UWB Theory and Applications

UWB Theory and Applications
UWB Theory and Applications
Edited by
Ian Oppermann, Matti Hämäläinen and Jari Iinatti
All of CWC, University of Oula, Finland
UWB Theory and Applications
UWB Theory and Applications
Edited by
Ian Oppermann, Matti Hämäläinen and Jari Iinatti
All of CWC, University of Oula, Finland
Copyright Ó 2004
John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester,
West Sussex PO 19 8SQ, England
Telephone
(þ44) 1243 779777
Email (for orders and customer service enquiries): cs-books@wiley.co.uk
Visit our Home Page on www.wileyeurope.com or www.wiley.com
All Rights Reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted
in any form or by any means, electronic, mechanical, photocopying, recording, scanning or otherwise, except
under the terms of the Copyright, Designs and Patents Act 1988 or under the terms of a licence issued by the
Copyright Licensing Agency Ltd, 90 Tottenham Court Road, London W1T 4LP, UK, without the permission
in writing of the Publisher. Requests to the Publisher should be addressed to the Permissions Department,
John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex PO19 8SQ, England, or emailed
to permreq@wiley.co.uk, or faxed to (þ44) 1243 770620.
Designations used by companies to distinguish their products are often claimed as trademarks. All brand
names and product names used in this book are trade names, service marks, trademarks or registered
trademarks of their respective owners. The Publisher is not associated with any product or vendor
mentioned in this book.
This publication is designed to provide accurate and authoritative information in regard to the subject
matter covered. It is sold on the understanding that the Publisher is not engaged in rendering professional
services. If professional advice or other expert assistance is required, the services of a
competent professional should be sought.
Other Wiley Editorial Offices
John Wiley & Sons Inc., 111 River Street, Hoboken, NJ 07030, USA
Jossey-Bass, 989 Market Street, San Francisco, CA 94103-1741, USA
Wiley-VCH Verlag GmbH, Boschstr. 12, D-69469 Weinheim, Germany
John Wiley & Sons Australia Ltd, 33 Park Road, Milton, Queensland 4064, Australia
John Wiley & Sons (Asia) Pte Ltd, 2 Clementi Loop #02-01, Jin Xing Distripark, Singapore 129809
John Wiley & Sons Canada Ltd, 22 Worcester Road, Etobicoke, Ontario, Canada M9W 1L1
Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may
not be available in electronic books.
British Library Cataloguing in Publication Data
A catalogue record for this book is available from the British Library
ISBN 0-470-86917-8
Typeset in 10/12pt Times by Integra Software Services Pvt. Ltd, Pondicherry, India
Printed and bound in Great Britain by TJ International Ltd, Padstow, Cornwall
This book is printed on acid-free paper responsibly manufactured from sustainable forestry in which at least
two trees are planted for each one used for paper production.
dedicated to
our families
Contents
Preface
Acknowledgements
Abbreviations
1
Introduction
1.1 Introduction
1.1.1 Scope of this Book
1.2 UWB Basics
1.2.1 Advantages of UWB
1.3 Regulatory Bodies
1.3.1 UWB Regulation in the USA
1.3.2 UWB Regulations in Europe
1.3.2.1 IEEE 802.15.3a
1.3.2.2 IEEE 802.15.4a
1.4 Conclusions
2
UWB Channel Models
2.1 Introduction
2.2 Channel Measurement Techniques
2.2.1 Frequency Domain Channel Sounding
2.2.1.1 Signal Analysis Using IFFT
2.2.1.2 Hermitian Signal Processing
2.2.1.3 Conjugate Approach
2.2.2 Calibration and Verification
2.2.3 Measurement Experimental Set-up
2.2.3.1 Modified Frequency Domain Sounding System
2.2.4 Time Domain Channel Sounding
2.2.4.1 Impulse Sounding
2.2.4.2 Direct Sequence Spread Spectrum Sounding
xiii
xv
xvii
1
1
2
2
3
4
4
5
7
7
7
9
9
9
10
11
11
12
13
15
15
19
19
20
Contents
viii
2.3
UWB Radio Channel Models
2.3.1 Modified Saleh–Valenzuela Model
2.3.2 Other Multipath Models
2.4 Path Loss Models
2.5 Conclusions
21
23
25
33
37
3
Modulation Schemes
3.1 Introduction
3.2 Impulse Radio Schemes
3.2.1 Impulse Radio UWB
3.2.2 Fast Stepped Frequency Chirps
3.3 Multi-Carrier Schemes
3.3.1 Multi-carrier Spread Spectrum Schemes
3.3.2 Multiband UWB
3.4 Data Modulation
3.4.1 Pulse Amplitude Modulation
3.4.2 On–Off Keying
3.4.3 Pulse Position Modulation
3.4.4 Pulse Shape Modulation
3.4.5 Theoretical Bounds
3.5 Spectrum ‘Spreading’
3.5.1 TH-UWB
3.5.2 Data Modulation with Time Hopping
3.5.3 Multiple Access with TH-UWB
3.5.4 Direct Sequence UWB
3.5.4.1 Data Modulation with DS-UWB
3.5.5 Comparison of TH and DS BPAM UWB
3.6 Conclusions
39
39
40
40
44
45
45
47
47
48
48
49
51
53
54
55
57
60
61
61
62
65
4
Receiver Structures
4.1 Introduction
4.2 Rake Receiver
4.2.1 Rake Receiver Types
4.2.2 Detection Techniques
4.3 Synchronization in UWB Systems
4.3.1 Basics
4.3.1.1 Synchronization Schemes
4.3.2 Performance Measures
4.3.2.1 Performance of CLPDI
4.3.2.2 AWGN Channel Performance
4.3.2.3 Performance in Saleh–Valenzuela Channels
4.4 Conclusions
67
67
68
68
71
75
75
76
78
80
80
82
85
5
Integrated Circuit Topologies
5.1 Introduction
5.2 Ultra Wideband Basic Architectures
87
87
88
Contents
5.3
5.4
5.5
5.6
5.7
5.8
6
ix
Review of Existing UWB Technologies
5.3.1 Time Domain Corporation: PulsOn Technology
5.3.2 Time Domain Corporation: Sub-Carrier Technology
5.3.3 MultiSpectral Solutions, Inc.
5.3.4 XtremeSpectrum Inc.: Trinity
5.3.4.1 Pulse Generation by Avalanche Transistor
5.3.5 Coplanar Waveguides
Integrated Circuit Topologies
5.4.1 Source Coupled Pair
5.4.2 The Gilbert Multiplier
5.4.3 Analogue Addition/Subtraction
5.4.4 Integrator
5.4.5 Current Source
IC Processes
Example Implementation
5.6.1 Transceiver
5.6.2 Pulse Generator
5.6.3 The Analogue Correlator
5.6.4 Timing Circuit
Simulation Results
5.7.1 Transmitter
5.7.2 Receiver
Conclusions
UWB Antennas
6.1 Introduction
6.2 UWB Antenna Characteristics
6.3 Antenna Types
6.3.1 General Requirements
6.3.1.1 Base Station Antenna
6.3.1.2 Portable Antenna
6.3.2 TEM Horn
6.3.3 TEM Horn Variants
6.3.4 Impulse Radiating Antenna
6.3.5 Folded-horn Antenna
6.3.6 Dipoles and Monopoles
6.3.7 Loop Antennas
6.3.8 Antenna Arrays
6.4 Simulation Techniques
6.4.1 Finite-Element Method
6.4.1.1 Method of Moments
6.4.1.2 Finite-Difference Time-Domain
6.5 Simulation examples
6.5.1 Perfectly Conducting Dipole
6.5.2 Capacitively Loaded Dipole
6.5.3 Resistively Loaded Dipole
90
91
92
92
94
95
95
97
98
102
103
105
107
107
109
110
111
115
116
117
118
124
126
129
129
129
131
131
132
132
132
133
133
133
134
135
135
136
136
136
136
137
137
140
140
Contents
x
7
8
6.5.4 Conical Dipole
6.5.5 Log-periodic Dipole Array
6.5.6 TEM Horn
6.6 Measured examples
6.6.1 Measurement Techniques
6.6.1.1 Frequency-domain Measurements
6.6.1.2 Time-domain Measurements
6.6.2 TEM Horn
6.6.3 Small Antennas
6.7 Conclusions
141
142
143
144
144
144
147
148
150
156
Medium Access Control
7.1 Introduction
7.2 Multiple Access in UWB Systems
7.2.1 MAC Objectives
7.2.1.1 Coexistence
7.2.1.2 Interoperability
7.2.1.3 Positioning/Tracking Support
7.2.2 Structure of the UWB Signal
7.2.3 Modulation and Multiple Access
7.2.4 Multiuser System Capacity
7.3 Medium Access Control for Ultra-Wideband
7.3.1 Constraints and Implications of UWB Technologies
on MAC Design
7.3.2 Resource Allocation in UWB Systems
7.4 IEEE 802.15.3 MAC
7.4.1 Introduction
7.4.2 Applications
7.4.3 Main Features
7.4.3.1 UWB Considerations
7.5 Conclusions
157
157
158
158
158
158
159
159
160
161
163
Positioning
8.1 Introduction
8.2 Positioning Techniques
8.2.1 Time-based Positioning
8.2.2 Overview of Position Estimation Techniques
8.2.3 Direct Calculation Method
8.2.4 Optimization Based Methods
8.2.4.1 Objective Function
8.2.4.2 Gauss–Newton Method
8.2.4.3 Quasi-Newton Method
8.2.5 Simulation Results
8.3 Delay Estimation Techniques
8.3.1 General Approaches
8.3.2 Inter-path Cancellation
175
175
176
176
176
177
180
180
180
181
182
185
186
187
164
166
167
167
168
169
172
173
Contents
8.4
xi
NLOS Conditions
8.4.1 Sources of Uncertainty
8.4.2 Delay Through Walls
8.5 Metrics for Positioning
8.5.1 Identifying NLOS Channels
8.5.1.1 Use of Confidence Metrics
8.6 Conclusions
Appendices
References and Bibliography
188
188
189
190
191
194
196
197
209
Index
217
Preface
The work covered in this book has been undertaken at the Centre for Wireless Communication (CWC) at the University of Oulu, Finland. The authors have been involved
with ultra-wideband (UWB) projects for several years, which have included fundamental studies as well as design–build–test projects. A substantial number of propagation
measurements have been undertaken as well as work developing simulators, antenna
components and prototypes.
The book focuses very much on impulse radio UWB techniques rather than multiband systems. The reasons for this are both practical and historical. The promise of
UWB was low complexity, low power and low cost. Impulse radio, being a baseband
technology, holds the most promise to achieve these three benefits. The newer multiband proposals may potentially offer the most spectrally efficient solutions, but they are
substantially more complex and it is potentially more difficult to ensure compliance
with Federal Communications Commission (FCC) requirements.
The historical reason for the focus on impulse radio techniques is that CWC has been
working on UWB devices based on impulse radio techniques since 1999. Much of the
work in this book has been performed as part of projects carried out at CWC.
At the time of writing, the European regulatory bodies are still to decide on the
spectrum allocation mask for Europe. It is expected to be very similar to the FCC mask,
but with more stringent protection for bands below 3.1 GHz. Europe’s decision on
UWB will have a dramatic impact on the size and shape of the market for UWB devices
worldwide. Europe is definitely aware of the historical battles of wireless Local Area
Network systems (Hiperlan versus IEEE 802.11) and is seeking a harmonized, global
approach to standardization and regulation. The race for UWB consumer devices is
moving quickly but is definitely not over yet.
Ian Oppermann
Matti Hämäläinen
Jari Iinatti
Oulu, July 2004
Acknowledgements
The work in this book has predominantly been carried out in projects at CWC in the last
several years. The contributing projects include FUBS (future UWB systems), IGLU
(indoor geo-location solutions), ULTRAWAVES (UWB audio visual entertainment
systems) and URFA (UWB RF ASIC). More information about each of these projects
may be found on the CWC WWW site http://www.cwc.oulu.fi/home.
The UWB projects at CWC have been funded by the National Technology Agency of
Finland (TEKES), Nokia, Elektrobit, Finnish Defence Forces and European Commission. We are most grateful to the financiers for their interest in the subject.
Many researchers have also contributed to this work. The editors would like to thank
Ulrico Celentano, Lassi Hentilä, Taavi Hirvonen, Veikko Hovinen, Pekka Jakkula,
Niina Laine, Marja Kosamo, Tommi Matila, Tero Patana, Alberto Rabbachin, Simone
Soderi, Raffaello Tesi, Sakari Tiuraniemi and Kegen Yu for their contributions.
The authors also offer a special thanks to Mrs. Therese Oppermann for many hours
of proof reading and Ms. Sari Luukkonen for taking care of the proofread corrections.
Abbreviations
A
Apeak
ao
a
ðtÞ
an
B
Be
Bf
ðkÞ
cj
C
Cgd
Cgs
Cint
CL
Cox
Cn
c(t), C(t)
Ci
Cu
Dopt
D
d
do
dj
Ew
f
f
fc
f0
fH
fi
fL
fmax
fn
fPRF
Gain
Peak amplitude
first signal component
power scaling constant
amplitude gain for nth multipath component
Bandwidth
Bandwidth expansion factor
Fractional bandwidth
Pseudo random time-hopping code
Capacitance, capacitor
Gate to drain capacitance
Gate to source capacitance
Integrating capacitor
Load capacitance
Gate capacitance
Node capacitance
Spreading code
Chip (bit of the spreading code)
number of cells in uncertainty region
Optimum number of rake branches
number of rake fingers
distance
reference distance
data bit
Energy
Frequency
Frequency shift
Centre frequency
Nominal centre frequency
Upper frequency
Carrier frequency
Lower frequency
Maximum frequency
Node frequency
Pulse repetition frequency
Abbreviations
xviii
fT
gi
gm
Gm
hRX
hTX
h(t)
idj
iout
isj
I
Ibias
ID1, ID2 . . .
I0
Iout
ISS
j
k
k
k
K
L
L
Lr
Lc
M
M1, M2,..
n
N
NP10dB
NP (85%)
Nsmp
NF
Nfft
N
NU
Pd
Pfa
Pov
d
Pov
m
PG4
PG2
PG
PTX,av
PTX,fr
PL
Transit frequency
Chip (bit of the spreading code)
Transconductance, small signal
Transconductance, large signal
receiver antenna height
transmitter antenna height
Impulse response
Small signal drain current
Small signal output current
Small signal source current
Current
Bias current
Drain currents of transistor M1, M2 . . .
The zeroth order modified Bessel function of the first kind
Output current
Current of tail current source
Number of current monocycle
User
Transconductance parameter
k-factor of Ricean faded signal
SNR value for Ricean fading channel
Number of multipath
Length of a MOS transistor, Number of multipath
Number of rake branches
number of clusters
number of simultaneous users
MOS transistor 1, 2, . . .
Number of bits, body effect factor, attenuation factor
Noise power
Number of paths within 10dB of the peak
Number of paths capturing 85% of the energy
Number of sample points
Noise figure
Length of IFFT
Number of pulses per data symbol
Number of users
probability of detection
probability of false alarm
overall probability of detection
overall probability of missing a code
Processing gain from the pulse repetition
Processing gain due to the low duty cycle
Processing gain
Average single pulse power
Average power over time hopping frame
Attenuation factor
Abbreviations
r
rn
rs1
r(t)
R
RS
s(k)
s (t)
Str
S
SRX
STX
t
tcoh
t(k)
ttr
tsw
DT
Tc
Tf
Tli
Tp
TPRF
Ts
Ti
Tacq
Tfa
Th
TMA
Ts
Tw
y(t)
V(t)
V(f)
Vin
Viþ
Vi
V
Vbias
Veff
VGS
Vi
Vint
Vout
VRF
VT
Distance
Node resistance
Source resistance of M1
Transmitted signal
Resistance, data rate
Symbol rate
Laplace operator, non-centrality parameter of Ricean distribution function
information signal
Received waveform
Signal power
Received power spectral density
Transmitted power spectral density
Time
coherence time
Clock for user k
Time of flight
Sweeping time
Time delay used in PPM, modulation index
Time hopping interval inside a frame (thus, chip length)
Time hopping frame
Delays of the lth cluster
Pulse width
Pulse repetition interval, length of a time frame
Symbol time
time to evaluate a decision variable
acquisition time
penalty time
threshold
mean acquisition time
time limit
time delay between the pulses in doublet
Received signal
pulse waveform
pulse spectrum
Small signal input voltage
Positive single-ended, small signal input voltage
Negative single-ended, small signal input voltage
Voltage
Bias voltage
Effective gate to source voltage
Gate to source voltage
Input voltage
Voltage across integrating capacitor
Output voltage
RF input voltage
Threshold voltage of a MOS transistor
xix
xx
Abbreviations
Upper input voltage for Gilbert cell
Lower input voltage for Gilbert cell
ith derivative of the Gaussian pulse
Transmitted waveform
Width of a MOS transistor
Shadowing effect
Exponential decay coefficient
Multipath gain coefficients
Exponential decay constant
Early/late data modulation
optimal modulation index
modulation index
Ray decay factor
Cluster decay factor
Ray arrival rate
Cluster arrival rate
The mean energy of the first path of the first cluster
0
Standard deviation
standard deviation for cluster lognormal fading
1
standard deviation for ray lognormal fading
2
Variance of the acquisition time
Tacq
standard deviation for lognormal shadowing
x
i
Delays for the kth multipath component
k;l
Maximum detectable delay
max
ðtÞ
excess delay
n
e
code phase
Electron mobility near silicon surface
n
Unity-gain frequency of integrator
!ti
Fading term
l
signal-to-noise ratio
a.c.
Alternating Current
AC
Absolute Combining
ACF
Autocorrelation Function
ACK
Acknowledgement
ADSL
Asymmetric Digital Subscriber Line
ALT PHY Alternative Physical
AOA
Angle of arrival
A-rake
all rake receiver
ARQ
Automatic Repeat reQuest
AWGN
Additive White Gaussian Noise
BER
Bit Error Rate
BiCMOS
Bipolar Complementary Metal-Oxide-Semiconductor process
BJT
Bipolar Junction Transistor
BPAM
Binary Pulse Amplitude Modulation
CAP
Contention Access Period
CLPDI
chip level post detection integration algorithm
VX
VY
wgi
wtr
W
Xi
ik;l
ðkÞ
dbj=Ns c
opt
Abbreviations
CTA
CTAP
CTS
CDMA
CEPT
CIR
CMOS
CFAR
CMF
CSMA/CA
DAB
DARPA
DC
DCOP
DEV
DM
DME
DS
DSO
DSSS
DUT
DVB
EGC
ETSI
FCC
FAA
FCC
FD
FEM
FDTD
FET
FFT
FIR
GaAs
GaN
GaP
Gm-C
GPR
GPS
GSM
HBT
HDR
HEMT
IC
IEEE
IF
Channel Time Allocation
Channel Time Allocation Period
clear to send
Code Division Multiple Access
European Conference of Postal and Telecommunications
Channel Impulse Response
Complementary Metal-Oxide-Semiconductor
constant false alarm rate
Code Matched Filter
Carrier sense multiple access with collision avoidance
Digital Audio Broadcasting
Defense Advanced Research Projects Agency, USA
Direct Current
Direct Current Operation Point
Device
Deterministic Model
Device Management Entity
Direct Sequence
Digital Sampling Oscilloscope
Direct Sequence Spread Spectrum
Device Under Test
Digital Video Broadcasting
Equal GainCombining
European Telecommunications Standards Institute
US Federal Communications Commission
Federal Aviation Administration, USA
Federal Communications Commission, USA
Frequency Domain
Finite element method
Finite difference time domain
Field Effect Transistor
Fast Fourier Transform
Finite Impulse Response,
Gallium Arsenide
Gallium Nitride
Gallium Phosphide
Transconductor-Capacitor
Ground Penetrating Radar
Global Positioning System
Global System for Mobile Communications
Hetero-junction Bipolar Transistor
high data rate
High Electron Mobility Transistor
Integrated Circuit
The Institute of Electrical and Electronics Engineers
Intermediate Frequency
xxi
Abbreviations
xxii
IFFT
I/H
InP
InSb
IR
IRA
I-Rake
ISI
ISM
ISO
ITU
LDR
LLC
LLNL
LNA
LO
LOS
LPD
LPDA
LPI
MAC
MBT
MC
MCTA
MESFET
MF
ML
MLME
MMIC
MoM
MRC
MT
NLOS
NMOS
NOI
OOK
OSI
PA
PAM
PCB
PE
PER
PG
PHY
PLL
PMOS
Inverse fast Fourier transform
Integrate and Hold
Indium Phosphide
Indium Antimonide
Impulse Radio
Impulse radiating antenna
Ideal rake receiver
Inter Symbol Interference
Industrial, Scientific and Medical
International Standards Organization
International Telecommunications Union
Low data rate
Logical Link Control
Lawrence Livermore National Laboratory
Low Noise Amplifier
Local Oscillator (frequency)
Line-of-Sight
Low Probability of Detection
Log-periodig dipole array
Low Probability of Interception
Medium Access Control
Modified Bowtie
Multi-Carrier
Management Channel Time Allocations
Metal Semiconductor Field Effect Transistor
Matched Filter
Maximum Likelihood
MAC Layer Management Entity
Microwave/Millimetre-wave Integrated Circuit
Methods of moments
Maximum Ratio Combining
Multi-tone
Non-Line-of-Sight
N-channel Metal-Oxide-Semiconductor FET
Notice of Inquiry
On-off Keying
Open System Interconnection
Power Amplifier
Pulse Amplitude Modulation
Printed circuit board
Power Estimation
Packet Error Rate
Processing Gain
Physical layer
Phase Locked Loop
P-channel Metal-Oxide-Semiconductor
Abbreviations
PN
PNC
PPM
PR
P-rake
PRI
PRF
PSD
PSM
PTD
PVT
PPM
QoS
RC
RF
RFIC
RMS
RTS
RX
S1, S2
SC
S/H
SIFS
SiGe
SINR
SIR
SM
SNR
S-rake
SRD
SS
SV
TD
TDMA
TDOA
TH
TH-PPM
TM
TOA
TR
TSA
TX
UMTS
UMB
UWBWG
VCO
Pseudo-random Noise
PicoNet Coordinator
Pulse Position Modulation
Pseudo Random
Partial rake receiver
Pulse Repetition Interval
Pulse Repetition Frequency
Power Spectral Density
Pulse Shape Modulation
Programmable Time Delay
Process, power supply Voltage and Temperature
Pulse Position Modulation
Quality of Service
Resistor-Capacitor (circuit)
Radio Frequency
Radio Frequency Integrated Circuit
Root Mean Square
Request to send
Receiver, Receiver port
Short Pulses 1, 2
Switched Capacitor
Sample and Hold (circuit)
Short Inter-frame spacing
Silicon Germanium semiconductor process
Signal-to-Interference-plus-Noise Ratio
Signal-to-Interference Ratio
Statistical Model
Signal-to-Noise Ratio
Selective rake receiver
Step Recovery Diode
Spread Spectrum
Saleh-Valenzuela Channel Model
Time Domain
Time division multiple access
Time difference of arrival
Time-Hopping
Time-Hopping Pulse Position Modulation
Time Modulated
Time of Arrival
Transistor
Tapered slot antenna
Transmitter, Transmitter port
Universal Mobile Telecommunications System
Ultra-WideBand
Ultra-Wideband Working Group
Voltage Controlled Oscillator
xxiii
Abbreviations
xxiv
VHF
VNA
VSWR
WLAN
WO
WP
WPAN
WS
WZ
Very high frequency
Vector Network Analyser
Voltage standing wave ratio
Wireless Local area network
Worst case One
Worst case Power
Wireless Personal Area Networks
Worst case Speed
Worst case Zero
1
Introduction
Ian Oppermann, Matti Hämäläinen, Jari Iinatti
1.1 Introduction
The world of ultra-wideband (UWB) has changed dramatically in very recent history. In
the past 20 years, UWB was used for radar, sensing, military communications and niche
applications. A substantial change occurred in February 2002, when the FCC (2002a,b)
issued a ruling that UWB could be used for data communications as well as for radar
and safety applications. This book will focus almost exclusively on the communications
aspects of UWB.
The band allocated to communications is a staggering 7.5 GHz, by far the largest
allocation of bandwidth to any commercial terrestrial system. This allocation came hot
on the heels of the hotly contested, and very expensive, auctions for third generation
spectrum in 2000, which raised more than $100 billion for European governments. The
FCC UWB rulings allocated 1500-times the spectrum allocation of a single UMTS
(universal mobile telecommunication system) licence, and, worse, the band is free to use.
It was no wonder, therefore, that efforts to bring UWB into the mainstream were
greeted with great hostility. First, the enormous bandwidth of the system meant that
UWB could potentially offer data rates of the order of Gbps. Second, the bandwidth sat
on top of many existing allocations causing concern from those groups with the primary
allocations. When the FCC proposed the UWB rulings, they received almost 1000
submissions opposing the proposed UWB rulings.
Fortunately, the FCC UWB rulings went ahead. The concession was, however, that
available power levels would be very low. If the entire 7.5 GHz band is optimally
utilized, the maximum power available to a transmitter is approximately 0.5 mW. This
is a tiny fraction of what is available to users of the 2.45 GHz ISM (Industrial, Scientific
and Medical) bands such as the IEEE 802.11 a/b/g standards (the Institute of Electrical
and Electronics Engineers). This effectively relegates UWB to indoor, short-range,
UWB Theory and Applications Edited by I. Oppermann, M. Hämäläinen and J. Iinatti
Ó 2004 John Wiley & Sons, Ltd ISBN: 0-470-86917-8
UWB Theory and Applications
2
communications for high data rates, or very low data rates for substantial link distances. Applications such as wireless UWB and personal area networks have been
proposed, with hundreds of Mbps to several Gbps and distances of 1 to 10 metres.
For ranges of 20 metres or more, the achievable data rates are very low compared with
existing wireless local area network (WLAN) systems.
One of the enormous potentials of UWB, however, is the ability to move between the
very high data rate, short link distance and the very low data rate, longer link distance
applications. The trade-off is facilitated by the physical layer signal structure. The very
low transmit power available invariably means multiple, low energy, UWB pulses must be
combined to carry 1 bit of information. In principle, trading data rate for link distance
can be as simple as increasing the number of pulses used to carry 1 bit. The more pulses
per bit, the lower the data rate, and the greater the achievable transmission distance.
1.1.1
Scope of this Book
This book explores the fundamentals of UWB technology with particular emphasis on
impulse radio (IR) techniques. The goals of the early parts of the book are to provide
the essential aspects of knowledge of UWB technology, especially in communications and in control applications. A literature survey examining books, articles
and conference papers presents the basic features of UWB technology and current
systems. A patent database search provides a historical perspective on the state-of-art
technology.
1.2 UWB Basics
Other terms associated with ‘ultra-wideband’ include ‘impulse’, ‘short-pulse’, ‘nonsinusoidal’, ‘carrierless’, ‘time domain’, ‘super wideband’, ‘fast frequency chirp’ and
‘mono-pulse’ (Taylor, 1995).
Impulse radio communication systems and impulse radars both utilize very short
pulses in transmission that results in an ultra-wideband spectrum. For radio applications, this communication method is also classified as a pulse modulation technique
because the data modulation is introduced by pulse position modulation (PPM). The
UWB signal is noiselike which makes interception and detection quite difficult. Due to
the low-power spectral density, UWB signals cause very little interference with existing
narrow-band radio systems. Depending on the attitude of national and international
regulatory bodies, this should allow licence-free operation of radio systems.
Time-modulated (TM) impulse radio signal is seen as a carrier-less baseband transmission. The absence of carrier frequency is the very fundamental character that
differentiates impulse radio and impulse radar transmissions from narrow-band applications and from direct sequence (DS) spread spectrum (SS) multi-carrier (MC) transmissions, which can also be characterised as an (ultra) wideband technique. Fast slewing
chirps and exponentially damped sine waves are also possible methods of generating
UWB signals.
At the end of the book there is an extensive bibliography of UWB technology in general,
and particularly impulse radio and impulse radar systems. Impulse radars, sensors, etc.,
are touched on in this book, but the main focus is in the communication sector.
Introduction
1.2.1
3
Advantages of UWB
UWB has a number of advantages that make it attractive for consumer communications
applications. In particular, UWB systems
.
.
.
.
have potentially low complexity and low cost;
have noise-like signal;
are resistant to severe multipath and jamming;
have very good time domain resolution allowing for location and tracking applications.
The low complexity and low cost of UWB systems arises from the essentially baseband
nature of the signal transmission. Unlike conventional radio systems, the UWB transmitter produces a very short time domain pulse, which is able to propagate without the
need for an additional RF (radio frequency) mixing stage. The RF mixing stage takes a
baseband signal and ‘injects’ a carrier frequency or translates the signal to a frequency
which has desirable propagation characteristics. The very wideband nature of the UWB
signal means it spans frequencies commonly used as carrier frequencies. The signal will
propagate well without the need for additional up-conversion and amplification. The
reverse process of downconversion is also not required in the UWB receiver. Again, this
means the omission of a local oscillator in the receiver, and the removal of associated
complex delay and phase tracking loops.
Consequently, TM-UWB systems can be implemented in low cost, low power,
integrated circuit processes (Time Domain Corporation, 1998). TM-UWB technique
also offers grating lobe mitigation in sparse antenna array systems without weakening
of the angular resolution of the array (Anderson et al., 1991). Grating lobes are a
significant problem in conventional narrowband antenna arrays.
Due to the low energy density and the pseudo-random (PR) characteristics of the
transmitted signal, the UWB signal is noiselike, which makes unintended detection quite
difficult. Whilst there is some debate in the literature, it appears that the low power,
noise-like, UWB transmissions do not cause significant interference to existing radio
systems. The interference phenomenon between impulse radio and existing radio systems is one of the most important topics in current UWB research.
Time-modulation systems offer possibility for high data rates for communication.
Hundreds of Mbps have been reported for communication links. It is estimated (Time
Domain Corporation, 1998; Kolenchery et al., 1997) that the number of users in an
impulse radio communication system is much larger than in conventional systems. The
estimation is claimed to be valid for both high- and low-data-rate communications.
Because of the large bandwidth of the transmitted signal, very high multipath resolution is achieved. The large bandwidth offers (and also requires) huge frequency diversity
which, together with the discontinuous transmission, makes the TM-UWB signal
resistant to severe multipath propagation and jamming/interference. TM-UWB systems
offer good LPI and LPD (low probability of interception/detection) properties which
make it suitable for secure and military applications.
The very narrow time domain pulses mean that UWB radios are potentially able to
offer timing precision much better than GPS (global positioning system) (Time Domain
Corporation, 1998) and other radio systems. Together with good material penetration
UWB Theory and Applications
4
properties, TM-UWB signals offer opportunities for short range radar applications
such as rescue and anti-crime operations, as well as in surveying and in the mining
industry. One should however understand that UWB does not provide precise targeting
and extreme penetration at the same time, but UWB waveforms present a better choice
than do conventional radio systems.
1.3 Regulatory Bodies
One of the important issues in UWB communication is the frequency allocation. Some
companies in the USA are working towards removing the restrictions from the FCC’s
regulations for applications utilising UWB technology. These companies have established an Ultra-Wideband Working Group (UWBWG) to negotiate with the FCC.
Similar discussion on frequency allocation and radio interference should also emerge
in Europe. Currently, there are no dedicated frequency bands for UWB applications in
the ETSI (European Telecommunications Standards Institute) or ITU (International
Telecommunications Union) recommendations.
1.3.1
UWB Regulation in the USA
Before the FCC’s first Report and Order (Federal Communications Commission,
2002a,b), there was significant effort by industrial parties to convince the FCC to
release UWB technology under the FCC Part 15 regulation limitations, and to allow
licence-free use of UWB products. The FCC Part 15 Rules permit the operation of
classes of radio frequency devices without the need for a licence or the need for
frequency coordination (47 C.F.R. 15.1). The FCC Part 15 Rules attempt to ensure a
low probability of unlicensed devices causing harmful interference to other users of the
radio spectrum (47 C.F.R. 15.5). Within the FCC Part 15 Rules, intentional radiators
are permitted to operate within a set of limits (47 C.F.R. 15.209) that allow signal
emissions in certain frequency bands. They are not permitted to operate in sensitive
or safety-related frequency bands, which are designated as restricted bands
(47 C.F.R.15.205). UWB devices are intentional radiators under FCC Part 15 Rules.
In 1998, the FCC issued a Notice of Inquiry (NOI) (Federal Communications
Commission, 1998). Despite the very low transmission power levels anticipated, proponents of existing systems raised many claims against the use of UWB for civilian
communications. Most of the claims related to the anticipated increase of interference
level in the restricted frequency bands (e.g. TV broadcast bands and frequency bands
reserved for radio astronomy and GPS). The Federal Aviation Administration (FAA)
expressed concerned about the interference to aeronautical safety systems. The FAA
also raised concerns about the direction finding of UWB transmitters.
The organizations that support UWB technology see large scale possibilities for new
innovative products utilizing the technology. The FCC Notice of Inquiry and comments
can be found on the Internet (Ultra-Wideband Working group, 1998, 1999, 2004).
When UWB technology was proposed for civilian applications, there were no definitions for the signal. The Defense Advanced Research Projects Agency (DARPA)
provided the first definition for UWB signal based on the fractional bandwidth Bf of
Introduction
5
the signal. The first definition provided that a signal can be classified as an UWB signal
if Bf is greater than 0.25. The fractional bandwidth can be determined as (Taylor, 1995).
Bf ¼ 2
fH fL
fH þ fL
ð1:1Þ
where fL is the lower and fH is the higher 3 dB point in a spectrum, respectively.
CURRENT UWB DEFINITION
In February 2002, the FCC issued the FCC UWB rulings that provided the first
radiation limitations for UWB, and also permitted the technology commercialization.
The final report of the FCC First Report and Order (Federal Communications Commission, 2002a,b) was publicly available during April 2002. The document introduced
four different categories for allowed UWB applications, and set the radiation masks for
them.
The prevailing definition has decreased the limit of Bf at the minimum of 0.20, defined
using the equation above. Also, according to the FCC UWB rulings the signal is
recognized as UWB if the signal bandwidth is 500 MHz or more. In the formula above,
fH and fL are the higher and lower 10 dB bandwidths, respectively. The radiation limits
by FCC are presented in Table 1.1 for indoor and outdoor data communication
applications.
1.3.2
UWB Regulations in Europe
At the time of writing, regulatory bodies in Europe are awaiting further technical input
on the impact of UWB on existing systems. The European approach is somewhat more
cautious than that of the USA, as Europe requires that a new technology must be shown
to cause little or no harm to existing technologies. The European organizations have, of
course, been heavily influenced by the FCC’s decision. Currently in Europe, the recommendations for short-range devices belong to the CEPT (European Conference of
Postal and Telecommunications) working group CEPT/ ERC/ REC 70- 03 (UltraWideband Working Group, 1999). Generally, it is expected that ETSI/CEPT will follow
Table 1.1
FCC radiation limits for indoor and outdoor communication applications
Frequency in MHz
960–1610
1610–1990
1990–3100
3100–10600
Above 10600
Indoor
Outdoor
EIRP in dBm
EIRP in dBm
75:3
53:3
51:3
41:3
51:3
75:3
63:3
61:3
41:3
61:3
UWB Theory and Applications
6
the FCC’s recommendations but will not necessarily directly adopt the FCC’s regulations.The ITU limits (ITU 2002) for indoor and outdoor applications are defined by the
formulas represented in Table 1.2.
Figure 1.1 shows the current proposal for the European spectral mask limits as well as
the FCC masks. The upper plot represents the masks for data communication applications for indoor and outdoor use. The lower plot gives the FCC radiation mask for
radar and sensing applications. In all cases the maximum average power spectral density
Table 1.2
ITU radiation limits for UWB indoor and outdoor applications
Frequency range [GHz]
Indoor mask
Outdoor mask
Figure 1.1
f < 3:1
3:1 < f < 10:6
f > 10:6
51:3 þ 87 logð f =3:1Þ
61:3 þ 87 logð f =3:1Þ
41:3
41:3
51:3 þ 87 logð10:6=f Þ
61:3 þ 87 logð10:6=f Þ
UWB radiation mask defined by FCC and the existing CEPT proposal
Introduction
7
follows the limit of FCC Part 15 regulations (Federal Communications Commission,
2004).
The working groups for UWB include ERM/TG31A covering generic UWB, and
ERM/TG31B, which covers UWB for automotive applications at higher bands.
1.3.2.1
IEEE 802.15.3a
The IEEE established the 802.15.3a study group to define a new physical layer concept
for short range, high-data-rate, applications. This ALTernative PHYsical (ALT PHY)
is intended to serve the needs of groups wishing to deploy high-data-rate applications.
With a minimum data rate of 110 Mbps at 10 m, this study group intends to develop a
standard to address such applications as video or multimedia links, or cable replacement. The study group has been the focus of significant attention recently, as the debate
over competing UWB physical layer technologies has raged. The work of the study
group also includes analysing the radio channel model proposal to be used in the UWB
system evaluation.
The purpose of the study group is to provide a higher speed PHY for the existing
approved 802.15.3 standard for applications which involve imaging and multimedia
(IEEE, 2004). The main desired characteristics of the alternative PHY are:
.
.
.
.
.
coexistence with all existing IEEE 802 physical layer standards;
target data rate in excess of 100 Mbits/s for consumer applications;
robust multipath performance;
location awareness;
use of additional unlicensed spectrum for high rate WPANs (wireless personal area
network).
1.3.2.2
IEEE 802.15.4a
The IEEE established the 802.15.4a study group to define a new physical layer concept
for low data rate applications utilizing UWB technology at the air interface. The study
group addresses new applications that require only moderate data throughput, but long
battery life such as low-rate wireless personal area networks, sensors and small networks.
1.4 Conclusions
The fact that UWB technology has been around for so many years and has been used
for a wide variety of applications is strong evidence of the viability and flexibility of the
technology. The simple transmit and receiver structures that are possible make this a
potentially powerful technology for low-complexity, low-cost, communications. As will
be discussed in later chapters, the physical characteristics of the signal also support
location and tracking capabilities of UWB much more readily than do existing narrower
band technologies.
8
UWB Theory and Applications
The severe restrictions on transmit power (less than 0.5 mW maximum power) have
substantially limited the range of applications of UWB to short distance–high data rate
or low data rate–longer distance applications. The great potential of UWB is to allow
flexible transition between these two extremes without the need for substantial modifications to the transceiver.
Whilst UWB is still the subject of significant debate, there is no doubt that the
technology is capable of achieving very high data rates and is a viable alternative to
existing technology for WPAN; short-range, high-data-rate communications; multimedia applications, and cable replacement. Much of the current debate centres around
which PHY layer(s) to adopt, development of a standard, and issues of coexistence and
interference.
2
UWB Channel Models
Matti Hämäläinen, Veikkò Hovinen, Lassi Hentilä
2.1 Introduction
In many respects, the UWB channel is very similar to a wideband channel as may be
experienced in spread spectrum or CDMA systems. The main distinguishing feature of
an ‘ultra’ wideband channel model is the extremely multipath-rich channel profile.
With a bandwidth of several GHz, the corresponding time resolution of the channel
is of the order of fractions of a nanosecond. When translated to the spatial
domain, this means that it is possible to distinguish reflecting surfaces separated by
mere centimetres. Many everyday objects now can be seen to act as distinct specular
reflectors rather than contributing to ‘lumped’ multipath components. The significantly
greater time resolution is accompanied by substantially lower power per multipath
component.
This chapter examines common UWB channel models, provides methods to measure
UWB channels, and introduces the channel model adopted by the IEEE 802.15.3a study
group, which will be used as a reference model in UWB system performance studies.
2.2 Channel Measurement Techniques
There are two possible domains for performing the channel sounding to measure the
UWB radio channel. First, the channel can be measured in the frequency domain (FD)
using a frequency sweeping technique. With FD sounders, a wide frequency band is
swept using a set of narrow-band signals, and the channel frequency response is
recorded using a vector network analyser (VNA). This corresponds to S21-parameter
measurement set-up, where the device under test (DUT) is a radio channel.
Second, the channel can be measured in the time domain (TD) using channel sounders
that are based on impulse transmission or direct sequence spread spectrum signalling.
UWB Theory and Applications Edited by I. Oppermann, M. Hämäläinen and J. Iinatti
Ó 2004 John Wiley & Sons, Ltd ISBN: 0-470-86917-8
UWB Theory and Applications
10
With impulse based TD sounders, a narrow pulse is sent to the channel and the channel
impulse response is measured using a digital sampling oscilloscope (DSO).
The corresponding train of impulses can also be generated using a conventional direct
sequence spread spectrum (DSSS) based measurement system with a correlation
receiver. The performance of the DSSS sounder is based on the properties of the autocorrelation function (ACF) of the spreading code used as an overlay signal. The drawback of using the DSSS technique is that it needs very high chip rates to achieve
bandwidths required for UWB.
In this chapter, the frequency and time domain measurement concepts are presented.
Theoretically, both techniques give the same result if there is a static measurement
environment and an unlimited bandwidth.
2.2.1
Frequency Domain Channel Sounding
With frequency domain sounders, the RF signal is generated and received using a vector
network analyser (VNA) which makes the measurement set-up quite simple. The sounding
signal is a set of narrow-band sinusoids that are swept across the band of interest. The
frequency domain approach makes it possible to use wideband antennas, instead of
special impulse radiating antennas. As will be discussed in a later chapter, UWB
antennas have restrictions, for example, with ringing leading to pulse shape distortion.
The UWB channel models can then be generated at the data post-processing stage.
When the FD sounder approach is used, the channel state during the soundings must
be static to maintain the channel conditions during the sweep. The maximum sweep
time is limited by the channel coherence time. If the sweep time is longer than the
channel coherence time, the channel may change during the sweep. For fast changing
channels, other sounding techniques are needed.
The performance of the frequency domain sounder is also limited by the maximum
channel delay. The upper bound for the detectable delay max can be defined by the
number of frequency points used per sweep and the bandwidth B (frequency span to be
swept). This is given by
max ¼ ðNsmp 1Þ=B
ð2:1Þ
where Nsmp is the number of frequency points.
Another possible source of error in the measurement process is the frequency shift
caused by the propagation delay when long cables are used, or when the flight time of
the sounding signal is long. In frequency-sweep mode, the sounding signal is rapidly
swept across the whole band of interest. For a transmitter and receiver that are in lockstep sweeping across the frequency band of interest, very long propagation delays can
cause the receiver to take samples at a frequency that is higher than the received
frequency. This frequency shift f is a function of the propagation time ttr (time of
flight), the frequency span B and the sweep time tsw as
f ¼ ttr
B
:
tsw
ð2:2Þ
UWB Channel Models
11
Frequency
f2
f1
time
Network analyzer
Amplitude
IFFT
time
Figure 2.1 Vector network analyser-based frequency domain channel sounding system
In general, f has to be smaller than the analyser IF bandwidth to obtain reliable
results. The idea used in FD measurements is presented in Figure 2.1. After the channel
frequency response has been measured, the time domain representation (impulse
response) can be achieved by inverse Fourier transform (IFFT).
2.2.1.1
Signal Analysis Using IFFT
The signal measured using a VNA is a frequency response of the channel. The inverse
Fourier transform is used to transform the measured frequency domain data to the time
domain. The IFFT is usually taken directly from the measured raw data vector. This
processing is possible since the receiver has a down-conversion stage with a mixer
device. This method is referred to as the complex baseband IFFT, and is sufficient for
modeling narrow- and wideband systems.
There are two common techniques for converting the signal to the time domain,
which both lead to approximately the same results. The first approach is based on
Hermitean signal processing, which results in a better pulse shape. The second approach
is the conjugate approach. Tests show that the conjugate approach is an easier and more
efficient way of obtaining approximately the same pulse shape accuracy. These two
approaches are introduced next.
2.2.1.2
Hermitian Signal Processing
Using Hermitian processing, the pass-band signal is obtained with zero padding from
the lowest frequency down to DC (direct current), taking the conjugate of the signal,
and reflecting it to the negative frequencies. The result is then transformed to the
time domain using IFFT. This Hermitian method is shown in Figure 2.2. The signal
UWB Theory and Applications
12
Zero-padding
Complex
conjugate
Conjugate transformation
IFFT
Figure 2.2
Zero padding, conjugate reflection and resulting impulse response
spectrum is now symmetric around DC. The resulting doubled-sided spectrum corresponds to a real signal. The time resolution of the received signal is more than twice that
achieved using the baseband approach. This improvement in accuracy is important,
since one purpose in UWB channel modelling is to separate accurately the different
signals paths.
2.2.1.3
Conjugate Approach
The conjugate method involves taking the conjugate reflection of the passband signal
without zero padding. Using only the left side of the spectrum, the signal is converted
using the IFFT with the same window size as the Hermitian method. The technique is
presented in Figure 2.3. The conjugate result is very likely to be the same as the
Zero-padding
Complex
conjugate
Conjugate transformation
IFFT
Figure 2.3 Conjugate reflection with zero padding and resulting impulse response
UWB Channel Models
Figure 2.4
13
Impulse responses of the different IFFT methods
Hermitian result with zero padding. However, the conjugate method is more efficient in
terms of data processing complexity, since the matrix calculations in the post-processing
stage become easier to manipulate due to the smaller memory requirements.
The impulse responses of these two methods and the baseband method are shown in
Figure 2.4, where Hamming window is used in data processing.
From Figure 2.4 it can be seen that for both methods, the main reflection positions
are the same and the amplitudes are very close. Thus, the approach based on the leftside conjugate produces an adequate pulse shape with lower processing complexity.
2.2.2
Calibration and Verification
The vector network analyser system, like all measurement systems, requires calibration
with the same cables, adapters and other components that will be used for the measurements before the soundings. An enhanced response calibration is required to be able to
determine both the magnitude and phase of the transmitted signal (Balanis, 1997).
Amplifiers must be excluded from calibration because they are isolated in the reverse
direction. The amplifiers’ frequency response can be measured independently and their
effects can be taken into account in the data post-processing. Long cables and the
adapters connected to the ports of the analyser cause a frequency dependent variation in
UWB Theory and Applications
14
the sounding signal. The deviation is directly proportional to the quality of the used
equipments. This variation can, however, be compensated in the calibration procedure.
The calibration process moves the time reference points from the analyser ports to the
calibration points at the ends of the cables. When the time references are at the cable
ends (at the antenna connectors), the resulting delay profiles only includes the propagation delays that result from the radio channel. Due to their dimensions, delays due to the
antennas themselves are insignificant.
The performance of the measurement system was verified in a corridor where major
delays could be readily calculated from the room geometry (Hovinen et al., 2002). The
physical dimensions of the reference corridor are presented in Figure 2.5. The corresponding impulse response that is calculated from the recorded frequency response is
presented in Figure 2.6. Reflections coming from points END I, END II and END III
of the corridor (marked on Figure 2.5), as well as the main reflections from the walls, the
36.5 m
5.6 m
1.5 m
END III
END II
10.5 m
END I
RX
18.0 m
TX
8.0 m
END II and END III are glass doors with metal grids
Figure 2.5 Dimensions for the calibration test measurement made in a corridor (figure is
not to scale)
Figure 2.6
Impulse response measured at the corridor presented in Figure 2.5
UWB Channel Models
15
floor and from the ceiling, can be found with simple calculations. END II and END III
are glass doors with a metal grid inside. Comparing the estimated propagation delays
plotted at the top of Figure 2.6 with the measured impulse response, a clear conformity
between the results can be observed.
2.2.3
Measurement Experimental Set-up
The following section introduces the radio channel measurement system, which consists
of a vector network analyser, a wideband amplifier, a pair of antennas and a control
computer with LabVIEW TM controlling software. This particular installation has been
used in the measurements carried out by CWC.
The network analyser is operating in a response measurement mode, where PORT1 is
a transmitter port (TX) and PORT2 is a receiver port (RX). An external amplifier is
connected to PORT1 to increase the transmitted power level. The antennas used in the
measurement system are CMA-118/A conical antennas developed by Antenna Research
Associates, Inc. Typical features of conical antennas are an approximately omnidirectional radiation pattern with a constant phase centre. Both of these features are
important in radio channel sounding. The constant phase centre means that the exact
radiation point is independent of the frequency. The omnidirectional radiation pattern
makes it possible to receive all of the reflections generated by the channel. A low noise
amplifier at the input of the VNA can be used to improve the receiver’s noise figure, and
to increase the energy of the received channel probing signal.
