Ultra-Wideband Radio Propagation Channels

Ultra-Wideband Radio
Propagation Channels
A Practical Approach
Pascal Pagani
Friedman Tchoffo Talom
Patrice Pajusco
Bernard Uguen
Series Editor
Pierre-Noël Favennec
This page intentionally left blank
Ultra-Wideband Radio Propagation Channels
Ultra-Wideband Radio
Propagation Channels
A Practical Approach
Pascal Pagani
Friedman Tchoffo Talom
Patrice Pajusco
Bernard Uguen
Series Editor
Pierre-Noël Favennec
First published in France in 2007 by Hermes Science/Lavoisier entitled: “Communications ultra large
bande : Le canal de propagation radioélectrique”
First published in Great Britain and the United States in 2008 by ISTE Ltd and John Wiley & Sons, Inc.
Apart from any fair dealing for the purposes of research or private study, or criticism or review, as
permitted under the Copyright, Designs and Patents Act 1988, this publication may only be reproduced,
stored or transmitted, in any form or by any means, with the prior permission in writing of the publishers,
or in the case of reprographic reproduction in accordance with the terms and licenses issued by the CLA.
Enquiries concerning reproduction outside these terms should be sent to the publishers at the
undermentioned address:
ISTE Ltd John Wiley & Sons, Inc.

27-37 St George’s Road 111 River Street
London SW19 4EU Hoboken, NJ 07030
UK USA
www.iste.co.uk www.wiley.com
© ISTE Ltd, 2008

© LAVOISIER, 2007
The rights of Pascal Pagani, Friedman Tchoffo Talom, Patrice Pajusco and Bernard Uguen to be
identified as the authors of this work have been asserted by them in accordance with the Copyright,
Designs and Patents Act 1988.
Library of Congress Cataloging-in-Publication Data
Communications ultra large bande. English

Ultra-wideband radio propagation channels / Pascal Pagani ... [et al.].
p. cm. -- (A practical approach)
Includes bibliographical references and index.
ISBN 978-1-84821-084-4
1. Broadband communication systems. I. Pagani, Pascal. II. Title.
TK5103.4.C6213 2008
621.382--dc22
2008030345
British Library Cataloguing-in-Publication Data

A CIP record for this book is available from the British Library
ISBN: 978-1-84821-084-4
Cover image created by Atelier Isatis, based on an original photograph by Denis Stenderchuck.
Printed and bound in Great Britain by CPI Antony Rowe Ltd, Chippenham, Wiltshire.
Contents
Foreword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Chapter 1. UWB Technology and its Applications . . . . . . . . . . . . . 21

1.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.2. Definition and historical evolution . . . . . . . . . . . . . . . . . . 22
1.2.1. Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.2.2. Historical evolution . . . . . . . . . . . . . . . . . . . . . . . . 23
1.3. Specificities of UWB . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.4. Considered applications . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.5. Regulation evolution . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.5.1. Regulation in the USA . . . . . . . . . . . . . . . . . . . . . . 31
1.5.2. Regulation in Europe . . . . . . . . . . . . . . . . . . . . . . . 32
1.5.3. Regulation in Asia . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.6. UWB communication system and standardization . . . . . . . . . 34
1.6.1. Impulse radio . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
1.6.1.1. Pulse position modulation . . . . . . . . . . . . . . . . . 35
1.6.1.2. Pulse amplitude modulation . . . . . . . . . . . . . . . . 38
1.6.2. Direct sequence UWB . . . . . . . . . . . . . . . . . . . . . . 39
1.6.3. Multiband OFDM . . . . . . . . . . . . . . . . . . . . . . . . . 40
1.7. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Chapter 2. Radio Wave Propagation . . . . . . . . . . . . . . . . . . . . . 43

2.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.2. Definition of the propagation channel . . . . . . . . . . . . . . . . 43
2.2.1. Free space propagation . . . . . . . . . . . . . . . . . . . . . . 44
2.2.2. Multipath propagation . . . . . . . . . . . . . . . . . . . . . . 45
2.2.3. Propagation channel variations . . . . . . . . . . . . . . . . . 47
6 Ultra-Wideband Radio Propagation Channels
2.2.3.1. Spatial selectivity . . . . . . . . . . . . . . . . . . . . . . 48

2.2.3.2. Frequency selectivity . . . . . . . . . . . . . . . . . . . . 48
2.2.3.3. Doppler effect . . . . . . . . . . . . . . . . . . . . . . . . 50
2.3. Propagation channel representation . . . . . . . . . . . . . . . . . 51
2.3.1. Mathematical formulation . . . . . . . . . . . . . . . . . . . . 51
2.3.2. Characterization of deterministic channels . . . . . . . . . . 52
2.3.2.1. The time varying impulse response . . . . . . . . . . . . 53
2.3.2.2. The frequency domain function . . . . . . . . . . . . . . 53
2.3.2.3. The time varying transfer function . . . . . . . . . . . . 54
2.3.2.4. The delay-Doppler spread function . . . . . . . . . . . . 54
2.3.3. Characterization of linear random channels . . . . . . . . . . 54
2.3.4. Channel classification . . . . . . . . . . . . . . . . . . . . . . . 55
2.3.4.1. Wide sense stationary channels . . . . . . . . . . . . . . 55
2.3.4.2. Uncorrelated scattering channels . . . . . . . . . . . . . 56
2.3.4.3. Wide sense stationary uncorrelated scattering channels 57
2.4. Channel characteristic parameters . . . . . . . . . . . . . . . . . . 58
2.4.1. Frequency selectivity . . . . . . . . . . . . . . . . . . . . . . . 58
2.4.1.1. RMS delay spread . . . . . . . . . . . . . . . . . . . . . . 59
2.4.1.2. Coherence bandwidth . . . . . . . . . . . . . . . . . . . . 59
2.4.1.3. Delay window and delay interval . . . . . . . . . . . . . 60
2.4.1.4. Exponential decay constants . . . . . . . . . . . . . . . . 61
2.4.1.5. Cluster and ray arrival rates . . . . . . . . . . . . . . . . 61
2.4.2. Propagation loss . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.4.3. Fast fading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
2.4.4. Spectral analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Chapter 3. UWB Propagation Channel Sounding . . . . . . . . . . . . . 67

3.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.2. Specificity of UWB channel sounding . . . . . . . . . . . . . . . . 67
3.3. Measurement techniques for UWB channel sounding . . . . . . . 70
3.3.1. Frequency domain techniques . . . . . . . . . . . . . . . . . . 71
3.3.1.1. Vector network analyzer . . . . . . . . . . . . . . . . . . 71
3.3.1.2. Chirp sounder . . . . . . . . . . . . . . . . . . . . . . . . 72
3.3.2. Time domain techniques . . . . . . . . . . . . . . . . . . . . . 73
3.3.2.1. Pulsed techniques . . . . . . . . . . . . . . . . . . . . . . 73
3.3.2.2. Correlation measurements . . . . . . . . . . . . . . . . . 75
3.3.2.3. Inversion techniques . . . . . . . . . . . . . . . . . . . . . 78
3.3.3. Multiple-band time domain sounder for dynamic channels . 78
3.3.3.1. Principle of multiple-band time domain sounding . . . 80
3.3.3.2. Description of the SIMO channel sounder . . . . . . . . 81
3.3.3.3. Extension towards UWB . . . . . . . . . . . . . . . . . . 81
3.3.3.4. Experimental validation . . . . . . . . . . . . . . . . . . 84
Contents 7
3.4. UWB measurement campaigns . . . . . . . . . . . . . . . . . . . . 85

3.4.1. Overview of UWB measurement campaigns . . . . . . . . . . 85
3.4.2. Illustration of channel sounding experiments . . . . . . . . . 91
3.4.2.1. Static measurement campaign over the 3.1–10.6 GHz
band . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.4.2.2. Static measurement campaign over the 2–6 GHz band 95
3.4.2.3. Dynamic measurement campaign over the 4–5 GHz
band . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3.5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
Chapter 4. Deterministic Modeling of the UWB Channel . . . . . . . . 99

4.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.2. Overview of deterministic modeling . . . . . . . . . . . . . . . . . 99
4.2.1. FDTD based approach . . . . . . . . . . . . . . . . . . . . . . 100
4.2.2. MoM based approach . . . . . . . . . . . . . . . . . . . . . . . 100
4.2.3. Ray based approach . . . . . . . . . . . . . . . . . . . . . . . . 101
4.3. Specificity of deterministic modeling in UWB . . . . . . . . . . . 101
4.4. Overview of UWB deterministic modeling . . . . . . . . . . . . . 102
4.4.1. Qiu model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.4.2. Yao model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.4.3. Attiya model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
4.4.4. Uguen and Tchoffo Talom model . . . . . . . . . . . . . . . . 104
4.5. Illustration of a deterministic model formalism . . . . . . . . . . . 104
4.5.1. Received signal synthesis . . . . . . . . . . . . . . . . . . . . . 105
4.5.2. Ray impulse response without delay . . . . . . . . . . . . . . 105
4.5.3. Ray channel matrix without delay . . . . . . . . . . . . . . . 108
4.5.4. Described model results . . . . . . . . . . . . . . . . . . . . . 110
4.5.4.1. Emitted waveform and considered scenario . . . . . . . 110
4.5.4.2. Channel matrix of each emitted waveform in the LOS
case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
4.5.4.3. Received signal with ideal antennas . . . . . . . . . . . 114
4.6. Consideration of real antenna characteristics in deterministic
modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
4.7. Building material effects on channel properties . . . . . . . . . . . 120
4.8. Simulation and measurement comparisons . . . . . . . . . . . . . 124
4.8.1. Evaluation of real antenna consideration . . . . . . . . . . . 124
4.8.2. Evaluation of impulse response reconstruction . . . . . . . . 125
4.9. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
Chapter 5. Statistical Modeling of the UWB Channel . . . . . . . . . . 133

5.1. Experimental characterization of channel parameters . . . . . . . 134
5.1.1. Propagation loss . . . . . . . . . . . . . . . . . . . . . . . . . . 134
5.1.1.1. Frequency propagation loss . . . . . . . . . . . . . . . . 134
5.1.1.2. Distance propagation loss . . . . . . . . . . . . . . . . . 136

5.1.2. Impulse response characterization . . . . . . . . . . . . . . . 137
5.1.2.1. Delay spread . . . . . . . . . . . . . . . . . . . . . . . . . 137
5.1.2.2. Power delay profile decay . . . . . . . . . . . . . . . . . 141
5.1.2.3. Ray and cluster arrival rate . . . . . . . . . . . . . . . . 145
5.1.3. Study of small-scale channel variations . . . . . . . . . . . . 148
5.1.4. Effect of moving people . . . . . . . . . . . . . . . . . . . . . . 151
5.1.4.1. Observation of temporal variations . . . . . . . . . . . . 151
5.1.4.2. Slow fading . . . . . . . . . . . . . . . . . . . . . . . . . . 152
5.1.4.3. Fast fading . . . . . . . . . . . . . . . . . . . . . . . . . . 153
5.1.4.4. Spectral analysis . . . . . . . . . . . . . . . . . . . . . . . 156
5.2. Statistical channel modeling . . . . . . . . . . . . . . . . . . . . . . 157
5.2.1. Examples of statistical models . . . . . . . . . . . . . . . . . 158
5.2.1.1. IEEE 802.15.3a model . . . . . . . . . . . . . . . . . . . 158
5.2.1.2. IEEE 802.15.4a model . . . . . . . . . . . . . . . . . . . 159
5.2.1.3. Other models . . . . . . . . . . . . . . . . . . . . . . . . . 160
5.2.2. Empirical modeling principles . . . . . . . . . . . . . . . . . . 162
5.2.2.1. Propagation loss model . . . . . . . . . . . . . . . . . . . 162
5.2.2.2. Modeling the channel impulse response over an infinite
bandwidth . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
5.2.2.3. Modeling the channel impulse response over a limited
bandwidth . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
5.2.2.4. Simulation results . . . . . . . . . . . . . . . . . . . . . . 166
5.3. Advanced modeling in a dynamic configuration . . . . . . . . . . 169
5.3.1. Space variation modeling . . . . . . . . . . . . . . . . . . . . . 169
5.3.2. Modeling the effect of people . . . . . . . . . . . . . . . . . . 172
5.4. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
Appendices
A. Baseband Representation of the Radio Channel . . . . . . . . . . . 177
B. Statistical Distributions . . . . . . . . . . . . . . . . . . . . . . . . . 181
B.1. Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
B.1.1. Rayleigh distribution . . . . . . . . . . . . . . . . . . . . . 181
B.1.2. Rice distribution . . . . . . . . . . . . . . . . . . . . . . . 182
B.1.3. Nakagami distribution . . . . . . . . . . . . . . . . . . . . 183
B.1.4. Weibull distribution . . . . . . . . . . . . . . . . . . . . . 184
B.1.5. Normal distribution . . . . . . . . . . . . . . . . . . . . . . 184
B.1.6. Log-normal distribution . . . . . . . . . . . . . . . . . . . 185
B.1.7. Laplace distribution . . . . . . . . . . . . . . . . . . . . . 185
B.2. Kolmogorov-Smirnov goodness-of-fit test . . . . . . . . . . . . 186
C. Geometric Optics and Uniform Theory of Diffraction . . . . . . . . 189
C.1. Geometric optics . . . . . . . . . . . . . . . . . . . . . . . . . . 189
C.1.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 189
Contents 9
C.1.2. Field locality principle . . . . . . . . . . . . . . . . . . . . 190

C.1.3. Field expression in geometric optics . . . . . . . . . . . . 191
C.1.4. Change of local basis . . . . . . . . . . . . . . . . . . . . . 192
C.1.5. Incident field . . . . . . . . . . . . . . . . . . . . . . . . . . 192
C.1.6. Reflected field . . . . . . . . . . . . . . . . . . . . . . . . . 193
C.1.7. Refracted and transmitted field . . . . . . . . . . . . . . . 197
C.2. Uniform theory of diffraction . . . . . . . . . . . . . . . . . . . 200
C.2.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 200
C.2.2. Diffracted field . . . . . . . . . . . . . . . . . . . . . . . . . 200
C.2.3. UTD 2D coefficient . . . . . . . . . . . . . . . . . . . . . . 201
C.2.4. UTD 3D coefficient . . . . . . . . . . . . . . . . . . . . . . 204
D. Ray Construction Techniques . . . . . . . . . . . . . . . . . . . . . . 209
D.1. Ray launching . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
D.2. Ray tracing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
D.3. Other techniques . . . . . . . . . . . . . . . . . . . . . . . . . . 211
E. Description of the Time-Frequency Transform . . . . . . . . . . . . 213
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
Foreword
Although the origins of distant signal transmission are ancient, the

theoretical foundations of modern telecommunication techniques are
owed to Claude Shannon’s 1948 publications. Since then, the field of
telecommunications has not stopped evolving. In particular, spread spectrum
techniques enabled unprecedented improvements in the quality and security of
digital communications under harsh transmission conditions.
Today, the main players in the telecommunication world are facing an

increasing demand for multimedia wireless applications, linked to a real need
for very high throughput radio communication systems. Among the most recent
innovations in this field, the scientific community is particularly interested
in ultra-wideband (UWB) technology. This technique consists of transmitting
radio signals spreading over very large frequency bandwidths, typically in the
order of 500 MHz to several GHz. Initially developed in the field of radar
localization, UWB technology is now seen as a promising candidate for future
wireless transmission systems. As such, it is considered with an increasing
interest in both scientific and industrial communities.
UWB technology undeniably offers numerous advantages and

unprecedented possibilities in the design of radio systems. Its very large
spectral bandwidth optimally exploits the benefits of spread spectrum
techniques, by increasing the transmission capacity and by improving the
jamming immunity. Simultaneously, UWB presents a high resolution in
the time domain, which may be used to efficiently process the multiple
propagation paths or handle localization issues. In particular, the capacity
of UWB pulses to travel through different materials allows us to consider
through-wall imaging applications. It is also noteworthy that UWB systems
present a low power spectral density, which not only increases the discretion
and security of wireless communications, but also reduces the potential
jamming experienced by other spectrum users. Finally, the complexity of
UWB transceivers may be considerably reduced with respect to traditional

architectures.
These characteristics are unique to UWB technology and enable the

design of communication systems offering very high data rates, up to several
hundred Mbps. In 2002, the Federal Communication Commission (FCC) –
an American regulation body – authorized the transmission of UWB signals
in the 3.1–10.6 GHz band, encouraging the research efforts in this field in
all continents. In Europe, a transmission mask for UWB signals has been
defined and the coexistence between UWB systems and other applications is
under study. In the current context of high demand for wireless multimedia
applications, UWB seems to be an innovative and attractive solution for
future radio communication systems. As an illustration, important industrial
consortiums such as UWB Forum and WiMedia Alliance are currently
involved in the design of UWB based equipment.
These systems are particulary well suited to ad hoc communication

networks, but a number of other applications may be envisioned by exploiting
the unique characteristics of UWB. The possibility of accurate localization
enable the development of sensor networks known as radio frequency
identification (RFID), for industrial environment applications or for mass
market geographical information services. Radar identification techniques
based on UWB signals may be exploited to design anti-collision systems for
vehicles, but may also be used in the fields of civil engineering, medicine and
imaging. Numerous applications still need to be explored, associating UWB
with advanced techniques such as multiple-input multiple-output (MIMO)
antennas techniques or time reversal techniques.
In order to develop such systems, many challenges are still to be

encountered. The short duration of UWB signals requires a high accuracy in
the system synchronization procedure. Also, electronic components need to be
adapted in order to process such a wide frequency band, while maintaining
a tractable complexity. In particular, the antenna characteristics largely vary
with increasing frequency and require an accurate characterization.
Among these scientific challenges, a complete knowledge of the radio

channel properties is fundamental. Indeed, the performance of a wireless
transmission system is directly linked to the propagation conditions between
the emitter and the receiver. These devices need to be designed in order
to benefit from the channel characteristics and to mitigate the channel
impairments. For instance, modeling the propagation loss allows us to estimate
the radio system coverage, while link level simulations may be used to assess
the communication robustness. Owing to the width of its frequency band, the
Foreword 13
UWB propagation channel is intrinsically different from traditional wideband

channels. For instance, the interactions between the radio waves and their
environment need to be described more accurately and the variations of the
material properties with frequency need to be taken into account. It is thus
necessary to closely study the propagation channel in order to evaluate the
potential and the constraints attached to UWB communication systems.
This textbook results from an intense collaboration between France

Telecom’s Research and Development Division and the IETR–UMR CNRS
6164 (Institut d’Electronique et de Télécommunications de Rennes/Institute
of Electronics and Telecommunications in Rennes). These research teams
conducted joint studies on UWB techniques and on the impact of the
transmission channel on UWB communication systems. Through its didactic
presentation and the detailed illustration of the discussed topics, this
document is an excellent introduction for engineers and communication
systems designers as well as for researchers and lecturers willing to expand
their knowledge to the field of the UWB transmission channel. The originality
of this textbook lies in its experimental approach, which allows the reader
to follow step by step the theoretical and practical aspects of radio channel
characterization and modeling. This approach may easily be adapted to
different contexts: for different applications, in other environments or in
different frequency bands, such as the available bands around 17 GHz and
60 GHz. It should also be noted that the accurate knowledge of the UWB
channel in the 3.1–10.6 GHz band gives access to all the useful information for
developing systems included in this frequency band, such as WiFi systems
operating around 5 GHz.
This book is divided into five distinct parts.
Chapter 1 presents UWB technology. Its historical evolution, the

envisioned applications and its main characteristics are detailed. UWB
spectrum regulation issues and the proposed communications techniques are
also discussed.
Chapter 2 describes the propagation of electromagnetic waves in general.

All large scale and small scale radioelectric phenomena at play are highlighted.
The reader is then introduced to the area of mathematical representation and
its characteristic parameters. This didactic presentation covers both indoor and
outdoor environments and is applicable in both contexts of mobile radio and
wireless networks.
Chapter 3 presents an overview of the channel sounding techniques adapted

to UWB technology. A distinction is made between frequency domain and
time domain techniques, with a discussion on the application domains and the
limitation of each solution. The main UWB channel measurement campaigns

available in the literature, including those conducted by the authors, are listed
and illustrated by a few examples.
For the study of the propagation phenomena and the design of

communication systems, the UWB propagation channel is simulated using
deterministic or statistical models. These two modeling approaches are detailed
and illustrated in the last two chapters. Chapter 4 focuses on deterministic
UWB modeling. A literature review of the proposed deterministic models
is given, comparing the advantages and drawbacks of each proposal. The
fundamental issues and the theoretical formalism of a UWB deterministic
model are then detailed. To illustrate this presentation, some examples are
given, where the authors describe a comprehensive deterministic simulator for
the UWB channel.
Chapter 5 is dedicated to the statistical modeling of the UWB transmission

channel. The characterization of the most representative radio channel
parameters is first presented. Different results available in the literature
are compared and commented upon, hence providing a rich experimental
database on the UWB propagation channel characteristics. After a description
of different statistical models, the principles of statistical modeling and the
different related issues are presented following an experimental method. This
approach is illustrated using a model designed by the authors, allowing for the
simulation of the UWB channel in both static and dynamic environments.
Finally, the interested reader will find useful additional information

regarding channel analysis and modeling in the appendices. The discussed
technical material includes the fields of signal processing, statistics, geometrical
optics and diffraction theory. It should also be noted that the numerous
bibliographical references constitute an abundant source of information, which
may easily be exploited to learn more about the last advances in this field.
This book examines all the characteristics of the UWB transmission

channel and provides useful tools for designing efficient UWB systems.
Non-specialists will be introduced to UWB technology in general and more
particularly to the propagation channel, which is the key element in a
communication system. Specialists will find valuable information and a
practical approach in order to design simulators, set-up measurements, study
the channel characteristics and define models for UWB or in other contexts.
This reference document is also an excellent basis for further research on
advanced techniques, such as time reversal or UWB transmission in a MIMO
Foreword 15
configuration. I am convinced that this textbook will prove essential reading

in future research in the field of UWB communications.
Professor Ghaı̈s El Zein

Deputy Director at IETR
Acronyms
ADC Analog-to-digital converter

AGC Automatic gain control
BER Bit error rate
BPSK Binary phase shift keying
CDF Cumulative density function
CDMA Code division multiple access
CEPT Conference of European Postal and Telecommunications
DAA Detect and avoid
DAC Digital-to-analog converter
DCS Digital communication system
DOA Direction of arrival
DOD Direction of departure
DSO Digital sampling oscilloscope
DS-UWB Direct sequence ultra-wideband
ECC Electronic Communication Committee
EIRP Effective isotropic radiated power
ESD Energy spectral density
ETSI European Telecommunications Standards Institute
FCC Federal Communication Commission
FDML Frequency domain maximum likelihood
FDTD Finite difference time domain
GO Geometrical optics
GPS Global positioning system
GSM Global system for mobiles
GTD Geometrical theory of diffraction
IDA Infocom Development Authority of Singapore
IEEE Institute of Electrical and Electronics Engineers
IF Intermediate frequency
IMST Institut für Mobil- und Satellitenfunktechnik
IR Impulse response
ISB Incident shadow boundary
ISM Industrial, scientific and medical
ITU International telecommunication union
LDC Low duty cycle
LNA Low noise amplifier
LO Local oscillator
LOS Line of sight
MB-OFDM Multiband orthogonal frequency division mutliplexing
MIC Ministry of Internal Affairs and Communications
MIMO Multiple-input multiple-output
MoM Method of moments
NLOS Non-line of sight
OFDM Orthogonal frequency division mutliplexing
OOK On-off keying
PAM Pulse amplitude modulation
PDA Personal digital assistant
PDF Probability density function
PDP Power delay profile
PLL Phase locked loop
PN Pseudo-noise
PPM Pulse position modulation
PSD Power spectral density
QAM Quadrature amplitude modulation
QPSK Quadrature phase shift keying
RF Radio frequency
RFID Radio frequency identification
RIR Ray impulse response
RSB Reflection shadow boundary
SAGE Space alternating generalized expectation
SHF Super high frequencies
SIMO Single-input multiple-output
SISO Single-input single-output
SWR Standing wave ratio
UHF Ultra high frequencies
UMTS Universal mobile telecommunications system
UNII Unlicensed national information infrastructure
US Uncorrelated scattering
UTD Uniform theory of diffraction
UWB Ultra-wideband
VCO Voltage control oscillator
VNA Vector network analyzer
WBAN Wireless body area network
Acronyms 19
WiFi Wireless fidelity

WLAN Wireless local area network
WPAN Wireless personal area network
WSS Wide-sense stationary
WSSUS Wide-sense stationary uncorrelated scattering
802.11 IEEE task group on WLAN
802.15.3 IEEE task group on high rate WPAN
802.15.4 IEEE task group on low rate WPAN
Chapter 1
UWB Technology and its Applications
1.1. Introduction
By the end of the 20th century, studies in the telecommunication field had
made significant progress. The advent of new radiocommunication technologies
allowed telephony to change from a telegraphic transmission support to a
radio transmission support. Over the last few years, the processing speed and
storage size of the computers have increased considerably. This explains the
general public’s passion for communicating objects, which require the high
speed transfer of a great amount of information.
One of the current scientific challenges, where significant research efforts

are engaged, is related to the use of very high data rate radio transmissions
techniques on relatively short ranges. In this context, ultra-wideband (UWB)
technology, initially used in radar, appears to be an ideal candidate for future
wireless communication systems.
This chapter presents the UWB technology and its applications for wireless
communication systems. After a definition of UWB and of its historical
evolution, its characteristics are first outlined, then the considered applications
and the regulation spectrum in the USA, Asia and Europe are detailed. The
chapter ends by presenting the modulation techniques proposed for UWB and
by a state of the art of standardization today.
1.2. Definition and historical evolution
1.2.1. Definition
The generic term UWB is used to represent a radio technique which was
studied under various names. In the earliest writings on this field, we can find
the terms impulse radio, carrier-free radio, baseband radio, time domain radio,
non-sinusoid radio, orthogonal function radio and large relative bandwidth
radio [BAR 00]. The relative bandwidth is defined by:
fH − fL
Bf,3 dB = 2 · [1.1]
fH + fL
where fH and fL respectively represent the upper and lower cut-off frequencies
of the band defined at −3 dB. Initially, UWB signals were defined by a relative
bandwidth of 25% or more [TAY 95]. In 2002, the American regulation
authority, the Federal Communication Commission (FCC), extended this
definition to a broader category of signals, by including signals with a relative
bandwidth Bf,10 dB higher than 20% or with a frequency band higher than
500 MHz [FCC 02]. Typically, the bandwidth of UWB signals is about
500 MHz to several GHz. Thus, the denomination UWB not only includes
impulse techniques, but also all the modulations presenting an instantaneous
band higher than or equal to 500 MHz.
GHz
MHz
Conventional kHz
Power spectral density
narrowband
(dBm.MHz -1)
modulation
Spread
spectrum
Ultra-wideband -41.3 dB m.MHz -1
Bandwidth (Hz)
Figure 1.1. Comparison of various radio system spectrums
Figure 1.1 illustrates the comparison between conventional radio systems,

which generally modulate a narrowband signal on a carrier frequency, wideband
UWB Technology and its Applications 23
systems, with spreading spectrum for example, and ultra-wideband systems,

which show a weak power spectral density. As a comparison, the bandwidth of
UMTS signals is 5 MHz.
1.2.2. Historical evolution
The study of electromagnetism in the time domain began 40 years ago. The
first research was concentrated on radar applications because of the broadband
nature of the signals, which implies a strong resolution1 in the time domain.
It was in 1960 that impulse radars were developed by the American and
Soviet armies. Indeed, impulse systems have good space resolution properties.
The resolution in distance of a system is conversely proportional to its
bandwidth; the brevity of an impulse signal determines its spectrum width.
In the 1970s, Bennett and Ross presented a complete study of the first
research carried out on UWB [BEN 78]. Two decades later, Taylor described
the bases of the UWB technology applied to radar [TAY 95]. Since the middle
of the 1960s, research on this field regularly progressed, as mentioned in the
historical bibliography published by Barrett [BAR 00]. However, the use of
UWB signals for radio communication was not really considered before the end
of the 20th century. In 1990, the Department of Defense of the US government
published the results of its evaluation of UWB technology. These results were
mainly concentrated on radar systems, since no application of UWB technology
to communication systems was considered at that point [FOW 90].
More recently, research focused on UWB signals for radio communication

[SCH 93, SCH 97a], by using the main characteristics of this technique: a time
resolution around one nanosecond due to the huge frequency bandwidth, a short
duty cycle allowing for modulations such as time hopping and the management
of multi-users, as well as a possible transmission without carrier, which can
lead to a simplification of the radio system architecture [FOE 01a].
In 1998, the FCC launched the first study on UWB. In February 2002, a
first regulatory report was published, allowing particularly for the transmission
of signals in the 3.1–10.6 GHz band for wireless communications, with strong
constraints on the power spectral density [FCC 02]. From this date, intensive
academic and industrial research was undertaken with the aim of proposing a
powerful communication systems using UWB technology.
1. The resolution of a system is its capacity to separate very close energy paths.
1.3. Specificities of UWB
With the need for increasing the data rate of wireless systems, the UWB
technology seems to be an ideal candidate for future radio communication
systems in various types of networks, which can be residential, office, ad hoc,
etc.
As shown in section 1.2.1, the main characteristic of the UWB signal is the
width of the occupied frequency band, typically about 500 MHz to several GHz.
Information theory teaches us that with the use of an appropriate code, it is
possible to transmit data with a binary error rate (BER) lower than a fixed
arbitrarily low threshold, that is, if the data rate is lower than the theoretical
value of the transmission channel capacity. Thus, channel capacity C gives an
indication of the theoretical maximum data rate reachable with a given channel.
It can be obtained using Shannon’s theorem [SHA 49]:

S
C = Bw · log2 1 + [1.2]
N
where C is the channel capacity (bit/s), Bw is the signal bandwidth (Hz), S is

the signal power level (W) and N is the noise power level (W).
We can note that for a given bandwidth the channel capacity increases in a
logarithmic way with the signal to noise ratio. In the case of a constant signal
to noise ratio, the capacity increases linearly with the signal bandwidth.2
In the context of an increasing demand for communication systems with

high data rates, radio technologies using a broad spectral band are more able
to propose convenient data rates. So, with a frequency band reaching several
GHz, the UWB is well adapted to an increase of data rate than the systems
showing a high constraint on the bandwidth [FOE 01a].
The main properties of UWB systems are described below:
• High temporal resolution capability

Because of their great bandwidth, UWB signals have a high temporal
resolution, typically about one nanosecond. A first implication of this
property is related to the localization: knowing the delay of a signal with
2. We can note that for a given signal power level S, the capacity increases nonlinearly
with the bandwidth and reaches the asymptotic value of Clim = NS0 · log2 (e), where
N0 is the noise power spectral density.
a precision of about 0.1 to 1 ns, it is possible to obtain information on

the position of the transmitter with an accuracy of 3 to 30 cm.
• Robustness against fading related to multipath propagation

In usual propagation channels, narrowband systems suffer from fading
related to the multipath which combine in a destructive way. In the case of
impulse signals, the transmitted waveforms can have a great bandwidth,
so the multipath presenting delays lower than one nanosecond can be
resolved and added in a constructive way. This recombination causes some
complications on the system implementation, as it leads to the design of
a receiver with a great number of diversity branches.
• Low power spectral density

This characteristic is not intrinsic to UWB signals as they were defined
(see equation [1.1]), but it is imposed by the radio spectrum regulation
authorities. Indeed, as UWB signals present a wide spread spectrum,
the occupied frequency band necessarily covers the frequencies already
allocated to existing radio systems. To allow a peaceful coexistence of
UWB with existing narrowband radio technologies, the FCC has limited
the power spectral density of UWB signals to −41.3 dBm.MHz−1 ,
which corresponds to the power spectral density limit of authorized
non-intentional radio transmissions.3 This low power spectral density
improves the safety of UWB radio communications, since it becomes
more difficult to detect the transmitted signals. Another consequence
of this characteristic concerns the propagation distance, which is thus
limited to about ten meters. So, the applications considered for UWB
systems are short range and high data rate, and are particularly adapted
to the development of ad hoc networks.
• Less sensitivity to jamming

UWB systems offer a great processing gain.4 Thus, the interference UWB
systems may have on other systems is reduced, thanks to the low level of
the power spectral density authorized by the FCC. On the contrary, the
interference caused by narrowband systems on UWB systems is a priori
minimized by the bandwidth covered by the impulse signals.
3. Thus, we can consider that UWB radio signals are transmitted “under the noise
level”, although the imposed limits remain above the thermal noise.
4. The processing gain of a system is a parameter which gives an indication of the
resistance of this system against the jamming caused by the other systems.
• Protected and secured communications

UWB signals are signals that are by nature difficult to detect. Indeed, they
are spread over a broad band and transmitted with a low power spectral
density close to the noise floor level of traditional radiocommunication
receivers. These characteristics enable secured transmissions with a weak
probability of detection and a weak probability of interception.
• Relative simplicity of the systems design
In terms of implementation, the conventional radio systems are generally
of heterodyne design: the signal which codes the transmitted data is
generated in baseband and is then transposed to higher frequencies
before being emitted. The UWB technology allows the use of impulse
generated in baseband and directly transmitted on the radio channel
without modulation. This possibility of transmission without carrier may
simplify the architecture of the radio systems. Indeed, it is possible to
design UWB transmitter-receivers without any synthesizer using a phase
locked loop (PLL), any voltage control oscillator (VCO) or any mixer.
Thus, this can lead to the design of systems with low manufacturing and
marketing costs.
• Good obstacle penetration properties
UWB signals offer good capabilities of penetration in the walls and the
obstacles, in particular for the low frequencies part of the spectrum. This
makes it possible to have a good precision in terms of localization and
tracking [DEN 03b]. The American company Time Domain, pioneer in
the topic of UWB for communication, has developed a broad activity
around radar systems of vision through walls.
1.4. Considered applications

For a few years, the world of telecommunications has faced an increasing
demand for wireless numerical applications, in the industrial environment as
well as from the general public. Moreover, we can add to this tendency an
important need for total connectivity, the information having to be available
for anybody, anywhere and at any time [POR 03b]. This increasing need for
wireless connectivity leads to the development of many standards for wireless
and short range communication systems. We can mention Bluetooth, the WiFi
standards family (IEEE 802.11 a, b and g), Zigbee (IEEE 802.15.4) and the
recent standard 802.15.3. We may note that the majority of these technologies
for wireless local area networks (WLAN) and personal area networks (WPAN)
uses free frequencies in ISM and UNII bands, with a maximum bandwidths of
about 10 MHz.
A comparative study of the characteristics of some wireless technologies is

presented in Table 1.1. Taking into consideration other wireless systems like
Bluetooth or WiFi, the UWB technology presents a very low level of emission.
However, we can reach transmission data rates that are definitely higher than
the two other technologies.
Figure 1.2 shows the position of UWB in comparison with the leading
WLAN and WPAN standards in terms of rate and maximum achievable
ranges. We can note that contrarily to WiFi standards, UWB technology
mainly addresses short range WPAN networks. However, its potential data
rate exceeds the performances of all current WLAN and WPAN standards.
1000
Maximum bit rate (Mbps)
UWB
100
802.15.3 WiFi 802.11a WiFi 802.11g
10 WiFi 802.11b
1 Bluetooth
0.1 Zigbee 802.15.4a (Low rate UWB)
10 20 30 40 50 100
Maximum indoor range (m)
Figure 1.2. WLAN and WPAN main standards: rate and maximum ranges
In order to provide a high data rate anywhere, the future networks will have
to be designed considering an optimization of the space capacity, namely the
global available data rate per unit of area.
In narrow band, we generally regard the spectral capacity of the systems

in (bps/Hz) as one of the main parameters for a transmission. The various
elements of a communication system, like the modulation, the coding, the
implementation, etc. make it possible to improve it. By increasing only the
transmission power or the signal band, the capacity also increases. However,
the spectral resource for these systems is limited and their power cannot be
increased indefinitely. It is limited by medical or commercial considerations,
for example the pollution of the spectrum or the life duration of the batteries.
For UWB systems, the transmission level or the transmitted power spectral
density must be kept sufficiently low because these systems operate in already
28
Technology Data rate Frequency band EIRP Modulation Specification

UWB ≥ 100 Mbps 3.1–10.6 GHz −41.3 dBm/MHz PPM, OFDM, CDMA, . . . IEEE 802.15.3a
≥ 500 kbps 3.1–10.6 GHz −41.3 dBm/MHz PPM, OFDM, CDMA, . . . IEEE 802.15.4a
Bluetooth ≤ 700 kbps ISM 2.4 GHz type 1: 20 dBm GMSK IEEE 802.15.1
type 2: 0 dBm
WiFi ≤ 54 Mbps 5 GHz 0.2–1 W BPSK, 16-QAM, QPSK, IEEE 802.11a
64-QAM
≤ 11 Mbps ISM 2.4 GHz 0.1–2 W CCK, BPSQ, QPSK, DSS IEEE 802.11b
≤ 54 Mbps ISM 2.4 GHz 0.1–1 W BPSK, 16-QAM, QPSK, IEEE 802.11g
64-QAM, OFDM
Table 1.1. Comparison of some wireless communication technologies

occupied bands. These low levels are compensated by the use of a broad band.
So, compared to the existing wireless systems (see Table 1.2), UWB technology
has a low spectral capacity. Thus, it is more correct to speak about the spatial
capacity [GHA 04b]. This parameter corresponds to the maximum data rate of
a system divided by the surface covered by this system.
Rate Distance Spatial Spectral

capacity capacity
2
(Mbps) (m) (kbps/m ) (bps/Hz)
UWB 100 10 318.3 0.013
IEEE 802.11a 54 50 6.9 2.7
Bluetooth 1 10 3.2 0.012
IEEE 802.11b 11 100 0.350 0.1317
Table 1.2. Comparison of the spatial and spectral capacity of some wireless systems
Wireless and very high data rate radio technologies like UWB will make
it possible to considerably increase the spatial capacity, by the development
of dynamic ad hoc networks [POR 03b]. Finally, we can note that a
standardization work is currently in progress in the IEEE 802.15.4a work
group to use UWB spectrum within the framework of low data rate and short
range radio links. The expected data rate is typically the same as that of the
Zigbee standard, with a range of about a hundred meters.
Thus, the potential applications of UWB radio technology are mainly related
to two techniques: high data rate for short range systems (typically 200 Mbps up
to 10 m), and low data rate for long range systems (typically 200 kbps at 100 m).
These two ways of using the UWB radio spectrum allow us to consider a
given number of typical applications for UWB systems [YAN 04, POR 03b].
Firstly, UWB technology will make it possible to increase the data rate of
traditional personal wireless networks. So, it will be useful for WiFi networks
which make wireless access to the Internet network possible, or for connections
between various peripherals (printer, readers, etc.) in limited size environments
of, for example, one or more rooms. Because of a potential of very high data
rate in short range, applications requiring a higher data rate are also possible
with a range from 1 to 4 meters, for example a high quality multi-media transfer
between a DVD player and a screen. In the same manner, the UWB promoters
also proposed a wireless alternative for the Ethernet standard.
In addition, UWB applications are expected to be used for house

automation, where a great number of devices able to communicate at a
distance of several dozen meters are deployed in an office or residential
environment. In this usage, the exploited characteristics of UWB systems are
the low costs of the equipment and the possibility of obtaining localization
information. Such potential house automation applications include detection
of intrusion, or the electronic reception (owner detection and launching of
services like the unlocking of doors).
In outdoor scenarios, UWB technology will be used for point-to-point

communication applications. An example usage is the transfer of data between
several personal devices. In addition, some studies are in progress concerning
services of multimedia contents diffusion from electronic kiosks. An example
of application typical of such electronic kiosks is the download of stockmarket
information on a portable digital assistant (PDA) every time the user is
waiting at his regular subway station.
Finally, UWB applications are also envisioned in the industrial context.

