Factor”,” IEEE Trans. Circuits Syst., vol.
cas-34, no. 9, pp. 1125–1127, Sept. 1987.

[550] ——, “Peak Factor Minimization of Input and Output Signals of Linear
Systems,” IEEE Trans. Instrum. Meas., vol. 37, no. 2, pp. 207–212, June 1988.

[551] ——, “Peak Factor Minimization Using a Time-Frequency Domain Swapping

Algorithm,” IEEE Trans. Instrum. Meas., vol. 37, no. 1, pp. 145–147, Mar. 1988.

[552] R. van Nee, G. Awater, M. Morikura, H. Takanashi, M. Webster, and K. W.

Halford, “New High-Rate Wireless LAN Standards,” IEEE Commun. Mag., pp.
82–88, Dec. 1999.

[553] R. van Nee and A. de Wild, “Reducing the Peak-to-Average Power Ratio of
OFDM,” in Proc. IEEE VTC, vol. 3, May 1998, pp. 2072–2076.

[554] R. van Nee and R. Prasad, OFDM for Wireless Multimedia Communications.
Boston: Artech House, 2000, ch. 6.

[555] ——, OFDM for Wireless Multimedia Communications. Boston: Artech House,
2000.

[556] R. D. van Nee, “OFDM Codes for Peak-to-Average Power Reduction and Error
Correction,” in Proc. IEEE Globecom, London, Nov. 1996, pp. 740–744.

[557] L. Vandendorpe, “Multitone Spread Spectrum Multiple Access Communications

System in a Multipath Rician Fading Channel,” IEEE Trans. Veh. Technol.,
vol. 44, no. 2, pp. 327–337, May 1995.

[558] G. Vandersteen, J. Verbeeck, Y. Rolain, and J. Schoukens, “Accurate

Bit-Error-Rate estimation for OFDM based telecommunication systems schemes
in the presence of nonlinear distortions,” in Proc. IEEE IMTC, vol. 1, Baltimore,
May 2000, pp. 80–85.

[559] G. Vandersteen, P. Wambacq, Y. Rolain, J. Schoukens, S. Donnay, M. Engels,

and I. Bolsens, “Efficient Bit-Error-Rate Estimation of Multicarrier
Transceivers,” in Proc. Auto. Test Europe, Munich, Mar. 2001, pp. 164–168.

[560] A. Vannucci and R. Raheli, “Sequence Detection in Nonlinear Channels: A

Convenient Alternative to Analog Predistortion,” IEEE Trans. Commun.,
vol. 50, no. 9, pp. 1515–1524, Sept. 2002.

[561] S. J. Vaughan-Nichols, “OFDM: Back to the Wireless Future,” IEEE Computer,

pp. 19–21, Dec. 2002.
 189

[562] E. Viterbo and K. Fazel, “How to combat long echos in OFDM transmission
schemes: Sub-channel equalization or more power channel coding,” in Proc.
IEEE Globecom, vol. 3, Nov. 1995, pp. 2069–2074.

[563] T. Walzman and M. Schwartz, “Automatic Equalization Using the Discrete

Frequency Domain,” IEEE Trans. Inform. Theory, vol. IT-19, no. 1, pp. 59–68,
Jan. 1973.

[564] L. Wan and V. K. Dubey, “BER Performance of OFDM System Over Frequency
Nonselective Fast Ricean Fading Channels,” IEEE Commun. Lett., vol. 5, no. 1,
pp. 19–21, Jan. 2001.

[565] N. Wang and S. D. Blostein, “Power Loading for CP-OFDM over

Frequency-Selective Fading Channels,” in Proc. IEEE Globecom, vol. 4, Dec.
2003, pp. 2305–2309.

[566] T. Wang, J. G. Proakis, and J. R. Zeidler, “Performance Analysis of High QAM

OFDM System Over Frequency Selective Time-Varying Fading Channel,” in
Proc. IEEE PIMRC, Beijing, Sept. 2003, pp. 793–798.