The sweep time is automatically adjusted by the analyser, depending on the bandwidth and on the number of measured frequency points within the sweeping band.
Figure 2.7 presents the block diagram of the VNA based measurement set-up. The
operating measurement system is presented in Figure 2.8. Figure 2.9 gives examples of
the environments where the measurements have been carried out. The left-hand figure
introduces the assembly hall at the main building at the University of Oulu, and the
right-hand one represents a typical classroom with furniture.
The operation of the measurement set-up, and the applicability of IFFT, was verified
by recording the channel frequency response using short cables. The verifications also
show that the measurements using short cables (about 1 m) and long cables (maximum
30 m) give the same result for the channel impulse response which proves that the
calibration procedure is correct. UWB measurements with high delay resolution generally demonstrate similar structure to wideband measurements for indoor environments. In Figure 2.10, the measured frequency response and the corresponding impulse
response are presented. The measurement was performed with one obstructing wall
between transmitter and receiver.
2.2.3.1
Modified Frequency Domain Sounding System
The VNA-based measurement system introduced in Figure 2.7 is limited in range by the
length of the antenna cables. Due to the fact that the same device works both as a
transmitter and receiver, the distance between the antennas is limited to short or
UWB Theory and Applications
16
LabView(TM)
Data
storage
Postprocessing
in Matlab
Network Analyzer
PA
LNA
TX
RX
Figure 2.7 The VNA based radio channel measurement set-up
medium range. The measurement set-up can be improved to make it possible to increase
the link distance and to make the system more flexible, by changing the topology whilst
maintaining the remaining VNA as a receiver (Hämäläinen et al. 2003).
The modified UWB radio channel sounder based on the VNA is presented in Figure 2.11.
The sinusoidal probing signal is now generated by an external signal generator.
The wired connection between the transmitting and receiving ends is avoided by sending
only the triggering signal to the transmitting part via radio. The channel frequency
response is again measured using a set of tones much like the approach presented in
Section 2.2.1. The functionality of the modified sounding system is presented in Figure
2.11. All of the measurement procedures are again controlled by LabViewTM software.
The main component in the modified channel measurement system is a vector network analyser. This device receives the probing signal and makes the calculations for the
channel S21 values in the same way as described in Section 2.2.1. Rather than coming
from the VNA itself, the probing signal is now sent by the external sweeping signal
generator, which is SMIQ in the reference system. These two devices are frequency
synchronized by using the external reference clock. The transmitting port must be
terminated with 50 impedance to avoid unwanted reflections from the unused RFport of the VNA.
A low noise amplifier (LNA) can be used at the receiver port of the VNA to improve
the noise figure of the receiver, and of course, to amplify the received probing signal
level. A power amplifier (PA) can be used at the transmitter end. Any linear power
amplifier can be used to have the required power level at the receiver input.
In the test setup, a Rohde & Schwartz AFGU function generator is used to generate
the triggering pulses that are fed to the SMIQ and VNA. If the propagation delay
UWB Channel Models
Figure 2.8
17
Frequency domain measurement system
between the devices is long, the trigger pulse timing for SMIQ and VNA must be tuned
to make the SMIQ probing signal and the VNA reference signal arrive at the VNA
detector simultaneously. However, the maximum detectable delay max cannot be
exceeded. The devices are operating in single-sweep stepping mode, sweeping the predefined frequency band using 1601 points. Again, the frequency difference needs to be
less than the VNA IF-bandwidth for reliable detection.
In the reference system settings, the sweep time is approximately 5 s, and the sweep
time depends only on the number of measured frequency points. The number of
UWB Theory and Applications
18
Figure 2.9
Different environments covered during the channel measurement campaign
(a)
(b)
Figure 2.10 Example of an ultra-wideband through wall radio channel measurement:
(a) frequency response and (b) corresponding impulse response
frequency steps and therefore the sweeping time can be decreased if the maximum delay
of the channel is small.
LabView delivers almost all of the timing information and all the commands and
adjustments for the devices. A constant clock reference is used to maintain high
frequency stability in the system. An external 5 MHz clock reference, based on the
TV stripe frequency signal, is fed to the external clock reference inputs in the VNA and
the sweeping signal generator. After this procedure, a maximum frequency stability of
1012 is guaranteed. During the frequency sweeps, the phase difference between the
generator signal and the analyser signal is not known. However in UWB radio-channel
models, the phase information does not play as significant role as it does in wideband
models because there is no carrier involved. The carrier would normally provide the
reference phase. A scalar network analyser or spectrum analyser can also be used
because phase information is not utilized in data post-processing. If the objective is to
use a continuous sweep during the measurements, the sweep time in this construction
increases significantly, to approximately 30 s.
UWB Channel Models
19
LabView(TM)
Radio Channel
Probing signal
PA
External frequency
reference
5/10 MHz
AFGU
Sweeping signal
generator
GPIB
Network Analyzer
triggering
External
trigger
External frequency
reference
5/10 MHz
LNA
50Ω
RX
External triggering pulse
Figure 2.11 Modified frequency domain radio channel measurement system
In the test system, LV217 VHF combat net radios are used to transfer the triggering
signal from AFGU to SMIQ. The output pulse of AFGU is transformed to sound wave
and sent via LV217. At the SMIQ-end, the sound wave is converted back to an electrical
signal and used as an external trigger for the SMIQ. The use of VHF (very high
frequency) radios in a triggering link is justified by their frequency band that is outside
of the band under the measurements.
The frequency band that can be measured depends on the frequency range of the
VNA and SMIQ. The current devices used in the experiment allow the measurements in
a selected band between 50 MHz and 20 GHz. The antennas may be selected by the
frequency range requirements. However, the antennas should have a constant phase
centre and they should be omnidirectional if used in the radio channel measurements in
order to minimize distortion caused to the probing signal.
The limitation of the frequency domain channel sounding is that there needs to be a
static environment during the recordings.
2.2.4
Time Domain Channel Sounding
As discussed in Section 2.2.1, the frequency domain channel sounder excludes the
measurements of the non-stationary channel. However, movement can be supported
to a certain extent if the soundings are made in time domain. The following section
introduces time domain sounding systems that can be utilized for UWB.
2.2.4.1
Impulse Sounding
One way to carry out time-domain UWB radio channel soundings is to use of very short
impulses. The receiver in this case is a digital sampling oscilloscope. The bandwidth of
UWB Theory and Applications
20
Amplitude
<ns
Amplitude
time
time
Impulse generator
Probe
antenna
Digital sampling
oscilloscope
Trigger pulse
Figure 2.12 Impulse sounder
the sounder depends on the pulse shape and the pulse width used. By changing the pulse
width, the spectral allocation can also be changed. However, the simpler the pulse
shape, the easier it is to perform the deconvolution during the post-processing, where
the channel impulse response is calculated by removing the transmitted pulse waveform
from the results.
The most exact channel model can be generated if the pulse waveforms used in the
sounding correspond to the waveforms of the application. The topology for the time
domain measurements is presented in Figure 2.12.
The impulse-based measurement system requires an additional antenna to be used for
triggering purposes. The sounding distance is still limited due to the probing antenna,
which is close to the TX antenna needing to be connected to the DSO. However, the
long cable is only needed to transfer the triggering pulse for the sounding pulse, so the
high quality (e.g., expensive) cable is not needed. In the modified version, the triggering
pulse can also be sent via radio link.
2.2.4.2
Direct Sequence Spread Spectrum Sounding
The original wideband radio channel sounding system is based on direct sequence
spread spectrum technique and a correlation receiver. Theoretically, a train of impulses
can be generated using the maximum length code (m-sequence) and calculating its
autocorrelation function at the receiver. The idea of the use of spreading code as a
channel probing signal is presented in Figure 2.13(a) where s(t), r(t), c(t) and h(t) are
transmitted signal, received signal, pseudo-random code and the channel impulse
response, respectively. m-sequence is used to spread the transmitted carrier signal
energy over the wide frequency band. The bandwidth of the sounding signal is twice
the chip rate, which is the bit rate of the m-sequence.
The well known m-sequence is the optimal pseudo-random code for channel sounding
due to the low side-lobe level of its even autocorrelation function (see Figure 2.14). If
the length of the m-sequence is N, the normalized even and periodic autocorrelation
function has a value of 1 if codes are synchronized, and has a value of 1=N in all the
other code phases. The dynamic range of the signal increases as a function of N.
UWB Channel Models
21
s(t)
r(t)
h(t)
Channel
(a)
r(t)
t
t
s(t)
t
(b)
r(t)
h(t)
c(t)
Spreading code
Correlator
Channel
r(t)
t
Spreading code
c(t)
(c)
c(t)
h(t)
Spreading code
Channel
r(t)
Correlator
r(t)
t
Spreading code
c(t)
Figure 2.13
technique
Radio channel sounding system based on direct sequence spread spectrum
The correlation operation can be performed after the channel [see Figure 2.13(c)] as
the radio channel is linear, e.g., the channel does not saturate the propagating signal or
create new frequencies.
The correlator output is sampled at the chip rate, and each value represents a certain
delay caused by the channel. The magnitude of the sample is related to the strength of
the corresponding propagation path. All of this creates a delay resolution which is
inversely proportional to the chip rate. The maximum unaliased delay using this
sounding method is NTc, where Tc is the chip length.
In principle, this channel sounding technique can also be applied in the UWB context.
Because the bandwidth of the sounding signal depends on the used chip rate, the
minimum 500-MHz bandwidth can be achieved using a chip rate of 250 MHz. If the
bandwidths are larger than 1 GHz, the chip rate needs to exceed 500 MHz. This high
chip-rate requirement limits the utilization of this DSSS in UWB channel sounding.
However, if the frequency band to be sounded is located in the lower UWB band, which
is allocated for several radar applications, the fractional bandwidth requirements can
easily be met.
2.3 UWB Radio Channel Models
Due to the very large bandwidth of the impulse radio signal, the propagation phenomena
are different in the lower band and the upper band of the signal spectrum. The size of
UWB Theory and Applications
22
Autocorrelation of m–sequence
1.2
L=5
L=8
Normalised auto–correlation function
1
0.8
0.6
0.4
0.2
0
–0.2
300
350
400
450
500
550
600
650
700
750
index
Figure 2.14
Even autocorrelation function of m-sequence
the Fresnel zone, which is a function of wavelength, will be markedly different at the
lower and higher frequency ends. The larger Fresnel zone size of the lower frequencies
will mean the lower end of the signals will become obstructed more easily, and the signal
components will attenuate more than the higher signal components. Also, the natural
and man-made interference sources are different in different parts of the spectrum.
For example, the impulse-transmission-based time-domain experimental results show
that the radio channel impulse response for a sub-nanosecond pulse is a few hundred
nanoseconds long in a typical American laboratory/office building such as that presented by (Win, 1998). The path-loss factor, according to the same measurements, was
seen to be between 2:2 and 3:3, which means that the signal power attenuates 22 to
33 dB/decade when the distance increases. According to the same experimental study,
the number of resolvable multipaths for 1 GHz UWB signal bandwidth in an office
building is between 5 and 50.
The measurement system used in the indoor measurements (Win, 1998) is also used
outdoors to find out UWB’s suitability to narrowband vegetation loss models. (Win
et al., 1997a) give a comparison between the results from narrowband vegetation loss
models (Weissoerger, 1992) and the results based on UWB channel soundings. The
results show that the narrowband loss models can be applied to the UWB case.
UWB Channel Models
2.3.1
23
Modified Saleh–Valenzuela Model
During 2002 and 2003, the IEEE 802.15.3 Working Group for Wireless Personal Area
Networks, and especially its channel modelling subcommittee decided to use the so
called modified Saleh–Valenzuela model (SV) as a reference UWB channel model
(Foerster et al., 2003; Foerster, 2003).
The real valued model is based on the empirical measurements originally carried out
in indoor environments in 1987 (Saleh and Valenzuela, 1987). Due to the clustering
phenomena observed at the measured UWB indoor channel data, the model proposed
by IEEE 802.15 is derived from Saleh and Valenzuela using a log-normal distribution
rather than an original Rayleigh distribution for the multipath gain magnitude. An
independent fading mechanism is assumed for each cluster as for each ray within the
cluster. In the SV models, both the cluster and ray arrival times are modelled independently by Poisson processes. The phase of the channel impulse response can be either 0
or . Therefore the model contains no imaginary component.
The following analytical representation was originally presented by Foerster (2003),
who introduces the model characterization in more detail, and also gives the Matlab
functions to create the channel realizations. According to Foerster, a discrete time
multipath channel impulse can be presented as
1
hi ðtÞ ¼ Xi
1
LC KLC
X
X
i
ik;l t Tli k;l
ð2:3Þ
l¼0 k¼0
i
where ik;l represents the multipath gain coefficients, Tli the delays of the lth cluster, k;l
gives the delays for the kth multipath component relative to the lth cluster arrival time
ðTli Þ. Shadowing effect is log-normal distributed and is represented by Xi and i refers to
the ith realization.
The proposed modified Saleh-Valenzuela model uses the following definitions:
Tl ¼ the arrival time of the first path of the lth cluster;
k,l ¼ the delay of the kth path within the lth cluster relative to the first path arrival
time, Tl ;
¼ cluster arrival rate;
¼ ray arrival rate, i.e. the arrival rate of path within each cluster.
The definition assumes that 0;l ¼ 0. The cluster arrival time distribution can be presented by (Foerster, 2003)
pðTl jTl1 Þ ¼ exp½ðTl Tl1 Þ;
l>0
ð2:4Þ
and the ray arrival times by
p k;l ðk1Þ;l ¼ exp k;l ðk1Þ;l ;
k > 0:
ð2:5Þ
UWB Theory and Applications
24
The channel coefficients are defined by (Foerster, 2003)
k;l ¼ pk;l l k;l ; and
ð2:6Þ
20 log 10ðl k;l Þ / Nðk;l ; 21 þ 22 Þ; or
ð2:7Þ
l k;l ¼ 10ðk;l þn1 þn2 Þ=20 ;
ð2:8Þ
where n1 / Nð0; 21 Þ and n2 / Nð0; 22 Þ are independent and correspond to the fading on
each cluster and ray, respectively, and
h
2 i
E l k;l ¼ 0 eTl = ek;l =
;
ð2:9Þ
where 0 is the mean energy of the first path of the first cluster, and pk;l is equiprobable
f1; þ1g to account for the signal polarity inversion due to the reflections. The k;l is
given by (Foerster, 2003)
k;l ¼
10 lnð
0 Þ 10Tl =G 10k;l =
ð21 þ 22 Þ lnð10Þ
:
lnð10Þ
20
ð2:10Þ
The variables in the above equations represent the fading associated with the lth cluster,
l , and the fading associated with the kth ray of the lth cluster, k;l . As noted above, the
impulse responses given by the model are real valued.
Due to the fact that the log-normal shadowing of the total multipath energy is
captured by the term Xi , the total energy contained in the terms fik;l g is normalized
to unity for each realization. The shadowing term is characterized by (Foerster, 2003)
20 log 10ðXi Þ / Nð0; 2x Þ:
ð2:11Þ
The model derives the following channel parameters as an output:
.
.
.
mean and root mean square (RMS) excess delays;
number of multipath components;
power decay profile.
In addition to cluster and ray arrival rates, and l the model inputs cluster and ray
decay factors
and g, respectively, and standard deviation terms in dB for cluster
lognormal fading, ray lognormal fading and lognormal shadowing term for total multipath realization 1 ; 2 and x , respectively.
Four different channel implementations are suggested, which are based on the average distance between transmitter and receiver, and whether a LOS component is present
or not.
UWB Channel Models
25
Table 2.1 The different SV-models and their main parameters like presented in the IEEE
802.15.3 proposal
Target channel characteristics
SV-1
SV-2
Mean excess delay (nsec) ðm Þ
RMS delay (nsec) ðrms Þ
NP10dB
NP (85%)
Model parameters
Lð1=nsecÞ
ð1=nsecÞ
G
1 ðdBÞ
2 ðdBÞ
x ðdBÞ
Model characteristics
Mean excess delay (nsec) ðm Þ
RMS delay (nsec) ðrms Þ
NP10dB
NP (85%)
Channel energy mean (dB)
Channel energy standard (dB)
5.05
5.28
10.38
8.03
24
0.0233
2.5
7.1
4.3
3.3941
3.3941
3
5.0
5
12.5
20.8
0.4
2.9
36.1
0.4
0.5
5.5
6.7
3.3941
3.3941
3
9.9
8
15.3
33.9
0.5
3.1
SV-3
14.18
14.28
35
61.54
0.0667
2.1
14.00
7.9
3.3941
3.3941
3
15.9
15
24.9
64.7
0.0
3.1
SV-4
25
0.0667
2.1
24.00
12
3.3941
3.3941
3
30.1
25
41.2
123.3
0.3
2.7
SV-1: line-of-sight (LOS) model for 0–4-m
SV-2: Non-LOS (NLOS) model for 0–4-m
SV-3: NLOS for 4–10-m model
SV-4: NLOS for 4–10-m model. This model represents an extreme NLOS multipath
channel condition.
NP10dB: Number of paths within 10dB of the peak
NP(85%): Number of paths capturing 85% of the energy
The four channel models and their parameters are listed in Table 2.1 (Foerster, 2003).
Figure 2.15 gives an example of 100 channel realizations that are based on SV-3 model.
Figure 2.16 illustrates the difference between the different SV-models in the delay
domain. The average profiles are calculated from 100 independent channel realizations,
which is the approach recommended by the IEEE 802.15. The delay resolution in the
models is 167 ps, which corresponds to a spatial resolution of 5 cm. In Figure 2.17, the
number of the distinguishable propagation paths inside a 10-dB dynamic range if
compared to the highest path is presented for each SV-model calculated from 100
channel realizations. When the link occupation increases, the number of distinguishable
paths also increases. This can easily be seen from the figure.
2.3.2
Other Multipath Models
Radio channel experiments have been carried out at the University of Oulu using the
measurement system represented in Figures 2.7 and 2.8. The measured frequency band
26
UWB Theory and Applications
Figure 2.15 Delay profiles of modified Saleh–Valenzuela channel 3, 100 channel realizations
was 2–8 GHz which was covered with 1601 frequency points. Transmitted power
measured at the input of transmitter antenna at 2 GHz was PTX ¼ þ10 dBm 1 dB
and the typical antenna gain was 0 dBi. Based on the maximum noticeable delay
definition and the used parameters, the reflections exceeding the delay m ¼ 1600=
ð6 GHzÞ ¼ 266:6 ns, which corresponds to 80 m, cannot be detected.
Figures 2.18 and 2.19 show the measured impulse responses as a snapshot given by a
VNA which has a TD-measurement option available. The delay grid is 10 ns and the
power scale grid is 10 dB. Because the complex samples can be recorded when the
original VNA measurement system is used, the propagation delay, and therefore also the
link distance, can be calculated from the results, as can be seen from the example figures.
The room sizes examined are 7430 mm 4100 mm and 5940 mm 6280 mm for the
results presented in Figures 2.18 and 2.19, respectively, and the heights are typically
around 3.5 m. In Figure 2.18, the 7.438-ns delay for the first detected path corresponds
to a distance of 2.23 m. The room where the measurements were carried out is the one
presented in Figure 2.9 (right). The 40-dB dynamic range gives the 53 ns and 23 ns delay
spreads for the given examples. However, for the communication applications, the
maximum 10-dB dynamic range limits the exploitable multipaths for Rake receiver as
is the case also in IEEE802.15.3 model definition (Foerster, 2003; Foerster et al., 2003).
The following results have been calculated for antenna heights hTX ¼ 2:2 m and
hRX ¼ 1:1 m, and vice versa (Hovinen et al., 2002). The link separations in this data
set were between 1.25 m and 8.10 m, which corresponds to initial propagation delays
from 4.2 ns to 27.0 ns. The first measurement campaign covered situations where the
UWB Channel Models
27
Average impulse responses of different SV-models
0.02
SV-1
0
–0.02
0
50
100
150
200
250
0.02
SV-2
0
–0.02
0
50
100
150
200
250
0.02
SV-3
0
–0.02
0
50
100
150
200
250
0.02
SV-4
0
–0.02
0
50
100
150
200
250
Time [ns]
Figure 2.16 Average delay profiles for SV-models 1–4
transmitting and receiving antennas were located in the same room. Both the LOS and
NLOS links were studied. The measurements where TX-antenna and RX-antenna were
in the different rooms (through-wall measurement) were also carried out.
The amplitude response data were inverse Fourier transformed into delay domain.
A Hanning window was applied before transformation to make it easier to locate the
line-of-sight component of the signal. As phase information was available, the propagation delay was able to be observed in the measured data. The data was simultaneously
transformed into delay domain using no windowing in order to avoid data manipulation. The system noise level was estimated from the absolute delay range 0 –3 ns
which covers the time of flight, and was found to be around 105 dBm, typically,
limiting the dynamic range to 40 dB. The time-of-arrival (TOA, initial delay) of the
LOS component was extracted for each of the radio links using the average delay profile
of 500 impulse responses (small-scale statistics). Initial delay was removed from the
results, and the statistical parameters were extracted from this data.
The maximum excess delay was limited to 70 ns, which corresponds to 420 samples in
delay domain. The limit was found by removing the strongest reflections and plotting
average delay profiles from various data sets. This corresponds to large-scale statistics
represented in Figure 2.20, where the reflections are plotted in dark grey tint.
UWB Theory and Applications
28
Figure 2.17 Number of significant paths within 10 dB dynamic range for different SV-models
A discrete model for the output of a time-variant fading multipath channel is given by
X
hðtÞ ¼
an ðtÞsðt n ðtÞÞ;
ð2:12Þ
n
where s(t) is the transmitted signal, an ðtÞ is the amplitude gain of the nth multipath
channel and n ðtÞ is the corresponding excess delay. In an indoor environment with no
moving scatterers and with fixed antenna positions, an ðtÞ and n ðtÞ are assumed to be
constant during the observation time. Measurements recorded in various antenna
positions give an estimate of the average static indoor channel. Movement of scatterers
or altering the length of the radio link will introduce Doppler shift, which in turn
reduces the coherence time tcoh of the channel. The channel can be measured if the
measurement time is shorter than the coherence time. This assumption is valid in static
indoor measurements.
A multipath channel is typically modelled as a linear tapped delay line (finite impulse
response, FIR, filter), with complex tap coefficients (Bello, 1963). In computer simulation, the time variance of the channel filter is realized in various ways, the simplest and
most straightforward being a FIR filter, whose coefficients are updated from previously
stored complex-valued channel data. The drawback of this approach is the limited
randomness, since the data eventually has to be reused. Statistical channel description
UWB Channel Models
29
Figure 2.18 Impulse response, hTX and hRX ¼ 2200 m
gives more freedom for simulation. In the tapped delay line model, the time variation is
realized by mixing the multiplicative white noise through a bandpass filter directly to the
tap coefficients.
Nevertheless, a realistic assumption for a static indoor channel is a Ricean fading
model. Ricean fading signals have amplitude an ðtÞ that is distributed according to
(Proakis, 1995)
2
a
a þ s2
as
pR ðaÞ ¼ 2 exp I0 2 ; a 0
2
2
ð2:13Þ
where is the standard deviation and I0 is the zeroth order modified Bessel function of
the first kind. The non-centrality parameter s is defined by
s2 ¼ kaðtÞk2
ð2:14Þ
where a is mean complex amplitude.
k-factor of a Ricean fading signal is defined as
k¼
s2
s2
¼
22 2
ð2:15Þ
UWB Theory and Applications
30
Figure 2.19 Impulse response, hTX ¼ 2200 mm and hRX ¼ 600 mm
which in logarithmic scale is
2
:
k ¼ s2dB dB
ð2:16Þ
Rayleigh fading channel is a special case of Ricean channel with k ¼ 0. It has been
shown (Talvitie, 1997) that the Ricean fading channels become effectively Rayleigh
fading when k becomes smaller than 5 dB.
A large-scale (long-term) model is constructed from data that has been collected in
various locations. Because antenna positions and room sizes vary, we need to separate
pure reflections from the random part of the model. In other words, the channel model
contains a deterministic environment-dependent ray-tracing part (deterministic model,
DM) and a statistical environment-independent Ricean fading part (statistical model,
SM) as presented in Figure 2.21.
The DM contains static reflections, delay and power estimates of which can be
modeled either with ray tracing tools or calculated from a simplified reflection model.
The power of the LOS signal is normalized to unity, and it is contained in the DM
part.
Parameters for SM are extracted from the measurement data after normalizing to
average power level and removing the most significant reflections. Figure 2.22 (a–b)
UWB Channel Models
31
Figure 2.20 Average delay profile of 32 500 independent impulse responses
Channel model
x (t )
y (t )
DM
SM
Figure 2.21 Channel model blocks
shows the statistical parameters of SM as functions of antenna separation at delays 0 ns
and 5 ns, respectively. In other words, k decreases exponentially in linear scale. Estimates for kð; dÞ can be tabulated directly, but here they have been expressed using
estimates of signal power s2 ð; dÞ and noise power ð; dÞ. Figure 2.23 (a–b) shows the
values for regression parameters as ; a and bs ; b as functions of excess delays. Again, we
can fit piecewise linear regression lines to this data and get the following expressions
as ðÞ ¼ 0s þ s
a ðÞ ¼ 0 þ ð2:17Þ
UWB Theory and Applications
32
(a)
Figure 2.22
(b)
Ricean parameters for line-of-sight path and at delay 5 ns
(a)
Figure 2.23
parameters
(b)
Coefficients (a) as and a and (b) bs and b for exponential decay of Ricean
and
bs ðÞ ¼ s0 þ s ;
b ðÞ ¼ 0 þ ð2:18Þ
where subscripts s and are associated to s2 ðd; Þ and to 2 ðd; Þ, respectively. Table 2.2
and Table 2.3 list the regression parameters for the signal power and the noise power.
The signal power and the noise power can be expressed in compact matrix forms
S ¼ aTs d þ bTs ;
N ¼ aT d þ bT
ð2:19Þ
UWB Channel Models
33
Table 2.2
Delay range
0–6 ns
6–60 ns
Parameters for exponential decay coefficient Signal power
0s
0.095
0.035
Table 2.3
Delay range
0–10 ns
10–60 ns
Delay range
s
1.20
1.98
0
0.0
0.015
0.42
0.84
Parameters for exponential decay constant Signal power
s0
0.00
0.47
0–28 ns
28–60 ns
Noise power
Delay range
s
15.2
10.6
0–28 ns
28–60 ns
Noise power
0
0.00
0.15
42.7
38.5
where T denotes transpose operation, and dimensions of a and b depends on delay, and
dimension of d depends on antenna separation. Signal-to-noise ratios can in dB be
expressed as
K ¼ S N:
ð2:20Þ
Matrices S, N and K contain the signal power (delay profile), noise power and the SNR
values for Ricean fading channel, respectively. Each of the columns describes the fading
statistics as a function of delay at the given antenna separation d. When K is below 5 dB,
a Rayleigh channel with corresponding noise power applies (Talvitie, 1997).
The channel model presented can be applied to computer simulations for performance studies of UWB systems.
The impulse responses based on the indoor measurements carried out at the University of Oulu also support the clustered ray arrival phenomenon presented in Chapter 3.1.
UWB radio channel characteristics have also been studied lately (for example, Kunisch
and Pamp, 2002; Cassioli et al., 2002; Gramer et al., 2002).
2.4 Path Loss Models
The measurement data collected in a series of multipath propagation studies can also be
used to model UWB path loss. The impact of the link distance on the received signal
energy can be determined by propagation loss calculations which define the fraction of
the transmitted power that can be received at the distance d. General propagation
physics approaches are valid in the case of UWB transmission, which means that the
longer the link distance is, the lower is the received signal energy. This connection can
easily be derived by the well known path loss.
n
d
PLðdÞ d0
ð2:21Þ
UWB Theory and Applications
34
where d is the link separation, d0 is the reference distance that is usually 1 m, and n is
environment dependent constant. In free-space, n ¼ 2.
At the University of Oulu premises, path-loss focused measurements have been
carried out using the VNA measurement system from Figure 2.8. The received signal
power as a function of distance was studied in two different sized corridors and a lecture
hall.
Figure 2.24 and Figure 2.25 present the layouts and dimensions for the corridors
where the soundings were carried out. The frequency range was 2–8 GHz, the antenna
heights were 110 mm. The heights of corridors 1 and 2 were 3890 and 2280 mm,
respectively. The height of the lecture room was 3540 mm.
The soundings were made at the lecture hall at distances between the 152 cm at
minimum and 1052 cm at maximum, see Figure 2.26.
The received signal powers were studied as a function of the distance, which indicates the
attenuation of the signal. The exponential correspondence can be noticed in an absolute
2800 mm
1475 mm
TX2 X
44410 mm
Distance from
TX (cm)
X
RX2
19540 mm
RX4 X
RX5 X
4330 mm
35410 mm
X
RX3
Re f
X -- ---- ---- -- ----- --- ---X
RX1
---------- ----------------------
2270 mm
1335 mm
TX1 X
RX1
100
RX2
150
RX3
200
RX4
250
RX5
RX10
300
350
400
450
500
550
RX11
600
RX12
RX13
650
RX14
750
800
RX6
RX7
RX8
RX9
RX15
RX16
700
850
RX17
900
RX18
950
RX19
1000
RX20
RX21
1050
1100
Figure 2.24 Layout for the path loss measurements carried out at corridor 1
UWB Channel Models
714 cm
35
RX 21
Distance from
TX (cm)
RX 20
RX 19
RX1
100
RX 18
RX2
150
RX17
RX3
200
RX4
250
RX5
300
RX 13
RX6
350
RX 12
RX7
400
RX8
450
RX9
500
RX10
550
RX11
600
RX12
650
RX13
700
RX4
RX14
750
RX3
RX15
800
RX2
RX16
850
RX17
900
RX18
950
RX19
1000
RX20
1050
RX21
1100
RX16
RX15
2500 cm
RX14
RX 11
RX9
RX8
RX7
RX6
RX5
78 cm
RX1
764 cm
703 cm
TX
1023 cm
RX 10
155 cm
Figure 2.25
Layout for the path loss measurements carried out at corridor 2
distance scale. The attenuation corresponds to the straight line in a logarithmic scale. A
linear regression line is then fitted to the measured power data points using the equation
PðdÞ ¼ n10 log10 ðdÞ þ a
ð2:22Þ
where n corresponds to the path loss factor, d is distance and a is a power scaling constant.
UWB Theory and Applications
36
X
Y
RX3
RX7
RX2
RX6
RX4
RX5
RX10
RX8
RX9
RX1
RX15
cm
RX17
RX14
862 cm
50
0
cm
RX11
RX12
50
0
RX13
RX16
RX20
90 cm
TX
90 cm
RX19
RX18
I
I
875 cm
Figure 2.26 Layout for the path loss measurements carried out at a lecture hall
There are two possible ways of fitting the regression line to the data points. The more
accurate method is to use piecewise fitting where the path loss factor is different for the
shorter distances than the longer distances. The other way is to use only one regression
line. Figure 2.27 presents the example of the curve fitting. In this example, the piecewise
fitting matches the data points almost perfectly, but the single line gives the general
slope for the line. The locations of the measured data points depend on the environment, the materials used and the link parameters.
Table 2.4 collates the final results based on the various path loss measurements. In
addition to the LOS measurements, a link through the wall was measured using
different antenna heights. The measurements showed that the UWB signal attenuation
is smaller than in freespace when the propagation media includes the line-of-sight
component.
Figure 2.28 presents the measurement results and the fitted linear regression lines
for the NLOS data. The soundings were done between two rooms that were
UWB Channel Models
37
Figure 2.27 Linear regression lines for the path loss measurements carried out at corridor 1
separated by the corridor. The link distances were from 3 m to 13 m. The transmitting antenna height was set to 220 cm. The measurements were carried out using
three different receiver antenna heights: 60, 110 and 220 cm. When the signal was
obstructed by walls, the signal was also significantly attenuated leading to reduced
received power.
2.5 Conclusions
This chapter has explored some common channel measurement techniques as well as
presenting commonly used channel models.
Table 2.4
Path loss factors for different environments
Location
Lecture hall
Corridor 1
Corridor 2
Through wall (NLOS)
Through wall (NLOS)
Through wall (NLOS)
TX height (cm)
RX height (cm)
k
110
110
110
220
220
220
110
110
110
60
110
220
1.0454
1.7952
1.4386
3.8514
3.3009
3.1797
38
UWB Theory and Applications
Figure 2.28 Path loss in NLOS measurements
Due to the large frequency band that the UWB signal occupies, the frequency
diversity of the signal is extremely high. The UWB signal is therefore not very sensitive
to deep notches in any part of the signal spectrum and the fading mechanism differs
greatly in the opposite edges of the signal band.
Due to the large bandwidth of the UWB signal, the propagation phenomena are
different in the lower and upper bands of the signal spectrum. The size of the Fresnel
zone, which is a function of wavelength, will be markedly different at the lower and
higher frequency ends. Surface roughness and reflection coefficients may also be significantly different as the wavelength varies significantly over the available signal bandwidth.
In general, the UWB channel model is very similar to well-known tap-delay line
models for wideband systems. One characteristic that is different is the relatively large
number of multipath components.
3
Modulation Schemes
Matti Hämäläinen, Raffaello Tesi, Ian Oppermann
3.1 Introduction
Many different pulse generation techniques may be used to satisfy the requirements
of an UWB signal. As discussed in the previous chapters, the FCC requires that
the fractional bandwidth is greater than 20 %, or that the bandwidth of the transmitted signal is more than 500 MHz, whichever is less. The FCC also stipulates peak
power requirements (Federal Communications Commission, 2002a) Many possible
solutions may be developed within these restrictions to occupy the available bandwidth.
UWB systems have historically been based on impulse radio concepts. Impulse radio
refers to the generation of a series of very short duration pulses, of the order of
hundreds of picoseconds. Each pulse has a very wide spectrum, which must adhere to
the spectral mask requirements. Any given pulse will have very low energy because of
the very low power levels permitted for typical UWB transmission. Therefore, many
pulses will typically be combined to carry the information for one bit. Continuous pulse
transmission introduces a complication in that, without further signal processing at the
transmitter, strong spectral lines will be introduced into the spectrum of the transmitted
signal. Several techniques are available for minimizing these spectral lines, the most
common of which are described later in this chapter.
Impulse radio has the significant advantage of being essentially a baseband technique.
The most common impulse radio based UWB concepts are based on pulse position
modulation with time hopping (TH-PPM). Time hopping, direct sequence techniques
and multi-carrier schemes are described in this chapter. However, the focus will be on
impulse radio modulation schemes.
UWB Theory and Applications Edited by I. Oppermann, M. Hämäläinen and J. Iinatti
Ó 2004 John Wiley & Sons, Ltd ISBN: 0-470-86917-8
UWB Theory and Applications
40
3.2 Impulse Radio Schemes
3.2.1
Impulse radio UWB
Time-modulated ultra wideband (TM-UWB) communication is based on discontinuous
emission of very short Gaussian pulses or other types of pulse waveforms (monocycles), as
in Figure 3.1(a). Each pulse has the ultra wide spectral requirement in frequency domain
like that in Figure 3.1(b). This type of transmission does not require the use of additional
carrier modulation as the pulse will propagate well in the radio channel. The technique is
therefore a baseband signal approach. This radio concept is referred as impulse radio (IR).
One transmitted symbol is spread over N monocycles to achieve processing gain that may
be used to combat noise and interference. This is similar to the approach used for spread
spectrum systems. The processing gain in dB derived from this procedure can be defined as
PG1 ¼ 10 log10 ðNÞ:
ð3:1Þ
The monocycle waveform can be any function which satisfies the spectral mask regulatory requirements. Common pulse shapes include Gaussian, Laplacian, Rayleigh or
Hermitian pulses. Data modulation is typically based on pulse position modulation
(PPM). Conroy et al. (1999) present pulse amplitude modulation (PAM) in UWB
transmission. The UWB receiver is a homodyne cross-correlator that is based on the
architecture that utilizes a direct RF-to-baseband conversion. Intermediate frequency
conversion is not needed, which makes the implementation simpler than in conventional
(super-)heterodyne systems.
Unlike spread spectrum systems, the pulse (chip) does not necessarily occupy the
entire chip period. This means that the duty cycle can be extremely low. The receiver is
only required to ‘listen’ to the channel for a small fraction of the period between pulses.
The impact of any continuous source of interference is therefore reduced so that it is
only relevant when the receiver is attempting to detect a pulse. This leads to processing
gain in the sense of a spread spectrum system’s ability to reduce the impact of interference. Processing gain due to the low duty cycle is given by
Tf
PG2 ¼ 10 log10
;
Tp
ð3:2Þ
where Tf is time hopping frame and Tp is impulse width.
Total processing gain PG is a sum of the two processing gains (Time Domain
Corporation, 1998):
PG ¼ PG1 þ PG2 :
ð3:3Þ
As the transmitted signal is not continuous, UWB communication is resistant to severe
multipath propagation. If the time between pulses is greater than the channel delay
spread, there is no ISI between pulses and so no ISI between bits. In discontinuous
Modulation Schemes
41
(a)
(b)
Figure 3.1 Gaussian monocycle in time domain (a) and in frequency domain (b)
UWB Theory and Applications
42
transmission, consecutive pulses are sent within a time frame (Tf ) which is defined by a
Pseudo random (PR) time-hopping code. Due to the short pulse width and a relative
long pulse repetition time (compared with pulse width), the transmitted pulse is
attenuated before the next pulse is sent. This reduces interpulse interference.
In the time domain, the transmitted Gaussian monocycle can be defined as the first
derivative of the Gaussian function. Figure 3.1 shows the time and frequency domains
of sample monocycles of different duration.
The Gaussian monocycle in time domain v(t) can be mathematically defined using
the formula (Time Domain Corporation, 1998)
rffiffiffiffiffi
ep t 6pðt=Tp Þ2
e
;
vðtÞ ¼ 6A
3 Tp
ð3:4Þ
where A is pulse amplitude, Tp is pulse width and t is time.
The corresponding function in the frequency domain is the Fourier transform of v(t),
which can be defined as
FfvðtÞg ¼ Vðf Þ ¼ j
Af Tp2
3
rffiffiffiffiffi
ep p f 2 Tp2
e 6
;
2
ð3:5Þ
where f is frequency.
The nominal centre frequency and the bandwidth of the monocycle depends on the
monocycle’s width. The 3 dB bandwidth is approximately 116 % of the monocycles
nominal centre frequency f0 ¼ 1=Tp (Time Domain Corporation, 1998). Considering an
example monocycle duration of 0.75 ns (as shown in Figure 3.1(a)), the nominal centre
frequency is 1.33 GHz and the half power bandwidth is 1.55 GHz. The spectrum of
Gaussian monocycle is asymmetrical, as can be seen in Figure 3.1(b).
The ideal Gaussian monocycle has a single zero crossing. If additional derivatives of
the Gaussian pulse are taken, the relative bandwidth decreases, and the centre frequency
increases for a fixed time decay constant Tp . Additional zero crossings of the
pulse will reduce the space and time resolution of the system due to the decreased
bandwidth.
Figure 3.2 presents a block diagram of a time-hopping UWB impulse radio concept
that utilizes pulse position modulation (early concept by Time Domain Corporation,
USA).
The pulse waveforms and their spectra are collected in Figures 3.3 and 3.4. In
Figure 3.3, the solid line represents the generated pulse, and the dashed line represents
the pulse waveform in the channel. The UWB antenna acts as a high pass filter, and may
be thought of as a differentiation block in the time domain (Ramirez-Mireles and
Scholtz, 1998a). The transmitted pulse is therefore the first derivative of the generated
pulse waveform.
In Figures 3.3 and 3.4, a Gaussian doublet is also presented as a potential waveform
candidate. However, this waveform is more suitable for geolocation and positioning
applications than for communication applications because of the longer total bi-pulse
Modulation Schemes
43
Impulse Radio Transmitter
Modulation
Impulse Radio Receiver
Pulse
Generator
Cross
Correlator
Variable
Delay
Variable
Delay
Code
Generator
Code
Generator
Time
Base
Time
Base
LPF
BPF
DSP
Modulation
Output
Figure 3.2 Block diagram of the TH-PPM UWB impulse radio concept presented by Time
Domain Corporation
0
1
0.1
0.2
0.3
0.4
0.5
0.6
0.7 0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
2
1.5
Amplitude
0.5
1
0.5
0
0
–0.5
–1
–0.5
–1.5
Gaussian pulse
–1
nd
the 2
derivative of Gaussian pulse
Tw
Amplitude
1
–2
1
0.5
0.5
0
0
–0.5
–0.5
–1
the 3rd derivative of Gaussian pulse
0
0.1
0.2
0.3 0.4 0.5
Time [ns]
Figure 3.3
0.6
Tw = 0.75 ns
Gaussian doublet
0.7 0
0.5
–1
1
Time [ns]
Gaussian-pulse-based waveforms
1.5
UWB Theory and Applications
44
0
Normalized power level [dB]
–20
–40
–60
Tw = 2 ns
–80
–100
Gaussian pulse
the 2nd derivative of Gaussian pulse
the 3rd derivative of Gaussian pulse
Gaussian doublet, Tw = 1 ns
Gaussian doublet, Tw = 2 ns
–120
–140
0
1
2
3
4
5
6
7
8
9
10
Frequency [GHz]
Figure 3.4
Spectra of the pulse waveforms used (pulse width 0.5 ns)
width. The doublet consists of two amplitude reversed Gaussian pulses with a time delay
Tw between the pulses. This limits the use of the doublet in high speed data communication applications. In specific cases, Tw can be used to generate nulls to the spectrum
in order to avoid intentional interference. This can be seen from Figure 3.4.
3.2.2
Fast Stepped Frequency Chirps
Both the bandwidth and spectral mask limitations must be satisfied by the UWB signal.
There are many means of generating a very wideband signal. One technique is based on
fast frequency chirps, which are commonly used in impulse radar applications. It is
possible to generate a wideband transmission by sweeping the transmitter’s oscillator in
the frequency domain. A bandwidth of several hundred MHz can be achieved with
10 ns sweep time (Stickley et al., 1997). Wider bandwidths can be achieved using this
technique. For example, ground penetrating radar (GPR) systems with 50–1200 MHz
bandwidth have been documented (Carin and Felsen, 1995).
GPRs based on UWB technology are suitable for object detection in, for example,
landmine sweeping and avalanche rescue operations because of the good signal penetration ability and fine space resolution.
Modulation Schemes
45
3.3 Multi-Carrier Schemes
3.3.1
Multi-carrier Spread Spectrum Schemes
Another approach is to extend the techniques utilized for direct sequence spread
spectrum or code division multiple access (CDMA) schemes that are used for thirdgeneration mobile systems. Wideband CDMA systems with optional multi-carrier (MC)
techniques can be used to fill the available spectral mask.
There are three main techniques for generating a SS-MC transmission: multi-carrier
CDMA, multi-carrier DS-CDMA, and multitone (MT) CDMA (Prasad and Hara, 1996).
Each of these techniques relies on, and benefits from, the properties of conventional spread
spectrum signals. However, multi-carrier systems are reasonably complex to implement.
In particular, multi-carrier systems require several mixers or digital fast Fourier (FFT)
transform techniques to place the different signal components in the required bands.
Figure 3.5 shows a block diagram for a MC-CDMA system. The original data stream
is spread over the different sub-carriers fi with each chip of pseudo random (PR) code
Ci . The spectrum spreading is done in the frequency domain (Prasad and Hara, 1996).
The signal is de-spread at the receiver using corresponding chips gi of the spreading code
Ci. In UWB applications, the individual modulated carrier needs to fulfil the 500 MHz
bandwidth requirements.
Figure 3.6 shows a block diagram for a multi-carrier DS-CDMA system. This
method involves spreading the original data in the time domain after serial-to-parallel
conversion of the data stream (Prasad and Hara, 1996). The system needs to obtain the
500 MHz minimum bandwidth requirement to be treated as UWBs.
Figure 3.7 shows the block diagram of a multi-tone CDMA system. The bandwidth
of the MT-CDMA system is smaller than in the previous multi-carrier systems because
of the small sub-carrier spacing. This kind of multi-carrier approach is the closest to the
original UWB idea. However, it causes the highest self-interference due to the overlapping spectra.
Multi-carrier technology is currently used in high data rate applications, for example
in WLAN systems such as Hiperlan2, Digital Audio or Video Broadcasting (DAB and
DVB, respectively) and in asymmetric digital subscriber line (ADSL).
C1
cos(2πf1t)
C2
cos(2πf2t)
Cn
cos(2πf1t) g1
cos(2πf2t) g2
Σ
LPF
LPF
cos(2πfnt) gn
cos(2πfnt)
Σ
LPF
f1
Figure 3.5
f2
fn
f
Block diagram and spectrum for multi-carrier CDMA system
UWB Theory and Applications
46
cos(2πf1t) C(t)
C(t)
cos(2πf1t)
Serial
to
parallel
converter
LPF
C(t)
cos(2πf2t)
C(t)
cos(2πf2t) C(t)
Σ
cos(2πfnt) C(t)
cos(2πfnt)
Figure 3.6
Parallel
to
serial
converter
LPF
f1
f2
fn
f
Block diagram and spectrum for multi-carrier DS-CDMA system
C1
Serial
to
parallel
converter
LPF
C2
cos(2πf1t)
cos(2πf2t)
cos(2πf1t)
Rake
Combiner 1
cos(2πf2t)
Rake
Combiner 2
cos(2πfnt)
Rake
Combiner n
Σ
Cn
cos(2πfnt)
f1 f2 fn
Figure 3.7
Parallel
to
serial
converter
f
Block diagram and spectrum for multi-tone- CDMA system
The advantage of multi-carrier technologies over single carrier systems is that the
data rate in each sub-carrier is lower than for single carrier systems. This eases synchronization of spreading sequence at the receiver, and helps to avoid ISI. The disadvantage
is the increased complexity of the receiver requiring either multiple mixing stages or fast
Fourier transform processing.
MC-CDMA schemes spread the signal in the frequency domain. The signal consists
of overlapping, relatively narrow, carriers which fill the available UWB signal spectrum.
MC-DS-CDMA and MT-CDMA schemes apply spreading in the time domain.
MT-CDMA has a similar bandwidth to a basic DS-CDMA scheme (Prasad and Hara,
1996) with relatively small separation of the tones used ( f1 to fn in Figure 3.7). Consequently, the spreading factor employed for MT-CDMA schemes are much higher than
for the MC schemes. The increased spreading factor increases the processing speed
required at the receiver.
The conventional DS-SS technique without multi-carrier properties can also be
characterized as an UWB technique if the chip rate is high enough. This calls for
extremely fast digital signal processing which may be impractical. Issues such as synchronization will also be a significant challenge.
Modulation Schemes
3.3.2
47
Multiband UWB
A current proposal within the IEEE 802.15.3 working group for UWB signals (IEEE,
2004) utilizes overlapping groups of UWB signals which each have a bandwidth of
approximately 500 MHz. This so called multiband UWB ensures adherence to the FCC
minimum bandwidth requirements and allows efficient utilization of the available
spectrum.
Figure 3.8 shows the spectrum plan for the first group of UWB signals. The subbands are spaced 470 MHz apart, and any number of 500 MHz signals may be
utilized. This allows for flexible coexistence with existing communications systems
(such as WLAN systems). Each sub-band is generated by an OFDM symbol with
10 dB bandwidth of 520 MHz. Figure 3.9 shows the second group of UWB signals,
which overlap the first group of UWB signals by 235 MHz. This potentially enhances
the system’s flexibility with respect to coexistence, interference mitigation and multiple access.
Each group of UWB signals is divided into lower sets (sub-bands 1–8) and upper sets
(sub-bands 9–15). Only seven sub-bands are used in the lower set, which means one subband can be avoided for coexistence. The upper set is used in parallel with the lower set
to increase the bit-rate.
3.4 Data Modulation
A number of modulation schemes may be used with UWB systems. The potential
modulation schemes include both orthogonal and antipodal schemes.