By exploiting the possibilities of long range localization combined with
information transfer, sensor networks could be deployed in the productions
lines or warehouses, in order to follow-up and automatically manage the
operations. This kind of application is well suited for UWB low data rate and
long range communication. The main challenge to be raised for this type of
applications is the control of radio communication under difficult propagation
and interference conditions.
1.5. Regulation evolution

The bandwidth of UWB signals requires a strict regulation of their
transmission spectrum. Indeed, many systems, whether licensed or not,
are presented in UHF and SHF bands, which are very favorable for radio
systems deployment. To allow the use of UWB signals over several GHz,
regulatory authorities imposed a strict limitation on the transmission power.
Figure 1.3 shows some radio systems existing in UHF and SHF bands.
We can note that there are reserved bands for several systems like the
standards of cellular telephony GSM (900 MHz), DCS (1.8 GHz) and UMTS
(2 GHz). The global positioning system (GPS) also occupies a reserved band
around 1.5 GHz. Other frequency bands are already used for unlicensed
communication systems. For example, the ISM band is used by systems such
as Bluetooth, WiFi and DECT, and is also authorized for domestic devices
such as microwave ovens. The UNII band is the frequency band where the
WiFi 802.11a and HiperLAN standards operate.
In order to limit the UWB signals effects on the other radio systems, the
regulatory agencies agree on the use of the 3.1–10.6 GHz band for UWB
signals [AIE 03b].5 This part of the radio spectrum allows us to use a
bandwidth as wide as 7.5 GHz, while avoiding telephony systems and GPS.
The very low authorized power spectral density, located under the level of
unintentional emission fixed by the FCC (−41.3 dBm.MHz−1 ), is compensated
by the bandwidth, allowing it to emit a total power of 0.6 mW.
UMTS
GSM
GPS
DCS
UNII
ISM
Bluetooth, 802.11a,
Power spectral density
802.11b, HiperLAN
DECT,
(dBm.MHz-1)
microwave
ovens Unintentional
radiation limit
(-41.3 dBm.MHz-1)
UWB
0.9 1.5 2.0 2.4 3.1 5 10.6

1.8 Frequency (GHz)
Figure 1.3. Radio systems in UHF and SHF bands
1.5.1. Regulation in the USA
In the USA, the regulatory agency FCC launched its first works on UWB
radio technology as early as 1998 [MOR 03, POR 03a]. In May 2000, a first
proposal for a regulation was published (Notice of Proposed Rule Making),
which led to the current regulatory text Report & Order of February 2002
[FCC 02].
The FCC rules of the UWB spectrum regulation enable it to emit signals
mainly in the frequency band 3.1–10.6 GHz, by respecting a power spectral
density lower than the one applied to non-intentional radio transmissions. Three
different classes of equipment are considered:
• Visualizing systems: ground penetrating radars, through-wall visualizing
systems, medical systems and monitoring systems.
5. We will see that in certain regions of the world (Europe, Asia), only a part of the
3.1–10.6 GHz band is considered.
• On-board radar systems: for example, the radars for cars in the 24–29 GHz
band.
• Communication and measurement systems.
Each class of equipment has its own emission mask. Figure 1.4 presents the
emission mask of the communication systems, for indoor use. The spectrum
was defined to ensure a protection of the sensitive systems, more particularly
the GPS (1.2–1.5 GHz), and the bands dedicated to civil aviation.
-35
UWB EIRP emission level (dBm.MHz-1)
7.25 10.25 FCC (indoor)

-40
CEPT
-45
MIC
1.99 (proposal)
-50
3.1 10.6 Singapore
-55
(UWB
-60 Friendly
Zone)
-65
GPS
2.7 3.8
-70 band
6 8.5
-75
0.96 1.61
1 10 Frequency (GHz)
Figure 1.4. UWB systems emission masks
1.5.2. Regulation in Europe
In Europe, the European Telecommunications Standards Institute (ETSI)

has operated since 2001 on the attribution of the frequencies in Europe for UWB
systems. The studies are carried out in close cooperation with group SE24 of
the Conférence Européenne des Postes et Télécommunications (CEPT), which
more particularly analyzes the possible impact of the UWB systems on the
existing ones [POR 03b]. Normally, these European authorities aim to reach a
given accord for all the European Union States, but each national regulatory
agency keeps the right to manage its radio spectrum independently.
Compared to the American regulation, a more restrictive position was

adopted by the CEPT. In a decision presented in March 2006 [ECC 06],
the regulatory organization Electronic Communication Committee (ECC)
proposed a spectral mask limiting the emission of UWB signals to the
6–8.5 GHz band with a power spectral density of −41.3 dBm.MHz−1 , are
shown in Figure 1.4. The emission is also authorized in the band 3.8–6 GHz
with a power spectral density of −70 dBm.MHz−1 . However, we should
note that such a power limitation does not make it possible to carry out
reliable communication systems for a distance of about one meter. In
order to encourage a fast emergence of UWB systems in Europe, the ECC
currently considers the possibility of using mitigation techniques to ensure the
compatibility of UWB systems with the other radio services in the 3.1–4.8 GHz
band. Among these mitigation mechanisms, we can find the detection and
avoidance (DAA) systems which allow us to avoid bands already used by other
systems and the use of low duty cycle (LDC). Hence, all UWB systems using
satisfactory mitigation mechanisms could be authorized to emit with a power
spectral density of −41.3 dBm.MHz−1 in the 3.1–4.8 GHz band. Temporarily
– until June 2010 – the ECC authorizes the emission in the 4.2–4.8 GHz band
with a power spectral density of −41.3 dBm.MHz−1 .
1.5.3. Regulation in Asia
In Asia, the regulation of UWB is particularly well advanced in Japan and

Singapore.
In Japan, from September 2002, the Information and Technology

Communication Sub-Council working group presented its first investigations
on UWB technology at the ministry for telecommunications, in order to
prepare the regulation of UWB. Moreover, the Communications Research
Laboratory (CRL) is developing a project with many industrial partners
in order to design marketable UWB systems. However, the emission mask
proposed by their ministry of internal affairs and communications, MIC,
remains more restricted than the American mask. As illustrated in Figure 1.4,
a preliminary proposal presented in August 2005 suggests a limitation of
UWB emissions to the 7.25–10.25 GHz frequency bands with a power spectral
density of −41.3 dBm.MHz−1 .
At the beginning of 2003, the Singaporean regulatory agency named

Infocom Development Authority (IDA) created a UWB research area,
called UWB Friendly Zone, which makes it possible to deploy tests and
demonstrators in Singapore with experiments using an emission power of
about 10 dB above the FCC limit and a band spreading from 2 GHz to
10 GHz [POR 03b]. With that, the IDA tries to give an significant advance
in Singapore to new telecommunication technologies, in order to remain
scientifically and economically competitive.
To conclude, it should be specified that the greatest constraint of regulation

comes from management of the interference. The problems of UWB regulation
do not come from the effect of only one UWB system, but from the aggregation
of hundreds of devices using this technology, creating a sum of signals which
can possibly interfere with other systems, like the navigation or safety systems.
So, the UWB scientific community currently works to test and define systems
which remain inoffensive, even when using several colocalized equipments.
1.6. UWB communication system and standardization
The emission mask of UWB radio signals established by the FCC allows
the use of various signals. Figure 1.5 presents various solutions which can
be considered. For each approach, the used frequency band as well as the
emission mask of the FCC are presented in the left-hand graph. In each case,
the right-hand graph presents the time domain signal corresponding to the
band represented in solid line. As we can observe, the duration of the obtained
impulse is inversely proportional to the bandwidth used.
The mono-band approach consists of using all the frequency band available.
It is characterized by very short impulses, therefore resistant to the effects of
superimposed multipaths, and the signals can be created from an arbitrary
impulse modeled by an adequate filter. However, this approach allows little
flexibility in the use of the radio spectrum, and requires us to use very powerful
radio frequency (RF) components.
Another solution consists of dividing the allocated UWB spectrum into

two parts: it is the dual bands approach. It makes it possible to use cheaper
integrated circuit technologies, especially for the low band (typically 3–6 GHz),
the high band being used according to the development of new RF components.
The flexibility of the spectrum radio remains moderate, but this solution allows
it to avoid an arbitrarily sensitive band, as the UNII band around 5 GHz for
example.
Finally, the multiband approach consists of using frequency bands of

minimal width about 500 MHz. This solution has a very great flexibility
for radio spectrum management. For example, if the emission mask is
more restricted in a given country, we can avoid the subbands which are
not authorized. The management of the multi-user communications is also
simplified, as many frequency combinations or temporal multiplexings are
possible.
Many modulation techniques were developed from these various UWB

signals. Historically, the first form of modulation suggested for UWB was the
impulse radio [SCH 93]. Regarding standardization, the American institute
IEEE began the work of defining a high data rate communication system
using the UWB spectrum in 2002. The choice of the type of modulation for
this system was the subject of a procedure of very strict selection, within the
802.15.3a work group. The debate for a single solution was mainly articulated
around two proposals: Direct Sequence UWB (DS-UWB), proposed by the
consortium UWB Forum [FIS 04] and MultiBand Orthogonal Frequency
Division Multiplexing (MB-OFDM), proposed by the consortium MultiBand
OFDM Alliance/WiMedia Alliance [BAT 04b]. In January 2006, as no
consensus could be found for a single solution, the IEEE 802.15.3a group
was disbanded. Today, only one industrial standard exists concerning UWB
systems: the standard ECMA-368 [ECM 05], developed by the consortium
WiMedia Alliance and based on MB-OFDM.
1.6.1. Impulse radio
The impulse radio concept, developed from radar studies, corresponds to the
emission of impulses of very short duration (around 100 ps to 1 ns). Typically,
this type of impulses occupies a very broad spectrum (about 1 to several
GHz). It is thus a mono-band approach. The impulse signals generally adopted
for UWB communications include the Gaussian impulse, its first derivative
(Gaussian monocycle), and its second derivative, as represented in Figure 1.6.
The drawback of the Gaussian impulse lies in its non-zero mean value, which
corresponds in the frequency domain to an important continuous component.
Thus, the Gaussian impulse cannot be propagated without deformation, and
we generally prefer the Gaussian monocycle [BAT 03].
1.6.1.1. Pulse position modulation

Relation [1.3] gives the typical expression of a transmitted impulse radio
signal, using a pulse position modulation (PPM) [SCH 97a]:

s(t) = w t − j · Tf − cj · Tc − Δ · d j [1.3]
Ns
j
where w(t) represents the waveform of a transmitted monocycle, normally

starting from t = 0.
Thus, the transmitted signal corresponds to a sequence of impulses

transmitted at different times, the j th impulse being transmitted at
j · Tf + cj · Tc + Δ · d j .
Ns
The term j · Tf allows a uniform spacing of the impulses. Indeed, the signal
defined by:

s(t) = w t − j · Tf [1.4]
j
(a )
-40 1
density (dBm.MHz -1 )
Normalized signal
-50 0.5
Power spectral
-60 0
-70 -0.5
-80 -1
0 5 10 15 0 0.1 0.2 0.3 0.4
(b)
-40 1
Normalized signal
-50 0.5
Power spectral
-60 0
-70 -0.5
-80 -1
0 5 10 15 0 0.5 1 1.5 2
(c)
-40 1
Normalized signal
-50 0.5
Power spectral
-60 0
-70 -0.5
-80 -1
0 5 10 15 0 2 4
Frequency (GHz) Time (ns)
Figure 1.5. UWB spectrum and signals: mono-band approach (a), dual-band
approach (b) and multiband approach (c)
1
Gaussian
pulse
0.5 Gaussian
Normalized signal
monocycle
Second
0
derivative
-0.5
-1
0 0.2 0.4 0.6 0.8 1
Time (ns)
Figure 1.6. UWB impulse waveforms
corresponds to a sequence of impulses uniformly distributed, with a pulse

spacing equal to Tf seconds.
The impulse sequence of relation [1.4] is represented in Figure 1.7(a). Tf is

generally called “frame duration” and is about 100 to 1000 times the impulse
duration. This makes it possible to obtain signals with a low duty cycle, and
therefore with a low power spectral density. However, we should note that the
periodicity of this signal generates parasitic spikes in the radio spectrum. In
addition, impulses cyclically transmitted by the users of the network are very
sensitive to collision when the signals access the channel.
These two problems are solved by the use of a temporal pseudo-random

hopping code. The frame of duration Tf is subdivided in a given number
of time intervals (chip) of duration Tc . Each user is provided with a
pseudo-random code {cj } of length Ns which indicates in which chip of each
frame the impulse must be transmitted. The use of a pseudo-random code
makes it possible to decrease the effect of appearance of spikes due to the
periodicity of the frame, and the spectrum appears much more smoothed
[PEZ 03]. If the pseudo-random sequence is sufficiently long, the UWB
signal can be compared, on the occupied band, to a Gaussian white noise.
In addition, the pseudo-random code allows the management of multi-users
on the radio channel. The transmitted signal with the considered code is
expressed by:

s(t) = w t − j · Tf − cj · Tc [1.5]
j
and its shape is given by Figure 1.7(b).

Tf
(a)
t
Tc
(b)
t
(c)
t
Figure 1.7. Sequences of radio impulses: (a) sequence of uniformly distributed

impulses, (b) spreading code at position 0, and (c) spreading code at position 1
In relation [1.3], the modulation used to transmit data is PPM. Indeed, the
term Δ represents a time interval of about T2c , and the dk terms are the 0 and
1 symbols to be transmitted. The index d j , where · indicates the integer
Ns
part, shows that the same symbol is used over the whole length of the code.
There is thus a redundancy of information, which makes it possible to increase
the processing gain. Under these conditions, when zero is transmitted, there
is no temporal shift in the emission of the data, while a duration shift of Δ
is applied over the entire code duration when one is transmitted. These two
states are shown in Figure 1.7(b) and (c). We can note that PPM modulations
using a greater number of states are possible.
Concerning impulse radio, the signals reception is done by correlation. The

basic idea is to multiply the received signal by a model signal, which makes it
possible to demodulate the transmitted data. So, the model signal corresponds
to a given pseudo-random emitted code. Thus, the use of time codes allows the
management of multi-access [SCH 93]. If several users emit simultaneously by
using orthogonal pseudo-random codes, only the signal corresponding to the
selected code will be demodulated, the other users will appear as noise. This
scheme corresponds to the traditional code multiple division access (CDMA).
1.6.1.2. Pulse amplitude modulation
The pulse amplitude modulation (PAM) is an alternative to the pulse

position modulation. This technique consists of varying the amplitude of the
transmitted impulses to code the different states.
Theoretically, an unlimited number of different values can be used for the

signal amplitude. In practice, the PAM modulation is often reduced to two
states, 1 and -1. Under these conditions, 2-PAM modulation may be regarded
as a form of binary phase shift keying (BPSK). This BPSK modulation has a
good robustness to the effects of the channel and simplifies the synchronization.
Indeed, the position of the impulse remains fixed, and it is only its phase which
varies.
Another alternative of the PAM modulation consists of transmitting two

states: 1 and 0. This corresponds to an on-off keying (OOK) modulation. In each
defined transmission time-slot, an impulse is emitted to code 1, and nothing is
emitted to code 0.
Finally, we may also consider hybrid modulations. We can for example

create a modulation of 512 states by using a combination of 256-PPM with
a modulation 2-PAM.
The impulse radio modulation technique was implemented by the American

company Time Domain, and was marketed in an electronic chip, named PulsOn
210. However, because of the difficulty of implementation and the low spectral
flexibility of this type of system, the standardization authorities have chosen
other types of modulations, presented in the following subsections.
1.6.2. Direct sequence UWB

The direct sequence ultra-wideband (DS-UWB) modulation is a solution
proposed by the industrial group UWB Forum [FIS 04]. This modulation
uses the frequency band allocated to UWB under two dual bands covering
respectively 3.1–4.85 GHz and 6.2–9.7 GHz. This configuration make it
possible to avoid the UNII band at 5 GHz used by WiFi systems. On these
dual bands, the transmitted impulses have a time duration of about 0.3–0.5 ns
and present several cycles (see Figure 1.5(b)). In a first step, only the low
band is used, in order to simplify the architecture of the radio transmission
systems.
As with the PPM modulation, the DS-UWB modulation uses a frame

divided in chips. However, an impulse can be transmitted in each chip of
the frame. So, the signal is transmitted continuously and there is no low
duty cycle as in the case of the impulse radio. The transmitted symbols are
represented by ternary spreading codes of one frame length, composed of a
series of 1, 0 or −1. According to the specifications proposed by the UWB
forum, all the DS-UWB systems will be able to provide these codes using a
BPSK baseband modulation. Optionally, a more efficient modulation, named
M-BOK, is defined in order to ensure a more robust transmission. The length
of the spreading code varies from 1–24 chips according to the considered
data rate. The transmission data rate proposed for the DS-UWB systems
range from 28–1,320 Mbps. The management of multiple users, clustered in

sub-networks called piconets, is performed by the use of orthogonal codes.
In this prospect, the DS-UWB modulation is similar to the CDMA system
used in UMTS. Finally, we can note that the isolation of users belonging to
different piconets is improved by the use of slightly different chip frequencies
into each piconet. More details on this technology can be found in [WEL 03].
Compared to impulse radio, the DS-UWB systems seem easier to

implement, because the considered frequency bands are narrower, which
imposes less constraints on RF components. The used modulation still being
based on impulses, this radio access technique is robust against the multipath
effects of the channel. In terms of regulation, the separation into two dual
bands makes it possible to protect the sensitive frequency bands, but the
transmitted spectrum is not very flexible. In terms of standardization, we may
notice that following the disbanding of the IEEE 802.15.3a working group,
no official norm is currently based on the DS-UWB modulation. However,
some industrial companies are developing this technical solution, starting
from the specifications proposed by the UWB Forum group. As an example,
the company Freescale Semiconductor, founder of the UWB Forum group,
implemented the DS-UWB modulation in the component Freescale XS110.
1.6.3. Multiband OFDM
Multiband orthogonal frequency division multiplexing (MB-OFDM) is a

solution based upon the UWB spectrum proposed in standardization and used
by the industrial groups MBOA and WiMedia Alliance [BAT 04b]. It is a
multiband approach, where the FCC spectrum is subdivided into 14 subbands
of 528 MHz. Figure 1.8 shows these subbands which are divided into different
groups. First of all, the industrial companies focussed their development efforts
on group 1 (3.1–4.9 GHz). The other groups should be quickly exploited in order
to produce systems in conformity with the European and Asian regulation.
In each subband, an OFDM modulation is applied, the signal being

distributed over 100 carriers with narrow bandwidth. For each carrier, the
baseband modulation is either BPSK or QPSK. This configuration facilitates a
very flexible management of the radio spectrum. Indeed, to avoid the jamming
of any frequency band, we can forbid a series of carriers or the totality of
a subband. This management can possibly be performed according to the
legislation of the country of use or in a dynamic way according to the potential
jammers. The management of the multi-users of the same group of subbands
is performed using time-frequency code techniques. In a group of subbands,
the communication of a given user regularly hops from one band to another,
according to a cycle of approximately 1 μs duration. The hopping pattern
from one band to another corresponds to the considered time-frequency code,

which is unique to each user. The data rates offered by this technology extend
from 53.3–480 Mbps. References [BAT 04a, BAT 04b] give all the necessary
details for the implementation of this system.
Group 1 Gr. 2 Gr. 3 Gr. 4 Gr. 5

← →← →← →← →←→
-40
density (dBm.MHz-1)
-50
Power spectral
-60
-70
-80
0 2 4 6 8 10 12
Frequency (GHz)
Figure 1.8. Subbands for the MB-OFDM solution
The advantages of the MB-OFDM radio access technique are mainly related
to its low complexity of implementation, as the OFDM modulation shows a high
degree of maturity. Moreover, it is currently the only UWB communication
technique developed as an international standard, ECMA – 368, available since
December 2005. The restriction on the used frequency band to the first group
of subbands allows it to use existing systems and RF components. However,
in order to achieve an international deployment of the MB-OFDM solution,
it will be necessary to develop systems operating under the European and
Asian frequency regulations. In technical terms, it should be noted that signals
are no longer impulsional, and the technology no longer benefits from the
advantages related to a very wide frequency band, such as the robustness
to the radio channel effects or the possibilities of localization. The integrated
circuit UBLink proposed by the company Wisair is an example of commercially
available MB-OFDM design.
1.7. Conclusion
Initially used for radar localization applications, UWB technology has

been studied for the last 15 years in the field of wireless communications.
The main characteristics of this technology, like the width of the spectrum
and its strong temporal resolution, made it possible for the scientific and
industrial communities to propose a certain number of interesting applications:
high data rate WLAN networks, home automation applications, etc. As
developed in this chapter, various types of UWB modulations were proposed
in standardization, in particular DS-UWB and MB-OFDM. UWB regulation
has been effective in the USA since 2002. In Europe, a mask of emission for
UWB signals was established in 2006, but the coexistence of UWB systems
with other applications is still under study.
As for any radio access technique, an accurate knowledge of the propagation

channel as well as the antennas is predominant for the development of
UWB communication systems. Thus, the following section presents the radio
waves propagation phenomena inside buildings, and describes the principal
parameters for UWB radio channel characterization. In this book, we address
antennas by including them in the channel. A thorough study of the antennas
in the specific case of UWB could be the subject of a specific book. Indeed,
the realization of an antenna presenting good characteristics in terms of
adaptation and radiation on such a broad band is still a challenging task
[SCH 03c, SCH 03b, OSS 03].
Chapter 2
Radio Wave Propagation
2.1. Introduction
The existence of electromagnetic waves was theoretically predicted by
J.C. Maxwell as early as 1855 [MAX 55]. The German physicist Hertz tried to
demonstrate that electromagnetic waves travel at a finite speed, and in 1886
he performed the very first radio propagation experiments. The oscillating
circuit designed by Hertz consisted of two discharging metallic spheres, which
produced an observable spark on an open wire loop [SCH 86]. Interestingly,
the signal used by Hertz consisted of a short duration pulse, which could
hence be regarded as an ultra-wideband signal [AIE 03a].
In order to facilitate the industrial use of radio transmission, a large

amount of research has been conducted to characterize the electromagnetic
wave propagation mechanisms. This research first focused on signals confined
in narrow frequency bands and then extended to wideband signals. This
chapter details the definition of the propagation channel and its mathematical
representation. The characteristic parameters of the radio-mobile channel are
then presented.
2.2. Definition of the propagation channel

By definition, a radio transmission system transforms an emitted electrical
signal e(t) into a received electrical signal s(t) by means of electromagnetic
waves. The propagation channel corresponds to the system converting the
signal e(t) into the signal s(t) and thus accounts for the interactions between
the electromagnetic waves and their environment. At this point, a distinction
should be made between the propagation channel and the transmission
channel. The propagation channel represents the transformation of the

electromagnetic waves throughout their propagation, while the transmission
channel also includes antenna radiation patterns (cf. Figure 2.1). In the
literature, some authors assimilate these two concepts, but the distinction
comes fully into play when considering multiple-input multiple-output
(MIMO) channels [COS 04]. The full study of antennas in the specific
UWB context would require a dedicated book. Indeed, the development of
a UWB antenna presenting adequate characteristics in terms of adaptation
and radiation over such a wide frequency band is still a challenging issue
[SCH 03c, SCH 03b, OSS 03].
Transmission channel
Emitted Received
signal signal
e(t) s(t)
Propagation channel
Figure 2.1. Propagation channel and transmission channel
2.2.1. Free space propagation
Let us first consider an ideal case where the transmission system is placed
in free space, i.e. in an environment where no obstruction is present. Denoting
by GE the emission antenna gain and PE the emitted signal power, the power
density observed at a distance d is given by [PAR 00]:
PE GE
W = [2.1]
4πd2
The power PR collected at the output of a receiving antenna with gain GR

relates to the power density W as follows:
λ2 GR
PR = W AR = W [2.2]
4π
where AR represents the effective area of the receiving antenna, and λ represents
the wavelength at the working frequency.
Radio Wave Propagation 45
Equations [2.1] and [2.2] yield the Friis formula, giving the signal attenuation
in free space:
2
PR c
= GT GR [2.3]
PT 4πf d
where we used the relation c = f λ existing between the wavelength λ, the
frequency f and the propagation speed c.
It should be noted that this relation holds for a distance d large enough for
the receiving antenna to be considered in the far field region with respect to
the transmitting antenna [AFF 00]. A receiver is situated in the far field if the
distance d is larger than the Fraunhofer distance dF , which is related to the
largest dimension of the transmitting antenna D and to the wavelength λ as
follows:1
2D2
dF = [2.4]
λ
Free space propagation is a theoretical or reference situation. In realistic

propagation conditions, the transmitted wave is affected by the system
environment through different mechanisms, which are presented in the
following section.
2.2.2. Multipath propagation
In a realistic environment, signal transmission follows not only the direct

path, but also a number of distinct propagation paths. These paths undergo
various effects depending on the type of interaction between the wave and the
surrounding objects. At the output port of the receiving antenna, the observed
signal corresponds to a combination of different waves, each of them presenting
a different attenuation and a different phase rotation. Moreover, each wave
reaches the receiver with a distinct delay, linked to the length of the propagation
path. Multipath propagation mechanisms may lead to a significant distortion
of the received signal. On the other hand, a direct path, known as the line
of sight (LOS) path, is not always available, which is particularly frequent in
indoor configurations. In this case, so called non-line of sight (NLOS) paths
are necessary to enable efficient radio communication. Figure 2.2 illustrates the
concept of multipath propagation, as well as the main propagation phenomena.
1. It should be noted, though, that for systems operating at different frequencies, the
antenna dimension D is generally adapted to the wavelength. For a wire antenna, the
antenna length is for instance defined as D = λ2 , and hence the Fraunhofer distance
follows dF = λ2 .
Diffuse
reflection
Ȝ
Diffuse
scattering
Transmission
Receiver
Diffraction
Specular
reflection
Transmitter Waveguide
effect
Figure 2.2. Main propagation mechanisms
• Reflection: reflection takes place on obstacles of large dimensions with

respect to wavelength. When two different materials are separated by a plane
surface (i.e. a surface where possible rough spots are small with respect to the
wavelength), the reflection is said to be specular. In this case, the direction
and the amplitude of the reflected ray are governed by the Snell-Descartes
and Fresnel laws. When the surface separating the two materials presents
non-negligible random rough spots, the reflection is called diffuse reflection.
Most of the energy is then directed along the reflected ray, but part of the
energy is diffused in neighbor directions.
• Transmission: when the medium where a reflection takes place is not

perfectly radio-opaque, part of the incident wave travels through the material
following a so-called transmission mechanism. Most of the building materials
used in indoor environments significantly attenuate the transmitted wave. For
a given material, the attenuation and the direction of the transmitted signal
are related to the wavelength, since the refractive index of the material varies
with the frequency. Finally, for layered materials such as plasterboard, multiple
reflection may occur within the material.
• Diffraction: diffraction takes place on the edges of large sized

obstacles with respect to the wavelength. This mechanism accounts for the
electromagnetic field continuum on either side of the optical line of sight. The
calculation of a diffracted field uses Huygens’ principle, which considers every
point of a wavefront as a secondary spherical source. Hence, diffracted waves
are distributed along a geometrical cone, with an angle corresponding to the
incident angle.
• Diffusion: when an electromagnetic wave travels towards a group of

obstacles of small dimensions with respect to the wavelength, the observed
phenomenon corresponds to the superimposition of a large number of random
diffractions. In this case, the behavior of the incident wave is handled in a
statistical way and the resulting phenomenon is called diffusion. In general, the
electromagnetic wave is directed in all directions with a variable attenuation.
This phenomenon is generally observed in outdoor environments, for instance
in the presence of tree foliage. In indoor environments, diffusion may occur on
a group of small objects.
• Waveguide effect: the waveguide effect occurs in indoor environments,

between two corridor walls for instance. Successive reflections on two parallel
obstacles lead to a global wave motion along the guiding direction. This
phenomenon also occurs in urban environments, for instance between two
buildings lining a narrow street.
2.2.3. Propagation channel variations
Owing to the different interactions between the radio waves and their
propagation medium, significant variations of the channel characteristics are
observable at different scales. When isotropic antennas are used in free space,
the resulting power attenuation is proportional to the square of the distance
d between the antennas (cf. section 2.2.1). However, actual observations
present large scale power variations linked to the propagation environment.
In practice, the obstacles within the environment as well as the combination
of multiple propagation paths lead to an additional attenuation of the
transmitted waves. In general, this attenuation is a function of the distance d

between the transmitter and the receiver. It is characterized by the path loss
exponent Nd , the received signal power decreasing proportionally to d−Nd .
The parameter Nd is equal to 2 in free space, and varies between 2 and 5 in
NLOS configurations. In LOS configurations, waveguide effects may lead to a
value of Nd lower than 2. Finally, we may note differences in the order from
1 to several dB between the received power and the mean trend following a
d−Nd law. This is due to occasional obstructions known as shadowing or slow
fading.
Small scale fluctuations are directly linked to multipath propagation

mechanisms. Multiple versions of a single signal presenting different
attenuations and different phase delays may eventually combine at the
receiver, which produces significant fast fluctuations in the order of 10 to
20 dB.
2.2.3.1. Spatial selectivity

Let us consider a signal composed of a single carrier frequency propagating
along two paths, the direct path and a reflected path. If the reflection occurs in
the vicinity of the LOS path, we may consider that these two paths will cause a
similar attenuation. However, depending on the wavelength of the transmitted
signal and on the length difference between the two paths, the two versions
of the signal may arrive in phase or antiphase. This concept is illustrated in
Figure 2.3. In the first case, the signals are adding constructively and some
power gain is observed. In the second case, the signals are adding destructively
and the total received power is strongly attenuated. When the mobile device is
displaced, the phase rotation linked to each path leads to a series of signal peaks
and signal fades, which is called a fast fading signal. When this phenomenon
occurs with a large amount of paths, the received signal may be considered as
a random process.
2.2.3.2. Frequency selectivity

Let us now consider a more realistic signal, spreading over a given
frequency bandwidth. As explained in the previous section, the spatial
selectivity phenomenon depends on the phase difference between multiple
paths. Thus, spatial selectivity also depends on signal frequency. When the
frequency band is narrow, all frequencies forming the transmitted signal
undergo similar phase variations. Hence, the possible power fades are constant
over the whole considered frequency band. This phenomenon is known as flat
fading.
For signals occupying wider frequency bands, the various frequencies

may be affected in a different way, and thus the received signal is somehow
V t1
direct path
t
V t2 reflected path
(case 1)
t
V t2 + ǻt reflected path
(case 2)
t
case 1
V
global signal
t
case 2
Figure 2.3. Constructive and destructive addition of two propagation paths
distorted with respect to the transmitted signal. In this case, we may

observe frequency selective fading, which consists of a variation in the power
received with increasing frequency. The bandwidth over which the frequency
components are affected in a similar way is called the coherence bandwidth or
correlation bandwidth.
In the delay domain, frequency selectivity corresponds to a delay between

the various versions of the signal propagating along different paths. This delay is
in the order of a few nanoseconds in indoor environments, and in the order of a
few microseconds in outdoor environments. Depending on the signal bandwidth,
these echoes may superimpose on each other, which results in significant fades.
For signals presenting a very wide frequency band, and particularly for UWB
signals, the time domain resolution is very high, which limits the possible
interference between different delayed versions of the transmitted signal. In
this case, signal fading is less pronounced. It is then possible to apply advanced
reception techniques, such as channel equalization or RAKE reception,2 in order
to collect as much energy as possible from the multiple propagation paths
[CRA 98, FOE 01b, GAU 03].
2. A RAKE receiver combines the signals from different multipaths, by using several
reception branches in parallel [HAY 01].
Finally, it should be noted that in the case of wideband signals, frequency

selectivity leads to a time domain spreading of the transmitted signal.
Characterizing this spreading is essential to calibrate communication systems
and mitigate intersymbol interference.3
2.2.3.3. Doppler effect

The spatial selectivity phenomenon demonstrates that the properties of
the radio propagation channel may vary significantly when the receiving
antenna is positioned at different locations. It is thus questionable how the
propagation channel is affected when the transmitting antenna, the receiving
antenna or even the environment are moving. The Doppler effect corresponds
to an observed shift in the frequency of an electromagnetic signal due to the
variation of its propagation path. As a simple example, let us consider a
mobile receiver moving at a speed v and receiving a radio signal as a plane
wave arriving with an angle α with respect to the mobile direction. In this
case, the observed Doppler shift is [PAR 00]:
v
ν = f cos(α) [2.5]
c
where f is the signal frequency and c is the wave propagation speed.
The statistical case where the ray distribution is represented by the

probability density function (PDF) of their angle of arrival pα (α) was
developed in [CLA 68]. For a simple case where the handset moves at a
constant speed, the PDF of the Doppler shift pν (ν) is given by:

1 ν ν
pν (ν) = ν 2 · pα − arccos νmax +pα arccos νmax [2.6]
1− νmax
where νmax is the maximum Doppler shift given by:
v
νmax = f [2.7]
c
This theoretical PDF of the Doppler shift may be linked to the Doppler
spectrum of a measured signal. In particular, for a uniform distribution of the
arrival angles, the PDF of the Doppler shift is proportional to the Jakes Doppler
spectrum [JAK 93]:
1
pν (ν) = ν 2 [2.8]
π 1 − νmax
3. Intersymbol interference consists of the corruption of a transmitted symbol by the

previous symbols due to the time domain spreading generated by the propagation
channel. It is one of the main sources for the degradation of the transmission quality.
In what follows, a distinction will be made between the concepts of temporal

and spatial variations. Spatial channel variations are due to the displacement
of at least one of the antennas in an otherwise static environment. Temporal
variations are linked to a change in the close environment of a fixed radio link
[HAS 94a]. In the radio-mobile context, both types of variation are generally
observed, but one or the other may be considered as predominant, depending
on the situation. In the indoor environment, temporal variations are mainly
due to moving people.
In both cases of spatial and temporal variations, propagation paths between

the transmitter and the receiver may appear, disappear or undergo successive
transformations. In the case of a stationary channel, only the existing paths
are considered, and the channel variations consist of an evolution of the path
length, which may be positive or negative. Hence, both spatial and temporal
channel variations lead to an observed Doppler effect.
2.3. Propagation channel representation
2.3.1. Mathematical formulation
Owing to the multipath propagation phenomena, the received signal s(t)

is composed of a number of superimposed replicas of the transmitted signal
e(t), each of them presenting a different delay and a different attenuation. The
propagation channel thus behaves like a linear filter. Hence, the propagation
channel may be represented by its impulse response (IR) h(τ ), corresponding
to the output of this filter when the excitation is a Dirac function. The received
signal may thus be written as:
∞
s(t) = e(t − τ )h(τ )dτ [2.9]
−∞
In most radio systems, the transmitted signal spreads over a frequency band
which is not centered around zero. A given signal x(t) may thus be represented
by its complex envelope γx (t) defined as:

x(t) = γx (t)ej2πf0 t [2.10]
where {·} represents the real part of a complex number and f0 represents
a given frequency in the considered band. The complex envelope γx (t) is also
named the equivalent baseband term of x(t).
It can be shown that there are two ways of defining the baseband filter
heq (τ ) equivalent to the passband filter h(τ ) (cf. Appendix A):
1
heq1 (τ ) = γh (τ )
2 [2.11]
−j2πf0 τ
heq2 (τ ) = h(τ )e
The three representations in Figure 2.4 are thus equivalent to observing

the channel effect on the transmitted signal. In the remainder of this book, we
will use, unless otherwise stated, the channel impulse response in its complex
envelope form γh (τ ), which will be denoted h(τ ) for the sake of simplicity.
e(t) s(t)
- h(τ ) -
γe (t) γs (t)
- heq1 (τ ) = 1 -
2 γh (τ )
γe (t) γs (t)
- heq2 (τ ) = h(τ )e −j2πf0 τ -
Figure 2.4. Equivalent representations of a static radio-mobile channel
2.3.2. Characterization of deterministic channels
The representation of the radio channel in the form of an IR h(τ ) is valid for
static channels only. In practice, the environment or the antenna position may
be modified, and hence the radio channel may vary with time. The IR h(t, τ )
is thus dependent on both time and delay. The inputs and outputs of a linear
filter may be described in the time domain or in the frequency domain. This
leads to a set of four transfer functions that may be used to describe the radio
channel [BEL 63]. Figure 2.5 illustrates the relations between these functions.
time-delay
h(t, τ ) @
I
@
F −1 @ @ F −1
F F@ @
@ @
delay-Doppler @ @ frequency-time
R @
@
S(τ, ν) T (f, t)
@ I
@
@ @
@ @ F −1 F −1
F@ @ F
@ @
@R
@ H(f, ν)
frequency-Doppler
Figure 2.5. Characteristic functions for a deterministic channel. Arrows represent a

Fourier transform (F ) or an inverse Fourier transform (F −1 )
2.3.2.1. The time varying impulse response

The function h(t, τ ) is called the time-variant impulse response. It relates
the received signal s(t) to the transmitted signal e(t) according to the following
filtering operation:
∞
s(t) = e(t − τ )h(t, τ )dτ [2.12]
−∞
The magnitude of this impulse response may be observed to distinguish

between different signal echoes according to their propagation delay. Equation
[2.12] thus provides a physical representation of the channel as a continuum
of scatterers that are fixed – since their delay is constant – and scintillating –
which corresponds to the channel’s temporal evolution.
2.3.2.2. The frequency domain function

The function H(f, ν) is also called the output Doppler-spread function and
reflects the Doppler shift phenomenon caused by the channel. It is the dual
function of the function h(t, τ ) in the frequency-Doppler shift space. It thus
relates the received signal spectrum S(f ) to the transmitted signal spectrum
E(f ) as follows:
∞
S(f ) = E(f − ν)H(f − ν, ν)dν [2.13]
−∞
Using this representation, the output signal spectrum S(f ) is considered

as the sum of superimposed input signal spectrum replicas E(f ), each replica
being Doppler shifted and filtered.
2.3.2.3. The time varying transfer function

Another approach to characterizing the radio channel consists of describing
the relation between the time domain output signal s(t) to the input signal
spectrum E(f ), using the time varying transfer function T (f, t):
∞
s(t) = E(f )T (f, t)ej2πf t df [2.14]
−∞
The function T (f, t) is related to the functions h(t, τ ) and H(f, ν) through a
simple Fourier transform. If the bandwidth of the considered channel is narrow
enough, the time-variant transfer function may be directly measured using a
network analyzer.
2.3.2.4. The delay-Doppler spread function

A final approach consists of representing the radio channel in the
delay-Doppler shift space. The corresponding function allows us to
simultaneously observe the channel dispersion in both the time domain and
the frequency domain. For this reason it is referred to as the delay-Doppler
spread function. The function S(τ, ν) relates the output signal s(t) to the
input signal e(t) through the following relation:
∞ ∞
s(t) = e(t − τ )S(τ, ν)ej2πνt dνdτ [2.15]
−∞ −∞
Equation [2.15] represents the output signal s(t) as a sum of superimposed

replicas the input signal e(t), each replica being delayed and Doppler-shifted.
The function S(τ, ν) is related to the functions h(t, τ ) and H(f, ν) through a
simple Fourier transform.
2.3.3. Characterization of linear random channels

In practical situations, the propagation channel fluctuations are due to a
number of superimposed phenomena, which cannot be measured individually.
The radio channel variations may thus be regarded as a random process and it
is no longer valid to describe them deterministically. The propagation channel is
thus characterized in a statistical way. In practice, the statistical description of
the channel is limited to the second order, and we consider the autocorrelation
functions of the channel system functions only. These functions are defined as
follows:

Rh (t, u; τ, η) = E h(t, τ )h∗ (u, η)

RH (f, m; ν, μ) = E H(f, ν)H ∗ (m, μ)
[2.16]
RT (f, m; t, u) = E T (f, t)T ∗ (m, u)

RS (τ, η; ν, μ) = E S(τ, ν)S ∗ (η, μ)
where E[·] represents the mathematical expectation and (·)∗ represents the
complex conjugation operation.
These four autocorrelation functions are linked through double Fourier

transforms in a similar scheme to the one linking the channel system functions
(cf. Figure 2.5). The second order moments of the input and output signals
are then given by:
∞ ∞
Rs (t, u) = e(t − τ )e∗ (u − η)Rh (t, u; τ, η)dτ dη
−∞ −∞
[2.17]
∞
Res (t, u) = Re (t, u − η)h(u, η)dη
−∞
The first and second order moments are sufficient to completely describe
the output signal s(t) in the case of a Gaussian signal.
2.3.4. Channel classification
The representation of the random radio channel may be simplified by

considering different assumptions about the channel characteristics.
2.3.4.1. Wide sense stationary channels

A channel is considered as wide sense stationary (WSS) if its temporal (or
spatial) variation meets the statistical conditions of second order stationarity.
In other words, the expectation of the channel impulse response needs to be
invariant with time, and its autocorrelation Rh (t, u; τ, η) will depend on the
variables t and u through the difference ξ = t − u only. In practice, these

conditions mean that the channel fluctuation statistics are constant over a
short time interval ξ, which is a reasonable assumption for traditional channels
(e.g. indoor or urban environment). Under these conditions, the autocorrelation
functions of the time-variant impulse response and the time-variant transfer
function may be written as:
Rh (t, t + ξ; τ, η) = Rh (ξ; τ, η)
[2.18]
RT (f, m; t, t + ξ) = RT (f, m; ξ)
Considering the double Fourier transform relationship between Rh (t, u; τ, η)

and RS (τ, η; ν, μ), and using ξ = t − u, this yields:
∞ ∞
RS (τ, η; ν, μ) = Rh (t, u; τ, η)ej2π(uμ−tν) dtdu
−∞ −∞
∞
[2.19]
= δ(ν − μ) Rh (ξ; τ, η)e−j2π(ξν) dξ
−∞
= δ(ν − μ)PS (τ, η; ν)
where the term PS (τ, η; ν) may be identified as a power spectral density

obtained from Rh (ξ; τ, η) by applying the Wiener-Kinchine theorem.
This last relation indicates that the spectral content of the signal is
uncorrelated for different Doppler shifts. Physically, this means that echoes
generating different Doppler shifts are uncorrelated. Similarly, we may write
the autocorrelation of the frequency domain function as:
RH (f, m; ν, μ) = δ(ν − μ)PH (f, m; ν) [2.20]
2.3.4.2. Uncorrelated scattering channels

Under the uncorrelated scattering (US) assumption, we consider that the
contributions from elemental scatterers corresponding to different delays are
uncorrelated. This condition is equivalent to an assumption of second order
stationarity in the frequency domain. As in the WSS case, we may simplify the
autocorrelation functions of the frequency domain function and the time-variant
transfer function as follows (and noting by Ω the frequency difference m − f ):
RH (f, f + Ω; ν, μ) = RH (Ω; ν, μ)
[2.21]
RT (f, f + Ω; t, u) = RT (Ω; t, u)
The time-variant impulse response and the delay-Doppler spread function

may be re-written in the form of power spectral densities:
Rh (t, u; τ, η) = δ(η − τ )Ph (t, u; τ )

[2.22]
RS (τ, η; ν, μ) = δ(η − τ )PS (τ ; ν, μ)
2.3.4.3. Wide sense stationary uncorrelated scattering channels

A wide sense stationary uncorrelated scattering (WSSUS) channel follows
both the WSS and US assumptions. It corresponds to the simplest class of
channels, which presents an uncorrelated spread in both the delay domain and
the Doppler shift domain. In this case, the autocorrelation functions of the four
channel system functions may be simplified as follows:
Rh (t, t + ξ; τ, η) = δ(η − τ )Ph (ξ; τ )
RH (f, f + Ω; ν, μ) = δ(ν − μ)PH (Ω; ν)

[2.23]
RT (f, f + Ω; t, t + ξ) = RT (Ω; ξ)
RS (τ, η; ν, μ) = δ(η − τ )δ(ν − μ)PS (τ ; ν)
Figure 2.6 illustrates the simple Fourier transform relationships linking the
four autocorrelation functions for a WSSUS channel.
Two of the obtained functions are of particular interest. The function

PS (τ ; ν) is called a scattering function. It physically represents the Doppler
spectrum of radio wave paths as a function of their propagation delay.
For ergodic signals, the function Ph (ξ; τ ) may be written as:

Ph (ξ; τ ) = E h(t + ξ, τ )h∗ (t, τ )
T [2.24]
1 2
= lim h(t + ξ, τ )h∗ (t, τ )dt
T →∞ T − T2
Hence, the term Ph (0, τ ) corresponds to the temporal average of the impulse
response power. This function is called the power delay profile (PDP).
In practice, it is generally assumed that observed channels fall under the

WSSUS assumption. For measured channels, it is sometimes necessary to
time-delay
Ph (ξ, τ ) @
I
@
F −1 @ @ F −1
F F@ @
@ @
delay-Doppler @ @ frequency-time
R @
@
PS (τ, ν) RT (Ω, ξ)
@ I
@
@ @
@ @ F −1 F −1
F@ @ F
@ @
@R
@ PH (Ω, ν)
Doppler-frequency
Figure 2.6. Autocorrelation functions of a WSSUS random channel
confirm this postulation. A thorough discussion about the validation of the

WSSUS assumption for experimental measurements has been published in
[BUL 02]. In particular, the RUN method is based on assessments of the
process variance over successive sub-intervals [BEN 66].
2.4. Channel characteristic parameters
In order to evaluate the characteristics of a propagation channel, a set

of measured impulse responses need to be analyzed. This section presents
a number of parameters describing different aspects of the radio channel.
In general, their computation requires the WSSUS assumption. As it is not
practically feasible to collect a set of statistical channel realizations, we
generally assume that the channel system functions are ergodic. Hence, it
is for instance possible to approximate the statistical expectation using a
temporal (or spatial) average over a set of successive measurements.
2.4.1. Frequency selectivity
Frequency selectivity is characterized by the presence of multiple paths,

particularly observable on the channel’s impulse response. The power
repartition as a function of the delay is given by the PDP, defined by equation
[2.24]. In order to mitigate the local effects of fast channel fading, the PDP is
calculated from a set of M impulse responses, successively measured over a
path length in the order of 20–40 wavelengths [LEE 85]. This calculation is
valid under the stationarity assumption. The following equation presents the
PDP calculation:
1 2
M
Ph (0, τ ) = h tm , τ [2.25]
M m=1
2.4.1.1. RMS delay spread
The root mean squared (RMS) delay spread τRMS , sometimes simply
referred to as delay spread, represents the standard deviation of the PDP. It is
calculated as follows:

∞ 2
τ − τm Ph (0, τ )dτ
τRMS = −∞
∞ [2.26]
P (0, τ )dτ
−∞ h
where τm is the mean delay given by:

∞
τ Ph (0, τ )dτ
τm = −∞
∞ [2.27]
−∞
Ph (0, τ )dτ
2.4.1.2. Coherence bandwidth
The RMS delay spread is a significant parameter for the analysis of the
intersymbol interference. It is also closely linked to the correlation between the
different frequencies of the signal spectrum. In order to quantify this frequency
dependence, the n% coherence bandwidth is defined from the autocorrelation
of the channel transfer function RT (Ω, ξ):

RT (Ω, 0)
Bc,n%
= min Ω : = n [2.28]
RT (0, 0) 100
The function RT (Ω, 0) is referred to as the frequency correlation function

and is obtained from the PDP Ph (0, τ ) by a simple Fourier transform. The
coherence bandwidth is thus the frequency lag above which the frequency
autocorrelation function crosses a given threshold. Thresholds generally given
in the literature are 90% and 50% [BAR 95].
Relative magnitude (dB)
q% of the total energy
Noise level
Ĳ0 Ĳ1 Ĳ2 Ĳ3 Delay (ns)
Figure 2.7. Power delay profile defining the delay window
2.4.1.3. Delay window and delay interval