[567] T. Wang and J. Ilow, “Compensation of Nonlinear Distortions with Memory

Effects in OFDM Transmitters,” in Proc. IEEE Globecom, vol. 4, Dec. 2004, pp.
2398–2403.

[568] X. Wang, P. Ho, and Y. Wu, “Robust Channel Estimation and ISI Cancellation
for OFDM Systems With Suppressed Features,” IEEE J. Select. Areas Commun.,
vol. 23, no. 5, pp. 963–972, May 2005.

[569] X. Wang, T. T. Tjhung, and C. S. Ng, “Reduction of Peak-to-Average Power

Ratio of OFDM Systems using a Companding Technique,” IEEE Trans.
Broadcast., vol. 45, no. 3, pp. 303–307, Sept. 1999.

[570] ——, “Replay to the Comments on: “Reduction of peak-to-average power ratio of
OFDM system using a companding technique”,” IEEE Trans. Broadcast., vol. 45,
no. 4, pp. 420–423, Dec. 1999.

[571] X. Wang, T. T. Tjhung, and Y. Wu, “On the SER and Spectral Analyses of
A-Law Companding Multicarrier Modulation,” IEEE Trans. Veh. Technol.,
vol. 52, no. 5, pp. 1408–1412, Sept. 2003.

[572] Z. Wang and G. B. Giannakis, “Wireless Multicarrier Communications: Where

Fourier Meets Shannon,” IEEE Signal Processing Mag., pp. 29–47, May 2000.

[573] ——, “Complex-Field Coding for OFDM Over Fading Wireless Channels,” IEEE
Trans. Inform. Theory, vol. 49, no. 3, pp. 707–720, Mar. 2003.

[574] Z. Wang, X. Ma, and G. B. Giannakis, “OFDM or Single-Carrier Block

Transmissions?” IEEE Trans. Commun., vol. 52, no. 3, pp. 380–394, Mar. 2004.
 190

[575] W. D. Warner and C. Leung, “OFDM/FM Frame Synchronization for Mobile

Radio Data Communications,” IEEE Trans. Veh. Technol., vol. 42, no. 3, pp.
302–313, Aug. 1993.

[576] W. J. Weber, III, “Performance of Phase-Locked Loops in the Presence of Fading

Communication Channels,” IEEE Trans. Commun., vol. com-24, no. 5, pp.
487–499, May 1976.

[577] L. Wei, “Commonalities of Several Modulation Techniques,” IEEE Trans.

Commun., vol. 44, no. 3, pp. 298–300, Mar. 1996.
[578] L. Wei and C. Schlegel, “Synchronization Requirements for Multi–user OFDM on
Satellite Mobile and Two–path Rayleigh Fading Channels,” IEEE Trans.
Commun., vol. 43, no. 2/3/4, pp. 887–895, Feb./Mar./Apr. 1995.
[579] S. B. Weinstein and P. M. Ebert, “Data Transmission by Frequency-Division
Multiplexing Using the Discrete Fourier Transform,” IEEE Trans. Commun.
Technol., vol. 19, no. 5, pp. 628–634, Oct. 1971.

[580] E. W. Weisstein, CRC Concise Encyclopedia of Mathematics, 2nd ed. Boca

Raton, FL: Chapman & Hall/CRC, 2003.
[581] C.-K. Wen, Y.-Y. Wang, and J.-T. Chen, “A Low-Complexity SpaceTime OFDM
Multiuser System,” IEEE Trans. Wireless Commun., vol. 4, no. 3, pp. 998–1007,
May 2005.
[582] T. A. Wilkinson and A. E. Jones, “Minimisation of the Peak to Mean Envelope
Power Ratio of Multicarrier Transmission Schemes by Block Coding,” in Proc.
IEEE VTC, vol. 2, Chicago, IL, July 1995, pp. 825–829.

[583] T. J. Willink and P. H. Wittke, “Optimization and Performance Evaluation of

Multicarrier Transmission,” IEEE Trans. Inform. Theory, vol. 43, no. 2, pp.
426–440, Mar. 1997.