1
Power Spectrum Magnitude (dB)
0
–1
–2
–3
–4
–5
–6
–7
–8
–9
–10
3
4
5
6
7
8
9
Frequency
Figure 3.8
Multiband frequency plan—Group A
10
11
×109
UWB Theory and Applications
48
1
Power Spectrum Magnitude (dB)
0
–1
–2
–3
–4
–5
–6
–7
–8
–9
–10
3
4
Figure 3.9
3.4.1
5
6
7
Frequency
8
9
10
11
×109
Multiband frequency plan—Group A and Group B
Pulse Amplitude Modulation
The classic binary pulse amplitude modulation (PAM) can be presented using e.g. two
antipodal Gaussian pulses as shown in Figure 3.10. The transmitted binary baseband
pulse amplitude modulated information signal (t) can be presented as
xðtÞ ¼ dj wtr ðtÞ;
ð3:6Þ
where wtr ðtÞ represents the UBW pulse waveform, j represents the bit transmitted (‘0’ or
‘1’) and
1; j ¼ 0
dj ¼
:
ð3:7Þ
1; j ¼ 1
In Figure 3.10, the first derivative of the Gaussian pulse is shown, analytically defined as
2
t
t
ð3:8Þ
wG1 ðtÞ ¼ pffiffiffiffiffiffi e 22 ;
2p3
where deviation is directly related with the pulse length Tp by ¼ Tp =2p.
3.4.2
On–Off Keying
The second modulation scheme is binary on–off keying (OOK). Using the following
definitions
0; j ¼ 0;
dj ¼
ð3:9Þ
1; j ¼ 1
Modulation Schemes
49
0.8
0.6
Amplitude
0.4
0.2
0
–0.2
–0.4
–0.6
–0.8
0
Bit ‘1’, Ptype:1, Tp = 1 ns
Bit ‘0’, Ptype:1, Tp = 1 ns
0.2
0.4
0.6
0.8
1
1.2
Time [ns]
1.4
1.6
1.8
2
Figure 3.10 BPAM pulse shapes for ‘1’ and ‘0’ bits
the waveform used for this modulation can be defined in equation 3.8 (Figure 3.11). The
major difference between OOK and PAM is that nothing is transmitted in OOK when
bit ‘0’ is chosen.
3.4.3
Pulse Position Modulation
With pulse position modulation (PPM), the chosen bit to be transmitted influences the
position of the UWB pulse. That means that while bit ‘0’ is represented by a pulse
originating at the time instant 0, bit ‘1’ is shifted in time by the amount of from 0.
Analytically, the signal can be represented as
xðtÞ ¼ wtr t dj ;
ð3:10Þ
where dj assumes the following values, depending on the bit chosen to be transmitted,
dj ¼
0;
1;
j¼0
j ¼ 1;
ð3:11Þ
and the other variables have been defined previously.
The value of could be chosen according to the autocorrelation characteristics of the pulse.
The autocorrelation function can be analytically defined as (Proakis and Salehi, 1994)
ðtÞ ¼
Z
þ1
wtr ðÞwtr ðt Þd:
1
ð3:12Þ
UWB Theory and Applications
50
0.8
0.6
0.4
Amplitude
0.2
0
–0.2
–0.4
–0.6
–0.8
0
Bit ‘1’, Ptype:1, Tp = 1 ns
Bit ‘0’
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Time [ns]
Figure 3.11 OOK pulses used for ‘1’ and ‘0’ bits
For instance, if we want to implement a standard PPM with orthogonal signals, the
optimum value for (which we call opt ) will be the one which satisfies
opt ¼
Z
þ1
wtr ðÞwtr opt d ¼ 0:
ð3:13Þ
1
Figure 3.12 shows a particular case of PPM transmission where data bit ‘1’ is sent
delayed by a fractional time interval < 1, and data bit ‘0’ is sent at the nominal
time.
The optimal modulation changes by using different pulse waveforms. The theoretical
performance in an additive white Gaussian noise channel can be achieved with nonoverlapping, orthogonal pulses, specifically, the modulation index Tp . However,
optimal bit error rate (BER) performance and higher data rates will be achieved if the
modulation index < Tp is as shown in Figure 3.13. The optimal modulation index is
independent of pulse width because of the definition of relative fraction of pulse width.
As the order of the derivative increases, the minimum bit error rate is reached for a
lower value of , and better BER performance is achieved. The justification for this
behaviour is related to the cross-correlation of the pulses related to the data-bits ‘0’ and
‘1’. Figure 3.14 shows the autocorrelation values for different kinds of pulses. As
previously mentioned, the pulse width does not affect the results related to the different
values of .
Modulation Schemes
51
0.8
δ
0.6
0.4
Amplitude
0.2
0
–0.2
–0.4
–0.6
Bit ‘0’, Ptype:1, Tp = 1 ns
Bit ‘1’, Ptype:1, Tp = 1 ns
–0.8
0
0.5
1
1.5
2
2.5
Time [ns]
Figure 3.12
PPM pulse shapes for ‘1’ and ‘0’ bits
Two peculiarities should be noted for PPM modulation:
.
.
The autocorrelation functions of the Gaussian waveforms have both positive and
negative values. This explains why it is possible to achieve a better BER performance
than the BER performance for time-orthogonal pulses with values less than 1 ( 1
implies time-orthogonal signals, as seen in Figure 3.14).
The autocorrelation minima occur at values, which correspond to the best BER
performances.
The behaviour of the cross-correlation provides a means of selecting the optimal value
of for the AWGN channel case. The value of can be fixed a priori once the UWB
pulse waveform has been chosen. The best value to use for can be selected once the
cross-correlation of the selected pulse waveform is calculated. The optimal value of for
each pulse waveform is presented in Table 3.1.
3.4.4
Pulse Shape Modulation
Pulse shape modulation (PSM) uses different, orthogonal waveforms to represent bit ‘0’
and ‘1’. The transmitted pulse can be represented as
ð0Þ
ð1Þ
xðtÞ ¼ 1 dj wtr ðtÞ þ dj wtr ðtÞ;
ð3:14Þ
UWB Theory and Applications
52
Figure 3.13
Bit error rate for different pulse waveforms as a function of Table 3.1 Optimal time shift values for
PPM modulation in an AWGN channel
Waveform
Optimal Second Derivative
Third Derivative
Fourth Derivative
Fifth Derivative
0.292683Tp
0.243902Tp
0.219512Tp
0.195122Tp
ð0Þ
ð1Þ
where dj is defined as (2.11) and wtr and wtr represent two different waveforms.
Figure 3.15 shows an example pair of PSM pulses. This case uses the first and the
second derivatives of the Gaussian pulse. These two waveforms are orthogonal
according to the definition given by the cross correlation of the two waveforms
c ðtÞjt¼0 ¼
Z
þ1
ð0Þ
ð1Þ
wtr ðÞwtr ðt Þd ¼ 0:
1
ð3:15Þ
Modulation Schemes
53
1
2nd derivative of the Gaussian pulse, Tp = 0.5ns
3rd derivative of the Gaussian pulse, Tp = 0.5 ns
4th derivative of the Gaussian pulse, Tp = 0.5 ns
5th derivative of the Gaussian pulse, Tp = 0.5 ns
0.8
0.6
Correlation factor
0.4
0.2
0
–0.2
–0.4
–0.6
–0.8
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
δ - Time shift of bit ‘1’ (Fraction of the pulse width)
Figure 3.14
3.4.5
Autocorrelation of different pulse waveforms
Theoretical Bounds
The theoretical bounds in AWGN can be calculated for the modulation schemes
presented. It is clear that BPAM is an antipodal modulation. Let x0 ðtÞ and x1 ðtÞ the
signal representing bit 0 and 1 respectively. The BPAM signals are related by
Z
Tp
x0 ðtÞx1 ðtÞdt ¼ 1;
ð3:16Þ
0
where Tp is the time duration of the signal. It can be shown that OOK and PPM are
orthogonal using that same approach, that is,
Z
Tp
x0 ðtÞx1 ðtÞdt ¼ 0:
ð3:17Þ
0
The orthogonality of the PSM signals depends on the pulse waveforms chosen to describe
bit 0 and 1. For example, in the case of the first and second derivatives of the Gaussian
pulse, it can be shown that the signals are orthogonal (da Silva and de Campos, 2002).
UWB Theory and Applications
54
0.8
0.6
0.4
Amplitude
0.2
0
–0.2
–0.4
–0.6
–0.8
–1
Bit ‘0’, Ptype:1, Tp = 1 ns
Bit ‘1’, Ptype:2, Tp = 1 ns
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Time [ns]
Figure 3.15 Examples of the pulse waveforms used for PSM modulation
The probability of error of an antipodal signal and of an orthogonal signal in an
AWGN channel as a function of the signal-to-noise ratio has values (Proakis, 1995)
ðantÞ
Pb
ðortÞ
Pb
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 SNR ;
pffiffiffiffiffiffiffiffiffiffiffi
¼ Q SNR ;
¼Q
ð3:18Þ
ð3:19Þ
as depicted in Figure 3.16.
3.5 Spectrum ‘Spreading’
Continuous pulse generation leads to strong spectral lines in the transmitted signal at
multiples of the pulse repetition frequency. Data modulation typically occurs in a
number of conceptual stages. First, a pulse train is generated. Second, a randomizing
technique is applied to break up the spectrum of the pulse train. Third, the data
modulation is applied to carry the information. The two main approaches to randomizing the pulse train are time hopping (TH) and direct sequence (DS) techniques.
The spectrum of a pulse train with and without randomizing techniques is depicted in
Figure 3.17. Figure 3.17(a) shows the spectrum of a simple pulse train. As can be seen,
the spectrum contains strong spectral lines at multiples of the pulse repetition frequency.
The envelope of the spectrum is that of a single pulse. The regularity of these energy
Modulation Schemes
55
100
Probability of error (Pb)
10–1
10–2
10–3
10–4
10–5
10–6
Antipodal signals
Orthogonal signals
0
1
2
3
4
5
6
7
8
9
10
Signal-to-noise ratio (SNR)
Figure 3.16 Probability of error for signals in AWGN channel
spikes may interfere with other communication systems at short range (Proakis, 1995).
Randomizing the position in time of the generated pulses using data modulation and
other randomizing techniques will affect the spectrum in such a way that the energy
spikes are spread all over the spectrum, which is therefore smoothed. Figure 17(b) shows
the spectrum of a pulse train that includes time hopping based randomization. A number
of randomizing techniques may be found in the literature (Withington et al., 1999).
Randomizing is typically realized in UWB systems by a pseudo-random sequence.
3.5.1
TH-UWB
The data modulation is typically based on PPM using TH-UWB as the basis for a
communication system. This approach allows matched filter techniques to be used in
the receiver. Values of time shift (which is the modulation index for this form of
modulation) have been reported as approximately one-quarter of a pulse width
(Withington et al., 1999). The optimum time shift depends on the cross-correlation
properties of the pulses used. The concept of TH-PPM is presented in Figure 3.18 where
only monocycle per data symbol is used (no processing gain is achieved). Figure 3.19
shows the construction of a single bit for TH-UWB systems.
TH-PPM monocycles spread the RF energy across the frequency band, reducing the
large spikes in the pulse train spectrum. When a PR code is used to determine the
transmission time within a large time frame, the spectra of the transmitted pulses
UWB Theory and Applications
56
Figure 3.17
‘1’ ‘1’
Spectrum of pulse train without (a) and with (b) randomizing techniques
‘0’
‘0’ ‘1’
‘0’
‘1’ ‘0’
‘1’ ‘0’
‘1’
‘0’
‘1’ ‘0’ ‘0’
‘0’
t
Time hopping frame
t
Actual transmission time
due to PPM
Nominal transmission time
Figure 3.18
Time-hopping pulse position modulation technique
Modulation Schemes
57
Time Hopping
1 bit (Td = N ∗Tf)
Data bit
Tp
Tf = M ∗Tp
Figure 3.19 Time hopping system concept
become much more white-noise-like. The time hopping randomizes the signal in both
time and frequency domains (Withington et al., 1999). Pseudo random time hopping
also minimises collisions between users in multiple access systems, where each user has a
distinct pulse shift pattern (Win and Scholtz, 1997a).
However, a consequence of the PR time-modulation is that the receiver needs accurate knowledge of the PR code phase for each user. One can imagine impulse radio
systems as time hopping spread spectrum systems. UWB waveforms are generated
without any additional spreading. This simplifies the transceivers relative to conventional SS transceivers (Fontana et al.). The data rate of the transmission can be selected
by modifying the number of pulses used to carry a single data bit (Kolenchery et al.,
1997). This in turn has an effect on the processing gain.
The pulse repetition time (or frame time) typically ranges from a hundred to a thousand
times the pulse (monocycle) width. The symbol rate can be defined in terms of the number
of monocycles used to modulate one data symbol in fixed frame time as
Rs ¼
1
1
¼
½Hz;
Ts Ns Tf
ð3:20Þ
where Rs ¼ symbol rate, Ts ¼ symbol time, Tf ¼ time hopping frame and Ns ¼ number
of monocycles/data bit.
If the data rate in (3.20) is reduced whilst the time hopping frame remains constant,
the number of monocycles per data bit is increased. This leads to an increased processing gain.
3.5.2
Data Modulation with Time Hopping
In TH-mode, the pulse transmission instant is defined by the pseudo-random code.
One data bit is spread over the multiple pulses to achieve a processing gain due to
the pulse repetition (3.1). The processing gain is also increased by the low transmission duty cycle (3.2).
UWB Theory and Applications
58
The TH spreading approach has been studied for PAM, PPM and PSM. However,
OOK cannot take advantage of TH spreading because of the blank transmission in case
of bit ‘0’, and because it would create further problems for synchronization.
The information signal s(t) for the mth user can be analytically described for PAM
modulation as
sðmÞ ðtÞ ¼
1 N
1 X
X
ðmÞ
ðmÞ
w t kTd jTf ðcw Þj Tc dk ;
ð3:21Þ
k¼1 j¼0
for PPM modulation as follows
sðmÞ ðtÞ ¼
1 N
1 X
X
ðmÞ
ðmÞ
w t kTd jTf ðcw Þj Tc dk ;
ð3:22Þ
k¼1 j¼0
and for PSM modulation as follows
sðmÞ ðtÞ ¼
1 N
1
X
X
k¼1 j¼0
ðmÞ
wd ðmÞ t kTd jTf ðcw Þj Tc :
ð3:23Þ
k
Figures 3.20, 3.21 and 3.22 show a single data bit for PAM, PPM and PSM respectively.
Note that in the case of PAM and PPM the same TH sequence has been used to
represent either data bit ‘0’ and ‘1’, while in the PSM case, two different sequences have
Figure 3.20 Time window of a transmitted data bit for BPAM modulation with TH
spreading
Modulation Schemes
59
Figure 3.21 Time window of a transmitted data bit for PPM modulation with TH spreading
4
3
2
Amplitude
1
0
–1
–2
–3
–4
Databit ‘0’, TH-PSM, Ptype:1, Tp = 1ns
Databit ‘1’, TH-PSM, Ptype:2, Tp = 1ns
–5
0
5
10
15
20
25
30
35
40
45
50
Time [ns]
Figure 3.22 Time window of a transmitted data bit for PSM modulation with TH spreading
UWB Theory and Applications
60
been considered. This results in a different allocation of the single pulses within the data
bit frame. The pulse repetition frame is assumed to be 10 ns long in all of these figures.
3.5.3
Multiple Access with TH-UWB
In TH systems, users are separated using different PR codes of length N. In a frame,
there are N possible transmission instants, so under ideal conditions a maximum of
M ¼ N users can be allocated into the system without creating interference.
In time-hopping mode (TH-UWB), the modulated information signal s(t) for the mth
user can be written as: (3.21)
In TH-UWB Tf Tp producing a low duty cycle. In the following discussion, M is
assumed to be the same as N. The input signal for the receiver in one user case is given by:
rðtÞ ¼
L
X
Al sð1Þ ðt l Þ þ nðtÞ
ð3:25Þ
l¼1
where Al is amplitude of radio channel path l, l is delay of radio channel path l, L is the
number of resolvable multipath components, nðtÞ is additive white Gaussian noise.
The delay can be presented as a portion of Tc as l ¼ Tc . The effect of the antenna
should be taken into account in the received signal waveform in s(t).
Each pulse in a pulse train has a nominal transmission time, which is determined by the
pulse repetition frequency (PRF). The actual transmission instant is varied from the
nominal position by pulse position modulation. This pulse position modulation carries
information, so that an early pulse represents a ‘0’ and a delayed pulse represents a ‘1’.
The determination of transmission for a given user is determined by the unique PR code.
The pulse repetition interval (which defines the length of each time hopping frame) is
defined by the number of users multiplied by the length of a single time slot within the
time hopping frame as given by
TPRF ¼ NU TC ;
ð3:26Þ
where TC is the length of each time slot, and NU is the number of users in the channel.
The maximum number of non-overlapping users is determined by the length of the
PR code as seen in
NU ¼ 2n 1;
ð3:27Þ
where n is the number of bits in the PN sequence generator.
The length of each of the time slots must be more than twice the length of a single
pulse, since there must be enough time within the slot to transmit either a ‘0’ or a ‘1’.
This is assuming that the modulation index is no less than the length of a single pulse.
This is assumed to be the minimum so as to avoid the overlapping of the pulses. In
general, the length of a time slot is determined by
TC > 2Tp þ ;
where TP is the length of the time delay (modulation index) used in PPM.
ð3:28Þ
Modulation Schemes
61
1 bit (Td = Ns∗Tc)
DS-UWB
Data bit
Spreading code in DS-SS system
τc
Figure 3.23
Direct Sequence system concept
For example, if the number of users is 31 and the pulse width is 800 ps, the length of
the time slot within the pulse to be transmitted must be at least 1.6 ns. This results in a
pulse repetition frequency of less than 21 MHz. By choosing the number of pulses per
symbol to be 200, the processing gain is more than 41 dB.
3.5.4
Direct Sequence UWB
When utilizing DS techniques, a PR code is used to spread the data bit into multiple chips,
much as in conventional DS spread spectrum systems. In the case of UWB systems, the
pulse waveform takes the role of the chip in DS. The DS spreading approach has been
studied for PAM, OOK and PSM modulation schemes. PPM modulation is intrinsically
a time hopping technique since the bit value is given by the position of the pulse in a
transmission slot. The use of the DS technique as a spreading approach would create a
hybrid DS/TH configuration of the signal. Figure 3.23 shows the bit structure for a DS
signal. The rectangular waveform indicates the individual chip elements.
The PAM and OOK information signal s(t) for the mth user can be presented as
sðmÞ ðtÞ ¼
1 N
1
X
X
ðmÞ ðmÞ
wðt kTd jTc Þðcp Þj dk ;
ð3:29Þ
k¼1 j¼0
where dk is the kth data bit, ðcp Þj is the jth chip of the PR code, wðtÞ is the pulse
waveform, N represents the number of pulses to be used per data bit, Tc is the chip
length, The pseudo random code is bipolar assuming values f1; þ1g, The bit length is
Td ¼ NTc ¼ NTp .
3.5.4.1
Data Modulation with DS-UWB
Figure 3.24 shows a single data bit for binary PAM modulation when a data bit ‘1’ or ‘0’
is transmitted. The square wave represents the random code, which affects the polarity
of individual pulses which make up the DS waveform. For clarity, only a small part of
the data bit waveform is shown. Pulse type (Ptype) defines the pulse waveform used.
Ptype2 is the second derivative of the Gaussian pulse.
UWB Theory and Applications
62
Figure 3.24
spreading
Time window of a transmitted data bit for BPAM modulation with DS
Figure 3.25 shows a single data bit for OOK modulation when data bit ‘1’ or ‘0’ is
transmitted. As mentioned above, in case of data bit ‘‘0’’, the OOK spreading signal is
characterized by no transmission.
For PSM modulation, the information signal s(t) for the mth user can be presented as
sðmÞ ðtÞ ¼
1 N
1
X
X
k¼1 j¼0
ðmÞ
wd ðmÞ ðt kTd jTc Þðcp Þj ;
ð3:30Þ
k
where the chosen bit dk for the mth user, defined according to equation (3.14), determines the choice of the UWB pulse waveform to be transmitted.
Figure 3.26 shows a single data bit using for PSM modulation when data bit ‘1’ or ‘0’
is transmitted. Ptype1 and Ptype2 in the figure represent the first and the second
derivatives of the Gaussian pulse.
3.5.5
Comparison of TH and DS BPAM UWB
This section examines the relative performance of BPAM for DS and TH techniques
in an AWGN channel. The BER for up to 60 simultaneous users is examined. The
Modulation Schemes
63
Figure 3.25 Time window of a transmitted data bit for OOK modulation with DS spreading
Figure 3.26
Time window of a transmitted data bit for PSM modulation with DS spreading
UWB Theory and Applications
64
processing gain is 127 (approximately 21 dB). The DS system uses 127 pulses per bit,
which is equal to the processing gain or the spreading sequence length, whilst the TH
system uses 10 pulses per bit, which is the remainder of the processing gain coming
from channel activity factor. Both synchronous and asynchronous transmissions are
considered. All users have the same power. The pulse shape used is the fourth
derivative of the Gaussian pulse. The pulse duration is 0.5 ns and the data rate is
16 Mbps.
Figure 3.27 shows the performance of the DS UWB, while Figure 3.28 shows the
performance of the TH system. There is a substantial difference between synchronous
(solid) and asynchronous (dushed) performance for the DS system. The synchronous
DS system has the pulses for all users transmitted at the same time. The only feature
which can be used to suppress multiple access interference is the de-spreading (pulse
combining) process in the receiver. When the asynchronous system is used, the lower
cross correlation values that occur at different pulse alignments means that there is
substantially less interference per bit to be suppressed.
The performance of the synchronous and asynchronous systems for the TH system is
very similar. This is because each user has a different pulse transmit instant associated
with their PR sequence, so the pulses are offset even if the time hopping frames are
aligned.
The performance of the TH and DS asynchronous systems are very similar. This is to
be expected in an AWGN channel with low duty cycle pulses.
10–1 ++
+
+
+
+
+
+
+
+
+
+
14
16
+
+
+
10–2
Bit error rate
+
10–3
10–4
0
+
3 users synch.
3 users asynch.
10 users synch.
10 users asynch.
30 users synch.
30 users asynch.
+ 60 users synch.
+ 60 users asynch.
Theor. BPAM in AWGN
2
4
6
8
Eb/N0 [dB]
10
12
Figure 3.27 DS-BPAM in AWGN channel Tp ¼ 0:5 ns
Modulation Schemes
65
10–1
+
+
+
+
+
+
+
+
+
10–2
Bit error rate
+
+
+
10–3
10–4
+
+
+
+
3 users synchronous
3 users asynchronous
10 users synchronous
10 users asynchronous
30 users synchronous
30 users asynchronous
+ 60 users synchronous
+ 60 users asynchronous
Theor. BPAM in AWGN
0
2
4
6
8
10
12
14
16
Eb/N0 [dB]
Figure 3.28 TH-BPAM in AWGN channel Tp ¼ 0:5 ns
3.6 Conclusions
UWB systems may be primarily divided into impulse radio (IR) systems and multiband
systems. Multiband systems offer the advantage of potentially efficient utilization of
spectrum.
However, IR systems have the significant advantage of simplicity, and so are potentially lower cost. In addition, IR is essentially a baseband technique. The IR UWB
concepts investigated support many modulation schemes including orthogonal and
antipodal schemes. Basic modulation must also include some form of spectrum randomization techniques to limit the interference caused by the transmitted pulse train. Both
TH and DS randomization techniques were examined. Which modulation scheme to use
depends on the expected operating conditions and the desired system complexity.
4
Receiver Structures
Matti Hämäläinen, Jari Iinatti, Raffaello Tesi, Simone Soderi,
Alberto Rabbachin
4.1 Introduction
UWB systems can be characterized as an extension of traditional spread spectrum (SS)
systems. One of the major differences between UWB systems and traditional SS
systems is the radio channel which they use. As discussed in Chapter 2, the UWB
channel is extremely multipath rich. The multipath components that are combined
increase the total signal power. The multipath components that are not combined lead
to interference. The significantly greater number of resolvable multipath components
associated with UWB system’s much greater bandwidth means that many more
receiver elements need to be considered. This chapter examines some of the more
popular receiver structures for UWB systems, particularly the rake and modifications
of the rake.
Synchronization is one of the main problems in telecommunications, navigation and
radar applications. Different synchronization levels operate for carrier, code, symbol,
word, frame and network synchronization. As in SS systems, code synchronization
should be performed in UWB systems. When the receiver is synchronized, the received
spreading code and reference code are aligned with the same phase. All of these levels of
synchronization can be split into two phases—acquisition (coarse synchronization,
initial synchronization) and tracking (fine synchronization). It is only possible to communicate using spread-spectrum systems if all of the necessary synchronization levels
are performed with sufficient accuracy. This chapter examines code synchronization for
DS and TH UWB systems.
UWB Theory and Applications Edited by I. Oppermann, M. Hämäläinen and J. Iinatti
Ó 2004 John Wiley & Sons, Ltd ISBN: 0-470-86917-8
UWB Theory and Applications
68
4.2 Rake Receiver
The received signal energy can be improved in a multipath fading channel by utilizing a
diversity technique, such as the rake receiver (Proakis, 1995). Rake receivers combine
different signal components that have propagated through the channel by different
paths. This can be characterized as a type of time diversity. The combination of different
signal components will increase the signal-to-noise ratio (SNR), which will improve link
performance. We will consider three main types of rake receivers.
4.2.1
Rake Receiver Types
The ideal rake receiver structure captures all of the received signal power by having a
number of fingers equal to the number of multipath components. The so called idealrake or all rake (I-rake and A-rake) is such a receiver. (Win et al., 1999; Win and
Scholtz, 2003). The problem with this approach is the need for an infinite number of
rake branches, which also means an infinite number of correlators. Consequently,
implementation of the A-rake is not possible. The performance close to the performance
in AWGN channel can be met by using the maximum ratio combining (MRC). MRC
involves coherently combining all of the signal components to achieve optimal performance (Proakis, 1995). Some channel estimation needs to be used to obtain channel
information. The delay resolution of the receiver depends on the signal bandwidth.
A different number of distinguishable propagation paths can be separated by the
receiver based on the channel estimate. Figure 4.1 illustrates the concept behind the
A-rake including the envelope of the channel delay profile and the propagation paths
that can be separated. All of the paths that arrive within the receiver’s time resolution
will be seen as a single path. The energy of the single path is a combination of the energy
of all of the undistinguishable paths.
A practical rake receiver implementation is a selective rake, S-rake. The S-rake
only uses the Lr strongest propagation paths. Information on the channel impulse
response is required in order to use the S-rake. Channel estimation algorithms must
be used to obtain this a-priori information. The SNR is maximized when the
strongest paths are detected. The link performance will be improved relative
to the single path receiver. Figure 4.2 shows the multipath components used by the
S-rake. The channel profile equals to the one presented in Figure 4.1. The complexity
of the S-rake receiver is greatly reduced relative to the A-rake by only selecting those
multipath components that have significant magnitude. In the example channel
profile shown in Figure 4.2, only Lr ¼ 7 of the strongest paths are selected. Depending on the channel delay profile, the selected paths can be consecutive, or spread over
the profile.
The partial-rake receiver, P-rake, is a simplified approximation to the S-rake. The
P-rake involves combining the Lr first propagation paths. The principle behind this
approach is that the first multipath components will typically be the strongest and
contain the most of the received signal power. The disadvantage is that the multipath
components that the P-rake receiver combines are not necessarily the strongest multipath components, so optimum performance will not be achieved. Figure 4.3 shows the
P-rake for Lr ¼ 7 branches using the same channel profile as earlier examples. The
Receiver Structures
Delay profile
69
Delay, τ
Principle of the all-rake receiver
Delay profile
Figure 4.1
Delay, τ
Figure 4.2
Principle of the selective-rake receiver
figure indicates that there are stronger multipath components at later delays than those
which have been combined by the P-rake receiver.
Under ideal conditions, the A-rake outperforms the S-rake, which typically outperforms the P-rake. However, if the strongest propagation paths are at the beginning of
the channel impulse response, the S-rake and P-rake will give the same performance.
Figure 4.4 presents the bit-error-rates for different rake receivers in modified
Saleh–Valenzuela channel 2 (SV2). The number of rake fingers used, Lr , is 1, 4, 10 and 20
UWB Theory and Applications
Delay profile
70
Delay, τ
Figure 4.3
Principle of partial rake receiver
Figure 4.4 The performance of different rake receivers in modified Saleh–Valenzuela
channel 2. Number of fingers Lr ¼ 1; 4; 10 and 20
Receiver Structures
71
for both the P-rake and S-rake. The BER results for A-rake are also presented as a
reference. A significant difference can be seen between the P-rake and S-rake performance because the SV2 is a non-line-of-sight model.
Figure 4.5 presents the average delay profile used in the simulations. The energy
collected by the P-rake is tens of decibels less than the energy collected by the S-rake,
which uses the strongest paths. The single-path S-rake uses the strongest path, so it has
significantly better performance than the single-path P-rake. The system performance
of A-rake is not as good as the ideal system performance in the AWGN channel because
of the limited delay resolution of the A-rake receiver.
4.2.2
Detection Techniques
The rake-based receiver algorithms can be implemented in different ways, depending on
the knowledge of the gain of the channel taps, in both amplitude and phase. These
schemes are characterized as coherent if the phase of the channel tap is recovered, or
non-coherent if the phase of the channel tap is not recovered. Channel tap amplitude
recovery can be considered as a possible choice for non-coherent schemes.
Equal gain combining (EGC) is a coherent detection scheme. EGC requires a perfect
estimation of the phase of each of the channel taps in order to correct the offset at the
Figure 4.5
Average power delay profile of the modified Saleh–Valenzuela model 2
UWB Theory and Applications
72
received signal before the detection block. The received signal for a single databit may
be defined as follows
Lr
X
rðtÞ ¼
an sðt n Þ þ nðtÞ;
ð4:1Þ
n¼1
where Lr is the number of recovered paths, sðtÞ the transmitted signal, nðtÞ the Gaussian
noise in the channel, an ¼ jan j e jn the gain, n the delay of the n th multipath. n will only
assume values 0 and as in the modified SV model (See Chapter 2).
Thus, the EGC decision variables will be
ðEGCÞ
Ui
¼
Lr
X
ejn
Z
Td
rðt n Þwi ðtÞ dt
ð4:2Þ
0
n¼1
where Td is the databit length and wi ðtÞ the pulse waveform (e.g., train of pulses)
representing databit ‘i’ at the receiver.
A second coherent scheme is maximal ratio combining. MRC involves phase recovery
and estimating the received power level for each multipath. The decision variable will
then assume the following form
ðMRCÞ
Ui
¼
Lr
X
an
n¼1
Z
Td
rðt n Þwi ðtÞdt:
ð4:3Þ
0
An absolute combiner (AC) can be implemented in non-coherent detection by adding
the absolute values of the outputs of all the matched filters before feeding the detector.
In this case the decision variables will be
ðACÞ
Ui
Lr Z
X
¼
n¼1
0
Td
rðt n Þwi ðtÞdt:
ð4:4Þ
The implementation of the S-rake receiver is based on the non-coherent power
estimation (PE) of each single channel path. The knowledge of the channel amplitude
can be used to improve the performance of the system, by weighting the output of each
single correlator. This is similar to the MRC approach, except that it does not require
knowledge of the channel phase. This assumption leads to a more sophisticated implementation of AC, analytically defined as
ðACþPEÞ
Ui
Z
Lr X
¼
jan j n¼1
Td
0
rðt n Þwi ðtÞdt :
ð4:5Þ
Receiver Structures
73
For more information upon combining schemes, refer to Proakis (1995).
The simulations results presented below are based on a UWB system having
PG ¼ 20 dB. The length of the pulse waveforms is 0.5 ns, leading to a data rate of
20 Mbit/s. In the TH case, the pulse integration gain PG1 is 10 dB, that is, each data bit
is composed of 10 pulse waveforms, whose time location is defined by the PR code.
The binary modulation schemes are PPM (for TH only), PSM and PAM (for both
TH and DS). The combining approach is one of the techniques depicted above: MRC
and EGC for coherent detection, AC and AC þ PE for non-coherent. PE is assumed
perfect in all cases. The total average signal-to-noise ratio at the receiver, Eb =N0; , is
fixed to two values, 8 dB and 15 dB, respectively. However, the signal-to-noise ratio in
the decision variable is less, since not all the paths have been combined.
The detection block is defined by a selective chip-spaced receiver, with a time resolution of 0.5 ns, that is, equal to the length of the pulse waveform. The channel models
were simulated using at least 100 channel realizations for each case (number of fingers)
examined. The channel power has been normalized to 1 over all the channel realizations
used for each simulation point.
Figures 4.6 and 4.7 show the behaviour of some of the results of the simulations, which
evaluated the BER of the system as a function of the number of rake fingers D for
DS PSM, SV1, 4th, Tp = 0.5 ns, selective rake
100
Bit error rate
10 –1
10 –2
Eb/N0 = 8 dB, EGC
Eb/N0 = 8 dB, MRC
Eb/N0 = 8 dB, AC
Eb/N0 = 8 dB, AC + PE
Eb/N0 = 15 dB, EGC
Eb/N0 = 15 dB, MRC
Eb/N0 = 15 dB, AC
Eb/N0 = 15 dB, AC + PE
10 –3
10 –4
0
Figure 4.6
10
20
30
Number of rake fingers
40
50
BER as a function of the number of fingers for DS-PSM in SV-1
60
UWB Theory and Applications
74
TH-PPM, SV3, 4th, Tp = 0.5 ns, selective rake
100
Bit error rate
10 –1
10 –2
Eb/N0 = 8 dB, EGC
Eb/N0 = 8 dB, MRC
Eb/N0 = 8 dB, AC
Eb/N0 = 8 dB, AC + PE
Eb/N0 = 15 dB, EGC
Eb/N0 = 15 dB, MRC
Eb/N0 = 15 dB, AC
Eb/N0 = 15 dB, AC + PE
10 –3
10 –4
0
10
20
30
Number of rake fingers
40
50
60
Figure 4.7 BER as function of the number of fingers for TH-PPM in SV-3
Eb =N0 ¼ 8 and 15 dB. In particular, Figure 4.6 represents DS-PSM for SV-1, and Figure 4.7
is TH-PPM in SV-3 case. The absence of power estimation in both the coherent approaches
(EGC) and non-coherent approaches (AC) generates a minimum value of BER for a defined
Dopt , which can be clearly chosen as the minimum. This behaviour is more evident in SV-1,
where the presence of a LOS component makes the minimum BER appear for a lower D,
relative to SV-3. In the MRC case, the performances of the systems are continuously
improving as D increases due to the perfect weight used in estimation. Thus, the optimal
value has been chosen where the BER performance tends to saturate. The improvement
given by the use of coherent detection is characterized by a BER approximately 10-times
lower than the equivalent non-coherent implementation for a fixed value of SNR and D.
Table 4.1 depicts the optimal number of fingers for each of the examined systems. The
table gives some general trends, such as Dopt generally increasing with the Eb =N0 . However, this effect is more remarkable for non-coherent systems. Dopt is higher in SV3, due to
absence of LOS component. Dopt is also clearly lower for non-coherent systems. Antipodal
modulations show the same results for a fixed Eb =N0 and channel model, both in terms of
number of fingers and BER. Among orthogonal modulations, TH-PPM the one that
shows the lowest values of Dopt , on average, despite the poorer BER results.
Receiver Structures
Table 4.1
SV
75
Optimal number of rake fingers for different receiver algorithms
Eb =N0 [dB]
UWB system concept
Optimal number of fingers (Dopt)
Coherent
1
Non-coherent
MRC
EGC
AC þ PE
AC
8
DS-PSM
TH-PSM
TH-PPM
DS-BPAM
TH-BPAM
DS-PSM
TH-PSM
12
12
10
12
12
14
20
8
8
10
10
10
8
10
4
4
4
–
–
10
10
4
4
2
–
–
8
6
15
TH-PPM
DS-BPAM
TH-BPAM
DS-PSM
TH-PSM
14
15
15
18
20
8
10
10
16
18
8
–
–
8
6
6
–
–
6
6
8
TH-PPM
DS-BPAM
TH-BPAM
DS-PSM
TH-PSM
20
15
15
20
16
14
15
20
16
16
4
–
–
16
16
4
–
–
12
12
15
TH-PPM
DS-BPAM
TH-BPAM
20
18
18
18
15
20
10
–
–
6
–
–
3
4.3 Synchronization in UWB Systems
4.3.1
Basics
A survey of the published literature reveals a number of publications relating to
acquisition in direct sequence spread spectrum systems (Katz, 2002). Assuming an
AWGN channel, the input signal for the receiver in one user case is given by
rðtÞ ¼
L
X
Al sð1Þ ðt l Þ þ nðtÞ
ð4:6Þ
l¼1
where Al is amplitude of radio channel path l; l is delay of radio channel path l; L is the
number of resolvable multipath components, nðtÞ is additive white Gaussian noise.
UWB Theory and Applications
76
The delay can be presented as a portion of Tc as l ¼ T. The signal waveform s(t) is
dependent on the system concept, that is, TH or DS. In both cases, the signal waveform
includes the effect of the spreading code.
The synchronization stage must provide an estimate e for the timing offset (i.e. e ).
The maximum likelihood (ML) algorithm generates several values e to evaluate when
conditional probability density function of the received signal pðrjÞ achieves the maximum value (Iinatti and Latva-aho, 2001). So,
^e ¼ arg max pðrje Þ:
e
ð4:7Þ
As we assume white Gaussian noise channel, the conditional probability density function is readily evaluated. For the optimal estimation e , we have to maximize
ðe Þ ¼
Z
To
rðtÞsðt; e Þ dt
0
¼
Z
To
Al
0
1 N
1
X
X
wðt kTb jTf Þ:
k¼0 j¼0
1 N
1
X
X
ð4:8Þ
wðt kTb jTf e Þ dt þ nc ðtÞ
k¼0 j¼0
where nc ðtÞ is the additive noise component, and the information signal is in the form
of TH-UWB as described in Chapter 3. In practice, the algorithm is too complex
to implement, and some approximations must be used. In the so-called serial
search strategy, the phase of the local code ðe Þ is changed step-by-step in equal
increments. In this way, the correct position within the uncertainty region can be
found (Iinatti, 1997). Serial search acquisition can be implemented using active
correlation measurement techniques or passive correlation measurement based on
matched filtering (MF).
4.3.1.1
Synchronization Schemes
UWB systems use very large bandwidths, and therefore a dense channel multipath
profile where many components can be distinguished from the received signal.
The multipath channel then introduces more than one correct synchronization cell.
From the perspective of code synchronization, this phenomenon causes problems:
the energy of the signal is spread over many multipath components, and the energy
of each path is very low. Therefore, the paths are difficult to acquire. Depending on
the receiver structure (type and length of rake), a number of paths should be
acquired.
In DS systems, the uncertainty region corresponds to a multiple of the code length.
In TH systems, the uncertainty region is divided into a number of cells (Cu). The number
of cells depends on the number of possible pulse positions combinations in a bit interval.
Recently, an algorithm called chip-level post-detection integration (CLPDI) was
Receiver Structures
77
proposed for code acquisition in direct sequence CDMA systems (Iinatti and Latvaaho, 2001). Algorithm is suitable for synchronization in multipath environments. Once
the CLPDI algorithm finds one of the possible synchronization cells, an additional
sweep has to be performed to acquire the necessary number of paths. The aim of this
initial code acquisition in a multipath channel is to find a starting point to reduce the
multipath search time. In this chapter, the method is applied for time hopping UWB
system due to nature of UWB signals.
Figure 4.8 presents the synchronisation algorithm for a multipath environment
proposed for DS CDMA (Iinatti and Latva-aho, 2001). In DS systems, the impulse
response of the matched filter is the time-reversed replica of the spreading code. This
means that the impulse response has coefficients given by the spreading code, and the
delays between the consecutive coefficients are Tc . The MF output signal is proportional to the autocorrelation function (ACF) of the spreading code. The sampling at
the output of the MF is made at least at the chip rate. The MF is followed by a
threshold comparison, and if the threshold Th is crossed, the acquisition process ends.
There is detection if the threshold is crossed by the ACF at the zero delay. This occurs
with probability of detection, Pd . If the threshold is crossed with some other delay,
a false alarm occurs. This happens with the probability of false alarm, Pfa . False
alarms may be catastrophic, and cause total miss of the correct code phase. Because
the MF gives its peak when the code is inside the filter, a multipath channel leads to
several peaks.
Chip-level post-detection integration performs post-detection integration at the chip
level, i.e., a number of consecutive samples at the output of the MF are combined.
CLPDI is performed over m samples and CLPDI output is sampled at multiples of mTc
as is presented in Figure 4.8. Because the sampling is done as multiples of mTc ,
consecutive samples at the output of CLPDI are uncorrelated. In addition, the uncertainty region, i.e., the number of cells to be tested in the acquisition, is decreased relative
to the pure MF acquisition. The number of cells in the uncertainty region is now
reduced to Cm ¼ Cu=m. In addition, the number of potentially correct cells to be tested
is reduced, as can be seen by comparing Figures 4.9 and 4.10.
In time hopping systems, the MF collects the samples together according to delays
between consecutive chips (pulses as presented in Chapter 3), i.e., the spreading code
sets the delays between consecutive ‘1’s in the impulse response. Therefore, the MF
waits until the chips (pulses) arrive in predetermined time slots inside frames.
nTc
MF
n(mTc)
sync
r(t)
CLPDI(m)
Th
Figure 4.8
MF with CLPDI
UWB Theory and Applications
78
600
500
400
300
200
100
0
0
50
100
150
200
250
300
350
400
450
Figure 4.9 MF output without noise as function of time in channel model 1, number of
pulses in bit is N ¼ 10 and number of pulse position in frame is M ¼ 16
1400
1200
1000
800
600
400
200
0
0
50
100
150
200
250
300
350
400
450
Figure 4.10 CLPDI output without noise as function of time in channel model 1, N ¼ 10,
M ¼ 16 and m ¼ 4
4.3.2
Performance Measures
The specifications of satisfactory performance measures for synchronization will
depend on the particular application. The main issue is the time that elapses between
the time the synchronization starts and the time of acquisition. There are two basic
scenarios—when there is an absolute time limit and when there is no absolute time
limit. If there is no time limit, the most interesting parameter is the mean acquisition
Receiver Structures
79
time, TMA ¼ EfTacq g and possibly the variance of the acquisition time 2Tacq . This is
the case when a data or pilot signal is always present, that is, when the link operates
continuously (Polydoros, 1982). The mean acquisition time is defined as the expected
value of the time that elapses between the initiation of acquisition and the completion of acquisition. If there is a time limit, the parameters are the acquisition time
Tacq and the time limit Ts , If there is a time limit, a better performance measure is
the probability of prompt acquisition PrfTacq Ts g, which is called overall probability of detection Pov
d . The complement of the probability of prompt acquisition is
ov
the overall probability of missing the code Pov
m ¼ 1 Pd . Missing the code occurs in
systems where data transmission starts after a certain time interval Ts from the initial
system start up, i.e., burst communication (Polydoros, 1982).
The acquisition time, and therefore also TMA and Pov
d , are functions of several
parameters. The most important of these are the probability of detection Pd and the
probability of false alarm Pfa . The probability of detection is the probability that the
decision is the ‘correct cell’ when the correct cell of the uncertainty region is being tested.
The probability of false alarm is the probability that the decision is the ‘correct cell’
when a false cell is being tested. If the decision is based on threshold comparison,
threshold Th plays an important role concerning Tacq because both Pd and Pfa are
functions of this threshold. The other parameters are the time spent evaluating one
decision variable Ti , the penalty time Tfa , and the number of cells in the uncertainty
region Cu.
In this context, mean acquisition time is used as a performance measure. It can be
calculated for the DS CLPDI case as (Iinatti and Latva-aho, 2001):
L=m TMA ¼
PM
LTc þ ðN LÞðTc þ Tfa Pfa Þ
L=m
1 PM
L=m1
P
NTc þ ðN LÞðTc þ Tfa Pfa Þ
iPiM
i¼0
þ
N
L=m1
P
i¼0
þ
þ mTc
ð4:9Þ
PiM =m
ðN LÞðN L þ mÞðTc þ Tfa Pfa Þ
þ NTc
2N
where the uncertainty region N is the length of the code, no a-priori information of the
correct code phase exists at the beginning of acquisition, Pfa is the probability of false
alarm at the output of the CLPDI, PM ¼ 1 Pd where Pd is the probability of detection
in the correct code phase, Tfa ¼ Kp Tc is the penalty time caused by a false alarm, Tc is
chip interval.
In a multipath propagation environment without the CLPDI block, the TMA is evaluated with a similar equation (Ramirez-Mireles and Scholtz, 1998a).
UWB Theory and Applications
80
4.3.2.1
Performance of CLPDI
Figure 4.11 presents the TH UWB simulation model with matched filter synchronization. The figure also describes the outputs from different parts of the system.
Chip correlation (point A in Figure 4.11) is performed by correlating the received
signal with the template waveform which has the same shape as the received signal (after
the antenna). The pulse waveform in the transmitter in the simulations is a Gaussian
pulse. Therefore, the reference pulse in the receiver is the second derivative of the
Gaussian pulse because of the effect of the transmitting and receiving antenna
(Ramirez-Mireles and Scholtz, 1998a). Only one sample per Tc is used so as to reduce
the complexity in the simulations. This can be done when the pulse width ðTp Þ is smaller
than chip length ðTc Þ, and when chip synchronism is assumed. The output of the
chip correlator is a sequence that contains the information about the pulse locations
within the received sequence in every frame. The chip correlator output is fed into a
code matched filter (CMF). The filter is matched to the whole spreading code (point
B in Figure 4.11). In the TH system this means, that delays between consecutive ‘1’s
in impulse response is not fixed, but instead depends on the code phase (as seen
in Figure 4.12. After the filter, a threshold comparison is used (point C in
Figure 4.11). Figure 4.12 presents an example of the output of code matched filter, as
well as describing the correct synchronization points.
In the CLPDI simulations, the CLPDI block is inserted after the code matched filter
in the block diagram of the TH-UWB synchronizer (Figure 4.11).
4.3.2.2
AWGN Channel Performance
We will first consider a channel with static multipath. For a static channel, the equal
power path case represents the worst scenario (Iinatti and Latva-aho, 2001). One sample
per chip is used to achieve lower simulation times, and chip synchronism is assumed.
We use a code length N ¼ 80 and penalty time Tfa ¼ 100Tb . A constant false alarm rate
Code_CRR
sqn_com
Tr
sig_rx
bsrx
Tc
A
awn
C
B
Ctemplate0
Code Synchronization
CODE
MF
CHIP
CORRELATOR
filtro
Th
PROBABILITY
FUNCTION
41
Uindex_slot(1)
length of data bit in Tc
Pd_t
Pd Pfa_t Pfa
Figure 4.11 Simulation model for TH-UWB synchronizer in AWGN channel
Receiver Structures
81
Code_CRR
Th
1
(1 – q)
21
= sychronization point
N Tc slots
Figure 4.12
Output of the code matched filter when six bits are transmitted
criterion (CFAR) is used when setting the threshold of the comparator ðPfa ¼ 102 Þ.
The Th was found before simulations, so that the required Pfa was obtained for each
SNR value. Simulations were performed over 100 bits, i.e., the number of the correct
code phases is 100L and the number of false code phases is 100ð80 LÞ.
Figure 4.13 presents TMA performance of the synchronizer with and without CLPDI
for two channels, and Figure 4.14 presents TMA performance of the synchronizer with
and without CLPDI for four-path channels. In a four-path channel, m in CLPDI
350
MF, 1- path
MF, 2 - paths, NO CLPDI
CLPDI(2), 2 - paths
300
TMA/Tb
250
200
150
100
50
0
2
3
4
5
6
7
8
9
10
Eb/N0 [dB]
Figure 4.13 TMA with and without CLPDI, N ¼ 80, L ¼ 2, Tb ¼ NTc , Tfa ¼ 100, Pfa ¼ 102
UWB Theory and Applications
82
2000
MF, 1- path
MF, 4 - paths, NO CLPDI
CLPDI(4), 4 - paths
CLPDI(2), 4 - paths
1800
1600
1400
Tma/Tb
1200
1000
800
600
400
200
0
2
3
4
5
6
Eb/N0 (dB)
7
8
9
10
Figure 4.14 TMA with and without CLPDI, N ¼ 80, L ¼ 4, Tb ¼ NTc , Tfa ¼ 100,
Pfa ¼ 102
is either 2 or 4, Pd and Pfa are obtained by simulation, and TMA is calculated from
equation (4.9) utilizing the uncertainty region of TH system. TMA is also presented for a
one-path channel. In TH-UWB systems, CLPDI is an improvement over conventional
MF synchronizers because there are fewer total cells to be examined, and because of the
increased Pd in correct cells. In a four-path channel, performance is improved by using
both m values 2 and 4. The behaviour is similar to WCDMA systems (Iinatti and
Latva-aho, 2001).