Two other parameters are also used to give a more precise idea of the PDP
spread [COS 89]. The q% delay window is the duration of the central part of
the PDP containing q% of the total energy. Referring to the delays represented
in Figure 2.7, the delay window is given by:

Wq% = τ2 − τ1 q% [2.29]
The delays τ1 and τ2 are defined by:

τ2 τ3
q
Ph (0, τ )dτ = Ph (0, τ )dτ [2.30]
τ1 100 τ0
and the energy which is outside the window is split into two equal parts. The
delays τ0 and τ3 are the delays at which the PDP first rises and finally falls
through the signal noise level.
The p dB delay interval refers to a threshold positioned at p dB below the

maximum of the PDP. It simply corresponds to the interval from the first delay
at which the PDP exceeds this threshold to the last delay at which it falls below
this threshold.
2.4.1.4. Exponential decay constants

A number of analyses of the UWB radio propagation channel agree on
a representation of the impulse response in the form of a discrete sum of
individual contributions. Each contribution, known as ray, corresponds to a
propagation path and is characterized by its distinct delay and amplitude.
This representation is widely used for wideband channels [HAS 93]. In order to
account for the possible presence of ray clusters, this discrete impulse response
is represented using the formalism proposed by Saleh and Valenzuela [SAL 87]:

L(t) Kl (t)

h(t, τ ) = βk,l (t)ejθk,l (t) δ τ − Tl (t) − τk,l (t) [2.31]
l=1 k=1
where L(t) is the number of clusters, Kl (t) is the number of rays in the lth
cluster, and Tl (t) is the arrival time of the lth cluster. Parameters βk,l (t),
θk,l (t) and τk,l (t) respectively represent the magnitude, the phase and the
arrival time associated with the kth ray within the lth cluster. Theoretically,
all these parameters vary with time. Defining the PDP from a set of
measurements according to equation [2.25], and assuming that the delay of
each ray is approximately constant during the measurements, the Saleh and
Valenzuela formalism yields the following formula:

L
Kl

Ph (0, τ ) = 2
βk,l δ τ − Tl − τk,l [2.32]
l=1 k=1
As shown in Figure 2.8, the magnitude of the PDP rays generally follows
a decay which is close to an exponential function. This exponential decay may
be observed at both the cluster level and the ray level within a single cluster.
The inter- and intra-cluster exponential decay constants, respectively denoted
Γ and γ, are thus defined so that the magnitude of the rays obeys the following
rule:
Tl −T1 τk,l
2
βk,l 2
= β1.1 e− Γ e− γ [2.33]
These constants may be retrieved from measurements by applying a linear

least squares fitting procedure on the PDP represented in dB.
2.4.1.5. Cluster and ray arrival rates

Without more accurate knowledge about the environment, it is generally
assumed that ray arrivals correspond to independent events, and that the
number of these events is dependent on the observation duration only. By
Ĳk1
Power (mW)
Cluster
e-Ĳ/ī
Ĳk2
Ĳk3
e-Ĳ/Ȗ
e-Ĳ/Ȗ
e-Ĳ/Ȗ
T1 T2 T3 Delay (ns)
Figure 2.8. Power delay profile following the Saleh and Valenzuela formalism
definition, the ray – or cluster – arrivals may reasonably be regarded as a

Poisson process. In the Saleh and Valenzuela formalism, it is thus proposed that
the arrival probability for a new cluster or a new ray follows an exponential
law, with respective parameters Λ and λ. Parameters Λ and λ are respectively
called cluster and ray arrival rates, and the numbers Λ1 and λ1 represent the
mean duration between two successive clusters or rays. These parameters are
assessed by studying the inter-cluster and inter-ray duration distributions.
2.4.2. Propagation loss
When expressed in dB, equation [2.3] corresponding to the Friis formula for
free space propagation is written:

4πf d
P L(f, d) = 20 log − GT (f ) − GR (f ) [2.34]
c
where P L(f, d) represents the ratio between the transmitted power and the
received power.
The slow variations of the propagation channel are mainly due to

propagation loss and shadowing effect. For a practical channel, in order to
characterize the double frequency and distance dependence of the path loss,
the parameter P L(f, d) is approximated by the following formula:

f d
P L(f, d) = P L f0 , d0 + 10Nf log + 10Nd log + S(f, d) [2.35]
f0 d0
where f0 and d0 correspond to an arbitrary frequency and an arbitrary
distance.4 Nf and Nd are called frequency and distance dependent path loss
exponents. Parameter Nd accounts for the interactions between the radio wave
and the environment, such as the transmission phenomenon or the waveguide
effect, and can significantly differ from the theoretical value Nd = 2. Parameter
Nf accounts for the frequency dependence of the propagation phenomena,
and also includes the variations of the effective area for an ideal isotropic
antenna. By considering the propagation channel without antennas in a LOS
situation, this parameter should be close to its theoretical value Nf = 2. Some
authors consider antennas as a part of the propagation channel; in this case,
additional variations of the parameter Nf may be observed, which is linked to
the measurement antenna gain. Finally, the term S(f, d) corresponds to the
slow variations of the propagation channel. As the parameters Nf and Nd are
calculated by linear regression, the parameter S(f, d) presents a zero average.
When expressed in dB, this variable is generally considered as Gaussian, and
it is characterized by its standard deviation σS .
In order to calculate the parameter Nf from measurements, we introduce the

power transfer function PT (f ), equivalent to the PDP Ph (0, τ ) in the frequency
domain. This takes the average received power as a function of frequency for a
number of locally measured transfer functions:
1 2
M
PT (f ) = T f, tm [2.36]
M m=1
2.4.3. Fast fading
The fast fading of the propagation channel corresponds to the magnitude

fluctuations in the received signal. It may be assessed for a narrowband signal
or for a given delay in the case of a wideband signal. Mathematically, it is
characterized by the statistical law of the random variable |h(t, τ )| for a
given τ . In practice, we study a set of samples |h(tm , τ )| corresponding to M
measurements taken at close locations (or in a time-variant environment),
4. We may use the central frequency in the considered band for f0 , and the distance
d0 is generally 1 m.
while verifying that the channel stationarity condition holds. The distributions
generally used to assess fast fading include Rayleigh, Rice, Nakagami,
Weibull, and log-normal laws (cf. Appendix B). In order to select a specific
theoretical distribution, a goodness-of-fit test is employed, such as the
Kolmogorov-Smirnov test (cf. Appendix B.2).
Some useful tools also exist to characterize the temporal behavior of a signal.
The level crossing rate consists of determining the frequency at which the signal
magnitude decreases below a given threshold. The average fade duration is an
estimate of the time during which the signal magnitude stays below a given
threshold [PAR 00].
2.4.4. Spectral analysis
Another way to characterize temporal variations in the channel consists of

studying the Doppler spectrum for the main impulse response paths. These
spectra may be observed on the scattering function PS (τ, ν). In order to
characterize the channel variations independently of the delay, we use the
mean Doppler spectrum defined as:
∞
PH (0, ν) = PS (τ, ν)dτ [2.37]
−∞
As in the PDP case, we may define the Doppler spread as:

∞ 2
ν − νm PH (0, ν)dν
νRM S = −∞
∞ [2.38]
−∞ H
P (0, ν)dν
where νm is the mean Doppler shift defined by:

∞
νPH (0, ν)dν
νm = −∞ ∞ [2.39]
−∞ H
P (0, ν)dν
As explained earlier (section 2.2.3.3), the Doppler spectrum is closely linked

to the arrival direction of the multiple propagation paths. Parameters such as
the angular spectrum may be used to characterize the signal’s departure and
arrival directions. This is of particular interest for MIMO applications [COS 04].
2.5. Conclusion
In order to design and optimize wireless communication systems, an accurate

knowledge of the radio propagation channel is required. The performance of a
wireless transmission system is indeed dictated by the propagation conditions

between the transmitter and the receiver. These devices need to be designed in
order to benefit from the channel characteristics and to mitigate its negative
effects.
In this chapter, the propagation channel characteristics have been presented

in terms of power attenuation and multipath propagation. The spatial and
frequency selectivity concepts as well as the Doppler effect were introduced.
Multipath propagation may be regarded as a filter causing a distortion and a
temporal variation to the transmitted signal. A mathematical framework was
presented in order to describe both deterministic and random channels. For
this, the particular case of WSSUS channels has been thoroughly discussed.
A number of parameters are available to characterize the propagation

phenomena. We generally model the propagation loss with respect to
frequency and distance, the channel impulse response’s general shape and its
variations in terms of fast fading. The physical concepts and mathematical
tools presented in this chapter will serve as a basis to describe the simulation
and modeling techniques detailed in Chapters 4 and 5.
Chapter 3
UWB Propagation Channel Sounding
3.1. Introduction
Experimental characterization of the radio channel requires the analysis

of a large number of propagation measurements. In order to set up a
measurement campaign, a large number of sounding techniques are available
[GUI 99]. Selecting one method or the other depends on the measurement
environment, on the observed frequency band, and on the acquisition speed
constraints.
Sounding UWB channels raises some particular issues, which are developed
in this chapter. The different sounding methods for wideband channels are
then exposed, by distinguishing between frequency domain and time domain
techniques. In order to observe the UWB channel fluctuations in more detail,
it is necessary to employ a real time measurement technique. For this purpose,
we present an advanced sounding technique, exploiting the performances of a
wideband, single-input multiple-output (SIMO) sounder.
The end of the chapter illustrates the different channel sounding techniques
by presenting a few measurement campaigns. First, the most significant UWB
channel measurement campaigns are listed. For each experiment, the equipment
and the measurement conditions are described. We then present the setting up
of a sounding campaign through a few examples.
3.2. Specificity of UWB channel sounding
The purpose of channel sounding is to measure the impulse response

h(t, τ ) linking the received signal s(t) to the transmitted signal e(t) through
a filtering operation (see equation [2.9]). Such measurements characterize

both the frequency selectivity and the time variability of the channel. In
practice, there is no inverse convolution operation available, that would allow
for the extraction of the impulse response from arbitrary signals e(t) and s(t).
However, a number of techniques using specific exciting signals e(t) may be
used, where the impulse response is obtained by processing the signals e(t)
and s(t). The transmitted signal needs to fulfill some properties depending on
the method used for calculating the impulse response.

Sounder impulse Channel impulse
response response over an infinite
bandwidth
Delay (ns) Delay (ns)

Convolution
Measured impulse
Unresolved response
path
Delay (ns)
Figure 3.1. Time resolution of a wideband sounder
The main characteristics of a wideband sounder are the following:
• Analyzed bandwidth: the analyzed bandwidth corresponds to the

frequency band over which the impulse response is estimated. It generally
corresponds to the bandwidth of the transmitted probe signal. Propagation
channel measurements over bandwidths in the order of 100 MHz is well
supported by the currently available sounders [GUI 99]. In the case of UWB
signals, the bandwidth of several GHz may represent a challenge for channel
measurements.
• Time resolution: the time resolution characterizes the sounder’s

capability to distinguish between two paths with very close delays. The
impulse response estimated by the sounder corresponds to a convolution
UWB Propagation Channel Sounding 69
between the channel impulse response over an infinite bandwidth and the
sounder’s impulse response over the analyzed band (see Figure 3.1). The
latter may be obtained by directly cable-connecting the transmitter and the
receiver. The time resolution is generally defined as half the width of the
sounder impulse response peak. As an approximation, it may also be defined
as the inverse of the analyzed bandwidth. In order to avoid the shadowing of
some paths by the side lobes of the sounder impulse response, a weighting
window may be applied [HAR 78]. Such a window is selected as a compromise
between the level of the side lobes and the width of the main lobe.
• Maximum Doppler shift: when the propagation channel varies with time,
we can measure its frequency dispersion by studying its Doppler spectrum (see
section 2.4.4). For this purpose, the sounder needs to be able to quickly measure
successive impulse responses. The measurement repetition duration ΔT (meas)
is defined as the duration separating two successive channel measurements. It
encompasses the measurement acquisition duration t(meas) and some time for
data processing and storage. It is then possible to measure a maximum absolute
(meas) 1
Doppler shift νmax = 2ΔT (meas) . However, it should be noted that during the
measurement of a single impulse response, the channel should be considered as
static. Thus, a sounder capable of measuring the time varying channel needs
to present a very low acquisition duration t(meas) .
• Dynamic and length of the channel impulse response: the impulse

response dynamic corresponds to the power ratio between the maximum
impulse response and the noise level. A high dynamic allows the sounder to
detect strongly attenuated paths. The length of the channel’s impulse response
corresponds to the maximum delay that can be measured.
In the case of the UWB propagation channel, the time resolution is generally
high, due to the wide analyzed bandwidth. Consequently, developing UWB
sounders with a low measurement duration is a challenging task.
Some constraints are common in UWB channels, because of the wide

measured frequency band. Over an analyzed bandwidth of several GHz, the
behavior of the sounding equipment may vary strongly, and this needs to
be accounted for in the measurement process. In particular, the antennas
characteristics need to be stable across the measured frequency band.
Narrowband antennas (dipoles, horns, etc.) and dispersive wideband antennas
(spiral and log-periodical antennas, etc.) may thus not be used for this purpose
[SCH 03a]. In practice, biconical, monoconical or planar UWB antennas are
used. Some examples of these antennas are given in Figure 3.2. In any case, it
is necessary to properly characterize the antennas used for the measurement
[SIB 04]. The properties of some other pieces of equipment, such as cables or
amplifiers, may also vary in frequency, and it is important to characterize each
item accurately before the measurement.
Monoconical
antenna
Bowtie planar
antenna
Biconical
Ideal biconical antenna
antenna
Figure 3.2. UWB measurement antennas
Finally, a UWB receiver is sensitive to any radio emission at frequencies

near or within the analyzed frequency band. Before the measurement starts,
we need to make sure that the analyzed radio spectrum is free of any jamming
from external systems. Out-of-band emissions, such as the GSM and WiFi
signals, should also be filtered. This can be done using a high-pass filter. This
filter should be taken into account during the calibration phase in order not to
affect the measured signal.
3.3. Measurement techniques for UWB channel sounding
This section presents the main UWB radio channel measurement methods:
frequency domain methods and time domain methods. For each method, the
basic principles are presented, and the advantages and drawbacks of this
method are detailed. We finally present an innovative hybrid method for the
real time measurement of UWB channels.
3.3.1. Frequency domain techniques
The frequency domain method is the most commonly used UWB channel
sounding technique, on account of its ease of implementation. It consists of
sampling the channel transfer function T (f, t). This is done by transmitting a
narrowband signal at a fixed frequency, and by measuring the attenuation and
the relative phase of the received signal [GUI 99]. In practice, the analyzed band
is divided into N samples separated by a frequency step Δf (meas) . The impulse
response is obtained using an inverse Fourier transform along the frequency
axis.
The obtained time resolution1 is:
(meas) 1
Rt = [3.1]
N Δf (meas)
and the length of the channel impulse response is:
(meas) N −1
τmax = [3.2]
N Δf (meas)
3.3.1.1. Vector network analyzer

The tool best suited to characterizing the channel using a frequency
sweeping method is the vector network analyzer (VNA). This device is
generally used to characterize high frequency quadripoles through the
measurement of S-parameters. For the purpose of channel sounding, port 1 is
connected to the transmitting antenna and port 2 is connected to the receiving
antenna. The channel transfer function is derived from parameter S21 (f ).
In general, a sine signal is used to sweep the analyzed band. The received
signal is down-converted towards an intermediate frequency (IF), where it
is pass-band filtered around a fixed frequency and analyzed. Hence, this
device is capable of sweeping a very wide frequency band. Using narrowband
filters at the receiver leads to a very good measurement dynamic, but also
increases the overall measurement duration. In addition, the measurement of
the transfer function phase requires a very good synchronization between the
local oscillators (LO) at the transmitter and at the receiver. Figure 3.3 gives a
schematic view of this device.
1. This time resolution is calculated without applying any frequency domain

windowing.
Propagation
channel
Reference
signal
LO LO
S21 measurement
at IF
Analysis
and storage
Figure 3.3. Propagation measurement using a frequency domain sounder
This technique offers advantages in terms of bandwidth and dynamic.

As a result, it has been frequently used for the purpose of UWB channel
sounding [GUN 00, OPS 01, GHA 02a, CHE 02, KEI 02, HOV 02, KUN 02a,
GHA 03b, BUE 03, DAB 03, ALV 03, CHA 04, SCH 04, BAL 04a, KAR 04b,
JAM 04, CHO 04b, CAS 04a, HAN 05]. However, the measurement duration
is proportional to the number of measured frequency tones. For an analyzed
bandwidth of several GHz, the measurement duration is in the order of 10
seconds. Hence, this technique cannot be used for time varying channels.
When conducting VNA-based measurements, we should thus ensure that the
environment remains static for the measurement duration. In addition, the
distance from transmitter to receiver is limited to approximately 20 m, due to
power attenuation in the feeding cables.
3.3.1.2. Chirp sounder

A chirp sounder is a frequency domain method providing a solution to the
issue of acquisition duration. Indeed, the sounding signal is no longer a pure sine
wave, but a frequency chirp. Such a signal is generated using digital frequency
synthesis components [COS 04].
This frequency chirp may be characterized by the sequence duration Tc and

the covered frequency range Bc , as illustrated in Figure 3.4. For a Bc Tc product
larger than 100, over 98% of the energy is confined in the analyzed bandwidth
Bc .
The channel impulse response may be calculated either by matched

filtering or using a heterodyne receiver and a low-pass filter. In the first case,
the impulse response of the matched filter corresponds to a time reversed
version of the emitted chirp signal [ART 95]. The channel impulse response
is directly obtained at the output of the matched filter. In the case of a
heterodyne reception, the output of the low-pass filter provides the channel
transfer function, represented on a compact frequency band. Its bandwidth
depends on the parameters Bc and Tc [COS 04]. The signal acquisition may
thus be performed at a reduced sampling frequency. However, generating
a chirp signal with a bandwidth larger than a few hundred MHz is still a
challenging task. For this reason, chirp sounders have not yet been used for
UWB channel sounding.
Tc Tc Magnitude Bc
Frequency
Magnitude
Bc > 98%
energy
Time Time Frequency
Figure 3.4. Chirp signal
3.3.2. Time domain techniques
The common characteristic of all time domain techniques is the use of a

wideband excitation signal at the emitter. This way, the receiver processes
the whole frequency band simultaneously, which drastically reduces the
measurement duration. This section presents the most common time domain
methods, and in particular those that have been used for UWB channel
sounding.
3.3.2.1. Pulsed techniques

The pulsed technique is mathematically the most straightforward
time domain measurement method. Indeed, if the transmitted signal e(t)
corresponds to a Dirac pulse, the received signal s(t) is directly proportional
to the channel impulse response h(t, τ ). However, such a Dirac pulse would
present an infinite flat spectrum and is not physically implementable. In
practice, we use pulse generators enabling the emission of short signals with a
duration Δt in the order of 100 picoseconds. Denoting by ΠΔt (t) the emitted
pulse, the received signal is given by:
s(t) = h(t, τ ) ⊗ ΠΔt (t) [3.3]
and represents a close approximation of the channel impulse response.
At the receiver, a very fast acquisition of the signal is necessary. A digital

sampling oscilloscope (DSO) is generally used, with sampling rates up to
20 Gsamples.s-1 . Other types of receivers are also implemented: for instance,
a correlation receiver has been developed by the company Time Domain
[YAN 02].
This technique is particularly interesting for the measurement of UWB

channels. Indeed, pulse generators capable of emitting directly in the FCC
frequency band (3.1–10.6 GHz) are now available. Generally, UWB sounders
based on this principle do not require an up-converter stage, which simplifies the
experimental setup. It should be noted, though, that the signals synthesized by
the current pulse generators are limited to a maximum bandwidth of 1–2 GHz,
with an upper frequency limited to 5 GHz. Hence, only parts of the FCC band
may be measured using a direct pulse generator technique. In order to sound
the upper part of the FCC spectrum, it is possible to add an up-converter stage,
composed of an LO and a mixer at both the transmitter and the receiver. This
solution is presented as a dashed line in Figure 3.5.
The main advantage of the pulsed sounding technique is its low acquisition
duration, the channel impulse response being recorded in real time. This
technique should thus be selected for the measurement of space or time
channel variations. However, this method also presents a number of drawbacks.
Firstly, generating short duration pulses requires the amplifiers to deliver
a high power directly followed by idle periods. The resulting low average
power leads to a poor signal to noise ratio, and this method is not suited
to large distances in NLOS configurations. Secondly, the power dynamic at
the amplifier stage does not allow for an accurate control of the pulse shape.
Finally, this technique requires a perfect synchronization between the emitter
and the receiver, which may be solved by connecting the terminals using a
cable.
The attractive simplicity of this solution led a number of scientists to use

this method for UWB channel sounding, but with a sounded frequency band
generally below the FCC band [WIN 97a, WIN 97b, GUN 00, OPS 01, CHE 02,
YAN 02, TER 03, LI 03]. Two measurement programs have been established
using this technique within the 3.1–10.6 GHz band [PEN 02, BUE 03].
Propagation
channel
Reference
signal
0°
LO LO
90°
I Q
Pulse
Synchronization DSO
generator
Analysis
and storage
Figure 3.5. Propagation measurement using a pulsed technique
3.3.2.2. Correlation measurements

A possible way to increase the signal to noise ratio at the receiver is to use
the autocorrelation properties of pseudo-noise (PN) sequences. Let Ree (t) and
Res (t) respectively denote the input signal autocorrelation and the correlation
function of the input signal e(t) with the output signal s(t). According to the
convolution and correlation function properties, these functions are also linked
by a convolution equation:
Res (t) = h(t, τ ) ⊗ Ree (t) [3.4]
If the input signal is close to a white noise, its autocorrelation function is

close to a Dirac pulse. In this case, the identification problem is similar to the
one occurring in the pulsed technique. In practice, we use maximum length PN
sequences, also called m-sequences. This corresponds to a good approximation
of a colored noise with zero mean and limited bandwidth. An m-sequence may
be generated using a shift register with m bits and an adder, which results in
a sequence of length 2m − 1. Its autocorrelation function presents a theoretical
dynamic of 20 log(2m − 1) dB. In the frequency domain, the −3 dB analyzed
bandwidth is equal to the clock frequency fc used at the PN sequence generator.
At the transmitter, a correlation sounder is thus composed of a PN sequence

generator and an up-converter stage towards the working frequency. Typically,
wideband correlation sounders transmit signals in the order of 100 MHz. For
UWB channel sounding, the frequency band may be widened by increasing the
clock frequency, but the sounder implementation becomes more complex. The
University of Ilmenau, Germany, presented a MIMO UWB sounder working in
the 0.8–5 GHz frequency band [SAC 02]. For this purpose, an integrated circuit
shift register has been specifically developed.
At the receiver, once the received signal has been down-converted in

baseband, several acquisition techniques may be used. The first consists of
using an analog filter matching the transmitted PN sequence. This principle
enables the measurement of the impulse response in real time. However,
the analyzed bandwidth is limited by the passband of the filter, and the
measurement dynamic is low. Another technique consists of sampling the
received signal and computing the convolution function using digital filters.
The maximum analyzed bandwidth is then limited by the speed of the
analog-to-digital converters. This solution is presented in Figure 3.6. It has
been used by the University of Rome “Tor Vergata”, Italy, to sound the
3.6–6 GHz band [DUR 04]. The receiver was simply composed of a DSO.
Time domain correlation sounders present the advantages of a high

impulse response dynamic, and a possible functioning in real time. Their
implementation is however complex, and the quality of the measured impulse
response largely depends on the performance of the sounder components.
In particular, an accurate synchronization between the transmitter and the
receiver is necessary. This may be achieved using a stable reference (e.g. a
rubidium oscillator) or a direct cable connection between the two terminals.
Sliding correlation technique
The sliding correlation sounder is a similar technique that proceeds the

convolution operation in the analog domain while increasing the measurement
dynamic. The convolution is performed using a replica of the original PN
sequence with a clock frequency fc2 slightly different from the initial clock
frequency fc1 . This leads to a time domain sliding of the PN sequences with
respect to each other. Hence, using an integrator, one point of the correlation
Propagation
channel
Reference
signal
0°
OL OL
90°
I Q
Arbitrary
sequence Synchronization Fast acquisition
generator
(e.g. shift
register)
Analysis
Correlation
and storage
or inversion
Figure 3.6. Propagation measurement using a PN sequence. The illustrated

acquisition technique consists of digitizing the received signal
prior to convolution processing
function between the transmitted signal and the received signal is calculated
at each instant.
The obtained impulse response is expanded by a ratio k = fc2f−f c1

c1
in
the time domain and compressed by a ratio k in the frequency domain.
This compression enables the use of narrower filters at the receiver, which
improves the dynamic. In addition, a lower sampling rate may be used for the
analog-to-digital conversion.
This technique has been used by the University of Kyoto Sangyo (Japan)
to develop a UWB sounder operating in the 5.5–8.5 GHz frequency band
[TAK 01]. The main drawback of this technique is the duration of the impulse
response calculation, which precludes any Doppler analysis of fast time varying
channels.
3.3.2.3. Inversion techniques

Inversion techniques are based on the transmission of an arbitrary sequence
and are thus similar to correlation sounding techniques (see Figure 3.6). The
main difference lies in the impulse response estimation, which is performed
using an estimation filter. Filtering is generally applied in the digital domain
after the acquisition of the analog signal. The used filter minimizes the squared
error between the impulse response and the estimated response. In particular,
such a filter accounts for the imperfections of the measurement device.
The inversion technique is notably used in the sounder developed at France

Telecom Research and Development [CON 03]. It uses a Wiener filter, with a
frequency spectrum given by:
E ∗ (f )I ∗ (f )
GWiener (f ) = [3.5]
|E(f )I(f )|2 + α
where E(f ) represents the transmitted signal spectrum, I(f ) represents the
sounder response spectrum, and α is a non-zero constant linked to the signal
to noise ratio [BAR 95].
As a main advantage, the inversion technique improves the impulse

response calculation by accounting for the imperfections of the sounder
components. The used PN sequence is optimized in order to spread over
a wide band with a relatively constant envelope. The limitations of this
technique are directly linked to the performances of the digital-to-analog and
analog-to-digital converters. In particular, the analyzed bandwidth is in the
order of a few hundred MHz.
For the sake of comparison, Table 3.1 presents the main advantages and
drawbacks of the presented frequency domain and time domain techniques.
3.3.3. Multiple-band time domain sounder for dynamic channels
Classical frequency or time domain sounding techniques are not intrinsically

well adapted to the measurement of time varying UWB channels. On the
one hand, frequency domain techniques lead to a long measurement duration.
On the other hand, time domain techniques are fast enough to enable the
measurement of time varying channels, but their implementation within the
FCC band is challenging. This section presents a hybrid solution consisting
of using a wideband time domain sounder and successively measuring several
adjacent partial bands. This principle was initially exploited by the University
of Kassel (Germany). They developed a sounder capable of measuring a channel
of 600 MHz divided into 10 partial bands of 60 MHz, with a measurement
Technique Dynamic Doppler analysis Maximum bandwidth a Implemen- Use for UWB
tation sounding
VNA very high impossible 10 GHz and above easy frequent
(limited by the (static channel)
measurement duration)
Frequency
Chirp significant possible for low 100–300 MHz (limited by complex no experiment
Doppler spreads signal generation)
Pulsed low possible 1–2 GHz (limited by easy frequent
pulse generator) out of FCC band
Correlation significant possible 1–5 GHz (limited by complex marginal
clock frequency and
acquisition)
Time
Sliding high possible for low 1–3 GHz (limited by complex marginal
correlation Doppler spreads clock frequency)
Inversion significant possible 500 MHz–1 GHz (limited complex no experiment
by DAC and acquisition)
a
For practically implemented solutions.
Table 3.1. Comparison of the different sounding techniques

79
repetition duration of 300 μs [KAT 00]. In the following, we present the different
issues linked to the development of such a sounder, and we illustrate how it can
be implemented on the basis of a wideband SIMO sounder [PAJ 03].
3.3.3.1. Principle of multiple-band time domain sounding

The technique of multiple-band time domain sounding exploits the
performance of a time domain sounder, such as the sounder developed by
France Telecom Research & Development. In its standard version, this sounder
allows for the measurement of 250 MHz frequency bands. At the up-converter
stage, an LO signal is responsible for the up-conversion of the transmitted
sequence towards the analyzed band. By appropriately driving this LO signal,
it is possible to sound several adjacent bands. The different challenges are the
following:
• Analyzed band: the number of measured partial bands determines the

width of the global analyzed band. At both the transmitter the receiver, the
global analyzed band needs to be filtered and cleared from any interfering
signal, such as the LO signal and its harmonics.
• Measurement repetition duration: in an indoor configuration, the

maximum speed of mobile terminals is about 2 m.s−1 . Considering an extreme
case, i.e. a mobile terminal and a maximum frequency of 10.6 GHz, the
maximum Doppler shift is λv = 70 Hz. This maximum Doppler shift increases
to 2v
λ = 140 Hz if we consider mobile scatterers, taking into account the paths
with one reflection as significant paths. Thus, to characterize time varying
channels, the equipment has to sample the channel every 3.5 ms. This time
includes the necessary switching from one band to another, which needs to be
synchronized with the transmitted sequence.
• Calibration: when measuring adjacent bands, a simple process needs to be

set up for an accurate calibration of the sounder in terms of phase, magnitude
and absolute delay, for each of the adjacent bands.
• Measurement dynamic: when working as a multiple-band sounder, the

dynamic performance of the original wideband sounder needs to be preserved,
in order to obtain a reliable result.
These different issues are similar to those occurring when characterizing

the wideband channel with multiple antennas at the receiver, in a SIMO
configuration. A SIMO sounder may thus be easily adapted to perform
multiple-band time domain sounding. For a better understanding, the next
section reviews the general structure of the SIMO sounder available at France
Telecom. A further description of this sounder is available in [CON 03].
3.3.3.2. Description of the SIMO channel sounder

The sounder developed at France Telecom Research & Development is a
time domain sounder allowing for radio channel measurement over different
frequency bands between 2 GHz and 60 GHz. The sounding method is the
Wiener inversion technique, applied on a PN sequence exhibiting a flat
spectrum over the analyzed bandwidth. For frequency bands below 17 GHz,
a multiple sensor mode enables the simultaneous measurement of the radio
channel over 10 receiving antennas.
In terms of performance, the sensitivity of this sounder is about −85 dBm.

The impulse response dynamic depends on the length of the PN sequence, the
signal to noise ratio at the sounder input, and the oscillator phase noise. For a
received signal of about −60 dBm over a band of 250 MHz around 5 GHz and
a sequence of 8192 samples, the measurement dynamic is larger than 40 dB.
Figure 3.7 presents a simplified block diagram of the receiver. To repeat the
measurement as quickly as possible, a fast switch selects the following antenna
every other sequence. The first PN sequence is used for the measurement,
while the antennas switch during the next PN sequence. The signal is then
amplified using a wideband (3–18 GHz) low noise amplifier (LNA). It is then
pass-band filtered around the RF carrier frequency. A first down-converter
stage driven by an external frequency synthesizer displaces the signal around
an IF at 1.5 GHz. Automatic gain control (AGC) is then applied using variable
attenuators to compensate for the power variations in the received signal.
AGC is regularly performed, before the measurement of the signal received
by the different sensors. Thus, the AGC duration limits the measurement
repetition duration. In a traditional configuration with 10 antennas, the
minimum repetition duration is about 1.2 ms. A second down-converter stage
displaces the signal around a second IF at 250 MHz. The received signal is
then sampled using an analog-to-digital converter (ADC) at a sampling rate
of 1 Gsample.s−1 . We may note that the transmitter and the receiver use the
same IF at 1.5 GHz. Thus, the LO frequency is the same for the external
synthesizers at the transmitter and the receiver. All LO are synchronized
using a reference rubidium at 10 MHz.
3.3.3.3. Extension towards UWB

In order to perform measurements over adjacent frequency bands, it is
necessary to modify the LO frequency at the transmitter and at the receiver.
Traditional oscillators present a locking time in the order of 10–50 ms. This
transition duration is mainly due to the positioning duration of the internal
phase locked loop, and cannot be reduced. Directly driving a single external
synthesizer thus leads to a total measurement duration in the order of
40–200 ms for an analyzed band of 1 GHz. Taking into account the constraints
Fast switch
10 1
RF
LNA AGC
A 250 1.5
External D MHz GHz
synthesizer
LO
Figure 3.7. Receiver block diagram
linked to real time measurements (see section 3.3.3.1), this solution cannot be
used.
The extension of the SIMO sounder towards a real time UWB sounder
exploits the duality between multiple-input measurements and multiple-band
measurements. The main idea is to reuse the fast antenna switching module,
in order to switch between the carrier frequencies of each partial band to be
measured.
A practical implementation of this concept is presented in Figure 3.8. As

can be seen, the input connector is now directly connected with the LNA,
and a single antenna is available. The sounder is thus in a single-input
single-output (SISO) measurement configuration. The fast switching module
is used to sequentially feed the mixer of the first down-converter stage with
one of the 10 LO signals (f1 to f10 ) in turn. These signals are generated
by external synthesizers and tuned so that the receiver actually sweeps the
selected partial bands. As each synthesizer is locked at a fixed frequency, no
delay is experienced for the locking of the LO frequency. In this configuration,
the sounder is capable of sweeping up to 10 partial bands of 250 MHz each.
Hence, UWB measurements are theoretically possible up to a bandwidth of
2.5 GHz.
On account of its development concept, this sounder fulfills most of

the required criteria for real time UWB channel measurement. Indeed, the
multiple-input architecture is directly used for the measurement over multiple
bands. For the calibration of each partial band in terms of phase, magnitude
External
synthesizers LNA AGC
f1 Fast switch
10 1
A 250 1.5
f10 D MHz GHz
Towards
LO
the emitter
up-converter
Figure 3.8. Block diagram of the modified receiver with UWB extension
and delay, the procedure used in the SIMO configuration is still valid. In
addition, regarding the fast switching of partial bands, the overall switching
time through all partial bands may be as low as 1.2 ms, which allows for the
measurement of channel variations with a maximum Doppler shift of 416 Hz.
In the original version of the SIMO sounder, the filter preceding the first
down-converter stage was only a few hundred MHz wide to reject undesirable
signals and reduce the noise level. The same type of filtering was performed at
the emitter. In the UWB configuration, however, the sounded frequency band
needs to be filtered by one single pass-band filter, in order to avoid unnecessary
filter switching. In its current version, the UWB sounder is equipped with filters
of 1 GHz bandwidth.
Unlike the SIMO channel sounding case, UWB measurements using

the sweeping method necessitates a periodical modification of the emitted
signal central frequency. For this reason, the frequency switch performed
at the receiver and at the transmitter need to be perfectly synchronized.
To solve this problem, the same LO signal may be used to feed both the
receiver down-converter and the transmitter up-converter stages. The main
drawback of this solution is that the transmitter and the receiver need to be
cable connected. However, in most indoor configurations, this connection is
realizable using a cable of acceptable length. Advantageously, this solution
allows us to use only one set of external synthesizer, which considerably
reduces the global cost of the equipment.
3.3.3.4. Experimental validation

The modifications presented above have been applied to the sounder
developed by France Telecom for its extension to real time UWB
measurements. To sweep the 5–6 GHz frequency band, 5 external synthesizers
have been connected to the fast switching module. In this configuration, the
impulse response dynamic was measured at 40 dB. Using 5 partial bands,
the minimum measurement repetition duration is about 1 ms and the actual
measurement duration is 20.5 μs. With its 256 MB memory, the acquisition
card is able to sound the time varying UWB channel for a duration of 80 s in
standard conditions (bandwidth of 1 GHz and observable Doppler spread of
150 Hz).
Figure 3.9 presents the results obtained during a real time radio channel
measurement in a dynamic environment. During this experiment, the
transmitting antenna was fixed and the receiving antenna was held by a
moving person. This configuration corresponds to the practical situation where
a user with a mobile terminal walks in the proximity of a fixed access point.
0
(a)
2 (b) -5
4 -10
Relative attenuation (dB)

6 -15
Distance-delay (m)
(c)
8 -20
10
(d) -25
12
-30
14
-35
16 (e)
-40
18
-45
0 2 4 6 8 10 12
Time (s)
Figure 3.9. Time varying impulse response in a dynamic environment
The impulse response represented in Figure 3.9 corresponds to a 12.5 s

record of the signal obtained at the receiving antenna moving during
the experiment. Acquisitions were performed every 10 ms. For ease of
interpretation, the path delay on the y-axis has been converted to path length
in meters. In the first part of the trajectory (between t = 0 s and t = 7 s), the
person is moving towards the emitter antenna, reducing its relative distance
from 6 m to 2 m. Hence, we can observe a main path with increasing power
(a). In the second part of the trajectory (between t = 7 s and t = 11 s), the
person is moving away from the transmitter antenna, partially obstructing
the line of sight. This explains the shadowing experienced by the shortest
path (b). In this part of the curve, three other main paths are observable, one
with increasing length (c) and two with decreasing lengths (d, e). These two
last paths might correspond to echoes transmitted via a reflection on a wall
opposite the transmitter location.
This feasibility study shows that it is possible to develop the UWB channel
sounder by applying minor modifications to a wideband SIMO sounder. By
exploiting the duality between multiple inputs and multiple bands, most of the
SIMO-capable sounders could be extended towards UWB channel sounding.
The prototype presented here allows for the measurement of UWB channels
over a band of 1 GHz with a dynamic of 40 dB.
3.4. UWB measurement campaigns
3.4.1. Overview of UWB measurement campaigns
A number of UWB radio channel measurement campaigns were cited in

this chapter. These diverse campaigns used different sounding techniques,
but also differ in the measured frequency band, the type of environment, the
antennas used and the measurement set-up. In order to compare the analysis
results from these experiments, it is necessary to recall the conditions of
each campaign. Table 3.2 summarizes the inventoried UWB measurement
campaigns and provides some useful information for each of them.
Among the first UWB studies, the UltRaLab laboratory from the University
of South California undertook two measurement programs in 1997, but most
UWB experiments started from the year 2000 onwards. Most measurement
programs used the VNA frequency domain method. Other experimental set-ups
are based on a time domain method, using pulses or PN sequences. Different
indoor environments, such as the office or residential environments, and some
outdoor environments (forest, urban) were sounded. We may also note some
original experiments, on a metal ship or in an industrial environment. Finally,
regarding the analyzed band, only seven measurement campaigns cover the
entire FCC band [KUN 02a, BUE 03, ALV 03, CAS 04c, KAR 04b, HAN 05,
PAG 06b]. Within the FCC band, three experiments only permitted real time
UWB measurements [PEN 02, CAS 03, PAG 06a].
86
Measurement Date Reference Sounding method Antennas Analyzed band Environment Link Spatial
campaign configuration
UltRaLab 1997 [WIN 97a, Pulse generator and Patch diamond Width Forest; office LOS/ 49 locations
WIN 97b] DSO dipoles >1 GHz NLOS on a 90 cm ×
90 cm grid
UltRaLab SS 2000 [GUN 00] Pulse generator and Patch diamond 700–1.8 GHz, Metal ship LOS/ —
Curtis DSO; VNA; Pulse dipoles with 1 MHz NLOS
generator and step
correlation receiver
Stanford 2001 [OPS 01] Pulse generator and Wideband 1–1.8 GHz, Office; LOS/ —
University DSO; VNA antennas with 2 MHz laboratory NLOS
step
AT&T Labs - 2002 [GHA 02a, VNA Monoconical 4.375– Residential LOS/ —
MIT GHA 03c] omnidirectional 5.625 GHz, NLOS

antennas with 3.1 MHz
step
Intel Labs 2002 [CHE 02] Pulse generator and Biconical 2–8 GHz, with Residential; LOS/ Two
DSO; VNA antennas (MEC) 3.75 MHz step office NLOS perpendicular
and discone positioners
antennas (length 25 cm)
(Pulsicom with 20
Technol.) locations
Time 2002 [PEN 02] Pulse generator and Omnidirectional 3–5 GHz, with Residential; LOS/ —
Domain DSO antennas 3.75 MHz step office NLOS
Corporation
(DSO)
Time 2002 [YAN 02] Pulse generator and Omnidirectional Width Office NLOS —
Domain correlation receiver antennas >1.5 GHz
Corporation [WIT 99] around 2 GHz
(correlator)
UCAN - 2002 [KEI 02, VNA Omnidirectional 2–3 GHz, with Residential; LOS/ 100 locations
CEA LETI KEI 03] monoconical 2.5 MHz step office NLOS on a 90 cm ×
antennas 90 cm grid
(CMA118/A)
Ultrawaves - 2002 [HOV 02] VNA Omnidirectional 2–8 GHz, with Office LOS/ —
Oulu monoconical 3.75 MHz step NLOS
University antennas
(CMA118/A)
Whyless.com 2002 [KUN 02a] VNA Biconical 1–11 GHz, Office LOS/ 4500 locations
antennas (ETH with 6.25 MHz NLOS on a 149 cm ×
Zürich) step 29 cm grid
AT&T - 2003 [GHA 03b] VNA Omnidirectional 2–8 GHz, with Residential; LOS/ 25 locations
WINLAB monoconical 3.75 MHz step shopping NLOS on a 20 cm ×
antennas center 20 cm grid
87
88
CNAM - 2003 [TER 03] Pulse generator and Wideband 0–3 GHz Corridor LOS —
Thales DSO antennas
France 2003 [PAG 03] VNA Patch 4–6 GHz, with Office LOS/ 60 locations
Telecom omnidirectional 2 MHz step NLOS on a rotating
R&D antennas arm (27 cm
radius)
Hong Kong 2003 [LI 03] Pulse generator and Omnidirectional 1.2–1.8 GHz Office; LOS/ x-positioner
University DSO dipole laboratory; NLOS (length
meeting 120 cm) with
room 61 locations
NETEX - 2003 [BUE 03, Pulse generator and Omnidirectional 100–12 GHz Office LOS/ 49 locations
Virginia Tech BAY 04] DSO; VNA biconical NLOS on a 90 cm ×
antennas and 90 cm grid
TEM horns
New Jersey 2003 [DAB 03] VNA Omnidirectional 2–6 GHz Office; LOS —
Instit. of monoconical laboratory
Tech. antennas and
log-periodical
antennas
UCAN - 2003 [ALV 03] VNA Omnidirectional 1–13 GHz, Office; LOS/ 9 locations on
Cantabria monoconical with 2.5 MHz laboratory NLOS a 6 cm × 6 cm
University antennas step grid
(EM6865)
Ultrawaves - 2003 [CAS 03, PN sequence and Omnidirectional 3.6–6 GHz; Office; LOS/ 625 locations
Rome Tor CAS 04c] DSO; VNA; PN monoconical 2–12 GHz, laboratory NLOS on a 48 cm ×
Vergata sequence and antennas with 5 MHz 48 cm grid; 73
University correlation receiver (EM6865); step positions on a
prototype non-uniform
antennas grid
(ENSTA)
Chiao-Tung 2004 [CHA 04] VNA Not 3–5 GHz, with Office LOS/ 64 locations
University communicated 2.5 MHz step NLOS on a 26 cm ×
26 cm grid
ETH Zürich 2004 [SCH 04] VNA Planar 2–8 GHz, with Corridor; LOS/ 45 locations
omnidirectional 1.875 MHz lobby NLOS on a 28 cm ×
antennas step 56 cm grid
(Skycross)
Instit. for 2004 [BAL 04a] VNA Omnidirectional 3–6 GHz, with Urban; LOS/ 9 or 49
Infocomm conical antennas 1.875 MHz office NLOS locations on a
Research step 10 cm × 10 cm
or 30 cm ×
30 cm grid
Lund 2004 [KAR 04b] VNA Omnidirectional 3.1–10.6 GHz, Industrial LOS/ Two
University monoconical with 6 MHz NLOS x-positioners
antennas step (transmitter/
receiver) with
7 locations
along 30 cm
89
90
Oulu 2004 [JAM 04] VNA Omnidirectional 0.5–10 GHz, Office LOS/ x-positioner
University monoconical with 6.25 MHz quasi- with 7.9 mm
antennas step NLOS step
(CMA118/A)
Samsung 2004 [CHO 04b] VNA Planar dipoles 3–10 GHz, Residential LOS/ 25 locations
with NLOS on a 60 cm ×
4.375 MHz 60 cm grid
step
Tokyo Instit. 2004 [TSU 04, VNA Monopoles and 3.1–10.6 GHz, Residential; LOS/ Up to 10 × 10
of Tech. HAN 05] biconical with 10 MHz office NLOS × 7 locations
antennas step at transmitter
and receiver
France 2005 [PAG 06b] VNA Omnidirectional 3.1–11.1 GHz, Office LOS/ 90 locations
Telecom monoconical with 2 MHz NLOS on a rotating
R&D antennas step arm (20 cm
radius)
France 2005 [PAG 06a] Multiple-band time Omnidirectional 4–5 GHz, with Office LOS —
Telecom domain sounder monoconical 2 MHz step corridor
R&D antennas
Table 3.2. UWB propagation channel measurement campaigns

3.4.2. Illustration of channel sounding experiments
The previous study of the different UWB channel sounding techniques

shows that the frequency domain technique using a VNA is the only available
method that enables a full characterization over the whole FCC band. In the
following, two measurement campaigns using this method are presented. The
first campaign covers the 3.1–10.6 GHz band [PAG 06b], and the second one
covers the 2–6 GHz band. A third campaign is then described, presenting real
time measurements in the 4–5 GHz band.
3.4.2.1. Static measurement campaign over the 3.1–10.6 GHz band

The measurement system used for the static measurement campaign over
the 3.1–10.6 GHz band is presented in Figure 3.10.
Propagation
channel
Rotating arm
Port Port
Webcam USB
1 2
GPIB
VNA
Rotating HP8510C
arm driver RS 232
Driving and
storage
Figure 3.10. Equipment setup for the static UWB sounding

over the 3.1–10.6 GHz band
Measurements were taken using a VNA (see Figure 3.11(a)) in the

frequency band 3.1–11.1 GHz with a frequency step Δf (meas) of 2 MHz.
The full acquisition process consists of measuring 4005 frequency tones in a
duration of about 15 seconds. Using a Hanning window, side lobes may be
reduced to −32 dB. This compromise corresponds to an impulse response
(meas)
resolution Rt of 0.25 ns.
(a) (b)
Figure 3.11. Measurement equipment: (a) HP8510C sounder and (b) rotating arm
In order to estimate the local PDP and assess the signal spatial fluctuations,
a rotating arm with radius 20 cm was used to measure the radio channel at 90
different locations (see Figure 3.11(b)). This configuration corresponds to a
circular path of 45λ and to a distance between two successive sensors below λ2
at the maximum frequency of 10.6 GHz.
Measurements were performed using two CMA118/A antennas from the

company Antenna Research Associates. These antennas are monoconical
antennas with a metallic ground plane. Their radiation pattern is
omnidirectional in the azimuth plane and a standing wave ratio (SWR) lower
than 2 in the 1–18 GHz frequency band. It should be noted that the radiation
pattern of these antennas strongly varies, especially in the elevation plane, as
the frequency increases several octaves. Their complex gain has thus been
characterized in 3D in the 1–10 GHz band with a 1 GHz step. Figure 3.12
illustrates the antenna characterization process.
Figure 3.12. Measurement antennas. CMA 118/A antennas

(from the ARA company) during their characterization
This sounding equipment was complemented with up to three wideband

power amplifiers. The sounder calibration consists of a measurement sample
taken while the receiver and the transmitter are directly cable connected.
In order to achieve an accurate time reference, all cables involved in the
measurement process need to be taken into account at the calibration stage.
As active components, the amplifiers are not included in this measurement.
Instead, they are characterized separately and their transfer function is
subtracted from the measurements at the digital post-processing stage.
Measurements were taken in an indoor office environment. External

walls are made of brick and concrete, and internal walls consist of thinner
plaster and plastic boards. Two typical radio access configurations have been
studied. In both cases, the transmitter was situated at a height of 1.40 m and
was considered as the mobile terminal. The receiver was considered as the
access point and was first placed in a meeting room at a height of 2.19 m,
and then in a corridor at a height of 2.45 m. Figure 3.13 presents these two
configurations. In both cases, LOS and NLOS configurations were assessed.
The transmitter-receiver distance varied between 1 m and 20 m. Over the
whole measurement campaign, the rotating arm was placed at more than 120
different locations, which corresponds to a set of over 10,000 UWB impulse
responses available for statistical characterization. The analysis of these
measurements is presented in Chapter 5.
94
4m
Corridor configuration Meeting room configuration
Figure 3.13. Measurement locations during the UWB sounding campaign over
the 3.1–10.6 GHz band. Black squares represent fixed locations (receiver)
and white circles correspond to the rotating arm locations (transmitter)
3.4.2.2. Static measurement campaign over the 2–6 GHz band

The schematic description of the equipment used for the static measurement
campaign over the 2–6 GHz band is presented in Figure 3.14. Measurements
were performed using the VNA HP8753 D over the frequency band 2–6 GHz
with a frequency step Δf mes of 2.5 MHz. The acquisition process for a
full measurement consists of the measurement of 1601 frequency tones in a
duration of 4 seconds. The CMA 118/A antennas have also been used for this
measurement. The measurement did not require the use of power amplifiers
at the transmitter or at the receiver. Indeed, the link budget without the
amplifiers was high enough to support transmitter-receiver distances up to
12 m.
CMA – 118/A
MATLAB/SCILAB Tx (Port 1) Cable 20 m
Tx Ant
Post processing
Analysis Vector
Network
Analyzer Propagation channel
HP 8753 D
PC GPIB Cable 20 m
Labview Rx Ant
Rx (Port 2)
CMA – 118/A
Figure 3.14. Equipment setup for the static UWB

sounding over the 2–6 GHz band
The environments sounded during these measurements are residential

houses: the first house is a modern building and the second one is an older
building. The external walls of the first house consist of concrete blocks,
bricks and insulation material made of glass wool. The external walls of the
older house do not incorporate any insulating material. Internal walls of both
houses are made of brick and plasterboards. Both dwellings are two-floor
houses. However, measurements have been performed at the ground level only.
The transmitter and the receiver are placed at the same height of 1.40 m.
Figure 3.15(a) represents the first house with LOS measurement locations.
Figures 3.15(b) and 3.15(c) represent the second house with both LOS and
NLOS measurement locations. In these two environments, the number of
measurements is respectively 89 and 230.
3.4.2.3. Dynamic measurement campaign over the 4–5 GHz band

Finally, in order to study the effect that mobile people have on the UWB
channel, we present a real time measurement campaign performed over a
bandwidth of 1 GHz [PAG 06a].
(a)
(b)
(c)
Figure 3.15. Measurement locations during the UWB sounding campaign over
the 2–6 GHz band. Squares represent fixed locations (Tx: transceiver)
and circles represent mobile locations (receiver): (a) modern house;
(b) old house in LOS; and (c) NLOS
A few channel sounders only are currently capable of measuring the UWB
channel in real time. This experimental study has been performed using the
UWB sounder presented in section 3.3.3. The channel impulse response was
measured every 10 ms, which enables the measurement of Doppler shifts
up to 50 MHz. The analyzed band extends from 4 GHz to 5 GHz, with a
resolution of 2 MHz, which corresponds to a maximum delay of 500 ns. At the
transmitter and the receiver, CMA118/A antennas were used, as presented in
section 3.4.2.1.
Rx antenna
Rx
Tx
B
F
F
Tx antenna
People
motion
Figure 3.16. Environment of the real time experiment. F: forward movement,

B: backward movement, Tx: transmitter, Rx: receiver
The experiment took place in a typical indoor office environment. All

measurements were taken near a bend of the building’s main hallway, as
depicted in Figure 3.16. The receiving antennas was fixed on a wall at a height
of 2.10 m. The transmitting antenna was placed in the middle of the corridor
at a distance of 11 m from the receiving antenna and at a height of 1.35 m.
In order to assess the time variations of the UWB channel, measurements
were taken over the fixed radio link as groups of people walked towards the
end of the hallway and back. During this displacement, people occasionally
obstructed the LOS path and other main paths of the channel impulse
response. The number of people within each group varied from 1 to 12. This
experiment corresponds to a practical situation where the user of a WLAN
is standing in a crowd of people, in a subway corridor for instance. The
measurement campaign consisted of a collection of 27 records of the time
varying UWB radio channel, containing about 3,000 impulse responses each.
Chapter 5 presents a statistical analysis of this experimental data.
3.5. Conclusion
In this chapter, the different channel sounding techniques adapted to UWB

measurements have been presented. Frequency domain methods use either a
vector network analyzer or a chirp sounder. Time domain methods include
pulsed techniques, correlation measurements and inversion techniques. Due to
the wide analyzed bandwidth, the vector network analyzer method seems to
be the most appropriate technique for static UWB channel sounding. However,
its main limitation lies in its long measurement duration. Several seconds are
indeed required to measure the 3.1–10.6 GHz band. For the acquisition of real
time UWB channel measurements, an innovative sounding technique based on
the extension of a wideband SIMO sounder has been presented. This adaptation
corresponds to minor modifications of the original equipment and allows for real
time measurements over a 1 GHz band.
A literature survey made it possible to inventory about 20 UWB

measurement campaigns since 1997. Most of these campaigns used a
VNA-based frequency domain method. The reported campaigns cover
the indoor residential and office environment, as well as some outdoor
environments. Finally, the experimental setup of a typical sounding campaign
has been illustrated through three examples: two measurement campaigns
using a VNA over the 3–10.6 GHz and 2–6 GHz frequency bands respectively,
and a series of real time measurements performed over the 4–5 GHz band
using a multiple-band time domain sounder.
Chapter 4
Deterministic Modeling of the UWB Channel
4.1. Introduction
Deterministic models are often used as site specific models allowing us to

realistically predict signal propagation in a given environment using simulation
software. Currently, little simulation software of this type is proposed for UWB
in comparison to narrowband systems, where they are widely used to determine
radio system deployment.
The use of classical deterministic modeling software for UWB needs a

particular adjustment in order to cover the entire frequency band of impulse
signals. Concerning the total calculation field, it is mandatory to consider
not only the power loss amplitude but also the phase information in order to
easily reconstruct in the time domain the received signal corresponding to the
modeled UWB link.
This chapter describes the UWB deterministic modeling. After the

presentation of classical techniques used for deterministic modeling,
the specificities of UWB applications are described. Then, the different
deterministic models proposed in the literature for UWB signal propagation
are presented. Finally, theoretical formalisms are detailed and compared to
measurement results in order to illustrate the deterministic modeling in the
UWB context.
4.2. Overview of deterministic modeling
Propagation models considered to be deterministic are obtained from

simulations made in simplified propagation environments. They are based on
electromagnetic wave propagation theory. Their use requires a good knowledge

on the propagation environment and allows us to obtain precise as well as
accurate predictions of signal propagation in the channel corresponding to the
considered environment.
For a theoretical point of view, the wave propagation characteristics can

be calculated by solving Maxwell equations. However, this requires us to make
complex mathematical operations and to use powerful numerical calculations.
The most frequently used approaches are based on frequency difference in time
domain (FDTD) techniques, methods of moments (MoM) or ray techniques.
4.2.1. FDTD based approach
The FDTD based approach allows us to obtain a complete cartography of

the field on all points of a given map. It is typically used for electromagnetic
field coverage in a given environment [LAU 95, SCH 97b, KON 99].
The FDTD used for modeling consists of solving Maxwell equations using
regular spatial discretization in time. All derivations are simplified by algebraic
systems of equation. The studied volume is discretized in various elementary
cells, the size of which is related to wavelength of typically λ/10 or λ/15. The
solving method fixes some restrictions in term of the considered cells sizes and
time grid in order to ensure computation stability. The studied environment is
limited by absorbing walls limiting the reflection phenomena on the boundaries
[IBA 00].
These approaches demand huge memory spaces to obtain the solutions

in all points of the considered environment and to update all iterative field
calculations for the different instants. This approach is often used for small
environments with respect to the wavelength or complementary to other
approaches [YIN 00].
4.2.2. MoM based approach
Similarly to the FDTD approach, the MoM based approach is a numerical

method which requires a large numerical memory for field calculation and radio
coverage in a given environment [DEB 96, YAN 99].
The MoM is based on the use of an integral form of Maxwell equations. The
integral system is transformed by an impedance matrix discretization which
represents the interactions between elementary cells of the environment. The
accuracy of the solutions obtained with this approach depends on the size of
the considered cells.
Deterministic Modeling of the UWB Channel 101
This approach is used when the structure size is roughly a few wavelengths.
It can also be considered complementary to other approaches like the ray based
approaches in order to obtain hybrid models. Nevertheless, the MoM based
approach is mainly used to validate results obtained with other approaches like
the ray based approach [YAN 98].
4.2.3. Ray based approach
Based on geometric optic (GO) combined with uniform theory of diffraction

(UTD) (see Appendix C), the ray based approach is well suited for the study
of radio wave propagation. The GO considers that the energy is radiated
along infinite tubes called rays. These rays define the propagation directions
and can reflect or refract on all the surfaces they encounter. The use of UTD
complements the GO as it introduces the diffracted rays and ensures the
continuity of the field on regions where the GO predicts a non-existence of a
field.
Before the determination of the propagated field, we need to find the rays
from which the GO and the UTD will be applied. This step of ray finding can
use various techniques. There are two main techniques of ray determination: ray
launching is considered to be a forward technique and ray tracing is considered
to be a backward technique [SAR 03] (see Appendix D). When the rays are
obtained, the propagated field can thus be calculated from the transmission to
reception side.
The ray based approach needs less numerical resources than the FDTD and
MoM approaches for wave propagation prediction. Consequently, it is mostly
used for deterministic propagation modeling [IKE 91, MCK 91, TAM 95,
TAN 95, SAN 96, CHE 96, RIZ 97, VIL 99, ZHA 00]. Such approaches are
not valid for low frequencies (typically lower than 100 MHz), when the size
of objects interacting with rays become small or have the same order as the
wavelength.
An improvement of the propagation prediction with the ray based approach

can be made by combining it with exact approaches such as those using FDTD
or MoM.
4.3. Specificity of deterministic modeling in UWB
When studying narrowband transmission techniques using deterministic

software, much is made of the power loss of the transmitted signal. So, modeling
allows the establishment of radio coverage maps on any considered environment
at a given central frequency corresponding to the transmitted narrowband
signal. Indeed, we consider that the bandwidth is narrow enough to only focus
the propagation study on the central frequency.
In UWB, the frequency components of the transmitted signal cover a

vast bandwidth which can reach 7.5 GHz. So, the performed simulation will
necessarily address all the phenomena appearing on the entire bandwidth.
Thus, we cannot simply focus on the propagation around the central frequency.
This implies that we have to consider in the synthesized link both antennas
and material properties of the considered environment on the whole covered
frequency band.
Another particularity of the UWB deterministic modeling is the need to

consider the phase information related to the propagation as well as combining
the effects of antennas and material properties. This phase information is
important as it allows the identification of potential deformations (such
as attenuation, dispersion, etc.) appearing on the received signal during
the transmission. Deterministic modeling can thus be used in specific
environmental configuration in order to better understand the effects of
physical phenomena observed on results of channel propagation measurement
campaigns.
So, the use of site specific tool for deterministic modeling allows us on the
one hand to study the waves propagation in various types of environments and
in specific configurations, and on the other hand to physically understand the
effects observed on measurements.
4.4. Overview of UWB deterministic modeling
In this section, the four principal deterministic models which appear in the
literature are described, with a main focus on their respective specificities.
4.4.1. Qiu model
Some authors consider the study proposed by Qiu [QIU 02] on the diffraction
undergone by an UWB signal as a UWB deterministic model. In the study made
by Qiu, he mainly focuses on the distortions and dispersions introduced in a
link and on system performances by single and multiple diffractions.
4.4.2. Yao model
The particularity of the model proposed by Yao is that it takes into

account the GO and UTD coefficients in the time domain in order to account
for the effects of the propagation channel interactions [YAO 03a, YAO 03b].
These time domain coefficients allow us to directly express the impulse
response of each interaction [VER 90, BAR 91, ROU 95, YAO 97]. This
approach can appear appropriate for the UWB channel propagation modeling
as the reconstructed signal is a sum of each ray contribution. The contribution
of each ray at the receiver side is a successive convolution of the impulse
responses of each interaction of the considered ray with the transmitted signal.
Nevertheless, a deterministic modeling tool using this approach will need

significant numerical resources. In fact, a numerical operation consisting of
a convolution needs great memory resources. So, although this model gives
acceptable results, its use can become very restrictive in term of calculation
time.
4.4.3. Attiya model
The deterministic model proposed by Attiya [ATT 04] considers the GO

and the UTD interactions coefficients expressed in the frequency domain (see
Appendix C). The received signal is obtained in the time domain thanks to
the use of an inverse Fourier transform. This allows us to directly use all
the classical formulations corresponding to the frequency behavior of the
propagation phenomena.
Another difference between Yao and Attiya’s deterministic model lies in the
consideration of antennas, something Yao neglected entirely. Attiya’s proposed
model takes into account the antennas by inserting analytical formulations of
antenna radiation in the expression of the reconstructed signal. In his model,
he focuses on the case of horn antennas for which he proposes a time domain
characterization technique [ATT 03].
The same impulse response obtained from measurement in an anechoic

chamber between a transmitting antenna and a receiving antenna is
directly applied in the time domain on each ray. For each ray, the relation
corresponding to the effect of ray interactions is obtained after inverse
Fourier transformation. So, the determination of the received signals after
antennas are made by adding all the contributions of each ray. Each of these
contributions is obtained by convolution of the impulse response between
antennas with the relation corresponding to effect of a channel without
antenna. So, it appears that this method shows two limits. The first limit is
the consideration of the impulse response of antennas in order to insert their
behavior in the model. The second limit is the application of the same impulse
response for all rays, although they do not go and reach transmitting and
receiving antennas respectively from the same direction.
4.4.4. Uguen and Tchoffo Talom model
The other deterministic model proposed in the literature is that of Uguen

and Tchoffo Talom [TCH 04, UGU 05]. The first contribution of this model
mainly presents the synthesis of a received signal which adopts a formalism
considering the channel ray by ray [UGU 02]. In this first contribution,
some important elements of the transmission channel were not considered,
such as the indoor multi-layered materials and antennas. Eventually, a more
complete description was made [TCH 04, UGU 05]. In comparison to the first
proposition, the introduced improvement concerns a better description of the
consideration of each channel element in the model.
This model is quite similar to the one proposed by Attiya [ATT 04].
However, unlike the Attiya model, the antennas are better taken into
account in the model proposed by Uguen and Tchoffo Talom. Moreover,
the construction of each ray contribution is made separately, which allows
us to naturally access to the direction of departure (DoD) and direction of
arrival (DoA) information. So, each ray is affected by the antenna functions
corresponding to the actual DoD and DoA of the ray, defined in the frequency
domain.
4.5. Illustration of a deterministic model formalism
In this section, we illustrate the UWB deterministic modeling by presenting

the theoretical formalism of the model proposed by Uguen and Tchoffo Talom
[TCH 04, UGU 05]. This model allows us to synthesize a received signal for an
indoor UWB link.
We consider the application of an impulse signal p(τ ) on the transmitting

antenna feeding port. As explained previously, it is not helpful to sum all the ray
contributions before the transformation from the frequency domain to the time
domain (see Appendix E). It is better to apply the transformation separately
on each ray and then to sum each ray contribution directly in the time domain.
This approach allows us to reduce the number of frequency points to consider
and avoids unneeded frequency domain truncation of the complex exponential
corresponding to the phase shift introduced by the propagation.
The signal after the receiving antenna is obtained by summing all the
vectorial and complex component of the field coming from the rays arriving
on the receiving antenna. After the frequency sweep on the band of interest,
the signal is expressed in the time domain using an inverse discrete Fourier
transformation. So, this signal can be considered as the convolution of the
signal p(τ ) with the SISO impulse response of the propagation channel.
The presentation of the theoretical formalism of the deterministic model is

made by the detailed synthesis of the received signal. This detail mainly focuses
on the ray by ray treatment of the propagation and antenna information as well
as the proper consideration of antennas in the modeling.
4.5.1. Received signal synthesis
The received signal r(τ ) is constructed as the sum of the appropriately

shifted contribution rk (τ ) of the Nray rays obtained from the ray tracing step
(see Figure 4.1) by:

Nray

r(τ ) = rk τ − τk [4.1]
k=1
The parameter τk is the delay corresponding to the free space propagation

of the k th ray.
The non-delayed signal rk (τ ) (see relation [4.2]) corresponds to the

convolution of the transmitted signal p(τ ) applied at the transmitting antenna
port with the non-delayed ray impulse response (RIR) hk (τ ).
rk (τ ) =
hk (τ ) ∗ p(τ ) [4.2]
In relation [4.1], Nray is one of the important parameters for the

reconstruction of the received signal. It determines the realism of the obtained
synthesized signal r(τ ).
Deterministic models using rays are strongly dependent on the number of

significant rays contributing to the total field. So, it is important to use rapid
and appropriate techniques allowing us to obtain the main contributors in LOS
or NLOS situations. Most of the time, this last situation requires us to consider
a great number of rays in order to improve the realism of the synthesized signal.
4.5.2. Ray impulse response without delay
The non-delayed ray impulse response (RIR) is noted hk (τ ) and defined in

the time domain by:

hk (τ ) = f r − τ, ŝrk
ck (τ ) f t τ, ŝtk [4.3]
106
r(τ) τ
=
r̃1 (τ − τ1 ) τ r̃1 (τ)
τ1
r̃2 (τ − τ2 ) τ r̃2 (τ)
τ2
r̃k (τ − τk ) τ r̃k (τ)
τk T FFT
Figure 4.1. Building of received signal r(τ ) by the sum of shifted rk (τ )
f r (τ, ŝrk ) is a line vector corresponding to the impulse response of the

received antenna in the arrival direction ŝrk of the k th ray.
f t (τ, ŝtk ) is a column vector corresponding to the impulse response of the

transmitted antenna in the departure direction ŝtk of the k th ray.

ck (τ ) is a 2×2 matrix corresponding to the consideration of attenuations and
distortions introduced by the reflection, transmission or diffraction interactions
appearing on the k th ray when propagating through the channel. This term does
not consider the delay introduced by propagation τk . This delay is directly used
when adding all the rk (τ ) at the appropriated time position (see Figure 4.1).
ŝtk and ŝrk are the directions of departure (t) and arrival (r) of the k th ray.
They are respectively the couples of polar coordinates (θkt , φtk ) and (θkr , φrk ) in
the entire spherical base.
Nevertheless, as reflection, transmission and diffraction phenomena have a

simple analytical expression in the frequency domain, rk (τ ) is obtained from
an inverse Fourier transform applied on its frequency expression R k (f ):

k (f ) = Fr ∗ f, ŝrk C
R k (f ) Ft f, ŝt P (f ) [4.4]
k
P (f ) corresponds to the Fourier transform of the transmitted signal.
Ft,r are the complex vectors corresponding respectively to the transmitting

and receiving antenna frequency behavior. They allow us to consider the
directivity Dt,r , the return loss Γt,r , the radiation efficiency η t,r and the
antenna polarization state Ut,r [LO 88]:

Ft f, ŝtk = Gt1 f, ŝtk Ut f, ŝtk [4.5]

λ
Fr f, ŝrk = −j Gr2 f, ŝrk Ur f, ŝrk [4.6]
4π
2
Gt1 f, ŝtk = η t (f ) 1 − Γt (f ) Dt f, ŝtk [4.7]

Gr2 f, ŝrk = η r (f ) Dr f, ŝrk [4.8]

Ut,r f, ŝt,r
k = Uθt,r θ̂t,r + Uφt,r φ̂t,r [4.9]
The terms Ut,r are unitary vectors defined in the basis corresponding to the
directions of departure ŝtk and arrival ŝrk for a given ray. These terms follow the
relation:
U∗ U = 1 [4.10]
The term −j 4π
λ
in the expression of Fr corresponds to the integration
operation made at the receiver side.
C k is the 2×2 matrix corresponding to the frequency filter presented by the

propagation channel on the k th ray. This matrix is defined from the GO/UTD
field and is expressed in a vectorial form:
⎡ ⎤
θ,θ θ,φ

k (f ) = Ek = ⎣
C C
C ⎦ [4.11]
Einc
k
C
φ,θ
φ,φ
C
k the field at the receiving antenna and Einc the field coming from the
with E k
transmitting antenna in the direction ŝtk by:
t
Einc
k = F f, ŝk P (f )
t
[4.12]
k = Ek e+j2πf τk
E [4.13]
4.5.3. Ray channel matrix without delay
The ray channel matrix without delay is expressed in a vectorial way by

k . It depends on the expression of the GO and the UTD field by the use
C
of the field Ek (see relation [4.14]). The superscript tilde corresponds to the
extraction of the delay related to the propagation in Ek expression.
k Einc e−j2πf τk
Ek = C [4.14]
k
Typically, there can appear on a k th ray approximately Nk interactions of

reflection, transmission or diffraction nature. So, Ek is expressed by the relation:
1
Ek = Ak Gk e−j2πf τk Einc
k [4.15]
s0k
1
corresponds to the spherical nature of the wave coming from the
s0k
transmitting antenna.
τk is the delay related to the propagation of the k th ray. It is the sum of

delays associated with the distances between the interactions appearing on the
considered ray.

Nk
τk = τki [4.16]
i=0
Ak corresponds to the product of all the divergence factors of each

interaction appearing on the ray. The expression of each divergence Aik is
related to the considered interaction’s nature:

Nk
Ak = Aik [4.17]
i=1
Gk is the interaction matrix representing the consecutive reflections,

transmissions or diffractions appearing on the k th ray. The term Gik of
each interactions is related to the frequency, the polarization or ⊥ and
the reflection, transmission and diffraction incidence angles. These terms
correspond to 2 × 2 matrices and are defined according to the incidence basis
of reflection, transmission and diffraction interaction considered by:
⎡ ⎤
, G
G ,⊥
Gik = ⎣ ⊥, ⎦ [4.18]
G ⊥,⊥
G
In other terms, the matrix Gik corresponds to the R, T or D matrix

according to the interaction nature (see Appendix C).
So, it is necessary to insert, in the global expression of Gk [4.19], the matrix

used to change the basis related to the possible existence of an angle between the
incidence plane of an interaction and the reflection, transmission or diffraction
plane of the previous interaction.
N
out,Nk k
out,i−1
Bk →Brk i Bk →Bin,i t in,1
Gk = M Gk M k G1k MBk →Bk [4.19]
i=2
t in,1
MBk →Bk is a matrix used to change the description of a field from a basis
t
Bk related to a given departure direction of a k th ray to the incoming basis
Bin,1
k of the first ray interaction.
out,Nk
→Brk
MBk is a matrix used to change the field from the outgoing basis
Bout,N
k
k
of the last interaction to the incoming basis Brk of the arrival direction
th
of the k ray.
out,i−1 in,i
MBk →Bk
is a matrix allowing for a k th ray to express the field,
initially in an outgoing basis Bout,i−1
k of the (i−1)th interaction in the incoming
in,i
basis Bk of the i interaction.
th
4.5.4. Described model results
4.5.4.1. Emitted waveform and considered scenario

We consider the impulse waveform transmission p(τ ), which is obtained
using the relation:
p(τ ) = Ep p̂(τ ) [4.20]
with Ep the energy of the impulse signal p(τ ) and p̂(τ ) a Gaussian impulse
normalized in energy.
p̂(τ ) can be expressed using relation [4.21]:

! √
2 2 τ 2
p̂(τ ) = √ sin 2πfc τ e−( β ) [4.21]
β π
" √
2 ln 10 α
Bα,β = α ≈ 0.216 [4.22]
πβ 20 β
where Bα,β is the band of p̂(τ ) given at −α dB, β is a scaling factor allowing
us to adjust the time domain support of p̂(τ ) and fc is the central frequency of
p̂(τ ).
In the frequency domain, the applied impulse signal is given by:
P (f ) = Ep P̂ (f ) [4.23]
! √
β π −[π β (f +fc )]2 2

P̂ (f ) = √ e − e−[π β (f −fc )] [4.24]
2
So, |P (f )|2 is the energy spectral density (ESD) of the impulse signal. The
impulse energy Ep is then obtained by integrating either the square of the
signal p(τ ) or |P (f )|2 , according to the following Parseval-Plancherel relation

[PRO 83]:

Ep = p2 (τ ) dτ = P (f )2 df [4.25]
R R
To specify Ep , relation [4.26] proposed by [UGU 04] can be used:

1MHz
Tr Pmax
Ep = 2 1MHz
[4.26]
σ γmax
So, the impulse energy can be determined with respect to the emission
1 MHz
limits specified by regulation using Pmax , the pulse repetition period Tr ,
1 MHz
the considered modulation given by σ, and γmax the capability of the
transmitting antenna to emit an amplified signal in the channel over a 1 MHz
band.
According to the FCC specifications concerning the transmitted UWB

1 MHz
signal, the authorized maximum PSD is Pmax = −41.3 dBm/MHz. In this
case, Ep is directly expressed by:
Tr 10−4.13
Ep = [4.27]
σ 2 γ 1 MHz
max
Figure 4.2 shows two impulses obtained for a central frequency fc = 4 GHz
and an ideal1 transmitting antenna. We can note that the signal time spreading
conversely increases with the band as well as the maximum level.
The previously described model is applied for an indoor link of a LOS

configuration in the environment represented in Figure 4.3. The rays reported in
the figure are obtained using a 3D ray determination technique which combines
ray launching and ray tracing [TCH 05a].
The material properties of the environment shown in Figure 4.3 are reported
in Table 4.1. These properties are obtained from material characterizations
[TCH 05a]. The frequency dependence of the materials is introduced by the
permittivity expression in the interactions coefficients (see Appendix C). In
this table, we can note that the floor and the ceiling are in reinforced concrete
because the environment is the ground floor of a house with two levels.
1. An ideal antenna is an omnidirectional isotropic antenna with a unitary gain in all

directions.
(a)
(b)
Figure 4.2. Impulses applied on an ideal transmitting antenna with fc = 4 GHz:

B10 dB,β = 0.5 GHz (a) and B10 dB,β = 2 GHz (b)
Figure 4.3. 3D rays obtained for an indoor link in LOS configuration
Elements Materials Material properties

r r μr μr σ (S/m) Δδ (m) e (cm)
Wall brick 3.8 0 1 0 0.05 0 10
Windows glass 3.1 0 1 0 0 0 1.5
Doors wood 2.84 −0.02 1 0 0 0 3
Ceiling reinforced concrete 7.7 0 1 0 1 0 10
Floor reinforced concrete 7.7 0 1 0 1 0 —
Table 4.1. Properties and structure of various elements of the considered environment
4.5.4.2. Channel matrix of each emitted waveform in the LOS case

Considering the two waveforms (see Figure 4.2) for p(τ ) and the LOS
configuration (see Figure 4.3), Figures 4.4 and 4.5 represent the four
components of the channel matrix c(τ ):

cθ,θ (τ ) cθ,φ (τ )
c(τ ) = φ,θ [4.28]
c (τ ) cφ,φ (τ )
These illustrations are an artifice of representation which allows us to report

on the same time axis all the contributions of the rays associated with each
component of the matrix c. Each component of c(τ ) is obtained by adding the
electric field contribution of each of the Nray for the considered component.
This representation allows us to underline the time domain behavior of the

Nray ray fields arriving at the receiver without considering any transmitting or
receiving antenna.
We can note in these figures that the field level is weak on the cross
components (cθ,φ and cφ,θ ). So, in order to obtain the received signal, the
antennas considered at the transmitting and receiving sides will project these
c(τ ) matrix components according to their characteristics [TCH 05b]. The
shape of the final obtained signal will strongly depend on the components of
c(τ ) modified by the chosen couple of antennas.
Moreover, when the frequency band of the impulse p(τ ) decreases,

each contribution of all rays can no longer be clearly isolated. For a band
B10 dB,β = 500 MHz and for each component of c(τ ), the obtained signal
corresponds to a compact grouping of the ray contributions arriving at a
time lap with approximately the same width as the transmitted signal. This
does not allow us to isolate the contribution of each ray but underlines the
progressive rise of shadowing as the transmitted signal bandwidth decreases.
4.5.4.3. Received signal with ideal antennas

In this part, from the components of channel matrix c (see Figure 4.5) and
for the LOS case as well as the transmitted pulse p(τ ) of Figure 4.2(b), we
focus on the reconstructed waveform obtained in the case where the antennas
at transmission (Tx) and reception (Rx) sides are ideal omnidirectional, with
unitary gain.
Figures 4.6(a) and 4.6(b) show the reconstructed signal r(τ ) obtained
respectively for a couple of antennas polarized along θ and φ.
In the considered configurations, the transmitting and receiving antennas

show a good accordance in term of polarization. So, the reconstructed signals
(see Figures 4.6(a) and 4.6(b)) are respectively a projection of the channel
matrix c(τ ) components cθ,θ (see Figure 4.5(a)) and cφ,φ (see Figure 4.5(d)).
In the case of a real antenna, the received signal r(τ ) will probably be a
less simple projection of different matrix c components. So, it will no longer be
easily identified to one of the matrix c(τ ) components.
One of the interests of the deterministic model detailed in section 4.5 is the
ease of identifying DOA and DOD information. This is because the signal r(τ )
is reconstructed thanks to the rays obtained after ray tracing which directly
gives the information corresponding to the direction of the rays. The other
interest is that no sum is beforehand applied on the contributions of each ray.
So, it is easy to extract the contribution associated with each ray.
(a) (b)
(c) (d)
Figure 4.4. Channel matrix component c(τ ) for the scenario of Figure 4.3 with the
impulse p(τ ) of Figure 4.2(a): cθ,θ (τ ) (a), cθ,φ (τ ) (b), cφ,θ (τ ) (c) and cφ,φ (τ ) (d)
115
116
(a) (b)
(c) (d)
Figure 4.5. Channel matrix component c(τ ) for the scenario of Figure 4.3 with the
impulse p(τ ) of Figure 4.2(b): cθ,θ (τ ) (a), cθ,φ (τ ) (b), cφ,θ (τ ) (c) and cφ,φ (τ ) (d)
(a)
(b)
Figure 4.6. Received signal r(τ ) for the LOS scenario of Figure 4.3, the pulse p(τ )
of Figure 4.2(b) and the ideal omnidirectional antennas at Tx and Rx sides:
Tx-Rx antennas polarized along θ (a) and Tx-Rx polarized antennas along φ (b)
Figure 4.7 illustrates the extraction of DOA information corresponding to

the signal reported in Figure 4.6(a). We can note that the obtained rays are
well described in 3D. The direct path is clearly identified. It arrives at the
receiving antenna side in the direction corresponding to the angles θ ≈ 95◦ and
φ ≈ 245◦ . Here, the θ direction is not 90◦ because the transmitter and receiver
antennas are not placed at the same height.
4.6. Consideration of real antenna characteristics in deterministic modeling
Previously when ideal antennas were considered, we could easily identify

the matrix c(τ ) component corresponding to the signal received after the Rx
antenna. Here, real couple of antenna radiation characteristics are considered.
In Figure 3.12, the used antennas which are called the CMA (Conical Monopole
Antenna) are represented. The characteristic of these antennas are obtained
from measurements performed on 4 π steradian in the Stargate 32 near-field
antenna measurement engine of Satimo (see Figure 4.8) [TCH 05b].
The considered modeling configuration is the case illustrated in Figure 4.3.

The received signal is illustrated in Figure 4.9(a). We can note that the
omnidirectional characteristic of the CMA allows us to obtain a signal different
from those obtained previously with ideal antennas. So, the obtained signal is
no longer easily identified to one component of the channel matrix c(τ ) (see
Figure 4.5).
Figure 4.9(b) illustrates the amplitude and phase contribution applied by

the antennas on each ray. The rays corresponding to the main contribution on
the received signal are represented by a dark line. The contribution of a ray
is thus proportional to the gray intensity used for its drawing. The darkest
rays are those for which the combined directivity of transmitting and receiving
antennas are important. If a rotation is applied along θ in order to fit both
antenna directions of maximum radiation in elevation, the received signal will
have a higher level than in the previous case (see Figure 4.10) and the other
directions will contribute less to the obtained signal.2
We can note that the consideration of real antennas characteristics in the

deterministic model is crucial and contributes to the validation and realism of
the received signal synthesized with the modeling.
2. By comparing the two figures, it seems that the second RI is less dense. In fact, it
is the first and main ray which has more energy in this case.
Figure 4.7. Illustration of r(τ ) DOA (see Figure 4.3) and

119
the Tx-Rx ideal antennas polarized along θ

Figure 4.8. Satimo Stargate 32 near-field antenna measurement engine
4.7. Building material effects on channel properties
The previously described studies were made for materials whose properties
are reported in Table 4.1. So, the received signal reported in Figure 4.11(a)
is obtained for a LOS link (see Figure 4.3), the waveform transmission in
Figure 4.2(b) and the use of a couple of CMA antennas. This signal is the
same as that reported in Figure 4.9(a).
Considering the properties reported in Table 4.2, the effect of building

material properties on the received signal synthesized by deterministic modeling
is underlined. The new adopted properties correspond to a reduction of the
previous wall thickness values (see Table 4.1), to a change of window structure
and especially to an increase of floor and ceiling conductivity. In fact, the
reinforced concrete used for constructing floors and ceilings at various stages of
house building are made with metallic rods which can contribute to the increase
in conductivity.
The signal received with these new material properties shows important
differences in comparison to the previous signals (see Figure 4.11). There are
rays arriving after the direct path which show a higher level. These rays are
those which connected the transmitter and receiver and reflected on the floor or
the ceiling. So, the conductivity increase σ of the reinforced concrete introduces
a high level contribution on the received signal for the rays which interact with
the ceiling or the floor.
(a)
(b)
Figure 4.9. Received signal and ray contributions corresponding to the CMA
antennas: received signal after the receiving antenna (a) and 3D rays
with their corresponding contribution (b)
(a)
(b)
Figure 4.10. Received signal and ray contributions corresponding to the CMA
antennas (rotation of 35◦ in θ for both Tx and Rx antennas): received signal after
receiving antenna (a) and 3D rays with their corresponding contributions (b)
(a)
(b)
Figure 4.11. Received signal obtained with CMA antennas for two configurations
of building material properties: case of materials in Table 4.1 and
case of materials in Table 4.2
Elements Materials Material properties

r r μr μr σ (S/m) Δδ (m) e (cm)
Walls brick 3.8 0 1 0 0.05 0 7
Windows glass 3.1 0 1 0 0 0 0.5
air 1 0 1 0 0 0 0.5
glass 3.1 0 1 0 0 0 0.5
Doors wood 2.84 −0.02 1 0 0 0 3
Ceiling reinforced concrete 7.7 0 1 0 10 0 10
Floor reinforced concrete 7.7 0 1 0 10 0 —
Table 4.2. Properties and structure of various building environment

considered for the study of material influence
In addition to the antennas, the change of material properties strongly

affects the received signal shape. In order to increase the realism of the
waveforms obtained with a deterministic modeling tool, the characterization
of building materials is needed to insert into the modeling the good
material properties for the environment considered in link modeling
[KRA 93, SAT 95, HUA 96, COU 98, MUQ 03a].
4.8. Simulation and measurement comparisons
In this section, comparisons are made between the results of the realized
measurements (see section 3.4.2.2) and the performed simulations using the
deterministic model described in section 4.5. These comparisons allow us to
evaluate the impact of the antenna characteristics in the described model and
the received signal building for various LOS links.
These comparisons consider a transmitted signal covering a band of

2–6 GHz. As the measurements were made in the frequency domain, no
impulse signal is applied. Nevertheless, the transmitted signal corresponds to
a sinc function in the time domain.
4.8.1. Evaluation of real antenna consideration
To evaluate the consideration of antenna radiation characteristics in

the deterministic model described in section 4.5, direct link measurements
were made in an anechoic chamber. This configuration allows us to be free
from multipath and to obtain a direct link with only one ray [TCH 06]. For
this evaluation, two antenna couples are used: directive horn antennas and
omnidirectional CMA antennas.
In the anechoic chamber, the antennas are placed at the same height and
directed, in the case of horn antennas, so that both antenna directions of
maximum gain face each other. This configuration is reproduced in simulation
for the same departure and arrival directions as in the measurement.
The simulation uses antenna characteristics data obtained from wideband
measurement made in the Stargate 32 engine [TCH 05b].
Figure 4.12 illustrates the impulse response and transfers functions obtained
with measurement and simulation in the case of the couple of horn antennas.
We can observe a good matching between measurements and simulation

results. Although there is only one ray, the obtained transfer functions are
not constant. This can be explained by the frequency dependencies of antenna
characteristics which affect the channel transfer function.
Figure 4.13 illustrates the impulse responses and transfers functions

obtained with measurement and simulation in the case of the couple of
CMA antennas. We can also notice in this case a good matching between
measurements and simulation results.
The comparison of these two configurations testify to the relevance of

antenna characteristics in deterministic link modeling. Indeed, according to
their characteristics, antennas will affect the shape of the received impulse
responses. Moreover, the comparison between measurement and simulation
validates the adopted vectorial insertion of antenna characteristics in the
described deterministic model.
4.8.2. Evaluation of impulse response reconstruction
To evaluate the impulse response reconstruction, the focus here is on LOS

links for which the measurement configurations are described in section 3.4.2.2.
In each of the shown figures, the measured and simulated impulse responses as
well as transfer functions are superimposed. The measurement and simulation
results are respectively represented in gray and black.
These illustrations concern four typical LOS situations extracted from 126
measurements points made in the environment described in section 3.4.2.2.
Figures 4.14 and 4.15 represent the results obtained considering the four
following positions corresponding to 1.14 m, 4.38 m, 6.78 m and 8.94 m
distances.
The measured and simulated transfer functions show a great similarity

in terms of level and fluctuations in form (see Figure 4.14). The same
high frequency decrease is observed in measurement and simulation.
Nevertheless, some differences are presented on the transfer function reported
in Figure 4.14(a). These differences can be explained by the fact that for the
presented results the transmitter was placed near a stone made fireplace. In
simulation, a normal wall made with concrete was considered. Moreover, some
differences can also be observed in phase information. This can be explained
by the rays considered for the modeling, which are insufficient to provide
access to the degree of measurement detail.
Although the transfer functions seem different, we can note a great

accordance between measured and simulated impulse responses (see
Figure 4.15). Moreover, the simulated impulse responses are less consistent
in term of rays compared to the measurement impulse responses. This can
be explained by the simplified description of the environment used in the
modeling. This simple description of the environment leads us to discard
furniture and the inappropriate material properties for the considered
environment walls.
We may keep in mind that in such situations of very good concordance

between simulation and measurement results, the simulation represents the
benefit of giving additional information like departure and arrival angles. The
channel is here known on various dimensions and with less cost than using a
sounder with a single sensor.
4.9. Conclusion
For a long time, the deployment of narrowband technologies has used the
deterministic modeling of the radio propagation channel to establish coverage
maps. With the use of UWB technology, some deterministic models have
been proposed for signal prediction in new context. These models allow signal
prediction and the rapid study of various environments with low costs in
terms of channel transmission effects on the UWB link.
From all the deterministic models presented, those which use frequency
domain formalisms clearly seem better suited for the study of UWB wave
propagation. More especially, the detailed Uguen and Tchoffo Talom model
allows us to easily insert the antenna radiation information in the model.
Moreover, this model naturally enables the access to DoD and DoA information
as the electromagnetic calculations are performed ray by ray. The contributions
of all the rays are then summed to obtain the overall received signal.
When deterministic modeling is used in UWB, it is important to take into

account antenna characteristics and to use appropriate material properties for
the considered environment. The reconstructed signal is strongly dependent on
the number and relevance of the rays used in modeling. This firstly depends
on the techniques used for ray determination and secondly on the details with
which the environment is described.
(V)
delay (ns)
(a)
Magnitude
(level)
Freq (GHz)
Phase
(radian)
Freq (GHz)
(b)
Figure 4.12. Measurement (gray) and simulation (black) received signals for
the couple of horn antennas: (a) impulse response and (b) transfer function:
magnitude and phase
(V)
delay (ns)
(a)
Magnitude
(level)
Freq (GHz)
Phase
(radian)
Freq (GHz)
(b)
Figure 4.13. Measurement (gray) and simulation (black) received signals for
the couple of CMA antennas: (a) impulse response and (b) transfer function:
magnitude and phase
Magnitude Magnitude
130
(leve l)
(leve l)
Freq (GHz) Freq (GHz)
Phase Phase
(radian)
(radian)
(a) (b)
Magnitude Magnitude
(leve l)
(leve l)

Phase Phase
(radian)
(radian)
(c) (d)
Figure 4.14. Measured (gray) and simulated (black) transfer functions for 4 distances
between Tx and Rx: (a) d = 1.14 m, (b) d = 4.38 m, (c) d = 6.78 m and
(d) d = 8.94 m
(V)
(V)
delay (ns) delay (ns)
(a) (b)
(V)
(V)

(c) (d)
Figure 4.15. Measured (gray) and simulated (black) normalized impulse responses for
4 distances between Tx and Rx: (a) d = 1.14 m, (b) d = 4.38 m, (c) d = 6.78 m
and (d) d = 8.94 m
131
Chapter 5
Statistical Modeling of the UWB Channel
Deterministic modeling makes a relatively accurate reproduction of the

UWB channel properties possible for a given configuration in a known
environment. Assuming that the elementary propagation phenomena are well
modeled, the obtained impulse responses can be very realistic. The main
constraints linked to this type of model lie in their long calculation time and
the necessity of describing the considered environment in detail.
Statistical models represent an interesting alternative for the simulation of

UWB communication systems. They consist of reproducing a possible behavior
of the propagation channel in a given type of environment. They are based on
a large number of measurements, from which each parameter of the model is
defined using a statistical law. These models enable the random generation of
different impulse responses.
This chapter presents the statistical modeling of the UWB channel

through a practical approach. The principles of UWB propagation channel
characterization is first illustrated from a series of measurements performed in
an indoor office environment. For each step of the characterization process,
experimental results are compared to the main anaylses published in the
literature. Following a description of the different statistical models of the
UWB channel, a full channel model is detailed. Its practical conception
is based on a set of experimental characteristics. In particular, advanced
techniques are presented for the modeling of spatial and temporal variations
of the UWB channel.
5.1. Experimental characterization of channel parameters
In this section, the characterization process of the UWB propagation

channel is illustrated from a measurement campaign covering the whole FCC
frequency band. The description of this measurement campaign is given
in section 3.4.2.1. This study also compares the experimental results from
different analyses of the UWB channel.
5.1.1. Propagation loss
5.1.1.1. Frequency propagation loss

One of the channel properties characteristic to UWB signals is the
power decay observed with increasing frequency. In an ideal free space
configuration, the propagation loss is given by the Friis formula, as
recalled in equation [2.34]. This formula implies that for a given distance
d, the channel power transfer function (see equation [2.36]), expressed
in dB, should follow a frequency variation in the form of −20 log(f ). It
should be noted that this frequency dependence is linked to the effective
area of an isotropic antenna and is not strictly speaking a characteristic
of the propagation channel. In the literature, the attenuation of the
received power as the frequency increases has been observed in different
studies dedicated to antennas [KOV 03, HOF 03], or to the UWB channel
[CHE 02, KUN 02a, ALV 03, BAL 04b, HAN 04, CHO 05, PAG 05]. As an
example, Table 5.1 presents different estimates of the frequency dependent
path loss exponent Nf . We may observe a relative diversity in the obtained
values, as some of the authors even reported an increase in the received power
with increasing frequency [KOV 03]. This may be explained by the high
dependence between the parameter Nf and the measurement antenna.
Measurement Whyless.com Aalborg University Instit. for Samsung

campaign [KUN 02a] [KOV 03] Infocomm [CHO 05]
Research
[BAL 04b]
Nf 1.6 to 2.8 −0.3 to 5.5 1.0 to 3.7 2.0 to 3.1
Table 5.1. Estimation of the frequency dependent path loss exponent for different
analyses of the UWB channel. The published values have been adapted to the
definition of the parameter Nf used in this book
In order to estimate the frequency dependent path loss, we studied the

channel attenuation at different frequencies regularly spaced between 4 GHz
and 10 GHz. For each measurement location, the path loss P L(f, d) was
Statistical Modeling of the UWB Channel 135
extracted from the power transfer function at the selected frequencies. In order
to remove the distance dependency, each power transfer function was initially
normalized by arbitrarily setting the attenuation of the total received power
over the whole measurement bandwidth to 0 dB. The resulting normalized
path loss P Lnorm (f ) can be compared to a model in the form:

f
P Lnorm (f ) = P Lnorm f0 + 10Nf log + S(f ) [5.1]
f0
where S(f ) represents a residual term with zero mean expressing the difference
in dB between the measurement and the model.
10
8
6
(f) (dB)
4
norm
2
PL
0
-2
-4
3 4 5 6 7 8 9 10 11
Frequency (GHz)
Figure 5.1. Average normalized path loss vs. frequency
The normalized path loss averaged over all measurement locations is

represented as a function of frequency in Figure 5.1. It should be noted that
the antenna gains GT (f ) and GR (f ), measured in an anechoic chamber, were
subtracted from each power transfer function at the selected frequencies, prior
to path loss calculation. For each antenna, the antenna gain was selected by
taking the direction of the direct transmitter-receiver path into account. This
approach may seem simplistic a priori, as the total power is not only received
via the direct path, but also via numerous multipaths arriving from other
directions. However, with no further knowledge of the departure and arrival
direction of these secondary paths, this method makes a sensible compensation
of the antenna effect possible. We may observe that the dispersion of the
measured plots around the linear approximation is reduced to a standard
deviation of σS = 0.6 dB. The frequency dependent path loss exponent is
Nf = 2.28. With respect to the error level inherent to the measurement
and calculation methods, this value may be considered close to the theory
(Nf = 2). Hence, we recommend using the theoretical frequency loss of
20 log(f ) in the modeling of UWB propagation channels, as suggested in
[BUE 03].
5.1.1.2. Distance propagation loss

In the previous section, we observed that the used antennas may have a
non-negligible impact on the radio parameters. In the following analyses, in
order to minimize this effect, the collected data were corrected by accounting for
the gain of both antennas, measured in the direction of the transmitter-receiver
path. The antenna radiation patterns, measured every GHz, were interpolated
in frequency.
According to the previous analysis on the frequency path loss, all

measurements were fitted to a general formula as follows:

f d
P L(f, d) = P L f0 , d0 + 20 log + 10Nd log + S(f, d) [5.2]
f0 d0
where f0 represents the central frequency of 6.85 GHz and d0 is an arbitrary

distance of 1 m.
40
LOS
50
NLOS
Attenutation (dB)
60
70
80
90
100
110
0.6 0.8 1 2 3 4 5 6 8 10 20 30 40
Distance (m)
Figure 5.2. Path loss vs. distance. Each point represents

the median attenuation in the FCC band
For each measurement location, the accounted path loss is the median
attenuation in the FCC band, after shifting each measurement to the reference
frequency f0 . Path loss exponents were obtained by linear fit in each LOS and
NLOS situation. Figure 5.2 presents the obtained results. The measurement
plots mainly follow a linear decay in log-scale, which corresponds to an
exponential decay of the received power with respect to the distance. In the
LOS situation, a path loss exponent Nd = 1.62 was recorded, with a standard
deviation σS = 1.7 dB.1 In the NLOS case, the measurement plots are
somewhat more dispersed, with a distance dependent path loss exponent Nd
= 3.22, and a standard deviation σS = 5.7 dB.2 The values of the parameter
P L(f0 , d0 ) were respectively evaluated at 53.7 dB and 59.4 dB. Table 5.2
shows that these values are in line with other analyses of the UWB channel
published in the literature. For comparison purposes, the UWB path loss
parameters proposed by the standardization organization ITU [ITU 04] are
also reported.
More details regarding the influence of frequency on the path loss

coefficients are given in [PAJ 07]. In general, these parameters undergo a
negligible variation when they are observed on partial frequency bands within
the 3.1–10.6 GHz band.
5.1.2. Impulse response characterization
Figure 5.3 presents typical PDP measured in LOS and NLOS configurations,
as well as one of the 90 impulse responses. The delay on the x-axis has been
converted in path length in meters, to ease the interpretation of the main paths.
In both LOS and NLOS situations, we may observe one or several clusters,
corresponding to a main echo followed by an exponential decay of diffuse power.
In the LOS case, walls or pieces of furniture in the vicinity of the radio link
generate significant reflected or diffracted echoes, which explains the presence
of peaks in the PDP. An attenuation of 10 dB to 20 dB has thus been observed
between the power of the main path of each cluster and the power of the
secondary paths. The general shape of the PDP is globally smoother in the
NLOS case.
5.1.2.1. Delay spread

The RMS delay spread τRM S was calculated for each measured PDP,
over the whole 3.1–10.6 GHz band. In order to minimize the effect of noise,
a threshold placed 20 dB below the maximum value of the PDP was used.
1. In the LOS case, the value of Nd may be lower than in the theoretical case of free
space (Nd = 2). This is due to the waveguide effect, frequently observed in indoor
configurations (see section 2.2.2).
2. This significant value of the standard deviation in the NLOS case may be explained
by the diversity of NLOS configurations, as the obstruction between the transmitter
and the receiver may result from a single plasterboard or several concrete walls.
Measurement Nd σS (dB) P L(f0 , d0 ) (dB)

campaign LOS NLOS LOS NLOS LOS NLOS
UltRaLab [CAS 01] 2.4 5.9
AT&T Labs - MIT [GHA 02a] 1.7 3.5 1.6 2.7 47 51
Time Domain Corporation 2.1 3.55
[YAN 02]
Intel Labs [CHE 02] 1.72 4.09 1.48 3.63
Whyless.com [KUN 02a] 1.58 1.96
UCAN - CEA LETI [KEI 03] 1.6 to 3.7 to
1.7 7.2
Ultrawaves - Oulu University 1.04 to 3.18 to
[HOV 03] 1.80 3.85
UCAN - Cantabria University 1.4 3.2 to
[ALV 03] 4.1
New Jersey Instit. of Tech. 1.55 to 0.77 to
[DAB 03] 1.72 1.98
NETEX - Virginia Tech 1.58 to 2.41 to 1.6 to 3.3 to
[MUQ 03b] 1.60 2.60 1.9 6.1
NETEX - Virginia Tech 1.3 2.3 to 2.8 to 2.8 to
[BUE 03] 2.4 3.6 5.4
AT&T - WINLAB [GHA 03b] 2.01 to 2.95 to 2.3 to 3.8 to 43.7 to 47.3 to
2.07 3.12 3.2 4.1 45.9 50.3
Hong Kong University [LI 03] 1.8 3.4 0.6 3.2
Ultrawaves - University of 1.92 3.66 1.42 2.18 48.68 47.24
Rome “Tor Vergata”
[CAS 04a]
Instit. for Infocomm Research 1.8 1.8 to 1.5 2.4 to 36.6 46.4 to
[BAL 04b] 2.1 4.2 52
ETH Zürich [SCH 04] 1.2 to 2.2 49 to 45
1.6 51
Samsung [CHO 05] 1.18 to 2.18 to 0.93 to 1.43 to 46.5 to 41.3 to
2.48 2.69 1.50 4.69 50.1 47.3
France Telecom - INSA 1.62 3.22 1.7 5.7 53.7 59.4
[PAG 06b]
ITU Recommendation 1.7 3.5 to 7 1.5 2.7 to 4
[ITU 04]
Table 5.2. Estimate of the distance dependent path loss for different analyses
of the UWB channel
(a)
0 IR
PDP
Relative power (dB)

-10
-20
-30
-40
-50
0 20 40 60 80
Distance-delay (m)
(b)
0 IR
PDP
Relative power (dB)
-10
-20
-30
-40
-50
0 20 40 60 80
Distance-delay (m)
Figure 5.3. Typical PDP and impulse response.

(a) LOS and (b) NLOS configurations
Over the whole set of measurements performed in LOS configuration, the

mean delay spread is τRMS = 4.1 ns, with a standard deviation στ = 2.7 ns.
In the NLOS configuration, the mean delay spread is τRMS = 9.9 ns, with
a standard deviation στ = 5.0 ns. Table 5.3 presents the results published
in other analyses of the UWB radio channel. Values obtained for τRMS
are in accordance with some of the previous experiments, although this
parameter may vary from one experiment to the other. This is due to the
high sensitivity of this parameter to the measurement environment and to the
used experimental setup. The large scope of values in the published results
may also be explained by the different thresholds used when calculating τRMS
[CAS 04a].
140
Measurement τRMS (ns) στ (ns)

campaign LOS NLOS LOS NLOS
UCAN - CEA LETI [KEI 03] 10.9 to 12.2 9.9 to 22.4 1.9 to 2.1 1.4 to 9.5
Time Domain Corporation (DSO) 5.27 8.78 to 14.59
[PEN 02]
Time Domain Corporation (correlator) 5.72 5.22
[YAN 02]
AT&T Labs - MIT [GHA 02a] 4.7 8.2 2.3 3.3
New Jersey Instit. of Tech. [DAB 03] 7.71 to 17.34
NETEX - Virginia Tech [BUE 03] 0.53 to 4.55 2.30 to 18.50
AT&T - WINLAB [GHA 03a] 3.38 to 5.49 7.31 to 8.15 1.58 to 1.63 2.45 to 3.47
Hong Kong University [LI 03] 19.9 14.3 1.8 2.8
Lund University [KAR 04b] 28 to 31 34 to 40

Instit. of Infocomm Research [BAL 04b] 15.6 18.7 to 23.6
ETH Zürich [SCH 04] 21.08 to 53.62 31.11 to 74.08 1.63 to 3.37 1.87 to 7.04
Samsung [CHO 05] 12.48 to 14 26.51 to 38.61 1.53 to 1.87 5.22 to 8.03
France Telecom - INSA [PAG 06b] 4.1 9.9 2.7 5
Table 5.3. Estimate of the delay spread for different analyses of the UWB channel
In order to observe the evolution of these parameters with frequency, the

mean values of τRM S calculated for seven partial bands of 528 MHz each are
represented in Figure 5.4 as a function of the central frequency of each band.
For each band, the value of στ is represented by the length of the vertical line.
The values of τRMS and στ obtained in each partial band are relatively close
to the values calculated from the global UWB frequency band. As can be seen,
the delay spread is not affected by the frequency. This was also observed by
the University of Oulu from measurements performed in the 1–11 GHz band
[JAM 04].
15
RMS delay spread (ns)
10
LOS NLOS
0
3 4 5 6 7 8 9 10 11
Central frequency (GHz)
Figure 5.4. Mean delay spread for different partial bands. The length of the vertical
line represents the corresponding standard deviation
5.1.2.2. Power delay profile decay

Exponential decay constants
The typical PDP presented in Figure 5.3 shows that the received power
is grouped in different clusters, corresponding to the main propagation
echoes. The decay of the received power with increasing delay is generally
characterized with the inter- and intra-cluster exponential decay constants,
respectively denoted Γ and γ (see section 2.4.1.4).
To evaluate these parameters, the time intervals corresponding to the

clusters of each measured PDP are identified by visual inspection. This
technique was also used in [KAR 04a]. In each of the identified interval, a
linear fit is performed at the delays included between the maximum and the
minimum values of the PDP (expressed in dB). This enables the extraction
of the intra-cluster exponential decay constant γ. The inter-cluster decay
constant Γ is obtained using a linear fit on the maximum of each cluster.

In all cases, only the parts of the PDP presenting a power larger than 5 dB
above the noise level were considered. Figure 5.5 illustrates this parameter
extraction.
0 Inter-cluster decay
Intra-cluster decay
Relative power (dB)
-10
-20
-30
-40
-50
-60
0 30 60 90 120
Delay (ns)
Figure 5.5. Extraction of the inter- and intra-cluster exponential decay constants
Among all PDP measured in a LOS situation, between 3 and 8 clusters (5.6
on average) were identified. The mean exponential decay constants have been
evaluated as Γ = 15.7 ns and γ = 7.5 ns. In NLOS configuration, the PDP
encompass between 1 and 4 clusters (2.4 on average). The mean exponential
decay constants have been assessed as Γ = 16.5 ns and γ = 12.0 ns.
Table 5.4 compares these experimental values with results published from
similar experiments. In some analyses [CAS 02, ALV 03], the whole PDP was
considered as a single cluster, which explains the lack of results regarding Γ.
The value of the parameters Γ and γ is generally between 7 ns and 30 ns,
even if some higher values have occasionally been reported [CRA 02, ALV 03,
CHO 04a]. The inter-cluster decay is generally stronger than the intra-cluster
decay. Results published in [KAR 04b] are a particular case where this tendency
is inversed, and where the values of Γ and γ are quite low. This may arise from
the measurement environment, as the experiment took place in a factory. For
this experiment, a dependency of the parameter γ with the delay has been
observed. In [CRA 02], the authors suggest that the constant Γ is linked to
the building architecture, while γ is determined by the objects in the vicinity
of the receiving antenna. The diversity of the sounded environments may thus
explain the variety of the obtained results.
Measurement Γ (ns) γ (ns)
Whyless.com [KUN 02a] 13.6
UltRaLab [CAS 02] 16.1
UltRaLab [CRA 02] 27.9 84.1
UCAN - CEA LETI [KEI 03] 14.5 to 21 9 to 20 6 to 8 5 to 15
UCAN - Cantabria University [ALV 03] 100 125 to 167
NETEX - Virginia Tech 7.1 21 2 8
[BUE 03, MCK 03]
Intel Labs [FOE 03b] 7.6 16 1.6 8.5
Ultrawaves - University of Rome 13 10.83 to 13.97 7 to 58
“Tor Vergata” [CAS 04b]
Lund University [KAR 04b] 2.63 4.94 4.58 5.58
Samsung [CHO 04a] 22.1 to 24.0 36.9 to 51.5 14.3 to 30.8 27.4 to 38.6
Instit. of Infocomm Research [BAL 04b] 27.8 24.6 to 30.4 14.1 25.3 to 33.8
France Telecom - INSA [PAG 06b] 15.7 16.5 7.5 12.0
Table 5.4. Estimate of the PDP exponential decay constants for different analyses
of the UWB channel. The published values have been adapted to the definition
of the parameters Γ and γ used in this book
143
Power decay constants

The assumption of an exponential decay for the cluster and ray amplitudes
was first introduced by Saleh and Valenzuela from their observation of
the indoor wideband radio channel [SAL 87]. However, the analysis of the
results from the measurement campaign shows that exponential decay is not
completely satisfactory to model the slope of the PDP. Indeed, following
this assumption, the PDP expressed in log-scale should present a linear
decrease with increasing delay. The whole PDP as well as each constitutive
cluster should hence fit into a triangular shape. This general shape is not
representative of the experimental observations (see Figure 5.5).
By considering successive echoes of the main path, we may identify two main
sources of attenuation. On the one hand the propagation of the wavefront over
a longer path induces a stronger power loss. On the other hand, delayed echoes
undergo more propagation phenomena, which can be of a different nature, such
as reflection or diffraction. This physical interpretation leads us to model the
multipath attenuation following a similar approach to that used for distance
path loss. We may recall that in this case, the observed attenuation at a
transmitter-receiver distance d is proportional to d−Nd , where Nd represents
the distance dependent path loss exponent. Regarding the different rays of the
impulse response, the length of a propagation path is proportional to its delay.
Hence, we suggest an adaptation of the Saleh and Valenzuela model, where
the cluster and ray amplitudes decrease according to a power function. In the
classical formalism presented in section 2.4.1.4, the amplitude βk,l of the k th
ray in the lth cluster (see equation [2.33]) is replaced by the following formula:
−Ω −ω
2 2 Tl τk,l + Tl
βkl = β11 [5.3]
T1 Tl
where Tl represents the delay associated with the lth cluster and τk,l is the delay
of the k th ray within the lth cluster. The parameters Ω and ω are respectively
called inter-cluster and intra-cluster power decay constants.
As for the case of exponential decay, the values of the parameters Ω and ω
have been assessed by linear fit on the PDP clusters. In each case, the standard
deviation σε of the error in dB between the model and the measurement
has been calculated, in order to validate the proposed approach. Figure 5.6
illustrates the extraction of the parameters Ω and ω. It can be compared to the
results reported in Figure 5.5.
Regarding inter-cluster decay, using a power function instead of an

exponential function leads to a decrease in the average standard deviation
σε from 4.8 dB to 2.9 dB in the LOS case, and from 2.4 dB to 1.7 dB in the
0 Inter-cluster decay
Intra-cluster decay
Relative power (dB)

-10
-20
-30
-40
-50
-60
0 30 60 90 120
Delay (ns)
Figure 5.6. Extraction of the inter- and intra-cluster power decay constants
NLOS case. Regarding the intra-cluster decay, the average modeling error σε
decreases from 1.9 dB to 1.8 dB in the LOS case, and from 1.7 dB to 1.6 dB
in the NLOS case. This validates the proposed model, which is closer to our
experimental measurements. Finally, in the LOS configuration, we observed a
significant power attenuation G between the main path of each cluster and the
following rays, as may be seen in the example of Figure 5.6. This phenomenon
was already observed for UWB channels in [CAS 02] and [KUN 03].
Over the whole set of experimental measurements, we observed the average

values Ω = 4.4 and ω = 11.1 in the LOS case, and Ω = 3.9 and ω = 10.2 in the
NLOS case. In the LOS configuration, the average attenuation G was measured
at 12 dB.
5.1.2.3. Ray and cluster arrival rate

Study of the clusters
In order to estimate the arrival time statistics for a new cluster, we consider
the arrival time of the lth cluster, noted Tl , over the whole set of measured
PDP presenting more than one cluster. This corresponds to the delay of the
maximum value of the PDP in each time interval representing a cluster. The
statistical distribution of the inter-cluster durations ΔT is then studied. ΔT is
calculated as follows:
ΔT = Tl+1 − Tl , l ∈ [1; L − 1] [5.4]
where L represents the number of clusters in the PDP.
Over the whole set of PDP presenting more than one cluster, the average
inter-cluster duration was ΔT = 27.4 ns in the LOS case and ΔT = 40.1 ns in
the NLOS case. The graphs in Figure 5.7 are percentile-percentile diagrams,
representing the experimental distribution percentiles of ΔT on the x-axis, and
the theoretical percentiles of an exponential distribution with parameter Λ on
the y-axis. The best fit leads to a value Λ = 27.41 ns = 36.5MHz in the LOS case,
and Λ = 40.11 ns = 24.9 MHz in the NLOS case. In both cases, the alignment of
the plots on the diagram diagonal shows that the exponential distribution is a
reasonable approximation to modeling inter-cluster duration.
(a) (b)
120 180
LOS NLOS
100 150
Theoretical ΔT (ns)
80 120
60 90
40 60
20 30
0 0
0 20 40 60 80 100 120 0 30 60 90 120 150 180
Experimental ΔT (ns) Experimental ΔT (ns)
Figure 5.7. Percentile-percentile diagrams for the inter-cluster duration.

Experimental percentiles vs. theoretical percentiles corresponding to
an exponential distribution with parameter Λ = 36.5 MHz in the
LOS case (a) and Λ = 24.9 MHz in the NLOS case (b)
Study of the rays

Individual echoes due to the different multiple paths are not directly
observable on the measured PDP. Because of the limited frequency band
of the measurement, each echo is received in the form of an impulse with
non-zero duration. The impulse response is a continuous waveform composed
of the sum of all individual contributions, each one presenting a different
attenuation and a different phase rotation.
Different methods are available to extract the delay and amplitude

information of the main paths constituting an impulse response. The
CLEAN method, initially used in radio astronomy [HÖG 74], was adopted
by different researchers for the characterization of the UWB radio channel
[YAN 02, PEN 02]. This method was modified by the University of South
California for the study of the arrival direction [CRA 02]. The Tokyo Institute
of Technology used an algorithm based on the high-resolution method called
space alternating generalized expectation (SAGE), also allowing the analysis
of the arrival direction [HAN 03]. It should be noted that these two studies
were based on measurements performed with a large number of co-located
sensors.
In order to illustrate the process of ray identification we present

here the frequency domain maximum likelihood (FDML) algorithm
[DEN 03a, LEE 02]. The search of ray locations is performed in an iterative
way, from the representation of the impulse response in its real form h
(τ ).
As an assumption, the channel is thus described in the form:

K

h
(τ ) = βk δ τ − τk [5.5]
k=1
where K represents the number of rays, and βk and τk are the real amplitude
and delay linked to the k th ray. The parameter βk may take negative values,
in order to account for the phase inversion linked to some interactions, such as
the reflections.
At each iteration of the process, a new ray is detected by finding the

correlation peak between the measured impulse response and a template
signal, corresponding to the sounding waveform. Contrarily to the CLEAN
algorithm, the FDML method proposes to optimize the whole set of detected
rays. This requires a minimization of the squared error between the measured
frequency spectrum and the synthetic spectrum, built from the identified
rays. This optimization presents two main advantages. First, it enables the
detection of superimposed rays, which does not always lead to a correlation
peak. Second, the processing is performed in the frequency domain, which
avoids the resolution limitations linked to the sampling rate in the time
domain. More details on adapting the FDML algorithm to accelerate the
calculation time are available in [PAG 05]. The results of the FDML algorithm
allow us to study the arrival time of the k th ray, over the whole set of selected
impulse responses. As with the clusters case, we study the distribution of the
inter-ray durations, defined as follows:
Δτ = τk+1 − τk , k ∈ [1; K − 1] [5.6]
where K represents the number of rays in the impulse response.
Over the whole set of measurements performed in both LOS and NLOS
configurations, the average inter-ray duration was respectively evaluated at
Δτ = 0.168 ns and Δτ = 0.161 ns. Figure 5.8 presents a percentile-percentile
diagram used to compare the experimental distribution of Δτ to an exponential
1
distribution with parameter λ = 0.168
1
ns = 5.95 GHz in the LOS case and
λ = 0.161 ns = 6.19 GHz in the NLOS case. In this case, some differences may
be noted between the theoretical and experimental data, but the exponential
approximation still provides an acceptable fit to the measurements.
(a) (b)
0.8 0.8
LOS NLOS
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
Experimental ΔT (ns) Experimental ΔT (ns)
Figure 5.8. Percentile-percentile diagrams for the inter-ray duration. Experimental

percentiles vs. theoretical percentiles corresponding to an exponential distribution
with (a) parameter λ = 5.95 GHz in the LOS case and (b) λ = 6.19 GHz
in the NLOS case
The experimental values obtained for the cluster arrival rate Λ and the ray
arrival rate λ are compared with the results available in the literature, shown in
Table 5.5. Regarding the cluster arrival rate, the observed values are generally
in the order of 10 to several hundreds of MHz. The average duration between
two clusters is thus in the order of 10 ns to 100 ns. It should be recalled that a
cluster within the PDP corresponds to a main path, arising from transmissions
or reflections on walls, on the ceiling or on the building ground. The parameter
Λ is thus dependent on the structure of the building where the measurement
took place. The ray arrival rate λ presents variable values depending on the
experiment. Indeed, values obtained depend highly on the ray identification
technique that was used in the analysis. For this reason, researchers from the
German institute IMST advise to arbitrarily setting the inter-ray duration at
the temporal resolution value obtained [KUN 02a]. In this case, the ray arrival
rate would be equal to the width Bw of the analyzed band.
5.1.3. Study of small-scale channel variations
The last characteristic of the UWB radio channel studied from the
experimental campaign (see section 5.1) is related to the fast fading of the
impulse response. During the campaign, a rotating arm was used, which
enabled the measurement of the channel impulse response at 90 locations
Measurement Λ (MHz) λ (MHz)
Whyless.com [KUN 02a] 100 to 1000 Bw
UltRaLab [CRA 02] 21.98 434.78
UCAN-CEA LETI [KEI 03] 10 to 25 10 to 800 45 to 180 1500 to 5500
NETEX - Virginia Tech 200 100 1429 714
[BUE 03, MCK 03]
Intel Labs [FOE 03b] 16.7 90.9 2000 2857
Ultrawaves - University of Rome 26 59
“Tor Vergata” [CAS 04b]
Lund University [KAR 04b] 70.9 89.1
Samsung [CHO 04a] 85 to 115 47 to 64 1160 to 1960 1390 to 1790
Instit. of Infocomm Research [BAL 04b] 18.6 2.4 to 13.4 280 270 to 360
France Telecom - INSA [PAG 06b] 36.5 24.9 5946 6194
Table 5.5. Estimate of the average cluster and ray arrival rates for different
analyses of the UWB channel. The published values have been adapted
to the definition of the parameters Λ and λ used in this book
149
situated around a circle of 20 cm in radius. For each measurement, the space

fluctuation statistics were studied by comparing the 90 impulse responses
collected locally. The study consisted of analyzing the statistical distribution
of the amplitude of the received signal at each delay. In the literature,
previous studies showed that this distribution was well represented using a
Nakagami function [CAS 02]. The Nakagami m parameter was thus estimated
for each measured PDP, at delays separated by 0.5 ns on the time scale (see
Appendix B.1.3).
An example of values for the parameter m is given for a specific measurement

in Figure 5.9. As can be noted, the best fitted Nakagami distribution presents
a parameter m close to 1 for the vast majority of the delays within the PDP.
Thus, the amplitude distribution of the impulse response for a given delay may
be correctly described using a Rayleigh distribution. This was observed for most
of the measured PDP. In addition, we may observe that the cases where the
parameter m takes higher values correspond to the main paths of the PDP. In
this case, the value of the parameter m may increase to m = 4 or m = 5, and
up to m = 8 in the extreme cases.
The characterization of amplitude distribution in the UWB impulse

response is a controversial issue in the literature. The Rayleigh distribution
has frequently been observed in the study of UWB radio channel fast
fading statistics [CRA 02, KUN 02a, SCH 04, KAR 04b]. Researchers
from ETH Zürich [SCH 04] and from the Lund University [KAR 04b] also
noted a modification of this distribution for the main path. Other studies
show that the fast fading is well represented by a Nakagami distribution
[CAS 02, CAS 04b, BAL 04b]. In this case, the parameter m may vary with
the delay. However, it has been demonstrated that a theoretical impulse
response presenting a Rayleigh fast fading may be observed as following a
Nakagami distribution, depending on the duration over which the received
power is integrated in the time domain [KUN 03]. Finally, it should be
noted that some analyses recommend that we use a log-normal distribution
[KEI 03, FOE 03b, LI 03, BUE 03, MCK 03] or a Rice distribution [HOV 02].
In practice, we may distinguish between two sources of channel fluctuations

[HAS 94a]. Spatial variations arise when at least one of the antennas is moved in
an otherwise static environment. Temporal variations are due to environmental
modifications and may be observed in fixed radio links. A study of the channel
variations when the transmitting antenna is moved on a 1 m2 grid was presented
in [PAG 04]. Results show that in the case of the UWB channel, the delay
linked to the main paths of the impulse response significantly varies while the
antenna is moved. This shift of the main paths in the time domain is due to
the high resolution of UWB signals and needs to be taken into account when
modeling the channel spatial variations. More details on these observations are
Relative power (dB)

0
PDP
-10
-20
-30
-40
-50
0 20 40 60 80 100 120 140
Delay (ns)
4
Nakagami m
0
0 20 40 60 80 100 120 140
Delay (ns)
Figure 5.9. Nakagami m parameter analysis. PDP measured in an NLOS situation

and parameter value m for each delay
available in [PAG 05]. In indoor environments, the temporal variations of the

propagation channel are mainly due to moving people. This phenomenon is
studied more particularly in the following section.
5.1.4. Effect of moving people
The study of the effect of moving people on the UWB channel is based on the
analysis of a measurement campaign performed in an indoor office environment.
For this campaign, a UWB sounder enabling real time measurements over the
4–5 GHz frequency band was used. During the measurement process, a group
of 1 to 12 people was walking in a corridor in the vicinity of a fixed radio link.
A complete description of this campaign is given in section 3.4.2.3.
5.1.4.1. Observation of temporal variations

A typical impulse response collected during the experiment is given in
Figure 5.10. In this graph, the delay has been converted to path length in meters
to ease the interpretation of multipaths. After the direct path (a), whose length
corresponds to the transmitter-receiver distance (11 m), we may distinguish two
echoes (b) and (c) corresponding to reflections on the corridor walls.
(a)
70
Power attenuation (dB)
(b)
80
(c)
90
100
110
0 10 20 30 40 50 60
Distance-delay (m)
Figure 5.10. Typical impulse response
A first observation of the effect that moving people have on the UWB
channel is given in Figure 5.11. The time varying impulse response is
represented in the case of 12 people walking back and forth through the radio
link. Successive measurements are represented from left to right, while the
vertical axis represents the excess delay converted in path length (m). The
influence that moving people have on the CIR appears clearly on this graph.
The main paths (a) and (b) are regularly obstructed by moving people, during
both forward (t = 15 s to t = 27 s) and backward (t = 63 s to t = 74 s)
displacements along the corridor. At other values of the excess path, we can
observe strong signal fluctuations, with respect to the stationary part of the
diagram.
5.1.4.2. Slow fading

As a first step in the analysis, we observed the slow fading generated by
human beings in the vicinity of the radio link. For this purpose, we studied the
slow temporal evolution of the mean power received in the LOS path in the
presence of people. Fast fading fluctuations were eliminated by averaging the
received power using a sliding window.
Figure 5.12 presents the aggregate effect of several people passing through
the main signal path. In general, the progression of one person through the
LOS path yields a maximum attenuation of about 8 dB, the shadowing effect
lasting for about 4 s. The obstruction duration increases with the number of
people, up to about 15 s for a group of 12 people. In this case, the maximum
attenuation of the mean power is about 15 dB. We clearly see that the
shadowing pattern obtained for groups of people is composed of superimposed
0 0
5 ← Forward mov. → ← Backward mov. →

5
(a)
10
10
Relative attenuation (dB)

15
Distance-delay (m)
15
20 (b)
20
25
30 25
35 30
40 (c) 35
45
40
0 20 40 60 80
Time (s)
Figure 5.11. Time varying impulse response in the case of 12 moving people
individual contributions. However, the particular effect that each person has
on the received signal is not always observable, as, for example, in the case
involving 4 people.
5.1.4.3. Fast fading

In addition to the large-scale fading generated by moving people,
small-scale fading is observed while people are walking in the vicinity of the
radio link. The signal received at a delay corresponding to the main paths
of the impulse response may be thought of as a vector summation of several
multipath components, as would be the case in a typical situation exhibiting
Rician fading. As depicted in Figure 5.13, we may distinguish between two
components of the received signal. The dominant component accounts for the
signal normally received in a static environment. The random component
represents the summation of all waves scattered by moving people.
In order to accurately analyze the fast fading observed during

experimentation, the random component was isolated by filtering. Figure 5.14
represents both magnitude and phase fluctuations of the random component
extracted from the signal received via the main path, when obstructed by
154
1 people 4 people 12 people

5 5 5
0 0 0
-5 -5 -5
-10 -10 -10
-15 -15 -15
-20 -20 -20
Relative power (dB)

-25 -25 -25
Received power
-30 -30 Slow fading -30

0 4 8 12 16 0 4 8 12 16 20 24 28 32 36 0 4 8 12 16 20 24 28 32 36
Time (s) Time (s) Time (s)
Figure 5.12. Typical large-scale fading patterns for the LOS component.
Effect of 1, 4 and 12 people
Q
Random
component
I
Dominant
component
time
Figure 5.13. Vector decomposition of the received signal
12 moving people. This situation corresponds to the measurement depicted

in Figure 5.12 (graph on the right). We may notice that the magnitude of
the small-scale fading presents a high degree of regularity, compared to the
non-stationary large-scale fading pattern.
Using a Kolmogorov-Smirnov testing procedure (see Appendix B.2), the

amplitude distribution of the random component is well suited to a Rayleigh
distribution. This observation could be made for most of the available
measurements, independently of the number of moving people. The signal
received via the main paths of the impulse response, composed of a Rayleigh
random component and a dominant component of greater amplitude, the
resulting signal hence follows a Rician distribution (see Appendix B.1.2).
The Rician K parameter is defined as the power ratio between the dominant
component and the random fluctuations. As the dominant component presents
a slowly time-variant amplitude, the parameter K varies over time accordingly,
depending on the obstruction of the main path. We may notice in this example
that the mean signal power of the random component is about 12 dB below the
unobstructed signal level. Depending on the attenuation level of the dominant
signal, the parameter K varies from about 12 dB down to less than −20 dB.
In this case, the total received signal follows a Rayleigh distribution. These
results were observed on other measurement records, with maximum values of
the parameter K varying between 8 dB and 13 dB.
In addition, the phase of the random component depicted in Figure 5.14

presents a rather unstable behavior while people are interfering with the main
path of the impulse response, with distinct periods of linear progress. However,
Relative power (dB)

-10
-20
-30
-40
↑
-50 ↑
14 16 18 20 22 24
Time (s)
↓ ↓
2
Phase (rad)
-2
14 16 18 20 22 24
Time (s)
Figure 5.14. Fast fading of the random component. LOS path, 12 people
we may note rapid and significant phase shifts at instants corresponding to

fading nulls (indicated by arrows on the figure), due to the rapidly decreasing
power.
5.1.4.4. Spectral analysis

This section presents a spectral analysis the analysis of the temporal signal
variations received via the main paths of the impulse response. Figure 5.15
represents the average scattering function PS (τ, ν) of the random component,
for the delay τ corresponding to the LOS path. Measurements involving 1
to 12 moving people are taken into account, and the collected records were
distinguished according to the direction of the movement. The general shape
of the scattering function is triangular, with an average Doppler shift centered
around 0 Hz. The spectrum width was calculated in terms of Doppler spread
νRM S , defined in equation [2.38]. The calculated Doppler spread varied
between 0.6 and 3.3 Hz, with no marked influence from the number of moving
people. Similar 0 Hz-centered, triangular shapes of the Doppler spectrum
are reported from continuous wave measurements of the channel temporal
variations performed at frequencies around 1 GHz [HAS 94b, BUL 87].
0
Forward
Backward
-5
Relative power (dB)
-10
-15
-20
-25
-15 -10 -5 0 5 10 15
Doppler shift (Hz)
Figure 5.15. Average scattering function of the random component.

LOS path, forward and backward motions
Some asymmetry can be observed in Figure 5.15 between measurements

performed during forward and backward movements. This can be explained
by the location of the antennas, which emphasize either the lengthening or
the shortening of propagation paths, depending on the direction of the human
motion. More details on the interpretation of these results can be found in
[PAG 06a].
5.2. Statistical channel modeling
In section 5.1, the characterization of the radio channel parameters

was presented from a series of experimental measurements. The frequency
dependent and distance dependent attenuation linked to the radio signal
propagation was evaluated, as well as a number of parameters regarding the
impulse response. The RMS delay spread, the PDP magnitude decay and the
rays arrival rate are a few examples. Statistical models use these experimental
parameters to realistically reproduce the channel effects. This section presents
a few statistical models for the UWB propagation channel available in the
literature. The analyses presented in section 5.1 are then exploited to illustrate
in more detail the design of a statistical model based on experimental data.
5.2.1. Examples of statistical models

In order to provide a unique channel model for the evaluation of different
UWB systems proposals during standardization meetings, the IEEE 802.15
work group made several calls for contributions. Two statistical models were
defined, the first one for short range, high rate, indoor applications (IEEE
802.15.3a model), and the second one for applications with longer range in
both indoor and outdoor environments (IEEE 802.15.4a model). These models
are briefly presented in the following sections.
5.2.1.1. IEEE 802.15.3a model

The IEEE 802.15.3a model [FOE 03a, MOL 03] was developed from
around 10 contributions, all referring to distinct experimental measurements,
performed in indoor residential or office environments [GHA 02a, PEN 02,
FOE 03b, HOV 03, KUN 02b, GHA 02b, CAS 02, CRA 02, SIW 02].
In order to reflect the phenomenon of ray clustering that was observed

in several measurement campaigns, the model is based on the Saleh and
Valenzuela formalism (see equation [2.31]). Parameters are provided to
characterize the clusters and ray arrival rates (Λ and λ), as well as the inter- and
intra-cluster exponential decay constants (Γ and γ). Four sets of parameters
are provided to model the four following channel types:
• the channel model CM 1 corresponds to a distance of 0–4 m in a LOS
situation;
• the channel model CM 2 corresponds to a distance of 0–4 m in an NLOS
situation;
• the channel model CM 3 corresponds to a distance of 4–10 m in an NLOS
situation;
• the channel model CM 4 corresponds to an NLOS situation with a large
delay spread τRMS = 25 ns.
Regarding channel attenuation, the IEEE 802.15.3a model proposes a

theoretical approach using a path loss exponent Nd = 2 for the LOS situation,
which is equivalent to a free space propagation. The NLOS case was not
addressed.
Finally, the fluctuations of ray amplitude modeled using a log-normal

law (see Appendix B.1.6), and a random inversion coefficient is introduced
to simulate the phase inversion observed in the impulse response due to
reflections.
This comprehensive model is a reference for the study of UWB systems. It

can be applied in indoor environments and short range conditions. However,
modeling of the path loss is not practically addressed. It can also be noted that
the measurements used for the model calibration [PEN 02, CHE 02] are limited
to at most 6 GHz bandwidth (2 GHz for the models CM 1 and CM 2).
5.2.1.2. IEEE 802.15.4a model

In order to be representative of a larger number of potential applications,
the IEEE 802.15.4a working group proposed a model with a wider scope
in terms of both frequency and environments [MOL 04]. The targeted
applications are low rate communications (from 1 kbps to a few Mbps), in
the following environments: indoor (residential and office), outdoor, industrial
(factory, etc.) and on-body (for WBAN applications). Two UWB frequency
bands are considered: 2–10 GHz and 0.1–1 GHz. We present here the model
corresponding to the first frequency band.
The general structure of this statistical model is similar to the IEEE

802.15.3a model (see equation [2.31]). Some differences are however noteworthy
regarding the shape of the impulse response:
• The phase θk,l of each ray is no longer limited to the values 0 or π, but
is uniformly distributed between 0 and 2π. Thus, this model reproduces the
complex envelope of the baseband impulse response.
• Ray arrival follows a dual law composed of two Poisson processes.
Accordingly, the model proposes two ray arrival rates λ1 and λ2 , as well as
a mixing parameter.
• The exponential decay of each cluster increases with the delay. The
intra-cluster exponential decay constant is thus of the following form:
γl = kγ Tl + γ0 [5.7]
where Tl represents the time of arrival of the lth cluster and kγ accounts for
the increase of the coefficient γl with increasing delay.
The main difference between the IEEE 802.15.4a model and the IEEE
802.15.3a model is that the former accounts for a realistic path loss, in both
the distance and frequency domains. The proposed model is independent of
the transmitter and receiver antennas. After some changes in the proposed
variables, this path loss model can be expressed in the form given in equation
[2.35], and the model provides the parameter values equivalent to Nf , Nd ,
P L(f0 , d0 ) and σS .
In addition, the small scale variations of the ray amplitude are modeled using
a Nakagami distribution (see Appendix B.1.3). As for the Cassioli-Win-Molisch
model, the parameter value m is a delay function.
Nine channel types were identified within the IEEE 802.15.4a work group.
Each channel type is defined by a given set of parameters as follows:
• the CM 1 and CM 2 models correspond to the indoor residential
environment (respectively in LOS and NLOS configurations);
• the CM 3 and CM 4 models correspond to the indoor office environment
(respectively in LOS and NLOS configurations);
• the CM 5 and CM 6 models correspond to the outdoor environment
(respectively in LOS and NLOS configurations);
• the CM 7 and CM 8 models correspond to the indoor industrial
environment (respectively in LOS and NLOS configurations);
• the CM 9 model corresponds to the outdoor environment in an NLOS
configuration, in the specific case of a farm area or an area covered with snow.
This model is more detailed than the IEEE 802.15.3a model, but is
also somewhat more complex. The provided parameter sets are based on
experimental measurements for each environment: residential [CHO 04b], office
[BAL 04a, SCH 04], industrial [KAR 04b] and outdoor [BAL 04a, KEI 04].
A survey of the results from the literature complements these measurement
campaigns. It should be noted, though, that the measurements performed in
the office and outdoor environments covered frequency bands limited to 3 or
6 GHz.
5.2.1.3. Other models

The Cassioli-Win-Molisch model
The Cassioli-Win-Molisch model [CAS 02] is the result of a joint research
work performed by the University of Rome “Tor Vergata” (Italy), Vienna
University (Austria) and the UltRaLab laboratory from the University of
South California (USA). This model is one of the first statistical models
describing the UWB propagation channel. Despite some limitations, it is thus
frequently cited in the studies on UWB. This model is based on a series
of measurements performed by the UltRaLab laboratory in an indoor office
environment, over a frequency band of about 1 GHz [WIN 97b]. 686 impulse
responses were used: 14 antenna locations were selected, and at each location,
49 measurements were taken over an area of about 1 m2 .
The Cassioli-Win-Molisch model is based on a discrete scale on the delay

axis, with a delay bins defined with a step Δτ of 2 ns. The overall impulse
response power received between the delays kΔτ and (k + 1)Δτ is integrated,
and it is assumed that a ray exists at each delay kΔτ . This corresponds to a
1
ray arrival rate λ = Δτ . The power of each ray follows an exponential decay,
but a single cluster is observed. Following the Saleh and Valenzuela formalism
(see section 2.4.1), the PDP model can be expressed in the form:

K
d
Ph (0, τ ) = βk2 δ τ − − (k − 1)Δτ [5.8]
c
k=1
where d indicates the distance between the transmitter and the receiver.
The PDP exponential decay is characterized by the coefficient γ (see

equation [2.33]), but the model also introduces an additional coefficient r in
order to account for a significant attenuation between the 1st ray and the 2nd
ray.
The propagation loss is characterized as a function of distance according to

a dual-slope law:
⎧
⎪
⎪P L d0 + 20.4 log d + S(d)
⎪ d ≤ 11 m
⎨ d0
P L(d) = [5.9]
⎪
⎪ d
⎪
⎩P L d0 − 56 + 74 log + S(d) d > 11 m
d0
The fast fading due to the antenna displacement is characterized by a ray

amplitude distribution following a Nakagami distribution (see Appendix B.1.3).
The Nakagami m parameter decreases with the delay to approach the value of
1 for the last rays in the PDP, where the ray amplitude follows a Rayleigh
distribution.
This model provides an accurate and reproducible description of the

observed measurements. Its main limitations lie in the low number of
measurements on which the statistical study is based, and in the reduced
frequency band.
Frequency domain approach

Most of the research efforts regarding statistical modeling for the UWB
propagation channel concentrate on a time domain approach, where the
model provides a description of the channel impulse response. However, other
approaches are possible. The model proposed by the researchers from the
AT&T Research Laboratory and MIT [GHA 04a] is an interesting example
of a frequency domain approach. This model is designed from an extensive
measurement campaign performed in 23 residential houses.
The main concept of this model is to reproduce the channel transfer function
T (f, t) in a statistical way. As for the time domain approach, each parameter
can be described using a statistical law. The value of the frequency domain
approach lies in its opening up the possibility of describing the transfer function
components in a regressive way. The model can be expressed in the form of a
filter with infinite impulse response, which can be mathematically represented
by:

T fi , t + a1 T fi−1 , t + a2 T fi−2 , t = ni [5.10]
where ni represents the input white Gaussian noise. The model can be
represented using five variables: the parameters a1 and a2 , the input conditions
T (f1 , t) and T (f2 , t), and the standard deviation of the Gaussian noise σn .
Each of these parameters is then described in a statistical way as a function
of the distance. The model is complemented with a power attenuation law.
Other approaches were proposed regarding UWB channel modeling in the

frequency domain, such as the Prony method used by the Virginia Polytechnic
Institute (USA) [LIC 03]. The value of such a model lies in its low complexity:
only a few parameters are required for its description. However, there is no
direct knowledge of the traditional channel characterization parameters, such
as the shape of the PDP or the delay spread. For this reason, the comparison
between the two model types is not straightforward.
5.2.2. Empirical modeling principles
This section aims at illustrating UWB channel statistical modeling based

on experimental data. A statistical model based on the Saleh and Valenzuela
formalism is designed from the characteristic parameters observed in
section 5.1 [PAG 06d]. The following sections present a propagation loss model
and an impulse response model. Finally, simulation results are compared to
the experimental measurements.
5.2.2.1. Propagation loss model

The first step in the design of the UWB channel model is the definition of
attenuation due to signal propagation. In section 5.1.1.1, it was shown that
when the antenna effect is correctly compensated, the channel attenuation in
the frequency domain is close to the theoretical loss of 20 dB per decade. The
path loss model can thus be expressed in dB in the following form:

f d
P L(f, d) = P L f0 , d0 + 20 log + 10Nd log + S(d) [5.11]
f0 d0
where d0 = 1 m represents a reference distance, f0 = 6.85 GHz corresponds
to the center frequency of the FCC analyzed band and S is a Gaussian
random variable with zero mean. The parameters of this model are defined by
experimental characterization (see section 5.1.1) and are given in Table 5.6.
LOS NLOS
Nd 1.62 3.22
σS (dB) 1.7 5.7
P L(f0 , d0 ) (dB) 53.7 59.4
Table 5.6. Path loss model parameters. Values of P L(f0 , d0 )

are given for f0 = 6.85 GHz and d0 = 1 m
5.2.2.2. Modeling the channel impulse response over an infinite bandwidth

The principle of the UWB impulse response model consists of generating
all constituent rays while maintaining the characteristics observed from the
experimental measurements. Among these characteristics, we mainly aim at
reproducing the clustering of multipath echoes, the ray and cluster arrival rates,
and the decreasing magnitude of the received power with increasing delay. For
a baseband representation, a ray is described by its delay τ , its amplitude β
and its phase θ. By generating these parameters, it is possible to describe the
impulse response over an infinite bandwidth. We then process these parameters
in the frequency domain in order to include the effect of the limited observation
bandwidth.
In order to account for the clustering of multiple paths, the impulse response
is modeled using the Saleh and Valenzuela formalism (see section 2.4.1). At a
given instant, the UWB channel impulse response is thus described by the
following formula:

L
Kl

h(τ ) = βk,l ejθk,l δ τ − Tl − τk,l [5.12]
l=1 k=1
where L represents the number of clusters, Kl is the number of rays within

the lth cluster and Tl corresponds to the arrival time of the lth cluster. The
parameters βk,l , θk,l and τk,l represent the amplitude, phase and arrival time
associated with the k th ray within the lth cluster.
The experimental observations given in section 5.1.2.3 show that the

inter-clusters duration follows an exponential distribution. Hence, the arrival
of a new cluster can be modeled by a Poisson process, and the number of
clusters in the impulse response can be generated by drawing a random
variable L according to the following law [MOL 04]:
(L̄)L exp(−L̄)
pL (L) = [5.13]
L!
where L̄ represents the average number of clusters. During measurement

analysis, we observed an average number of clusters of L̄ = 5.6 in the LOS
case and L̄ = 2.4 in the NLOS case.
Having selected a transmitter-receiver distance d for the simulation, the

time of arrival of the first cluster is given by T1 = dc , where c is the speed of
light. The time of arrival Tl of the L − 1 remaining clusters is then calculated
by generating inter-cluster durations following an exponential law [SAL 87]:

p Tl | Tl−1 = Λ exp − Λ Tl − Tl−1 [5.14]
where Λ is the cluster arrival rate.
The inter-cluster power decay is accurately modeled by a power function

(see section 5.1.2.2). This approach differs from the one followed by Saleh
and Valenzuela [SAL 87], but presents a closer fit to the experimental
measurements. The amplitude of the first ray within each cluster is thus given
by:
−Ω
2 2 Tl
β1,l = β1.1 [5.15]
T1
where Ω represents the inter-cluster power decay constant.
In accordance to the results of the static UWB channel study, it is

recommended to use the values Λ = 36.5 MHz and Ω = 4.4 in the LOS case.
In the NLOS case, the recommended values are Λ = 24.9 MHz and Ω = 3.9.
The rays are iteratively generated for each cluster. The arrival time of
each ray is calculated using inter-ray durations following an exponential law
[SAL 87]:
p τk,l | τk,l−1 = λ exp − λ τk,l − τk,l−1 [5.16]
A power function is used to calculate the ray amplitude (see section 5.1.2.2):
−ω
τk,l + Tl
= 10− 10 β1,l
G
2 2
βk,l [5.17]
Tl
where ω represents the intra-cluster power decay constant and G accounts for
the observed attenuation between the first path of each cluster and the following
multipaths. For each cluster, the ray generation stops when the ray amplitude
reaches a given threshold D, fixed at −50 dB.
According to the characterization study presented in section 5.1 and to

the observation of the measured PDP, we recommend using the parameters
λ = 5.95 GHz, ω = 11.1 and G = 12 dB in the LOS case. The recommended
values in the NLOS case are λ = 6.19 GHz and ω = 10.2.
Finally, the phase θk,l of each ray is generated using a uniform law over
the interval [0, 2π[. This approach was also adopted in the IEEE 802.15.4a
model [MOL 04]. Figure 5.16 presents the rays obtained in a LOS situation
for a transmitter-receiver distance of 6 m, where we may observe 5 clusters. It
may be noted that this representation corresponds to an infinite observation
bandwidth, each ray being represented by a Dirac function.
0
Relative power (dB)
-10
-20
-30
-40
-50
0 20 40 60 80 100
Delay (ns)
Figure 5.16. Impulse response simulated over an infinite bandwidth. LOS situation,
with the parameters d = 6 m, Λ = 36.5 MHz, λ = 5.95 GHz, Ω = 4.4, ω = 11.1 and
G = 12 dB
Table 5.7 summarizes all experimental parameters used in our statistical

model of the UWB channel.
LOS NLOS
fmin (GHz) 3.1≤ fmin < fmax
fmax (GHz) fmin < fmax ≤ 10.6
D (dB) 50
Λ (MHz) 36.5 24.9
L̄ 5.6 2.4
λ (GHz) 5.95 6.19
Ω 4.4 3.9
ω 11.1 10.2
G (dB) 12 0
Table 5.7. Parameters of the UWB impulse response model

5.2.2.3. Modeling the channel impulse response over a limited bandwidth

In practice, impulse response is observed on a limited bandwidth, situated
within the 3.1–10.6 GHz band. We note by fmin and fmax the minimum and
maximum frequencies in the observation band, and by fc = 12 (fmin + fmax )
the center frequency. The channel transfer function limited to the observation
bandwidth hence appears (for positive frequencies):
⎧
⎪ L Kl
⎨fc 1
βk,l ej(θk,l −2πf (Tl +τk,l )) if fmin ≤ f ≤ fmax
'L 'Kl 2
Tlim (f ) = f β
⎪
⎩ l=1 k=1 k,l l=1 k=1
0 otherwise
[5.18]
P L(fc ,d)
where the coefficient ' L ' Kl 2
normalizes the power received at the
l=1 k=1 βk,l
frequency fc according to the proposed path loss model. In addition, the
term ffc accounts for the power decrease in −20 log(f ) described previously.
This transfer function normalization procedure was also proposed in
[FOE 03a, ALV 03].
The impulse response observed over a limited bandwidth hlim (τ ) is simply

obtained from the transfer function Tlim (f ) using an inverse Fourier transform.
In order to limit the side lobes level, it is possible to use a Hanning window for
instance at this stage [HAR 78].
Figure 5.17(a) represents an impulse response example observed over the

3.1–10.6 GHz frequency band. We may note that the low delay between two
consecutive rays (see Figure 5.16) naturally induces constructive or destructive
interferences. Figure 5.17(b) presents the same impulse response observed over
3.1–4.1 GHz frequency band: we may note that the multipath resolution is less
accurate when the observation bandwidth decreases.
5.2.2.4. Simulation results

A series of simulations was conducted using the static UWB radio channel
model presented in section 5.2.2. A set of 119 PDP was generated using
the same transmitter-receiver distances as in our measurement campaign
(see section 3.4.2.1), for the LOS and NLOS situations. Each PDP was
calculated from a set of 90 impulse responses, simulating the motion of the
antenna around a circle of radius 20 cm. The generation of the fast fading
characteristics for the 90 locally simulated impulse responses was performed
using an advanced algorithm, taking the ray arrival angle into account. This
algorithm is fully described in section 5.3.1.
Figure 5.18 presents two typical PDP obtained by simulation. One of the
90 constituting impulse responses is also represented. Note that on the x-axis,
(a)
-60
-70
Relative power (dB)

-80
-90
-100
-110
-120
-130
0 20 40 60 80 100
Delay (ns)
(b)
-60
-70
Relative power (dB)
-80
-90
-100
-110
-120
-130
0 20 40 60 80 100
Delay (ns)
Figure 5.17. Impulse response simulated over a limited bandwidth. Identical set of
rays, observed over the 3.1–10.6 GHz band (a) and the 3.1–4.1 GHz band (b)
the delay was converted in path length. As a general observation, simulations

are similar to the measured PDP (see Figure 5.3).
Two parameters representative of the channel temporal dispersion were

calculated for the entire set of generated PDP: the delay spread τRM S and the
75% delay window W75% . The average values obtained for these parameters in
measurement and simulation are compared for the LOS and NLOS situations
in Table 5.8. We may note that the proposed model correctly reproduces the
dispersion parameters experimentally measured over the UWB radio channel.
The delay spread of the simulated PDP is particularly close to the measurement,
in both LOS and NLOS configurations. The values obtained for the delay
window are somewhat different from the experimental data, but stay in the
same range as the measured characteristics. The proposed model therefore
allows us to reproduce not only the impulse response structure, but also channel
dispersion.
(a)
0 IR
Relative power (dB) PDP
-10
-20
-30
-40
-50
0 20 40 60 80
Distance-delay (m)
(b)
0 IR
PDP
Relative power (dB)
-10
-20
-30
-40
-50
0 20 40 60 80
Distance-delay (m)
Figure 5.18. PDP and impulse responses obtained by simulation.

(a) LOS and (b) NLOS configurations
LOS NLOS
Parameter Measurement Simulation Measurement Simulation
τRMS (ns) 4.1 4.0 9.9 9.7
W75% (ns) 7.6 9.7 23.7 21.2
Table 5.8. Comparison of the dispersion parameters: measurement vs. simulation

5.3. Advanced modeling in a dynamic configuration
Section 5.2.2 describes how to simulate an impulse response from statistical

parameters obtained through experimental measurements, in the case of a static
channel. In instances of practical use for future UWB systems, the antenna
displacement or the motion of people may generate significant variations in the
channel impulse response. This section describes two adaptations to the statical
model for an advanced modeling of the spatial and temporal fluctuations of the
UWB radio channel [PAG 06c].
5.3.1. Space variation modeling
Spatial variations are due to the displacement of at least one of the

transmitting or receiving antennas. In this case, the lengths of most
propagation paths undergo significant variation. Only a few studies on the
statistical modeling of the UWB channel spatial variations are available. The
IMST model is one example where a simplified ray tracing algorithm is used
to calculate a fixed delay variation for each cluster [KUN 03]. We here propose
to model these spatial fluctuations by taking the angle formed between the
antenna motion and each propagation path into account. In the remainder
of this section, we will assume a displacement of the receiving antenna and
consider the arrival angle of each path. For the sake of clarity, the presented
model allows an antenna displacement in the (O, x, y) plane only, by taking
the ray azimuth ϕ into account. A 3D model including ray elevation is
forwardly derivable.
The first step of this spatial fluctuation model consists of generating an

impulse response over an infinite bandwidth h(x0 , y0 , τ ), corresponding to a
location (x0 , y0 ), as described in section 5.2.2. This impulse response observed
over an infinite bandwidth may be expressed as in equation [5.12]. In order to
calculate the impulse response observed over an infinite bandwidth at a location
(x, y) close to the location (x0 , y0 ), we consider that each ray corresponds to
a plane wave. This assumption holds when the receiving antenna is placed in
the far field with respect to the surrounding walls and furniture where the
transmission, reflection or diffraction phenomena occur. Figure 5.19 illustrates
the obtained configuration for a given ray. The path elongation Δl of the
incident ray is given by:

Δl = − x − x0 cos(ϕ) − y − y0 sin(ϕ) [5.19]
where ϕ represents the incident ray azimuth.
A limited number of statistical studies are available in the literature

regarding the departure and arrival angles for UWB channels [CRA 02,
ĳ
Į
(x0, y0) x
Figure 5.19. Path elongation related to an antenna displacement
HAN 05, VEN 05]. To determine the azimuth of each ray in the impulse
response, we can use the model derived by Spencer et al. from wideband
indoor measurements [SPE 97]. In this model, the arrival azimuth of the
k th ray within the lth cluster is decomposed in Φl + ϕk,l , where Φl is the
mean arrival azimuth in the lth cluster. Without further knowledge about the
environment, we may consider that Φl is uniformly distributed in the [0, 2π[
interval. ϕk,l represents the distribution of the ray azimuth within a cluster,
and follows a Laplace distribution with zero mean and standard deviation σϕ :
√
1 −|
2ϕk,l
|
pϕk,l ϕk,l =√ e σϕ
[5.20]
2σϕ
In [HAN 05], the standard deviation of the arrival azimuth σϕ is lower than
or equal to 6.7◦ .3 As a first approximation, we recommend using this value
for the parameter σϕ . This value could be refined as more experimental data
becomes available. For instance, [VEN 05] proposes a 2-cluster model, where
the azimuth distribution follows a Laplace law in the first cluster (σϕ = 5◦ )
and a uniform law in the second cluster.
From the impulse response generated at the location (x0 , y0 ) and knowing
the arrival azimuth Φl + ϕk,l of each ray, we can calculate the transfer function
3. For the indoor LOS and NLOS configurations on a single floor.

over a limited bandwidth at a location (x, y) close to location (x0 , y0 ):
Tlim (x, y, f )
⎧
⎪ L Kl
⎨fc 1
βk,l ej(θk,l −2πf (Tl +τk,l +Δτk,l )) if fmin ≤ f ≤ fmax
'L 'Kl 2
= f β
⎪
⎩ l=1 k=1 k,l l=1 k=1
0 otherwise
[5.21]
with:
1
Δτk,l = − x − x0 cos Φl + ϕk,l + y − y0 sin Φl + ϕk,l [5.22]
c
The impulse response hlim (x, y, τ ) observed over a limited bandwidth is

simply obtained using an inverse Fourier transform.
Figure 5.20 presents a simulation of the space varying impulse response

observed over the 3.1–10.6 GHz band. The delay, amplitude and phase of each
ray correspond to the parameters already used in the previous examples. The
spatial variations of the impulse response were calculated for a displacement of
the receiving antenna along the Ox axis over a distance of 2 m with 1 cm step.
We may clearly distinguish the evolution of the main echoes (a-e), depending
on their arrival direction. For instance, the length of the first path shortens
from approximately 6 m to 4 m.
0 0
(a) -5
5
-10
Distance-delay (m)
Relative power (dB)
10 (b) -15
-20
15 (c)
-25
20 (d) -30
-35
25
(e) -40
-45
0 50 100 150 200
Location x (cm)
Figure 5.20. Simulation of a space variant impulse response. Displacement of the

receiving antenna along the Ox axis over a distance of 2 m
5.3.2. Modeling the effect of people
This section describes a method to model the effect of people, based on the
experimental analysis given in section 5.1.4. The main idea is to reproduce the
time variance observed on each ray of a given impulse response. This model
is based on the results of a real time UWB measurement campaign performed
over the 4–5 GHz band (see section 3.4.2.3). As a result, the presented model
is valid for simulating UWB channels with bandwidths up to 1 GHz only.
However, the modeling methodology may be used in other configurations as
more experimental characteristics of the time-variant UWB channel will become
available.
The extension of the model to temporal variations is based on an initial

impulse response h(t0 , τ ) corresponding to a static environment at a given
instant t0 . The method described in section 5.2.2 can be used to obtain a
description of the impulse response over an infinite bandwidth. The initial
impulse response can thus be expressed according to equation [5.12], where
αk,l (t0 ) = βk,l eθk,l denotes the initial complex magnitude of the k th ray within
the lth cluster.
As observed in our experimental characterization, we will differentiate

between (a) the main path of each cluster, where each person generates a
shadowing pattern in addition to fast amplitude fluctuations, and (b) the
clusters of dense multipath where Rayleigh fading occurs [PAG 06a]. When
observing a radio channel over a relatively low bandwidth, in our case lower
than 1 GHz, the observed main path encompasses not only the first ray in
each cluster, but also several rays situated within the observation resolution
R. This resolution R is equal to 2 ns for a bandwidth of 1 GHz when using
a Hanning window [HAR 78]. For all rays within the main path of each
cluster, the time-variant amplitude αk,l (t) is generated according to the
algorithm below. A more detailed description of this algorithm may be found
in [PAG 05]. Assuming that a group of Np people is interfering with the UWB
radio link, the algorithm steps are as follows:
i) Select Np instants {tn }n=1...Np corresponding to the instants where each
of the Np people are crossing the main path of the cluster. These passing
instants may be randomly chosen, or calculated according to the environmental
geometry.
ii) Generate Np slow shadowing patterns {sn (t)}n=1...Np corresponding to
the individual effects of Np moving people, using the following Gaussian-shaped
attenuation function:
( )
2 2
sn (t) = −As exp − 2 t − tn [5.23]
Ts
where As represents the maximal shadowing attenuation in dB, Ts represents

the shadowing duration in seconds, and tn represents the shadowing instant.
According to our experimental characterization, the parameter values may be
selected in the range of 5 dB to 10 dB for As , and in the range of 3 s to 5 s for
Ts .
iii) Calculate the amplitude variations of the dominant component dl (t) in
the linear scale as follows:

Np
sn (t−tn )
dl (t) = 10 20 [5.24]
n=1
iv) Generate additional fast fading fluctuations rk,l (t) following a Rayleigh
amplitude distribution and having a Laplacian scattering function. As observed
from the experimental data, the Doppler spread νRMS should be in the range
of 1–3 Hz. The mean power Pr of the random component rk,l (t) should be 8 dB
to 13 dB below the power level of the dominant component αk,l (t0 ).
v) Finally, calculate the time-variant amplitude αk,l (t) for rays within the
main path of the lth cluster as:

αk,l (t) = αk,l t0 dl (t) + rk,l (t) [5.25]
Regarding step iv), several methods for the generation of a random signal
presenting a Rayleigh distribution and an arbitrary Doppler spectrum are
thoroughly discussed in [PAE 02].
The algorithm presented above allows us to generate the effect that

moving people have on the main path of each cluster. For the remaining
rays, corresponding to regions of dense multipath, we propose to model
the amplitude αk,l (t) variations by using a Rayleigh distribution. More
experimental analysis should be performed to study the exact shape of the
Doppler spectrum in the regions of dense multipath. As a first step, we
propose using a Laplacian distribution. Thus, the amplitude αk,l (t) may be
generated as in step iv) of the algorithm.
Having calculated the amplitude αk,l (t) for all impulse response rays, the
time variant transfer function is given as:
Tlim (t, f )
⎧
⎪ L Kl
⎨ fc 1
αk,l (t)e−j2πf (Tl +τk,l ) if fmin ≤ f ≤ fmax
'L 'Kl 2
= f β
⎪
⎩ l=1 k=1 k,l l=1 k=1
0 otherwise
[5.26]
The impulse response hlim (t, τ ) observed over a limited bandwidth is

obtained using an inverse Fourier transform.
The proposed model was used to simulate the motion of 4 people in the
vicinity of a radio link in a LOS situation. For selected realistic passing instants
with respect to the main path of each cluster, the delays Tl and the passing
instants tn were calculated using a ray tracing tool, from which only the
most significant rays were extracted. Figure 5.21 illustrates the simulation
configuration. The five selected paths correspond to the main path and to
the reflections on the four walls. The dashed arrow indicates the motion of the
group of 4 people and circles present locations where the group crosses one of
the main echoes.
Rx
Tx
Figure 5.21. Simulation configuration for the effect of people
The time varying impulse response was simulated according to this

information, and the results are presented in Figure 5.22. The impulse
response is observed over the 4–5 GHz frequency band, as in an experiment.
The following parameter values were used for the temporal variations model:
As = 5 dB, Ts = 4 s, Pr = −13 dB and νRMS = 1 Hz.
As indicated by the circle locations on the graph, the group of people

interferes with each main path at a different instant. For each shadowing
instant, we may observe a power attenuation of the corresponding path, for
a duration in the order of 3 s to 6 s. In addition, the presence of mobile people
generates fast fading, observable at all impulse response delays. It can be noted
that on occasions, this fast fading significantly interferes with the main paths,
as for instance on the second cluster. This proposed model could be refined by
analyzing additional experimental results. In particular, further research could
be conducted by sounding the UWB channel over a larger bandwidth.
Figure 5.22. Simulation of a time variant impulse response. Circles indicate the group
passing instants for the five main paths of the impulse response
5.4. Conclusion
Statistical modeling of the UWB channel consists of reproducing the

main channel characteristics observed during the measurements using a
mathematical description. The experimental analyses described in this chapter
illustrated the characterization of the UWB propagation channel in an indoor
office environment. Path loss was analyzed in both frequency domain and
distance domain. The measured PDP were characterized through the delay
spread, as well as the arrival rate and amplitude decay of the rays and clusters.
The fast fading analysis shows that the signal amplitude received at a given
delay follows a Rayleigh distribution, except for the main paths, where the
situation is more deterministic.
Several models of the UWB propagation channel were presented in the

literature. Among the statistical models, the proposals presented in the
standardization groups IEEE 802.15.3a and IEEE 802.15.4a, based on a
Saleh and Valenzuela approach, are frequently used for the analysis of UWB
communication systems. The principles of statistical modeling were presented
using a practical approach based on experimental data. The described path
loss model can be used directly for dimensioning studies and to evaluate the
jamming generated by an UWB terminal. A detailed model has also been given
for the static impulse response. Simulations show that this model accurately
reproduces the channel dispersion. It can be used in system simulations, for
the design and development of UWB based radio transceivers.
Two extensions of the impulse response were finally described to account for
the spatial and temporal variations of the UWB channel. The spatial variations
model emulates the effects of antenna displacement, by including the arrival
direction of the delayed wavefronts. The temporal fluctuations model integrates
the effects linked to the motion of people, and can be used to simulate a UWB
radio link in realistic conditions.
Appendix A
Baseband Representation of the Radio Channel
A real signal x(t) with spectral components covering a bandwidth centered

on the frequency f0 = 0 can be represented by its complex envelope γx (t)
defined by:

x(t) = γx (t)ej2πf0 t [A.1]
There are several ways to define a complex envelope γ(t) able to represent
the signal x(t) according to relation [A.1]. We will use here the complex envelope
defined by the analytical form of the signal x(t) as follows [BAR 95]:
γx (t) = x(t)e−j2πf0 t [A.2]
where x(t) is the analytical representation of signal x(t).
A real signal possesses a complex spectral structure with a Hermitian

symmetry. All the information concerning this signal is known in the
positive part of the frequency spectrum. The analytical signal is the signal
representation using only its positive spectral frequency components. In order
to keep the same signal power, the power spectral density (PSD) of the
spectrum’s positive side is multiplied by two. The obtained analytical signal is
thus defined in the frequency domain by:

X(f ) = F x(t) = 2U (f )X(f ) = 1 + sign(f ) X(f ) [A.3]
where F{·} is the Fourier transform operation and U (f ) represents the unit
step function (Heaviside step function).
By using the inverse Fourier transform, it leads to the following relation:

1
x(t) = x(t) + j ⊗ x(t) = x(t) + jH x(t) [A.4]
πt
where ⊗ represents the convolution operator and H{·} represents the Hilbert
transform.
The signal s(t) is obtained after a convolution of e(t) and h(t):
s(t) = e(t) ⊗ h(t) [A.5]
By keeping the same notations as previously, we have [BAR 95]:

1
S(f ) = 2U (f )S(f ) = 2U (f )E(f )H(f ) = E(f )H(f ) [A.6]
2
and then:

γs (t) = s(t)e−j2πf0 t = F −1 S(f + f0 )

1
= F −1 E(f + f0 )H(f + f0 ) [A.7]
2
1
= γe (t) ⊗ γh (t)
2
There is a second way to represent the equivalent baseband filter [GUI 96].
Indeed, relation [A.6] allows us to write:
S(f ) = E(f )H(f ) [A.8]
So, γs (t) can also be expressed by:

γs (t) = F −1 E f + f0 H f + f0
[A.9]
= γe (t) ⊗ h(t)e−j2πf0 t
So, the baseband filtered heq (t) which is equivalent to the real band filter
h(t) can be expressed in two ways:
1
heq1 (t) = γh (t)
2 [A.10]
heq2 (t) = h(t)e−j2πf0 t
Baseband Representation of the Radio Channel 179
In the frequency domain, these two baseband filters Heq (f ) equivalent to

the real band filter H(f ) are expressed by:

Heq1 (f ) = U f + f0 H f + f0
[A.11]
Heq2 (f ) = H f + f0
From a practical viewpoint, the filter Heq1 (f ) corresponds to a translation

of the positive side frequency components of the filter H(f ) so that they are
centered around 0 Hz. The filter Heq2 (f ) corresponds to a simple translation of
all the filter component H(f ). These filters are similar when they are applied
to signals expressed in baseband.
Appendix B
Statistical Distributions
B.1. Definition
This section defines the main distribution laws presented in the book. Unless
otherwise mentioned, most of these laws are generally used to characterize
the magnitude of the channel impulse response for a given delay. For each
distribution of the random variable X, we give the probability density function
(PDF) pX (x), the cumulative density function (CDF) F (x) = P (X ≤ x), the
first and second order moments E[X] and E[X 2 ], and the variance V ar[X]
when it can be simply expressed.
B.1.1. Rayleigh distribution
The Rayleigh distribution [PAR 00] is defined from the parameter σ which
is related to the standard deviation of the distribution by a constant value.
⎧x x2
⎨ e− 2σ2 if x ≥ 0
2
pX (x) = σ [B.1]
⎩
0 otherwise
⎧ x2
⎨1 − e− 2σ2 if x ≥ 0
F (x) = [B.2]
⎩
0 otherwise
"
π
E[X] = σ [B.3]
2

E X 2 = 2σ 2 [B.4]

4−π
Var[X] = σ2 [B.5]
2
B.1.2. Rice distribution
The Rice distribution [PAR 00, LAU 94] is defined from two parameters, s
and σ.
⎧
⎪ 2 2
⎨ x e− x 2σ+s2 I0 xs if x ≥ 0
pX (x) = σ 2 σ2 [B.6]
⎪
⎩0 otherwise
where I0 represents the modified Bessel function of type 1.

⎧
⎪
⎨1 − Q s , x if x ≥ 0
F (x) = σ σ [B.7]
⎪
⎩0 otherwise
where Q is the Marcum function given by [MAR 60]:

∞
x2 +y 2
Q(x, r) = ye− 2 I0 (xy)dy [B.8]
r
"
π s2
E[X] = |σ|L 12 − 2 [B.9]
2 2σ
where L 12 is the Laguerre function, the solution of the differential equation:
d2 y dy 1
x + (1 − x) + y=0 [B.10]
dx2 dx 2

E X 2 = s2 + 2σ 2 [B.11]
A typical parameter of this distribution is the parameter k expressed by:
s2
k= [B.12]
2σ 2
Statistical Distributions 183
Numerous estimators exist for the k parameter. The k parameter estimator

considered in the book is the one based on the second and fourth order moments,
and uses the following m parameter:
2

x2
m= 2 [B.13]
x4 − x2
where (·) corresponds to the empirical mean value. So, k is expressed by

[ABD 01]:

1− m1
k [B.14]
1− 1− m 1
The Rice distribution tends to a Rayleigh distribution when s tends to 0.
B.1.3. Nakagami distribution
The Nakagami distribution [LAU 94] is defined from two parameters, m and
Ω.
⎧ m
⎪
⎨ 2m mx2
m
x2m−1 e− Ω if x ≥ 0
pX (x) = Γ(m)Ω [B.15]
⎪
⎩0 otherwise
where Γ represents the Gamma function defined for x > 0 by:

∞
Γ(x) = e−t tx−1 dt [B.16]
0
⎧
⎪ 2
⎨γ mx , m if x ≥ 0
F (x) = Ω [B.17]
⎪
⎩0 otherwise
where γ represents the incomplete Gamma function defined for x > 0 by:
a
1
γ(a, x) = e−t tx−1 dt [B.18]
Γ(x) 0
"
Γ m + 12 Ω
E[X] = [B.19]
Γ(m) m

E X2 = Ω [B.20]
Numerous estimators exist for the parameter m. The one used in the book
is the following [ABD 00]:
2
x2
m 2 [B.21]
x4 − x2
The Nakagami distribution tends to a Rayleigh distribution when m tends

to 1.
B.1.4. Weibull distribution
The Weibull distribution [LAU 94] is defined from the two parameters, a
and b.
*
abxb−1 e−ax if x ≥ 0
b
pX (x) = [B.22]
0 otherwise
*
1 − e−ax if x ≥ 0
b
F (x) = [B.23]
0 otherwise

− 1b 1
E[X] = a Γ 1 + [B.24]
b

2 2
E X 2 = a− b Γ 1 + [B.25]
b
The Weibull distribution tends to a Rayleigh distribution when b tends to 2.
B.1.5. Normal distribution
The normal distribution is defined from two parameters, the mean μ and
the standard deviation σ.
1 (x−μ)2
pX (x) = √ e− 2σ2 [B.26]
2πσ

1 x−μ
F (x) = 1 + erf √ [B.27]
2 2σ
where erf represents the error function defined by:

x
2 2
erf(x) = √ e−t dt [B.28]
π 0
E[X] = μ [B.29]
2
E X = μ2 + σ 2 [B.30]
Var[X] = σ 2 [B.31]
B.1.6. Log-normal distribution
The log-normal distribution [WIE 03] is defined from two parameters, μ and
σ. This distribution corresponds to the normal distribution of a signal complex
envelope expressed in dB.
⎧
(10 log(x)−μ)2
⎨ √ 10 −
e 2σ 2 if x ≥ 0
pX (x) = 2πxσ ln(10) [B.32]
⎩0 otherwise
⎧
⎨ 1 1 + erf 10 log(x) −μ
√ if x ≥ 0
F (x) = 2 2σ [B.33]
⎩
0 otherwise
μ 1 σ 2
E[X] = 10 · 10 10 + 2 ( 10 ) [B.34]
μ σ 2
E X 2 = 10 · 102( 10 +( 10 ) ) [B.35]
We can estimate the parameters μ and σ by calculating the first and second
moments of the variable expressed in dB:

E XdB = μ [B.36]
2
E XdB = μ2 + σ 2 [B.37]

Var XdB = σ 2 [B.38]
B.1.7. Laplace distribution
The Laplace distribution [SPE 97] is often used to model the arrival angle
associated with the cluster rays. In the book, this distribution is also used to
describe the Doppler spectrum observed on the fast variations of the signals
generated by moving people. This distribution is defined from two parameters,

μ and σ.
√
1 |x−μ| 2
pX (x) = √ e− σ [B.39]
2σ
1 |x−μ|√2
F (x) = 1 + sgn(x − μ) e− σ [B.40]
2
E[X] = μ [B.41]

E X 2 = μ2 + σ 2 [B.42]
Var[X] = σ 2 [B.43]
B.2. Kolmogorov-Smirnov goodness-of-fit test
Considering an empirical CDF Fn (x) based on n samples and the theoretical

CDF F0 (x) of the random variable from which the random draw is held, we
can study the following variable:

Dn = max Fn (x) − F0 (x) [B.44]
Figure B.1 illustrates the calculation of the variable Dn , for which the
distribution has been studied by Kolmogorov [KOL 33]. For a decision threshold
α and the critical value dα , we note:
P (Dn > dα ) = α [B.45]
or:
P (Fn (x) − dα ≤ F0 (x) ≤ Fn (x) + dα , ∀x) = 1 − α [B.46]
The critical value dα has been tabulated for various values of α and n. We
can demonstrate for example that for n > 80:
1
d0.05 1.3581 · n− 2
1
[B.47]
d0.01 1.6276 · n− 2
For a sample size n = 100, for example, we can conclude that the probability
is:
– 95% that F0 (x) is totally situated between F100 (x) − 0.13581 and
F100 (x) + 0.13581;
Fn(x)
Dn
P(X x)
F0(x)
Figure B.1. Kolmogorov-Smirnov test: theoretical and empirical CDF
– 99% that F0 (x) is totally situated between F100 (x) − 0.16276 and
F100 (x) + 0.16276.
The Kolmogorov-Smirnov test consists of calculating the maximum

deviation between an empirical CDF and a theoretical CDF, for a given decision
threshold α:
– if this deviation is smaller than the critical value dα , we conclude that
the empirical CDF follows the same law as the theoretical CDF;
– if this deviation is higher than the critical value dα , we conclude that the
empirical CDF does not follow the same law as the theoretical CDF.
Formally, a statistical decision process like the Kolmogorov-Smirnov test can

lead to two types of errors. If the test erroneously concludes that the sample
set does not follow the theoretical distribution law, we are faced with an error
of type I. If the test erroneously concludes that the sample set follows the
theoretical distribution law, we are faced with an error of type II. In the case of
the Kolmogorov-Smirnov test, the type I probability of error is known (α), but
the type II probability of error cannot be directly calculated. In other words,
the probability of error is not known when we conclude that some samples
follow a given law. So, this test is to be used sparingly.
Appendix C
Geometric Optics and Uniform

Theory of Diffraction
C.1. Geometric optics
C.1.1. Introduction
Geometric optics (GO) have been developed for the analysis of light wave
propagation which corresponds to high frequencies. In propagation problems,
the GO is valid only for frequencies higher than 100 MHz.
If we consider a monochromatic electromagnetic pulsation wave ω, the

s, ω) and magnetic field H(
electric field E( s, ω) expressions at an observation
point P, the spatial position of which is defined by a vector s, correspond to
the Maxwell equation expressed in an inhomogenous medium free from charges,
with a permittivity (s) and a permeability μ(s) [PET 93] (see equation [C.1]).
The Maxwell equations imply a time domain field dependence with e−jωt .
s, ω) + jωμ(s)H(
∇ × E( s, ω) = 0
s, ω) − jω(s)E(
∇ × H( s, ω) = 0
[C.1]
s, ω) = 0
∇ · (s)E(

s, ω) = 0
∇ · (s)E(
For a homogenous dielectric medium, the permittivity and the permeability

are not dependent on the position ((s) = and μ(s) = μ).
The study of propagation problems in homogenous media is conducted by

using Helmholtz vectorial equations which are derived from Maxwell equations:
(s, ω) + k 2 U
∇2 U (s, ω) = 0 [C.2]
with U (s, ω) the electric or magnetic field and k = ω √μ = 2π = 2π √r μr the
λ λ0
wavenumber of the propagation medium. λ and λ0 are the wavelengths in the

propagation medium and in the free space respectively. r = r − j(r + 60σλ)

and μr = μr − jμr correspond to the relative permittivities and the
permeabilities of the medium. The conductivity of the medium is represented
by σ.
C.1.2. Field locality principle
In a homogenous medium, energy propagates on paths which are straight

and orthogonal to the wavefront. These wavefronts are defined by the wave
surfaces which are plane, spherical, cylindrical or any other shape. A group of
rays is a beam or tube which starts and ends with two caustic segments (see
AB and CD in Figure C.1). The energy carried by a ray is persistent in the
time and space domains as well as in magnitude and phase.
Caustics
φ̂ θ̂
A
C
ŝ propagation
P(s) direction
P(0)
D
B
ρ2 ρ1
Figure C.1. Ray tubes, caustics and local base
As there is always a scale for which a wave can be considered as locally plane,
the properties of transverse electromagnetic plane waves can thus be generalized
to all the electromagnetic waves [GAR 87]. This is the electromagnetic wave
Geometric Optics and Uniform Theory of Diffraction 191
locality principle. The electric and magnetic waves are strictly transverse to the
propagation direction s. The electric field vector is in the plane orthogonal to
the propagation direction. The trihedron (E, H,
s) is orthonormal, direct and
follows relation [C.3].
"
= μ
E (H × s) [C.3]

The study can thus be restricted to the case of the electric field E which
is expressed in a local basis B = (ŝ, θ̂, φ̂). This local basis depends on the
propagation direction. For reasons of readability, a simplified notation is
adopted in the following text which allows us to mix up the vector E with its
vectorial representation E.
C.1.3. Field expression in geometric optics
According to the work of Luneberg and Kline, the electric field is expressed
in GO by [GLO 99]:
E(s) = A(s, ρ1 , ρ2 ) E(0) e−jkΨ(s) [C.4]

⎡ ⎤ ⎡ ⎤
0 0
E(s) = ⎣ Eθ (s) ⎦ = A s, ρ1 , ρ2 ⎣ Eθ (0) ⎦ e−jkΨ(s) [C.5]
Eφ (s) B Eφ (0) B
A(s, ρ1 , ρ2 ) corresponds to the ratio between the field magnitudes E(s) and
E(0). It is also called the divergence factor:
"
ρ1 ρ2
A(s, ρ1 , ρ2 ) = [C.6]
(ρ1 + s)(ρ2 + s)
– s is the covered distance between P (0) and P (s).
– Ψ(s) is a phase function at the observation point P (s) and is equal to s.
– E(0) is the electric field at the point P (0).
ρ1 and ρ2 are the two main curvature radii of the wavefront measured on the
central ray at the reference point P (0). This point corresponds to the source
position or the interaction point between the wave and a surface.
As the electric field is transverse to the propagation direction ŝ, the

component along ŝ is zero. In the following, the field will be expressed in its
basis B without its component along ŝ (see equation [C.7]):

Eθ (s) Eθ (0)
E(s) = = A s, ρ1 , ρ2 e−jks [C.7]
Eφ (s) B Eφ (0) B
The curvature radii ρ1 and ρ2 vary along the ray trajectory. They have to
be recalculated whenever the ray is obstructed because the interactions modify
the ray trajectory. An interaction can be a reflection, a transmission (or double
refraction) or a diffraction.
C.1.4. Change of local basis
Considering an electric field expressed in a given basis B1 , if the wave

corresponding to this field undergoes an interaction, the field has to be
expressed in a new local basis. This new local basis corresponds to the input
basis B2 of the interaction. To obtain the new expression of the field, we need
to use the basis change matrix MB1 →B2 (see equation [C.8]).

Eα2 (0) B1 →B2 Eα1 (0)
E(0) = =M [C.8]
Eβ2 (0) B Eβ1 (0) B
2 1
B1 = (ŝ1 , α̂1 , β̂1 ) and B2 = (ŝ2 , α̂2 , β̂2 ) are the direct orthonormal bases in
which the incident field is expressed at the interaction input.
MB1 →B2 is the transition matrix allowing us to express the incident field in
the new basis. This matrix corresponds to a projection of the vectors α̂1 and
β̂1 in the basis B2 :

B1 →B2 α̂2 · α̂1 α̂2 · β̂1
M = [C.9]
β̂2 · α̂1 β̂2 · β̂1
C.1.5. Incident field
The incident field is the field radiated by a source S in the direction of

an observation point P placed at a distance si . The expression of this field is
i r
deduced from equation (C.7) by applying the transition matrix MB →B :

i i

E s i i Bi →Br Ei (0)
e−jks
i
Er si = i = A s, ρ ,
1 2ρ M i [C.10]
E⊥ si Br E⊥ (0) Bi
ρi1 and ρi2 are the main curvature radii of the incident wavefront.
Bi = (ŝi , êi , êi⊥ ) is a direct orthonormal local basis at the emission. It

corresponds to a basis described by the direction ŝi . If we consider a spherical
frame, êi and êi⊥ can be assimilated to vectors êθ and êφ which form with ŝi a
direct orthonormal basis.
Br = (−ŝi , êr , êr⊥ ) is a direct orthonormal local basis at the receiver side.
It corresponds to the basis described by the direction −sî . If we consider a
spherical frame, êr and êr⊥ can be assimilated to the vectors êθ and −êφ which
form with −sî a direct orthonormal basis.
MB →B is the transition matrix which allows us to express the incident

i r
field at the point P in the good basis. This matrix corresponds to a projection
of the vectors êi and êi⊥ in the basis Br .

Bi →Br êr · êi êr · êi⊥ 1 0
M = r i = [C.11]
ê⊥ · ê êr⊥ · êi⊥ 0 −1
Most of the time, we consider that the incident wavefront is spherical. So,
the field is expressed by:
1 Bi →Br i
E (0)e−jks
i
Er (si ) = M [C.12]
si
C.1.6. Reflected field
The reflected field is the field received at a point P after the reflection of
a ray at a point Qr placed at a distance si from the source point S and at a
distance sr to the point P (see Figure C.2).
C.1.6.1. Expression of the reflected field in GO

The expression of the reflected field Er (sr ) is derived from the fundamental
expression of the field in GO (see equation [C.7]). The reflection law which
comes from the Fermat principle defines the reflection point Qr . The incident
ŝi and reflected propagation ŝr directions are fixed by the interaction surface
by the normal vector n̂ (see equation [C.13]) [MCN 90].

n̂ × ŝi − ŝr = 0̂ [C.13]
The incidence and reflection planes are the same and the incidence angle θi
and reflection angle θr are equal (see Figure C.2). The incidence plane is defined
S
ŝr
êr
êi
êr⊥ local basis Br
êi⊥ n̂
ŝi
local basis Bi θr
θi
Qr
tangent plane of a surface
Figure C.2. Incident and reflection local bases
by the normal n̂ and the incident ray given by the vector ŝi . The reflection plane
is defined by the normal n̂ and the reflected ray is given by the vector ŝr . The
use of the locality principle in the case of the reflected field corresponds to
the introduction of a reflection dyad R which is expressed according to the
reflection coefficients R and R⊥ by:

R 0
R= [C.14]
0 R⊥
Thus, the reflected field Er (0) at the point Qr and defined in the basis
i
B only depends on the dyad R and on the incident field Ei (si ) defined in the
basis B at this same point. It is thus necessary to express the field in the correct
incident local basis Bi = (ŝi , êi , êi⊥ ).
i
Er (0) = R MB→B Ei (si ) [C.15]
In relation [C.15], B = (ŝi , α̂, β̂) corresponds to the orthonormal basis in

which the incident field Ei (si ) is defined. α̂ and β̂ are an arbitrary couple
i
of orthonormal vectors, such that B is direct. The transition matrix MB→B
allows us to describe the field Ei (si ) in the basis B i .

B→B i êi · α̂ êi · β̂
M = i [C.16]
ê⊥ · α̂ êi⊥ · β̂
The reflected field Er (sr ) at the observation point P is expressed in the

reflected local base B r = (ŝr , êr , êr⊥ ) by the equation [C.17]. The main curvature
radii ρr1 and ρr2 of the reflected wave depend on those of the incident wave. In
the case of a planar surface, the following equalities are obtained: ρr1 = ρi1 and
ρr2 = ρi2 .
i
Er sr = A sr , ρr1 , ρr2 RMB→B Ei si e−jks
r
[C.17]
C.1.6.2. Incident and reflected ray bases

The local bases Bi and Br are defined according to the incident ray
propagation direction ŝi and the normal n̂. In fact, êi and êr are parallel to
the incidence and reflection planes, while êi⊥ = êr⊥ are orthogonal to the same
planes (see Figure C.2).
C.1.6.3. Reflection coefficients

The dyad R depends on the reflection coefficient components which are
parallel R and perpendicular R⊥ to the incidence plane (see equation [C.14]).
These coefficients are for the phase and the magnitude changes introduced
by the reflection phenomenon on each component of the field. For multi-layer
interfaces with one or more layers, these coefficients have to be adapted in
order to take the stratification into account. Equation [C.18] corresponds to the
reflection coefficient in the case of a surface with a thickness e and a permittivity
r . The permeability is considered to be set at 1.

1 − e−2jδ e2jδ
R,⊥ = Γ,⊥ [C.18]
1 − Γ2,⊥ e−2jδ e2jδ
Γ,⊥ are the Fresnel coefficients in the case of an infinite plane surface with
a permittivity r :
r cos θi − r − sin2 θi
Γ = [C.19]
r cos θi + r − sin2 θi
cos θi − r − sin2 θi
Γ⊥ = [C.20]
cos θi + r − sin2 θi
In the case of a perfect conductor surface (σ → ∞), the reflection coefficients

are expressed with a phase difference multiple of π by:
Γ = +1 Γ⊥ = −1 [C.21]

δ = kr l and δ k0 d2 are the phase terms associated with the delays
√
introduced by each reflected field (see Figure C.3). kr = k0 r corresponds to
the wavenumber of the medium with permittivity r and l = e/ cos θt is the
one-way distance in the medium. The incidence θi and refraction θt angles are
defined by:
1
sin θt = √ sin θi [C.22]
r
0 r 0
Transmitted
rays
Reflected
rays
θi
θt
θi
e
Incident ray
0 r 0
d d
d = 2l sin θi sin θt
l
e
Figure C.3. Reflection of a plane wave on a dielectric interface with thickness e
The expressions of R,⊥ do not consider the possible roughness of the

encountered surfaces (see equation [C.18]). The roughness is introduced by
multiplying the coefficients Γ,⊥ by the roughness factor ρ (see equation [C.23])
[BEC 87] [COC 05]. h is the height standard deviation in the rough interface.
2
ρ = e−2(kr h cos θi ) [C.23]
C.1.7. Refracted and transmitted field
The refracted (or transmitted) field is the received field at the point P after
refraction (or multiple refractions) of a wave at the point Qt placed on an
interface at a distance si of the source point S and at a distance st of the point
P.
S
ê i
ê⊥i
ŝ i
local basis B i
ê t
θi θt
n̂ Qt
ê⊥t ŝ t P
t
local basis B
Tangent plane
at the refraction
surface
Figure C.4. Transmission and incidence local bases
C.1.7.1. Expression of refracted and transmitted field in OG

The expression of the transmitted field Et (st ) at the observation point P,
placed at a distance st of the transmission point, also follows from the GO field
main expression (see equation [C.7]).
The refraction law, also called the Snell-Descartes law, comes from the
Fermat principle. It establishes the relation between the incidence angle θi and
the transmission angle θt (see Figure C.4) at the interface and in the direction
air → medium (see equation [C.22]). The Fermat principle implies that the
incidence and transmission planes are the same and defined by the incident
ŝi ray and transmitted ŝt ray directions, and the reflection surface normal
n̂. Identically to the reflection example, the transmission can be considered

as a local phenomenon, so the transmitted field is expressed by introducing
a transmission dyad T. This dyad is obtained according to the transmission
coefficients (see equation [C.24]).

T 0
T= [C.24]
0 T⊥
The field Et (0) after the transmission point Qt depends on the coefficient
T and on the incident field Ei (si ) at this same point, taking care to define the
incident field in the incident local basis Bi = (ŝi , êi , êi⊥ ) (see equation [C.25]).
i
Et (0) = TMB→B Ei si [C.25]
B = (ŝi , α̂, β̂) corresponds to the orthonormal basis in which the incident
field Ei (si ) is defined. As in the reflected field case, the transition matrix
i
MB→B allows the expression of the incoming field Ei (si ) in the incident local
basis B i .
The transmitted field Et (st ) at the observation point P is expressed in the

transmission local basis Bt = (sˆt , êt , êt⊥ ) by:
i
Et st = A st , ρt1 , ρt2 TMB→B Ei si e−jks
t
[C.26]
In the simple refraction case, st corresponds to the distance between P and

Qt . As the point P is inside the interface, the propagation constant considers
the propagation speed related to the dielectric properties at the interface. For
a multiple refraction (or transmission), st corresponds to the distance between
P and Qt at which the propagation distance in the interface is removed. This
last distance is obtained for propagation in free space. Indeed, in the case of
multiple refraction, the phase delay associated with propagation in the interface
is directly introduced in T.
The divergence of a transmitted ray is given by the following relation:

!
t t t ρt1 ρt2
A s , ρ1 , ρ2 = [C.27]
(ρ1 + st )(ρt2 + st )
t
ρt1 is the first main curvature radius of the transmitted wave. Its expression
is different from the case of simple refraction (see equation [C.28]) and the case
of double refraction (see equation [C.29]).
ρt1 = ρi1 α−1 [C.28]

ρt1 = ρi1 + αl [C.29]
ρt2 is the second main curvature radius of the transmitted wave. Its
expression is also different from the case of the simple refraction [C.30] and
the case of the second refraction [C.31] [PLO 03].
ρt2 = ρi2 α−1 [C.30]

2
ρt2 = ρi2 + αγ l [C.31]
e 1 cos θi
l= α= √ γ= [C.32]
cos θt r cos θt
C.1.7.2. Incident and transmitted ray bases

The knowledge of the incident ray propagation direction ŝi and the normal n̂
allows us to define the local bases Bi and Bt . In fact, êi and êt are respectively
parallel to the incidence and transmission planes and êi⊥ = êt⊥ are perpendicular
to the same planes. êi and êi⊥ are defined according to the normal n̂ (see
Figure C.2). êt⊥ is obtained from equation [C.33]. Generally, after a double
refraction Bi and Bt are the same because sî = sˆt .
êt = êt⊥ × sˆt [C.33]
C.1.7.3. Transmission coefficient

In the expression of the transmission dyad, the two components T and
T⊥ are respectively the transmission coefficients parallel and orthogonal to
the incidence plane. They correspond to the phase and magnitude changes
introduced by the transmission phenomenon on each component of the field. In
the case of a simple refraction and a plane surface, their expressions are given
by equations [C.34] and [C.35]. Γ and Γ⊥ correspond to the Fresnel reflection
coefficients defined at equations [C.19] and [C.20].
√
1 + Γ 2 r cos θi
T = √ = [C.34]
r r cos θi + r − sin2 θi
2 cos θi
T⊥ = 1 + Γ⊥ = [C.35]
cos θi + r − sin2 θi
In the case of a double refraction (transmission), T and T⊥ are expressed by

equation [C.36] (see Figure C.3). In [SAG 03], the expression of the transmission
coefficient for an interface with more than two materials is proposed.

1 − Γ2,⊥ e−2jδ e2jδ
T,⊥ = [C.36]
1 − Γ2,⊥ e−2jδ e2jδ
C.2. Uniform theory of diffraction
C.2.1. Introduction
Electromagnetic waves are continuous, in magnitude and phase, in the time

and space domain. However, the GO does not ensure the total continuity field.
It predicts areas where the field is zero (shadow zone). In 1953, Keller proposed
a generalization of the GO in order to consider diffracted rays. This led to the
birth of general diffraction theory (GTD) [KEL 62].
The GTD allows us to solve the problem of field existence in the shadow
zone of the GO. However, it presents a singularity at the boundary between
a lighted zone and a shadow zone. So in 1962, the theory has been completed
by Kouyoumjian and Pathak in order to ensure the total field continuity in all
space points [KOU 74]; henceforth we talk about UTD.
Thereafter, Burnside and Burgener have proposed a formalism of the

uniform theory of diffraction (UTD) in the case of the 3D diffraction on a
corner of small thickness [BUR 83]. Other authors have also studied the
simple diffraction, the double diffraction as well as the slope diffraction in
the frequency domain [LUE 89, DES 84, ROU 96, ROU 99] and in the time
domain [VER 90, ROU 95]. Then, only the expression of the diffracted field in
the frequency domain is addressed, considering the case of the diffraction by
dielectric or metallic dihedron.
C.2.2. Diffracted field
The expression of the diffracted field given by equation [C.37] corresponds to

the diffraction by a dihedron. For this kind of diffraction, the discontinuity line
is mixed up with the one of the diffracted beam caustics and so the curvature
radius ρd2 , also called minor ray, becomes zero.
i
Ed sd = A sd , ρd1 , ρd2 → 0 DMB→B Ei si e−jks
d
[C.37]
A(s) is the divergence factor of the ray diffracted by the dihedron and is
expressed by:
!
d d d ρd
A s , ρ1 , ρ2 → 0 = d 1 d d [C.38]
ρ1 + s s
D is the dyad introducing the diffraction coefficients. This dyad can be

written differently according to the diffracted ray’s nature, which can be
two-dimensional (2D) or three-dimensional (3D).
i
The matrix MB→B makes it possible to describe the incident field,
initially expressed in the basis B, in the incidence basis of the diffraction
Bi = (ŝd , êi , êi⊥ ). The diffracted field Ed (sd ) is defined in the diffraction basis
Bd = (ŝd , êd , êd⊥ ).
The diffraction law, coming from the generalized Fermat principle, connects
the Keller angle β0 to the incident ray direction ŝi and the diffracted ray
direction sd as well as the tangent t̂ of the dihedron wedge (see Figure C.5) at
the diffraction point Qd by:
ŝi · t̂ = ŝd · t̂ = cos β0 [C.39]
The incidence and diffraction planes are thus defined by the tangent of the
wedge t̂ and by the incident and diffraction ray directions. Generally, these two
planes are not merged, as illustrated in Figure C.5.
The vectors of the bases Bi and Bd are obtained using the following
relations:
t̂
êi = −ŝi × êi⊥ = ŝi × êi
sin β0
[C.40]
t̂
êd = ŝ ×
d
êd⊥ = ŝ ×
d
êd
sin β0
C.2.3. UTD 2D coefficient
In the case of a 3D ray, the dyad D is diagonal and directly expressed from
the two components of the diffraction coefficients D and D⊥ :

D 0
D= [C.41]
0 D⊥
The terms D and D⊥ are the diffraction coefficients initially introduced by

Keller and then modified by Kouyoumjian, Burnside and Luebbers.
ne
pla
n P
tˆ tio
rac
ff
Di
Incidence plane ŝd
êd
local basis Bi êd⊥
β0
êi local basis Bd
ŝi Qd
S φd
β0
êi⊥
φi
(2 − n)π
Figure C.5. Incidence and diffraction local bases
C.2.3.1. UTD coefficient – dihedron conductor

Contrary to the GTD, the UTD introduced by Kouyoumjian and Pathak
[KOU 74] ensures the field continuity in all the space, especially at the optic
boundaries ISB (incident shadow boundary) and RSB (reflection shadow
boundary). These authors have taken an interest in the case of a perfect
conductor dihedron and have defined two new diffraction coefficients which
introduce a correcting factor compared to the GTD coefficients:
D//,⊥ = D1 + D2 ± (D3 + D4 ) [C.42]

+ ,
e−j
π
4 π + (φd − φi )
D1 =− √ cot F kLa+ (φd − φi )
2n 2πk sin βo 2n
+ ,
e−j 4
π
π − (φd − φi )
D2 =− √ cot F kLa− (φd − φi )
2n 2πk sin βo 2n
+ ,
[C.43]
e−j 4
π
π + (φd + φi )
D3 =− √ cot F kLa+ (φd + φi )
2n 2πk sin βo 2n
+ ,
e−j 4
π
π − (φd + φi )
D4 =− √ cot F kLa− (φd + φi )
2n 2πk sin βo 2n
β0 is a semi-angle at the diffraction cone head. The normal incidence

corresponds to βo = π2 (see Figure C.5).
φi and φd are respectively the incidence and diffraction angles. They are
specified from the wedge face 0 (see Figure C.5).
n is a parameter defined so that the dihedron inner angle is given by (2−n)π.
± corresponds to the reflection coefficient in the case of a conductor [C.21].
F (x) corresponds to the transition function which uses the Fresnel integral
[LEG 95]:
∞
√ 2
F (x) = 2j xe jx
√
e−jt dt [C.44]
x
L is a distance parameter which corresponds to the incidence type on the

wedge. It brings into play the incident wave main curvature radii (ρd1 and ρd2 )
as well as the wedge of curvature radius (ρde ) (see equation [C.45]). In the case
of a spherical wave, L is given by equation [C.46].
ρde + sd ρd1 ρd2

L = sin2 β0 sd d d d
[C.45]
ρe ρ1 + s ρ2 + sd
d
sd ρd1
L = sin2 β0 [C.46]
ρd1+ sd
a± is a function defined by:

2nπN ± − (φd ± φi )
a± = 2 cos2 [C.47]
2
with N ± an integer corresponding (outside the optic boundaries) to relation

[C.48]:
2πnN ± − (φd ± φi ) = ±π [C.48]
When the cotangent (cot) of one of the Di (i = 1, 2, 3, 4) terms becomes

singular for the
√
optic boundary angles φi and φd , the corresponding term is
replaced by − 2 sign() [PLO 00].
L
C.2.3.2. UTD coefficients – dielectric dihedron

Luebbers proposes a more general formulation for the diffraction coefficients.
This formulation allows us to consider the dielectric nature of the wedge as well
as its roughness [LUE 89]. These coefficients are heuristic solutions as they are
not obtained by solving Maxwell equations:

D//,⊥ = Gn//,⊥ D1 + ρn R//,⊥
n
D3 + G0//,⊥ D2 + ρ0 R//,⊥0
D4 [C.49]
Di (i = 1, 2, 3, 4) are terms reported on equations [C.43].

0,n
R//,⊥ corresponds to the Fresnel reflection coefficients [C.19] and [C.20].
The incidence angles θi0 = π2 − φi and θin = π2 − (nπ − φi ) depend on the faces
0 and n of the considered wedge.
ρ0,n corresponds to the roughness coefficient of the considered face [C.23].
G0,n
//,⊥ are the correcting coefficients which allow us to consider the grazing
incidences on the faces 0 and n of a dihedron.
⎧
⎪ 1
⎪
⎪ for φi = 0 and |1 + R//,⊥
0
|>0
⎪ 0
⎨ 1 + R//,⊥
G0//,⊥ = 1 [C.50]
⎪
⎪ for φi = nπ
⎪
⎪ 2
⎩
1 otherwise
⎧
⎪ 1
⎪
⎪ for φi = nπ and |1 + R//,⊥
n
|>0
⎪
⎨ 1 + R//,⊥
n
Gn//,⊥ = 1 [C.51]
⎪
⎪ for φi = 0
⎪2
⎪
⎩1 otherwise
C.2.4. UTD 3D coefficient
Contrarily to the case of a 2D ray, the dyad D of a 3D ray is no longer

diagonal and is expressed only by two diffraction coefficients, as in the case of
2D ray (see equation [C.49]). This 3D dyad has been introduced by Burnside
and Burgener [BUR 83] and reused by Rouvière [ROU 99] in the case of a
dielectric half-plane with very small thickness.
To obtain the 3D dyad of [BUR 83], field continuity has to be ensured at

each optic limit. So, the boundary conditions at the reflection in incidence
boundaries have to consider relations [C.52] and [C.53]. In these relations, Ui

and Ur correspond to the incident and reflected fields respectively.
*
e−jks −1/2Ur area I
d
i
RU (Qd )D(φ + φ ) √d i
= [C.52]
sd 1/2Ur area II
*
e−jks −1/2(1 − T)Ui area II
d
(1 − T)Ui Qd D φd − φi √ = [C.53]
sd 1/2(1 − T)Ui area III
with the areas denoted I, II and III illustrated by Figure C.6.
RSB
Zone I Zone II
incident + reflected + diffracted incident + transmitted
ISB
Zone III
diffracted + transmitted
diffracted wedge
Figure C.6. Field area around a diffracting wedge
Using these boundary conditions, the following expression is obtained for

the dyad D:
D = (I − T )D(φd − φi ) + R D(φd + φi ) [C.54]

D(φd − φi ) = D1 + D2 [C.55]
D(φd − φi ) = D1 + D2 [C.56]
Each of the dyads (D, R and T ) are expressed in their respective local
basis. It is thus necessary to calculate the base change matrices in order
to express all the fields in the local basis of the diffraction. α is the angle
between the incident-reflection plane as well as the incidence-transmission
plane and the incident plane. The diffraction plane is given by π2 − α. In 2D,
the incidence-reflection and incidence-transmission planes are perpendicular to

the incidence-diffraction plane (α = 0). In 3D, these planes have no particular
direction (see Figure C.7).
ion
flect
ne - re
ce pla
den
inci
P
tˆ
Di
ffr
ac
ŝd
tio
n
êd
In
pla
cid
π
−α n̂
ne
êd⊥
en
2 θr
ce
êi
pla
θi
ne
ŝi Qd
S φd
β0
êi⊥
φi
(2 − n)π
Figure C.7. Reflection and diffraction incidence planes
The relations between the reflection and diffraction local bases are
illustrated at Figure C.8 allow us to establish the matrix of basis change
necessary to express the reflection R (see equation [C.57]) and transmission
T dyads in the diffraction local basis (see equation [C.57]). In these equations,
the exponent terms correspond to the plane nature (i for incident and d for
diffraction) as well as the interaction specifying the plane (r for reflection and
d for diffraction).
⎡ ⎤
êi,d i,r
êi,d i,r

· ê · ê⊥ − sin α
MiR→D (α) = M(α) = ⎣ i,d i,r ⎦ = cos α [C.57]
ê⊥ · ê êi,d · êi,r sin α cos α
⊥ ⊥
⎡ ⎤
êd,d d,r
êd,d d,r

· ê · ê⊥ cos α sin α
MdR→D (α) = M(−α) = ⎣ d,d d,r ⎦= [C.58]
ê⊥ · ê êd,d d,r − sin α cos α
⊥ · ê⊥
êi,r
⊥ êd,d
⊥
êi,d
⊥ êd,r
⊥
α α
êi,d

α
êi,r
êd,d

α
êd,r

before diffraction after diffraction
Figure C.8. Incidence reflection and diffraction bases in 3D case
In the case of transmission, the same basis change matrices are used. They
are denoted M(−α) because we turn conversely. The dyadic matrices are thus
expressed by R and T by using equations [C.59] and [C.60].
−1
R = MdR→D (α) R MiR→D (α)
= M(−α) R M(−α) [C.59]

R cos2 α − R⊥ sin2 α (R + R⊥ ) cos α sin α
=
−(R + R⊥ ) cos α sin α R⊥ cos2 α − R sin2 α
−1
T = MdT →D (α)T MiT →D (α)
= M(α)TM(−α) [C.60]

T cos2 α + T⊥ sin2 α (T − T⊥ ) cos α sin α
=
(T − T⊥ ) cos α sin α T⊥ cos2 α + T sin2 α

1 − T cos2 α − T⊥ sin2 α −(T − T⊥ ) cos α sin α
(I − T ) = [C.61]
−(T − T⊥ ) cos α sin α 1 − T⊥ cos2 α − T sin2 α
Expressions [C.59] and [C.61] inserted in equation [C.54] underline the

non-diagonal structure of the dyad D which is expressed by [BUR 83] [GLO 99]:

Da Db
D= [C.62]
Dc Dd

Da = 1−T cos2 α−T⊥ sin2 α D φi − φd −[R cos2 α − R⊥ sin2 α D φi + φd
[C.63]
Db = cos α sin α [(T⊥ − T ) D(φi − φd ) + (R + R⊥ ) D(φi + φd )] [C.64]

Dc = cos α sin α [(T⊥ − T ) D(φi − φd ) − (R + R⊥ ) D(φi + φd )] [C.65]
Dd = [1−T sin2 α−T⊥ cos2 α]D(φi − φd )−[R⊥ cos2 α−R sin2 α]D(φi + φd )
[C.66]
In equations [C.63] [C.64] [C.65] and [C.66], the terms R,⊥ and T,⊥ are
the reflection and transmission coefficients, the expressions of which are given
by [C.18] and [C.36].
Appendix D
Ray Construction Techniques
D.1. Ray launching
Ray launching is a forward approach which consists of sending rays in all the
directions from a transmission position of coordinates (Tx) with an incremental
angular step (Δα) which can be set to various values. Then, the receiver position
coordinates (Rx) are specified (see Figure D.1) [CHE 96, KIM 97, STA 98,
LIN 89].
The speed of ray determination with the ray launching approach is directly
related to the chosen Δα. When the Rx position is fixed, the rays connecting
the transmitter Tx and the receiver Rx are obtained considering a receiver
sphere of diameter Δd which can be set to various values. So, the launched rays
connecting Tx to Rx will be considered if they intersect the receiver sphere (see
Figure D.1). The number of rays obtained increases with the sphere diameter
Δd (for a given Δα), but the precision of the calculated field and the received
signal decreases.
D.2. Ray tracing
Ray tracing is a backward approach which consists of applying the image

principle from Tx and Rx position in order to obtain reflected rays (see
Figure D.2) [HUM 03, MCK 91, CHA 03].
The specificity of this approach is the increase of the number of the potential
images combinations with the complexity of the considered environment. This
can lead to important ray calculation time.
δα
Tx
(a)
δα
δd Rx
Tx
(b)
Figure D.1. Ray launching principle: (a) rays launched from the transmitter
and (b) rays obtained considering intersection of receiver sphere

Im Im Tx /s1 /s2

Im Tx /s1
s1
s2
Tx Rx
Figure D.2. Ray tracing principle

Ray Construction Techniques 211
D.3. Other techniques
In the literature, other techniques are also proposed and used for ray
determination. Some of these techniques are hybrid and combine ray launching
and ray tracing [TAN 95, TCH 03, AVE 04], while others require a visibility
tree to be built [AGE 97, AGE 00], or rely upon calculation time speed
improvement techniques borrowed from the image processing domain. These
last techniques use octree, raster or Voxel matrices for ray determination
[FOL 84, LEG 05].
Appendix E
Description of the Time-Frequency Transform
The determination of a time domain signal x(τ ) from its frequency domain
expression X(f ) is generally made using an inverse Fourier transform (see
equation [E.1]). As the frequency domain expression X(f ) of the signal x(τ )
shows a Hermitian symmetry (see equation [E.2]), the obtained signal x(τ ) is
real.
+∞
x(τ ) = X(f ) ej 2 π f τ df [E.1]
−∞
∗
X (f ) = X(−f ) ∀ f ∈ R [E.2]
From frequency domain measurements, the obtained channel transfer

function H(f ) is defined on a finite frequency band from fmin to fmax , for
Nf frequency points. As H(f ) is discrete, the determination of h(τ ) is made
using an inverse Fourier transform (see equation [E.3]). δf = fmaxN−f
f
min
is the
frequency sampling step and fk corresponds to a frequency value given by
fk = fmin + k δf .
Nf −1

h(τ ) = δf H(fk ) ej 2 π fk τ [E.3]
k=0
The transfer function H(f ) does not show a Hermitian symmetry because it
is only defined between fmin and fmax . The signal obtained after discrete inverse
Fourier transform is complex and corresponds to h(τ ) + j ĥ(τ ) (see Figure E.1).
214
real part imaginary part
0.08 0.08
0.06 0.06
0.04 0.04
0.02 0.02
0 0
−0.02 −0.02
−0.04 −0.04
−0.06 −0.06
−0.08 −0.08
0 10 20 30 40 50 0 10 20 30 40 50
Figure E.1. Inverse Fourier transform of H(f )

Description of the Time-Frequency Transform 215
To obtain a real signal h(τ ), it is mandatory to use a transfer function H2 (f )

presenting by construction a Hermitian symmetry. Moreover, to improve the
precision of the reconstructed h(τ ), we can expand the spectral support of
H2 (f ) in comparison to H(f ). This consists of making an operation of zero
padding. Indeed, the product δτ δf Nf is equal to 1, so the increase of Nf
necessary leads to a reduction of δτ and consequently improves the time domain
precision of the signal h(τ ) obtained after the discrete inverse Fourier transform.
The reconstruction of a real h(τ ) from H(f ) is made by applying

two successive operations on H(f ) before the discrete inverse Fourier
transformation: zero padding and Hermitian symmetric forcing.
Operation of zero padding

The operation of zero padding consists of obtaining a new transfer function
H1 (f ) from H(f ) (see Figure E.3). The new transfer function H1 (f ) contains
far more samples than H(f ). H1 (f ) is constructed according to the following
relation with fe = 2 1δτ .
*
H(f ) f ∈ fmin , fmax
H1 (f ) = [E.4]
0 f ∈ δf, fmin − δf ∪ fmax + δf, fe
So, the new transfer function H1 (f ) is constructed as shown in equation

[E.5]) with N1 = Nr + Nf + Nl samples where Nl = fδfmin
− 1 and Nr = fe −f
δf
max
.

H1 (f ) = zeros(1, Nl ) H(f ) zeros(1, Nr ) [E.5]
Figure E.2 shows the time domain signal obtained after discrete inverse
Fourier transformation is applied on H1 (f ). We can notice that the sampling
of the obtained time domain signal is improved. Nevertheless, it is still complex.
Operation of Hermitian symmetric forcing

The operation of forcing the Hermitian symmetry consists of creating a
transfer function H2 (f ) from H1 (f ). This new transfer function H2 (f ) is defined
from −fe to fe (see equation [E.6]) as illustrated in Figure E.3. Figure E.4
represents a real impulse response h(τ ) obtained after a discrete inverse Fourier
transform applied on H2 (f ).
*
H2 (f ) = H1 (f ) for f > 0
∗
[E.6]
H2 (−f ) = H2 (f ) ∀f ∈ δf, fe
216
0.08 0.08
0.06 0.06
0.04 0.04
0.02 0.02
0 0
−0.02 −0.02
−0.04 −0.04
−0.06 −0.06
−0.08 −0.08
0 10 20 30 40 50 0 10 20 30 40 50
Figure E.2. Zero padding and inverse Fourier transform on H1 (f )

H (f ) + zero padding
±f f min f max fe f
Operation of symmetric f orcing
H *(f) + f lip zero padding H (f ) + zero padding

Description of the Time-Frequency Transform
−fe −f max −f min 0 f min f max fe f

Figure E.3. Illustration of the Hermitian symmetric forcing
217
218
0.08 0.08
0.06 0.06
0.04 0.04
0.02 0.02
0 0
−0.02 −0.02
−0.04 −0.04
−0.06 −0.06
−0.08 −0.08
0 10 20 30 40 50 0 10 20 30 40 50
Figure E.4. Reconstruction of h(τ ) after discrete inverse Fourier transform applied
on H2 (f )
Bibliography
[ABD 00] Abdi A., Kaveh M., “Performance comparison of three different
estimators for the Nakagami m parameter using Monte Carlo simulation”, IEEE
Communications Letters, vol. 4, no. 4, p. 119–121, April 2000.
[ABD 01] Abdi A., Tepedelenlioglu C., Kaveh M., Giannakis G., “On
the estimation of the K parameter for the Rice fading distribution”, IEEE
Communications Letters, vol. 5, no. 3, p. 92–94, March 2001.
[AFF 00] Affandi A., Caractérisation et modélisation de la propagation à l’intérieur
des bâtiments dans les bandes de 450 - 900 - 1800 - 2400 - 5800 MHz, PhD Thesis,
Institut National des Sciences Appliquées de Rennes, France, 2000.
[AGE 97] Agelet A., Fontan F., Formella A., “Fast ray-tracing for microcellular
and indoor environments”, IEEE Transactions on Magnetics, vol. 33, p. 1484–1487,
March 1997.
[AGE 00] Agelet A., Formella A., Rabanos J., De viente F., Fontan F.,
“Efficient ray tracing acceleration techniques for radio propagation modeling”,
IEEE Transactions on Vehicular Technologies, vol. 49, p. 2089–2104, November
2000.
[AIE 03a] Aiello R., Ellis J., Kareev U., Siwiak K., Taylor L., Understanding
UWB – principles and implications for low-power communications, Report
no. IEEE P802.15-03/157, IEEE P802.15 Working Group for WPANs, March
2003.
[AIE 03b] Aiello R., Rogerson G. D., “Ultra-wideband wireless systems”, IEEE
Microwave Magazine, vol. 4, no. 2, p. 36–47, June 2003.
[ALV 03] Alvarez A., Valera G., Lobeira M., Torres R., Garcia J. L., “New
channel impulse response model for UWB indoor system simulations”, IEEE
Vehicular Technology Conference, VTC Spring, vol. 1, Seoul, Korea, p. 1–5, April
2003.
[ART 95] Arthur J., “Modern SAW-based pulse compression systems for radar
applications.I. SAW matched filters”, Electronics & Communication Engineering
Journal, vol. 7, no. 6, p. 236–246, December 1995.
[ATT 03] Attiya A., Safaai-Jazi A., “Time domain characterization of receiving
TEM horn antennas”, IEEE Antennas and Propagation Society International
Symposium, Columbus, OH, USA, p. 233–236, June 2003.
[ATT 04] Attiya M., Saffai-Jazi A., “Simulation of ultra-wideband indoor
propagation”, Microwave and Optical Technology Letters, vol. 42, no. 2, p. 103–108,
May 2004.
[AVE 04] Aveneau L., Combeau P., “Path finding based on Monte Carlo techniques
compared with a full ray-tracing approach in narrow and wide bands”, IEEE
Vehicular Technology Conference, Milan, Italy, May 2004.
[BAL 04a] Balakrishnan K., Wee K. C., Xu S., Chuan C. L., Chin F., Huat
C. Y., Choy C. C., Thiang T. T., Xiaoming P., Ong M., Krishnan S.,
Characterization of Ultra-Wideband Channels: Small-Scale Parameters for Indoor
& Outdoor Office Environments, Report no. IEEE 802.15-04/385, IEEE 802.15
Working Group for Wireless Personal Area Networks (WPANs), July 2004.
[BAL 04b] Balakrishnan K., Wee K. C., Xu S., Chuan C. L., Chin F., Huat
C. Y., Choy C. C., Thiang T. T., Xiaoming P., Ong M., Krishnan S.,
UWB Channel Characterization in Indoor Office Environments, Report no. IEEE
802.15-04/439, IEEE 802.15 Working Group for Wireless Personal Area Networks
(WPANs), August 2004.
[BAR 91] Barnes P., Teacher F., “On the direct calculation of a transient plane
wave reflected from a finitely conducting half space”, IEEE Transactions on
Electromagnetic Compatibility, vol. 33, no. 2, p. 90–96, May 1991.
[BAR 95] Barbot J. P., Propagation radioélectrique avec les mobiles pour les
communications personnelles à l’intérieur des bâtiments dans la bande 1-3 GHz,
PhD Thesis, Paris XI-Orsay University, France, 1995.
[BAR 00] Barrett T. W., “History of ultrawideband (UWB) radar &
communications: pioneers and innovators”, Progress in Electromagnetics
Symposium, Cambridge, MA, USA, July 2000.
[BAT 03] Bateman D., “RF aspects of ultra-wideband”, International Workshop on
Ultra Wideband Systems, Oulu, Finland, June 2003.
[BAT 04a] Batra A., Balakrishnan J., Dabak A., “Multi-band OFDM: a new
approach for UWB”, International Symposium on Circuits and Systems, vol. 5,
Vancouver, Canada, p. 365–368, May 2004.
[BAT 04b] Batra A., Balakrishnan J., Dabak A., Gharpurey R., Lin J.,
Fontaine P., Ho J. M., Lee S., Frechette M., March S., Yamaguchi H.,
Foerster J., Molisch A. F., Multi-band OFDM physical layer proposal for
IEEE 802.15 Task Group 3a, Report no. IEEE P802.15-03/268r3, IEEE P802.15
Working Group for Wireless Personal Area Networks (WPANs), March 2004.
[BAY 04] Bayram A., Attiya A. M., Safaai-Jazi A., Riad S. M.,
“Frequency-domain measurement of indoor UWB propagation”, IEEE Antennas
and Propagation Society Symposium, Monterey, CA, USA, p. 1303–1306, June
2004.
Bibliography 221
[BEC 87] Beckmann P., Spizzichino A., The Scattering of Electromagnetic Waves
from Rough Surface, Artech House, 1987.
[BEL 63] Bello P. A., “Characterization of randomly time-variant linear channels”,
IEEE Transactions on Communication Systems, vol. 11, p. 360–393, December
1963.
[BEN 66] Bendat J. S., Piersol A. G., Measurement and Analysis of Random
Data, Wiley, New York, 1966.
[BEN 78] Bennett C. L., Ross G. F., “Time-domain electromagnetics and its
application”, IEEE Proceedings, vol. 66, no. 3, p. 299–318, March 1978.
[BUE 03] Buehrer R. M., Davis W. A., Safaai-Jazi A., Sweeney D.,
“Characterization of the ultra-wideband channel”, IEEE Conference on Ultra Wide
Band Systems and Technologies, Reston, VA, USA, p. 26–31, November 2003.
[BUL 87] Bultitude R. J. C., “Measurement, characterization and modeling
of indoor 800/900 MHz radio channels for digital communications”, IEEE
Communications Magazine, vol. 25, no. 6, p. 5–12, June 1987.
[BUL 02] Bultitude R. J. C., “Estimating frequency correlation functions from
propagation measurements on fading radio channels: a critical review”, IEEE
Journal on Selected Areas in Communication, vol. 20, no. 6, p. 1133–1143, August
2002.
[BUR 83] Burnside W., Burgener K., “High frequency scattering by a thin lossless
dielectric slab”, IEEE Transactions on Antennas and Propagation, vol. 31, no. 1,
p. 104–109, January 1983.
[CAS 01] Cassioli D., Win M. Z., Molisch A. F., “A statistical model for the UWB
indoor channel”, IEEE Vehicular Technology Conference, VTC Spring, vol. 2,
Rhodes, Greece, p. 1159–1163, May 2001.
[CAS 02] Cassioli D., Win M. Z., Molisch A. F., “The ultra-wide bandwidth
indoor channel: from statistical model to simulations”, IEEE Journal on Selected
Areas in Communications, vol. 20, no. 6, p. 1247–1257, August 2002.
[CAS 03] Cassioli D., Ciccognani W., Durantini A., UWB Channel Model
Report, Report, IST Ultrawaves Project, November 2003.
[CAS 04a] Cassioli D., Durantini A., “Statistical characterization of UWB indoor
propagation channels based on extensive measurement campaigns”, International
Symposium on Wireless Personal Multimedia Communications, vol. 1, Abano
Terme, Italy, p. 236–240, September 2004.
[CAS 04b] Cassioli D., Durantini A., “A time-domain propagation model of the
UWB indoor channel in the FCC-compliant band 3.6-6 GHz based on PN-sequence
channel measurements”, IEEE Vehicular Technology Conference, VTC Spring,
vol. 1, Milan, Italy, p. 213–217, May 2004.
[CAS 04c] Cassioli D., Durantini A., Ciccognani W., “The role of path loss
on the selection of the operating bands of UWB systems”, IEEE International
Symposium on Personal, Indoor and Mobile Radio Communications, vol. 4,
Barcelona, Spain, p. 2787–2791, September 2004.
[CHA 03] Chaigneaud L., Effets du mobilier et des personnes sur la propagation
radio indoor à 60 GHz, PhD Thesis, Poitiers University, France, 2003.
[CHA 04] Chang W. J., Tarng J. H., Shen M. J., “Radio bandwidth on the
averaged power-delay profiles of UWB indoor channels”, COST 273 Workshop,
Gothenburg, Sweden, June 2004.
[CHE 96] Chen S., Jeng S., “SBR image approach for radio wave propagation in
tunnels with and without traffic”, IEEE Transactions on Vehicular Technology,
vol. 45, no. 3, p. 570–578, August 1996.
[CHE 02] Cheung D., Prettie C., A path loss comparison between the 5 GHz UNII
band (802.11a) and the 2.4 GHz ISM band (802.11b), Report, Intel Corporation,
January 2002.
[CHO 04a] Chong C. C., Kim Y., Lee S. S., UWB Channel Model for Indoor
Residential Environment, Report no. IEEE P802.15-04/452, IEEE P802.15
Working Group for Wireless Personal Area Networks (WPANs), September 2004.
[CHO 04b] Chong C. C., Kim Y., Lee S. S., “UWB indoor propagation channel
measurements and data analysis in various types of high-rise apartments”, IEEE
Vehicular Technology Conference, VTC Fall, vol. 1, Los Angeles, CA, USA,
p. 150–154, September 2004.
[CHO 05] Chong C. C., Kim Y., Lee S. S., “Statistical characterization of the UWB
propagation channel in various types of high-rise apartments”, IEEE Wireless
Communications and Networking Conference, vol. 2, New Orleans, LA, USA,
p. 944–949, March 2005.
[CLA 68] Clarke R. H., “Statistical theory of mobile-radio reception”, Bell System
Technical Journal, vol. 47, p. 957–1000, July 1968.
[COC 05] Cocheril Y., Vauzelle R., Aveneau L., Khoudeir M., “Modélisation et
prise en compte des surfaces rugueuses pour un simulateur de propagation d’ondes
à 60 GHz en milieu indoor”, Actes des XIVieme Journées Nationales MicroOndes
(JNM 2005), Nantes, France, May 2005.
[CON 03] Conrat J. M., Pajusco P., Thiriet J. Y., “AMERICC, the France
Telecom R&D wideband channel sounder from 2 to 60 GHz”, COST 273
Workshop, Barcelona, Spain, January 2003.
[COS 89] COST-207, Digital land mobile radio communications, Report, European
Commission, 1989.
[COS 04] Cosquer R., Conception d’un sondeur de canal MIMO – Caractérisation
du canal de propagation d’un point de vue directionnel et doublement directionnel,
PhD Thesis, Institut National des Sciences Appliquées de Rennes, France, 2004.
[COU 98] Courtney C., “Time-domain measurements of electromagnetic properties
of materials”, IEEE Transactions on Microwave Theory and Techniques, vol. 46,
no. 5, p. 517–522, May 1998.
[CRA 98] Cramer R. J. M., Win M. Z., Scholtz R. A., “Impulse radio
multipath characteristics and diversity reception”, IEEE International Conference
on Communications, vol. 3, Atlanta, GA, USA, p. 1650–1654, June 1998.
Bibliography 223
[CRA 02] Cramer R. J. M., Scholtz R. A., Win M. Z., “Evaluation of an

ultra-wide-band propagation channel”, IEEE Transactions on Antennas and
Propagation, vol. 50, no. 5, p. 561–570, May 2002.
[DAB 03] Dabin J. A., Nan N., Haimovich A. M., Niver E., Grebel H., “The
effect of antenna directivity on pathloss and multipath propagation in UWB
indoor wireless channels”, IEEE Conference on Ultra Wideband Systems and
Technologies, Reston, VA, USA, p. 305–309, November 2003.
[DEB 96] De Backer B., Borjeson H., Olyslager F., DeZutter D., “The study
of wave propagation through periodic structures”, IEEE Vehicular Technology
Conference, Atlanta, GA, USA, p. 165–169, April 1996.
[DEN 03a] Denis B., Keignart J., “Post-processing framework for enhanced UWB
channel modeling from band-limited measurements”, IEEE Conference on Ultra
Wide Band Systems and Technologies, Reston, VA, USA, p. 260–264, November
2003.
[DEN 03b] Denis B., Keignart J., Daniele N., “Impact of NLOS propagation
upon ranging precision in UWB systems”, Proceedings of IEEE Conference on
Ultra Wideband Systems and Technologies (UWBST 2003), Reston, VA, USA,
p. 261–264, November 2003.
[DES 84] Deschamps G., Boersma J., Lee S.-W., “Three-dimensional half-plane
diffraction: exact solution and testing of uniform theories”, IEEE Transactions on
Antennas and Propagation, vol. 32, no. 3, p. 264–271, March 1984.
[DUR 04] Durantini A., Ciccognani W., Cassioli D., “UWB propagation
measurements by PN sequence channel sounding”, IEEE International Conference
on Communications, vol. 6, Paris, France, p. 3414–3418, June 2004.
[ECC 06] ECC, ECC decision of 24 March 2006 on the harmonised conditions for
devices using Ultra-Wideband (UWB) technology in bands below 10.6 GHz,
Report no. ECC/DEC/(06)04, Electronic Communications Commitee, March
2006.
[ECM 05] ECMA, High Rate Ultra Wideband PHY and MAC Standards, Report
no. ECMA-368, ECMA International, December 2005.
[FCC 02] FCC, First report and order, revision of Part 15 of the Commission’s rules
regarding ultra-wideband transmission systems, Report no. ET Docket 98-153,
FCC, April 2002.
[FIS 04] Fisher R., Kohno R., Mc Laughlin M., Wellborn M., DS-UWB
physical layer submission to 802.15 Task Group 3a, Report no. IEEE
P802.15-04/0137r3, IEEE P802.15 Working Group for Wireless Personal Area
Networks (WPANs), July 2004.
[FOE 01a] Foerster J., Green E., Somayazulu S., Leeper D., “Ultra-wideband
technology for short- or medium-range wireless communications”, Intel Technology
Journal, Q2 2001.
[FOE 01b] Foerster J. R., “The effects of multipath interference on the

performance of UWB systems in an indoor wireless channel”, IEEE Vehicular
Technology Conference, VTC Spring, vol. 2, Rhodes, Greece, p. 1176–1180, May
2001.
[FOE 03a] Foerster J., Channel Modeling Sub-comitee Report Final, Report
no. IEEE P802.15- 02/490r1-SG3a, IEEE P802.15 Working Group for WPANs,
February 2003.
[FOE 03b] Foerster J., Li Q., UWB channel modeling contribution from Intel,
Report no. IEEE 802.15-02/287, IEEE P802.15 Working Group for Wireless
Personal Area Networks (WPANs), July 2003.
[FOL 84] Foley J., Van Dam A., Hughes J., Computers Graphics: Principles and
Practices, Addison-Wesley Publishing Companies, 1984.
[FOW 90] Fowler C., Entzminger J., Corum J. O., “Assessment of ultra
wideband (UWB) technology”, IEEE Aerospace and Electronic Systems Magazine,
vol. 5, no. 11, p. 45–49, November 1990.
[GAR 87] Gardiol F., Traité d’Electricité, d’Electronique et d’Electrotechnique:
Electromagnétisme, Dunod, 1987.
[GAU 03] Gaur S., Annamalai A., “Improving the range of ultrawideband
transmission using RAKE receivers”, IEEE Vehicular Technology Conference,
VTC Fall, vol. 1, Orlando, FL, USA, p. 597–601, October 2003.
[GHA 02a] Ghassemzadeh S. S., Jana R., Rice C. W., Turin W., Tarokh V.,
“A statistical path loss model for in-home UWB channels”, IEEE Conference on
Ultra Wideband Systems and Technologies, Baltimore, MD, USA, p. 59–64, May
2002.
[GHA 02b] Ghassemzadeh S. S., Tarok V., The ultra-wideband indoor multipath
loss model, Report no. IEEE 802.15-02/283, IEEE P802.15 Working Group for
Wireless Personal Area Networks (WPANs), July 2002.
[GHA 03a] Ghassemzadeh S. S., Greenstein L. J., Kavcic A., Sveinsson T.,
Tarokh V., “UWB indoor delay profile model for residential and commercial
environments”, IEEE Vehicular Technology Conference,VTC Fall, vol. 5, Orlando,
FL, USA, p. 3120–3125, October 2003.
[GHA 03b] Ghassemzadeh S. S., Greenstein L. J., Kavcic A., Sveinsson T.,
Tarokh V., “UWB indoor path loss model for residential and commercial
buildings”, IEEE Vehicular Technology Conference,VTC Fall, vol. 5, Orlando, FL,
USA, p. 3115–3119, October 2003.
[GHA 03c] Ghassemzadeh S. S., Tarokh V., “UWB path loss characterization in
residential environments”, IEEE Radio Frequency Integrated Circuits Symposium,
Philadelphia, PA, USA, p. 501–504, June 2003.
[GHA 04a] Ghassemzadeh S. S., Jana R., Rice C. W., Turin W., Tarokh V.,
“Measurement and modeling of an ultra-wide bandwidth indoor channel”, IEEE
Transactions on Communications, vol. 52, no. 10, p. 1786–1796, October 2004.
Bibliography 225
[GHA 04b] Ghavami M., Michael L., Kohno R., Ultra Wideband: Signals and
Systems in Communication Engineering, Wiley, 2004.
[GLO 99] Gloria B., Elaboration d’un outil de planification et d’optimisation pour
le DCS 1800 en milieux confinés, PhD Thesis, Insa de Rennes, November 1999.
[GUI 96] Guillet V., Caractérisation de la propagation multitrajets en vue
de la prédiction de la qualité de transmission GSM/DCS 1800 en petites
et micro-cellules urbaines, PhD Thesis, Ecole Nationale Supérieure des
Télécommunications de Paris, France, 1996.
[GUI 99] Guillouard S., Conception et réalisation d’un sondeur de canal très large
bande à 60 GHz - Mesures de propagation à l’intérieur des bâtiments, PhD Thesis,
Institut National des Sciences Appliquées de Rennes, France, 1999.
[GUN 00] Gunderson S. J., Scholtz R., Chugg K., Weaver R., Lee J. Y.,
Corrada C., Homier E., Wilson R., Shipboard environment characterization,
Appendix A: Naval total asset visibility tests on the USS Curtiss, Report,
UltRaLab, 2000.
[HAN 03] Haneda K., Takada J., “An application of SAGE algorithm for UWB
propagation channel estimation”, IEEE Conference on Ultra Wide Band Systems
and Technologies, Reston, VA, USA, p. 483–487, November 2003.
[HAN 04] Haneda K., Takada J. I., Kobayashi T., “Clusterization analyses of
spatio-temporal UWB radio channels in line-of-sight and non-line-of-sight indoor
home environments”, Joint COST 273/284 Workshop, Gothenburg, Sweden, June
2004.
[HAN 05] Haneda K., Takada J., Kobayashi T., “On the cluster properties in
UWB spatio-temporal residential measurement”, COST 273 Workshop, Bologna,
Italy, January 2005.
[HAR 78] Harris F. J., “On the use of windows for harmonic analysis with the
discrete fourier transform”, Proceedings of the IEEE, vol. 66, no. 1, p. 51–83,
January 1978.
[HAS 93] Hashemi H., “The indoor radio propagation channel”, Proceedings of the
IEEE, vol. 81, no. 7, p. 943–968, July 1993.
[HAS 94a] Hashemi H., “A study of temporal and spatial variations of the
indoor radio propagation channel”, IEEE International Symposium on Personal,
Indoor and Mobile Radio Communications, vol. 1, The Hague, The Netherlands,
p. 127–134, September 1994.
[HAS 94b] Hashemi H., McGuire M., Vlasschaert T., Tholl D., “Measurements
and modeling of temporal variations of the indoor radio propagation channel”,
IEEE Transactions on Vehicular Technology, vol. 43, no. 3, p. 733–737, August
1994.
[HAY 01] Haykin S., Communication Systems, Wiley, New York, 2001.
[HÖG 74] Högbom J. A., “Aperture synthesis with a non-regular distribution of
interferometer baselines”, Astronomy and Astrophysics Supplement Series, vol. 15,
p. 417–426, 1974.
[HOF 03] Hoff H., Eggers P. C. F., Kovacs I. Z., “Directional indoor ultra
wideband propagation mechanisms”, IEEE Vehicular Technology Conference,
VTC Fall, vol. 1, Orlando, FL, USA, p. 188–192, October 2003.
[HOV 02] Hovinen V., Hämäläinen M., Pätsi T., “Ultra wideband indoor radio
channel models: preliminary results”, IEEE Conference on Ultra Wide Band
Systems and Technologies, pp. 13-17, Baltimore, MD, USA, p. 75–79, May 2002.
[HOV 03] Hovinen V., Hämäläinen M., Tesi R., Hentilä L., Laine N., A
proposal for a selection of indoor UWB path loss model, Report no. IEEE
802.15-02/280, IEEE 802.15 Working Group for Wireless Personal Area Networks
(WPANs), July 2003.
[HUA 96] Huang Y., Parsons D., “A time domain approach for measuring the
dielectric properties and thickness of walls of a building”, Proceedings of IEE
Colloquium on Propagation Aspects of Future Mobile Systems, London, UK,
October 1996.
[HUM 03] Humbert C., Simulation du canal de propagation indoor, PhD Thesis,
Marne-la-Vallée University, December 2003.
[IBA 00] Ibazizen A., Extension de la méthode FDTD aux applications
électromagnétiques de grande taille, PhD Thesis, Pierre and Marie Curie
University, France, 2000.
[IKE 91] Ikegami F., Takeuchi T., Yoshida S., “Theoretical prediction of mean
field strength for urban mobile radio”, IEEE Transactions on Antennas and
Propagation, vol. 39, no. 3, p. 299–302, March 1991.
[ITU 04] ITU, Propagation prediction methods for the planning of ultra-wideband
applications in the frequency range 1 GHz to 10 GHz (draft recommendation),
Report no. 3K/TEMP/18-E, International Telecommunications Union, October
2004.
[JAK 93] Jakes W. C., Microwave Mobile Communications, IEEE Press, New York,
1993.
[JAM 04] Jamsa T., Hovinen V., Hentilä L., “Comparison of wideband and
ultra-wideband channel measurements”, COST 273, Gothenburg, Sweden, June
2004.
[KAR 04a] Karedal J., Wyne S., Almers P., Tufvesson F., Molisch A. F.,
“Statistical analysis of the uwb channel in an industrial environment”, IEEE
Vehicular Technology Conference, VTC Fall, vol. 1, Los Angeles, CA, USA,
p. 81–85, September 2004.
[KAR 04b] Karedal J., Wyne S., Almers P., Tufvesson F., Molisch A. F.,
“UWB channel measurements in an industrial environment”, IEEE Global
Telecommunications Conference, vol. 6, Dallas, TX, USA, p. 3511–3516, November
2004.
[KAT 00] Kattenbach R., Weitzel D., “Wideband channel sounder for
time-variant indoor radio channels”, Conference on Antennas and Propagation,
Davos, Switzerland, April 2000.
Bibliography 227
[KEI 02] Keignart J., Daniele N., “Subnanosecond UWB channel sounding in
frequency and temporal domain”, IEEE Conference on Ultra Wide Band Systems
and Technologies, Baltimore, MD, USA, p. 25–30, May 2002.
[KEI 03] Keignart J., Daniele N., “Channel sounding and modelling for indoor
UWB communications”, International Workshop on Ultra Wideband Systems,
Oulu, Finland, June 2003.
[KEI 04] Keignart J., Daniele N., UWB Channel measurements in Snow-Covered
Environment, Report no. IEEE 802.15-04/449, IEEE 802.15 Working Group for
Wireless Personal Area Networks (WPANs), August 2004.
[KEL 62] Keller J., “Geometrical theory of diffraction”, Journal Optical Society of
America, vol. 52, no. 2, p. 116–130, February 1962.
[KIM 97] Kimpe M., LeiB H., “Computerized indoor radio channel estimation using
ray tracing”, Annales des télécommunications, vol. 52, no. 5-6, p. 251–263, 1997.
[KOL 33] Kolmogorov A. N., “Sulla determinazione empirica di una legge di
distribuzione”, Giornale dell’ Istituto Italiano degli Attuari, vol. 4, no. 1, p. 83–91,
January 1933.
[KON 99] Kondylis G., De Flaviis F., Pottie G., Rahmat-Samii Y., “Indoor
channel characterization for wireless communications using reduced finite
difference time domain”, IEEE Vehicular Technology Conference, VTC Fall,
Amsterdam, The Netherlands, p. 1402–1406, September 1999.
[KOU 74] Kouyoumjian R., Pathak P., “A uniform geometrical theory of
diffraction for an edge in a perfectly conducting surface”, IEEE Transactions on
Antennas and Propagation, vol. 62, no. 7, p. 1448–1496, November 1974.
[KOV 03] Kovács I., Eggers P. C. F., “Short-range UWB radio propagation using
small terminal antennas”, International Workshop on Ultra Wideband Systems,
Oulu, Finland, June 2003.
[KRA 93] Kralj D., Carin L., “Ultra-wideband characterization of lossy materials:
short-pulse microwave measurements”, Proceedings of IEEE Microwave Theory an
Techniques Symposium, Atlanta, GA, USA, June 1993.
[KUN 02a] Kunisch J., Pamp J., “Measurement results and modelling aspects for
the UWB radio channel”, IEEE Conference on Ultra Wide Band Systems and
Technologies, Baltimore, MD, USA, p. 19–23, May 2002.
[KUN 02b] Kunisch J., Pamp J., Radio channel model for indoor UWB WPAN
environments, Report no. IEEE 802.15-02/281, IEEE 802.15 Working Group for
Wireless Personal Area Networks (WPANs), June 2002.
[KUN 03] Kunisch J., Pamp J., “An ultra-wideband space-variant multipath indoor
radio channel model”, IEEE Conference on Ultra Wide Band Systems and
Technologies, Reston, VA, USA, p. 290–294, November 2003.
[LAU 94] Laurenson D. I., Indoor radio channel propagation modelling by ray
tracing techniques, PhD Thesis, Edinburgh University, UK, 1994.
[LAU 95] Lauer A., Woff I., Bahr A., Pamp J., Kunish J., “Multi-mode FDTD
simulations of indoor propagation including antenna properties”, Proceeding of the
1995 IEEE 45th Vehicular Technology Conference, Chicago, IL, USA, p. 454–458,
1995.
[LEE 85] Lee W. C. Y., “Estimate of local average power of a mobile radio signal”,
IEEE Transactions on Vehicular Technology, vol. 34, no. 1, p. 22–27, February
1985.
[LEE 02] Lee J. Y., Scholtz R. A., “Ranging in a dense multipath environment
using an UWB radio link”, IEEE Journal on Selected Areas in Communications,
vol. 20, no. 9, p. 1677–1683, December 2002.
[LEG 95] Legendre J.-F., Etude de modèle de prédiction de la propagation basés
sur la théorie géométrique de la diffraction, PhD Thesis, INSA de Rennes, France,
January 1995.
[LEG 05] Legendre J.-F., Gesny O., Bayetto J.-C., Roger L., “New
propagation model using fast 3D ray tracing. Application to WIFI”, European
Conference on Propagation and Systems (ECPS 2005), Brest, France, March 2005.
[LI 03] Li Q., Wong W. S., “Measurement and analysis of the indoor UWB channel”,
IEEE Vehicular Technology Conference, VTC Fall, vol. 1, Orlando, FL, USA,
p. 1–5, October 2003.
[LIC 03] Licul S., Davis W. A., Stutzman W. L., “Ultra-wideband (UWB)
communication link modeling and characterization”, IEEE Conference on Ultra
Wideband Systems and Technologies, Reston, VA, USA, p. 310–314, November
2003.
[LIN 89] Ling H., Chou R., Lee S., “Shooting and Bouncing Rays: Calculating
Shaped Cavity”, IEEE Transactions on Antennas and Propagation, vol. 37,
p. 194–205, January 1989.
[LO 88] Lo Y., Lee S., Antenna Handbook, Theory, Applications and Design, Van
Nostrand Reinhold Company., New York, 1988.
[LUE 89] Luebbers R., “A Heuristic UTD Slope diffraction coefficient for rough
lossy wedges”, IEEE Transactions on Antennas and Propagation, vol. 37, no. 2,
p. 206–211, November 1989.
[MAR 60] Marcum J., “A statistical theory of target detection by pulsed radar”,
IEEE Transactions on Information Theory, vol. 6, no. 2, p. 59–267, April 1960.
[MAX 55] Maxwell J. C., “On Faraday’s lines of forces”, Cambridge Philosophical
Society Conference, Cambridge, UK, December 1855.
[MCK 91] McKown J., Hamilton R., “Ray Tracing as a Design Tool for Radio
Networks”, IEEE Network Magazine, vol. 5, no. 11, p. 27–30, November 1991.
[MCK 03] McKinstry D. R., Buehrer R. M., “UWB small scale channel modeling
and system performance”, IEEE Vehicular Technology Conference, VTC Fall,
vol. 1, Orlando, FL, USA, p. 6–10, October 2003.
Bibliography 229
[MCN 90] McNamara D., Pistorius C., Malherbe J., Introduction to the
Uniform Geometrical Theory of Diffraction, Artech House, 1990.
[MOL 03] Molisch A. F., Foerster J. R., Pendergrass M., “Channel models for
ultrawideband personal area networks”, IEEE Wireless Communications, vol. 10,
no. 6, p. 14–21, December 2003.
[MOL 04] Molisch A. F., Balakrishnan K., Chong C. C., Emami S., Fort
A., Karedal J., Kunisch J., Schantz H., Schuster U., Siwiak K., IEEE
802.15.4a channel model–final report, Report no. IEEE P802.15-04/0662, IEEE
802.15 Working Group for Wireless Personal Area Networks (WPANs), November
2004.
[MOR 03] Morche D., Helal D., “RF front-ends implementation aspects and
regulation status”, The Ultra Wide Band Summit, Paris, France, December 2003.
[MUQ 03a] Muqaibel A. H. andSafaai-Jazi A., “A new formulation for
characterization of materials based on measured insertion transfer unction”, IEEE
Transactions on Microwave Theory and Techniques, vol. 51, no. 8, p. 1946–1952,
August 2003.
[MUQ 03b] Muqaibel A. H., Safaai-Jazi A., Attiya A. M., Bayram A., Riad
S. M., “Measurement and characterization of indoor ultra-wideband propagation”,
IEEE Conference on Ultra Wideband Systems and Technologies, Reston, VA, USA,
p. 295–299, November 2003.
[OPS 01] Opshaug G. R., Enge P., “GPS and UWB for indoor navigation”, Institute
of Navigation’s GPS Conference, Salt Lake City, UT, USA, September 2001.
[OSS 03] Ossberger G., Buchegger T., Stelzer A., Diskus C., Weigel R.,
“A Comparison of CPW-to-CPS fed ultra wideband antennas”, Proceedings of
International Workshop on UWB Systems (IWUWBS 2003), Oulu, Finland, June
2003.
[PAE 02] Paetzold M., Mobile Fading Channels, Wiley, Chichester, 2002.
[PAG 03] Pagani P., Pajusco P., Voinot S., “A study of the ultra-wide band
indoor channel: propagation experiment and measurement results”, International
Workshop on Ultra Wideband Systems, Oulu, Finland, June 2003.
[PAG 04] Pagani P., Pajusco P., “Experimental assessment of the UWB channel
variability in a dynamic environment”, International Symposium on Personal,
Indoor and Mobile Radio Communications, vol. 4, Barcelona, Spain, p. 2973–2977,
September 2004.
[PAG 05] Pagani P., Caractérisation et modélisation du canal de propagation radio
en contexte Ultra Large Bande, PhD Thesis, Institut National des Sciences
Appliquées de Rennes, France, 2005.
[PAG 06a] Pagani P., Pajusco P., “Characterization and modeling of temporal
variations on a ultra-wideband radio link”, IEEE Transactions on Antennas and
Propagation, vol. 54, no. 11-1, p. 3198–3206, November 2006.
[PAG 06b] Pagani P., Pajusco P., “Experimental analysis of the ultra wideband
propagation channel over the 3.1 GHz–10.6 GHz frequency band”, IEEE
International Symposium on Personal, Indoor and Mobile Radio Communications,
Helsinki, Finland, September 2006.
[PAG 06c] Pagani P., Pajusco P., “Modeling the space- and time-variant
ultra-wideband propagation channel”, IEEE International Conference on Ultra
Wide Band, Waltham, MA, USA, September 2006.
[PAG 06d] Pagani P., Pajusco P., “Statistical modeling of the ultra wide
band propagation channel through the analysis of experimental measurements”,
Comptes Rendus de l’Académie des Sciences, Physique, vol. 7, p. 762–773,
September 2006.
[PAJ 03] Pajusco P., Pagani P., “Extension of SIMO wideband channel sounder
for UWB propagation experiment”, IEEE Conference on Ultra Wideband Systems
and Technologies, Reston, VA, USA, p. 250–254, November 2003.
[PAJ 07] Pajusco P., Pagani P., “Frequency dependence of the UWB indoor
propagation channel”, European Conference on Antennas and Propagation,
Edinburgh, UK, p. 1–7, November 2007.
[PAR 00] Parsons J. D., The Mobile Radio Propagation Channel, Wiley, Chichester,
2000.
[PEN 02] Pendergrass M., Beeler W. C., Empirically based statistical
Ultra-Wideband (UWB) channel model, Report no. IEEE 802.15-02/240, IEEE
802.15 Working Group for Wireless Personal Area Networks (WPANs), June 2002.
[PET 93] Petit R., Ondes electromagnétiques en radioélectricité et en optique,
Masson, 1993.
[PEZ 03] Pezzin M., Keignart J., Daniele N., De Rivaz S., Denis B., Morche
D., Rouzet P., Catenoz R., Rinaldi N., “Ultra Wideband: the radio link of
the future?”, Annales des Télécommunications, vol. 58, no. 3-4, p. 464–506, March
2003.
[PLO 00] Plouhinec E., Etude et Extension de modèles de Prédiction de la
Propagation: Elaboration d’un Serveur Expert Multi-Modèles, PhD Thesis, INSA
de Rennes, France, December 2000.
[PLO 03] Plouhinec E., Uguen B., Tchoffo Talom F., “A parametric study
of GO transmission applied to an ultra wide band signal”, Proceedings
of International Conference on Electromagnetics in Advanced Applications
(ICEAA’03), Turin, Italy, p. 57–60, September 2003.
[POR 03a] Porcino D., “UWB regulations and coexistence: from the FCC first
report and order to the path to approval in Europe”, International Workshop
on Ultra Wideband Systems, Oulu, Finland, June 2003.
[POR 03b] Porcino D., Hirt W., “Ultra-wideband radio technology: potential and
challenges ahead”, IEEE Communications Magazine, vol. 41, no. 7, p. 66–74, July
2003.
[PRO 83] Proakis J. G., Digital Communications, McGraw-Hill, New York, 1983.
Bibliography 231
[QIU 02] Qiu R., “A study of the ultra-wideband wireless propagation channel
and optimum UWB receiver design”, IEEE Journal on Selected Areas in
Communications, vol. 20, no. 9, p. 1628–1637, December 2002.
[RIZ 97] Rizk K., Wagen J. F., Gardiol F., “Two-dimensional ray-tracing
modelling for propagation prediction in microcellular environnments”, IEEE
Transactions on Vehicular Technology, vol. 46, no. 2, p. 508–518, May 1997.
[ROU 95] Rousseau P., Pathak P., “Time-domain uniform geometrical theory of
diffraction for a curved wedge”, IEEE Transactions on Antennas and Propagation,
vol. 43, no. 12, p. 1375–1382, December 1995.
[ROU 96] Rouvière J. F., Douchin N., Combes P., “Improvement of the UTD
formulation for diffraction of an electromagnetic wave by a dielectric wedge”,
Electronics Letters, vol. 33, no. 5, February 1996.
[ROU 99] Rouvière J. F., Douchin N., Combes P., “Diffraction by lossy dielectric
wedge using both heuristic UTD formulations and FDTD”, IEEE Transactions on
Antennas and Propagation, vol. 47, no. 11, p. 1702–1708, November 1999.
[SAC 02] Sachs J., Thomä R., Schultheiss U., Zetik R., Dvoracek J., Wolf
M., “Real-time ultra-wideband channel sounder”, General Assembly of the Union
of Radio Science, Maastricht, The Netherlands, August 2002.
[SAG 03] Sagnard F., Sabouroux P., Tchoffo Talom F., El Zein G.,
“Reflection and transmission of pulses at oblique incidence from stratified
dispersive media”, 17ème Colloque International Optique Hertzienne et
Diélectriques, Calais, France, September 2003.
[SAL 87] Saleh A. A. M., Valenzuela R. A., “A statistical model for indoor
multipath propagation”, IEEE Journal on Selected Areas in Communications,
vol. 5, no. 2, p. 128–137, February 1987.
[SAN 96] Sanchez M., DeHaro L., Pino A., “Exhaustive ray tracing algorithm for
microcellular propagation prediction models”, Electronics Letters, vol. 32, no. 7,
p. 624–625, March 1996.
[SAR 03] Sarkar T., Ji Z., Kim K., Medouri A., Salazar-Palma M., “A Survey
of Various Propagation Models for Mobile Communication”, IEEE Antennas and
Propagation Magazine, vol. 45, no. 3, p. 51–82, June 2003.
[SAT 95] Sato K., Kozima H., Masuzawa H., Manabe T., Ihara T., Kasashima
Y., Yamaki K., “Measurements of reflection characteristics and refractive indices
of interrior construction materials in millimeter-wave bands”, Proceedings of 45th
IEEE Vehicular Technology Conference, New York, NY, USA, p. 449–453, 1995.
[SCH 86] Schrank H., “A centenial of electromagnetics”, IEEE Antennas and
Propagation Society Newsletter, vol. 28, no. 6, p. 35–38, December 1986.
[SCH 93] Scholtz R. A., “Multiple access with time-hopping impulse modulation”,
IEEE Military Communications Conference, vol. 2, Bedford, MA, USA,
p. 447–450, October 1993.
[SCH 97a] Scholtz R. A., Win M. Z., “Impulse radio”, Glisic S., Leppänen P. A.,
Eds., Wireless Communications: TDMA vs. CDMA, Kluwer, Norwell, 1997.
[SCH 97b] Schuster J., Luebbers R., “Comparison of GTD and FDTD predictions
for UHF radio wave propagation in a simple outdoor urban environment”, IEEE
Antennas and Propagation Society International Symposium, Montreal, Canada,
p. 2022–2025, July 1997.
[SCH 03a] Schantz H. G., “Introduction to ultra wideband antennas”, IEEE
Conference on Ultra Wide Band Systems and Technologies, Reston, VA, USA,
p. 1–9, November 2003.
[SCH 03b] Schantz H., “A brief history of UWB antennas”, Proceedings of IEEE
Conference on Ultra Wideband Systems and Technologies (UWBST 2003), Reston,
VA, USA, November 2003.
[SCH 03c] Schantz H., “Introduction to ultra-wideband antennas”, Proceedings of
IEEE Conference on Ultra Wideband Systems and Technologies (UWBST 2003),
Reston, VA, USA, November 2003.
[SCH 04] Schuster U., Indoor UWB Channel Measurements from 2 GHz to 8
GHz, Report no. IEEE 802.15-04/447, IEEE 802.15 Working Group for Wireless
Personal Area Networks (WPANs), September 2004.
[SHA 49] Shannon C. E., “Communication in the presence of noise”, Proceedings of
the Institute of Radio Engineers, vol. 37, no. 1, p. 10–21, January 1949.
[SIB 04] Sibille A., Roblin C., Bories S., Lepage A. C., Begaud X., “Conception
et caractérisation d’antennes ULB pour communications multimédia haut débit”,
Revue de l’Electricité et de l’Electronique, vol. 4, p. 73–80, April 2004.
[SIW 02] Siwiak K., UWB propagation phenomena, Report no. IEEE
802.15-02/301, IEEE 802.15 Working Group for Wireless Personal Area
Networks (WPANs), July 2002.
[SPE 97] Spencer Q., Rice M., Jeffs B., Jensen M., “A statistical model for
angle of arrival in indoor multipath propagation”, IEEE Vehicular Technology
Conference, VTC Spring, vol. 3, Phoenix, AZ, USA, p. 1415–1419, May 1997.
[STA 98] Stanislawiak M., Elaboration d’un modèle 3D de prédiction basé sur la
technique de lancer de rayons, PhD Thesis, Lille University, France, 1998.
[TAK 01] Takeuchi T., Tamura M., “A ultra-wide band channel sounder for mobile
communication systems”, IEEE International Symposium on Personal, Indoor
and Mobile Radio Communications, vol. 2, San Diego, CA, USA, p. 111–115,
September 2001.
[TAM 95] Tam W., Tran V., “Propagation modeling for indoor wireless
communication”, Journal of Electronics & Communication Engineering,
p. 221–228, 1995.
[TAN 95] Tan S., Tan H., “Improved three-dimension ray tracing technique for
microcellular propagation”, Electronics Letters, vol. 31, no. 17, p. 1503–1505,
August 1995.
[TAY 95] Taylor J. D., Introduction to Ultra-Wideband Radar Systems, CRC Press,
Boca Raton, 1995.
Bibliography 233
[TCH 03] Tchoffo Talom F., Uguen B., Chassay G., “Une méthode
d’accélération de la détermination de rayons pour la caractérisation de canaux
radio”, Actes des XIIIième Journées Nationales MicroOndes (JNM 2003), Lille,
France, May 2003.
[TCH 04] Tchoffo Talom F., Uguen B., Plouhinec E., Chassay G., “A site
specific ultra wideband channel modeling”, IEEE International Workshop on Ultra
Wide Band Systems (IWUWBS2004) Joint with Conference on Ultra Wide Band
Systems and Technologies (UWBST2004), Kyoto, Japan, May 2004.
[TCH 05a] Tchoffo Talom F., Modélisation déterministe du canal de propagation

indoor dans un contexte UWB, PhD Thesis, Institut National des Sciences
Appliquées de Rennes, France, 2005.
[TCH 05b] Tchoffo Talom F., Doré J.-B., Uguen B., Plouhinec E.,
Chassay G., “UWB in-home modelling”, Proceedings of European Conference on
Propagation and Systems (ECPS 2005), Brest, France, March 2005.
[TCH 06] Tchoffo Talom F., Uguen B., Aubert L.-M., “Antennas link synthesis
using near field chamber measurements”, Proceedings of European Conference on
Antenna and Propagation (EUCAP), Nice, France, November 2006.
[TER 03] Terré M., Hong A., Guibé G., Legrand F., “Major characteristics
of UWB indoor transmission for simulation”, IEEE Vehicular Technology
Conference, VTC Spring, vol. 1, Jeju, Korea, p. 19–23, April 2003.
[TSU 04] Tsuchiya H., Haneda K., Takada J., “Double-directional channel
sounding with ultra wideband signal in an indoor office environment”, COST 273
Workshop, Duisburg, Germany, September 2004.
[UGU 02] Uguen B., Plouhinec E., Lostanlen Y., Chassay G., “A deterministic
ultra wideband channel modeling”, Proceedings of IEEE Conference on Ultra Wide
Band Systems and Technology (UWBST 2002), Baltimore, MD, USA, p. 1–5, May
2002.
[UGU 04] Uguen B., Tchoffo Talom F., Chassay G., “UWB radio link
modeling for multipath environment”, European Conference on Electromagnetics,
Magdeburg, Germany, July 2004.
[UGU 05] Uguen B., Tchoffo Talom F., “Implementation of UWB antenna
characteristics in 3D ray tracing”, Proceedings of International Conference on
Electromagnetics in Advanced Applications (ICEAA’05), Turin, Italy, September
2005.
[VEN 05] Venkatesh S., Bharadwaj V., Buehrer R. M., “A new spatial model for
impulse-based ultra-wideband channels”, IEEE Vehicular Technology Conference,
VTC Fall, vol. 4, Dallas, TX, USA, September 2005.
[VER 90] Veruttipong T., “Time domain version of the uniform GTD”, IEEE
Transactions on Antennas and Propagation, vol. 38, no. 11, p. 1757–1764,
November 1990.
[VIL 99] Villanese F., Scanlon W., Evans N., Gambi E., “Hybrid
image/ray-shooting UHF radio propagation predictor for populated indoor
environments”, Electronics Letters, vol. 35, no. 21, p. 1804–1805, October 1999.
[WEL 03] Welborn M., Shvodian W., “Ultra-wideband technology for wireless
personal area networks – the IEEE 802.15.3/3a standards”, IEEE Conference on
Ultra Wideband Systems and Technologies, Reston, VA, USA, November 2003.
[WIE 03] Wiesbeck W., Younis M., Baldauf M., Advanced radio communication,
Report, Institut für Höchstfrequenztechnik und Elektronik, Karlsruhe University,
Germany, 2003.
[WIN 97a] Win M. Z., Ramirez-Mireles F., Scholtz R. A., Barnes
M. A., “Ultra-wide bandwidth (UWB) signal propagation for outdoor wireless
communications”, IEEE Vehicular Technology Conference, vol. 1, Phoenix, AZ,
USA, p. 251–255, May 1997.
[WIN 97b] Win M. Z., Scholtz R. A., Barnes M. A., “Ultra-wide bandwidth
signal propagation for indoor wireless communications”, IEEE International
Conference on Communications, Montreal, Canada, p. 56–60, June 1997.
[WIT 99] Withington P., Reinhardt R., Stanley R., “Preliminary results of
an ultra-wideband (impulse) scanning receiver”, IEEE Military Communications
Conference, vol. 2, Atlantic City, NJ, USA, p. 1186–1190, October 1999.
[YAN 98] Yang C., Wu B., Ko C., “A ray-tracing method for modeling indoor wave
propagation and penetration”, IEEE Transactions on Antennas and Propagation,
vol. 46, no. 6, p. 907–919, June 1998.
[YAN 99] Yang C., Wu B., “Simulations and measurements for indoor wave
propagation through periodic structures”, IEEE International Symposium on
Antennas and Propagation, Orlando, FL, USA, p. 384–387, July 1999.
[YAN 02] Yano S. M., “Investigating the ultra-wideband indoor wireless channel”,
IEEE Vehicular Technology Conference, VTC Spring, vol. 3, Birmingham, AL,
USA, p. 1200–1204, May 2002.
[YAN 04] Yang L., Giannakis G. B., “Ultra-wideband communications: an idea
whose time has come”, IEEE Signal Processing Magazine, vol. 21, no. 6, p. 26–54,
November 2004.
[YAO 97] Yao Y., “A time-domain ray model for predicting indoor wideband
radio channel propagation characteristics”, Proceeding of IEEE 6th International
Conference on Universal Personal Communications Record, vol. 2, San Diego, CA,
USA, p. 848–852, October 1997.
[YAO 03a] Yao R., Gao G., Chen Z., Zhu W., “UWB multipath channel
model based on time-domain UTD technique”, IEEE Global Telecommunications
Conference, San Francisco, CA, USA, p. 1205–1210, December 2003.
[YAO 03b] Yao R., Zhu W., Chen Z., “An efficient time-domain ray model for
UWB indoor multipath propagation channel”, IEEE Global Telecommunications
Conference, San Francisco, CA, USA, p. 1293–1297, December 2003.
Bibliography 235
[YIN 00] Ying W., Safavi-Naeini S., Chaudhuri S., “A hybrid technique based on
combining ray tracing and FDTD methods for site-specific modeling of indoor
radio wave propagation”, IEEE Transactions on Antennas and Propagation,
vol. 48, no. 5, p. 743–745, May 2000.
[ZHA 00] Zhang W., “Fast two-dimensional diffraction modeling for site-specific
propagation prediction in urban microcellular environments”, IEEE Transactions
on Vehicular Technology, vol. 49, no. 2, p. 428–436, March 2000.
Index
A transfer function, 213

Antenna US, 56
diversity, 102 variations, 47, 148, 150
effective area, 44, 134 WSS, 55
isotropic, 111 WSSUS, 57
polarization, 107 Channel capacity, 24
radiation efficiency, 107 Channel equalization, 49
return loss, 107 Channel modeling
Arrival rate statistical, 133
cluster, 61, 145 Chirp sounder, 72
ray, 146 Coherence bandwidth, 49, 59
rays, 61 Correlation bandwidth, 49
Average fade duration, 64 Correlation sounder, 75
sliding correlation, 76
B
BER, 24 D
Bluetooth, 26, 27, 29, 30 DAA, 33
BPSK, 39, 40 Delay interval, 60
C Delay window, 60
CDMA, 38, 40 Delay-Doppler spread function, 54
Channel Deterministic model, 99
characteristic parameters, 58, 134, Diffraction, 47
137 Diffusion, 47
definition, 43 DoA, 104
deterministic, 52 DoD, 104
impulse response, 181 Doppler
linear random, 54 effect, 50
matrix, 108 shift, 50, 69, 156
propagation, 25 spectrum, 64, 173
representation, 51, 177 spread, 64, 156, 173
sounding, 67 DS-UWB, 35, 39, 40, 42
Duty cycle, 23 Kolmogorov-Smirnov test, 64
E L
ESD (Energy spectral density), 110 Laplace distribution, 170, 173, 185
Exponential decay constants, 61, 141 Level crossing rate, 64
Log-normal distribution, 64, 150, 185
F LOS, 105
Fading
fast, 48, 63, 148, 153, 172 M
flat, 48 M-BOK, 39
slow, 48, 152, 172 m-sequence, 75
Far field, 45 Maxwell equation, 189, 204
FCC, 22, 25, 31, 33, 34, 40, 111 MB-OFDM, 35, 40–42
FDTD, 100 Mean delay, 59
Fourier transform, 103, 178, 213 Measurement campaigns, 85
Fraunhofer distance, 45 examples, 91
Free space propagation, 44 overview, 85
Frequency correlation function, 59 Model
Frequency domain function, 53 statistical, 157
Fresnel law, 46 Cassioli-Win-Molisch, 160
Friis formula, 45, 134 dynamic, 169
frequency domain, 161
G IEEE 802.15.3a, 158
General theory diffraction, 200 IEEE 802.15.4a, 159
Geometric optic, 189, 200 principles, 162
field expression, 191 MoM, 100
field locality, 190
GO, 101 N
GPS, 31, 32 Nakagami distribution, 64, 150, 159,
GSM, 30 161, 183
NLOS, 105
H Normal distribution, 184, 185
Hermitian symmetry, 177, 213, 215
Hilbert transform, 178 O
Huygens’ principle, 47 OOK, 39
I P
Impulse radio, 35, 38–40 PAM, 38, 39
Impulse response, 51, 103, 137, 163, Path loss, 62, 134, 162
166 exponent
time varying, 53 distance dependent, 48, 63, 136
Intersymbol interference, 50, 59 frequency dependent, 63, 134
Inversion sounding technique, 78 Power decay constants, 144
ISB, 202 Power delay profile, 57, 58, 137
Power spectral density, 25, 33, 37, 177
K PPM, 35, 38, 39
Kolmogorov-Smirnov, 186 Processing gain, 25
Index 239
Propagation measurement techniques, 70

free space, 44, 134 frequency domain, 71
model, 99 multiple-band time domain, 78
multipath, 45, 48 time domain, 73
phenomena, 45 time resolution, 68
PSD (Power spectral density), 111 Standing wave ratio, 92
Pulse sounder, 73
T
Q Time hopping, 23
QPSK, 40 Transfer function, 125
time varying, 54
R
Transmission, 47
Radio channel, 26
RAKE reception, 49 U
Ray launching, 209, 211 UHF, 30
Ray tracing, 209, 211 UMTS, 23, 30, 40
Rayleigh distribution, 64, 150, 155, Uncorrelated scattering, 56
161, 173, 181, 183, 184 Uniform theory of diffraction, 200
Rays launching, 101 UNII, 30, 34, 39
Rays tracing, 101 UTD, 101
Reflection, 46 UWB
Rice distribution, 64, 150, 155, 182, applications, 27, 29, 30
183 characteristics, 24
RIR, 105 definition, 21, 22
RMS delay spread, 59, 137 history, 23
RSB, 202 regulation, 30
RUN method, 58
V
S Variations
S-parameters, 71 spatial, 51, 150, 169
Saleh and Valenzuela model, 61 temporal, 51, 150, 172
Scattering function, 57, 64, 156, 173 Vector network analyzer, 71
Selectivity
frequency, 48, 58 W
spatial, 48 Waveguide effect, 47, 137
Shadowing, 48 Weibull distribution, 64, 184
SHF, 30 Wide sense stationary, 55
SISO, 104 WiFi, 26, 27, 29, 30, 39
Snell-Descartes law, 46 WLAN, 27
Sounding WPAN, 27
analyzed bandwidth, 68
CIR dynamic, 69 Z
CIR length, 69 Zero padding, 215
maximum Doppler shift, 69 Zigbee, 26, 29

Ultra-Wideband Radio Propagation Channels

Uploaded by

Copyright:

Available Formats

Ultra-Wideband Radio Propagation Channels

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ultra-Wideband Radio Propagation Channels

Uploaded by

Copyright:

Available Formats

Ultra-Wideband Radio

ISTE Ltd John Wiley & Sons, Inc.

© ISTE Ltd, 2008

Library of Congress Cataloging-in-Publication Data

Communications ultra large bande. English

British Library Cataloguing-in-Publication Data

Chapter 1. UWB Technology and its Applications . . . . . . . . . . . . . 21

Chapter 2. Radio Wave Propagation . . . . . . . . . . . . . . . . . . . . . 43

2.2.3.1. Spatial selectivity . . . . . . . . . . . . . . . . . . . . . . 48

Chapter 3. UWB Propagation Channel Sounding . . . . . . . . . . . . . 67

3.4. UWB measurement campaigns . . . . . . . . . . . . . . . . . . . . 85

Chapter 4. Deterministic Modeling of the UWB Channel . . . . . . . . 99

Chapter 5. Statistical Modeling of the UWB Channel . . . . . . . . . . 133

5.1.1.2. Distance propagation loss . . . . . . . . . . . . . . . . . 136

C.1.2. Field locality principle . . . . . . . . . . . . . . . . . . . . 190

Although the origins of distant signal transmission are ancient, the

Today, the main players in the telecommunication world are facing an

UWB technology undeniably oﬀers numerous advantages and

UWB transceivers may be considerably reduced with respect to traditional

These characteristics are unique to UWB technology and enable the

These systems are particulary well suited to ad hoc communication

In order to develop such systems, many challenges are still to be

Among these scientiﬁc challenges, a complete knowledge of the radio

UWB propagation channel is intrinsically diﬀerent from traditional wideband

This textbook results from an intense collaboration between France

This book is divided into ﬁve distinct parts.

Chapter 1 presents UWB technology. Its historical evolution, the

Chapter 2 describes the propagation of electromagnetic waves in general.

Chapter 3 presents an overview of the channel sounding techniques adapted

limitation of each solution. The main UWB channel measurement campaigns

For the study of the propagation phenomena and the design of

Chapter 5 is dedicated to the statistical modeling of the UWB transmission

Finally, the interested reader will ﬁnd useful additional information

This book examines all the characteristics of the UWB transmission

conﬁguration. I am convinced that this textbook will prove essential reading

Professor Ghaı̈s El Zein

ADC Analog-to-digital converter

WiFi Wireless ﬁdelity

UWB Technology and its Applications

One of the current scientific challenges, where significant research efforts

1.2. Deﬁnition and historical evolution

Ultra-wideband -41.3 dB m.MHz -1

Figure 1.1. Comparison of various radio system spectrums

Figure 1.1 illustrates the comparison between conventional radio systems,

systems, with spreading spectrum for example, and ultra-wideband systems,

1.2.2. Historical evolution

More recently, research focused on UWB signals for radio communication

1.3. Speciﬁcities of UWB

where C is the channel capacity (bit/s), Bw is the signal bandwidth (Hz), S is

In the context of an increasing demand for communication systems with

The main properties of UWB systems are described below:

• High temporal resolution capability

a precision of about 0.1 to 1 ns, it is possible to obtain information on

• Robustness against fading related to multipath propagation

• Low power spectral density

• Less sensitivity to jamming

• Protected and secured communications

1.4. Considered applications

A comparative study of the characteristics of some wireless technologies is

0.1 Zigbee 802.15.4a (Low rate UWB)