[584] M. Z. Win, G. Chrisikos, and J. H. Winters, “MRC Performance for M -ary

Modulation in Arbitrarily Correlated Nakagami Fading Channels,” IEEE
Commun. Lett., vol. 4, no. 10, pp. 301–303, Oct. 2000.
[585] H. Witschnig, G. Oßberger, A. Springer, A. Koppler, L. Maurer, M. Huemer, and
R. Weigel, “A Different Look on Cyclic Prefix for SC/FDE,” in Proc. IEEE
PIMRC, vol. 2, Sept. 2002, pp. 824–828.
[586] ——, “The Effect of Blockwise Transmission on Higher-Order Modulation
Schemes for SC/FDE,” in Proc. IEEE Intern. Sympo. Personal Wireless
Multimedia Commun., Oct. 2002, pp. 1323–1327.

[587] S. Wu and Y. Bar-Ness, “A Phase Noise Suppression Algorithm for OFDM-Based

WLANs,” IEEE Commun. Lett., vol. 6, no. 12, pp. 535–537, Dec. 2002.
[588] ——, “OFDM Systems in the Presence of Phase Noise: Consequences and
Solutions,” IEEE Trans. Commun., vol. 52, no. 11, pp. 1988–1996, Nov. 2004.
 191

[589] Y. Wu and W. Y. Zou, “Orthogonal Frequency Division Multiplexing: A

Multi-Carrier Modulation Scheme,” IEEE Trans. Consumer Electron., vol. 41,
no. 3, pp. 392–399, Aug. 1995.

[590] D. Wulich, “Peak factor in orthogonal multicarrier modulation with variable

levels,” IEE Electr. Lett., vol. 32, no. 20, pp. 1859–1861, Sept. 1996.

[591] ——, “Reduction of peak to mean ratio of multicarrier modulation using cyclic
coding,” IEE Electr. Lett., vol. 32, no. 5, pp. 432–433, Feb. 1996.

[592] ——, “Comments on the Peak Factor of Sampled and Continuous Signals,” IEEE
Commun. Lett., vol. 4, no. 7, pp. 213–214, July 2000.

[593] D. Wulich and L. Goldfeld, “Reduction of Peak Factor in Orthogonal

Multicarrier Modulation by Amplitude Limiting and Coding,” IEEE Trans.
Commun., vol. 47, no. 1, pp. 18–21, Jan. 1999.

[594] D. Wulich, “Definition of Efficient PAPR in OFDM,” IEEE Commun. Lett.,

vol. 9, no. 9, pp. 832–834, Sept. 2005.

[595] D. Wulich, N. Dinur, and A. Glinowiecki, “Level Clipped High-Order OFDM,”

IEEE Trans. Commun., vol. 48, no. 6, pp. 928–930, June 2000.

[596] G. Wunder and H. Boche, “Upper Bounds on the Statistical Distribution of the
Crest-Factor in OFDM Transmission,” IEEE Trans. Inform. Theory, vol. 49,
no. 2, pp. 488–494, Feb. 2003.

[597] Xfig. [Online]. Available: http://www.xfig.org/

[598] P. Xia, S. Zhou, and G. B. Giannakis, “Bandwidth- and Power-Efficient

Multicarrier Multiple Access,” IEEE Trans. Commun., vol. 51, no. 11, pp.
1828–1837, Nov. 2003.

[599] C. Xiao, Y. R. Zheng, and N. C. Beaulieu, “Second-Order Statistical Properties

of the WSS Jakes’ Fading Channel Simulator,” IEEE Trans. Commun., vol. 50,
no. 6, pp. 888–891, June 2002.

[600] ——, “Statistical Simulation Models for Rayleigh and Rician Fading,” in Proc.
IEEE ICC, vol. 5, May 2003, pp. 3524–3529.

[601] F. Xiong, “M -ary Amplitude Shift Keying OFDM System,” IEEE Trans.
Commun., vol. 51, no. 10, pp. 1638–1642, Oct. 2003.