4.3.2.3
Performance in Saleh–Valenzuela Channels
There is a large number of resolvable paths in UWB systems, more than 100, because of
the fine path resolution of the UWB channel. The following simulation results only
consider the most powerful paths for the code acquisition. The number of cells or chip
positions on which the receiver can be considered synchronized is defined by the 20
strongest paths in the MF case. When m-CLPDI is used, we need to make sure that the
20 synchronization positions defined in the MF case are included in the synchronization
cells for the m-CLPDI algorithm. Since it is possible that the 20 synchronization cells
are non consecutive, the number of synchronization cells in the m-CLPDI case can be
bigger than 20/m. Once one of the synchronization cells is found, a new acquisition
process starts. Verification extra-time is added for each false alarm event occurred
during the acquisition process.
Receiver Structures
83
The simulations consider four modified Saleh–Valenzuela channel models presented
in Chapter 2. The pulse width Tp is fixed to 1 ns. The code length is 160, the number
of frames per bit is N ¼ 10, the number of possible pulse position in each frame is
M ¼ 16, the penalty time in the case of false alarm is Tfa ¼ 100Tb and the required
Pfa ¼ 0:01. The synchronization process is performed over a data packet of 30 bits
and the channel is static over the each packet. A new data block is transmitted every
time the synchronization is reached or the data packet has been explored without
synchronization.
Figures 4.15 to 4.18 present the mean acquisition time of the code MF for m ¼ 4, 10
and 20 and the CLPDI block versus SNR for each of the Saleh–Valenzuela channel
models. SNR is defined as the Eb =N0 at the receiver side. Eb is the collective bit energy
spread over all the paths in the channel. Using CLPDI, the reduced code acquisition
time is evident for all the channel models. The mean acquisition time increases in
absolute time from both MF and CLPDI passing from the CM1 to the CM4. The
reason is the reduced energy borne by the 20 strongest paths whilst moving from CM1
towards CM4. Otherwise, passing from the CM1 to the CM4 increases CLPDI’s
performance improvement relative to the MF case. As the channel moves from the
CM1 to the CM4, the channel gets closer to the situation of having paths with equal
average power that gives the best result, as has been shown for data detection using
diversity techniques (Proakis, 1995). In CM1, the extra gain achieved using m ¼ 20 is
not remarkable. This is because of the reduced CM1 delay spread relative to the other
channel models.
Mean acquisition time, Pfa = 0.01, CM = 1
500
MF
clpdi4
clpdi10
clpdi20
450
400
350
Tma/Tb
300
250
200
150
100
50
0
6
7
8
9
10
11
12
Eb/N0(dB)
13
14
15
16
Figure 4.15 Tma =Tb as a function of Eb =N0 in channel model 1
UWB Theory and Applications
84
Mean acquisition time, Pfa = 0.01, CM = 2
500
MF
clpdi4
clpdi10
clpdi20
450
400
350
250
200
150
100
50
0
6
7
Figure 4.16
8
9
10
11
12
Eb/N0(dB)
13
14
15
16
Tma =Tb as a function of Eb =N0 in channel model 2
Mean acquisition time, Pfa = 0.01, CM = 3
500
MF
clpdi4
clpdi10
clpdi20
450
400
350
300
Tma/Tb
Tma/Tb
300
250
200
150
100
50
0
6
Figure 4.17
7
8
9
10
11
12
Eb/N0(dB)
13
14
15
16
Tma =Tb as a function of Eb =N0 in channel model 3
Receiver Structures
85
Mean time code acquisition, Pfa = 0.01, CM = 4
500
MF
clpdi4
clpdi10
clpdi20
450
400
350
Tma/Tb
300
250
200
150
100
50
0
6
7
8
9
10
11
12
13
14
15
16
Eb/N0(dB)
Figure 4.18 Tma =Tb as a function Eb =N0 in channel model 4
4.4 Conclusions
This chapter examined the concepts behind some of the most common UWB receivers.
The chapter described the A-rake, S-rake and P-rake and analysed their relative performance. The rake combines the many multipath components in the multipath-rich
channel. The complexity of the A-rake makes it impractical for many applications. The
S-rake and P-rake combine only a subset of the resolvable multipath components.
Whilst being much less complex, the S-rake and P-rake are not able to combine all
the received signal energy and so have substantially worse performance than the A-rake.
The S-rake requires channel estimation to select the most significant multipath components. This added complexity greatly improves the performance compared with the
P-rake, especially in non-LOS channels.
This chapter also applied a chip-level post-detection integration code synchronization
method to a time-hopping UWB system. The method was used in order to utilize dense
multipath propagation in a UWB environment during the code acquisition process. The
method increases the probability of detection of the burst of multipath components, i.e.,
it decreases the mean acquisition time for finding the existence of multipath profile. The
simulation results indicate that the method improves the acquisition performance, and
therefore is usable in a UWB communication system.
5
Integrated Circuit Topologies
Sakari Tiuraniemi, Ian Oppermann
5.1 Introduction
UWB is a spread spectrum technology. However, UWB differs from conventional
spread spectrum technologies in that UWB transmits information through short pulses
or a ‘chirped’ signal rather than transmitting information on a modulated continuous
carrier signal. The information is typically superimposed by a pulse modulation
method. The bandwidth occupied by an UWB system is far greater than in traditional
spread spectrum systems. A UWB data communication system is often referred to
as impulse radio (IR). Typically, impulse radio utilizes time hopping pulse position
modulation (TH-PPM) scheme (see Appendix 1).
UWB systems have some significant advantages over conventional spread spectrum
systems. First, the transceiver has a relatively simple structure as some of the functional
parts that increase the complexity of traditional radio systems are not necessary. This
reduces the required human resources and financial investment. Second, the low transmission power means that less power is consumed. Third, the large bandwidth makes
the detection of the signal quite difficult for unintended receivers. Fourth, the achieved
bit rate of more than a few hundreds of Mbps is significantly more than conventional
spread spectrum systems.
In this chapter, some of the architectures used in UWB data communication systems
are introduced and investigated, based on the information set out in the current available literature.
In particular, we will examine the pulse generator, a critical part of the transmitter,
which generates the transmitted waveform. The transmitted waveform must satisfy the
frequency mask defined by the Federal Communications Commission (FCC) (Federal
UWB Theory and Applications Edited by I. Oppermann, M. Hämäläinen and J. Iinatti
Ó 2004 John Wiley & Sons, Ltd ISBN: 0-470-86917-8
UWB Theory and Applications
88
Communications Commission, 2002b) or satisfy other applicable local radiation regulations. One of the most common waveforms is the first derivative of the Gaussian
pulse, also referred to as the monocycle.
We will also examine the correlation-based receiver. A conventional SNR maximizing
based correlator receiver requires a template waveform that exactly matches the
received waveform. Such a distorted and delayed waveform is difficult to generate.
The multiplier and integrator of the correlating circuit must be extremely fast to process
the short time domain pulses. UWB receiver structures often ignore or coarsely approximate the pulse shape, which significantly reduces the complexity and required speed of
operation, even though it also reduces the received SNR. This approach seems to be
quite popular amongst current receiver designs (for example, see Lang, 2003) and is a
recommended approach.
5.2 Ultra Wideband Basic Architectures
Impulse radio is a UWB digital data communication system for low power, low range,
applications typically utilizing TH-PPM. At the block level, the transmitter is very
simple. It consists of a pulse generator and a digital timing circuit that controls the
timing of transmission. These blocks are presented in Figure 5.1.
The block marked by is the timing circuit that is responsible for PPM and PN coding.
It provides a timing signal, or a trigger, for the pulse generator. In some presentations the
timing circuit is replaced by a programmable delay (Withington, 2004).
The clock oscillator determines the pulse repetition frequency, PRF, of the system. It
can be either a crystal oscillator or a custom designed oscillator, the first being the most
feasible in high frequency systems such as impulse radio. The pulse generator produces
the desired waveform. The pulse generation can be realized in a number of ways.
One of the great benefits of a UWB transmitter relative to continuous-wave
transmitters is that there is no need for complex circuits such as the power amplifier
(depending on the application) and frequency synthesizer, which contain circuits such as
the phase-locked loop (PLL), voltage controlled oscillator (VCO) and mixers (Taylor,
1995; Foerster et al., 2001). These are the most complex components of conventional
transmitters, and these components make conventional transmitters relatively difficult
and expensive to design and implement. In contrast, an UWB transmitter is relatively
data
pulse generator
clock oscillator,
fPRF
τ
PN code
Figure 5.1
Transmitter top level schematic
Integrated Circuit Topologies
89
inexpensive and moderately easy to design and implement because the UWB transmitter
does not have these components.
The optimal receiver for signals transmitted over an AWGN channel is a correlation
or a matched filter receiver, since it maximizes the SNR (Proakis, 1995). The receiver
examined in this chapter consists of a low-noise amplifier (LNA), a correlation circuit
and a circuit to provide the template waveform for the correlation. These blocks are
presented in Figure 5.2.
After the received signal is amplified, it is correlated with the template waveform. The
output of the correlation is processed by a bit decision circuit that decides which bits are
carried by the received pulses. The baseband signal processing circuit is responsible for
the bit decision.
To maximize the processing gain and SNR, the template waveform should be the
same as that of the received signal. Such a signal is difficult to generate in practice, since
the transmitted pulse is distorted by the antennas and the channel. The need to make the
template waveform the same as that of the received signal also makes the receiving
circuit more complex. One way to avoid this complexity is to approximate the template
waveform by using the transmitted pulse, or to make very coarse approximations such
as a rectangular pulse. It is also possible to ignore the template waveform altogether and
rely on the pulse shaping caused by the finite bandwidth of the transmitting and
receiving antennas (Lang, 2003).
In an ideal environment, the received pulse shape is the second derivative of the signal
transmitted (third derivative of the Gaussian pulse) since both antennas act as differentiators. Figure 5.3 shows the autocorrelation functions of the first and third derivatives of the Gaussian pulse as well as the cross-correlation between the first and third
Gaussian pulses. The amplitude of the correlation result between the first and the third
derivatives is 80 % of the autocorrelation of the received waveform. This means that
using the transmitted waveform as the template reduces the performance of the correlator by less than 1 dB.
The correlation circuit consists of an integrator, and a multiplier that multiplies the
received signal with the template waveform. The result of the multiplier is integrated
correlation circuitry
LNA
baseband signal
processing
dt
template waveform
Figure 5.2
Receiver top level schematic
UWB Theory and Applications
90
1
1st derivative autocorr.
1st and 3rd derivatives
3rd derivative autocorr.
0.8
0.6
0.4
0.2
0
–0.2
–0.4
–0.6
–0.8
0
20
Figure 5.3
40
60
80
100
120
140
160
Correlation results of transmitted and received signals
over the bit duration to maximize the received signal power and to minimize the noise
component. Having a train of pulses to integrate over, the correlated signal is raised
from the noise and the possible signals of other users. From this it can be seen that the
more pulses there are in a pulse train, the better SNR is attained, since more correlated
energy is put into each symbol.
In addition to providing an accurate synchronization, performing the correlation at
the required speed is probably the biggest issue. Both the multiplier and integrator must
be fast enough to process each pulse.
5.3 Review of Existing UWB Technologies
This section reports some of the architectures used in UWB communication applications, with a particular focus on the pulse or waveform generation and its implementation. This section also considers the used data modulation scheme and the correlation
receiver. These topics, combined with the transceiver architecture and antennas, form
the basis for the communication system.
The information is gathered from patents and the publications listed in the references.
The architectures are chosen to give examples of the possible structures used in UWB
communications, with a particular focus on the pulse or waveform generation and
implementation.
Integrated Circuit Topologies
5.3.1
91
Time Domain Corporation: PulsOn Technology
Time Domain Corporation’s PulsOn technology (Time Domain Corporation, 2003)
is presented in Figure 5.4. This, or a variant of it, is the most common architecture used
in UWB communication systems. A pulse generator generates the waveform (e.g. a
Gaussian monocycle), and the waveform is then provided to the transmitting antenna.
The pulse transmitting time is controlled by a programmable time delay (PTD), which
uses the signal coming from the clock oscillator to create a timing signal for the pulse
generator by means of pulse position modulation. The programmable time delay gets its
control signal from the modulator and code generator. The modulator modulates the
incoming data using PPM or an additional modulation scheme, and the code generator
gives an individual pseudo-random (PR) code for the modulated data. Power amplification is not required because of the low transmit power (below the noise level).
The receiver is based on the correlation technique. The correlation technique
is the optimal technique for a signal of such power since it maximizes the SNR
(Proakis, 1995). The received signal is multiplied with a template waveform generated in the receiver, and the result is then integrated over several periods of the
received pulse train. The correlator converts the received signal directly into a
baseband signal, which may then be further processed using baseband techniques
to improve performance. The signal processor also provides acquisition and tracking control to the PTD.
Transmitter
Receiver
Correlator
Multiplier
Pulse
Generator
Pulse Generator
Code
Generator
Programmable
Time Delay
Programmable
Time Delay
Modulation
Clock
Oscillator
Figure 5.4
website)
Integrator
S/H
Baseband Signal
Processing
Acquisition & Tracking
Control
Data Out
Code Generator
Data In
Clock Oscillator
PulsOn transceiver top level schematic (From Time Domain Corporation
UWB Theory and Applications
92
5.3.2
Time Domain Corporation: Sub-Carrier Technology
The Time Domain Corporation patent also includes another architecture developed by
the company (Time Domain Corporation, 2003) (see Figure 5.5). The architecture’s
basic structure is the same as Time Domain Corporation’s original PulsOn technology,
except that the second architecture utilizes sub-carrier technology to realize an additional modulation scheme.
The time base in Figure 5.5 provides a clock oscillator, which again functions as the
PRF source. The code time modulator and code source create a coded timing signal
from the PRF source. In addition, the information source is fed to a sub-carrier
generator and modulator block, which creates a sub-carrier signal (i.e. a piece of sine
wave) and modulates it with the information signal to create a modulated sub-carrier
signal. A sub-carrier time modulator then mixes the resultant modulated sub-carrier
signal with the previously created coded timing signal from the PRF source, and
provides the result to an output stage as its time controlling signal.
This technique provides further possibilities for channel coding and signal modulation.
For example, by using different sub-carrier frequencies, more channels can be utilized or
different information can be transmitted simultaneously. If different sub-carriers are used,
the receiver must separate the carriers from the received signal. This may be achieved by
band-pass filtering for example.
5.3.3
MultiSpectral Solutions, Inc.
MultiSpectral Solutions introduces three schemes to generate pulses or bursts of pulses
in their patent (Parkway, 2001). They have divided the three pulse generation schemes
into two classes. The first class consists of two types of pulse generation scheme. The
first pulse generation scheme, presented in Figure 5.6, utilizes an impulse generator and
a mixer (switch) to chop the signal coming from an oscillator, thereby providing a train
of bursts to the subsequent circuitry.
time
base
code time
modulator
sub-carrier time
modulator
code
source
sub-carrier
generator and
modulator
output
stage
information
source
Figure 5.5 Transmitter utilizing sub carrier technology
Integrated Circuit Topologies
93
frequency/phase
control
mixer
digitally
controlled
attenuator
bandpass
filter
to gated power amplifier
or antenna
oscillator
low-level
impulse
generator
band-pass or
impulse shaping
filter
Figure 5.6
First class of transmitter
bandpass or
impulse shaping
filter
low-level
impulse
generator
Figure 5.7
to gated power amplifier
or antenna
Special case of the first class of transmitter
The second pulse generator scheme in the first class is shown in Figure 5.7.
A bandpass or a pulse shaping filter is directly excited by a low level impulse, so that
a mixer and oscillator are not needed. This is functionally equivalent to MultiSpectral
Solutions’s first pulse generation technique in the special case of having a zero oscillator
frequency (i.e., d.c. source).
The band-pass or pulse shaping filter (set out in Figures 5.6 and 5.7) shapes the
incoming impulse, and thus provides the wanted centre frequency or bandwidth for the
transmitted signal. This could also be achieved by adjusting the width of the impulse.
The transceiver is less complex since there is no need to generate the derivative. The
trade-off is that the circuit is relatively more expensive, since very short impulses require
fast switches and digital signal processors to deal with the fast calculations.
MultiSpectral Solutions’s second class consists of the third type of pulse generator
scheme. It is shown in Figure 5.8. The impulse generator and filter combination in
amplitude
control
frequency/phase
control
time
gating
circuit
bandpass
filter
digitally
controlled
attenuator
oscillator
Figure 5.8
Second class of transmitter
to gated power amplifier
or antenna
UWB Theory and Applications
94
MultiSpectral Solutions’s first class of pulse generator schemes is replaced with a timing
circuit. The timing circuit functions as a switch providing a short time frame for the
oscillating signal to propagate through it. The response can be described as an amplitude modulation of the oscillating signal. The amplitude is defined by the ramp generated by the gating circuit.
5.3.4
XtremeSpectrum Inc.: Trinity
XtremeSpectrum’s transceiver architecture is covered in a 2001 patent (McCorkle, 2001)
and consists of an interface and three parts: transmitter, receiver and radio controller.
The basic idea of the transceiver is the same as in the patented architectures in Sections
5.3.2. and 5.3.3 and will not be analysed further.
The waveform or pulse generation in this architecture is interesting. The idea is to
produce two short pulses, which are half the length of the desired monocycle pulse.
These short pulses, S1 and S2 in Figure 5.9, are then combined by a Gilbert cell, which is
used as a differential mixer. As a result, the mixer produces the monocycles presented in
Figure 5.9 which is redraw from (McCorkle, 2001). The polarization of the monocycle
depends on the data bit A, which is the other input of the Gilbert cell. When A is low,
the monocycle begins with negative amplitude, and when A is high, the monocycle
begins with positive amplitude. The Gilbert cell will be introduced in a later section.
In XtremeSpectrum’s patent, the Gilbert cell functions as a modulator. The short
pulses are multiplied either by 1 or 1 depending on the data bit ‘A’. An example of the
BPAM modulation attained by the mixing function at the Gilbert cell is persented in
patent (McCorkle, 2001).
S1
S2
w1
w2
Figure 5.9
Ideal pulses and waveforms
Integrated Circuit Topologies
5.3.4.1
95
Pulse Generation by Avalanche Transistor
In Morgan (1994), an avalanche transistor is used to generate a monocycle pulse. The
pulse generation is based on operating the transistor in avalanche mode, which requires
a high voltage. High-voltage solutions are not covered in this book, and therefore this
method is introduced for completeness only.
The bias circuit of the pulse generator (shown as a dashed line in Figure 5.10) provides
a fixed bias voltage across the capacitor C7. This voltage level is about 100–130 V,
which is near the transistor’s (TR3) avalanche breakdown voltage. This voltage is
also provided across charging capacitor C8, which dumps a large charge through
transistor TR3 when the avalanche breakdown voltage is reached. This is achieved by
the pulse position modulated message signal, which is fed to the base of the avalanche
transistor.
As the avalanche breakdown takes place, the emitter voltage of TR3 rises dramatically for a short time (10 ns), and then falls to negative equivalent voltage due to the
inductance connected between the emitter and earth. After the charging capacitor C8
has dumped its charge, it takes a few tens of microseconds to attain the level required for
a new avalanche break down. This pulse generation method is feasible for applications
with high voltage levels and monocycles with relatively large length, such as RADAR.
5.3.5
Coplanar Waveguides
A recent pulse generation method (Lee et al., 2001a) is based on step recovery diode
(SRD), Schottky diode and charging and discharging circuitry (see Figure 5.11). The
SRD provides an impulse, which is high-pass filtered in a RC-circuit. The result is a
Gaussian-like pulse, which is fed to a pair of transmission lines. The generated pulse is
divided in two and propagates in both branches after the capacitor C. The first half of
the pulse propagates directly to the load resistor, and the other half of the pulse
propagates to the short. The transmission lines are designed to have such a length that
the propagation delay of the second half of the pulse (the one propagating to the short)
is equal to the length of the pulse. The pulse will be inverted when the pulse reflects from
the short circuit in the end of the transmission line. The resulting pulse seen across the
PPM
+V
TR3
C7
C8
CLK
Figure 5.10 Transmitter driver including the pulse generator
UWB Theory and Applications
96
LO
RL = 50Ω
R
SRD
C
Schottky Diode
Section A
Section B
+Z
+X
–X
–Z
LA
LB
Figure 5.11 Monocycle generator
load is the superposition of the two branches. The pulse width is controlled by the SRD.
The voltage across the load is shown in Figure 5.12.
Another embodiment of this development has been introduced (Lee et al., 2001b).
The basic idea is the same as in (Lee et al., 2001a), except that there is an additional
MESFET as an amplifying unit.
33.7330 ns
35.3580 ns
A
37.0630 ns
C
B
Ch :
= 500.0 mvolts/div
Time base = 333 ps/div 3452
Start
= 32.7798 ns
Figure 5.12
Stop = 33.0668 ns
Offset = –22.50 mvolts
Delay = 33.7330 ns
Delta T = 287.0 ps
Monocycle generated by the circuit shown in Figure 5.11
Integrated Circuit Topologies
97
5.4 Integrated Circuit Topologies
This section introduces the basic circuit topologies used to implement the transceiver on
an integrated circuit (IC), and briefly introduces high-speed integrated circuit processes
and designs.
The basic design flow of high-speed circuits is presented in Figure 5.13 (Häkkinen,
2002). The target specifications are set by customer or system-level requirements. The
designer’s task is to produce a circuit meeting these specifications after a suitable
technology has been chosen.
The design process typically means iterating through the steps in Figure 5.13 several
times before the customer or system specifications are met. The most common iterative
operation is that of optimizing the circuit performance to meet the speed, distortion and
matching specifications. This may be very challenging in portable devices because of the
rigid power consumption requirements.
In radio frequencies (RF), the IC design is especially difficult if the IC process is not
optimized for RF. This especially concerns CMOS which is optimized for digital circuits.
In addition, the processes that are suitable for analogue design are pushed to their limits,
since the maximum operating frequency is close to the signal frequencies (Häkkinen, 2002).
The circuits are strongly sensitive to component parasitic and model errors at high
frequencies and near the limits of the process. This makes the designer’s task even more
difficult, since the designer has to rely on inaccurate device and parasitic models. The
designer’s expertise becomes very important.
RF circuits are also very sensitive to the layout design and the packaging that is used.
Layout parasitic (such as capacitance and inductance) cause mismatch, signal crosstalk
and changes in circuit response. In addition, individual circuits may disturb each other
Target Specification
Circuit Principles
Technology
Circuit Diagram
Circuit Simulation
Transistor
Models
Intrinsic circuit design
especially optimization of
resistances and transistors
Line Models
(on chip)
Layout (optim. placement)
metallization,
parasitics, crosstalk
Models for
mounting
parasitics
Checking the complete IC
including all parasitics
Figure 5.13 Basic steps in the design of high-speed circuits. For simplification the iteration
loops between the different steps are not shown
UWB Theory and Applications
98
complementary signals
with symmetrical transients
lower voltage swing
across the load resistors
steeper pulse edges
reduced crosstalk and
instability problems
simpler and more
reliable design
Figure 5.14
simpler mounting
technique
higher data rates at
improved eye diagr.
no reference voltage
for the CSs required
reduced time jitter
reduced power
consumption
Advantages of differential operation (compared with single-ended operation)
through crosstalk, inductive cross coupling and noise induced to the internal power
lines. These effects can be minimized by careful layout design and yield analysis. Again,
the designer’s own expertise is important. Further discussion of layout design and
packaging is set out in Häkkinen (2002).
Due to these issues , the designer has to use careful design techniques and topologies. In
RF circuits, it is crucial to have as little interference as possible, especially for wideband
circuits, in which the interference is more likely to appear in the signal band. One widely
used design practice is to use differential signals and structures. Using differential signals
instead of single-ended signals minimizes the effect of a number of external and internal
interfering signals, such as noise or variations of the power source. Some advantages of
using differential structures are presented in Figure 5.14 (Häkkinen, 2002).
Differential circuits are required for processing differential signals. Hence all topologies examined in this chapter are built using differential pairs, except for current
source and source follower. We will consider the differential-pair topologies first.
The topologies presented in this section are CMOS circuits. All these circuits may also
be realized by other technologies. However, the equations presented are not necessarily
applicable to non-MOS technologies. The equations for other types of active components can be found in the literature (Gray and Meyer, 1993; Johns and Martin, 1997).
5.4.1
Source Coupled Pair
The differential pair (i.e. emitter/source coupled pair or transconductance stage) is one
of the basic and most common building blocks used in integrated circuits. Differential
pairs can be used to process differential signals, and these circuits can be coupled to each
other easily, for example, without the use of coupling capacitors. They may be used as
amplifiers as well as an input stage or buffers (Gray and Meyer, 1993).
Figure 5.15 presents a source coupled pair and its input voltage versus output current
curve (Babanezhad and Temes, 1985). The transistors are assumed to be identical and
operating in the saturation region in all the schematics presented in this chapter.
Integrated Circuit Topologies
ID1
99
ID2
I = ID1 – ID2
ISS
M1
+
M2
–(ISS/k)1/2
Vi
Vi
–
(ISS/k)1/2
–ISS
ISS
(a)
(b)
Figure 5.15 Source coupled pair (a) and its input voltage vs. output current curve (b)
The operation of the transconductance stage can be illustrated with the (low frequency) small signal model (based on the T-model, Figure 5.16) (Johns and Martin,
1997), from which the output current can be determined as a function of input voltage.
Defining the differential input voltage as Vin ¼ Viþ Vi , the small signal drain
current is determined by (Johns and Martin, 1997)
id1 ¼ is1 ¼
Vin
Vin
¼ rs1 þ rs2 1 gm1 þ 1 gm2
ð5:1Þ
where id1 is the small signal drain current, is1 is the small signal current through resistor,
rs1 ; rs2 are the source resistances, Vin is the differential input voltage and gm1 ; gm2 are the
transconductance values of transistors M1 and M2.
id1 = is 1
id 2 = is 2
vi+
vi–
rs 2
is1
rs1
is 2
vi+ – vi–
rs 1 + r s 2
Figure 5.16 Small signal model of source coupled pair
UWB Theory and Applications
100
When the two transistors, M1 and M2, are identical, the transconductance is
gm ¼ gm1 ¼ gm2 which gives equations (Johns and Martin, 1997)
id1 ¼ is1 ¼
id2 ¼ id1
gm
Vin
2
gm
¼ Vin :
2
ð5:2Þ
ð5:3Þ
By defining the differential output current as iout ¼ id1 id2 , we have the following
iout ¼ gm Vin :
ð5:4Þ
The gain of the transconductor is defined by its transconductance gm
iout
¼ gm
Vin
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
W
gm ¼ 2n Cox ID :
L
A¼
ð5:5Þ
The corresponding large signal equations may be derived (Babanezhad and Temes, 1985)
ID ¼ kðVGS VT Þ2
ð5:6Þ
n Cox W
2 L
ð5:7Þ
k¼
where ID is the drain current, VGS is the gate-to-source voltage, VT is the threshold
voltage, k is the transconductance parameter, n is the mobility of electrons near silicon
surface, Cox is the gate capacitance per unit area, W is the width of the transistor, and
L is the length of the transistor.
Equation (5.6) is the simplified MOS square law characteristic in the saturation region.
The equation is used to determine the large signal drain currents and the differential
output current of the source coupled pair (Babanezhad and Temes, 1985)
ID1
k
¼
2
!2
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ISS Vi2 Vi
þ pffiffiffi
k
2
2
ð5:8Þ
ID2
k
¼
2
!2
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ISS Vi2 Vi
pffiffiffi
k
2
2
ð5:9Þ
Iout ¼ ID1 ID2
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2ISS
Vi2
¼ kVi
k
ð5:10Þ
Integrated Circuit Topologies
101
where ID1 ; ID2 are the drain currents of transistors M1 and M2, ISS is the current of tail
current source, Vi is the differential input voltage, and Iout is the differential output
current.
Equations (5.8)–(5.10) are valid if the input voltage is
rffiffiffiffiffiffiffiffiffi
rffiffiffiffiffiffiffi
ISS
ISS
:
Vi k
k
ð5:11Þ
This is because of the non-linearity of the input pair in all differential pairs, which sets
limits for the dynamic range of the input. The dynamics of the inputs may be improved
by using various linearization techniques. Some linearization techniques are presented
in the literature (Häkkinen, 2002; Gray and Meyer, 1993; Johns and Martin, 1997;
Babanezhad and Temes, 1985; Gill et al., 1994; Gilbert, 1968). The large signal equations are useful for example when determining the output current of a Gilbert cell.
The frequency response of a source coupled pair needs to be determined in order to
optimize the speed of a single transconductor to make sure that the circuit is capable of
operating in the required frequencies. An additional frequency analysis needs to be
performed for the circuits that are built from these differential pairs, such as for example
the Gilbert cell,. This additional frequency analysis is usually made by determining the
dominant pole (or zero) from the transfer function of the circuit. The dominant pole
usually determines the 3 dB frequency, and hence the maximum operating frequency
of the circuit.
The transfer function is defined by determining the node equation of the circuit from
the equivalent circuit. The determination of the poles and zeros may also be performed
simply by finding the pole (or zero) frequency of each of the nodes of the circuit, in case
the equivalent circuit is difficult to built (as is often the case in differential structures).
This is done by using the following equation (Gray and Meyer, 1993)
fn ¼
1
2ðrn Cn Þ
ð5:12Þ
where fn is the frequency of the node, rn is the resistance of the node, and Cn is the
capacitance of the node.
The pole, or zero, that has the lowest frequency is the dominant one (Gray and Meyer,
1993). The nodes, for example the nodes in the source coupled pair in Figure 5.15, are
the nodes on the input (gates) and output (drains). This kind approximation gives a
reasonable estimate of the frequency response. The actual frequency response of a
circuit is found by an AC analysis.
The unity-gain frequency of an individual transistor may be determined (Johns and
Martin, 1997)
fT ¼
gm
2ðCgd þ Cgs Þ
ð5:13Þ
where fT is the unity gain frequency, Cgd is the gate to drain capacitance, and Cgs is the
gate to source capacitance.
UWB Theory and Applications
102
As a rule of thumb, the maximum oscillating frequency of a circuitry is usually twice
the fT which determines the maximum peak frequency.
5.4.2
The Gilbert Multiplier
The Gilbert multiplier (Gilbert, 1968) is probably the most common multiplier structure
used in integrated telecommunication circuits today. The basic Gilbert cell is presented
in Figure 5.17. It is a double balanced (i.e. fully differential) multiplier, so it is highly
immune to certain disturbances such as even order harmonics and common mode noise
(Häkkinen, 2002).
In the following, the behaviour of the Gilbert cell is derived from the (large signal) drain
currents of a single differential pair (Babanezhad and Temes, 1985; Gill et al., 1994).
The differential output current can be expressed as (Babanezhad and Temes, 1985)
Iout ¼ ID7 ID8 ¼ ðID3 þ ID5 Þ ðID4 þ ID6 Þ
ð5:14Þ
¼ ðID3 ID4 Þ ðID6 ID5 Þ:
ðID3 ID4 Þ and ðID6 ID5 Þ are the differential currents of the upper source coupled
pairs M3-M6, which form the so called mixer core (Gilbert, 1997). The differential
currents can be calculated using (5.10). Note the signs of the currents in the latter term.
This is due to the inverse polarity of the input voltage of the left side differential pair.
ID7
M3
+
ID8
M4
M5
VX
–
M1
+
M2
VY
–
ISS
Figure 5.17 Gilbert multiplier
M6
Integrated Circuit Topologies
103
The tail current sources for these two differential pairs are the drain currents of the
lower differential pair, which can be calculated using (5.8) and (5.9). This results in
(Babanezhad and Temes, 1985)
Iout
2vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
!2
u rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
ISS VY2 VY
6t
þ pffiffiffi VX2 ¼ kVX 4
k
2
2
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi3
!2
u rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
ISS VY2 VY
7
t
pffiffiffi VX2 5 ð5:15Þ
k
2
2
where VX ; VY are the input voltages of upper and lower differential pairs.
It can be seen from equation (5.15) that there is a non-linear relationship between the
input voltages. By limiting the input voltage swing, the output can be approximated to
be linear
Iout
2vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 3
!2 u rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
!2
u rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
u
ISS VY2 VY
ISS VY2 VY 7
6t
t
þ pffiffiffi pffiffiffi 5
¼ kVX 4
k
2
k
2
2
2
) Iout ¼
ð5:16Þ
pffiffiffi
2kVX VY :
The dynamics of the input of a Gilbert cell may also be improved by using various
linearization techniques. Some linearization techniques are presented elsewhere
(Häkkinen, 2002; Gray and Meyer, 1993; Johns and Martin, 1997; Babanezhad and
Temes, 1985; Gill et al., 1994; Gilber, 1968).
When using Gilbert multipliers, the RF input is usually connected to the lower
differential pair (M1-M2 in Figure 5.17), and the LO input is usually connected to the
upper differential pair (mixer core) which acts as switches. The RF input is supposed to
be differential. In case of a single-ended input, the lower differential pair may be
replaced by a circuit that converts the single-ended voltage to differential current. One
example of this kind of circuit is presented by Gilbert (1997). It is called a bisymmetric
class AB input stage, or a micro-mixer. Figure 5.18 presents a CMOS variant of this
circuit.
The single-ended RF input voltage is a.c. coupled to the diode coupled transistor (M01)
on the left side. The input voltage causes an a.c. current through transistor M1 and M01.
This current is inversely copied to the right side branch (M2 and M02). The operation
point is set by the bias voltage provided to the gates of transistors M1 and M2.
Since the transconductance of the input stage is one term of (5.16), output current of
the multiplier, the performance of the input stage is essential and has therefore been
extensively studied to find better voltage to current converters. Another reason for
developing new input stages is the previously mentioned non-linearity of the input pair
of any differential circuit. Some of these results are briefly discussed by Häkkinen (2002).
5.4.3
Analogue Addition/Subtraction
Differential pairs can also be used to build addition and subtraction circuits. Figure 5.19
presents a schematic for analogue subtraction. The difference compared with addition
UWB Theory and Applications
104
ID1
ID2
M1
M2
M01
M02
Vbias
+
VRF
–
Figure 5.18 Bisymmetric class AB input stage in CMOS
ID5
+
M1
ID6
M2
V1
–
+
M3
M4
V2
–
Iss
Iss
Figure 5.19 Analogue subtraction circuit
is the coupling of the drains. This kind of cross-coupling is fairly usual in differential
circuits.
Notice the similarity of the analogue subtraction circuit to the multiplier core of the
Gilbert cell. However, there are two differences. First, the drains are coupled differently.
This can be seen in the last term of
Integrated Circuit Topologies
105
Iout ¼ ID5 ID6 ¼ ðID1 þ ID4 Þ ðID2 þ ID3 Þ
ð5:17Þ
¼ ðID1 ID2 Þ ðID3 ID4 Þ:
Second, the input voltage is different for each of the two differential pairs. These
voltages are added to, or subtracted from, each other. This can be seen from
Iout
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2ISS
2ISS
2
¼ kV1
V1 kV2
V22
k
k
Iout
) Iout
rffiffiffiffiffiffiffiffiffi
rffiffiffiffiffiffiffiffiffi
2ISS
2ISS
¼ kV1
kV2
k
k
pffiffiffiffiffiffiffiffiffiffiffiffi
¼ ðV1 V2 Þ 2kISS :
ð5:18Þ
ð5:19Þ
Equation (5.19) is valid in the linear region, where (Babanezhad and Temes, 1985)
V12 ; V22 <<
2ISS
:
k
ð5:20Þ
As we can see from (5.19), the output current has a linear relationship to the difference
of the input voltages. This is when (5.20) is satisfied. The use of linearization techniques
may also increase the dynamic range of the input.
Equations (5.17)–(5.19) are the result of the same approach used to derive
(5.14)–(5.16) for the Gilbert cell in Babanezhad and Temes (1985).
5.4.4
Integrator
The first stage of the integrator used in the implementation of the correlator is a basic
structure in which a capacitor is used for the integration of current. A source coupled
pair is again used as an input stage. The drains of the differential pair are coupled with a
parallel capacitor as seen in Figure 5.20 (on the left side of the schematic) (Khorramabadi and Gray, 1984). This kind of current integrator may be referred to as a fully
differential Gm-C integrator, in which the source coupled pair is used as the transconductor (Gm-stage). Gm-C integrators are discussed in more detail elsewhere (Johns and
Martin, 1997; Khorramabadi and Gray, 1984).
The voltage difference across the integrating capacitor is first reset by closing the
switches. After a short while, when the voltage across the capacitor has reached the zero
value (i.e., when both ends of the capacitor are at the same potential), the switches are
opened. As the current flows through the transistors, the potentials of the drains
change. If the input voltages at the transistors are different, the change of potentials
in the ends of the capacitor is different in size. As a result, a voltage difference is
produced across the capacitor. The voltage increases in time, and thus integrates the
difference of the drain currents. The integration operation can be illustrated by (Johns
and Martin, 1997)
UWB Theory and Applications
106
Vint ¼
!ti ¼
Iout
Gm Vi
¼
sCint
sCint
ð5:21Þ
Gm
Cint
ð5:22Þ
where Vint is the output voltage of the integrator, Iout is the output current of the
transconductor, Cint is the capacitance of the integrating capacitor, !ti is the unity gain
frequency of the integrator, and s is the Laplace operator.
Only the voltage difference across the integrating capacitor is fed to an integrate and
hold circuit, I/H (on the right side in Figure 5.20). The voltage difference is fed by an
operational amplifier which acts as a switched capacitor (SC) inverting integrator
(Johns and Martin, 1997). The operational amplifier must have sufficiently large bandwidth to process the output of the first integrator.
The reason for using two integrators is that the first integrator is fast enough to
integrate the short pulses, but it only has a limited capacity, and therefore needs to be
reset after a certain time of integration. The second inverting integrator based on an
operational amplifier is too slow for the fast pulses, but fast enough to integrate the
output of the first integrator after, for example, 10 pulses. The first integrating capacitor
is reset again after the sampling. This procedure is repeated until the whole pulse train
(symbol) has been integrated.
The differential voltage at the output of the inverting integrator is fed to an analogue
comparator circuit which makes the bit decision. An inverter may be used after the
comparator since the integration result is inverted. Both integrators are reset after providing the voltage to the comparator. The integration of the next data symbol may then begin.
ISS/2
ISS/2
reset
reset
Cint2
Cint
+
V
– int
–
sample
M1 M2
+
+
Vi
–
sample
Cint2
ISS
Gm-C integrator
Figure 5.20
Integrate and Hold circuit
Integrator schematic
+
–
Vout
Integrated Circuit Topologies
5.4.5
107
Current Source
Current sources are used to provide bias current for the circuits designed. A current
mirror is one good way to provide bias from a global current source. The global current
source has a fixed value, but the individual circuits may need more or less current. With
a current mirror, the amount of current may be adjusted by changing the W/L ratio
(Gray and Meyer, 1993).
Figure 5.21 presents a simple current mirror. The d.c. current fed to the transistor on
the left side is ‘mirrored’ to the right side. The ratio of the currents on the right and left
side is approximated using (5.2) and (5.3) (Gray and Meyer, 1993) as
W1 L1
ID1
¼ :
ID2
W2 L2
ð5:23Þ
In practical implementations, the output impedance of the current source should be
increased to be as large as possible. This makes it perform more like an ideal current
source which does not act as a load to the circuit. Hence, the output impedance has no
impact on the a.c. signals, and the a.c. signals themselves have no impact on the current
level of the current source. This results in a more stable operation.
Some realizations of current sources with larger output impedance are presented in
Figure 5.22. More information on these circuits can be found in the literature (Gray and
Meyer, 1993; Gregorian and Temes, 1986; Johns and Martin, 1997).
5.5 IC Processes
There have been significant developments in RF IC processes during recent decades.
The size of the devices and the voltage levels of power supplies have decreased, whilst
the speed of the transistors has increased. The transit frequency (unity gain frequency)
has increased to several tens of GHz from the 10 GHz silicon bipolar technology
ID1
ID2
VD1
VD2
M1
M2
Figure 5.21 MOS current source
UWB Theory and Applications
108
ID1
M1
(a)
ID1
ID2
ID2
ID1
ID2
M3
M4
M3
M4
M3
M2
M1
M2
M1
M2
(b)
(c)
Figure 5.22 MOS current mirrors: (a) Wilson’s current source, (b) improved Wilson’s
current source, and (c) cascade current source
prevalent in the 1980s. The upper limit of the transit frequency of a silicon homojunction bipolar transistor (for 3 V operation) is reported to be around 30 GHz for the
fastest bipolar transistor today (Li, 2002).
The pure silicon bipolar process has decreased in popularity due to its low level of
integration as compared with BiCMOS processes. BiCMOS combines bipolar and
CMOS technologies. Digital circuits can now be integrated in CMOS and RF/analogue
circuits in bipolar transistors on the same chip. In that sense, BiCMOS has the highest
integration level of all IC processes, and is one of the best candidates for future RF IC
processes (Li, 2002).
New technologies were developed to increase the achievable transit frequency. One of
these new technologies is heterojunction bipolar technology (HBT), which is fabricated,
for example, in a silicon germanium (SiGe) bipolar process. The transit frequency
increases up to 50 GHz by using germanium in the base of the transistor,. Just like
silicon, SiGe processes are fully compatible with mainstream CMOS process. This
compatibility also allows the BiCMOS to be fabricated in SiGe. SiGe is the most
commonly used RF technology at present (Li, 2002).
Another class of high transit frequency technologies is compound semiconductors
(Li, 2002). The most commonly known (and used) compound semiconductors are III–V
compounds (III and V are columns of the periodic table of elements), which are, for
example, GaAs (gallium arsenide), GaP (gallium phosphide), InP (indium phosphide),
InSb (indium antimonide) and GaN (gallium nitride). These semiconductors have
physical properties such as large and direct band gap and high electron mobility. These
properties make compound semiconductors suitable for high frequency applications
and MMICs (microwave/millimeter-wave IC). However, compound semiconductors
have a few disadvantages such as low yield and low integration level. The most
commonly known GaAs processes used in RF applications are GaAS HBT and GaAS
MESFET (metal semiconductor field effect transistor).
There is still much debate about which of these IC processes is the optimum choice.
All of the IC processes have advantages and disadvantages. Each of the IC processes
Integrated Circuit Topologies
Table 5.1
109
Typical figures of merit of various RF IC technologies
Si BJT/
BiCMOS
Feature size
Peak fT
Peak fmax
Minimum
NF@2 GHz
1/f noise
corner
Breakdown
SiGe BJT/
BiCMOS
GaAs
HBT
GaAs
MESFET
Si
CMOS
0:6 m
27 GHz
37 GHz
1.0 dB
0:25 m
47 GHz
65 GHz
0.5 dB
2:0 m
50 GHz
70 GHz
1.5 dB
0:5 m
30 GHz
60 GHz
0.3 dB
0:25 m
40 GHz
50 GHz
1:5 dB
100–1000 Hz
100–1000 Hz
1–10 kHz
10 MHz
1 MHz
3.8 V
3.35 V
15 V
10 V
3–5 V
can be evaluated based on their properties such as transit frequency, maximum power
gain frequency, minimum noise figure, 1/f noise corner frequency, maximum power
added efficiency, linearity and reliability. Some key properties of IC processes typically
used in RF applications are compared in Table 5.1 (Li, 2002).
Silicon technologies are commonly selected as they cost less and have a higher
integration level. CMOS is the most available process and the dominant digital process.
CMOS has higher yield and therefore lower cost per die. That gives CMOS an advantage over BiCMOS. The selective cost issues are discussed by Li (2002).
BiCMOS has advantages such as better device matching, highly accurate/predictable
analogue characteristics, larger gain, and lower 1/f noise corner frequency. In addition,
the operating life of a bipolar transistor is longer than that of a MOS (except for digital
MOS circuits, which act as switches which are inactive for most of the time).
One of the operational advantages of MOS transistors is linearity, which is an
important property in designing mixers or other applications. A bipolar transmitter
needs emitter degeneration to achieve similar linear behaviour. Emitter degeneration
requires the use of inductors or resistors, which results in decreased gain and use of
larger silicon area. In UWB applications, an inductor may not be desirable, since an
inductor is likely to cause a resonant frequency inside the signal bandwidth that could,
for example, cause the generated pulse to oscillate.
5.6 Example Implementation
The UWB transceiver designed and discussed in this section is implemented in 0.35
micrometre CMOS. CMOS is probably the most desirable IC process, since it has the
lowest cost and the best availability. CMOS is also dominant in the area of digital
circuits. It is possible to avoid using multiple chips and processes by using CMOS in
both RF and analogue design. As mentioned in Section 5.5, the circuits may also be
implemented in other IC processes. Some of these other IC processes (e.g. BiCMOS)
may also provide larger transit frequency which may result in shorter pulses.
In our example, the target pulse repetition frequency is 20 M pulses/second, and the
target pulse length is less than 500 ps. This will not necessarily produce a system which is
UWB Theory and Applications
110
FCC compliant. However, it will serve to demonstrate the main functional blocks and
design approach for the UWB transceiver.
The transceiver implemented is part of a TH-PPM UWB system. The UWB signal is
generated by a monocycle waveform generator. The generated monocycle is used for
both the transmitter and receiver. Although PPM is the selected modulation scheme,
BPAM is also considered in the design process.
5.6.1
Transceiver
As mentioned in Section 5.2, the architecture of the transceiver is relatively simple. The
most complex part, the digital timing circuit, includes both the modulator and the pseudorandom noise coding. Together, these provide the timing for the pulse generation
(see Appendix 1) and transmission. In addition, the timing circuit may be responsible
for the synchronization, which increases its complexity. The synchronization is assumed
ideal. The timing circuit is presented only at a schematic level and is not used in the
simulations.
The transceiver components covered by this section are shown in Figures 5.24 and
5.25. They are the pulse generator (monocycle waveform generator) and the correlator
at the receiver, which includes a multiplier and an integrator. Any amplifiers required in
the receiver or in the transmitter are not presented. Amplifiers (or attenuators) may be
needed between some circuits to provide the desired voltage swing. As an example, an
attenuator is required after the pulse generator to fix the magnitude of the pulse in order
to keep the transmitted power below the limits set by the FCC (2002). A low noise
amplifier (LNA) is required at the front end of the receiver to amplify the received signal
for the correlator, and as an operational amplifier in the integrator.
The receiver architecture has some differences in the cases of PPM and BPAM. The
receiver in Figure 5.2 is for the BPAM and Figure 5.23 shows the implementation for the
PPM. The PPM receiver is implemented from two correlators, which are synchronized
correlation circuitry
LNA
baseband signal
processing
dt
Template
Timing signal
τ
dt
Figure 5.23
The PPM receiver architecture
Integrated Circuit Topologies
111
to two different time windows. The first correlator detects the ‘early’ signal, which
represents ‘0’, and the second correlator detects the ‘late’ pulse, which represents ‘1’. The
timing for the correlators is provided by a timing circuit which is similar to the timing
circuit in the transmitter. The correlator is synchronized to the received signal. The
timing signal is delayed by the time of the PPM modulation index ðÞ for the second
correlator, which is therefore synchronized to the late pulse. The bit decision is made by
determining which of the correlators produces the largest output.
The BPAM receiver has only one correlator. This correlator is synchronized to the only
time window set by the PR code. The bit decision is made from the output of the
correlator. The output of the correlator depends on the polarity of the transmitted pulse.
5.6.2
Pulse Generator
One method of producing a monocycle is to sum up two pulses with different polarities
with a time delay corresponding to the length of the pulses. This may be done in a number
of different ways. One method was introduced in an earlier Section 5.4 where a transmission line technique was utilised (Lee et al., 2001a). Another technique was introduced in
Section 5.3.4 in which two pulses with a time delay were combined in a multiplier so that
one of the pulses was multiplied by 1 and the other pulse was multiplied by 1 (McCorkle,
2001). By changing the multiplier’s control signals, the phase of the produced monocycle
may be changed 180 degrees, and the multiplier acts as a binary pulse amplitude modulator.
In the case of PPM, the modulator structure in McCorkle’s work (2001) is not
required, so the circuit may be simplified. Such a circuit is presented here (Tiuraniemi,
2002). The presented pulse generator subtracts two differential pulses with a time delay
corresponding to the length of the pulses from each other, thus producing a differential
monocycle. This is done by an analogue subtraction circuit which is much simpler than
any multiplier. BPAM may also be utilized even though the advantages of this pulse
generator mainly concern a PPM system.
The top level schematic of the pulse generator is presented in Figure 5.24. The
numbers in brackets indicate the waveforms in different phases of the pulse generation
shown in Figure 5.25.
Two short pulses are generated with a digital pulse generation circuit presented in
Figure 5.26 and Figure 5.27. The delay between the two pulses is realised by a simple
delay element consisting of two inverters and a PMOS varactor, which provides the
opportunity to tune the length of the delay. These short pulses are then fed as inputs to a
single-ended to differential converter presented in Figure 5.28. The resulting differential
pulses are fed to the inputs of two emitter or source coupled pairs of the analogue
subtraction circuit (see Figure 5.29). The outputs of these differential pairs are crosscoupled so that a monocycle is produced.
Figures 5.26 and 5.27 present two methods of generating short pulses. The first
method utilizes an XOR gate and the second method utilizes an AND gate. The XOR
operates so that whenever both input signals are at different logical levels, the output is
at a high level, i.e. logical ‘1’, and when both input signals have the same logical levels
the output is at low level, i.e., logical ‘0’ (Daniels, 1996). The length of the high-level
output can be adjusted by the phase difference between the inputs. This is also provided
by a simple delay element.
UWB Theory and Applications
112
PN coded
timing signal
digital pulse generation
a
b
τ
τ
analog monocycle
waveform generation
(3)
(1)
XORz
S_to_D
(4)
(2)
(5)
(6)
a
b
XORz
S_to_D
(4b)
τ
single-ended to
differential
conversion
Figure 5.24 Top level schematic of pulse generator
(1)
(2)
(3)
=TPRF
(4)
(4b)
(5)
(6)
Figure 5.25 Waveforms in different phases of monocycle generation presented in
Figure 5.24
The AND-gate operates so that whenever both the inputs are at a high level, the output
is at a high level (Daniels, 1996). An inverter is connected to one of the inputs of the AND
gate, and a clock signal is connected to both the inverter and the AND gate. The output
of the AND gate is now at low level at all times, but the output still reacts to the rising
edge of the clock. This is because the inverter has a delay during which its output is still at
Integrated Circuit Topologies
113
τ
PN coded
timing signal
(1)
(1)
(3)
(2)
(2)
(3)
Figure 5.26 Digital pulse generation by an XOR-gate
PN coded
timing signal
(1)
(1)
(3)
(2)
(2)
(3)
Figure 5.27
Digital pulse generation by an AND-gate
ID1
ID2
M1
M2
M01
M02
Vbias
+
VRF
–
Figure 5.28
Single-ended to differential converter
zero while the clock has already risen. This effect is called a glitch, or a hazard, which
results in a short pulse with length corresponding to the length of the inverter’s delay. The
delay can be adjusted by connecting a varactor from the output of the inverter to ground.
If the delay is very short, the resulting pulse might not have time to rise all the way up to
the logical ‘‘1’’, but it is still useful for generating a short pulse. When using this circuit, the
UWB Theory and Applications
114
PRF decreases to half that when using the XOR-circuit, since the short pulse will occur
only on the rising edge of the clock. The XOR reacts to both the rising and falling edge of
the clock. This effect can also be seen from Figure 5.25.
A NAND-gate may also be used. The operation is the same as in case of an AND
gate, the difference being that the pulse will occur on the falling edge of the clock and
the pulse is from high to low level. The NAND-gate is faster than the AND-gate and is
therefore used in the implementation of the transceiver. The shape of the pulse is defined
by the RC time constant of the used gate in all of the three circuits.
Two pulses separated by a controllable time delay are produced by combining two of
these digital pulse generators, with the input to one being delayed. These two pulses are
fed to the single-ended to differential converter to produce two differential pulses. One
example of such a converter is a transconductance stage, which inversely copies the AC
current (caused by the AC-coupled input voltage) from the left branch to the right
branch (see Figure 5.28). The two currents are then turned into voltage swings by the
resistors. This circuit is a variant of a BJT micro-mixer (Gilbert, 1997) and was
introduced in Section 5.4.2.
The two short differential pulses that have been generated are now fed to the differential pairs, which are presented in Figure 5.29. The outputs of the differential pairs
are cross coupled so that a differential monocycle (see Figure 5.25 for the theoretical
waveforms) is produced as a result of linear subtraction operation.
The differential pairs may consist of bipolar transistors or FETs. The main difference
between these two types of transistor is the operating speed, which is supposedly faster
for a bipolar transistor. Other transistors such as HEMT and HBT can also be used.
The load impedance of the differential pairs consists of resistors or an active PMOS
load. The active load increases the complexity and the number of transistors, but
decreases the used chip area. Active loads are commonly used in CMOS designs. One
ID6
ID5
+
M1 M2
Vi1
+
–
+
–
Vout
M3 M4
Vi2
–
Ibias
M5
M6
M7
M8
M9
Figure 5.29 Pulse generator’s analogue subtraction circuit
Integrated Circuit Topologies
115
advantage of active loads is that a very large impedance may be realized using a small
area of silicon. An active load is not necessarily suitable in a high frequency application,
since it is known to have poor high frequency characteristics. Therefore, resistive loads
are used in this example. Using an inductor may not be feasible since it may lead to the
existence of a resonant frequency inside the signal bandwidth, which may make the
circuit oscillate.
5.6.3
The Analogue Correlator
The multiplier of the analogue correlator used in this implementation is the Gilbert
multiplier cell. The choice is justified by the benefits explained in Section 5.4.2. Another
reason is the possible implementation of the BPAM in which only the polarity of the
received signal needs to be detected. Since a Gilbert cell is a four-quadrant multiplier,
i.e. the result of the multiplying may be 1 (opposite polarization), þ1 (equal polarization) or zero (no correlation), the Gilbert cell may be used to detect different polarities.
The integrator is realized by a simple differential pair with a capacitor connected
parallel to the drains. The voltage across the capacitor increases as the result of the
multiplication is fed to the integrator. Every time a pulse is received, the voltage across
the capacitor increases in proportion to the input of the differential pair. The polarity of
the voltage depends on the polarity of the product of the multiplication. The polarity of
the product of the multiplication depends on the polarity of the received pulse. This
correlator, presented in Figure 5.30, is therefore compatible in both cases of BPAM and
PPM. In PPM, two of these correlators are used so that the first correlator detects the
early pulses and the other correlator detects late pulses.
After the Gm-C integrator, an integrate and hold (I/H) circuit is used to improve the
performance of the integrator. The I/H circuit is realized by a discrete time inverting
integrator built from an operational amplifier.
Iss/2
ID7
Iss/2
ID8
reset
reset
Cint2
Cint
+
–
Vx
sample
M7 M8
+
–
+ V
int
–
M5 M6
M3 M4
M1
M2
+
+
–
–
sample
Vy
Ibias
Ibias2
M9
M10
Figure 5.30
M11
M12
CMOS correlator structure
Cint2
Vout
UWB Theory and Applications
116
The monocycle of the transmitter is used as the template waveform in the multiplier.
Even though the received waveform has been distorted from the transmitted monocycle,
the correlation still produces useful results. The correlation result is not as good as it
would be in the case of autocorrelation (in which the signal is multiplied by itself). The
mismatch decreases the SNR of the correlating receiver however the performance is still
adequate.
5.6.4
Timing Circuit
A timing circuit provides the trigger signal (or timing signal) for the pulse generator.
The timing circuit is responsible for the PN coding and modulation in the case of PPM.
The top level schematic of the timing circuit is presented in Figure 5.31.
The PN code is provided by a pseudo-random sequence generator, which may consist,
for example, of D flip-flops and an XNOR-gate (Daniels, 1996). This is presented in
Figure 5.32. The operation is such that the generator is given an initial state, which may
be any state except all ‘1’s. The generator steps in pseudo-random order from the initial
f = 1/TC
n-bit
counter
2n – 1
divider
comparator
PTD
timing signal
PPM + PN
data
f = 1/TPRF
Figure 5.31
PN code
generator
Top level schematic of timing circuit for generating TH-PPM signals
D Q
D Q
D Q
D Q
CLK
Figure 5.32 Pseudo-random sequence generator
Integrated Circuit Topologies
Table 5.2
117
Results of the processing gain calculations for the implementation
TPRF
TC
T
NU
NS
PG
50 ns
1.6 ns
0.8 ns
31 (5 bits)
200
41 dB
state through all the states, excluding the ‘all 1’ state, as these states cannot be
generated by the loop. In addition, the loop cannot continue from the ‘all 1’ state due
to the operation of the XNOR-gate. Therefore the loop may not be initiated in that
state.
Each time frame set by the PRF generates a new code. The number of states is
determined from the number of memory elements (i.e. the number of bits, n) to give a
sequence of length 2n 1 bits. The number of states corresponds to the maximum
number of unaliased users operating simultaneously in the network.
Any synchronous counter may be used as the counter in 5.31. The counter steps
through all the states starting from zero. When the counter reaches the last state, it
recommences from the beginning. The time spent for one cycle (count through all the
states) equals the length of the time frame. The number of states also matches the
maximum number of users.
The digital comparator in Figure 5.31 compares the PR code to each of the states of
the counter. A TRUE (i.e. logical ‘1’) signal is produced whenever the PR code and the
states of the counter match. The TRUE signal occurs once in every time frame set by the
PRF. These pulses form the PR coded timing signal.
In case of PPM, this timing signal is fed to a programmable time delay (PTD). The
PTD either delays the signal by T in case of ‘1’, or lets it through non-delayed in case
of ‘0’. This results in a TH-PPM signal which is provided to the pulse generation circuit
as the timing signal. There is no need for the PTD in the case of BPAM.
The same kind of timing circuit, without the PTD, may also be used in the receiver to
provide timing for the correlator. However, the delay experienced in the medium
between the antennas must be taken into account.
The processing gain calculations given in Table 5.2 provide results that are adequate
for this implementation. The results are presented in Table 5.2. The length of the PR
code, time frames, slots and delays (in seconds) and the pulse repetition frequency are
achieved from these results.
5.7 Simulation Results
In this section, the simulation results are presented and discussed. First, the functionality of the pulse generator and correlation receiver is confirmed by a transient analysis.
Second, a corner analysis is run for both the transmitter and receiver to evaluate the
performance of the system with respect to variations in process, temperature and power
supply. A commercially available simulator, Spectre, is used for the simulations. Spectre
is a Cadence version of the SPICE circuit simulator. The top level schematics of the
simulated circuits are depicted in Appendices 3–10.
UWB Theory and Applications
118
The effects of each of the antennas and the channel are not taken into account. This
means that the transmitted signal is received without distortion or delay. The performance of the correlator is more readily evaluated in such an arrangement. The correlation
between the template waveform (transmitted waveform) and the received waveform was
presented in Figure 5.3.
5.7.1
Transmitter
At this point, the timing circuit was not implemented and the timing for the pulse
generation was realized with a clock signal. The simulations of modulation and PN
coding are left for future work. The results of transient and corner analysis of the
transmitter (Appendix 3) are presented below.
The length and shape of the generated monocycle primarily depends on the digital
pulse generation (Appendix 4) circuit which produces the short pulse utilized in the
monocycle generation. The length of this short pulse is approximately one half of the
length of the monocycle, and may be adjusted by a PMOS varactor. However, the
varactor should be left out to minimize the delay, so that the shortest possible pulse
width is achieved. The shape of the short pulses is nearly Gaussian (or half of a sine
period) due to the nature of the glitch. Mathematically, the generated monocycle is not
equal to the first derivative of the Gaussian pulse. However the waveform may be
considered as a reasonable approximation to the first derivative.
The minimum pulse width is set by the delay of the inverter, assuming the NANDgate does not introduce additional delay. The use of a NAND-gate in the circuit leads to
faster operation compared with the use of AND- or XOR-gates. As seen in Figure 5.33,
the delay between the clock signal and the inverted clock signal is about 250 ps, and the
pulse width is about 280 ps. From this it can be seen that the NAND-gate has a small
effect on the pulse width. The pulse width of 280 ps is near the minimum pulse width
with this kind of a circuit in this process (CMOS 0:35 m). Shorter delays must be
generated for shorter pulses. This may be achieved by having more inverters of different
speed connected parallel to take advantage of the differences of the delays. Another
approach is to have faster NAND-gates and inverters available in another process.
Figure 5.34 shows the result of single-ended to differential conversion. The pulse width
is not affected by the circuit. The differential pulses generated with the single-ended to
differential conversion circuit (Appendix 5) can be connected directly to the input of the
pulse generator. However, an attenuator can be useful to adjust the amplitudes of the
input signals, since the peak-to-peak voltage of the differential pulses exceeds the dynamic
range of the input pair of the pulse generator. By attenuating the pulses, the pulse
generator operates in the linear range.
The monocycle generation is presented in Figures 5.35 and 5.36. The first figure
shows the input and output signals as single-ended signals to give a clear idea how the
monocycle waveform generation circuit (Appendix 6) works. The input signals are
marked as negative and positive pulses. By cross-coupling these pulses, two monocycles
with 180-degree phase difference are generated. The monocycles are not identical due to
some differences in the amplitudes of the negative and positive pulses of the output.
However, this is not a problem in a differential structure, since it is the difference of the
single-ended outputs that matters.
Integrated Circuit Topologies
Figure 5.33 Digital pulse generation
Figure 5.34 Single-ended to differential conversion
119
UWB Theory and Applications
120
Figure 5.35 Monocycle generation (single-ended signals)
Figure 5.36
Monocycle generation (differential signals)
Integrated Circuit Topologies
121
The differential inputs and output are depicted in Figure 5.36. Here the cross coupling
may be presented as a difference between two differential pulses, as seen in the upper
curves. As a result, a monocycle is generated. The pulse width of the monocycle is
approximately 650 ps. This pulse width indicates that some additional delay is introduced by the analogue pulse generator. By optimizing the circuit design, the pulse
generator may have a higher operating frequency and therefore would not introduce
additional delay. The delay between the two branches providing the input for the
waveform generator also affects the final pulse width. The pulse width may be further
shortened by minimizing this delay.
The pulse width corresponds to a centre frequency of about 1.54 GHz. The transmitted signal is not the same since it is differentiated due to the effect of the antenna.
The method of producing the monocycles is not perfect. A pulse is intended to be
produced only on rising (or falling) edge. In Figure 5.37, which presents a pulse train,
small imperfections can be seen between the monocycles. These are the undesirable
products of the digital pulse generation that occur when the clock signals change states.
By increasing the delay, the short pulse (glitch) increases in both amplitude and width,
and the undesirable product decreases. In other words, the problem is solved by lowering the operational frequency. The problem is also solved by using the XOR-gate in
the digital pulse generation, since every change in the clock signals is utilized. The XORgate would also increase the pulse width due to its slow operation. Such problems are
common in digital circuits that are pushed to their limits. This behaviour could also be
avoided by having faster digital cells in the process. Another way to remove these
undesirable effects is to use some kind of windowing technique to cut them off.
Figure 5.37 Monocycle pulse train
122
UWB Theory and Applications
The waveform after the transmitting antenna is depicted in Figure 5.38. It is a
derivative of the generated monocycle, since the antenna acts as a differentiator. The
differentiation is performed by the calculation tool in Cadence. The waveform may be
compared to the third derivative of the Gaussian pulse.
The power spectral density (PSD) of the transmitted signal (i.e., after the transmitting
antenna) is presented in Figure 5.39. The PSD was also produced using the calculation
tool in Cadence. The amplitude is normalized to 0 dBm. More important information is
the bandwidth and centre frequency. The centre frequency is 2.4 GHz, and the 3 dB
bandwidth is 1.8 GHz. This does not satisfy the FCC frequency mask (Fig 1.1). The
pulse width may be shortened and the FCC mask satisfied by using a faster IC process
to obtain reduced delays and higher operational frequency. The spectrum of the signal
may also be modified by filtering. This would lead to the desired spectrum, but the
waveform would be distorted.
In the high frequencies, from 5 to 8 GHz, a second maximum is seen. This is
supposedly caused by the undesirable products of the digital circuitry. The power
level of this spurious signal is more than 10 dB below that of the desired signal, so it
may be considered a non-harmful component. In any case, the second maximum is
placed inside the FCC’s frequency mask (41:25 dBm=MHz at 3.1–10.6 GHz). Even
if the spectrum is shifted 2 GHz to the right on the frequency axis (which would
make the spectrum FCC compliant), the power level of the unwanted side band is
low enough to satisfy requirements.
Figure 5.38 The waveform after the antenna
Integrated Circuit Topologies
123
Power Spectrol Density
Transient Response
:
PSD in the medium
0.0
(dBm)
–10
–20
–30
–40
0.0
1.0G
2.0G
3.0G
4.0G
5.0G
6.0G
7.0G
8.0G
Figure 5.39 Power spectral density of the pulse train after antenna
From the presented results, some characteristic figures may be calculated using
equations presented in Chapter 3 symbol time and rate are
TS ¼NS Tf ¼ 200 50 ns ¼ 10 s
1
¼ 100 kbits=s:
RS ¼
TS
Symbol rate could easily be increased by transmitting a smaller number of pulses per
symbol or having a smaller duty cycle, which may be achieved by increasing the Tf . The
trade-off is a decreased processing gain, PG, which is now
PG ¼ 10 log NS þ 10 log
Tf
¼ 23 dB þ 19 dB ¼ 42 dB
Tp
The fractional bandwidth of the transmitted signal is
Bf ¼ 2
fH fL
3:3 1:5
¼ 0:75 > 0:20
¼2
3:3 þ 1:5
fH þ fL
from which it is seen that it may be said to be an UWB signal. The 3 dB bandwidth of
the UWB signal is 1.8 GHz. The bandwidth expansion factor is as high as
Be ¼
B
1:8 GHz
¼
¼ 18 000 ¼ 43 dB;
R 100 kbit=s
UWB Theory and Applications
124
which is a large value. However, it is clearly expected bearing in mind that the data rate
of the designed system is sufficiently low compared with the possibilities of a UWB
system. A 2Mbps UMTS system has an expansion factor of 2:5 ¼ 4 dB. The UWB
system of this chapter would have an expansion factor of 900 ¼ 30 dB for a bit rate of
2 Mbps.
5.7.2
Receiver
The receiver (Appendix 7) was also simulated using the above-mentioned analysis. The
timing and synchronization for the correlator was provided by the transmitted pulse
train, which was used as both the received signal and template waveform (autocorrelation). The results of transient and corner analysis are presented in the following sections.
Figures 5.40 and 5.41 depict the inputs and outputs of the Gilbert multiplier (Appendix 8). The polarity of the received signal (RF input) is different in the two figures. As
may be seen, the output of the multiplier depends on the polarity of the received signal.
When the polarity of the received signal is the same as the polarity of the template
waveform, the output is positive, and when the polarity of the received signal is opposite
to the polarity of the template waveform, the output is negative. This also applies to the
output of the correlator.
The first integrator integrates every received pulse. As seen in Figure 5.42, the voltage
across the integrating capacitor increases by a certain amount every 50 nanoseconds,
Figure 5.40
Multiplier’s inputs and output (differential signals)
Integrated Circuit Topologies
Figure 5.41
125
Multiplier’s inputs and output with inverse polarity in RF input
Integrator
Transient Response
20m
: Integrator’s
output
(v)
10m
0.0
–10m
0.0
200n
400n
600n
800n
1.0u
time (s)
Figure 5.42 The output of the first integrator
1.2u
UWB Theory and Applications
126
(v)
Correlator's output
Transient Response
: Correlator output for ‘0’
: Correlator output for ‘1’
800m
700m
600m
500m
400m
300m
200m
100m
0.00m
–100m
–200m
–300m
–400m
–500m
–600m
–700m
–800m
0.0
2.0u
4.0u
6.0u
8.0u
10
time (s)
Figure 5.43 The outputs of the correlator in case BPAM
which is the pulse interval. After integrating a few pulses, the efficiency of the first
integrator starts to decrease. The voltage across the capacitor increases in a normal way,
but it is no longer able to hold its level. To avoid the loss of integrated energy, the voltage is
sampled and then reset. This procedure is repeated until the whole pulse train is integrated.
This takes 10 microseconds (200 pulses times 50 ns) with the used simulation parameters.
The sample is further integrated in the inverting integrator, which also provides some
extra gain. The output of the second integrator is depicted in Figure 5.43. The output
increases after each period of 500 ns (10 pulses). After the symbol is integrated, the
output of the correlator has reached a value that is detectable by semiconductor devices,
in this case a comparator.
The curves in 5.43 represent the outputs in case of ‘0’ and ‘1’ in BPAM. The output
for PPM is the same as ‘0’ for BPAM.
5.8 Conclusions
This chapter has explored some of the fundamental issues that need to be addressed
when attempting to implement an UWB transceiver. Various circuit fundamentals were
examined as were commercial solutions and solutions described in the open literature.
The chapter focuses on the main aspects of the UWB transceiver; the pulse generation
and key aspects of the receiver architectures. Several pulse generation techniques were
explored, including a digital generation technique.
The chapter highlighted some of the properties of the CMOS process which must be
taken into account when an UWB design is fabricated on a wafer. Parasitic components
Integrated Circuit Topologies
127
and other non-ideal characteristics are not modelled with adequate accuracy in typical
commercial CMOS design tools. The parasitic components must be modelled by the
designer, after which the transceiver must be re-simulated. The parasitic components are
likely to have a significant impact at the frequencies of interest. The parasitic components must be avoided by careful design selection. The parasitic components and nonideal characteristics are usually modelled more accurately in bipolar technologies. Some
cases still need to be modelled to verify the practical behaviour. The best way to
determine the behaviour is to measure the fabricated IC.
An example UWB transceiver implementation was shown based on relatively
straightforward CMOS technology. The pulse generation circuit developed was verified
as a very efficient and moderately reliable way to generate a monocycle. The example
UWB transceiver implementation provided an insight into the design issues for an UWB
transmitter, even though the transmitted signal did not satisfy the limits set by the
frequency mask specified by the FCC. There are at least two possible ways to try to
change the spectrum of the transmitted signal in the circuit example. The first way is to
have a shorter pulse, which may be realized by using a different pulse generation
technique, or different IC process. The second way is to shape the generated pulse using
a pulse-shaping filter so that the mask is satisfied. The latter method is much more
promising in practice, however is more likely to reduce the efficiency of the receiver.
One of the practical issues to consider when implementing UWB circuits is the
realization of synchronization. The realization of synchronization may be difficult to
achieve due to the ultra-short pulses and their fast rising and falling times. In this
chapter, the synchronisation was assumed ideal.
The integrator structure developed in the chapter demonstrated excellent performance in integrating fast pulse with fast rising times. The first stage of the integrator is
very sensitive and is able to integrate short pulses with more than adequate results. The
second stage provides additional capacity for the integrator to make long integration
times possible in cases where the received signal has low energy.
6
UWB Antennas
Tommi Matila, Marja Kosamo, Tero Patana, Pekka Jakkula,
Taavi Hirvonen, Ian Oppermann
6.1 Introduction
Whilst often ignored or assumed ideal in conventional narrower band system analysis,
antennas are a critical element in the signal flow of UWB systems. The antenna acts as a
filter for the generated UWB signal, and only allows those signal components that
radiate to be passed. The antenna is often approximated as a differentiator both at the
transmitter and receiver. This is the motivation behind the examination of the various
derivatives of the Gaussian pulse explored in Chapter 3.
The goal of this chapter is to explore antenna structures suitable for use in short
range, low power indoor UWB radio systems, and outdoor ‘base station’ communications. This chapter examines some of the main UWB antenna types (dipoles, bow-tie
and TEM horn), as well as discussing the performance of a small class of antennas that
are suitable for low-cost communications systems. Various practical antenna structures
are examined through simulation and measurements results of several prototypes are
presented. It is outside the scope of this work, however, to undertake a rigorous
theoretical performance evaluation of the antenna types.
6.2 UWB Antenna characteristics
The UWB antenna acts like a filter and is a critical component in UWB radio systems.
The basic effect of antennas is that they produce the derivative of the transmitted or
received pulse waveform (Funk et al., 1995a). This also has the effect of extending the
duration of the transmitted and received pulse. This extension of pulse duration
decreases the time resolution of the system. The antenna has a greater impact in
UWB than in narrower band systems because of the very large bandwidth of an
UWB signal.
UWB Theory and Applications Edited by I. Oppermann, M. Hämäläinen and J. Iinatti
Ó 2004 John Wiley & Sons, Ltd ISBN: 0-470-86917-8
UWB Theory and Applications
130
In antenna terminology, the frequency range demand must be 6:1 or greater in order
to be ‘ultra wide’ (Taylor, 1995) which means that the upper frequency must be at least
six times greater than the lower frequency of the band. For such very wideband
antennas, issues of linearity, radiation efficiency and impedance match across the band
present difficult problems.
One problem which will arise when a very short time domain (im)pulse (implying
large bandwidth) is used to excite the antenna, is the ringing effect. After the antenna,
the signal is no longer impulse like. Instead the pulse is spread in the time domain.
A typical antenna response is presented in Figure 6.1, where the ringing effect is
modelled using a simple Bessel function.
To avoid ringing, resistive antennas with low Q-values should be used. The resistive
loading will cause the unwanted signal component to die away quickly, leaving a pulse
much closer to the desired shape. The antenna bandwidth can also be increased by
making the Q-value small, since the bandwidth is inversely proportional to Q-value.
However, the low Q-value implies that the efficiency of a resistive antenna is generally
quite poor.
The Q-value for an antenna is given by
Q ¼ fo =ð fH fL Þ
where fo, fH, and fL are the centre frequency and the upper and lower 3 dB frequency
values of the antenna respectively.
0.05
0.04
0.03
Wanted pulse
Relative amplitude
0.02
Ringing
0.01
0
–0.01
–0.02
–0.03
–0.04
–0.05
0
50
100
150
200
250
300
Consecutive samples
Figure 6.1
The response of an antenna to impulse excitation showing the ringing effect
UWB Antennas
131
The frequency domain is also useful for describing the transient response of antennas
because the time and frequency domain are connected by the Fourier transform. The
ability of an antenna to preserve the waveform of the ultra-narrow pulse is investigated in
the time domain. Two of the most important time-domain properties of an antenna are
fidelity and symmetry. The fidelity is defined as the maximum cross-correlation of the
normalized incident voltage and the normalized electric field in the far field region
(Montoya and Smith, 1996). The symmetry is a measure of the symmetry of the waveform
in the far field region (Montoya and Smith, 1996). More theoretical approaches on timedomain antenna characterization can be found in the literature (Balanis, 1997; Montoya
and Smith, 1996; Shlivinsky et al., 1993; Allen et al., 1993; Lamensdorf and Susman, 1994).
UWB antennas differ from their narrowband counterparts in one basic concept.
Many antennas, especially in the telecommunication applications, are resonant elements
that are tuned to particular centre frequencies and have relatively narrow bandwidths.
In contrast, UWB antenna designs seek much broader bandwidths and require nonresonating operation.
In the literature, several antenna types have been presented for use in UWB systems.
Applications have used dipoles, log-periodic dipole arrays (LPDA), conical monopole,
spirals, notched, ridged and TEM horns antennas (Taylor, 1995). The main focus in
antenna technology in the existing UWB literature, is for high power radar antennas.
Table 6.1 presents a summary of typical dimensions of different antenna elements
(Taylor, 1995). The remainder of this chapter will explore the practical implementation
aspects of several antenna types.
6.3 Antenna types
In this section, the antenna types suitable for UWB operation are described. In Section 6.3.1
the general requirements for antennas are described. Sections 6.3.2 to 6.3.8 concentrate
on the different antenna types.
6.3.1
General Requirements
This section evaluates the transient response of antennas for base-station and indoorportable use. The short, sub-nanosecond pulses require special antenna structures
compared with typical narrowband systems. An ideal wideband antenna acts like a
high-pass filter, which means that the pulse waveform is differentiated when passing
through the antenna. Typical wideband antennas, such as log-periodic dipole arrays,
change the waveform even more due to dispersion. In general, UWB antennas should be
linear in phase and should have a fixed phase centre. Typical impedance circuits may
not be phase linear. For this reason the antennas should be inherently impedance
matched. The antenna’s radiation characteristics also have a significant impact on the
antenna’s performance. The antenna gain should be smooth across the frequency band
in order to avoid dispersion of the transmitted pulse. The antenna gain will typically
appear different from different angles. This will lead to different pulse shapes depending
on the angle to the receiver.
UWB Theory and Applications
132
Table 6.1
Frequency
Matching
Radiation efficiency (min.)
Directivity (typical)
Fidelity (typical)
Size
Target antenna values
Base station antenna
Portable antenna
2–10 GHz
typical VSWR <1.5
VSWR <2
50%
0–30 dB
>0.7
not specified
3–10 GHz
typical VSWR <2
VSWR <3
10%
0 dB
>0.7
area <100 cm2 on PCB
VSWR is voltage standing wave ratio.
Some target antenna specifications for base station and portable antennas are given in
Table 6.1. These are by no means definitive. However they give an indication of
appropriate parameter values.
6.3.1.1
Base Station Antenna
As discussed in previous chapters, the power of UWB transmitted signals is extremely
low. In this context, ‘base station’ antennas may be used for networks such as high
speed data kiosks or for low data-rate systems, including location and tracking systems.
The base station antenna may be designed for indoor or outdoor usage, depending on
the application. Outdoor usage allows the antenna to be relatively large, with dimensions of several wavelengths. In addition to the requirements outlined in Table 6.1, the
antennas must radiate efficiently. This may restrict the use of resistively loaded structures. Base station antennas may be either directive or omnidirectional. Directional
antennas could be used for example in radio links, whereas omnidirectional antennas
would be more favourable in mobile applications.
6.3.1.2
Portable Antenna
The requirements for the portable, short-range, UWB antenna differ slightly from the base
station antenna. Most importantly, the antenna must be small. It is also highly desirable
that the antenna be low cost and preferably constructed on a printed circuit board. The small
size implies that the antenna is omnidirectional. The radiation efficiency is not as critical
parameter as in base station antennas, which makes it possible to use resistive loading. If
possible, the transceiver will be embedded in the same circuit board as the antenna.
6.3.2
TEM Horn
The TEM horn and its variations are among the most commonly used antennas in
UWB applications. The basic structure consists of two tapered metal plates fed by a
two-wire TEM-mode transmission line. The structure can be considered as a bent bowtie dipole. The TEM horn preserves the pulse waveform very well and has a constant
phase centre (van Cappellen et al., 2000). The flaring and length of the antenna can
UWB Antennas
133
be adjusted to modify the radiation pattern, impedance matching, and the transient
behaviour of the antenna (Shlager et al., 1996). The gain of the TEM horn ranges from
5 to 15 dB, which is suitable for directive base-station operation.
The spatial dispersion caused by the non-planarity of the waveform from TEM horn
can be corrected by the use of lenses (Baum and Stone, 1993) or reflectors (Foster, 1993;
Baum and Farr, 1993). One such antenna is the impulse radiating antenna (IRA) (Baum
and Farr, 1993) which will be described in Section 6.3.4.
6.3.3
TEM Horn Variants
There are several modifications of the TEM horn. Septum plates can be put inside the
horn to improve the field uniformity (Buchenauer et al., 1999). Another technique
described by ( Yarovoy et al., 2000) is to fill the horn with dielectric material. One half
of the horn can also be replaced by a large ground plane such that the horn resembles a
bent bow-tie monopole. In this last case, a broadband balun is not required. It is also
possible to use resistive loading in the antenna to suppress the reflections from the end
of the horn (Shlager et al., 1996).
The idea behind a ridged horn and a tapered slot antenna (Yngvesson et al., 1989; Lewis
et al., 1974) is similar to that of the TEM horn: a TEM- or quasi-TEM mode is used to feed
the antenna. With a ridged horn structure, there is a ridge inside the TEM horn. This
approach appears to yield more constant gain with respect to frequency relative to standard
TEM horns. However, the structure is more difficult to produce. In addition, the antenna
appears not to be optimal for pulse radiation (van Capellen, 1998). The polarization diverse
antenna by Wicks and Antonik (1993) can also be considered as a variation of a ridged horn.
Tapered-slot antennas can be constructed using printed-circuit-board techniques.
With such an antenna, however, it may be difficult to achieve constant gain. The
tapered-slot antenna and its variations appear to be good candidates for portable
antennas. Another method for constructing TEM-horn-type antennas on printed circuit
boards is described by Nguyen et al., (2001).
6.3.4
Impulse Radiating Antenna
An impulse radiating antenna (IRA) consists of a TEM horn feeding a parabolic
reflector (Baum and Farr, 1993). With such an antenna, it is possible to obtain high
and almost frequency-independent gain. The reported gains are of the order of 25 dBi.
The antenna gain can be made adjustable (Farr et al., 1999) by moving the TEM horn
feed off the focal point of the parabolic reflector.
The high gain makes this antenna a good candidate for very long range base-station
applications. The high gain, narrow beamwidth and short pulses generated by this
antenna make it highly immune to interference.
6.3.5
Folded-horn antenna
A folded horn antenna for UWB high power applications is presented by Kardo-Sysoev
et al., (1999). The idea of the folded-horn antenna comes from sub-horns inserted in
UWB Theory and Applications
134
Folded horn with
apex angle φ
φ V
Not folded horn with same
aperture and angle at the apex φ
Figure 6.2
Folded horn antenna
a main horn. The sub-horns divide the initial horn aperture into two equal parts. Using
this technique, the size of the antenna can be reduced. This can be seen from Figure 6.2.
which shows the structure of folded horn antenna.
6.3.6
Dipoles and Monopoles
Dipoles and monopoles without resistive loading are based on resonance techniques,
therefore the ringing effect described earlier significantly degrades their performance.
A comprehensive review of differently loaded thin-wire monopoles is presented by
Montoya and Smith (1996). The results indicate that none of the monopoles examined
preserve the pulse characteristic. The resistive loading, however, seems to improve
performance significantly.
A popular design is the bow-tie antenna (also called the ‘bifin’ antenna, Stutzman
and Thiele, 1981). The bow-tie antenna is used in an ultra-wide antenna design Lai et al.,
(1992) which shows some examples of different ultra-wideband antennas.
The beamwidth and input impedance of a bow-tie antenna depend directly on the
antenna geometry, and they are nearly constant over the desired frequency range. To
ensure a balanced and wideband feed to the bow-tie antenna, a hybrid construction with
a slot line antenna is used (Lai et al., 1992). The bandwidth of a bow-tie antenna
depends on the length of the plate (see Figure 6.3). The flare angle and the length of
the plate define the lower frequency. The beamwidth of the bow-tie antenna varies
linearly with the flare angle.
The efficient use of resistive materials also appears to be beneficial in the bow-tie
antennas (Maloney and Smith, 1993a; Shlager et al., 1994).
There are other designs that can be considered as dipole structures, such as the
resistively loaded antenna by Chevalier et al. (1999) optimized for pulse radiation.
UWB Antennas
135
Biconical antenna
Bowtie antenna
Folded bowtie antenna
Plate length
Flare angle
Plate
angle
Cone
Figure 6.3
Flat
plate
A few wideband bow-tie antenna types
Lu and Shi (1999) demonstrated that dipole structures should be favoured over
monopole structures, because the size of the ground plane needed with the monopole
may be not be practical. On the other hand, the problems in developing an ultrabroadband dipole radiator is in controlling the unbalanced currents in the outer feed
cables. The currents are normally filtered by balun transformers or RF chokes. However transformers for UWB systems are difficult to design, and performance of ferrite
RF chokes degrade above 1 GHz.
6.3.7
Loop Antennas
There have been several attempts to use loop antennas in UWB communications. In
contrast to dipoles, the loop antennas radiate second derivatives of the incident electric
field (Harmuth, 1978), which may be advantageous for some waveforms. Yarovoy et al.
(2000) have presented a small loop antenna for UWB measurements. Such an antenna
structure appears to be a good candidate for the portable antenna.
The so-called large current radiators can also be considered to be loop antennas.
Their use in pulsed radiation was demonstrated by Pochain (1999).
6.3.8
Antenna Arrays
In UWB radar applications, linear and planar antenna arrays may be formed with
very sparsely spaced elements. This enables economical, high resolution phased array
antennas, with a beam which may be readily steered (Anderson et al., 1991). The ratio of
the wideband peak sidelobe level to the peak main lobe level is a function of the number
of antenna elements rather than the element spacing. As a consequence, sparsely spaced
wideband array antennas do not result in significant grating lobes. If the number of
antenna elements is increased, the side-lobe levels may be reduced to almost arbitrarily
low levels. In Lu and Shi (1997), an UWB omnidirectional monopole antenna is
UWB Theory and Applications
136
y
presented for the SPEAKeasy system, where the bandwidth requirement for the base
station antenna ranges from 30 to 500 MHz.
6.4 Simulation Techniques
The time and frequency domain are connected by the Fourier transform, which means
that time domain antenna simulations can be performed using standard frequency
domain electromagnetic simulators. However, conducting analysis in the frequency
domain requires the calculation of far-field values for both amplitude and phase. This
must be done over a very wide frequency range to be able accurately to extract the shape
of the radiated pulse. For example, if transmission between two antennas is investigated
(separated by 2 m), the calculation must be done over the range of approximately
10 MHz to 10 GHz, with a suitably small step size determined by phase resolution
requirements, in order to preserve the pulse characteristics. In practice, this may mean
a step size of 10–20 MHz, leading to a very large number of required calculation points.
In this work it is assumed, that the calculations are performed using standard personal
computers, which makes certain calculations very time consuming or, due to lack of
memory, impossible to perform. The limitations are less severe, if modern supercomputers with parallel architectures are used instead.
In this section, the transition from the frequency domain to the time domain and
visualization have been performed using MatlabTM.
6.4.1
Finite-Element Method
Software tools based on use of the finite-element method (FEM), such as Ansoft HFSS,
can be used to simulate antennas of arbitrary shape and material. However, FEM
techniques are very slow. Furthermore, the FEM mesh evaluation technique works
one single frequency at a time, implying that the mesh produced is only applicable for
a limited bandwidth around the single frequency used in the calculation.
6.4.1.1
Method of Moments
Software tools based on the method of moments (MoM) are much faster than FEM
tools. Most software tools, such as the Zeland IE3D and NEC-Win Pro from Nittany
Scientific do not use adaptive meshing. Instead, the user may define the density of the
grid. According to preliminary testing (see Section 6.5), moment-method-based tools
are applicable for predicting the transient antenna behaviour. However, most MoM
codes can only support infinite dielectric layers.
6.4.1.2
Finite-Difference Time-Domain
The finite-difference time-domain (FDTD) method and other time-domain methods are
inherently applicable for studying the ultra-short pulses. Software packages such as
y USA military software defined radio system (Cook and Bonser, 1999).
UWB Antennas
137
CST Microwave Studio can be used to simulate antennas of arbitrary shape and
material. This is demonstrated in Section 6.5. The transient solver of CST Microwave
Studio, which is based on finite integration, is restricted to rather simple and small
structures only. For this reason the calculations tend to become very slow for complex
structures. The improved grid generation methods, such as sub-gridding, together with
increased computing capacity makes time-domain methods the most promising for
UWB simulations.
6.5 Simulation examples
This section presents simulation examples. Structures that are known to preserve and
disperse the incident pulse waveform are presented in order to highlight the differences
between good and bad UWB antennas. Some of the structures are designed to work at
the different frequency range as defined by FCC (2002). However, each structure and
waveform can be scaled to meet the required frequencies.
The first three examples are simple dipole structures, including a perfectly conducting
dipole, a capacitively loaded dipole, and a resistively loaded dipole. The fourth example
is a conical dipole. The fifth example is a log-periodic dipole array, which is known not
to be suitable for pulsed radiation. All these five structures are simulated using the
NEC-Win Pro software, which models the structures using wire segments. There are
several non-commercial versions of NEC-code, which makes it easy to compare these
results. Students are advised to make comparative calculations with these codes. Dipole
structures are treated instead of monopoles, as their treatment is more reliable with the
used software.
Finally, a well known UWB antenna, the TEM horn, is simulated using both MoM
software Zeland IE3D and time-domain software CST Microwave Studio.
The incident waveform in the simulations is a differentiated Gaussian monopulse.
The centre frequency of the waveform is chosen to be 3 GHz. The time and frequency
domain representations are presented in Figures 6.4, 6.5 respectively.
6.5.1
Perfectly Conducting Dipole
As a first example, a thin dipole antenna was examined using the NEC-Win Pro
simulation tool. The results of the calculations are presented in Figures 6.6, 6.7 and
Table 6.2. The electric fields and the power in these figures are given in arbitrary units.
The same applies in subsequent figures. The reflection coefficient and the radiated
electric field into the bore sight direction, presented in Figure 6.6, indicate that the
dipole resonates in half-wavelength intervals. This causes severe ringing in the antenna
as seen in the time-dependent bore sight electric field and time-dependent power
spectrum (Figure 6.6). The ringing effect is very clear from the power spectrum. This
is demonstrated in Figure 6.7, which show magnification of the power spectrum with
respect to time. Note that the negative power is due to spline interpolation applied to the
calculated points. The spline interpolation has been done to resolve the positions of the
smaller peaks.
UWB Theory and Applications
138
1
0.8
0.6
0.4
0.2
0
–0.2
–0.4
–0.6
–0.8
–1
–5
–4
–3
–2
–1
0
1
2
3
4
5
Time (ns)
Figure 6.4
Incident voltage pulse used in time-domain simulations
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
0
1
2
3
4
5
6
7
8
9
10
Frequency (GHz)
Figure 6.5
Incident voltage pulse used in frequency domain simulations
In Figure 6.7, peaks 1, 3, and 5 correspond to (n þ 1/2) resonances, while the peaks
2, 4, and 6 to n resonances. The time differences from these peaks are presented in
Table 6.2. The ringing is caused by reflections from the ends of the dipole. The electric
field probe measuring the far field is located 2 metres from the antenna, which corresponds to a time delay of 6.7 ns.
UWB Antennas
139
Table 6.2 Time differences for peak positions
for power spectrum of thin dipole antenna
Peaks
Time delay (ns)
Radiated power
Radiated E-field
Radiated E-field
Reflection coeff.
1–2
2–3
3–4
4–5
5–6
Half dipole distance (0.1 m)
0.35
0.32
0.38
0.30
0.41
0.33
1
0.5
0
0
2
4
6
Frequency (GHz)
8
10
0
2
4
6
Frequency (GHz)
8
10
4
6
8
10
Time (ns)
12
14
4
6
8
10
Time (ns)
12
14
0.4
0.2
0
1
0
–1
1
0.5
0
Figure 6.6 Reflection coefficient (uppermost), radiated electric field with respect to
frequency and time, and radiated power of a thin dipole antenna
UWB Theory and Applications
140
2
1
0.8
Transmitted power
1
0.6
0.4
0.2
4
3
5
0
6
6.5
7
7.5
8
6
8.5
9
Time (ns)
Figure 6.7
6.5.2
Radiated power of a thin dipole antenna
Capacitively Loaded Dipole
As a second example, a thin capacitively loaded dipole antenna was studied using the NECWin Pro software. The capacitive loading can be used to improve impedance matching of thin
dipoles. For example, the impedance matching is improved compared with a typical dipole as
shown in Section 6.5.1. However, the capacitances act as high-pass filters in the dipole
antenna, which appears to increase the ringing effect here relative to that in the perfectly
conducting dipole. This is illustrated in Figure 6.8. The electric field probe measuring the far
field is located 1 metre from the antenna, which corresponds to a time delay of 3.3 ns.
The results for perfectly conducting and capacitively loaded dipole indicate that
typical dipole structures are not optimal for UWB applications. The most straightforward method to improve the performance is to use resistive loading. This is demonstrated in the following Section 6.5.3.
6.5.3
Resistively Loaded Dipole
A resistively loaded Wu-King dipole (Maloney and Smith, 1993b) was also studied
using the NEC-Win Pro software. The results, in Figure 6.9, indicate that the frequency
domain gain and reflection coefficient are smooth. As a result of this, the pulse waveform is well preserved. The radiation efficiency of this antenna is, according to Montoya
and Smith (1996), approximately 20%. In addition, the input impedance of the simulated antenna is as high as 660 , which is impractical in realistic applications. On the
other hand, the design clearly demonstrates the benefits of using resistive loading. In
case high efficiency is not needed, the resistive loading is a straightforward method of
improving the transient properties of small antennas.
Radiated power
Radiated E-field
Radiated E-field
Reflection coeff.
UWB Antennas
141
1
0.5
0
0
1
2
3
4
5
6
Frequency (GHz)
7
8
9
10
0
1
2
3
4
5
6
Frequency (GHz)
7
8
9
10
0
1
2
3
4
5
6
Time (ns)
7
8
9
10
0
1
2
3
4
5
6
Time (ns)
7
8
9
10
0.6
0.4
0.2
0
1
0
–1
1
0.5
0
Figure 6.8 Reflection coefficient (uppermost), radiated electric field with respect to
frequency and time, and radiated power of a capacitively loaded dipole
6.5.4
Conical Dipole
A conical dipole was studied using the NEC-Win Pro software as a example of a UWB
omnidirectional antenna. The model was also used to demonstrate that the time-domain
properties are generally direction dependent, as are the frequency-domain properties.
The wire model of the antenna is presented in Figure 6.10. The results for this antenna in
the bore sight direction are presented in Figure 6.11.
Figure 6.11 indicates that the frequency-domain radiation pattern of the bi-conical
antenna in the bore sight direction is not smooth, but contains several deep minima.
This appears to cause considerable ringing, although with much smaller amplitudes as
in standard and capacitively loaded dipole (Sections 6.5.1 and 6.5.2). The radiation
pattern is smoother in directions slightly above the bore sight plane. This is illustrated in
Figure 6.12. The smooth gain also results in decreased ringing.
UWB Theory and Applications
Radiated power
Radiated E-field
Radiated E-field
Reflection coeff.
142
1
0.5
0
0
1
2
3
4
5
6
Frequency (GHz)
7
8
9
10
0
1
2
3
4
5
6
Frequency (GHz)
7
8
9
10
0
1
2
3
4
7
8
9
10
0
1
2
3
4
7
8
9
10
0.06
0.04
0.02
0
1
0.5
0
–0.5
–1
5
6
Time (ns)
1
0.5
0
5
6
Time (ns)
Figure 6.9 Reflection coefficient (uppermost), radiated electric field with respect to
frequency and time, and radiated power for resistively loaded Wu-King dipole
It is possible to use resistive loading to decrease the ringing even further, as demonstrated by Maloney and Smith (1993a).
The time-domain behaviour of the bi-conical antenna wire model is in general better
than that of the thin dipole. The conical dipole, with its variants such as the 2D bow-tie
antenna, can thus be seen to be a good starting point for small UWB antenna designs.
6.5.5
Log-periodic Dipole Array
A log-periodic dipole array (LPDA, see Figure 6.13) was examined using the NEC-Win
Pro software package as an example of a dispersive wideband antenna. The results are
presented in Figure 6.14. The results indicate that the antenna examined has a very low
UWB Antennas
143
Figure 6.10 Wire-model of a bi-conical dipole
reflection coefficient and smooth gain. However, the LPDA does not have a constant
phase centre, which causes severe dispersion of the pulse. Figure 6.14 shows that the
high frequency components of the pulse are the first to arrive at the receiver as the
feeding point of the antenna is at the high frequency part of the antenna. The lower
frequencies are delayed, because the resonating longer dipoles are located behind the
feeding point with respect to the bore sight direction. On the basis of the results, the
LPDA is in general not suitable for UWB applications.