[602] W. Xu and L. B. Milstein, “On the Use of Interference Suppression to Reduce

Intermodulation Distortion in Multicarrier CDMA Systems,” IEEE Trans.
Commun., vol. 49, no. 1, pp. 130–141, Jan. 2001.

[603] B. Yang, K. B. Letaief, R. S. Cheng, and Z. Cao, “Timing Recovery for OFDM
Transmission,” IEEE J. Select. Areas Commun., vol. 18, no. 11, pp. 2278–2291,
Nov. 2000.
 192

[604] H. Yang, “A Road to Future Broadband Wireless Access: MIMO-OFDM-Based

Air Interface,” IEEE Commun. Mag., pp. 53–60, Jan. 2005.

[605] J. Yang, S. Roy, and N. H. Lewis, “Data-Driven Echo Cancellation For a

Multitone Modulation System,” IEEE Trans. Commun., vol. 42, no. 5, pp.
2134–2144, May 1994.

[606] K.-W. Yip and T.-S. Ng, “Efficient Simulation of Digital Transmission over
WSSUS Channels,” IEEE Trans. Commun., vol. 43, no. 12, pp. 2907–2913, Dec.
1995.

[607] ——, “Karhunen–Loéve Expansion of the WSSUS Channel and Its Application
to Efficient Simulation,” IEEE J. Select. Areas Commun., vol. 15, no. 4, pp.
640–646, May 1997.

[608] ——, “Matched Filter Bound for Multipath Rician-Fading Channels,” IEEE
Trans. Commun., vol. 46, no. 4, pp. 441–445, Apr. 1998.

[609] J. H. Yu and Y. T. Su, “Pilot-Assisted Maximum-Likelihood Frequency-Offset

Estimation for OFDM Systems,” IEEE Trans. Commun., vol. 52, no. 11, pp.
1997–2008, Nov. 2004.

[610] K. C. Yu and A. J. Goldsmith, “Linear Models and Capacity Bounds for

Continuous Phase Modulation,” in Proc. IEEE ICC, vol. 2, May 2002, pp.
722–726.

[611] J. H. Yuen, M. K. Simon, W. Miller, F. Pollara, C. R. Ryan, D. Divsalar, and

J. Morakis, “Modulation and Coding for Satellite and Space Communications,”
Proc. IEEE, vol. 78, no. 7, pp. 1250–1266, July 1990.

[612] M. Zakai, “The Representation of Narrow-Band Processes,” IEEE Trans. Inform.

Theory, vol. it-8, no. 4, pp. 323–325, July 1962.

[613] L. Zeger, P. Chen, and H. Kobayashi, “Analysis and Simulation of Multipath

Interference of FM Subcarrier Digital Signals,” in Proceedings of the IEEE
Symposium on Computers and Communications, Athens, July 1998, pp. 35–41.

[614] N. A. Zervos and I. Kalet, “Optimized Decision Feedback Equalization Versus

Optimized Orthogonal Frequency Division Multiplexing for High-Speed
Transmission Over the Local Cable Network,” in Proc. IEEE ICC, June 1989,
pp. 1080–1085.

[615] X. Zhang and M. P. Fitz, “Space–Time Code Design with Continuous Phase
Modulation,” IEEE J. Select. Areas Commun., vol. 21, no. 5, pp. 783–792, June
2003.

[616] Y. Zhao and S.-G. Häggman, “Sensitivity to Doppler Shift and Carrier Frequency
Errors in OFDM Systems—The Consequences and Solutions,” in Proc. IEEE
VTC, vol. 3, Apr. 1996, pp. 1564–1568.
 193

[617] ——, “BER Analysis of OFDM Communication Systems with Intercarrier

Interference,” in Proc. IEEE ICC, vol. 2, Oct. 1998, pp. S38–02–1 – S38–02–5.