It is possible to decrease dispersion by using special feeding structures for each of
the antenna elements of LPDA (Excell et al., 1998).
6.5.6
TEM Horn
A TEM horn model was designed in order to compare the available software to measured
data (see Section 6.6). Simulations were performed using a moment method based Zeland
IE3D program package and a time domain software tool from CST Microwave Studio.
The Zeland IE3D results for antenna-to-antenna setup are presented in Figure 6.15.
The results obtained using the CST Microwave Studio package are presented in
Figure 6.16. The antennas in Figure 6.15 are located 2 metres from each other, whilst
the antennas in Figure 6.16 are located 0.6 metres from each other. The longer interantenna distances in CST Microwave Studio are not practical due to computational
intensity. Note the different frequency scale for the uppermost figure. In the CST
simulations, the incident pulse differs slightly from the Gaussian monopulse. Otherwise
the two simulators give very similar results.
UWB Theory and Applications
Reflection coeff.
144
1
0.5
Radiated E-field
0
Radiated E-field
1
2
3
4
5
6
Frequency (GHz)
7
8
9
10
0
1
2
3
4
5
6
Frequency (GHz)
7
8
9
10
0
1
2
3
4
7
8
9
10
0
1
2
3
4
7
8
9
10
1.5
1
0.5
0
1
0
–1
Radiated power
0
5
6
Time (ns)
1
0.5
0
5
6
Time (ns)
Figure 6.11 Reflection coefficient (uppermost), radiated electric field with respect to
frequency and time, and radiated power for a bi-conical dipole (bore sight direction)
6.6 Measured examples
This section presents preliminary measurements for selected constructed prototypes.
The measurement set-up required to measure antenna transient characterization is also
briefly presented.
6.6.1
6.6.1.1
Measurement Techniques
Frequency-domain Measurements
Antenna-to-antenna time-domain responses can be evaluated from the corresponding
frequency response. Both amplitude and phase of the transmission must be recorded.
UWB Antennas
145
Reflection coeff.
1
0.5
Radiated E-field
0
0
1
2
3
6
5
4
Frequency (GHz)
7
8
9
10
0
1
2
3
6
5
4
Frequency (GHz)
7
8
9
10
0
1
2
3
4
5
Time (ns)
6
7
8
9
10
0
1
2
3
4
5
6
7
8
9
10
0.4
0.2
0
Radiated E-field
1
0
Radiated power
–1
1
0.5
0
Time (ns)
Figure 6.12 Reflection coefficient (uppermost), radiated electric field with respect to
frequency and time, and radiated power for a bi-conical dipole (40 degrees above the bore
sight plane).
This can be done using a standard vector network analyser (VNA). The VNA must be
calibrated for a broad frequency span to avoid artificial distortion of the waveform due
to aliasing. In a set of preliminary measurements, the VNA was calibrated to cover the
frequency range of 50 MHz–20 GHz. This yields a maximum resolution of approximately 10 MHz in frequency, which according to the Nyquist rate formula, restricts the
maximum measuring distance to about 5 metres. Longer distances can be handled by
using multiple calibrations, or by narrowing the frequency range. Examples of measurements using VNA are presented below.
UWB Theory and Applications
146
Figure 6.13 Wire model of log-periodic dipole array
Reflection coeff.
1
0.5
0
0
1
2
3
4
5
6
Frequency (GHz)
7
8
9
10
0
1
2
3
6
5
4
Frequency (GHz)
7
8
9
10
4
5
6
7
8
9
10
Time (ns)
11
12
13
14
4
5
6
7
8
11
12
13
14
Radiated E-field
3
2
1
0
Radiated E-field
1
0
Radiated power
–1
1
0.5
0
9
10
Time (ns)
Figure 6.14 Reflection coefficient (uppermost), radiated electric field with respect to
frequency and time, and radiated power for a log-periodic dipole array (bore sight direction)
UWB Antennas
Received E-field
1
147
×10–3
0.5
0
0
1
2
3
4
5
6
Frequency (GHz)
7
8
9
10
5
6
7
8
9
11
10
Time (ns)
12
13
14
15
5
6
7
8
9
11
10
Time (ns)
12
13
14
15
Received E-field
1
0.5
0
–0.5
Received power
–1
1
0.5
0
Figure 6.15 Received electric field with respect to frequency and time, and received timedomain power for TEM horn using the Zeland IE3D (bore sight direction)
6.6.1.2
Time-domain Measurements
A typical time domain antenna measurement consists of a pulse generator and a digital
sampling oscilloscope. Most of the available pulse generators are based on step generators with very short rise and fall times. Pulses and impulses are generated from the
step response using differentiation techniques. The pulses may also be shortened
through the use of step recovery diodes. In practice, the authors have found that
commercial pulse generators are expensive and difficult to embed in small designs.
The lack of suitable commercial pulse generators has led to many designs being
described in the open literature. For example, Lee et al. have presented small subnanosecond monocycle pulse generators (Lee and Nguyen, 2001a; Lee et al., 2001c).
UWB Theory and Applications
Received E-field
148
0.1
0.05
0
0
1
2
3
Frequency (GHz)
4
5
6
Received E-field
E-field of incident pulse
1
0.5
0
–0.5
–1
0
1
2
3
4
0
1
2
3
4
5
Time (ns)
6
7
8
9
10
5
6
7
8
9
10
0.05
0
–0.05
Time (ns)
Figure 6.16 Received electric field with respect to frequency, incident waveform and
received waveform for TEM horn, simulated using the CST Microwave Studio (bore sight
direction)
A digital sampling oscilloscope, together with a sampling, head is a critical part of the
measuring system.
6.6.2
TEM Horn
Two TEM horn prototypes were constructed and measured in order to demonstrate the
use of the VNA in the time domain antenna characterization. The model, ‘a bent bowtie monopole’ (see Section 6.5.6), is based on the simulated structure. For computational simplicity, the ground plate is infinite in the simulated model. The prototypes
were constructed with different ground plane sizes in order to investigate the effects of a
finite ground plane. The antennas are presented in Figure 6.17.
UWB Antennas
149
Figure 6.17 Constructed TEM horn with a small ground plane
The S11 of the TEM antenna using different prototypes and different simulation
methods are presented in Figure 6.18. There is good agreement between the simulations
and measurements for the antennas examined. In particular, the Zeland IE3D result
very closely matched the experimental results. The antenna with the large ground plane
is UWB according to our definition given in Table 6.1 — the VSWR is lower than 2 in the
frequency range of 1.65 GHz–10.9 GHz, and lower than 3 in the frequency range of
0.93 GHz–18 GHz.
As a next step, the frequency-domain antenna-to-antenna measurements were performed. All measurements were performed with a VNA. Measured S21 values of
amplitude and phase were post-processed with MatlabTM, resulting in a mathematically
calculated pulse shape. The measurements were done using two inter-antenna distances,
1 m and 1.5 m. In addition, a simulation in which a metal plate was placed between the
antennas to cause reflections was performed with inter-antenna distance of 1 m. In the
data analysis using MatlabTM, the antennas were ‘excited’ using the same Gaussian
mono-pulse as in the Zeland IE3D simulations.
Figure 6.19 shows the simulated and measured waveform in the antenna-to-antenna
set-up. The uppermost curve corresponds to the Zeland IE3D simulation, the second
curve is from the CST Microwave Studio, and the three lowest curves are measured
results (VNA measurements). The first measured result is obtained from the
UWB Theory and Applications
150
0
Reflection coefficient (dB)
–5
–10
Large ground
plane
Small ground
plane
IE3D
Microwave
Studio
–15
–20
–25
–30
0
2
4
6
8
10
Frequency (GHz)
Figure 6.18 Measured and simulated S11 of a TEM horn
inter-antenna distance of 1 m, the second of 1.5 m. In the third measurement, the 1 m
measurement is disturbed by a reflective plate.
The measured and simulated results indicate that the TEM horn is suitable for UWB
operation. The results also indicate that the modern simulation software can be used
reliably to estimate the transient behaviour of the UWB antennas as long as the antenna
is modeled accurately.
6.6.3
Small Antennas
In mobile applications, the size of the antenna is often restricted. In this section two
types of small antenna prototypes are designed and constructed.
The first of the prototypes is of modified bow-tie type. These antennas are later
referred to as MBT (modified bow-tie). As compared with conventional bow-tie, the
shape of the radiators is a half circle instead of triangle. This antenna uses a taperedslot-type feed point, but acts as a bow-tie dipole at lower frequencies. This modification
improves the impedance matching as compared to bow-tie antenna. The structure can
be considered at a high frequency region as a pair of tapered slot antennas. Constructed
antenna prototypes are presented in Figure 6.20. These antennas were manufactured on
PCB. Measured and simulated S11 are presented in Figure 6.21. The agreement in this
case is very good. The simulations were performed using CST Microwave Studio, as the
use of the FI method makes it possible to model the finite PCB structure.
The second studied antenna type is a tapered slot antenna (TSA). The constructed
TSA prototypes are shown in Figure 6.22. These antennas are more directive than the
MBT antennas. The directivity makes these antennas less favourable for portable
applications which need omni-directional radiation.
UWB Antennas
151
1
0
–1
3
4
5
6
7
8
3
4
5
6
7
8
3
4
5
6
7
8
3
4
5
6
7
8
3
4
5
6
7
8
1
Received waveform
0
–1
1
0
–1
1
0
–1
1
0
–1
Time (ns)
Figure 6.19 Simulated and measured waveform in the antenna-to-antenna set-up. See text
for details and explanations
Figure 6.23 shows the simulated and measured waveform in the antenna-to-antenna
set-up for MBT and TSA antennas. The results indicate that the ringing is not severe in
these antenna types. The spurious pulse in MBT results at about 6 ns was found to be
due to reflections from the antenna cables. The results indicate that these antennas can
be considered a good starting point for mobile UWB antennas.
In order to demonstrate further that time-domain properties are generally direction
dependent, the pulse shape of the MBT antenna was measured at different rotation
angles (see Figure 6.24). Figure 6.25 shows the time-domain pulse as a function of the
UWB Theory and Applications
152
Figure 6.20 MBT antennas manufactured on PCB
0
Reflection coefficient (dB)
–5
–10
Antenna 1
Antenna 2
–15
CST Microwave
Studio
–20
–25
–30
0
2
4
6
8
10
12
14
16
18
20
Frequency (GHz)
Figure 6.21
Measured and simulated S11 of MBT antennas
horizontal angle. The results indicate that the pulse shape varies as a function of the
horizontal angle, but remains satisfactory at all angles.
In order for the pulse to be transmitted properly, the antennas must be of
same polarization. Figure 6.26 shows the measurement set-up, in which the radiating
antennas are of different polarizations. In this case, not only the amplitude of the
UWB Antennas
153
Figure 6.22 Tapered slot antenna prototypes
1
Received waveform
0
–1
2
3
4
5
6
7
8
9
10
2
3
4
5
6
Time (ns)
7
8
9
10
1
0
–1
Figure 6.23 Measured waveform in the antenna-to-antenna setup for MBT (top) and TSA
(below) antennas
pulse is decreased, but the pulse shape is heavily distorted. Figure 6.27 shows the time
domain pulse as a function of the horizontal angle. Now the measured pulse shapes
show a clear ringing effect. In mobile applications the orientation of the antenna cannot
be guaranteed. For this reason, it is necessary to consider multiple polarizations, or
alternatively circular polarization, to guarantee good pulse shape at the receiver.
UWB Theory and Applications
154
30 cm
Figure 6.24 Measurement configuration for the MBT antennas. The horizontal angle is 0
0
1
0
–1
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
30
1
0
–1
60
1
0
–1
90
1
0
–1
Figure 6.25 Time domain response at different rotation angles (0 –90 )
UWB Antennas
155
30 cm
Figure 6.26 Cross-polarization measurement configuration for the MBT antennas. The
horizontal angle is 0
0
1
0
–1
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
0
1
2
3
4
5
6
30
1
0
–1
60
1
0
–1
90
1
0
–1
Figure 6.27
Time domain response at different rotation angles (0 –90 )
156
UWB Theory and Applications
6.7 Conclusions
This chapter has examined some of the more common structures and analysis methods
for UWB antennas. This has been done with the aid of simulations and prototype
measurements. The results indicate that the design of UWB antennas can be done in a
straightforward fashion using standard frequency domain simulators and measurements. The major design issues for UWB antenna are matching and ensuring constant
gain, i.e. smooth amplitude and phase radiation patterns, across a very wide band of
operation.
Structures for mobile antennas have been presented as well as various TEM horn
structures for ‘base station’ antennas. Of the examples presented, the bow-tie, or a
modification of it, is one of the candidates for mobile antennas due to its small size,
simplicity, and the fact that it can be constructed using simple PCB techniques. The
straightforward improvement of this antenna would be to use resistive loading in the
end of radiators to minimize reflections.
7
Medium Access Control
Ulrico Celentano, Ian Oppermann
7.1 Introduction
UWB technology was originally seen purely as a physical layer technology, with no or
little protocol to control the communication. It is now clear that Medium access control
(MAC) features play a major role in UWB communication systems.
UWB holds enormous potential for wireless ad-hoc and peer-to-peer networks. One
of the major potential advantages in impulse radio based systems is the ability to trade
data rate for link distance by simply using more or less concatenated pulses to define a
bit (see Chapter 3). Without dramatically changing the air interface, the data rate can be
changed by orders of magnitude depending on the system requirements. This means,
however, that high data rate (HDR) and low data rate (LDR) devices will need to
coexist. The narrow time domain pulse also means that UWB offers the possibility for
very high positioning accuracy. However, each device in the network must be ‘heard’ by
a number of other devices in order to generate a position from a delay or signal angleof-arrival estimate. These potential benefits, coupled with the fact that an individual low
power UWB pulse is difficult to detect, offer some significant challenges for the multiple
access MAC design.
UWB systems have been targeted at very HDR applications over short distances,
such as USB replacement, as well as very LDR applications over longer distances, such
as sensors and RF tags. Classes of LDR devices are expected to be very low complexity
and very low cost, implying that the MAC will also need to be very low complexity. The
potential proliferation of UWB devices means that a MAC must deal with a large
number of issues related to coexistence and interoperation of different types of UWB
devices with different capabilities. The complexity limitations of LDR devices may
mean that very simple solutions are required. HDR devices, which are expected to be
higher complexity, may have much more sophisticated solutions.
UWB Theory and Applications Edited by I. Oppermann, M. Hämäläinen and J. Iinatti
Ó 2004 John Wiley & Sons, Ltd ISBN: 0-470-86917-8
UWB Theory and Applications
158
Research on UWB MAC is at an early stage, and no comprehensive solution for the
unique difficulties for UWB systems have been proposed in the open literature. Existing
or emerging standards that allow direct peer-to-peer communications include the IEEE
802.15.3 (IEEE, 2003) and ETSI HiperLAN Type 2, although only in the context of a
centrally managed network. This chapter will examine some of the difficulties facing
UWB MAC design, and look at the impact of some of the key physical layer parameters
on MAC design. The chapter will also look at the emerging work of the IEEE working
group 802.15.3a.
7.2 Multiple Access in UWB Systems
7.2.1
MAC Objectives
The very wide bandwidth of UWB systems means that many potential solutions exist to
the issue of bandwidth usage. Devices may use all, or only a fraction of, the bandwidth
available in the 3.1 to 10.6 GHz band. These devices will still be classed as UWB
provided they use at least 500 MHz. Chapter 3 gave a summary of the major candidates
for the physical layer signal structure of UWB systems, which include impulse radio,
OFDM, multi-carrier and hybrid techniques. All of these possible techniques mean that
different UWB devices may or may not be able to detect the presence of other devices.
The main issues to be addressed by an UWB MAC include coexistence, interoperability
and support for positioning/tracking.
7.2.1.1
Coexistence
The potential proliferation of UWB devices of widely varying data rates and complexities will require coexistence strategies to be developed.
Strategies for ignoring or working around other devices of the same or different type
based on physical layer properties will reflect up to the MAC layer. Optimization of the
UWB physical layer should lead to the highest efficiency, lowest BER, lowest complexity transceivers. The assumptions of the physical layer will however have implications
on MAC issues such as initial search and acquisition process, channel access protocols,
interference avoidance/minimization protocols, and power adaptation protocols. The
quality of the achieved ‘channel’ will have implications on the link level, which may
necessitate active searching by a device for better conditions, which is what happens
with other radio systems.
7.2.1.2
Interoperability
The most common requirement of MAC protocols is to support inter-working with
other devices of the same type. With the potentially wide range of device types, the
MAC design challenge is to be able to ensure cooperation and information exchange
between devices of different data rate, QoS class or complexity. In particular, emphasis
must be placed on how low complexity LDR devices can successfully produce limited
QoS networks with higher complexity, HDR devices.
Medium Access Control
7.2.1.3
159
Positioning/Tracking Support
Positioning is integrally linked to the MAC. This includes strategies for improving
timing positioning accuracy and for exchanging timing information to produce positioning information.
It is possible for any single device to estimate the arrival time of a signal from another
device based on its own time reference. This single data point in relative time needs to be
combined with other measurements to produce a 3D position estimate relative to some
system reference. Exchange of timing information requires cooperation between devices.
Being able to locate all devices in a system presents a variation of the ‘hidden node’
problem. The problem is further complicated for positioning because multiple receivers need
to detect the signal of each node to allow a position in three dimensions to be determined.
Tracking requires that each device is able to be sensed/measured at a suitable rate to
allow reasonable update rate. This is relatively easy for a small number of devices, but
difficult for an arbitrarily large number of devices. Information exchange between devices
of timing and position estimates of neighbours (ad hoc modes) requires coordination,
and calculation of position needs to be done somewhere (centralized or distributed) and
the results fed to the information sink.
Finally, it is important to have the received signal as unencumbered by multiple
access interference as possible in order to allow the best estimation of time of arrival.
Every 3.3 ns error in delay estimation translates to a minimum 1 m extra error in
position estimation.
All of these issues (information exchange, device sampling rate, node visibility, signal
conditioning) require MAC support. These issues are significant obstacles to existing
WLAN and other radio systems offering reliable positioning/tracking when added on to
the MAC post-design.
7.2.2
Structure of the UWB Signal
As seen in Chapter 3, the impulse radio UWB signal is composed of a train of very short
pulses. A typical pulse width Tp, is of the order of 0.1–0.5 ns. Figure 7.1 shows the
structure of the basic UWB signal. Pulses are repeated with a (mean) pulse repetition
time or frame time, Tf, Tf Tp. The value of Tf may be hundreds or thousands of times
the pulse width (Scholtz, 1993, Scholtz and Win, 1997). The reciprocal of Tf, Rf, is called
pulse repetition frequency.
In the absence of coding or modulation, pulses are separated by Tf seconds. As seen in
Chapter 3, pulses might be placed within the time frame period depending on the coding
and modulation techniques applied.
Tp
Tf
Figure 7.1
Structure of the basic UWB signal
UWB Theory and Applications
160
In order to increase error resistance, repetition coding is often used. This means that a
number of pulses, N, is transmitted for each bit. This may be seen as over-sampling with
a factor of N when mapping data symbols to pulses. The bit period is given by
Ts ¼ NTf :
7.2.3
ð7:1Þ
Modulation and Multiple Access
As seen in Chapter 3, common multiple access techniques adopted for pulse based UWB
systems are TH and DS. Suitable modulation techniques include OOK (Foerster et al.,
2001) and especially PPM and PAM (Hämäläinen et al., 2002). A given UWB communication system will be a combination of these techniques, leading to signals based
on, for example, TH-PPM, TH-BPAM or DS-BPAM. TH-PPM is probably the most
frequently adopted scheme and will be used in the following as an example for determining the resources available in a UWB system, resources that are to be managed by
the medium access control.
To carry information, a modulation scheme is applied to a series of UWB pulses. If
PPM modulation is adopted, the position of the pulse inside the slot window is chosen
according to the data bit transmitted. A pulse is transmitted ‘on time’, or delayed by a
certain quantity, DT, depending on whether a ‘1’ or a ‘0’ is transmitted. It is clear that all
N pulses to which each bit is mapped are delayed by the same time shift.
In order to avoid collisions from different users, each pulse is placed inside a slot of
duration Tc positioned inside the frame period Tf. The slot position is chosen according to a
specific time hopping code which may be unique to each user (centralized system) or
selected randomly by each user (uncoordinated system). Nh slots of duration Tc are
available during a frame period Tf. The choice of Nh impacts the error performance and
transceiver complexity. In an uncoordinated system, if Nh is small, the probability of user
collisions is high. Conversely, if Nh is large, less time is available for reading the output and
resetting the monocycle correlator (Scholtz, 1993; Win and scholtz, 1998a). For the same
reason, the overall maximum time shift must be strictly less than the pulse repetition time Tf
ð7:2Þ
Nh Tc < Tf :
In Figure 7.2, the solid-line user has time hopping code {2,1,3,2} and data {0,1},
whereas the dashed-line user has {2,3,1,1} and {1,1}. A third interfering signal is also
shown in dot–dashed line.
For a TH-PPM system, the signal in the UWB channel in the multiuser case is given by:
Nu
X
k¼1
Ak
þ1 N
s 1
X
X
ðkÞ
ðkÞ
gðt iTb jTf cj Tc bi Þ
ð7:3Þ
i¼1 j¼0
where g(t) is the pulse waveform, Ak is the channel attenuation for user k, Tb is the
(k)
bit period, c(k)
j is the time hopping code for frame j for user k, and bi is the data bit i for
user k. The time hopping code chip is in the range from 0 to NhTc.
Medium Access Control
Tc
161
(Nh)
Tf
(Ns)
Example of TH-PPM. Nh ¼ 3; Ns ¼ 2
Figure 7.2
M-PAM is considered by Foerster et al., (2001) and Hämäläinen et al., (2002). The
TH-BPAM signal for the kth user is given by:
sðkÞ ðtÞ ¼
þ1 N
s 1
X
X
ðkÞ
ðkÞ
gðt iTb jTf cj Tc Þdi
ð7:4Þ
i¼1 j¼0
ðkÞ
where di is the data sequence.
The DS-BPAM signal for the k-th user is given by:
sðkÞ ðtÞ ¼
þ1 N
s 1
X
X
ðkÞ ðkÞ
gðt iTb jTc Þcj di
ð7:5Þ
i¼1 j¼0
where di(k) is the data sequence.
7.2.4
Multiuser System Capacity
The resistance against interference obtained with over-sampling, i.e., pulse repetition
coding, described above, leads to a processing gain in linear scale expressed by
PG1 ¼ N
ð7:6Þ
Another component of the processing gain is due to the duty cycle and can be called
pulse processing gain (Foerster et al., 2001)
PG2 ¼
Tf
Tp
ð7:7Þ
The total processing gain is the product of these two contributions (Time Domain
Corporation, 1998)
PG ¼ PG1 PG2
ð7:8Þ
Since the duty cycle for DS-UWB is 100 %, therefore Tp ¼ Tf. Hence, the processing
gain in DS-UWB comes only from the pulse repetition.
UWB Theory and Applications
162
The user’s symbol rate (bit rate) Rs is given by (3.20).
Rs ¼
1
:
NTf
ð7:9Þ
Using (7.6) (7.7) and (7.8), for TH-UWB and DS-UWB, the user’s symbol rate in
equation (7.9) can be written
Rs ¼
1
1
1 ¼
:
¼
Tp =PG Tp =PG1 TH Tp =PG DS
ð7:10Þ
With the same bandwidth, data rate and pulse width, the low duty cycle of TH-UWB
signals means that TH-UWB requires a lower N than DS-UWB to achieve the same
processing gain. The advantage is therefore a higher allowable peak (pulse) power, and
so improved interference compared with other systems.
The SNR for a TH-PPM system in the presence NU of users, has been evaluated by
Scholtz and Win (1997), assuming random hopping sequences and modelling the multiple access interference (MAI) as additive white Gaussian noise (AWGN), according to
the standard Gaussian approximation. Under the previous assumptions, the bit SNR is
given by (Scholtz and Win, 1997; Win 2000
1
1
ðNU Þ ¼ ð1Þ þ M
Nu X
Ak 2
k¼2
Al
ð7:11Þ
where (1) is the SNR in the single-link case and Ak is the signal attenuation along the
path. The parameter M is given by (Scholtz and Win, 1997; Win 2000; Scholtz, 1993)
M 1 ¼
Mmp 2
2a
ð7:12Þ
where mp and 2a depend on the pulse waveform and the modulation parameter
DT (Scholtz, 1993; Scholtz and Win, 1997) see (3.10) and (3.11). The expressions of
mp and 2a are (Scholtz, 1997)
mp ¼
Z
1
wrec ðt TÞVðtÞdt
ð7:13Þ
1
2a
1
¼
Tf
Z
1 Z 1
1
2
wrec ðt sÞVðtÞdt ds
ð7:14Þ
1
where wrec (t) is the received monocycle waveform and V(t) is the embedded correlation
template signal defined as
Vðt; Þ ¼ wrec ðtÞ wrec ðt Þ:
ð7:15Þ
Medium Access Control
163
Using (7.9), (7.12) can be rewritten as
M 1 ¼ N
mp 2
1 mp 2
¼
:
2
Rs Tf 2a
a
ð7:16Þ
Using (7.12), (7.11) becomes (Scholtz and Win, 1997; Win, 2000)
Nu 2a X
Ak 2
:
ðNU Þ ¼ ð1Þ þ
Nm2p k¼2 Al
1
1
ð7:17Þ
The ratio
P ¼
ðNÞ
ð1Þ
ð7:18Þ
is the increase in power required to keep performance constant with increasing number
of users, NU. It can be seen that NU (DP) is monotonically increasing with DP. Under
perfect power control (Scholtz and Win, 1997; Win and Scholtz, 1998a; Win, 2000
1
NU;max ¼ lim NðPÞ ¼ bM 1 req
cþ1
P!1
ð7:19Þ
regardless of increase in transmit power.
The expression of the SNR depends on the pulse waveform correlation properties.
One common type of pulse waveform is the Gaussian pulse and its derivatives. Different
pulse waveforms have been considered in Chapter 3.
The transmit power is given by (Cuomo et al., 1999)
Pt ¼
Ew
,
Ns T c
ð7:20Þ
where Ew is the energy.
The expressions reported here are discussed in the following section.
7.3 Medium Access Control for Ultra-Wideband
The typical structure of cellular networks or access-point based networks is to have a
central coordinating node which may be reached by each node in the network. In ad-hoc
networks, the coordinator is dynamically selected among capable devices participating
the network. In fully ad-hoc networks, or multi-hop networks, there is no coordinator at
all and access is entirely distributed.
Regardless of the network topology, the role of the MAC is to control access of the
shared medium, possibly achieving targeted goals such as high performance (quality of
service provision), low energy consumption (energy efficiency), low cost (simplicity),
and flexibility (for example with ad-hoc networking capabilities).
UWB Theory and Applications
164
For all mobile applications, energy efficiency is an important consideration, and it is
crucial in sensor networks. Energy efficiency affects the MAC design by requiring the
protocol overhead due to signalling (connection set-up, changing configuration, etc.) to
be minimized.
This section outlines the design foundations for a MAC suitable for operation using UWB
technologies. The focus of the section is on the resource allocation problem (Figure 7.3).
The physical layer provides a bit stream to the upper layers, utilizing techniques and
signals to optimize the usage of the available channel. At the physical layer, no distinction is made with respect to the significance of the bits carried. The role of the MAC is
to arbitrate access to the resources made available from the physical layer.
The previous section presented expressions for key system parameters (such as the
number of users, bit rates and achieved SNR) in terms of UWB parameters (such as
pulse repetition rate, number of pulses per bit). The system parameters represent the
resources made available by the physical layer. By adjusting these system parameters, it
is possible to design a MAC tailored for UWB.
The wide range of values of the UWB parameters enables the design of highly
adaptive software defined radios (di Sorte et al., 2002) which may be matched to a wide
range of service requirements (Cuomo et al.). In addition to focusing on the resources to
be managed, possible constraints imposed by the UWB technology itself must be taken
into account.
7.3.1
Constraints and Implications of UWB Technologies on MAC Design
Some qualities of UWB signals are unique and may be used to produce additional
benefit. For example, the accurate ranging capabilities with UWB signals may be
exploited by upper layers for location-aware services. Conversely, some aspects of
UWB pose problems which must be solved by the MAC design. For example, using a
NETWORK AND HIGHER LAYERS
CONVERGENCE LAYER
segmentation and reassembly
CTRL PLANE
DATA PLANE
error
control
radio link
control
traffic
server
PHY
signalling
sensing
measurements
Figure 7.3
data transfer
Reference model
Medium Access Control
165
carrier-less impulse radio system, it is cumbersome to implement the carrier sensing
capability needed in popular approaches such as carrier-sense, multiple access with
collision avoidance (CSMA/CA) MAC protocols.
As seen in Chapter 4, another aspect that affects MAC design is the relatively long
synchronization and channel acquisition time in UWB systems. In (di Sorte et al., 2002),
the performance of the CSMA/CA protocol is evaluated for an UWB physical layer.
CSMA/CA is used in a number of distributed MAC protocols and it is also adopted in the
IEEE 802.15.3 MAC.
Figure 7.4 shows the packet traffic between two users in an UWB ad-hoc network.
The approach assumes a very simple CSMA/CA protocol. Synchronization preambles
must be sent at the beginning of each transmission burst. In order to exchange data with
the CSMA/CA protocol, the receiver must first synchronize and decode the RTS
(request to send) packet. The transmitter then needs to synchronize and decode the
CTS (clear to send) packet, before starting transmission. In Figure 7.4, the UWB
preambles are highlighted and denoted by ‘p’. Processing times are neglected.
A long preamble is needed in order to achieve synchronization for both the RTS and
CTS packets. A shorter preamble is possible for data transmission since synchronization
may be assumed to be maintained after reception of the RTS packet. Further packets
may be received with only fine corrections or tracking. Depending on the allowed length
of the data packet, a longer preamble may be needed for the following ACK packet.
Three preamble lengths (nominal, long, short) are proposed by Roberts (2003).
The time to achieve bit synchronization in UWB systems is typically high, of the order
of few milliseconds (di Sorte et al., 2002). Considering that the transmission time of a
10 000 bit packet on a 100 Mbit/s rate is only 0.1 milliseconds, it is easy to understand
the impact of synchronization acquisition on CSMA/CA-based protocols. The efficiency loss due to acquisition time can be minimized by using very long packets.
However this may impact performance in other ways.
When the effects of acquisition are taken into account, simulation results (di Sorte et al.,
2002) show that the performance of CSMA/CA using UWB is poorer than for narrowband and even wideband systems, in terms of delay, throughput and channel utilization.
A TX
p
RTS
DIFS
p
A RX
B TX
B RX
p
S
I
F
RTS S
p
S p
I
F
CTS S
DATA
CTS
p
DATA
S
I
F
S
p
ACK
propagation delay
Figure 7.4
Messages between sender A and destination B using the CSMA/CA protocol
UWB Theory and Applications
166
Acquisition preambles are typically sent with higher transmit power than data packets
(Kolenchery et al., 1998). This impacts both the interference level and the energy
consumption in highly burst traffic. This effect must be taken into account when
determining the efficiency of the system.
The adoption of CSMA/CA as a distributed protocol must be jointly evaluated with
the performance of the underlying UWB physical layer. In general it may not be a
suitable choice for an UWB MAC unless proper synchronization techniques are developed. One solution to this problem is the exploitation of the very low duty cycle of impulse
radio. Synchronization can be maintained during silent periods by sending low power
preambles for synchronization tracking (Kolenchery et al., 1998). This approach is
feasible only for communications between a single pair of nodes, which is not the case
in peer-to-peer networks.
7.3.2
Resource Allocation in UWB Systems
The previous section gave an overview of multiple access in UWB. TH-PPM has been
used to illustrate in detail how the UWB parameters influence system performance. Timehopping multiple access is particularly interesting in ad-hoc networks since it does not
require network synchronization, thus allowing asynchronous users (Win et al., 1997b).
This property is also a desirable feature in structured networks since it may relax the
system requirements. The expressions describing system parameters, such as data rate
given in Section 7.2.4 are valid for devices using a single time-hopping code. The results
may be extended in a straightforward manner for the case of devices using multiple codes.
From the expressions given in Section 7.2.4, it is clear that there is a trade off between
SNR (range) and data rate (throughput). For example, it can be seen that decreasing the
frame time, Tf, implies a linear increase in the transmitted power (Foerster et al., 2001).
However, the pulse repetition frequency and the peak pulse power must be varied
inversely to maintain constant average power. Hence, for a fixed Tf, to increase the
receiver SNR, it is necessary to increase N and hence reduce Rs, trading off between
SNR and data rate.
UWB systems allow devices to vary rate and performance simply by varying the
number of pulses per bit. As seen above, N is related to the interference rejection
capability of the user. This means that for loss-tolerant traffic, or for low-interference
environments, N can be adjusted to trade between quality and bit rate. Multi-code
allocation may be used to offer a wider range of achievable rates to a single terminal.
The UWB parameters listed in Table 7.1 affect MAC performance (di Benedetto
2001). These parameters can be adjusted to achieve the service requirement goals.
Cuomo et al. (1999) propose a resource allocation technique that optimizes system
performance indices by varying the MAC parameters within given constraints. The
authors start from the identification of UWB-specific MAC parameters to develop a
distributed protocol, which supports guaranteed quality and best effort traffic. According to that protocol, a transmitter computes the allowed transmit power so that the
interference to other users is kept below a threshold. This threshold is computed using
the tolerable interference declared by each device in the network. The transmit rate is a
constraint for the guaranteed quality, reservation-based traffic. The protocol determines
whether a transmit power exists for a given user that satisfies the global requirements of
Medium Access Control
167
Table 7.1 Some UWB parameters, impact on data rate, Rs, and transmit power, Pt, and
other comments
Parameter
Symbol Comment
Pulse shape
—
Pulse duration
Pulse repetition time
Tp
Tf
Chip time
Tc
Number of pulses per
information bit
Number of allocated
TH codes
Number of chips per
pulse repetition frame
N
Time shift
Period of TH code
Transmission rate
Nc
Nh
Np
Family of hopping codes
Type of data modulation
Pulse shape can be used to control
interference
Impacts on level of external interference
Impacts on coverage area (range)
(Foerster et al., 2001)
Impacts on Rs, Pt (constrained to technology
implementation)
Impacts on Rs (simple to adjust);
Impacts maximum power
Impacts on data rate
Impacts on Rs, Pt (see text; being an integer, no
continuous values can be achieved: quantization effect)
Impacts on UWB interference
Impacts on external interference
Increasing Np the power spectral density is
controlled
Impacts on UWB interference
the network. For best-effort traffic, a solution always exists. The algorithm identifies
suitable pairs of rate and power so that inequalities are satisfied.
7.4 IEEE 802.15.3 MAC
7.4.1
Introduction
The IEEE 802.15.3 working group on wireless personal area networks is developing a
high data rate MAC capable of operation using UWB transmission. The current version
(version 17) of the draft standard defines the physical layer and the MAC. The
draft standard has been developed primarily to support ‘classical’ narrowband systems,
and needs major changes to address UWB technology.
Within the IEEE working group is the study group (SG) 3a, which is the so-called
alternate physical layer examining UWB solutions. At present, there are two main
candidates for the IEEE UWB standard, neither of which has yet lead to concrete
proposals at the MAC level. The first is a multiband approach (as described in Chapter 3),
which is based on the flexible utilization of a number of UWB channels each of
approximately 500 MHz. With this technique, the generated UWB signal is reshaped
according to country-specific regulatory constraints and the presence of known narrowband interferers (such as WLAN devices). The second is based on impulse radio or
pulse-based UWB techniques. In both cases, a flexible, dynamic MAC protocol must be
developed to manage the system.
UWB Theory and Applications
168
To date, little has been done for LDR UWB communication systems, and significant
innovation is required to develop a low complexity MAC for these devices. A commonly
shared view of future UWB systems is that LDR devices capable of achieving ultra-low
power and ultra-low cost should include a processor-less MAC protocol, which may be
operated using simple hardware finite state machines. Work within the IEEE working
group 802.15.4 will attempt to address MAC issues for LDR devices.
7.4.2
Applications
The IEEE 802.15.3 standard is being developed for high speed applications including
.
.
Video and audio distribution:
-high speed DV transfer from a digital camcorder to a TV screen;
-High definition (HD) MPEG2 (or better) between video players/gateways and
multiple HD displays;
-home theatre audio distribution;
-PC to LCD projector;
-interactive video gaming.
High-speed data transfer:
-MP3 players;
-personal home storage;
-printers and scanners;
-digital still cameras.
To meet the above goals, the main characteristics of the IEEE 802.15.3 MAC are being
defined as
.
.
.
.
.
high rate WPAN with multimedia QoS provision:
-short range (minimum 10 m, up to 70 m possible);
-high data rates (currently up to 110 Mbit/s, to be increased by TG3a to 100–800 Mbit/s);
-TDMA super-frame architecture.
ad-hoc network with support for dynamic topology:
-mobile devices may often join and leave the piconet;
-short time to connect (< 1 s);
-peer to peer connectivity.
centralized and connection-oriented topology:
-the coordinator maintains the network synchronization timing, performs admission
control, assigns time for connection between devices, manages PS requests.
flexible and robust:
-dynamic channel selection;
-transmit power control per link;
-handover of the piconet coordinator (PNC) role among capable devices;
-multiple power saving modes (support low power portable devices).
authentication, encryption and integrity:
-CCM authenticate-and-encrypt block cipher mode using AES-128 (IEEE, 2004);
-support for upper layer authentication protocols (e.g., public key).
The main features are described in more detail in the following section.
Medium Access Control
7.4.3
169
Main Features
Although the IEEE 802.15.3 MAC is under consideration for use with UWB radio
technologies, it has not been specifically designed for UWB. This section presents the
IEEE MAC and considers its suitability as a MAC for a UWB system. The discussion
suggests that the UWB MAC considerations described in Section 7.3.1 are only partially
supported by the IEEE 802.15.3 MAC.
A wireless personal area network (WPAN) is a wireless ad-hoc data communications
system, which allows a number of independent data devices to communicate with each
other. A WPAN is distinguished from other types of data networks in that communications are normally confined to a person or object that typically covers about 10 metres in
all directions, and envelops the person or a thing whether stationary or in motion.
The group of devices in the IEEE 802.15.3 MAC is referred to as a piconet as
illustrated in Figure 7.5. The piconet is centrally managed by the coordinator of the
network, referred to as the piconet controller. The PNC always provides the basic
timing for the WPAN. Additionally the PNC manages the quality of service (QoS)
requirements of the WPAN.
A device (DEV) willing to join a piconet first scans for an existing PNC. The presence
of a PNC is detected by the reception of a beacon sent from the PNC on a periodic basis.
If no PNC is found, and if the device is capable of the task, the new device becomes
a PNC and starts a piconet itself. The PNC periodically sends network information via a
beacon. Other devices that can receive the beacon may associate to the PNC. Associated
devices can exchange data directly amongst themselves, i.e. without using the central
node as a relay, but resources are managed centrally by the PNC. The role of PNC can
be handed over to an elected device according to a pre-determined protocol.
A compliant physical layer may support more than one data rate. In each physical
layer there is one mandatory base rate. In addition to the base rate, the physical layer
may support rates that are both faster and slower than the base rate. A DEV will send a
frame with a particular data rate to a destination DEV only when the destination DEV
is known to support that rate.
The IEEE 802.15.3 MAC is based on a dynamic TDMA structure. The channel time
is divided into periods referred to as ‘super-frames’ as shown in Figure 7.6. The
DEV
DEV
PNC
DEV
DEV
Figure 7.5
A piconet includes a piconet coordinator (PNC) and associated devices (DEVs)
UWB Theory and Applications
170
structure and duration of the super-frames may vary from frame to frame. The superframe is further divided into time slots and organized as below.
During the beacon time, the PNC broadcasts the beacon frame, which includes all the
necessary network information for all devices of the piconet. The information includes
allocation of the following time slots of the super-frame to a source-destination pair,
based on the channel request commands sent by the DEVs. Either the source or the
destination may be a broadcast address.
The contention access period (CAP) is used according to the distributed carrier
sensing multiple access with collision avoidance (CSMA/CA) protocol. The CAP can
be used for commands or asynchronous traffic.
The channel time allocation period (CTAP) is used for contention-less access of
asynchronous and isochronous data streams. Channel time allocation (CTA) periods
and management CTAs are allocated by the PNC. The CTAs allow devices to communicate without interference from other devices in the network.
The PNC may choose to allocate optional management CTAs (MCTAs), instead of
the CAP, for sending commands to and from the PNC. Commands from DEVs to the
PNC are sent in a so-called open MCTA, which has the broadcast address in the source
field. The access mechanism in open MCTAs and in association MCTAs is slotted
Aloha, whereas it is TDMA in other MCTAs. The association MCTAs are used by
channel time
superframe m-1
Figure 7.6
beacon
CAP
MCTA
MCTA
superframe m
superframe m + 1
CTAP
CTA
CTA
CTA
The super-frame structure includes contention and contention-less periods
Medium Access Control
171
DEVs wishing to join a piconet. Such DEVs use the MCTA to send their association
messages.
CTAs are generally dynamic, meaning that their timing and position in the superframe may change on a frame-by-frame basis. Conversely, pseudo-static CTAs are used
so that their position and duration remain relatively constant. A third type, private
CTAs, is reserved for uses other than communication, for example, to allow space for
dependent piconets. The IEEE 802.15.3 MAC includes the possibility of creating socalled ‘child piconets’. This is a feature that can be used to extend the coverage area by
multiple hops, reducing the transmit power to lower the interference level. The frame
structure for the child piconet is shown in Figure 7.7.
The PNC inserts guard times between consecutive CTAs (Figure 7.8). This guard time
is defined so that the time separation between CTAs is always at least one SIFS (short
inter-frame spacing). The required guard time depends on the maximum drift between
DEVs’ local time and the ideal time. It is therefore a function of the propagation delay
and of the time elapsed since the previous synchronizing event, which is the start of the
preamble of the beacon. The elapsed period from the synchronization event depends on
the super-frame length. This means that clock drifts are a constraint for the super-frame
length, and therefore for number and size of CTAs.
For data integrity, the IEEE draft standard includes three possible ACK policies.
With No-ACK, the sending device does not expect any kind of guarantee that the frame
is successfully received. With immediate Imm-ACK, the sending device requests and
waits for an acknowledgment from the destination before sending the subsequent
packet. Finally, the delayed Dly-ACK policy allows sending devices to send a number
of frames while waiting for acknowledgement of the first transmitted frame. The buffer
Parent piconet superframe
CTA2
(private)
CTA
1
CTA
3
CTA
n
be
ac
on
CAP
CAP
be
ac
on
CTA
1
CTA2
Child piconet superframe
Reserved time
be
ac
on
C
A
P
CTA
1
CTA
k
be
ac
Reserved time
on
Communications rules
None
C-P
None
C-C
C-P
None
C-P
None
Child network member
communication between: DEV &
DEV, or PNC & DEV
Communication between child PNC & parent (DEV or PNC)
No peer to peer communication during beacon times
Figure 7.7
Child piconet frame structure (source IEEE 802.15.3)
UWB Theory and Applications
172
drifts
CTA (n)
CTA (n + 1)
SIFS
SIFS
guard time
Figure 7.8 CTAs are separated by a guard time sufficient for compensating timing drifts
window size is negotiated between source and destination. The standard allows switching between Imm- and Dly-ACK
Figure 7.9 shows the reference model used in the IEEE draft standard. In the IEEE
802.15.3 MAC IEEE 2003, the resource reservation can be implemented exploiting the
functionalities of MAC layer management entity (MLME) and device management
entity (DME).
To reduce energy consumption, DEVs are allowed to enter a sleep mode, which may
span a number of super-frames. The collision-free period (Channel Allocation Time
Period) is TDMA based. The impact of synchronization acquisition time on CSMA/CA
performance indicates that the usage of the CAP for asynchronous data transfer should
be carefully considered.
7.4.3.1
UWB Considerations
As mentioned, the IEEE 802.15.3 MAC is not specifically designed for UWB. As a
consequence, a number of issues must be considered when attempting to utilize UWB
with the MAC.
Firstly, it is difficult for an UWB radio to detect when another DEV is transmitting.
This may make much of the signalling difficult or impractical to implement. It may also
mean that the CAP’s may be zero length as multiple access is less of a problem than for
conventional narrow-band systems.
FCSL
MAC
DME
MLME
PHY
Figure 7.9 The reference model used in the draft standard
Medium Access Control
173
Secondly, the high data rates supported by UWB will allow a large number of devices
to be in a piconet. Currently, the maximum number of unique DEV IDs in a piconet is
243, which will potentially limit the capacity of the UWB system.
Finally, high-data-rate UWB systems require efficient channel management to ensure
high throughput. To maximize throughput, a physical-layer aware MAC should allow
larger packet sizes and minimal spacing between frames from the same device type. The
MAC protocols also need to be optimized to minimize the time required to associate
with a piconet, quickly set up isochronous and asynchronous streams between DEVs,
and monitor channel quality for optimum performance. This is something that the
IEEE 802.15.3 MAC does not deal with particularly well for UWB systems.
7.5 Conclusions
At the time of writing, no standard exists for an UWB MAC. There is work in
standardization bodies to adapt and possibly optimize and enhance existing MAC for
use over UWB. Partly the reason is due to the inherent complications of using a
physical-layer signal, which is difficult to detect and difficult to synchronize with for
intended users. Efforts have been made to develop MACs that allow efficient operation
for large numbers of devices. However, this work is still at an early stage.
An example of this is IEEE 802.15.3 MAC. Its critical points as a MAC for UWB
have been outlined in this chapter. Whilst developed for high-speed, ad-hoc, network
applications, the MAC has been targeted for UWB by interested groups. The suitability
for UWB, either impulse based or for other physical layer UWB solutions, is yet to be
proven.
Research on MAC for UWB is needed and is going on in industry research centres
and in universities, although in research on MAC for UWB peculiarities of UWB
signalling are rarely taken into account. As has been shown, limitations due to specific
characteristics of UWB physical layer do exist and attention must be paid to these in
MAC design. The potential of UWB systems to offer flexible data rates, large numbers
of users and positioning services are critically dependent on the capabilities of the MAC.
8
Positioning
Keqen Yu, Ian Oppermann
8.1 Introduction
The very short time-domain pulses of UWB systems make them ideal candidates for
combined communications and positioning. The duration of a pulse is inversely proportional to the bandwidth of the transmitted signal. If the time of arrival of a pulse is
known with little uncertainty, then it is possible accurately to estimate the distance
travelled by the pulse from the source. By combining the distance estimates at multiple
receivers, it is possible to use simple triangulation techniques to estimate the position of
the source.
For UWB systems with potential bandwidths of 7.5 GHz, the maximum time
resolution of a pulse is of the order of 133 picoseconds. Therefore, when a pulse
arrives, it is possible to know to within 133 picoseconds the ‘time-of-flight’ of the
pulse. This time uncertainty corresponds to 4 cm spatial uncertainty. For more
modest bandwidths of 500 MHz, the corresponding time resolution is 2 nanoseconds,
which corresponds to a spatial uncertainty of approximately 60 cm. With any UWB
signal, therefore, it is potentially possible to achieve sub-metre accuracy in positioning, provided the time-space uncertainties can be combined from multiple sources
without significant loss. The corresponding location estimate will be subject to the
cumulative errors of each of the distance estimates as well as any uncertainty or
errors introduced by the positioning technique itself. Different techniques exist for
determining position from time-of-arrival or time-of-flight estimates, and each technique has strengths and weaknesses.
The goal of this chapter is to explore positioning techniques as well as to look at some
of the practical problems associated with positioning in UWB systems.
UWB Theory and Applications Edited by I. Oppermann, M. Hämäläinen and J. Iinatti
Ó 2004 John Wiley & Sons, Ltd ISBN: 0-470-86917-8
UWB Theory and Applications
176
8.2 Positioning Techniques
8.2.1
Time-based Positioning
Positioning based on the use of radio signals has a long history. There are many position
estimation techniques for various purposes under different scenarios. Signal strength,
angle-of-arrival (AOA), time measurements (TOA, time-of-flight, and time difference of
arrival (TDOA)) can all be exploited for the position estimation. Figure 8.1 shows
a general positioning system configuration with base stations or sensors and the device
to be located (tag). The example below is for a cellular system. However, the concept
is generally applicable.