[618] ——, “Intercarrier Interference Self-Cancellation Scheme for OFDM Mobile

Communication Systems,” IEEE Trans. Commun., vol. 49, no. 7, pp. 1185–1191,
July 2001.

[619] Y. Zhao and A. Huang, “A Novel Channel Estimation Method for OFDM Mobile
Communication Systems Based on Pilot Signals and Transform-Domain
Processing,” in Proc. IEEE VTC, vol. 3, Phoenix, May 1997, pp. 2089–2093.

[620] Y. Zhao, Jean-Damien, and S.-G. Häggman, “Intercarrier Interference

Compression of OFDM Communication Systems by Using Correlative Coding,”
IEEE Commun. Lett., vol. 2, no. 8, pp. 214–216, Aug. 1998.

[621] Y. R. Zheng and C. Xiao, “Improved Models for the Generation of Multiple
Uncorrelated Rayleigh Fading Waveforms,” IEEE Commun. Lett., vol. 6, no. 6,
pp. 256–258, June 2002.

[622] K. Zhong, T. T. Tjhung, and F. Adachi, “A General SER Formula for an OFDM
System With MDPSK in Frequency Domain over Rayleigh Fading Channels,”
IEEE Trans. Commun., vol. 52, no. 4, pp. 584–594, Apr. 2004.

[623] G. T. Zhou and J. S. Kenney, “Predicting Spectral Regrowth of Nonlinear Power

Amplifiers,” IEEE Trans. Commun., vol. 50, no. 5, pp. 718–722, May 2002.

[624] R. E. Ziemer and W. H. Tranter, Principles of Communications—Systems,

Modulation, and Noise, 4th ed. New York: John Wiley & Sons, 1995.

[625] M. S. Zimmerman and A. L. Kirsch, “The AN/GSC-10 (KATHRYN) Variable

Rate Data Modem for HF Radio,” IEEE Trans. Commun. Technol., vol. com-15,
no. 2, pp. 197–204, Apr. 1967.

[626] T. N. Zogakis and J. M. Cioffi, “The Effect of Timing Jitter on the Performance
of a Discrete Multitone System,” IEEE Trans. Commun., vol. 44, no. 7, pp.
799–808, July 1996.

[627] H. Zou, H. J. Kim, S. Kim, B. Daneshrad, R. Wesel, and W. Magione-Smith,

“Equalized GMSK, Equalized QPSK and OFDM, a Comparative Study for
High-speed Wireless Indoor Data Communications,” in Proc. IEEE VTC, vol. 2,
Houston, May 1999, pp. 1106–1110.

[628] W. Y. Zou and Y. Wu, “COFDM: An Overview,” IEEE Trans. Broadcast.,

vol. 41, no. 1, pp. 1–8, Mar. 1995.
Production Notes

This thesis was typeset using the LATEX document preparation system [348]. The
bibliography was managed using BibTeX (with help from bibtool). All numerical work,
including the computer simulations, was done with GNU Octave [188]. The block dia-
grams were drawn using Xfig [597] and all of the other figures were generated with Gnu-
plot [189] (using the pslatex driver). The source files were backed up and synchronized
among multiple computers using rsync. The LATEX output was converted to PostScript
using dvips; the PostScript was converted to PDF (portable document format) using
Ghostscript. The size of the PDF output is 1.9 megabytes.

The work was done at UCSD on a Dell Precision 370 workstation running the Debian
GNU/Linux operating system [131]. The X11 window system provided the graphical
user interface; the window manager used was IceWM. Typically the work was conducted
across several rxvt terminal emulators—arranged across multiple workspaces—running
bash. All work was done using the Vim (Vi improved) text editor [523]. PDF output
was viewed using xpdf. The work was also done at various locations throughout the San
Diego area on a Dell Inspiron 4000 laptop computer running the same software.

This thesis were printed on a Hewlett Packard LaserJet 1300n printer.

In terms of compilation time, this thesis takes roughly 10 s to compile on the work-
station which has a 3 gigahertz microprocessor and 1 gigabyte of memory.

194