8.2.2
Overview of Position Estimation Techniques
The most straightforward way to estimate the position is directly solving a set of
simultaneous equations (Fang, 1990) based on the TDOA measurements. Therefore,
exact solutions can be obtained for two dimensional (2-D) positioning with three
sensors/two TDOA measurements, and for three dimensional (3-D) positioning with
four sensors. For an overdetermined system (with redundant sensors), Taylor series
expansion may be used to produce a linearized, least-square solution iteratively to the
position estimate (Torrieri, 1984). However, to maintain good convergence, the Taylor
series method requires quite an accurate initial position estimate, which is often difficult
to obtain in some practical applications.
Base Station
d1
Base Station
d2
Base Station
Figure 8.1
General triangulation example
Measurement
error margin
Positioning
177
To avoid the convergence problem, several different approaches have been proposed
such as the spherical interpolation (Friedlander, 1987; Schau and Robinson, 1987;
Smith and Abel, 1987) and the double maximum likelihood (ML) method (Chan and
Ho, 1994). The hyperbolic positioning methods (except for the Taylor series method)
share one common drawback of producing multiple solutions. They also require knowledge of the variance and the distribution of the TOA estimation error (except for the
direct method). A different method for positioning is the use of non-linear optimization
theory. Gauss–Newton method, Levenberg–Marquardt method, and quasi-Newton
method, including the DFP formula (Fletcher and Powell, 1963) and the BFGS formula
(Broyden, 1970) can all be employed for position estimation iteratively. The DFP
algorithm has been used e.g. in the UWB precision assets location system developed
by Multi-spectral Solution, Inc.
For practical applications, the position-estimation algorithm should be robust and
easy to implement. To achieve this, we choose the direct calculation method and the
non-linear optimization algorithm for further investigation. These two methods do not
require knowledge of the variance or distribution of the TOA estimation error. They
also do not require an accurate initial position estimate.
8.2.3
Direct Calculation Method
In the Cartesian system, the range (distance) between sensor i and the tag is given by
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðx xi Þ2 þ ðy yi Þ2 þ ðz zi Þ2 ¼ cðti t0 Þ
i ¼ 1; 2; 3; 4
ð8:1Þ
where (x, y, z) and (xi, yi, zi) are the coordinates of the tag and the sensor respectively, c is
the speed of light, ti is the signal TOA at sensor i, and t0 is the unknown transmit time at
the tag/device. In the development of the expressions, we ignore the difference between
the true and the measured TOAs for simplicity. Squaring both sides of (8.1) gives
ðx xi Þ2 þ ðy yi Þ2 þ ðz zi Þ2 ¼ c2 ðti t0 Þ2
i ¼ 1; 2; 3; 4:
ð8:2Þ
Subtracting (8.2) for i ¼ 1 from (8.2) for i ¼ 2,3,4 produces
1
1
ði1 2xi1 x 2yi1 y 2zi1 zÞ i ¼ 2; 3; 4
ct0 ¼ cðt1 ti Þ þ
2
2cðt1 ti Þ
where
xi1 ¼ xi x1
yi1 ¼ yi y1
zi1 ¼ zi z1
i1 ¼ x2i þ y2i þ z2i ðx21 þ y21 þ z21 Þ:
ð8:3Þ
UWB Theory and Applications
178
Define the TDOA between sensors i and j as
tij ¼ ti tj
Eliminating t0 from (8.3) yields
a1 x þ b1 y þ c 1 z ¼ g1
ð8:4Þ
where
a1 ¼ t12 x31 t13 x21
b1 ¼ t12 y31 t13 y21
and
c1 ¼ t12 z31 t13 z21
1
g1 ¼ ðc2 t12 t13 t32 þ t12 31 t13 21 Þ:
2
a2 x þ b2 y þ c 2 z ¼ g2
ð8:5Þ
where
a2 ¼ t12 x41 t14 x21
b2 ¼ t12 y41 t14 y21
c2 ¼ t12 z41 t14 z21
1
g2 ¼ ðc2 t12 t14 t42 þ t12 41 t14 21 Þ:
2
Combining (8.4) and (8.5) yields
x ¼ Az þ B
ð8:6Þ
where
b1 c 2 b2 c 1
a1 b2 a2 b1
b2 g1 b1 g2
B¼
a1 b2 a2 b1
A¼
and
y ¼ Cz þ D
where
a2 c 1 a1 c 2
a1 b2 a2 b1
a1 g2 a2 g1
D¼
:
a1 b2 a2 b1
C¼
ð8:7Þ
Positioning
179
Then, substitution of (8.6) and (8.7) back into (8.3) with i ¼ 2 produces
cðt1 t0 Þ ¼ Ez þ F
ð8:8Þ
where
1
ðx21 A þ y21 C þ z21 Þ
ct12
ct12
1
þ
F¼
ð2ðx21 B þ y21 DÞ 21 Þ:
2ct12
2
E¼
Substituting (8.6), (8.7) and (8.8) back into (8.1) for i ¼ 1 followed by squaring
yields
Gz2 þ Hz þ I ¼ 0
ð8:9Þ
where
G ¼ A2 þ C 2 E 2 þ 1
H ¼ 2ðAðB x1 Þ þ CðD y1 Þ z1 EFÞ
I ¼ ðB x1 Þ2 þ ðD y1 Þ2 þ z21 F 2 :
The two solutions to (8.9) are
H
z¼
2G
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
H
I
:
2G
G
ð8:10Þ
The two estimated z values (if both are reasonable) are then substituted back into (8.6)
and (8.7) to produce the coordinates x and y, respectively. However, there is only one
desirable solution. We can dismiss one of the solutions if it has no physical meaning or it
is beyond the monitored area. If both solutions are reasonable and they are very close,
we may choose the average as the position estimate. Otherwise, an ambiguity occurs.
Other cases of no acceptable results include two complex solutions, or both solutions
are beyond the monitored area. To increase the probability of the existence of one
reasonable position estimate, we may add a fifth sensor. Then we have five different
combinations, producing five different results.
In practice, there may be more than five sensors. In this case, we may choose the five
sensors with the highest received signal powers. This simple method is particularly
suitable when the noise level is similar in the received signals. When a sequence of
measurements is available at each sensor, accuracy may be further improved by first
processing the sequence of measurements such as averaging.
UWB Theory and Applications
180
8.2.4
Optimization Based Methods
There are many schemes and techniques in non-linear optimization. In this section, we
are interested in examining several practical optimization methods and applying them to
our three-dimensional position estimation.
8.2.4.1
Objective Function
An objective function is normally required for any optimization algorithms. Since
the ultimate aim of positioning is to obtain an accurate position estimate, it is
natural to define the objective function as the sum of the squared range errors of all
sensors
Fðx; y; z; t0 Þ ¼
N
1X
f 2 ðx; y; z; t0 Þ:
2 i¼1 i
where N is the number of the active sensors/base stations, (x,y,z) is the unknown
position coordinates, t0 is the unknown transmit time (dummy variable), and
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
fi ðx; y; z; t0 Þ ¼ ðx xi Þ2 þ ðy yi Þ2 þ ðz z2i Þ cðti t0 Þ:
Here ti is the estimated TOA at the ith sensor and c is the speed of light. The optimization purpose is to minimize this objective function to produce the optimal position
estimate. For notational simplicity, we define
p ¼ ðx; y; z; t0 ÞT
fðpÞ ¼ ðf1 ðpÞ; f2 ðpÞ; . . . ; fN ðpÞÞT :
Then the objective function becomes
1
f ðpÞ ¼ jjfðpÞjj2 :
2
8.2.4.2
Gauss–Newton Method
Expanding the objective function in the Taylor series at the current point Pk and taking
the first three terms, we have:
1
Fðpk þ sk Þ Fðpk Þ þ gTk sk þ sTk Gðpk Þsk
2
ð8:11Þ
Positioning
181
where Sk is the directional vector (or increment vector) to be determined, gk is a vector
of the first partial derivatives (also called gradient) of the objective function at Pk
gk ¼ rf ðx; y; z; t0 Þjp ¼ pk
@f
@f
@f
@f
j
¼
; j
; j
;
j
@x p ¼ pk @y p ¼ pk @z p ¼ pk @t0 p ¼ pk
and G(Pk) is the Hessian of the objective function. Minimization of the right-hand side
of (8.11) yields
GðpÞk sk ¼ gk :
ð8:12Þ
The minimization in which Sk is defined by (8.12) is termed Newton’s method. To avoid
the calculation of the second order information in the Hessian, a simplified expression
can be approached from (8.12), resulting in
JkT Jk sk ¼ JkT fðpk Þ
ð8:13Þ
where Jk is the Jacobian matrix of f(P) at Pk. This is called the Gauss–Newton method.
When Jk is full rank, which is the usual case of an over-determined system, we have the
linear least-squares solution
sk ¼ ðJkT Jk Þ1 JkT fðpÞk :
ð8:14Þ
The Gauss–Newton method may get into trouble when the second-order information in
the Hessian is not trivial. A method that overcomes this problem is the Levenberg–
Marquardt method. The Levenberg–Marquardt search direction is defined as the solution of the equations
ðJkT Jk þ IÞsk ¼ JkT fðpk Þ
ð8:15Þ
where is a non-negative scalar that controls both the magnitude and direction of Sk.
8.2.4.3
Quasi-Newton Method
This type of method is like Newton’s method. The Hessian matrix G(Pk) in (8.12) is now
approximated by a symmetric positive definite matrix Bk, which is updated from
iteration to iteration. At the kth iteration, set
Sk ¼ Bk gk :
ð8:16Þ
Using line search along Sk to produce
pkþ1 ¼ pk þ sk
ð8:17Þ
UWB Theory and Applications
182
where is the step size. Then updating Bk yields Bkþ1. The initial matrix B1 can be any
positive definite matrix. It is usually set to be an identity matrix in the absence of any
better estimate. There exist different ways to update Bk. One well-known updating
formula is the DFP (Davidon–Fletcher–Powell) formula, in which Bk is updated
according to
Bkþ1 ¼ Bk þ
hk hTk Bk qk qTk Bk
T
qk B k qk
hTk qk
ð8:18Þ
where
hk ¼ pkþ1 pk
qk ¼ gkþ1 gk :
The BFGS quasi-Newton algorithm was also considered. This is significantly more
complicated than the DFP algorithm. Preliminary simulation results demonstrated that
the performance of the BFGS algorithm is not better than the DFP algorithm. As
a result, only results for the DFP algorithm are presented.
To start the iteration for any of the above mentioned algorithms, the initial position
coordinates and the initial transmit time are required. In the absence of any better
estimate, the initial estimated values of the position coordinates may be chosen to be the
mean position of all the active sensors or the area being monitored. The initial estimated
transmit time may be chosen to be some time point earlier than the earliest receive time,
which will depend on the dimension of the monitored area. Both the step size and the
first derivatives g (the gradient) are updated during each iteration. The performance of
the above optimization-based methods could be improved if a good initial position
estimate is available. This may be achieved by exploiting the results from the direct
calculation method or other non-iterative methods.
8.2.5
Simulation Results
A simulation model was developed to examine the relative performance of the
different positioning techniques. The tool models a closed region with a number of
sensors and examines the accuracy with which a device in the area may be positioned
in 3-D. The monitored area has been assumed to have a dimension of
30 m 40 m 5 m. The positions of the sensors (delay estimation points) and the
tag (device of interest) are randomly generated to obtain average performance. The
performance evaluation is first performed by assuming that the TOA measurement
error is an i.i.d. random-variable Gaussian distributed with zero mean and variance s2.
At each value of s examined, 1000 simulation runs are conducted with random positions
for the sensors and the tag at each run. The performance is then averaged. It is known
that the location of the base stations can have significant impact on the positioning
performance. The case of one particular location of the base stations is also examined.
This location is randomly chosen from 50 locations of the base stations which produce
the best results.
Positioning
183
The performance is evaluated in terms of the root-mean-square (RMS) error and the
failure rate. The RMS error is calculated according to
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
Np X
Ns
u 1 X
t
½ðxðiÞ x^ðijÞ Þ2 þ ðyðiÞ y^ðijÞ Þ2 þ ðzðiÞ z^ðijÞ Þ2 3Np Ns i¼1 j¼1
where Np is the number of different position combinations of the sensors and the tag
and Ns is the number of TOA samples at each s for each position combination. (x(i), y(i), z(i))
y(ij) ,^
z(ij) ) are the true and estimated position coordinates of the tag/device
and (^
x(ij) ,^
of interest, respectively. The failure rate accounts for the cases where there is no solution
or the solution is unreasonable. In the direct calculation, the cases include: the solutions
are beyond the monitored area; both solutions are complex valued; the two solutions are
reasonable but not close to each other. With the optimization-based methods, it
accounts for the situations where the algorithm does not converge to a solution, the
maximum number of function evaluations or iterations is exceeded, or the results are beyond
the area.
Figures 8.2 and 8.3 show the corresponding root-mean-square error of the three
coordinates and failure rate for a 4/5 sensor system with Gaussian distributed TOA
estimation error. DFPran and DFPsel denote the average RMS error of the x, y and z
coordinate estimation using the DFP algorithm with five sensors with randomly chosen,
and selectively chosen sensor positions respectively. LMran and LMsel are the
3
DIRECTran
DFPran
LMran
DIRECTsel
DFPsel
LMsel
(Root Mean Sqaure Error)
2.5
2
1.5
1
0.5
0
0
0.5
1
1.5
Standard Deviation of TOA Estimation Error (ns)
Figure 8.2
Root-mean-square error of position estimation
UWB Theory and Applications
184
1
DIRECTran(4)
DIRECTsel(4)
LMran
DFPran
DIRECTran(5)
DFPsel
LMsel
DIRECTsel(5)
Failure Rate
0.8
0.6
0.4
0.2
0
Figure 8.3
0
0.5
1
1.5
Standard Deviation of TOA Estimation Error (ns)
Position estimation failure rate for four- and five-sensor system
corresponding results using the Levenberg–Marquardt method, while DIRECTran and
DIRECTsel are the corresponding results with the direct calculation using five sensors.
The RMS error of the direct calculation method using four sensors is nearly the same as
the case of five sensors, so it is not plotted. Clearly, for the direction-calculation
method, the failure rate decreases dramatically with five sensors compared with the
case of four sensors, although the accuracy is nearly the same. Of course, the failure rate
improvement is achieved at the cost of increased computation complexity and system
complexity, due to the addition of one extra sensor. The relatively large position error
for small TOA estimation error in the DFP algorithm and the Levenberg–Marquardt
method may come from the very crude initial position estimate. It seems the direction
calculation method is suitable for quite small TOA estimation error, while the quasi-Newton
algorithm and the Levenberg–Marquardt method are suitable for relatively large TOA
estimation error. The Gauss–Newton method works very poorly so that the corresponding results are not presented.
Figures 8.4 and 8.5 show the results for the position estimation combined with delay
estimation. A two-step technique is employed to speed up synchronization. Each data
bit in the preamble assigned for synchronization consists of two separated pulse
sequences, one of which is not coded and the other is coded by a m-sequence, and
a large non-pulse region. The first step is to provide a rough timing message using the
uncoded pulse sequence based on a moving average scheme with a window equal to the
duration of the uncoded pulse sequence. Once the first step is completed, the approximate location of the coded pulses is available and the code acquisition proceeds to
achieve fine synchronization. We choose the direct-sequence spreading with spreading
gain 31 (i.e. 31 chips per data bit). The pulse width is 0.4 nanoseconds and the duty cycle
is 1/14. The sampling rate is 6 GHz.
Positioning
185
DIRECTran
DFPran
LMran
DIRECTsel
DFPsel
LMsel
Root mean square error
2
1.5
1
0.5
0
Figure 8.4
0
5
10
Average SNR (dB)
15
Root-mean-square error of position estimation using estimated TOA
0.18
DFPran
LMran
DIRECTran
DFPsel
LMsel
DIRECTsel
0.16
0.14
Failure rate
0.12
0.1
0.08
0.06
0.04
0.02
0
Figure 8.5
0
5
10
Average SNR (dB)
15
Position estimation failure rate for five-sensor system using estimated TOA
8.3 Delay Estimation Techniques
Conventional delay estimation uses a variety of techniques to improve the estimation of
the position of a multipath component in a received signal. The delay estimation
approaches typically attempt to detect the presence of ‘L’ multipath components.
Figure 8.6 shows an example impulse response measurement for LOS conditions. As
UWB Theory and Applications
186
Figure 8.6
Measured channel impulse response (with and without window filtering)
was seen in Chapter 2, the UWB channel is very multipath rich, which makes detection
of the direct component difficult.
Delay estimation approaches used for UWB are typically very simple due to the
multipath-rich nature of the channel. A simple threshold detector may be used to detect
the presence of the signal. However, much of the advantage of the small time uncertainty is lost with such a simple technique. The rich channel requires significant signal
processing to produce an accurate estimate of time of arrival of the direct component.
Several delay estimation techniques will be described here for completeness.
8.3.1
General Approaches
The delay estimation approaches generally used are interference/inter-path cancellation
based on recognizing the shape of the band limited transmitted pulse such as described
in (Moddemeijer, 1991). This approach is robust, but does not lead to enormous
improvements in initial delay position estimation.
Subspace techniques such as those presented by Jakobsson et al., (1998) are extremely
complex, requiring the generation of several correlation matrices and their inverses, and
ultimately performing a large number of matrix multiplications to achieve a delay estimate.
They also tend not to work well in static or slowly moving channels (Latva-aho, 1998).
Positioning
187
For example, eigenvector decomposition is a form of subspace technique (Manabe
and Takai, 1992). This delay estimation approach requires calculation of the eigenvectors of the channel correlation matrix. Again, it is very complex.
8.3.2
Inter-path Cancellation
When using short spreading sequences, the autocorrelation side-lobes have non-zero
values which contribute to each ‘apparent’ multipath component. As a consequence, all
multipath components within at least one symbol period of the first ‘detected’ multipath
will need to be considered.
The technique requires interpolation between samples and good knowledge of the
shape of the band-limited single path signal.
The technique may be described as:
(1)
(2)
(3)
(4)
identify shape of band limited autocorrelation (template);
perform autocorrelation of received signal (this is the channel impulse response);
estimate AWGN power;
over-sample/interpolate to desired temporal resolution;
Loop
(5)
(6)
(7)
(8)
identify largest multipath component in received signal;
scale main lobe of template to fit (5);
subtract (6) from (5);
repeat Loop until all paths identified in area of interest or signal are below noise.
Once an estimate of the AWGN had been removed, the individual ‘resolvable’ multipath
components are identified. This is achieved by selective identification and removal of the
largest multipath components. Identification of large multipath components uses the fact
that, given the finite bandwidth, each multipath has a characteristic shape. A template
with this shape is fitted to the channel magnitude, and each large multipath component
identified is subtracted from the impulse response magnitude. With some tolerance for
numerical and measurement accuracy, anything remaining after the large multipath
component (MPC) is removed must be due to other multipath components. The same
treatment is therefore applied to the magnitude of the residue of the impulse response.
This technique requires the centre of the multipath component to be estimated from
the samples that indicate the position of a given multipath component. Interpolated
‘impulse templates’, with resolutions higher than the received signal sampling rate, may
be used to improve the accuracy of this approach. These are then correlated with the
received magnitude until a maximum value is reached. The scaled template is then
subtracted from the received signal.
Figure 8.7 shows an example of an impulse response which has had noise removed
and multipath components detected. If two multipath components are separated by one
sample, and less than approximately 8 dB, then they will be resolved. The power
requirement comes from the error sensitivity mentioned earlier. Once an identified
impulse is removed, a margin of error is allowed for to account for system noise and
UWB Theory and Applications
188
10–2
potentially hidden
direct component
Received Amplitude [V]
10–3
10–4
10–5
10–6
10–7
20
Figure 8.7
30
40
50
60
70
Sample Number
80
90
100
Multipath components identified in a measured impulse response
finite resolution of the measurements. If the residual signal exceeds this margin of error,
the process will examine the residue signal for further impulses.
Since the measured channel profiles are relatively static, any multipath component should
be present in several consecutive symbols. If a given sample multipath is not detected in
several previous or later measured symbols, it can be considered as noise and removed.
One major weakness of this technique is the identification of the first multipath
component if the largest multipath is close to the direct component. In Figure 8.7, if
the direct component is smaller than approximately 8 dB and before the largest peak, it
will be missed. The largest peak would be presented as the direct component, and an
error would be introduced into the results.
8.4 NLOS Conditions
8.4.1
Sources of Uncertainty
The main physical implementation details which affect uncertainty regions are:
.
.
.
.
.
oscillator accuracy and drift;
walls and obstructions increasing apparent path length;
received signal strength;
accuracy of delay estimation technique;
false readings from interference and multipath.
Positioning
189
Propagation of the UWB signal through walls will lead to additional delay before
arriving at the receiver. Denser wall material will cause the transmitted signal to travel
more slowly than the signal would otherwise travel through free space. Transmission
through walls will also lead to significant additional attenuation, which will reduce the
received strength and lead to increased errors in estimation. The impact of the wall
depends on the thickness of the wall and the angle of arrival of the UWB pulse. For very
high accuracy positioning systems, this NLOS situation needs to be avoided or
compensated for.
The total timing uncertainty at the receiver may be given by
total ¼ osc þ
X
i ðL; Þ þ meas ðÞ
ð8:19Þ
ii
where tosc is the timing error introduced by oscillator mismatch; ti is the delay introduced by wall ii, and is a function of wall thickness, L and incident angle , and tmeas is
the limit of measurement technique and is a function of signal-to-noise ratio.
The triangulation techniques described above can be enhanced through the use of
a number of supporting techniques, including:
.
.
.
.
weighting error region based on confidence;
confidence based on received signal-to-noise ratio and impulse profile;
log-likelihood measure based on SNR of direct component used for weighting;
negative log-likelihood measure based on amplitude ratio of direct-to-largest multipath to detect NLOS conditions.
8.4.2
Delay Through Walls
The excess delays caused by the obstructing wall material can be estimated with the help
of the environment layout. The propagation ta delay is given by the expression
0 ¼ a þ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
L
L2 þ K 2
;
¼ a þ
c cosðÞ
c
ð8:20Þ
where 0 is the measured delay, L is the antenna separation, K is the distance from
normal to wall, c is the velocity of the radio wave and is the incident angle as shown in
Figure 8.8.
A good estimate for the total transmission delay through the wall is
¼ a þ
Lw
þ 0;
c cosðÞ
ð8:21Þ
where t0 is the excess delay caused by the interior wall, Lw is the wall thickness, and is
the incident angle, c is the speed of light in a vacuum.
UWB Theory and Applications
190
Lw
L1
L0
TX
α
RX10
RX00
α'
K0
K1
RX01.
β
RX11
RX'11
Figure 8.8
Propagation mechanism through isotropic medium
Excess delay t0 can be approximated by an inverse cosine function (from the curve fit):
0 ¼ m þ
A
;
cosðÞ
ð8:22Þ
where tm is the delay caused by the medium at incident angle 0 and A is an empirical
constant. Figures 8.9 and 8.10 show the results of a simple measurement for an indoor
environment with a single brick wall (University of Oulu lecture room). The measurements are performed over the range 2 to 8 GHz. Figure 8.9 shows the additional delay
compared with free space, while Figure 8.10 shows the additional path loss for the 2–3
GHz, 3–8 GHz and full band versus angle of incidence, where 0 degrees represents the
UWB pulse striking the wall at the perpendicular. The additional time delays are of the
order of 0.5 to 0.7 ns, which corresponds to 15 to 21 cm of spatial error. The additional
path loss values become quite high at higher angles, due to the substantially greater path
travelled through the wall.
8.5 Metrics for Positioning
The two most important issues for consideration of the accuracy of a received signal are the
signal-to-noise ratio and the number of obstructions the signal has passed through to arrive
at the receiver. Obstructions are important as each material affects the speed of light and so
increases propagation delay and hence alters the perceived distance from the receiver.
Positioning
191
0.8
SA118
L5
Curve fit
0.75
Excess Delay [ns]
0.7
Excess delay = 0.35 + 0.1/cos(alfa)
0.65
0.6
0.55
0.5
0.45
0.4
–20
–10
0
Figure 8.9
10
20
30
40
Incident Angle
50
60
70
80
Measured additional delay for UWB signal
At the receiver, we are interested in detecting the first received signal component
irrespective of the received signal strength. If we have LOS conditions, then the first
component will be the strongest and is unaffected by transmission through walls or
other obstructions. An example of a LOS channel response profile produced by a raytracing simulator is shown in Figure 8.11. In this case, the confidence with which we can
detect the correct position is closely related to the signal-to-noise ratio. A high signal-tonoise ratio implies very accurate estimation of the centre of the first received channel
path. Let us define a channel confidence metric as
P ¼ log10
j a0 j 2
2
!
ð8:23Þ
where a0 is the amplitude of the direct signal component and 2 is the background noise
power.
8.5.1
Identifying NLOS Channels
Later multipath components arrive at the receiver as a consequence of reflection or
diffraction caused by objects between the transmitter and the receiver. The additional
path travelled and the mechanisms, combined reflection and/or diffraction, introduce
UWB Theory and Applications
192
0
Received Direct Path Power [dBm]
–2
–4
–6
–8
–10
–12
–14
–16
–18
–20
20
40
60
50
Incident Angle [degrees]
30
70
80
Figure 8.10 Additional transmission loss for through-wall UWB signals (upper , lower +
and total band)
–75
–80
Received Signal Power (dBm)
–85
dn path loss
–90
–95
–100
–105
–110
–115
–120
0
50
100
150
200
250
Delay (ns)
Figure 8.11 Example of LOS multipath profile
300
350
Positioning
193
additional loss compared with the direct component. Therefore, later multipath components should have lower power than the direct component. Figure 8.12 shows an
example of a NLOS channel.
If later multipath components have similar or higher power than the direct component, then the direct component must have travelled through some obstruction, leading
to additional attenuation. If a NLOS condition is detected, some additional delay in the
propagation of the direct component can be expected, leading to an error in the
perceived position. We can include the detection of NLOS conditions in our confidence
metric by comparing the amplitude of the direct component to the largest received
multipath component.
Given the additional delay of the largest multipath component
P ¼ log10
j a0 j 2
2
!
þ log10
1
ja0 j2
A0 ði ; 0 Þ jai j2
!
ð8:24Þ
where a0 is the first signal compnent, ai is the amplitude of the largest signal component,
ti is the total path delay of the ith signal component, and A0 (ti, t0) is an amplitude
scaling factor which accounts for the additional path delay between the direct and the
largest signal component. If the direct component is the largest signal component, A0 is
equal to 1.
–85
dn path loss
Received Signal Power (dBm)
–90
–95
–100
–105
–110
–115
–120
0
50
100
150
200
Delay (ns)
Figure 8.12 Example of NLOS multipath profile
250
UWB Theory and Applications
194
In the event that the received signal is for a NLOS channel, it is possible to estimate
the amount by which the direct component has been attenuated and hence the additional delay experienced by the direct component. Let us define A0 by
2n
i
A0 ði ; 0 Þ ¼
0
ð8:25Þ
where n is the estimated path loss coefficient. The value of n may be between 2 and 3 for
practical systems. Figure 8.13 shows the value of the second term in the confidence
metric as a function of the ratio of delays for two equal power channel components.
8.5.1.1
Use of Confidence Metrics
Instead of degrading the confidence of the received signal based on the second term of
the confidence metric, the estimate A0 may be used to estimate the additional attenuation achieved.
In order to detect the position of a user in three dimensions, a minimum of four
detectors is required. If many sensors are available, the metrics may be used to select the
best detectors to be used for calculating position. Using this approach, the received
signal from each sensor can be ranked by confidence. The best four can be used to
determine the user position. This approach does assume that at least a crude channel
estimate is generated at each sensor.
0
Exponent 2
Exponent 3
Exponent 4
–0.5
–1
Confidence
–1.5
–2
–2.5
–3
–3.5
–4
–4.5
–5
1
1.5
2
2.5
3
3.5
Delay Ratio of Largest to First MPC (equal power)
4
Figure 8.13 Value of second term in confidence metric for equal power channel components
Positioning
195
A sensor configuration is illustrated in Figure 8.14, and Figures 8.15 and 8.16 show
the confidence weighting for LOS and NLOS conditions for that configuration. The
very large uncertainty regions result from the low power, poorly conditioned, impulse
response found for the NLOS location.
FLOORPLAN: Red = Wall. Blue = Transmitting Point. Green = Receiver Point.
20
Transmitter 2
Transmitter 1
15
LOS user
10
5
NLOS user
Transmitter 4
F
0
Transmitter 3
–5
–10
–5
5
0
10
15
20
Figure 8.14 Positioning scenario set-up
FLOORPLAN: Red = Wall. Blue = Transmitting Point. Green=Receiver Point
20
Transmitter 1
Transmitter 2
15
10
LOS user
5
NLOS user
Transmitter 4
0
Transmitter 3
Re
–5
–10
–5
0
5
10
15
20
Figure 8.15 Uncertainty regions for LOS user based on received signal strength
UWB Theory and Applications
196
FLOORPLAN: Red=Wall,blue=Transmitting Point,Green=Receiver Point
20
Transmitter 1
Transmitter 2
15
LOS user
10
5
NLOS user
Transmitter 4
Transmitter 3
0
–5
–10
–5
0
10
15
20
Figure 8.16 Uncertainty regions for NLOS user based on received signal strength
8.6 Conclusions
This chapter has examined some of the most common positioning techniques for radio
systems. The wideband nature of the signal allows UWB systems to offer potentially
very high positioning accuracy in simple propagation environments. However, the
multipath-rich channel and the very low power of the UWB pulse significantly complicate the positioning problem.
Some of the practical problems of UWB positioning have been addressed in this
chapter, and some possible means of detecting and possibly avoiding problems with
UWB positioning were examined. One very important issue is the identification, and
preferably avoidance of, NLOS conditions when performing delay estimation calculations for positioning.
Appendices
Appendix 1
Time hopping pulse position modulated signal in multiple-access case. The formulation
is presented in graphic form.
1
X
ðkÞ
str ðtðkÞ Þ ¼
ðkÞ
ðkÞ
wtr ðtðkÞ jTf cj Tc db j=Ns c Þ
j¼1
T(k) << Tf
j
=
1
2
Tf
3
2Tf
4
3Tf
5
4Tf
6
5Tf
2Tf
3Tf
4Tf
6Tf
8
9
7Tf
8Tf
10
.....
9Tf ......
Time
Select time/transmission frame (for single user case)
t(k)–jTf
Tf
7
5Tf
6Tf
7Tf
8Tf
9Tf ......
t(k) – jTf – cj(k)Tc Select nominal transmission time
(for multiple access case)
nTf
Tc
2Tc
3Tc
4Tc
5Tc
6Tc
7Tc
(n + 1)Tf
t(k)–jTf – cj(k)Tc – δdj/Ns(k)
nTf
Tc
2Tc
3Tc
4Tc
tj
5Tc
6Tc
Select actual transmission time
(with Pulse Position Modulation)
7Tc (n + 1)Tf
wtr(t(k) – jTf – cj(k)Tc – δdj/Ns(k))
Transmitted monocycle waveform
tj
nTc
(n + 1)Tc
UWB Theory and Applications Edited by I. Oppermann, M. Hämäläinen and J. Iinatti
Ó 2004 John Wiley & Sons, Ltd ISBN: 0-470-86917-8
Appendices
198
Meaning of each parameters in the formula.
Actual transmission time: t(k) – jTf – cj(k)Tc – δdj/Ns(k)
Data modulation (PPM): δdj/Ns(k)
Random time due to the code: cj(k)Tc
Current frame: jTf
nTf
tj
Tc
2Tc
3Tc
4Tc
5Tc
6Tc
7Tc (n + 1)Tf
Appendices
199
Appendix 2
The frequency mask for indoor UWB devices set by FCC (Federal Communications
Commission, 2002b)
–40
UWB EIRP Emission Level in dBm
–45
–50
–55
–60
–65
Indoor Limit
Part 15 Limit
–70
–75
100
101
Frequency in GHz
The unit used in the diagram above is dBm/MHz. The next calculations are to
convert the unit of signal level from dBm/MHz into power unit dBm/Hz and then into
watts (W),
41:3dBm
MHz
¼ 101:3dBm
PdBm ¼ 10 log
) Pwatt ¼ 10
pdBm
10
1 mW ¼ 10
101:3
10
Hz
Pwatt
1mW
1 mW ¼ 7:413 1014 W
From this (assuming the gain of antenna is 0 dB) the voltage is calculated in a 50 ohm
system (R ¼ 50 ),
Appendices
200
P¼
Urms ¼
Up ¼
pffiffiffiffiffiffiffi
U2
) U ¼ PR
R
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
7:413 1014 W 50 ¼ 1:925 mV
pffiffiffi
2 Urms ¼ 2:72 mV
The result is assumed to be the average voltage level of the transmission. The voltage
level of a single pulse is determined from previous by utilising the duty cycle (assumed to
be 1%),
Upulse ¼
Up 2:72 V
¼ 272 mV
¼
0:01
Appendices
Appendix 3
Schematic of UWB transmitter.
201
202
Appendix 4
Schematic of digital pulse generation circuit
Appendices
Appendices
Appendix 5
Schematic of single-ended to differential conversion circuit
203
204
Appendix 6
Schematic of waveform generator
Appendices
Appendices
Appendix 7
Schematic of UWB receiver
205
206
Appendix 8
Schematic of Gilbert multiplier
Appendices
Appendices
Appendix 9
Schematic of integrator
207
208
Appendix 10
Schematic of delay element
Appendices
References and Bibliography
Agee, F., C. Baum, W. Prather, J. Lehr, J. O’Loughlin, J. Burger, J. Schoenberg, D. Scholfield, R. Torres, J.
Hull and J. Gaudet (1998) ‘Ultra-wideband transmitter research’, IEEE Transactions on Plasma Science, 26,
pp. 860–873.
Allen O. E., Hill D. A., Ondrejka A. R. (1993) ‘Time-domain antenna characterisations’, IEEE Transactions
on Electromagnetic Compatibility, 35, pp. 339–346.
Anderson, F., W. Christensen, L. Fullerton and B. Kortegaard (1991) ‘Ultra-wideband beam forming in
sparse arrays’, IEE Proceedings H, 138, pp. 342–346.
Astanin, L. Y. and A. A. Kostylev (1992) ‘Ultra-wideband signals – a new step in radar development’, IEEE
AES Systems Magazine, March, pp. 12–15.
Babanezhad, J. N. and G. C. Temes (1985) ‘A 20-V four-quadrant CMOS analog multiplier’, IEEE Journal of
Solid State Circuits, 20, pp. 1158–1168.
Balanis C. A. (1997) ‘Antenna theory: analysis and design’, 2nd edition, John Wiley & Sons, Inc., 941 p.
Baum C. E., Farr E. G. (1993) ‘Impulse radiating antennas’, Ultra-Wideband Short Pulse Electromagnetics ,
H. L. Bertoni, L. Carin, and L. B. Felsen, Eds., New York: Plenum Press, pp. 139–147, 1993.
Baum C. E., Stone A. P. (1993) ‘‘Transient lenses for transmission systems and antennas’’, in Ultra-Wideband
Short Pulse Electromagnetics , H. L. Bertoni, L. Carin, and L. B. Felsen, Eds., New York: Plenum Press,
pp. 211–219, 1993.
Bello, P. A. (1963) ‘Characterization of randomly time-variant linear channels’, IEEE Transactions on
Communications Systems, 11, pp. 360–396.
Bennett, C. and G. Ross (1978) ‘Time-domain Electromagnetics and its applications’, Proceedings of IEEE, 66,
pp. 299–318.
Broyden C. G. (1970) ‘The convergence of a class of double-rank minimization algorithms’, Journal of the
Institute of mathematics and its applications, 6, pp. 76–90.
Buchenauer C. J., Tyo J. S., Schoenberg J. S. H. (1999), ‘Aperture efficiencies of Impulse radiating antennas’, in
Ultra-Wideband Short Pulse Electromagnetics 4 , E. Heyman, B. Mandelbaum, and J. Shiloh, Eds., Kluwer
Academic/Plenum Publishers, New York, pp. 91–108, 1999.
Buchwald, W., A. Balekdjian, J. Conrad, J. Burger, J. Schoenberg, J. Tyo, M. Abdalla, S. Ahern and
M. Skipper (1997) ‘Fabrication and design issues of bulk photoconductive switches used for ultra-wideband,
high-power microwave generation’, The 23rd International Conference on Power Modulators, Baltimore, MA
USA, pp. 970–974.
Burger, J., J. Schoenberg, J. Tyo, M. Abdalla, S. Ahern, M. Skipper and W. Buchwald (1997) ‘Development
and testing of bulk photoconductive switches used for ultra-wideband, high-power microwave generation’,
Proceedings of the 11th IEEE International Pulsed Power Conference, Baltimore, USA, pp. 965–969.
UWB Theory and Applications Edited by I. Oppermann, M. Hämäläinen and J. Iinatti
Ó 2004 John Wiley & Sons, Ltd ISBN: 0-470-86917-8
210
UWB Theory and Applications
Carin, L., N. Geng, M. McClure, J. Sichina and L. Nquyen (1999) ‘Ultra-wideband synthetic aperture radar
for minefield detection’, IEEE Antennas and Propagation, 41, pp. 18–33.
Carin L., Felsen L. (ed.) (1995) Ultra-wideband, short-pulse electromagnetics 2. Conference Proceedings.
Kluver Academic/Plenium Press, New York, 1995, 605 p.
Cassioli D., Win M. Z., Molisch A. F. (2002) The ultra-wide bandwidth indoor channel: from statistical model to
simulations. IEEE Journal on Selected Areas in Communications, 20, Issue: 6, Aug. 2002, Pages: 1247–1257.
Chambers, C., S. R. Cloude and P. D. Smith (1992) ‘Wavelet processing of ultra-wideband radar signals’, IEE
Colloquium on Antenna and Propagation Problems of Ultra Wideband Radar, London, UK, 4 p.
Chan Y. T., Ho K. C. (1994) ‘‘A simple and efficient estimator for hyperbolic location,’’ IEEE Transactions on
Signal Processing, 42, no. 8, pp. 1905–1915.
Chevalier Y., Imbs Y., Beillard B., Andrieu J., Jouvet M., Jecko B., de Legros E. (1999), ‘‘A new broadband
resistive wire antenna for ultra wideband applications’’, in Ultra-Wideband Short Pulse Electromagnetics 4,
E. Heyman, B. Mandelbaum, and J. Shiloh, Eds., Kluwer Academic/Plenum Publishers, New York,
pp. 157–164, 1999.
Cohen, M. N. (1991) ‘An overview of high range resolution radar techniques’, Proceedings of the National
Telesystem Conference (NTC ’91), Atlanta, GA USA, pp. 107–115.
Conroy, J., J. Locicero and D. Ucci (1999) ‘Communication techniques using monopulse waveforms’,
Proceedings of the IEEE Military Communications Conference (MILCOM ’99), Atlantic City, NJ USA,
pp. 1181–1185.
Cook, P. G. and W. Bonser (1999) ‘Architectural overview of the SPEAKeasy system’, IEEE Journal on
Selected Areas in Communications, 17, pp. 650–661.
Cramer, R. J.-M., M. Z. Win and R. A. Scholtz (1998) ‘Impulse radio multipath characteristics and diversity
reception’, Proceedings of the 1998 IEEE International Conference on Communications (ICC ’98), Atlanta,
GA USA, pp. 1650–1654.
Cramer, R. J.-M., M. Z. Win and R. A. Scholtz (1998) ‘Evaluation of the multipath characteristics of the
impulse radio channel’, Proceedings of Personal, Indoor, and Mobile Radio Communications (PIMRC ’98),
Boston, MA USA, pp. 864–868.
Cramer, R. J.-M., R. A. Scholtz and M. Z. Win (1999) ‘On the analysis of UWB communication channels’,
Proceedings of the IEEE Military Communication Conference (MILCOM ’99), Atlantic City, NJ USA,
pp. 1191–1195.
Cramer, R. J.-M., R. A. Scholtz and M. Z. Win (1999) ‘Spatio-temporal diversity in ultra-wideband radio’,
Proceedings of Wireless Communications and Networking Conference (WCNC ’99), New Orleans, LA USA,
pp. 888–892.
Cuomo, K., J. Piou and J. Mayhan (1999) ‘Ultra-wideband coherent processing’, IEEE Transactions on
Antennas and Propagation, 47, 1094–1107.
Daniels J. D. (1996) Digital design from zero to one, John Wiley & Sons, Inc. 640 p.
di Benedetto M.-G. (2001) ‘Ultra wide band radio system aspects in telecommunications sharing and networking’,
Workshop on regulatory issues regarding implementation of Ultra Wide Band Technology in Europe,
20 Mar 2001, Mainz, Germany.
Dickson, D. and P. Jett (1999) ‘An application specific integrated circuit implementation of a multiple
correlator for UWB radio applications’, Proceedings of the IEEE Military Communications Conference
(MILCOM ’99), Atlantic City, NJ USA, pp. 1207–1210.
di Sorte D., Femminella M., Reali G., Zeisberg S. (2002) ‘Network service provisioning in UWB open mobile
access networks’, IEEE Journal of Selected Areas in Communications, 20, no. 9, pp. 1745–1753
Edgar, D. L., N. I. Cameron, H. McLelland, M. C. Holland, M. R. S. Taylor, I. G. Thayne, C. R. Stanley
and S. P. Beaumont (1999) ‘Metamorphic GaAs HEMTs with fT of 200 GHz’, Electronics Letters, 35
pp. 1114–1115.
Engler, H. F., Jr (1991) ‘Systems considerations for large percent-bandwidth radar’, Proceedings of the
National Telesystem Conference (NTC ’91), Atlanta, GA USA, pp. 133–137.
Engler, H. (1993) ‘Advanced technologies for ultra-wideband system design’, Proceedings IEEE International
Symposium on Electromagnetic Compatibility, Dallas, TX USA, pp. 250–253.
Excell P. S., Tinniswood, A. D., Clarke R. W. (1998) ‘Log-periodic antenna for pulsed radiation’, Electronics
Letters, 34, pp. 1990–1991.
Fang B. T. (1990) ‘Simple solutions for and related position fixes’, IEEE Transactions on Aerospace and
Electronic Systems, 26, no. 5, pp. 748–753, Sept. 1990.
References and Bibliography
211
Farr E. G., Baum C. E., Prather W. D., Bowen L. H. (1999) ‘Multifunction impulse radiating antennas: Theory
and experiment’, in Ultra-Wideband Short Pulse Electromagnetics 4 , E. Heyman, B. Mandelbaum, and
J. Shiloh, Eds., Kluwer Academic/Plenum Publishers, New York, pp. 131–144, 1999.
Federal Communication Commission, FCC (1998) ‘Notice of Inquiry’, ET Docket No. 98–153, Rules Regarding Ultra-Wideband Transmission Systems.
Federal Communications Commission (2002a) http://www.fcc.gov/Bureaus/Engineering_Technology/News_
Releases/2002/nret0203.html, FCC press release, Feb 2002.
Federal Communications Commission (2002b) ‘First Report and Order in the matter of revision of Part 15 of the
Commission’s rules regarding ultrawideband transmission systems’, ET-Docket 98–153, FCC 02–48, released
April 22, 2002.
Fleming, B. (1999) ‘Integrated ultra-wideband localizers’, Proceedings of the 1999 International Ultra-wideband
Conference, Washington DC USA, Conference CD.
Fletcher R., Powell M. J. D. (1963) ‘A rapidly convergent descent method for minimization’, Computer Journal,
6, pp. 163–168, 1963.
Foerster J. (2003) ‘Channel modeling sub-committee report – Final’, EEE P802.15 Working Group for Wireless
Personal Area Networks (WPANs), Feb 7, 2003.
Foerster J., Green E., Somayazulu S., Leeper D. (2001) ‘Ultra-wideband technology for short- or mediumrange wireless communications’, Intel Technology Journal, Q2, 11 p.
Foerster, J., M. Pendergrass and A. F. Molisch (2003) ‘A channel model for ultra-wideband indoor communications’, Proceedings of the 6th International Symposium on Wireless Personal Multimedia Communications, Yokosuka, Japan, pp. 116–120.
Fontana R. J., Larrick J. F., Cade J. E.: ‘An ultra wideband communications link for unmanned vehicle applications’, Multispectral Solutions, Inc
Fontana, R. J., Larrick J. F. and J. E. Cade (1997) ‘An Ultra-wideband communications link for unmanned
vehicle applications’, Association for Unmanned Vehicle Systems International 1997 Conference (AUVSI ’97),
Baltimore, MD USA.
Foster P. R. (1993) ‘Reflector antennas for ultra wideband usage’, in Ultra-Wideband Short Pulse Electromagnetics , H. L. Bertoni, L. Carin, and L. B. Felsen, Eds., New York: Plenum Press, pp. 203–209, 1993.
Fowler, C., J. Entzminger and J. Corum (1990) ‘Assessment of ultra-wideband technology’, IEEE AES
Magazine, November, pp. 45–49.
Friedlander B. (1987) ‘A passive location algorithm and its accuracy analysis’, IEEE Journal of Oceanic
Engieering, 12, pp. 234–245, Jan. 1987.
Fullerton, L. (1991) ‘UWB waveforms and coding for communication and radar’, Proceedings IEEE National
Telesystem Conference (NTC ’91), Atlanta, GA USA, pp. 139–141.
Funk, E., S. Saddow, L. Jasper and A. Lee (1995a) ‘Time coherent ultra-wideband pulse generation using
photoconductive switching’, Digest of the LEOS Summer Topical Meetings, pp. 55–56.
Funk, E., S. Saddow, L. Jasper and C. Lee (1995b) ‘Coherent power combining of ultra-wideband pulsed
radiation in free space’, Microwave Symposium Digest, Vol. 3, IEEE MTT-S International, Orlando, FL USA,
pp. 1299–1302.
Funk, E., S. Ramsay, C. Lee and J. Craven (1998) ‘A photoconductive correlation receiver for wireless digital
communications’, International Topical Meetings on Microwave Photonics, Princeton, NJ USA, pp. 21–24.
Gilbert B. (1968) ‘A precise four quadrant multiplier with sub nanosecond response’, IEEE Journal of SolidStage Circuits, 3, pp. 365–373, Dec. 1968.
Gilbert B. (1997) ‘A highly linear variant of the Gilbert mixer using a bisymmetric class-AB input stage’, IEEE
Journal of Solid-Stage Circuits, 32, No. 9, pp. 1412–1223.
Gray P. R., Meyer R. G. (1993) ‘Analysis and design of analog integrated circuits’, Third Edition, John Wiley &
Sons, Inc. 792 p.
Gregorian R., Temes G. C. (1986) ‘Analog CMOS integrated circuits for signal processing’, John Wiley & Sons,
Inc. 598 p.
Gill, G. S., H. F. Chiang and J. Hall (1994) ‘Waveform synthesis for ultra-wideband radar’, Proceedings of the
IEEE Radar Conference, Atlanta, GA USA, pp. 240–245.
Glisic, S. and B. Vucetic (1997) ‘Spread Spectrum CDMA System for Wireless Communications’, Artech House
Publisher, Norwood, MA USA, pp. 383.
Griffiths J. (1987) ‘Radio Wave Propagation and Antennas: An Introduction’. Prentice-Hall International, (UK)
Ltd., London, UK, 384 s.
212
UWB Theory and Applications
Häkkinen J. (2002) ‘Integrated RF building blocks for base station applications’, Acta Universitatis Ouluensis,
C177, 102 p.
Hämäläinen, M., V. Hovinen and M. Latva-aho (1999a) ‘Survey on ultra-wideband systems’, COST262/
cwc_wg2_td013(99), Thessaloniki, Greece.
Hämäläinen, M., V. Hovinen and M. Latva-aho (1999b) ‘Introduction to impulse radio systems’, URSI/IEEE
XXIV Convention on Radio Science, Turku, Finland, 2 p.
Hämäläinen M., Hovinen V., Tesi R., Iinatti J., Latva-aho M. (2002) ‘On the UWB system co-existence with
GSM900, UMTS/WCDMA and GPS’, IEEE Journal on Selected Areas in Communications, 20, No. 9,
p. 1712–1721.
Hämäläinen M., Hentilä L., Pihlaja J., Nissinaho P. (2003) ‘Modified frequency domain radio channel measurement system for ultra wideband studies’, Finnish Wireless Communications Workshop (FWCW 2003), Oulu,
Finland.
Harmuth, H. F. (1978) ‘Frequency sharing and spread spectrum transmission with large relative bandwidth’,
IEEE Transactions on Electromagnetic Compatibility, Vol. 20, pp. 232–239.
Harmuth, H. E. and S. Ding-Rong (1993) ‘Large current, short length radiator for non-sinusoidal waves’,
Proceedings IEEE International Symposium on Electromagnetic Compatibility, Arlington, VA, USA, pp. 453–456.
Harmuth, H. E. (1984), ‘Antennas and Waveguides for Nonsinusoidal Waves’, New York: Academic.
Hovinen, V., M. Hämäläinen and T. Pätsi (2002) ‘Ultra-wideband indoor radio channel models: preliminary
results’, Proceedings of IEEE Conference on Ultra-wideband Systems and Technologies (UWBST ’02),
Baltimore, MD USA, pp. 75–79.
Hussain, M.(1996) ‘An overview of the principles of ultra-wideband impulse radar’, Proceedings of the CIE
International Conference on Radar, Beijing, China, pp. 24–28.
IEEE (2004) http://www.ieee802.org/15/pub/
ITU (2002) ‘Preliminary compatibility analysis between space scientific services and UWB’. ITU Report
SE24M16_50 Rev.2
Iinatti, J. (1997) ‘Matched filter code acquisition employing a median filter in direct sequence spread-spectrum
systems with jamming’, Acta Universitatis Ouluensis, C102, 54 p.
Iinatti J. (2000) ‘Performance of DS code acquisition in static and fading multi-path channel’, IEE Proceedings
Communications, vol.147, Issue 6, December 2000, pp.355–360.
Iinatti, J. and M. Latva-aho (2001) ‘A modified CLPDI for code acquisition in multipath channel’, Proceedings of Personal, Indoor, and Mobile Radio Communications (PIMRC’01), San Diego, CA USA, pp. F6–F10.
Iverson, D. E. (1994) ‘Coherent processing of ultra-wideband radar signals’, Radar, sonar, and navigation –
IEEE Proceedings, 141, pp. 171–179.
Jakobsson, A., A. Lee Swindlehurst and P. Stoica (1998) ‘Subspace-based estimation of time delays and
Doppler shift’, IEEE Transactions on Signal Processing, 46, 2472–2483.
Johns D. A., Martin K. (1997) ‘Analog integrated circuit design’, John Wiley & Sons, Inc. 706 p.
Kardo-Sysoev, A. F., V. I. Brylevsky, Y. S. Lelikov, I. A. Smirnova and S. V. Zazulin (1999) ‘Generation and
radiation of powerful nanosecond and sub-nanosecond pulses at high pulse repetition rate for UWB
systems’, 1999 International Ultra-wideband Conference, Washington, DC USA, Proceedings on CD.
Katz M. (2002) ‘Code Acquisition in advanced CDMA networks’, Acta Universitatis Ouluensis, C175, Finland,
85 p
Keignars, J. and N. Daniele (2003) ‘Channel sounding and modelling for indoor UWB communication’,
Proceedings of the First International Workshop on Ultra Wideband Systems, Oulu (IWUWBS ’03), Finland,
Proceedings on CD, 5 p.
Khorramabadi, H. and P. R. Gray (1984) ‘High-frequency CMOS continuous-time filters’, IEEE Journal of
Solid State Circuits, 19, 939–948.
Kim, A., L. Domenico, R. Youmans, A. Balekdjian, M. Weiner and L. Jasper (1993) ‘Monolithic photoconductive ultra-wideband RF device’, Proceedings IEEE International Microwave Symposium, Atlanta, GA
USA, pp. 1221–1224.
Kolenchery, S. S., J. K. Townsend and J. A. Freebersyser (1998) ‘A novel impulse radio network for tactical
military wireless communications’, Proceedings of the IEEE Military Communications Conference (MILCOM ’98),
Boston, MA USA, pp. 59–65.
Kolenchery, S. S., J. K. Townsend, J. A. Freebersyser and G. Bilbro (1997) ‘Performance of local power
control in peer-to-peer impulse radio networks with bursty traffic’, Proceedings of the Global Telecommunications Conference IEEE (GLOBECOM ’97), Phoenix, AZ USA, pp. 910–916.
References and Bibliography
213
Kunisch, J. and J. Pamp (2002) ‘Measurement results and modelling aspects for the UWB radio channel’, Digest of
papers, IEEE Conference on Ultra-wideband Systems and Technologies (UWB ST ’02), Baltimore, MD USA, 19–23.
König, U. (1999) ‘Progress in SiGe heterostructure devices’, IEE Colloquium on Advances in Semiconductor
Devices, London, UK, pp. 6/1–6/6.
Lai, A. K. Y., A. L. Sinopoli and W. D. Burnside (1992) ‘A novel antenna for ultra-wideband applications’,
IEEE Transactions on Antennas and Propagation, 40, pp. 755–760.
Lamensdorf D., Susman L. (1994) ‘Baseband-pulse-antenna techniques’, IEEE Antennas and Propagation
Magazine, 36, pp. 20–30.
Lang, J. (2003) ‘UWB chip design with embedded functionality’, UWB Summit, Paris, France, CD Proceedings.
Latva-aho, M. (1998) ‘Advanced receivers for CDMA Systems’, Acta Universitatis Ouluensis, C125, 179p.
Lee, J. S. and C. Nguyen (2001) ‘Novel low-cost ultra-wideband, ultra-short-pulse transmitter with MESFET
impulse-shaping circuitry for reduced distortion and improved pulse repetition rate’, IEEE Microwave and
Wireless Components Letters, 11, pp. 208–210.
Lee J.S., Ngyuen C., Sullicon T. (2001a) ‘New uniplanar subnanosecond monocycle pulse generator and
transformer for time-domain microwave applications’. IEEE Transactions on Microwave Theory and
Techniques, 49, No. 6.
Lee J.S., Nguyen C. (2001b) ‘Novel low-cost ultra-wideband, ultra-short-pulse transmitter with MESFET
impulse –shaping circuitry for reduced distortion and improved pulse repetition rate’. IEEE Microwave and
Wireless Components Letters, 11, No. 5.
Lestari, A. A., A. G. Yarovoy and L. P. Ligthart (2000) ‘Capacitively-tapered bowtie antenna’, Proceedings
Millennium Conference on Antennas and Propagation, Davos, Switzerland, Proceedings on CD.
Lewis, L. R., M. Fasset and J. Hunt (1974) ‘A broadband stripline array element’, Digest of the 1974 IEEE
Antennas and Propagation Society International Symposium, pp. 335–337.
Li X. (2002) ‘Evaluation of RF CMOS IC technology for wireless LAN applications’, University of Florida,
http://www.tec.ufl.edu/~xli. Page available at 25.08.2003.
Lu M., Shi C. (1999) ‘Quality ultra-wideband omni-directional antenna’, in Ultra-Wideband Short Pulse
Electromagnetics 4, E. Heyman, B. Mandelbaum, and J. Shiloh, Eds., Kluwer Academic/Plenum Publishers,
New York, pp. 122–125, 1999.
Lynch, W. C., K. Rahardja and S. Gehring (1999) ‘An analysis of noise aggregation from multiple distributed
RF emitters’, 1999 International Ultra-wideband Conference, Washington, DC USA, Proceedings on CD.
Maggio, G., N. Rulkov, M. Sushchik, L. Tsimring, A. Volkovskii, H. Abarbanel, L. Larson and K. Yao
(1999) ‘Chaotic pulse-position modulation for ultra-wideband communication systems’, 1999 International
Ultra-wideband Conference, Washington, DC USA, Proceedings on CD.
Maloney J. G., Smith G. S. (1993a) ‘A study of transient radiation from the Wu-King resistive monopole – FDTD
analysis and experimental measurements’, IEEE Transactions on Antennas and Propagation, 41, pp. 668–675.
Maloney J. G., Smith G. S. (1993b) ‘Optimization of conical antennas for pulse radiation: An efficient design
using resistive loading’ , IEEE Transactions on Antennas and Propagation, 41, pp. 940–947, Jul 1993.
Manabe, T. Takai H. (1992) ‘Superresolution of multipath delay profiles measured by PN correlation
method’, IEEE Transactions on Antennas and Propagation, 40, No 5, 1992, pp. 500–509
McCorkle J.W. (2001) ‘Ultra wideband communication system, method, and device with low noise pulse formation’. World Intellectual Property Organization WO 01/93520 A2.
Moddemeijer R. (1991) ‘On the determination of the position of extrema of sampled correlators’, IEEE
Transactions on Signal Processing, 39, No 1., 1991, pp. 216–291
Montoya T. P, Smith G. S. (1996) ‘A study of pulse radiation from several broad-band loaded monopoles’,
IEEE Transactions on Antennas and Propagation, 44, pp.1172–1182, Aug 1996.
Morgan, M. (1994) ‘Ultra-wideband impulse scattering measurements’, IEEE Transactions on Antennas and
Propagation, 42, 840–846.
Nunnally, N. (1993) ‘Generation and application of moderate power, ultra-wideband or impulse signals’,
Proceedings International Symposium on Electromagnetic Compatibility, Dallas, TX USA, pp. 260–264.
Nguyen C., Lee J.– S., Park J. – S. (2001) ‘‘Ultra-wideband microstrip quasi-horn antenna’’, Electronics
Letters, 37, pp. 731–732, Jun 2001.
Olhoeft, G. (1999) ‘Applications and frustrations in using ground-penetrating radar’, 1999 International Ultrawideband Conference, Washington, DC USA, Proceedings on CD.
Parkway P. (2001) ‘Ultra-wideband data transmission system’. World Intellectual Property Organization WO
01/39451 A1.
214
UWB Theory and Applications
Petroff, A. (1999) ‘Time modulated ultra-wideband: performance on a chip’, 1999 International Ultrawideband Conference, Washington, DC USA, Proceedings on CD.
Petroff, A. and P. Withington (2000) ‘Time modulated ultra-wideband: overview’, 1999 International Ultrawideband Conference, Washington, DC USA, Proceedings on CD.
Pochain G. P. (1999) ‘Large current radiator for the short electromagnetic pulses radiation’, in Ultra-Wideband
Short Pulse Electromagnetics 4 , E. Heyman, B. Mandelbaum, and J. Shiloh, Eds., Kluwer Academic/
Plenum Publishers, New York, pp. 149–155, 1999.
Polydoros, A. (1982) ‘On the synchronization aspects of direct-sequence spread spectrum systems’, Ph.D.
dissertation, University of Southern California, Los Angeles, California, USA.
Prasad, R. and S. Hara (1996) ‘An overview of multi-carrier CDMA’, Proceedings of the Fourth IEEE
Symposium on Spread Spectrum Techniques and Applications (IS SSTA 96), Mainz, Germany, 107–114.
Proakis J. G. (1995) ‘Digital communications’, McGraw-Hill Inc., Singapore, 928 s.
Proakis J. G., Salehi M. (1994) ‘Communication system engineering’, Prentice-Hall, Inc.
Qiu, R. (1998) ‘A theoretical study of the ultra-wideband wireless propagation channel based on the scattering
centres’, Proceedings of the IEEE Conference on Vehicular Technology (VTC 98), Ottawa, ON Canada,
pp. 308–312.
Raines, J. (1999) ‘Cumulative impact of TM-UWB devices and effects on three aviation receivers’, 1999
International Ultra-wideband Conference, Washington, DC USA, Proceedings on CD.
Ramirez-Mireles, F. and R. A. Scholtz (1997) ‘Performance of equicorrelated ultra-wideband pulseposition-modulated signals in indoor wireless impulse radio channels’, in Proceedings of the Sixth
IEEE Pacific Rim Conference on Communications, Computers and Signals, Victoria, BC Canada,
pp. 640–644.
Ramirez-Mireles, F. and R. A. Scholtz (1998a) ‘System performance analysis of impulse radio modulation’,
Proceedings of the Radio and Wireless Conference, Colorado Springs, AZ USA, pp. 67–70.
Ramirez-Mireles, F. and R. A. Scholtz (1998b) ‘Multiple access with time hopping and block waveform PPM
modulation’, Proceedings of the 1998 IEEE International Conference on Communications (ICC ’98), Atlanta,
GA USA, pp. 775–779.
Ramirez-Mireles, F. and R. A. Scholtz (1998c) ‘Multiple-access performance limits with time hopping and
pulse position modulation’, Proceedings of the IEEE Military Communications Conference(MILCOM ’98),
Boston, MA USA, pp. 529–533.
Ramirez-Mireles, F. and R. A. Scholtz (1998d) ‘Wireless multiple access using SS time-hopping and block
waveform pulse position modulation, Part 1: signal design’, Proceedings of the International Symposium on
Information Theory and Its Applications (ISITA), Mexico.
Ramirez-Mireles, F. and R. A. Scholtz (1998e) ‘Wireless multiple access using SS time-hopping and
block waveform pulse position modulation, Part 2: system performance’, Proceedings of the International
Symposium on Information Theory and Its Applications (ISITA), Mexico.
Ramirez-Mireles, F., M. Z. Win and R. A. Scholtz (1997a) ‘Signal selection for the indoor wireless impulse
radio channel’, Proceedings of the IEEE Conference on Vehicular Technology (VTC 97), Phoenix, AZ USA,
pp. 2243–2247.
Ramirez-Mireles, F., M. Z. Win and R. A. Scholtz (1997b) ‘Performance of ultra-wideband time-shift-modulated signals in the indoor wireless impulse radio channel’, Proceedings of the Thirty-first Asilomar Conference, Pasific Grove, CA USA, pp. 192–196.
Roberts, R. ‘‘XtremeSpectrum CFP Document’’, IEEE P802.15, IEEE P802.15–03/154r3, 124 p., Jul 2003.
Rothwell, E., K. Chen, D. Nyquist and J. Ross (1995) ‘Time-domain imaging of airborne targets using ultrawideband or short-pulse radar’, IEEE Transactions on Antennas and Propagation, 43, 327–329.
Rowe, D., B. Pollack, J. Pulver, W. Chon, P. Jett, L. Fullerton and L. Larson (1999) ‘A Si/Ge HBT timing
generator IC for high-bandwidth impulse radio applications’, Proceedings of the IEEE Custom Integrated
Circuits Conference, San Diego, CA USA, pp. 221–224.
Schau H. C., Robinson A. Z. (1987) ‘Passive source localization employing intersecting spherical surfaces
from time-of-arrival differences’, IEEE Transactions on Acoustics, Speech, and Signal Processing, 35,
pp. 1223–1225, Aug. 1987
Scholtz, R. A. (1993) ‘Multiple access with time-hopping impulse modulation’, Proceedings of Military
Communications Conference (MILCOM ’93), Boston, MA USA, pp. 447–450.
Scholz, R. A. 2nd M.Z. Win (1997) ‘Impulse radio’, in ‘Wireless communications – TDMA versus CDMA’,
S. Glisic and P. A. Leppänen (eds.), Kluwer Academic Publishers, London, pp. 245–263.
References and Bibliography
215
Scholtz, R. A., R. Cramer and M. Win (1998) ‘Evaluation of the propagation characteristics of ultra-wideband
communication channels’, Proceedings of the IEEE Antennas and Propagation Symposium, Atlanta, GA
USA, pp. 626–630.
Shlager K. L., Smith G. S., Maloney J. G. (1994) ‘Optimization of bow-tie antennas for pulse radiation’, IEEE
Transactions on Antennas and Propagation, 42, pp. 975–982.
Shlager K. L., Smith G. S., Maloney J. G. (1996) ‘Accurate analysis of TEM horn antennas for pulse
radiation’, IEEE Transactions on Electromagnetic Compatibility, 38, pp. 414–423.
Shlivinski, E. Heyman, and R. Kastner (1993) ‘Antenna characterization in the time domain’, IEEE Transactions on Antennas and Propagation, 45, pp.1140–1149.
Silva, J. A. N. da, and M. L. R. de Campos (2002) ‘Orthogonal pulse shape modulation for impulse radio’,
Proceedings of the International Telecommunications Symposium (ITS 2002), Brazil.
Smith J. O., Abel J. S. (1987) ‘Closed-form least-squares source location estimation from range difference
measurements’, IEEE Transactions on Acoustics, Speech, and Signal Processing, 35, pp. 1661–1669.
Staderini, E. (1999) ‘Medical applications of UWB radars’, 1999 International Ultra-wideband Conference,
Washington, DC USA, Proceedings on CD.
Stickley, G. F., D. A. Noon, M. Chernlakov and I. D. Longstaff (1997) ‘Preliminary field results of an ultrawideband (10–620 MHz) stepped-frequency ground penetrating radar’, Geosciences and Remote Sensing
(IGARSS ’97), Singapore, pp. 1282–1284.
Talvitie J. (1997) ‘Wideband radio channel measurement, characterisation and modelling for wireless local
loop applicatios’, Acta Universitatis Ouluensis Technica C99, 94 p.
Taylor J. D. (ed.) (1995) ’Introduction to ultra wideband radar systems’. CRC Press, Inc., Boca Raton, FL USA, 670 p.
Theron, I., E. Walton, S. Gunawan and L. Cai (1999) ‘Ultra-wideband noise radar in the VHF/UHF band’,
IEEE Transactions on Antennas and Propagation, 47, pp. 1080–1084.
Time Domain Corporation (1998) ‘Comments of time domain corporation*, Docket 98–154. In the Matter of
Revision of Part 15 of the FCC’s Rules Regarding Ultra wideband Transmission Systems.
Time Domain Corporation (2003) ‘PulsOn technology overview’, http://www.timedomain.com/. Page available
at 25.08.2003.
Tiuraniemi S. (2002) ‘Method and arrangement for generating cyclic pulses’. United States Patent Application
10/304915.
Torrieri D. J. (1998) ‘Statistical theory of passive location systems’, IEEE Transactions on Aerospace and
Electronic Systems, 20, no. 2, pp. 183–198.
Ultra Wideband Working Group (1998) ‘Comments of the ultra wideband working group, Docket 98–153’.
In the Matter of Revision of Part 15 of the FCC’s Rules Regarding Ultra wideband Transmission
Systems.
Ultra Wideband Working Group (1999) ‘The 1999 International Ultra Wideband Conference Proceedings-CD’,
Huntsville, AL USA.
Ultra Wideband Working Group (2004) http://www.uwb.org/regulatory/regulatory.html.
van Cappellen W. A., de Jongh R. V., Ligthart L. P. (2000) ‘Potentials of ultra-short-pulse time-domain
scattering measurements’, IEEE Antennas and Propagation Magazine, 42, pp. 35–45, Aug 2000.
Vaskelainen, L. and R. Pitkäaho (2002) Antennas Transmitting or Receiving Broadband Impulses’, VTT
Information Technology, Espoo, Finland, Internal report.
Vickers, R. (1999) ‘Design and applications of airborne radars in the VHF/UHF band’, 1999 International
Ultra-wideband Conference, Washington, DC USA, Proceedings on CD.
Weeks, G. and J. Townsend (1999) ‘Quantifying the covertness of impulse radio’, 1999 International Ultrawideband Conference, Washington, DC USA, Proceedings on CD.
Weeks, G., J. Townsend and J. Freebersyser (1999) ‘Performance of hard decision detection for impulse
radio’, Proceedings of the IEEE Military Communications Conference (MILCOM ’99), Atlanta, GA USA,
pp. 1201–1206.
Weissoerger, M. (1982) An Initial Critical Summary of Models for Predicting the Attenuation of Radio Waves by
Trees, ITT Research Institute, Report number: A343811, Annapolis, MD USA, 162 p.
Wicks M. C., Antonik P. (1993) ‘Polarization diverse ultra-wideband antenna technology’, in Ultra-Wideband
Short Pulse Electromagnetics , H. L. Bertoni, L. Carin, and L. B. Felsen, Eds., New York: Plenum Press,
pp. 177–187, 1993.
Win M.Z., Scholtz R. A. (2000) ‘Ultra-Wide Bandwidth Time-Hopping Spread-Spectrum Impulse Radio for
Wireless Multiple-Access Communications’, IEEE Transactions on Communications, 48, pp. 679–689.
216
UWB Theory and Applications
Win, M. Z. (1999) ‘Spectral density of random time-hopping spread-spectrum UWB signals with uniform
timing jitter’, Proceedings of the IEEE Military Communications Conference (MILCOM ’99), Atlanta, GA
USA, pp. 1196–1200.
Win, M. Z. and Z. Kostic (1999) ‘Impact of spreading bandwidth on rake reception in dense multipath
channels’, Proceedings of the 1999 IEEE International Conference on Communications (ICC ’99), Vancouver,
BC Canada, pp. 78–82.
Win, M. Z. and R. A. Scholtz (1997a) ‘Comparisons of analogue and digital impulse radio for wireless
multiple access communications’, Proceedings of the 1997 IEEE International Conference on Communications (ICC ’97), Montreal, Canada, pp. 91–95.
Win, M. Z. and R. A. Scholtz (1997b) ‘Energy capture vs correlator resources in ultra-wideband width indoor
wireless communications channels’, Proceedings of IEEE Military Communications Conference, Monterey,
CA USA, pp. 1277–1281.
Win, M. Z. and R. A. Scholtz (1998a) ‘Impulse radio: how it works’, IEEE Communication Letters, 2,
pp. 36–38.
Win, M. Z. and R. A. Scholtz (1998b) ‘On the energy capture of ultra-wideband width signals in dense
multipath environments’, IEEE Communications Letters, 2, pp. 245–247.
Win, M. Z. and R. A. Scholtz (1998c) ‘On the robustness of ultra-wideband width signals in dense multipath
environments’, IEEE Communications Letters, 2, pp. 51–53.
Win, M. Z. and J. H. Winters (1999a) ‘On maximal ration combining in correlated Nakagami channels with
unequal fading parameters and SNRs among branches: an analytical framework’, Proceedings of the
Wireless Communications and Networking Conference, New Orleans, LA USA, pp. 1058–1064.
Win, M. Z. and J. H. Winters (1999b) ‘Analysis of hybrid selection/maximal ratio combining of diversity
branches with unequal SNR in Rayleigh fading’, Proceedings of the IEEE Conference on Vehicular Technology (VTC ’99), Houston, TX USA, pp. 215–220.
Win, M. Z. and J. H. Winters (1999c) ‘Analysis of hybrid selection/maximal ration combining in Rayleigh
fading, Proceedings of the 1999 IEEE International Conference on Communications (ICC ’99), Vancouver,
BC Canada, pp. 6–10.
Win, M. Z., R. A. Scholtz and L. W. Fullerton (1996) ‘Time-hopping SSMA techniques for impulse radio with
an analogue modulated data sub-carrier’, Proceedings of the IEEE Symposium on Spread Spectrum Techniques and Applications (ISSSTA ’96), Mainz, Germany, pp. 359–364.
Win, M. Z., F. Ramirez-Mireles, R. A. Scholtz and M. A, Barnes (1997a) ‘Ultra-wideband width signal
propagation for outdoor wireless communications’, Proceedings of the IEEE Conference on Vehicular
Technology (VTC ’99), Phoenix, AZ USA, pp. 251–255.
Win, M. Z., J.-H. Ju, V. O. K. Li and R. A. Scholtz (1997b) ‘ATM-based ultra-wideband width multiple-access
radio network for multimedia PCS’, Proceedings of IEEE Fourth Annual Networld þ Interop Conference, Las
Vegas, Nevada USA, pp. 101–108.
Win, M. Z., R. A. Scholtz and M. A, Barnes (1997c) Ultra-wideband width signal propagation for indoor
wireless communications’, Proceedings IEEE International Conference on Communications (ICC ’97),
Montreal, Canada, pp. 56–60.
Win, M. Z., G. Chrisikos and N. Sillenberger (1999) ‘Impact of spreading bandwidth and diversity order on
the error probability performance of rake reception in dense multipath channels’, Proceedings of Wireless
Communications and Networking Conference, New Orleans, LA USA, pp. 1558–1562.
Withington P. (2004) ‘Impulse Radio Overview’. http://www.time-domain.com.
Withington, P., R. Reinhardt and R. Stanley (1999) ‘Preliminary results from ultra-wideband (impulse)
scanning receivers’, Proceedings of the IEEE Military Communications Conference(MILCOM ’99), Atlantic
City, NJ USA, pp. 1186–1190.
Yarovoy, A. G., R de Jongh and L. Ligthart (2000) ‘Ultra-wideband sensor for electromagnetic field
measurements in time domain’, Electronics Letters, 36, 1679–80.
Yngvesson, K. S., T. L. Korzniowski, Y.-S. Kim, E. L. Kollberg and J. F. Johansson (1989) ‘Tapered slot
antenna – a new integrated element for millimetre wave applications’, IEEE Transactions on Microwave
Theory and Techniques, 37, 365–74.
Ziolkowski, R. W. (1992) ‘Properties of electromagnetic beams generated by ultra-wideband width pulsedriven arrays’, IEEE Transactions on Antennas and Propagation, 40, pp. 888–905.
Index
A
AC analysis, 101
accuracy, 11, 12, 67, 127, 157, 175, 179,
182, 187, 189, 190, 196
acquisition, 67, 76, 77, 79, 82, 83, 85, 91,
165, 166, 172
additive white Gaussian noise, 50, 60,
75, 162
ad-hoc networks, 163, 166, 168
aeronautical, 4
air interface, 7, 157
amplifier, 13, 15, 98, 110
amplitude, 27, 29, 44, 72, 94, 122
amplitude gain, 28
analogue addition, 103
analogue subtraction, 103, 111
angle-of-arrival, 157, 176, 189
antenna, 14, 15, 19, 20, 27, 42, 118
aperture, 134
arrays, 135
beamwidth, 134
biconical, 141, 142
bow-tie, 132–134, 148, 150
capacitively loaded dipole, 140, 141
conducting dipole, 137
conical, 15
conical dipole, 141
conical monopole, 134
dipole, 134
horn, 133
impulse radiating antenna, 133
log-periodic dipole array, 142
loop, 135
planar antenna arrays, 135
radiation pattern, 15, 133, 141
resistively loaded dipole, 140
tapered slot antenna, 133, 150
TEM horn, 132, 133, 143, 148
wideband, 10, 131, 129–155
antenna positions, 28, 30
antenna separation, 31, 33, 189
anti-crime operations, 4
architecture, 40, 88–92, 94, 110
assets location system, 177
asymmetric digital subscriber line, 45
asynchronous, 64, 166, 170,
172, 173
attenuation, 34–36, 160, 162, 189,
193, 194
autocorrelation function, 10, 20, 49, 51,
77, 89
automotive applications, 7
avalanche breakdown voltage, 95
avalanche rescue, 44
avalanche transistor, 95
average power, 30, 83, 166
UWB Theory and Applications Edited by I. Oppermann, M. Hämäläinen and J. Iinatti
Ó 2004 John Wiley & Sons, Ltd ISBN: 0-470-86917-8
218
B
bandpass filter, 29, 92, 93
bandwidth, 1, 3, 5, 9–11, 17, 19–22, 38, 42,
44–47, 67, 76, 87, 89, 106, 109, 122,
123, 130, 131, 134, 136, 158, 175, 187
base stations, 129, 132, 133, 136,
156, 182
baseband, 3, 11, 39, 48
baseband signal, 3, 40, 91
baseband signal processing, 89
beacon, 169, 170, 171
Bessel function, 29, 130
bias circuit, 95
bit error rate, 50
bit rate, 20, 47, 87, 124, 162, 166
broadcast address, 170
buffer, 98, 171
burst, 79, 85, 92, 165, 166
C
capacitance, 97, 100, 101, 106, 140
capacitor, 95, 105, 106, 115, 124, 126
carrier-less, 2, 165
carrier modulation, 40
carrier sensing multiple access, 170
carrier signal, 20, 87, 92
Cartesian system, 177
centrally managed network, 158, 169
centre frequency, 42, 93, 122, 137
channel, 9, 10, 19, 51, 75, 76, 82,
158, 191
access, 158
acquisition, 165
coding, 92
correlation matrix, 186, 187
estimation, 68, 85
measurement techniques, 9
models, 20, 21, 25, 30, 73, 83
models, 9–37
profile, 9, 68, 188
request, 170
sounding, 9, 10, 15, 19–22
tap amplitude, 71
time allocation, 170
charging capacitor, 95
chip, 21, 40, 45, 61, 73, 77, 80, 82
Index
chip level post detection integration, 76,
77, 80, 85
chip period, 40
chip rate, 20, 21, 46, 77
chirp, 2, 87
clock oscillator, 88, 91, 92
cluster, 23, 24
code division multiple access, 45
code synchronization, 67, 76, 85
coexistence, 47, 158
coherence time, 10, 28
coherent, 71, 72, 74
collision avoidance, 165, 170
collision free period, 172
combining, 73, 114, 175
absolute combining, 72
equal gain combining, 71
maximum ratio combining, 68, 72
comparator, 106, 117, 126
complexity, 13, 46, 65, 68, 80, 85, 87–89,
114, 157, 158, 184
confidence metric, 191, 193, 194
conjugate approach, 11, 12
consumer communications, 3
contention access period, 170
continuous wave transmitter, 88
coplanar waveguides, 95
correlation matrices, 186, 187
correlation receiver, 10, 20, 117
coupling capacitor, 98
cross-correlation, 50–52, 55, 64,
89, 131
cross-correlator, 40
crystal oscillator, 88
current source, 103, 107
custom designed oscillator, 88
D
data rate, 1, 2, 7, 46, 57, 73, 157, 166
Davidon-Fletcher-Powell, 182
de-convolution, 20
Defense Advanced Research Projects
Agency, 4
delay, 10, 14, 18, 21, 26, 27, 33, 60, 68, 76,
77, 111–113, 118, 121, 185, 189
detectable, 10, 17
Index
estimation, 159, 182, 184–188,
profile, 14, 27, 68
resolution, 15, 21, 25, 71
spread, 26, 40, 83
deterministic model, 30
device management entity, 172
device under test, 9
differential pair, 98, 101–103, 105, 111,
114, 115
differentiator, 89, 122, 129
Digital Audio Broadcasting, 45
digital sampling oscilloscope, 10, 19,
147, 148
Digital Video Broadcasting, 45
direct sequence, 2, 9, 10, 20, 54, 61
discontinuous, 3, 40
distance, 2, 8, 15, 20, 24, 26, 33, 143, 145,
175, 177
distance estimates, 175
diversity, 68, 83
Doppler shift, 28
double maximum likelihood, 177
down-conversion, 3, 11
drain, 99–105
drift, 171
duty cycle, 40, 57, 123, 161, 162, 200
dynamic range, 20, 25–27, 101,
105, 118
E
efficiency, 126, 127, 130, 140, 165, 166
eigenvector decomposition, 187
emitter, 95, 109, 111
energy consumption, 166, 172
energy efficiency, 163, 164
error resistance, 160
European Conference of Postal and
Tele-communications, 5
European Telecommunications Standards
Institute, 4
excess delay, 28, 31, 189
F
fading, 23, 24, 33, 68
fast Fourier transform, 45, 46
fast frequency chirp, 2, 44
219
Federal Aviation Administration, 4
fidelity, 131
finite-difference time-domain, 136
finite-element method, 136
finite impulse response, 28
flexibility, 47, 163
Fourier transform, 42, 131, 136
fractional bandwidth, 4, 5, 21, 39, 123
frame time, 57, 159, 166
frequency, 2, 3, 9, 10, 13, 15, 19, 44, 101,
131, 136, 144
allocation, 4
chirps, 2, 44
diversity, 3, 38
mask, 87, 122, 127
response, 9, 11, 13–16, 101, 144
shift, 10
stability, 18
sweeping, 9
synthesizer, 88
Fresnel zone, 22, 38
gate, 100, 101, 103, 111, 112, 114,
118, 121
G
Gaussian doublet, 42, 44
Gaussian monocycle, 42
Gaussian pulse, 40, 42, 44, 52, 53, 61, 64,
80, 88, 89, 95, 129, 163
geolocation, 42
Gilbert cell, 94, 101–104, 115
Gilbert multiplier, 102, 115
glitch, 113, 118, 121
global positioning system, 3
gradient, 181
gradient-based algorithm, 182
ground penetrating radar, 44
guard times, 171
H
Hanning window, 27
Hermitean pulses, 40
Hermitean signal processing, 11
heterodyne, 40
HFSS, 136
high data rate, 2, 3, 7, 45, 157, 167, 173
Index
220
high-pass filtered, 42, 95, 131, 140
high-speed circuit, 97
Hiperlan2, 45, 158
homodyne, 40
homojunction bipolar transistor, 108
hopping sequences, 162
hyperbolic positioning methods, 177
I
IC process, 107
BiCMOS, 108, 109
CMOS, 97, 108, 109, 114, 127
Gallium Arsenide, 108
Gallium Nitride, 108
Gallium Phosphide, 108
heterojunction bipolar technology, 108
Indium Antimonide, 108
Indium Phosphide, 108
Silicon Germanium, 108
IE3D, 136, 143, 149
IEEE 802.15.3, 7, 23, 26, 47, 158,
167–173
impedance matching, 131, 140, 150
implementation, 24, 40, 68, 72, 105,
109–117, 188
impulse radar, 2, 44
impulse radio, 2, 3, 21, 39–44, 57, 87, 88,
157, 166, 167
impulse response, 10, 11, 13–15, 20, 22–24,
33, 68, 69, 77, 80, 187
indoor, 1, 5, 6, 15, 22, 23, 28, 29, 132
inductance, 95, 97
inductor, 109, 115
Industrial, Scientific and Medical, 1
integrate and hold circuit, 106, 115
integrated circuit, 87–126
integrator, 88–90, 105, 106, 115, 124,
126, 127
intentional radiators, 4
interception, 2
interference, 3, 4, 40, 45, 64, 65, 67, 98,
166, 170, 186
interference avoidance, 44, 158
interference mitigation, 47
International Telecommunications
Union, 4
inter-operability, 158
interpath cancellation, 187
inter-pulse interference, 42
inter-symbol interference, 40, 46
inverse Fourier transform, 11, 27
J
jamming, 3
K
K-factor, 29
L
landmine sweeping, 44
Laplace operator, 106
licence free, 1, 2, 4
linearized least-square solution,
176, 181
line-of-sight, 27, 30, 36, 74, 191
load resistor, 95
location, 7, 36, 73, 80, 182, 184
location-awareness, 7, 164
location estimate, 175
lognormal distribution, 23
low complexity, 3, 157, 158, 168
low cost, 3, 129, 132, 157, 163
low noise amplifier, 15, 16, 110
M
magnitude, 21, 68, 110, 157, 187
man-made interference, 22
matched filter, 55, 72, 76, 77, 80, 89
maximum excess delay, 27
maximum likelihood, 76
mean acquisition time, 78, 79, 83, 85
MESFET, 96, 108
method of moments, 136
microwave/millimeter-wave IC, 108
military communications, 1
mining industry, 4
mismatch, 97, 116, 189
mixer, 45, 88, 92–94, 102, 103, 109
modulation, 2, 3, 39–64, 73, 94, 160
antipodal, 47, 53, 65, 74
index, 50, 55, 60, 111
orthogonal, 47, 65, 74
Index
pulse amplitude, 40, 48
pulse position modulation, 2, 49, 60,
91, 95
pulse shape modulation, 40, 51
monocycle, 40, 42, 55, 57, 88, 94, 111, 116,
118, 121
monopoles, 134
mono-pulse, 2, 137, 143, 149
MOS, 100, 109
m-sequence, 20, 184
multi-band, 47, 167
multi-carrier, 45–47
multipath, 3, 9, 23–25, 28, 67, 68, 76, 77,
80, 85, 185, 187, 188
multipath energy, 24
multipath propagation, 3, 33, 40, 79, 85
multiple access, 57, 60, 157, 158,
160, 172
multiple access interference, 64, 159, 162
Multiple Access/Collision Avoidance,
165, 170
multiplier, 88, 89, 102, 111, 115
multi-tone, 45
multi-user, 161
N
narrowband, 2, 3, 9, 10, 22, 131, 167, 172
NEC, 136, 137, 140, 141, 142
network topology, 163
noise figure, 15, 16, 109
non line-of-sight, 71, 188–191
non-centrality parameter, 29
non-coherent, 71–74
non-sinusoidal, 2
non-stationary, 19
nonlinear optimization theory, 177
O
observation time, 28
obstructions, 190, 191, 193
omni-directional, 15, 19, 132, 135,
141, 150
on-off keying, 48
operational amplifier, 106, 110, 115
operational frequency, 121, 122
orthogonal, 47, 50–54
221
orthogonal frequency division
multiplexing, 47, 158
oscillator, 3, 44, 92
outdoor, 5, 6, 22, 132
overhead, 164
P
packaging, 97, 98
parabolic reflector, 133
parasitic model, 97
path loss, 33–37
peak frequency, 102
peak-to-peak voltage, 118
peer-to-peer networks, 157, 166
penetration properties, 4
personal area network, 2
phase, 15, 18, 23, 67, 71, 76, 131
phase-locked loop, 88
physical layer, 7, 157, 158, 164,
167, 169
piconet, 169, 171, 173
PMOS, 111, 114, 118
Poisson process, 23
pole, 101
portable device, 97, 132, 168
positioning, 159, 175–195
post detection integration, 76, 77, 85
power, 1, 2, 22, 26, 32–35, 37, 39, 73, 91,
98, 109, 122, 137, 163, 179, 187
adaptation, 158
amplifier, 16, 88
consumption, 97
control, 163, 168
decay profile, 24
estimation, 72, 74
spectral density, 2, 6, 122
preamble, 165, 166, 171, 184
probability, 54, 79, 160, 179
probability of detection, 3, 77, 79, 85
probability of false alarm, 77, 79
probability of interception, 3
probability probability density
function, 76
probing signal, 15, 16, 20
processing gain, 40, 57, 64, 89, 117, 161
programmable delay, 88, 91, 117
222
propagation, 3, 21, 25, 33, 36, 38, 68, 69,
189, 193
propagation delay, 10, 14, 15, 16, 26,
27, 95, 171, 190
propagation loss, 33
propagation time, 10
protocol, 157, 158, 164–166, 168, 173
pseudo random, 3, 57, 64
pseudo random code, 20, 42, 45, 55, 57,
60, 61, 73, 91
pseudo random noise, 110, 116
pulse generator, 87, 88, 91, 93, 95, 110,
111, 118, 147
pulse repetition frequency, 60, 61, 88, 117,
159, 166
pulse repetition time, 57, 159, 160, 167
pulse shape distortion, 10
pulse shaping filter, 93, 127
pulse train, 54, 60, 90
PulsOn technology, 91
Q
quality of service, 163, 169
quasi-Newton algorithm, 182, 184
Q-value, 130
R
radar, 4, 21, 95, 131, 135
radiation efficiency, 132, 140
radiation limitations, 5
radiation mask, 5, 6
radio astronomy, 4
radio channel, 9, 14–16, 19, 21, 25,
40, 67
radio channel model, 7, 18, 21
radio controller, 94
rake, 68–74
all-rake, 68, 69, 71, 85
partial-rake, 68, 71, 85
selective-rake, 68, 71, 72, 85
ray arrival time, 23
ray tracing, 30, 191
Rayleigh distribution, 23
Rayleigh fading, 30
Rayleigh pulses, 40
rectangular waveform, 61
Index
reflecting surfaces, 9
reflections, 12, 13, 15, 24, 26, 27, 30, 38,
137, 138, 140, 149, 156, 191
regression line, 31, 35, 36
repetition coding, 160, 161
repetition rate, 164
rescue, 4
resistive load, 130, 132–134, 140, 142
resonant frequency, 109, 115
resource allocation, 166
RF tag, 157
Ricean fading, 29, 30
ringing, 10, 130, 137, 138, 140–142, 151
root mean square, 183
S
S21-parameter, 9
Saleh-Valenzuela channel, 82
Saleh-Valenzuela model, 23
scatterers, 28
Schottky diode, 95
sensing, 1, 6, 165
sensor, 157, 164, 179, 182, 184, 194
short range, 4, 5, 7, 55, 129, 132
short-pulse, 2, 42, 87, 94, 111,
113, 118
signal crosstalk, 97
simple current mirror, 107
slotted ALOHA, 170
small signal model, 99
source coupled pair, 98
source follower, 98
spatial domain, 9
spatial resolution, 25
spatial uncertainty, 175
spectral allocation, 20
spectrum, 2, 4, 12, 18, 20, 39, 42, 45, 54,
65, 122, 127, 137
spectrum mask, 39, 45
specular reflector, 9
spherical interpolation, 177
spread spectrum, 9, 20, 45, 54
spreading code, 10, 20, 67, 77
staggering, 1
statistical model, 30
step recovery diode, 95, 147
Index
structured network, 166
sub-carrier technology, 92
subspace techniques, 186, 187
super-frames, 169–172
switched capacitor, 106
symbol rate, 57, 123, 162
symmetry, 131
synchronization, 67, 75–83, 127
system capacity, 161
T
tag, 157, 177, 182, 183
tapped delay line, 28, 29, 38
Taylor series expansion, 176
template waveform, 80, 88, 89, 124
terrestrial, 1
the Federal Communications
Commission, 4, 5, 39, 87
three dimensional positioning, 176, 182
threshold, 77, 79, 80, 166, 186
throughput, 7, 166, 173
time difference of arrival, 176
time diversity, 68
time division multiple access, 169, 170, 172
time domain, 2, 3, 9, 11, 19, 42, 88, 130,
131, 136, 147
time domain resolution, 3
time frame, 42, 55, 94, 117
time hopping, 39, 42, 54, 57, 60, 61, 77, 87,
160, 166
time hopping frame, 40, 57, 60
time of arrival, 27, 159, 175, 186
time of flight, 10, 27, 175, 176
time resolution, 9, 12, 42, 68, 73, 175
time-modulated ultra wideband, 3, 4, 40
time-variant, 28
timing, 17, 18, 76, 88, 91, 92, 110, 111, 117,
124, 189
timing circuit, 88, 94, 110, 111, 116
timing precision, 3
223
T-model, 99
tracking, 67, 159
transceiver, 87, 93, 94, 109, 110
transconductance stage, 98, 99, 114
transistor, 95, 98, 105, 107–109, 114
transit frequency, 107–109
transmission line, 95, 111
transmission power, 4, 8, 87
transmitted pulse, 20, 42, 51, 65, 89,
124, 186
transmitted signal, 3, 40, 118, 127, 132,189
triangulation techniques, 175, 189
trigger, 16, 17, 19, 88, 116
Trinity, 94
TV broadcast bands, 4
two-dimensional positioning, 176
U
uncertainty region, 76, 77, 79, 188, 195
unity-gain frequency, 101, 107
unlicensed spectrum, 7
USB replacement, 157
V
varactor, 111, 113, 118
vector network analyzer, 9, 10, 13, 15, 16,
145, 148, 149
voltage controlled oscillator, 88
voltage standing wave ratio, 132
W
wavelength, 22, 38, 132, 137
white noise, 29, 57
wireless local area network, 2, 45, 47, 159
wireless personal area network, 7, 23,
167, 169
Z
zero padding, 11, 12, 13
易迪拓培训
专注于微波、射频、天线设计人才的培养
网址:http://www.edatop.com
射 频 和 天 线 设 计 培 训 课 程 推 荐
易迪拓培训(www.edatop.com)由数名来自于研发第一线的资深工程师发起成立,致力并专注于微
波、射频、天线设计研发人才的培养;我们于 2006 年整合合并微波 EDA 网(www.mweda.com),现
已发展成为国内最大的微波射频和天线设计人才培养基地,成功推出多套微波射频以及天线设计经典
培训课程和 ADS、HFSS 等专业软件使用培训课程,广受客户好评;并先后与人民邮电出版社、电子
工业出版社合作出版了多本专业图书,帮助数万名工程师提升了专业技术能力。客户遍布中兴通讯、
研通高频、埃威航电、国人通信等多家国内知名公司,以及台湾工业技术研究院、永业科技、全一电
子等多家台湾地区企业。
易迪拓培训课程列表:http://www.edatop.com/peixun/rfe/129.html
射频工程师养成培训课程套装
该套装精选了射频专业基础培训课程、射频仿真设计培训课程和射频电
路测量培训课程三个类别共 30 门视频培训课程和 3 本图书教材;旨在
引领学员全面学习一个射频工程师需要熟悉、理解和掌握的专业知识和
研发设计能力。通过套装的学习,能够让学员完全达到和胜任一个合格
的射频工程师的要求…
课程网址:http://www.edatop.com/peixun/rfe/110.html
ADS 学习培训课程套装
该套装是迄今国内最全面、最权威的 ADS 培训教程,共包含 10 门 ADS
学习培训课程。课程是由具有多年 ADS 使用经验的微波射频与通信系
统设计领域资深专家讲解,并多结合设计实例,由浅入深、详细而又
全面地讲解了 ADS 在微波射频电路设计、通信系统设计和电磁仿真设
计方面的内容。能让您在最短的时间内学会使用 ADS,迅速提升个人技
术能力,把 ADS 真正应用到实际研发工作中去,成为 ADS 设计专家...
课程网址: http://www.edatop.com/peixun/ads/13.html
HFSS 学习培训课程套装
该套课程套装包含了本站全部 HFSS 培训课程,是迄今国内最全面、最
专业的 HFSS 培训教程套装,可以帮助您从零开始,
全面深入学习 HFSS
的各项功能和在多个方面的工程应用。购买套装,更可超值赠送 3 个月
免费学习答疑,随时解答您学习过程中遇到的棘手问题,让您的 HFSS
学习更加轻松顺畅…
课程网址:http://www.edatop.com/peixun/hfss/11.html
`
易迪拓培训
专注于微波、射频、天线设计人才的培养
网址:http://www.edatop.com
CST 学习培训课程套装
该培训套装由易迪拓培训联合微波 EDA 网共同推出,是最全面、系统、
专业的 CST 微波工作室培训课程套装,所有课程都由经验丰富的专家授
课,视频教学,可以帮助您从零开始,全面系统地学习 CST 微波工作的
各项功能及其在微波射频、天线设计等领域的设计应用。且购买该套装,
还可超值赠送 3 个月免费学习答疑…
课程网址:http://www.edatop.com/peixun/cst/24.html
HFSS 天线设计培训课程套装
套装包含 6 门视频课程和 1 本图书,课程从基础讲起,内容由浅入深,
理论介绍和实际操作讲解相结合,全面系统的讲解了 HFSS 天线设计的
全过程。是国内最全面、最专业的 HFSS 天线设计课程,可以帮助您快
速学习掌握如何使用 HFSS 设计天线,让天线设计不再难…
课程网址:http://www.edatop.com/peixun/hfss/122.html
13.56MHz NFC/RFID 线圈天线设计培训课程套装
套装包含 4 门视频培训课程,培训将 13.56MHz 线圈天线设计原理和仿
真设计实践相结合,全面系统地讲解了 13.56MHz 线圈天线的工作原理、
设计方法、设计考量以及使用 HFSS 和 CST 仿真分析线圈天线的具体
操作,同时还介绍了 13.56MHz 线圈天线匹配电路的设计和调试。通过
该套课程的学习,可以帮助您快速学习掌握 13.56MHz 线圈天线及其匹
配电路的原理、设计和调试…
详情浏览:http://www.edatop.com/peixun/antenna/116.html
我们的课程优势:
※ 成立于 2004 年,10 多年丰富的行业经验,
※ 一直致力并专注于微波射频和天线设计工程师的培养,更了解该行业对人才的要求
※ 经验丰富的一线资深工程师讲授,结合实际工程案例,直观、实用、易学
联系我们:
※ 易迪拓培训官网:http://www.edatop.com
※ 微波 EDA 网:http://www.mweda.com
※ 官方淘宝店:http://shop36920890.taobao.com
专注于微波、射频、天线设计人才的培养
易迪拓培训
官方网址:http://www.edatop.com
淘宝网店:http://shop36920890.taobao.com
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertising