You are on page 1of 385

Microwave and Millimeter-Wave

Remote Sensing
for Security Applications
For a list of recent related titles in the Artech House Antennas and Propagation Series,
please turn to the back of this book.
Microwave and Millimeter-Wave
Remote Sensing
for Security Applications
Jeffrey A. Nanzer

artechhouse.com
Library of Congress Cataloging-in-Publication Data
A catalog record for this book is available from the U.S. Library of Congress.

British Library Cataloguing in Publication Data


A catalog record for this book is available from the British Library.

ISBN-13: 978-1-60807-172-2

Cover design by Vicki Kane

© 2012 Artech House


685 Canton Street
Norwood, MA 02062

All rights reserved. Printed and bound in the United States of America. No part
of this book may be reproduced or utilized in any form or by any means, elec-
tronic or mechanical, including photocopying, recording, or by any information
storage and retrieval system, without permission in writing from the publisher.
All terms mentioned in this book that are known to be trademarks or service
marks have been appropriately capitalized. Artech House cannot attest to the
accuracy of this information. Use of a term in this book should not be regarded
as affecting the validity of any trademark or service mark.

10 9 8 7 6 5 4 3 2 1
Contents

Preface xi

CHAPTER 1
Introduction 1
1.1  Security Sensing 1
1.1.1  Needs for Remote Security Sensing 1
1.1.2 Advantages of Microwave and Millimeter-Wave
Remote Sensors 2
1.2  Overview of Remote Sensing Techniques 3
1.2.1  Radiometry 3
1.2.2  Radar Systems 4
1.2.3  Imaging Systems 4
1.2.4  Interferometric Angular Velocity Measurement 5
1.2.5 Microwave and Millimeter-Wave Remote Sensing
in Related Fields 5
1.3  The Microwave and Millimeter-Wave Spectrum 7
1.3.1  Frequency Designations 7
1.3.2  Propagation of Microwave and Millimeter-Wave Radiation 8
1.4  Examples of Remote Security Sensors 9
1.4.1  Active Imaging for Contraband Detection 10
1.4.2  Passive Imaging for Contraband Detection 10
1.4.3  Detection of Human Presence 12
1.4.4 Discrimination of Humans and Classification of
Human Activity 18
1.4.5  Through-Wall Detection 19
1.4.6  Biological Signature Detection 20
References 20

CHAPTER 2
Electromagnetic Plane Wave Fundamentals 27
2.1  Maxwell’s Equations 27
2.1.1  The Constitutive Parameters 30
2.2  Time-Harmonic Electromagnetic Fields 31
2.2.1  The Wave Equation 32
2.2.2  Plane Waves 33
2.2.2.1  Phase Velocity 34
2.2.2.2  Relationship Between E and H 35
2.2.3  Energy and Power 37


vi Contents

2.3  Wave Polarization 38


2.3.1  Linear Polarization 39
2.3.2  Elliptical Polarization 40
References 42

CHAPTER 3
Electromagnetic Waves in Media 43
3.1  Plane Wave Propagation in Unbounded Media 44
3.1.1  Good Conducting Media 46
3.1.2  Good Dielectric Media 47
3.1.3  Wave Impedance in Media 48
3.1.4  Complex Permittivity and Dispersion 48
3.2  Plane Wave Propagation in Bounded Media 51
3.2.1  Reflection and Transmission of Normally Incident Waves 52
3.2.2  Reflection and Transmission of Arbitrarily Incident Waves 54
3.2.2.1  Transverse Electric (Perpendicular) Incidence 54
3.2.2.2  Transverse Magnetic (Parallel) Incidence 57
3.2.3  Power Reflection and Transmission 58
3.2.4  Total Transmission and Total Reflection 60
3.2.5  Layered Media 61
3.3  Electromagnetic Propagation in Specific Media 63
3.3.1  Atmospheric Propagation Effects 63
3.3.2  Propagation Through Building Materials 69
3.3.3  Propagation Through Clothing and Garment Materials 70
3.3.4  Dielectric Properties of Explosives, Plastics, and Metals 71
3.3.5  Dielectric Properties of Human Tissue 72
References 81

chapter 4
Antennas 85
4.1  Electromagnetic Potentials 86
4.1.1  Electromagnetic Potentials Due to Electric Current Density J 86
4.1.2  Electromagnetic Potentials Due to Magnetic Current Density Jm 88
4.1.3  Infinitesimal Dipole Radiation 89
4.1.4  Far Field Radiation 90
4.1.5  Infinitesimal Dipole Far-Field Radiation 94
4.2  Antenna Parameters 95
4.2.1  Radiated Power Density and Total Radiated Power 95
4.2.2  Antenna Pattern 96
4.2.3  Antenna Pattern Beamwidth 97
4.2.4  Antenna Solid Angles 99
4.2.5  Directivity 99
4.2.6  Gain 101
4.2.7  Aperture Area and Pattern Solid Angle 102
4.2.8  Antenna Temperature and Noise Power 103
4.2.9  Polarization 103
Contents vii

4.3  Properties of Wire Antennas 104


4.3.1  Infinitesimal Dipole 104
4.3.2  Long Dipole 105
4.4  Aperture Antennas 107
4.4.1  Image theory 108
4.4.2  The Equivalence Principle 109
4.4.3  Radiation from a Rectangular Aperture 111
4.4.4  Radiation from a Circular Aperture 115
4.5  Antenna Arrays 117
4.5.1  Linear Array Theory 118
4.5.2  Planar Arrays 121
4.5.3  Array Beamwidth 122
4.5.4  Phased Arrays 123
4.5.5  Array Architectures 125
4.5.5.1  Signal Feeds 125
4.5.5.2  Beam Steering 127
4.6  Common Microwave and Millimeter-Wave Antennas 128
4.6.1  Horn Antennas 128
4.6.2  Slot Antennas 131
4.6.3  Microstrip Antennas 132
4.6.4  Reflector Antenna Systems 134
4.6.5  Lens Antenna Systems 136
References 137

chapter 5
Receivers 139
5.1  General Operation of Receivers 140
5.2  Receiver Noise 143
5.2.1  Sources of Receiver Noise 144
5.2.1.1  Thermal Noise 144
5.2.1.2  Shot Noise 145
5.2.1.3  Flicker Noise 146
5.2.2  Equivalent Noise Bandwidth 146
5.2.3  Thermal Noise at Millimeter-Wave Frequencies 148
5.3  Noise Figure and Noise Temperature 150
5.3.1  Noise Figure 150
5.3.2  Noise Temperature 152
5.3.3  Noise Figure of an Attenuator 153
5.3.4  Noise in Cascaded Systems 154
5.3.5  ADC Noise 157
5.4  Receiver Linearity 160
5.4.1  Gain Compression 162
5.4.2  Intermodulation Products 164
5.4.3  Third Order Intercept Point 166
5.4.4  Intercept Point of a Cascade 168
5.4.5  Dynamic Range 168
viii Contents

5.4.6  Spurious Free Dynamic Range 170


References 171

chapter 6
Radiometry 173
6.1  Radiometry Fundamentals 174
6.1.1  Brightness 174
6.1.2  Brightness and Distance 176
6.1.3  Flux Density and Source Distribution 178
6.1.4  Effect of the Antenna 179
6.2  Blackbody Radiation 180
6.2.1  Planck’s Blackbody Radiation Law 180
6.2.2  Approximations of Planck’s Law 184
6.2.3  Band-Limited Integration of Planck’s Law 185
6.3  Applied Radiometry 187
6.3.1  Source Resolution 188
6.3.1.1  Resolved Source 188
6.3.1.2  Unresolved Source 189
6.3.2  Received Power as a Convolution 190
6.3.3  Emissivity and Radiometric Temperature 191
6.3.3.1  Emissivities of Human Skin and Common Materials 192
6.3.3.2  Radiometric Temperature in an Environment 194
6.4  Radiometer Receivers 196
6.4.1  Sensitivity 197
6.4.2  Total Power Radiometer 200
6.4.2.1  Total Power Response 200
6.4.2.2  Sensitivity 201
6.4.3  Interferometric Correlation Radiometer 206
6.4.3.1  Spatial Point Source Response 207
6.4.3.2  Sensitivity 212
6.5  Practical Considerations 215
6.5.1  Receiver Instabilities 215
6.5.2  Dicke Radiometer 215
6.5.3  Radiometer Calibration 217
6.6  Scanning Radiometer Systems 218
6.6.1  Spatial Resolution 219
6.6.2  Dwell Time 222
6.6.3  Measurement Uncertainty 223
6.6.3.1  One-Dimensional Scanning 223
6.6.3.2  Two-Dimensional Scanning 225
References 226

chapter 7
Radar 229
7.1  Radar Fundamentals 230
7.1.1  Configurations and Measurements 231
Contents ix

7.1.2  Range Equation 233


7.2  Transmitter Systems 236
7.2.1  Transmitter Functionality 236
7.2.2  Transmitter Noise 239
7.2.3  Millimeter-Wave Oscillators 241
7.3  Radar Measurement Sensitivity 243
7.3.1  Measurement Error 243
7.3.1.1  Range Measurement Error 244
7.3.1.2  Frequency Measurement Error 245
7.3.1.3  Angle Measurement Error 245
7.3.1.4  Example 245
7.3.2  Impact of the Time-Bandwidth Product on Measurement Error 251
7.4  Micro-Doppler 253
7.4.1  Micro-Doppler in Security Radar 254
7.4.2  Micro-Doppler Theory 255
7.4.3  Human Micro-Doppler Signature 260
7.5  Continuous-Wave Radar 266
7.5.1  Continuous-Wave Doppler 267
7.5.2  Frequency-Modulated CW 271
7.5.3  Multifrequency CW 274
7.5.4  Moving Target Indication Radar 275
7.6  High-Range Resolution Radar 279
7.6.1  Pulse Radar 280
7.6.2  Linear Frequency Modulation 282
7.6.3  Stepped-Frequency Modulation 285
References 286

chapter 8
Imaging Systems 289
8.1  Scanning Imaging Systems 291
8.1.1  Types of Scanning Imagers 291
8.1.2  General Characteristics of Scanning Systems 292
8.1.2.1  Field of View and Spatial Resolution 292
8.1.2.2  Frame Rate 294
8.2  Interferometric Imaging Systems 295
8.2.1  Introduction 295
8.2.2  Image Formation 296
8.2.2.1  Visibility Function 297
8.2.2.2 Fourier Transform Relationship of Visibility and
Radiometric Temperature 299
8.2.2.3  The Correlation Interferometer as a Spatial Filter 301
8.2.3  Visibility Sampling 303
8.2.4  Two-Dimensional Visibility 308
8.2.5  Image Sensitivity 309
8.2.6  Image Resolution and Field of View 312
8.2.7  Interferometric Imaging Arrays 318
 Contents

8.2.7.1  Mills Cross Array 319


8.2.7.2  T-Array 321
8.2.7.3  Y-Array 322
8.2.7.4  Circular Arrays 323
References 325

chapter 9
Interferometric Measurement of Angular Velocity 329
9.1  Interferometer Response to an Angularly Moving Point Source 330
9.1.1  System Beam Pattern 331
9.1.2  Frequency Shift Induced by an Angularly Moving Object 332
9.1.3  Comparison to Doppler Frequency Shift 333
9.1.4  Frequency Uncertainty at Wide Angles 335
9.1.5  Small Angle Approximation 335
9.2  Interferometer Spectral Response 336
9.2.1  General Spectral Response 336
9.2.2  Response with a Sinc Function System Beam Pattern 337
9.2.3  Interferometer Response in the Time-Frequency Domain 341
9.3  Interferometric Measurement of Moving Humans 344
9.3.1  Narrow-Beamwidth Response to a Moving Human 344
9.3.2  Wide-Beamwidth Response to a Moving Human 346
References 349

List of Symbols 351

List of Abbreviations and Acronyms 355

About the Author 357

Index 359
Preface

Microwave and millimeter-wave remote sensing techniques are quickly becoming


important tools as threats to security in various situations become more sophisti-
cated and difficult to counter. Detecting intruders in exterior environments requires
all-weather capability; the ability to operate through fog, smoke, and other obscu-
rants; and the ability to detect features that can be used to discriminate humans and
nonhumans. Detecting concealed contraband, such as guns, knives, and explosives,
requires the ability detect through fabric and wall materials and generate images
with resolutions on the order of millimeters. The last decade has seen a dramatic in-
crease in the research and development of microwave and millimeter-wave sensors
for these situations, providing a rich body of work covering radiometric and radar
imaging systems for contraband detection, radiometry and Doppler radar systems
for intruder detection, and micro-Doppler radar analysis for the classification of
human activity. The field continues to develop, generating new techniques in re-
sponse to continually changing security-sensing needs. This book presents the
fundamental principles and advanced techniques used in remote security sensing.
The motivation for this book came to me during my time at the University of
Texas Applied Research Laboratories, where my task was to develop passive and
active millimeter-wave sensors for the detection of moving and stationary people,
with the goal of automated intruder detection from robotic platforms. I quickly
found that applying radiometry to security sensing was a novel idea, in that I could
not find sufficient literature to help guide me as I developed various millimeter-
wave radiometers. I found myself drawing from the fields of radio astronomy and
satellite remote sensing for radiometry literature (some of which was long out of
print), and I was frustrated by the lack of a textbook directed toward the basic
techniques needed for small-scale remote sensing applications. In addition, my col-
leagues were actively developing radar systems to exploit the micro-Doppler signature
of moving humans, with the goal of activity classification. Along with previous work
on millimeter-wave radiometric imaging by my colleagues at ARL and the emerging
body of literature on security imaging, it was apparent to me that the field of security
sensing was emerging as a new area for remote sensing and that a textbook presenting
the fundamentals of sensor design would be significantly useful.
The goal of this book is to present the principles of microwave and millimeter-
wave remote sensing for security applications. The two general areas comprising
the physical aspects of remote sensing are electromagnetic wave propagation and
remote sensing systems. The antenna represents the transition between these two
areas. Signal processing may be considered a third area, following the sensor hard-
ware; however, the implementation of signal processing algorithms is highly spe-
cific to the security sensing task and thus varies in implementation. As such, signal
processing is not specifically covered in this book. Chapter 1 gives an overview of

xi
xii Preface

remote sensing applications in security sensing, with examples of sensors from pub-
lished literature. Basic electromagnetic wave propagation is covered in Chapter 2,
and in Chapter 3 wave propagation through media is covered, focusing on media
encountered in security sensing applications, including propagation through air,
obscurants such as smoke and dust, precipitants including fog and rain, building
walls, clothing material, and interactions with human tissue. In Chapter 4 antenna
theory is presented, and Chapter 5 covers receiver theory. Principles of radiom-
etry are presented in Chapter 6, including blackbody and greybody radiation and
specific radiometer configurations including the total power and correlation radi-
ometers. Radar principles and systems are covered in Chapter 7, a large portion of
which is also devoted to the relatively new field of human micro-Doppler. Imaging
systems are covered in Chapter 8, with the majority of the chapter focusing on in-
terferometric imaging, a new and promising application in remote security sensing.
The book concludes with Chapter 9 presenting a new technique of measuring the
angular velocity of moving objects using a correlation interferometer.
This book is intended for practicing engineers and researchers who are develop-
ing remote sensors for security-related applications and for graduate and advanced
undergraduate students studying remote sensing. Many of the topics covered in the
book are relevant to remote sensing in general, and researchers in related remote
sensing fields may find the book to be a useful reference. The reader is assumed to
have a background in calculus and Fourier analysis, and although the material is
derived from basic principles, prior exposure to electromagnetic theory will also be
beneficial.
I am extremely grateful for the people who have provided support and guid-
ance in the development of this book. In particular, I would like to thank those
who have reviewed chapters of the manuscript: Andrew Temme at Michigan State
University, and Salvador Talisa and Keir Lauritzen at the Johns Hopkins University
Applied Physics Laboratory. I also thank the staff at Artech House for their profes-
sionalism and support, and their anonymous reviewer for thorough comments. My
thanks to Carl Nielson, for helping with some of the figures in Chapter 3, and to
my colleagues at the Johns Hopkins University Applied Physics Laboratory who
have provided encouragement and helpful discussions along the way. I am indebted
to those who helped shape my educational and professional career: Ed Rothwell at
Michigan State University, Bob Rogers at the University of Texas Applied Research
Laboratories, and Hao Ling at the University of Texas. Finally, I owe my greatest
thanks to my wife and children for supporting this endeavor and tolerating the long
hours spent developing this book.
Chapter 1

Introduction

1.1  Security Sensing

1.1.1  Needs for Remote Security Sensing


Increased security concerns in the past decade can be traced largely to the threat of
terrorism, a result of which has been heightened interest in the detection of humans,
the classification of human activity, and the detection of hidden objects that may be
carried by a person or concealed in walls. Threats to security are constantly chang-
ing, targeting weak areas of security as others are strengthened, and thus there
must be a corresponding evolution of security sensing to counter new threats. The
applications of security sensing expand beyond counter-terrorism: border security
involves detecting illegal immigration and intercepting drugs; military force protec-
tion involves detecting humans and determining their intent through activity clas-
sification; law-enforcement benefits from the detection and classification of humans
through walls; and search-and-rescue operations require the detection of people
hidden within buildings or beneath building materials. Addressing these situations
requires sensors that can operate in nearly all weather conditions, sense through
obscurants and clothing, and measure with resolution fine enough to detect small
hidden objects.
The increased interest in countering threats has led to the advance of sensor
technology to counter a wide range of threats, and the corresponding sensing tech-
niques are continually evolving and developing. Such techniques include x-ray
imagers, biological and chemical agent sensors, infrared cameras, spectroscopic
sensors, acoustic sensors, and THz imagers. A given sensor or sensing technique
often yields good performance in a specific application or setting but cannot coun-
ter all perceived threats. For instance, infrared imagers can detect the presence of
a human particularly well against a cool background, such as an indoor environ-
ment or at night; however, some environments, such as those encountered during
the daytime where the background is warm and solar reflections are more ubiq-
uitous, yield poorer human presence detection performance. Infrared sensors are
also ­incapacitated by obscurants such as fog, smoke, or dust. Detection of chemical
traces using spectroscopy includes the benefit of remote measurement; however,
it generally has lower performance than chemical trace sensors directly analyzing
air samples. THz imagers, while providing fine resolution, suffer from high signal
attenuation through the atmosphere, obscurants, clothing, and walls, resulting in
sensors with short operational ranges. Some sensors are undesirable due to their
nature: x-ray imagers, while successful at detecting contraband, can cause concern
due to the radiation levels required. Nonionizing radiation, such as microwave and


 Microwave and Millimeter-Wave Remote Sensing for Security Applications

millimeter-wave radiation, is often preferred where detection of people or objects


on people is the goal.
A single security sensing technique cannot address all security concerns, and
microwave and millimeter-wave sensors represent a unique and complementary
option to other methods. Microwave and millimeter-wave remote sensors have sig-
nificant potential to counter threats such as concealed objects (explosives, guns,
knives, drugs, and so on) under clothing or hidden in or behind walls, illegal border
crossings, intrusion into secured areas, and threatening activities by people. Remote
sensing, a mature and robust field, is being applied to these applications of security
sensing with increasing ingenuity: millimeter-wave radar systems are being applied
to high-range resolution concealed object detection, human presence detection and
through-wall sensing; and radiometry, long the domain of radio astronomy and
satellite remote sensing, is rapidly developing into the sensing method of choice for
detecting hidden contraband through imaging systems. In these implementations,
detection and discrimination can be accomplished from a distance, increasing the
safety of the operator of the sensor.

1.1.2  Advantages of Microwave and Millimeter-Wave Remote Sensors


Sensors operating in the microwave and millimeter-wave bands have a number of
unique characteristics that complement other security sensing methods. As such,
research in microwave and millimeter-wave technology has seen a dramatic increase
in the past decade. Fueled by reductions in component costs, increased bandwidth
capability of digitizers, and advances in signal and image processing techniques, the
benefits of microwave and millimeter-wave remote sensors in security applications
are now being realized in practice. These benefits will be covered in detail through-
out this book and include the following:

· Propagation through the atmosphere with minimal attenuation at most


­frequencies;
· Propagation through obscurants such as fog, haze, smoke, dust, and light to
medium rain with negligible attenuation;
· Propagation through clothing material, baggage material and some building
wall material with negligible or minimal attenuation;
· High thermal radiation power from humans relative to nonhuman objects;
· Fine range resolution due to wide-bandwidth signal capability;
· Fine angular resolution due to short wavelength radiation;
· Fine resolution of the radial velocity of objects due to high carrier
­frequencies;
· Physically small systems due to short wavelengths.

These characteristics support detection and imaging of humans and contraband


at long ranges and through walls, obscurants, and clothing. The capability to see
through obscurants and clothing is a primary benefit that has driven a large por-
tion of research in contraband detection, as well as sensors for imaging through fog
and dust for various imaging applications such as aiding in landing of helicopters
and other vehicles. The contrast between the thermal radiation emitted by humans
1.2  Overview of Remote Sensing Techniques 

and that emitted or reflected by other materials supports the detection of human
presence in cluttered environments and the detection of objects concealed beneath
clothing. The capability to detect humans in nearly all weather conditions, coupled
with the fine Doppler frequency resolutions that can be achieved, has pushed the de-
velopment of radar sensors for remote human presence detection and activity clas-
sification. Additionally, because the angular resolution of a sensor is proportional
to the physical size of the antenna in terms of the wavelength, systems operating at
higher frequencies can generate images with finer resolution without the need for
very large apertures.

1.2  Overview of Remote Sensing Techniques

Remote sensors determine information about an object or scene by transmitting a


signal and analyzing the signal reflected back to the sensor or by measuring a signal
intrinsically emitted by the object itself. The signals are in the form of electromag-
netic waves, which propagate through space and materials with some loss, which
may be negligible. Remote detection and interrogation of an object is beneficial in
security sensing, as it allows an operator to investigate a potential threat without
the need for physical inspection. This improves safety in the detection of explosives
and other weapons, reduces operator burden in border security and site security
monitoring, and allows discrete surveillance.
Remote sensors can be generally categorized as either active or passive systems.
Active systems transmit a signal and measure the signal that reflects off the object and
back to the sensor. By analyzing aspects of the return signal, such as the time between
the transmit and return signals or the power level or frequency of the return signal,
various characteristics of the remote object can be discerned. Passive systems rely on
measuring a signal that is intrinsic to the object, such as thermal radiation. Imaging
systems use active or passive sensors to generate two-dimensional images of a scene,
where image processing is generally required for detection and discrimination.

1.2.1  Radiometry
Radiometers are in essence highly sensitive receivers designed to measure the ther-
mal radiation that is intrinsically emitted by all objects. Radiometers are passive
systems: no signal is transmitted and reflected off the object. Because the signals
detected are intrinsic to the object, detection can be accomplished regardless of
whether the object is moving or stationary. Thermal signatures themselves may be
used for detection or discrimination, or a contrast between the thermal radiation of
humans and nonhumans may be used to detect people or objects.
Thermal radiation is generated by the motion of electrons in a material with a
nonzero temperature. The thermal power radiated by an object is determined by
Planck’s law, which states that the power is a function of temperature and frequency,
and is covered in detail in Chapter 6. An object at the temperature of the human
body (310 K) radiates the most power in the infrared region of the electromagnetic
spectrum, with lower power levels at lower and higher frequencies. Thermal radia-
tion in the microwave and millimeter-wave regions can be approximated by a linear
 Microwave and Millimeter-Wave Remote Sensing for Security Applications

relationship between the radiated power and the product of the temperature of the
object and the bandwidth of the receiver. In implementing radiometers at millimeter-
wave frequencies, thermal radiation from the body can be measured through cloth-
ing, allowing the detection of hidden people or objects hidden on the body. The
radiated power levels are low, on the order of 10–10 W or lower, and thus high-gain
systems must be designed in order to measure the differences in power.
A radiometer can be implemented in a number of ways. The most common
forms are the total power radiometer and the correlation radiometer, discussed
further in Chapter 6. The total power radiometer produces a voltage signal that is
a measure of the total power present in the bandwidth of the system. This power
level includes both power emitted by the object that is directly proportional to the
physical temperature of the object and any noise power generated in the system.
The correlation radiometer consists of two receivers, the outputs of which are cor-
related. The resulting signal response is proportional to the power emitted by the
object; ideally, uncorrelated noise responses from the receivers are removed by the
correlation process. Both the total power and correlation radiometers produce a
signal response that is proportional to the temperature of the object.

1.2.2  Radar Systems


Radar systems are active systems that transmit a signal and analyze the signal re-
flected off the object of interest. They consist of a transmitter and a receiver, which
may be collocated on the same platform and may use the same antenna. The trans-
mitted signal may be a continuous-wave signal or it may be pulsed, and it may
change in frequency. The received signal, collected after reflecting off the object,
is analyzed to determine various characteristics of the object. The difference in
time between transmitting a signal and the receipt of the reflected signal is propor-
tional to the distance to the object, and the shift in the frequency of the signal is
proportional to the radial velocity of the object. The range extent of objects can
be measured, the resolution of which is inversely proportional to the waveform
bandwidth, as discussed in Chapter 7. Millimeter-wave radars can support wide
bandwidths, yielding fine range resolution.
Radar systems can detect objects moving in a radial direction toward or away
from the sensor by measuring the frequency shift between the transmitted and re-
ceived signals, called the Doppler shift, and can thus be applied to the detection of
moving humans for intruder detection and border security applications and can be
used to measure biological signatures from respiration and heartbeat. Differences
in the power levels of the returned signals can also be used to classify different ma-
terials by their reflectivity and can thus be applied to concealed object detection.
Chapter 7 covers fundamental aspects of radar systems in detail.

1.2.3  Imaging Systems


Millimeter-wave imagers generate two-dimensional representations of a scene
through multiple active or passive measurements. Imaging systems are generally
implemented at millimeter-wave frequencies rather than microwave frequencies
because the shorter wavelengths of millimeter-wave radiation allows for finer
1.2  Overview of Remote Sensing Techniques 

resolution images with smaller systems than can be achieved at lower microwave
­frequencies. The images may be of the reflectivity in the case of an active imager or
the radiometric temperature profile in the case of a passive radiometric imager.
Formation of the images can be achieved using a scanning configuration or a
staring configuration. Scanning imagers are implemented using either mechanical
steering of a single antenna beam, electronic steering of the beam of an antenna
array, or a combination of the two. Mechanically steered imagers generally use a
single antenna, often a reflector antenna, and steer the beam across the image pixel
locations by placing either the feed antenna, the reflector, or the entire antenna
system on a rotator and physically moving the beam. Electrically steered imagers
use phased arrays or frequency-steered arrays to move the beam. Mechanically
steered imagers are generally slower to form images but benefit from their relative
simplicity of implementation and lower cost. Electrically steered imagers are gen-
erally faster than mechanically steered imagers; however, they include significant
complexity due to the implementation of an array and the associated beam-steering
controls. Staring imagers utilize multiple receiving elements corresponding to pixel
locations, where the electromagnetic wave is often focused onto the elements using
a lens to increase the aperture size and improve the resolution.
Interferometric imagers, long used in radio astronomy and satellite remote sens-
ing, have recently been applied to security imaging, and are discussed in detail in
Chapter 8. Interferometric imagers do not require steering and can be implemented
with a fraction of the elements needed in a full phased array. Generally implemented
as passive imagers, the image is formed through the cross–correlation of the received
signals between all pairs of antennas. Interferometric imagers thus require a large
number of correlation processes, often implemented digitally in modern systems.

1.2.4  Interferometric Angular Velocity Measurement


Whereas radar systems can measure the radial velocity of moving objects by detect-
ing the frequency shift of the return signal, measurement of the angular velocity of
moving objects is more difficult and is generally performed by tracking the object
through repeated estimates of the angle of the object. Recent demonstrations of a
method of measuring the angular velocity of moving objects using a two-element
correlation interferometer have shown the potential to measure the angular motion
of objects and humans directly, without the need for tracking. With this method,
an object moving through the interferometer beam pattern produces a frequency
shift on the sensor output signal that is proportional to the angular velocity of the
object, in a way that is mathematically similar to how the frequency shift in a radar
signal is proportional to the radial velocity. While demonstrating potential, the
interferometric measurement of angular velocity is a relatively new technique and
must be evaluated through future studies and experiments. The basic theory and
initial experimental results of this method are reviewed in Chapter 9.

1.2.5  Microwave and Millimeter-Wave Remote Sensing in Related Fields


The applications of radar and radiometry in remote sensing are broad. Since its
development during World War II, radar has become a very general and widespread
 Microwave and Millimeter-Wave Remote Sensing for Security Applications

remote sensing technique. It is used on ground- and air-based platforms, in large


phased arrays, and in small single-element sensors. Initially, radar was used for the
detection and tracking of vehicles, primarily airborne, where it is still used today in
both military and commercial applications. Measurement of the velocity of moving
objects through the induced Doppler shift is used in police radar for determining
vehicle speed and weather radar for measuring precipitation. The ranging capabil-
ity of radars provides one of the more ubiquitous functions of remote sensing in
general and is used for vehicle collision avoidance, ionospheric sounding, military
guidance applications, intragalactic astronomy, and industrial control, to name a
few. Uses of radar systems on airborne and spaceborne platforms include altimetry,
topographic mapping, vegetation coverage mapping, oil spill detection and moni-
toring, and measurements of soil moisture content.
Applications of radiometry are less widespread than radar. Traditionally, the
two major areas where microwave and millimeter-wave radiometry have found
significant use are radio astronomy and satellite remote sensing. Throughout the
middle of the twentieth century, radio telescopes and radio interferometers were de-
veloped to generate fine angular resolution images of the radio frequency radiation
of stellar objects. Radar measurements for this purpose are only practical within
our own galaxy due to the distances between our planet and extragalactic objects.
A primary development from radio astronomy was the interferometric receiver that
utilizes multiple antennas separated by a number of wavelengths to generate images
with resolutions equal to that of a single aperture with dimensions equivalent to the
distance between the two farthest antennas. The interferometric imaging technique
requires at least two antennas; they can be moved to form multiple baselines if the
viewing time is long enough. Radio astronomers have the benefit of stationarity
when viewing stellar objects: multiple measurements can be taken on successive
days because the objects do not move. Significant development of satellite remote
sensors occurred in the 1980s. Leveraging the interferometric technique developed
in radio astronomy, antenna arrays were developed using a small number of anten-
nas to synthesize a larger aperture. The benefit to satellite remote sensing came in
the reduction of necessary antenna hardware (and therefore weight) that needed
to be placed into orbit. Measurements that can be conducted using passive satel-
lite remote sensors include soil moisture content, vegetation coverage, surface tem-
perature maps, ocean surface salinity and temperature, ocean wind speed, sea ice
coverage, weather conditions, wildfire detection and monitoring, and ionospheric
measurements.
The application of remote sensing techniques to security is a relatively new en-
deavor, and one that can leverage developments made in the areas mentioned above.
Radar applications are clearly applicable to the detection of moving h­umans and
hidden contraband, as well as respiration and heartbeat detection. Security sensing
can be viewed as an emerging significant area of development for microwave and
millimeter-wave radiometry, following radio astronomy and satellite remote sens-
ing. Its uses are in the detection of stationary humans through the measurement
of intrinsic thermal radiation, and detection of concealed objects by measuring
physical temperatures and temperature differences between objects and the human
body.
1.3  The Microwave and Millimeter-Wave Spectrum 

1.3  The Microwave and Millimeter-Wave Spectrum

1.3.1  Frequency Designations


The region of the electromagnetic spectrum designated as the microwave band ex-
tends from 3 GHz to 30 GHz, while the millimeter-wave band extends from 30
GHz to 300 GHz. The radio frequency portion of the electromagnetic spectrum as
defined by the International Telecommunication Union (ITU) is shown in Table 1.1.
The microwave and millimeter-wave regions are covered by the super high fre-
quency (SHF) and extremely high frequency (EHF) bands on the ITU spectrum.
Commonly used in the United States are the IEEE letter designations [1], given in
Table 1.2. Under this definition, the microwave region may be considered to begin
in S-band and extend to Ka-band while the millimeter-wave region extends from Ka-
band to the millimeter-wave band, which does not have a standard letter designa-
tion. Figure 1.1 shows the microwave and millimeter-wave spectrum as defined by
the ITU and the IEEE. Frequencies between 300 GHz and 3,000 GHz do not have a
designation and are generally called simply terahertz (THz) frequencies, and while
the focus of this book is on the microwave and millimeter-wave regions, frequencies
extending into the THz region are considered where relevant.
Applications exploiting the microwave band are myriad and include commu-
nications, RFID, wireless networking, and satellite broadcasting, among others.
Remote sensing applications are widely used in the microwave band due to favor-
able atmospheric propagation characteristics. Modern radar systems are generally
implemented in the microwave region, as are radio astronomy arrays and satellite
remote sensors. Remote sensing applications in the millimeter-wave region have
traditionally consisted primarily of military sensors operating at the W-band; how-
ever, there has been significant work toward millimeter-wave remote sensors in
other areas in recent years, including radio astronomy arrays that are being devel-
oped at frequencies in the hundreds of gigahertz and sensors for imaging applica-
tions. Atmospheric attenuation is greater at millimeter-wave frequencies than at
microwave frequencies, making satellite remote sensing of the earth more difficult.
Driven by the wide bandwidths that can be supported, interest has increased for
the development of communication systems and remote sensors at millimeter-wave

Table 1.1  International Telecommunication Union Frequency Bands


Band Number Designation Frequencies
1 Extremely low frequency (ELF) 3–30 Hz
2 Super low frequency (SLF) 30–300 Hz
3 Ultra low frequency (ULF) 300–3,000 Hz
4 Very low frequency (VLF) 3–30 KHz
5 Low frequency (LF) 30–300 KHz
6 Medium frequency (MF) 300–3,000 KHz
7 High frequency (HF) 3–30 MHz
8 Very high frequency (VHF) 30–300 MHz
9 Ultra high frequency (UHF) 300–3,000 MHz
10 Super high frequency (SHF) 3–30 GHz
11 Extremely high frequency (EHF) 30–300 GHz
12 – 300–3,000 GHz
 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Table 1.2  IEEE Standard Letter Designations


for Radar Frequency Bands
Designation Frequencies
HF 3–30 MHz
VHF 30–300 MHz
UHF 300–1,000 MHz
L 1–2 GHz
S 2–4 GHz
C 4–8 GHz
X 8–12 GHz
Ku 12–18 GHz
K 18–26.5 GHz
Ka 26.5–40 GHz
V 40–75 GHz
W 75–110 GHz
mmw 110–300 GHz

carriers. The result has been component and system technology with increased ca-
pabilities and reduced cost, making millimeter-wave remote security sensor systems
more realizable.

1.3.2  Propagation of Microwave and Millimeter-Wave Radiation


Electromagnetic radiation in remote security sensing encounters various media be-
tween the object of interest and the antenna. Radiation travels through the atmo-
sphere in all cases, albeit it over short distances compared to some remote sensing
applications such as radio astronomy and satellite remote sensing. Various mecha-
nisms in the atmosphere cause radiation to be absorbed at some frequencies, whereas
other frequencies are passed through with negligible attenuation. Figure 1.2 shows
the absorption caused by the atmosphere over the microwave and millimeter-wave
bands. There are a number of regions, referred to as windows, where radiation can
pass through with minimal loss. Other regions, such as that around 60 GHz, have
very high losses, making remote sensing over long distances difficult.
Obscurants in the air can render sensors such as infrared and optical sensors
ineffective; in particular, fog, light rain, smoke, and dust absorb the majority of

ITU microwave and millimeter-wave band designations


SHF EHF
3 30 300
f (GHz)

IEEE microwave and millimeter-wave band designations


S C X Ku K Ka V W mmw
2 4 8 12 18 26.5 40 75 110 300
f (GHz)
Figure 1.1  ITU radio frequency band designations and IEEE standard letter band designations for
radar frequencies.
1.4  Examples of Remote Security Sensors 

0.8

Transmission (%)
0.6

0.4

0.2

0
0 100 200 300 400 500
f (GHz)
Figure 1.2  Atmospheric absorption in the microwave and millimeter-wave bands.

the radiation at optical and infrared frequencies. Microwave radiation, however,


transmits through such obscurants with negligible attenuation, as does the radia-
tion in most of the millimeter-wave band. Attenuation at some specific frequencies
can cause additional attenuation due to water absorption. Avoiding these regions,
sensors can be developed that can operate in nearly all conditions that would be
encountered in an outdoor environment.
The detection of human presence and concealed objects is enabled in the micro­
wave and millimeter-wave bands due to favorable propagation characteristics of
many building materials. Some materials, such as drywall, are nearly transparent
across these bands, whereas some, such as wood, have low transmission loss at low
frequencies but higher loss as the frequency increases. Concrete and brick generally
exhibit high loss, making detection through such materials difficult with a passive
system; however, active systems can be designed with increased transmit power to
overcome the loss incurred by the wall materials. The effectiveness of the sensor
thus generally depends on the type of building material. The loss incurred by micro­
wave and millimeter-wave radiation as it transmits through clothing materials is
also generally negligible, supporting contraband detection and passive radiometric
human presence detection where the intrinsic thermal radiation generated by the
human body is detected.

1.4  Examples of Remote Security Sensors

Remote sensing techniques have been applied to numerous applications in security


and related areas, and are continually being investigated for applicability in new
areas. In this book, the focus is on the most promising areas that have shown suc-
cess in practice and those areas of research that have shown potential for security-
related applications. Many remote security sensors can be grouped roughly into
the two categories of concealed object detection and human presence detection and
classification.
10 Microwave and Millimeter-Wave Remote Sensing for Security Applications

1.4.1  Active Imaging for Contraband Detection


Detection of objects hidden beneath clothing can be achieved using radar imaging
systems [2–11]. There are two approaches: measuring the reflectivity of the object
in each pixel of the image, and measuring the range to the object in each pixel.
Measurement of the reflectivity gives insight into the material composing the vari-
ous objects in the image. Metal objects are highly reflective, and the signal scattered
off metals will have high amplitude. Other materials are less reflective, such as
human tissue, and the scattered signals will be of lower amplitude. By considering
the contrast between pixels in the image, hidden objects can be detected; however,
objects with reflectivity similar to that of human tissue can be difficult to detect.
Another approach is to measure the range extent of objects. Radar imagers with
wide frequency bandwidths can resolve range extents short enough to detect objects
hidden under clothing by the thickness of the contraband; the resolution is inversely
proportional to the bandwidth. Images are thus formed by measuring the range
extent in each pixel; and objects hidden on the body can thus be detected if they are
thicker than the range resolution of the radar. Range and reflectivity measurements
can also be combined in an image. Active imagers are typically implemented at
­millimeter-wave frequencies in order to achieve fine resolution in angle and to sup-
port the bandwidths required for fine resolution in range.
Figure 1.3(a) shows a 350-GHz active imaging system with two reflectors, one
of which is rotated to steer the beam. The system had a 9.6-GHz bandwidth yield-
ing a longitudinal resolution of 1.5 cm and lateral resolution of 1 cm at ranges up to
10 m. Figure 1.3 shows images from the system of a concealed mock explosive on a
person taken at a range of 5 m [8]. A 72–80-GHz active imaging system is shown in
Figure 1.4. The antenna was a sparsely populated array, shown in Figure 1.4(a) and
consisting of a 4×4 array of antenna clusters, with each of the 16 clusters containing
46 transmit and 46 receive elements. Figure 1.4(b) shows images from the system of
a concealed handgun and pocketknife. The lateral resolution achieved was 2 mm,
while the 8-GHz bandwidth gives a longitudinal resolution of 1.87 cm [11].

1.4.2  Passive Imaging for Contraband Detection


Passive imaging techniques provide another useful solution for detecting concealed
objects [12–35]. Passive imagers detect intrinsic thermal radiation; different ma-
terials radiate with different efficiencies, and thus the detected power level of the
thermal radiation from one material will generally be different from that of another
material. Using this fact, concealed objects can be detected by the difference in radi-
ated power from the body. Metals and objects with low thermal radiation efficiency
reflect the majority of the incident radiation and will appear similar to the sur-
rounding environment. Passive imagers are generally operated at millimeter-wave
frequencies to achieve high resolution in angle.
Images taken from a 77-GHz video-rate passive imaging system with a frame
rate of 4 Hz are shown in Figures 1.5(a) and 1.5(b). This system consisted of a
linear array for azimuth scanning and achieved vertical scanning through a dual-
reflector design and a collimating lens, resulting in a lateral resolution of 20 mm,
and had a radiometric temperature resolution of 1.2 K [30]. Figure 1.6(a) shows the
1.4  Examples of Remote Security Sensors 11

Figure 1.3  (a) 350-GHz active imaging system with dual-reflector beam steering. (b) Image of a
concealed mock explosive beneath clothing. (© IEEE 2009 [8].)

antenna array of a video-rate interferometric passive radiometric imaging system


operating at 22 GHz, with a radiometric temperature sensitivity of 2 K and a frame
rate of 25 Hz [33]. The array elements, one of which is shown in Figures 1.6(b) and
1.6(c), are loop-fed waveguides. The radiometric image, shown in Figure 1.6(d),
is formed by cross-correlating the outputs of pairs of antennas and shows a metal
object concealed under a jacket. Figure 1.7(a) shows a 90-GHz passive imaging
system using a vertical mechanical track and a Cassegrain antenna system where the
subreflector rotates to scan the beam to form the image [35]. The spatial resolution
of this system was 2.5 cm at a range of 2.4 m, and the radiometric resolution was
between 0.06 K and 0.3 K. Images of concealed objects are shown in Figure 1.7(b)
and (c), where hidden metal objects are clearly visible.
12 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Figure 1.4  (a) 72–80-GHz active imaging system with a sparse array. The transmit elements are
placed along the top and bottom edges of each square cluster, with the receive elements placed
along the sides. (b) and (c) are images of objects concealed beneath clothing. (© IEEE 2011 [11].)

1.4.3  Detection of Human Presence


Detecting human presence is important in border security, military force protec-
tion, site security monitoring, search and rescue, and other related applications.
Detection of humans in indoor or outdoor environments can be accomplished us-
ing either radar or radiometer systems [36–46]. Radar systems can detect moving
1.4  Examples of Remote Security Sensors 13

Figure 1.5  Images of concealed objects taken from a 77-GHz video-rate passive radiometric imag-
ing system. (© 2010 SPIE [30].)

Figure 1.6  A 22-GHz video-rate interferometric passive radiometric imaging system: (a) Interfero-
metric imaging array. (b) Short waveguide antenna element showing the top of the loop feed and
(c) bottom of the loop feed. (d) Video image taken of a person with a foil square concealed under a
jacket. The foil presents as a dark spot in the signature, caused by reflecting the colder background
temperature. (© 2011 SPIE [33], images courtesy of N. Salmon.)
14 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Figure 1.7  (a) A 90-GHz passive radiometric imager using a Cassegrain antenna system with a
rotating subreflector and a vertical mechanical track. (b) and (c) Images showing concealed metal
objects. (© 2008 SPIE [35].)

humans from long distances by measuring the Doppler shift in the return signal,
whereas radiometers detect the thermal radiation emitted by the human body. Fig-
ure 1.8(a) shows a 36-GHz continuous-wave scanning-beam Doppler radar de-
signed for the detection of moving humans from a moving platform; the system
block diagram is given in Figure 1.8(b), and Figure 1.8(c) shows a plot in the time-
frequency domain of the radar return signal of three walking people. The rotation
rate of the sensor was approximately 1 rad·s–1. The continuous oscillatory line is the
return signal from the stationary environment, an artifact of rotation of the sensor,
while the point responses are the Doppler-shifted returns from the people walking
around the sensor. The background signal is due to the movement of the platform
and the rotation rate of the pedestal; it is deterministic, and filtering it out results
in the isolation of the responses from the people only. The antenna beamwidth was
1.4  Examples of Remote Security Sensors 15

Figure 1.8  (a) A 36-GHz continuous-wave Doppler radar for detecting moving people from a mov-
ing platform. (© IEEE 2009 [41].) (b) System diagram. (c) Signal response from three walking people
as the sensors rotates atop a moving platform. The point responses are due to walking people, while
the oscillation is due to the stationary environment. Filtering the environmental response isolates the
responses from the moving humans.
16 Microwave and Millimeter-Wave Remote Sensing for Security Applications

3.5º, and the system had a Doppler frequency resolution of 16 Hz. Figure 1.9(a)
shows a 27.4-GHz radiometer for detecting people from a moving platform. The
system consisted of both total power and correlation detection modes on a rota-
tor with a 1 rad·s–1 rotation rate. The total power radiometers had radiometric
temperature sensitivities of 0.5 K, while the correlation mode had a sensitivity of
0.27 K. F­igure 1.9(b) shows the signal response of a total power radiometer and a
correlation radiometer scanning across a stationary human and a pillar constructed
of metal and concrete [39]. The large peaks are due to the thermal radiation emit-
ted by the human. It can be seen that the correlation mode responds to the human,
but not the pillar; this is due to the lower coherence of the radiation from the

(a)

0.6 0.6
human
0.4 pillar
Amplitude

0.4
0.2 0.2 Total power
0 0 response
−0.2 −0.2
70 75 80 85 145 150 155 160 165

1 1
Amplitude

0 0 Correlation
response
−1 −1
70 75 80 85 145 150 155 160 165
θ θ
(b)
Figure 1.9  (a) A 27.4-GHz radiometer system atop a mobile robotic platform. The sensor included
both total power and correlation detection modes. (b) Signal response of a total power radiometer
and a correlation radiometer in an outdoor environment scanning across a stationary human and a
pillar constructed of concrete and metal. (© IEEE 2008 [39].)
1.4  Examples of Remote Security Sensors

Figure 1.10  (a) Measured and (b) simulated micro-Doppler signature of a walking human from a 2.4-GHz radar. (c) Simulated micro-Doppler signature of a walk-
17

ing human with a 12-GHz radar. (© IEEE 2008 [53], courtesy of H. Ling.)
18 Microwave and Millimeter-Wave Remote Sensing for Security Applications

pillar, which is largely reflected, r­elative to that of the human, which is intrinsically


generated and has higher spatial coherence.

1.4.4  Discrimination of Humans and Classification of Human Activity


Upon detection, it is important to determine that a nonhuman, such as an animal,
has not been inaccurately determined to be a human. Furthermore, it is desirable
to classify the type of activity the person is engaged in to determine whether the
person represents a threat. Of the various signatures that may be analyzed for these
purposes, the micro-Doppler signature present in the radar returns of humans and
animals is a prominent feature, and it has been the focus of target discrimination
and activity classification using microwave and millimeter-wave Doppler radar sys-
tems [47–62]. The micro-Doppler signature arises from the dynamic motion of the
various parts of the body, such as the different velocities of the torso, arms, and legs.
The arms and legs also move in a periodic fashion that lends itself to classification.
Through micro-Doppler signal classification, humans can be discriminated from
vehicles and animals, and various human activities can be classified. Figures 1.10(a)
and 1.10(b) show measured and simulated micro-Doppler signatures of a walking
human in the time-frequency domain, from a 2.4-GHz radar [53]. The large oscil-
lations are due to the motion of the legs, and the dominant signal with little oscil-
lation variability at the center of the signature is due to the torso. Figure 1.10(c)
shows a simulated micro-Doppler signature of a walking human from a 12-GHz
radar that has increased frequency resolution. The frequency responses of the foot,
leg, lower arm, and torso are more visible [53].

2.5
Reflecting plates

2
range (m)

1.5
Wall

0.5
0 0.5 1 1.5 2 2.5 3
cross−range (m)
Figure 1.11  Ultra-wideband through-wall radar image of two reflecting plates. The bright spots in
the wall are metal pylons. (© IEEE 2009 [74].)
1.4  Examples of Remote Security Sensors 19

1.4.5  Through-Wall Detection


Detection of humans and objects through walls is difficult due to attenuation of
electromagnetic waves in wall material, but it can be accomplished using either ra-
dar or radiometer systems [60, 63–78]. Losses due to most wall materials increase

Figure 1.12  (a) Block diagram of a Ka-band continuous-wave radar for biological signal measurements.
(b) Heartbeat and respiration detection in the frequency domain, plotted as a function of heartbeat rate,
taken from the front and back of the body from a distance of 1.5 m. (© IEEE 2006 [79].)
20 Microwave and Millimeter-Wave Remote Sensing for Security Applications

with frequency, and thus most through-wall sensors are operated at low microwave
frequencies, often 10 GHz or less. In order to achieve reasonable angular resolu-
tion, the required aperture sizes are thus physically large, which generally rules out
high-­resolution imaging techniques. However, the bulk movement of people can be
detected, and micro-Doppler signatures can also be measured. Figure 1.11 shows
a measured cross-range and down-range plot of a 5-GHz ultra-wideband through-
wall radar with 6 GHz of bandwidth, showing the wall material and two reflecting
plates behind the wall [74]. The longitudinal resolution was 2.5 cm. Images with
this system were formed by placing the antennas on a mobile robotic platform that
moved down the length of the wall, generating a synthetic aperture.

1.4.6  Biological Signature Detection


Radar systems with high frequency resolution can be used to detect and measure
minute biological movements of the human body, including movement of the torso
due to respiration and heartbeat [79–85]. The Doppler frequency shift is propor-
tional to the carrier frequency and the velocity of the object; therefore, sensing of
this type requires high carrier frequencies to measure the relatively small frequency
shifts. Figure 1.12(a) shows the diagram of a Ka-band continuous-wave radar for
the detection of respiration and heartbeat. Figure 12(b) shows a plot of the mea-
sured frequency response of a human on the front and back of the torso from a
Ka-band radar [79]. The axis shows the Doppler frequency as the heartbeat rate in
beats per minute.

References

  [1] “IEEE Standard Letter Designations for Radar-Frequency Bands,” IEEE Std 521-2002
(Revision of IEEE Std 521-1984), 2003, pp. 1–3.
  [2] Anderton, R. N., R. Appleby, P. R. Coward, P. J. Kent, S. Price, et al., “Security Scanning
at 35 GHz,” Proceedings of the SPIE, Vol. 4373, 2001, pp. 16–23.
  [3] Sheen, D. M., D. L. McMakin, W. M. Lechelt, and J. W. Griffin, “Circularly Polarized
Millimeter-Wave Imaging for Personnel Screening,” Proceedings of the SPIE, Orlando, FL,
2005, pp. 117–126.
  [4] Dallinger, A., S. Schelkshorn, and J. Detlefsen, “Short Distance Related Security Millimeter-
Wave Imaging Systems,” in German Microwave Conference, Ulm, 2005.
  [5] McMillan, R. W., “Terahertz Imaging, Millimeter-Wave Radar,” in Advances in Sensing
with Security Applications, J. Burnes, Ed., Dordrecht, the Netherlands: Springer, 2006.
  [6] Appleby, R., and R. N. Anderton, “Millimeter-Wave and Submillimeter-Wave Imaging for
Security and Surveillance,” Proceedings of the IEEE, Vol. 95, 2007, pp. 1683–1690.
  [7] Mizuno, K., Y. Wagatsuma, H. Warashina, K. Sawaya, H. Sato, et al., “Millimeter-Wave
Imaging Technologies and Their Applications,” Vacuum Electronics Conference, 2007,
IVEC ‘07, IEEE International, 2007, pp. 1–2.
  [8] Sheen, D. M., D. L. McMakin, T. E. Hall, and R. H. Severtsen, “Active Millimeter-Wave
Standoff and Portal Imaging Techniques for Personnel Screening,” in Technologies for
Homeland Security, 2009, HST ‘09, IEEE Conference on, 2009, pp. 440–447.
  [9] Goshi, D. S., Y. Liu, K. Mai, L. Bui, and Y. Shih, “Cable Imaging with an Active W-Band
Millimeter-Wave Sensor,” Microwave Symposium Digest (MTT), 2010 IEEE MTT-S Inter-
national, 2010, pp. 1620–1623.
1.4  Examples of Remote Security Sensors 21

[10] Abril, J., E. Nova, A. Broquetas, F. Torres, J. Romeu, et al., “Combined Passive and Ac-
tive Millimeter-Wave Imaging System for Concealed Objects Detection,” Infrared Millime-
ter and Terahertz Waves (IRMMW-THz), 2010 35th International Conference on, 2010,
pp. 1–2.
[11] Ahmed, S. S., A. Schiessl, and L. P. Schmidt, “A Novel Fully Electronic Active Real-Time
Imager Based on a Planar Multistatic Sparse Array,” Microwave Theory and Techniques,
IEEE Transactions on, Vol. 59, 2011, pp. 3567–3576.
[12] Blanchard, P. M., A. H. Greenaway, A. R. Harvey, and K. Webster, “Coherent Optical
Beam Forming with Passive Millimeter-Wave Arrays,” Lightwave Technology, Journal of,
Vol. 17, 1999, pp. 418–425.
[13] Lettington, A. H., M. R. Yallop, and D. Dunn, “Review of Super-Resolution Techniques
for Passive Millimeter-Wave Imaging,” Proceedings of the SPIE, Orlando, FL, 2002,
pp. 230–239.
[14] Yujiri, L., M. Shoucri, and P. Moffa, “Passive Millimeter Wave Imaging,” Microwave
Magazine, IEEE, Vol. 4, 2003, pp. 39–50.
[15] Martin, C., “Passive Millimeter-Wave Imaging for the Detection of Concealed Weapons,”
Air Force Research Laboratory Technical Report AFRL-IF-RS-TR-2005-37, 2005.
[16] Luthi, T., and C. Matzler, “Stereoscopic Passive Millimeter-Wave Imaging and Ranging,”
IEEE Transactions on Microwave Theory and Techniques, Vol. 53, 2005, pp. 2594–2599.
[17] Williams, T. D., and N. M. Vaidya, “A Compact, Low-Cost, Passive MMW Security Scan-
ner,” Proceedings of the SPIE, Orlando, FL, 2005, pp. 109–116.
[18] Nohmi, H., S. Ohnishi, and O. Kujubu, “Passive Millimeter-Wave Camera with Interfero-
metric Processing,” Proceedings of the SPIE, Vol. 6211, 2006, pp. 621104–621108.
[19] Lovberg, J. A., C. Martin, and V. Kolinko, “Video-Rate Passive Millimeter-Wave Imaging
Using Phased Arrays,” Microwave Symposium, 2007, IEEE/MTT-S International, 2007,
pp. 1689–1692.
[20] Yue, L., J. W. Archer, G. Rosolen, S. G. Hay, G. P. Timms, et al., “Fringe Management for
a T-Shaped Millimeter-Wave Imaging System,” Microwave Theory and Techniques, IEEE
Transactions on, Vol. 55, 2007, pp. 1246–1254.
[21] Nohmi, H., S. Ohnishi, and O. Kujubu, “Passive Millimeter-Wave Camera with Interfero-
metric Processing,” Proceedings of the SPIE, Vol. 6548, 2007, p. 65480C-8.
[22] Chen, C., C. A. Schuetz, R. D. Martin, J. Samluk, J. E. Lee Stein, et al., “Analytical Model
and Optical Design of Distributed Aperture Optical System for Millimeter-Wave Imaging,”
Proceedings of the SPIE, Vol. 7117, 2008, p. 711706.
[23] Dillon, T. E., C. A. Schuetz, R. D. Martin, J. E. Lee Stein, J. P. Samluk, et al., “Optical
Configuration of an Upconverted Millimeter-Wave Distributed Aperture Imaging System,”
Proceedings of the SPIE, Vol. 7485, 2009, p. 74850G.
[24] Persons, C. M., C. A. Martin, M. W. Jones, V. Kolinko, and J. A. Lovberg, “Passive Millimeter-
Wave Imaging Polarimeter System,” Proceedings of the SPIE, Vol. 7309, 2009, p. 730907.
[25] Stein, E. L., C. A. Schuetz, R. D. Martin, J. P. Samluk, J. P. Wilson, et al., “Passive Millimeter-
Wave Cross Polarization Imaging and Phenomenology,” Proceedings of the SPIE, Vol.
7309, 2009, p. 730902.
[26] Martin, R., C. A. Schuetz, T. E. Dillon, C. Chen, J. Samluk, et al., “Design and Performance
of a Distributed Aperture Millimeter-Wave Imaging System Using Optical Upconversion,”
Proceedings of the SPIE, Vol. 7309, 2009, p. 730908.
[27] Wikner, D., and E. Grossman, “Demonstration of a Passive, Low-Noise, Millimeter-Wave
Detector Array for Imaging,” Proceedings of the SPIE, Vol. 7309, 2009, p. 730909.
[28] Yue, L., J. W. Archer, J. Tello, G. Rosolen, F. Ceccato, et al., “Performance Evaluation of a
Passive Millimeter-Wave Imager,” Microwave Theory and Techniques, IEEE Transactions
on, Vol. 57, 2009, pp. 2391–2405.
[29] Ghasr, M. T., D. Pommerenke, J. T. Case, A. McClanahan, A. Aflaki-Beni, et al., “Rapid
Rotary Scanner and Portable Coherent Wideband Q-Band Transceiver for High-­Resolution
22 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Millimeter-Wave Imaging Applications,” Instrumentation and Measurement, IEEE Trans-


actions on, Vol. 60, 2011, pp. 186–197.
[30] Sato, H., K. Sawaya, K. Mizuno, J. Uemura, M. Takeda, et al., “Passive Millimeter-Wave
Imaging for Security and Safety Applications,” in Proceedings of the SPIE, Orlando, FL,
2010, p. 76710V-11.
[31] Sato, H., K. Sawaya, K. Mizuno, J. Uemura, M. Takeda, et al., “Development of 77 GHz
Millimeter Wave Passive Imaging Camera,” Sensors, 2009 IEEE, 2009, pp. 1632–1635.
[32] Hua-Mei, C., L. Seungsin, R. M. Rao, M. A. Slamani, and P. K. Varshney, “Imaging for
Concealed Weapon Detection: A Tutorial Overview of Development in Imaging Sensors
and Processing,” Signal Processing Magazine, IEEE, Vol. 22, 2005, pp. 52–61.
[33] Salmon, N. A., R. Macpherson, A. Harvey, P. Hall, S. Hayward, et al., “First Video Rate
Imagery from a 32-Channel 22-GHz Aperture Synthesis Passive Millimetre Wave Imager,”
Proceedings of the SPIE, Prague, Czech Republic, 2011, pp. 818, 805–818, 812.
[34] Salmon, N. A., I. Mason, P. Wilkinson, C. Taylor, and P. Scicluna, “First Imagery Gener-
ated by Near-Field Real-Time Aperture Synthesis Passive Millimetre Wave Imagers at 94
GHz and 183 GHz,” Proceedings of the SPIE, Vol. 7837, 2010, p. 78370I.
[35] Dill, S., M. Peichl, and H. Suss, “Study of Passive MMW Personnel Imaging with Respect
to Suspicious and Common Concealed Objects for Security Applications,” Proceedings of
the SPIE, Vol. 7117, 2008, p. 71170C.
[36] Brenda, A. K., “Selected Systems for the Detection of Human Stowaways in Air Cargo
Containers,” Carnahan Conferences Security Technology, Proceedings 2006 40th Annual
IEEE International, 2006, pp. 26–29.
[37] Nanzer, J. A., and R. L. Rogers, “Human Presence Detection Using Millimeter-Wave
­Radiometry,” Microwave Theory and Techniques, IEEE Transactions on, Vol. 55, 2007,
pp. 2727–2733.
[38] Nanzer, J. A., and R. L. Rogers, “A Ka-Band Correlation Radiometer for Human Presence
Detection from a Moving Platform,” Microwave Symposium, 2007, IEEE/MTT-S Interna-
tional, 2007, pp. 385–388.
[39] Nanzer, J. A., and R. L. Rogers, “Applying Millimeter-Wave Correlation Radiometry to
the Detection of Self-Luminous Objects at Close Range,” Microwave Theory and Tech-
niques, IEEE Transactions on, Vol. 56, 2008, pp. 2054–2061.
[40] Nanzer, J. A., and R. L. Rogers, “A Simple Model of Human Thermal Millimeter-Wave Ra-
diation for Radiometric Human Presence Detection,” 2009 USNC/URSI National Radio
Science Meeting, 2009.
[41] Nanzer, J. A., M. G. Anderson, T. M. Josserand, L. Kuan, G. A. Olinger, et al., “Detection
of Moving Intruders from a Moving Platform Using a Ka-Band Continuous-Wave Doppler
Radar,” Antennas and Propagation Society International Symposium, 2009, APSURSI ‘09,
IEEE, 2009, pp. 1–4.
[42] SangHyun, C., N. Mitsumoto, and J. W. Burdick, “An Algorithm for UWB Radar-Based
Human Detection,” Radar Conference, 2009 IEEE, 2009, pp. 1–6.
[43] Nanzer, J. A., E. Popova, and R. L. Rogers, “Applying Correlation Radiometry to Human
Presence Detection in Multiple Outdoor Environments,” 2010 CNC/USNC/URSI National
Radio Science Meeting, 2010.
[44] Nanzer, J. A., E. Popova, and R. L. Rogers, “Analysis of the Detection Modes of a Human
Presence Detection Millimeter-Wave Radiometer,” Antennas and Propagation Society In-
ternational Symposium (APSURSI), 2010 IEEE, 2010, pp. 1–4.
[45] SangHyun, C., M. Wolf, and J. W. Burdick, “Human Detection and Tracking via Ultra-
Wideband (UWB) Radar,” Robotics and Automation (ICRA), 2010 IEEE International
Conference on, 2010, pp. 452–457.
[46] Nanzer, J. A., and R. L. Rogers, “Frequency Estimation of Human Presence Detection
Signals From a Scanning-Beam Millimeter-Wave Correlation Radiometer,” Geoscience and
Remote Sensing Letters, IEEE, Vol. 8, 2011, pp. 78–82.
1.4  Examples of Remote Security Sensors 23

[47] van Dorp, P., and F. C. A. Groen, “Human Walking Estimation with Radar,” Radar, Sonar
and Navigation, IEE Proceedings, Vol. 150, 2003, pp. 356–365.
[48] Thayaparan, T., S. Abrol, E. Riseborough, L. Stankovic, D. Lamothe, et al., “Analysis of
Radar Micro-Doppler Signatures from Experimental Helicopter and Human Data,” Radar,
Sonar & Navigation, IET, Vol. 1, 2007, pp. 289–299.
[49] Chen, V. C., “Detection and Analysis of Human Motion by Radar,” Radar Conference,
2008, RADAR ‘08. IEEE, 2008, pp. 1–4.
[50] Smith, G. E., K. Woodbridge, and C. J. Baker, “Multistatic Micro-Doppler Signature of
Personnel,” Radar Conference, 2008, RADAR ‘08, IEEE, 2008, pp. 1–6.
[51] Anderson, M. G., “Design of Multiple Frequency Continuous Wave Radar Hardware and
Micro-Doppler Based Detection and Classification Algorithms,” Ph.D. Thesis, University
of Texas at Austin, 2008.
[52] Youngwook, K., and L. Hao, “Human Activity Classification Based on Micro-Doppler
Signatures Using an Artificial Neural Network,” Antennas and Propagation Society Inter-
national Symposium, 2008, AP-S 2008, IEEE, 2008, pp. 1–4.
[53] Sundar Ram, S., and L. Hao, “Simulation of Human MicroDopplers Using Computer
Animation Data,” Radar Conference, 2008, RADAR ‘08, IEEE, 2008, pp. 1–6.
[54] Zhaonian, Z., and A. G. Andreou, “Human Identification Experiments Using Acoustic
Micro-Doppler Signatures,” Micro-Nanoelectronics, Technology and Applications, 2008,
EAMTA 2008, Argentine School of, 2008, pp. 81–86.
[55] Nanzer, J. A., and R. L. Rogers, “Bayesian Classification of Humans and Vehicles Using
Micro-Doppler Signals from a Scanning-Beam Radar,” Microwave and Wireless Compo-
nents Letters, IEEE, Vol. 19, 2009, pp. 338–340.
[56] Youngwook, K., and L. Hao, “Human Activity Classification Based on Micro-Doppler Sig-
natures Using a Support Vector Machine,” Geoscience and Remote Sensing, IEEE Transac-
tions on, Vol. 47, 2009, pp. 1328–1337.
[57] Tahmoush, D., and J. Silvious, “Angle, Elevation, PRF, and Illumination in Radar Micro­
Doppler for Security Applications,” Antennas and Propagation Society International Sym-
posium, 2009, APSURSI ‘09, IEEE, 2009, pp. 1–4.
[58] Vignaud, L., A. Ghaleb, J. Le Kernec, and J. M. Nicolas, “Radar High Resolution Range &
Micro-Doppler Analysis of Human Motions,” Radar Conference—Surveillance for a Safer
World, 2009, RADAR. International, 2009, pp. 1–6.
[59] Silvious, J., J. Clark, T. Pizzillo, and D. Tahmoush, “Micro-Doppler Phenomenology of
Humans at UHF and Ku-Band for Biometric Characterization,” Proceedings of the SPIE,
Orlando, FL, 2009, pp. 73080X-9.
[60] Ram, S. S., C. Christianson, Y. Kim, and H. Ling, “Simulation and Analysis of Human
Micro-Dopplers in Through-Wall Environments,” Geoscience and Remote Sensing, IEEE
Transactions on, Vol. 48, 2010, pp. 2015–2023.
[61] Moulton, M. C., M. L. Bischoff, C. Benton, and D. T. Petkie, “Micro-Doppler Radar Sig-
natures of Human Activity,” in Proceedings of the SPIE, 2010, p. 78370L.
[62] Chen, V. C., The Micro-Doppler Effect in Radar, Norwood, MA: Artech House, 2011.
[63] Frazier, L., “Surveillance Through Walls and Other Opaque Materials,” Proc. SPIE, Vol.
2497, 1995, p. 115.
[64] Venkatasubramanian, V., and H. Leung, “A Novel Chaos-Based High-Resolution Imaging
Technique and Its Application to Through-the-Wall Imaging,” Signal Processing Letters,
IEEE, Vol. 12, 2005, pp. 528–531.
[65] Ram, S. S., and H. Ling, “Through-Wall Tracking of Human Movers Using Joint Dop-
pler and Array Processing,” Geoscience and Remote Sensing Letters, IEEE, Vol. 5, 2008,
pp. 537–541.
[66] Nilsson, S., A. Janis, M. Gustafsson, J. Kjellgren, and A. Sume, “Through-the-Wall High-
Resolution Imaging of a Human and Experimental Characterization of the Transmission
of Wall Materials,” Proceedings of the SPIE, Vol. 7117, 2008, p. 71170L.
24 Microwave and Millimeter-Wave Remote Sensing for Security Applications

[67] Chi-Wei, W., and H. Zi-Yu, “Using the Phase Change of a Reflected Microwave to Detect a
Human Subject Behind a Barrier,” Biomedical Engineering, IEEE Transactions on, Vol. 55,
2008, pp. 267–272.
[68] Narayanan, R. M., “Through-Wall Radar Imaging Using UWB Noise Waveforms,” Journal
of the Franklin Institute, Vol. 345, 2008, pp. 659–678.
[69] Dehmollaian, M., and K. Sarabandi, “Refocusing Through Building Walls Using Synthetic
Aperture Radar,” Geoscience and Remote Sensing, IEEE Transactions on, Vol. 46, 2008,
pp. 1589–1599.
[70] Johnson, J. T., M. A. Demir, and N. Majurec, “Through-Wall Sensing with Multifrequency
Microwave Radiometry: A Proof-of-Concept Demonstration,” IEEE Transactions on Geo-
science and Remote Sensing, Vol. 47, 2009, pp. 1281–1288.
[71] Gonzalez-Partida, J. T., P. Almorox-Gonzalez, M. Burgos-Garcia, B. P. Dorta-Naranjo,
and J. I. Alonso, “Through-the-Wall Surveillance with Millimeter-Wave LFMCW Radars,”
IEEE Transactions on Geoscience and Remote Sensing, Vol. 47, 2009, pp. 1796–1805.
[72] Venkatasubramanian, V., H. Leung, and L. Xiaoxiang, “Chaos UWB Radar for Through-
the-Wall Imaging,” IEEE Transactions on Image Processing, Vol. 18, 2009, pp. 1255–
1265.
[73] Hong, W., R. M. Narayanan, and Z. Zheng Ou, “Through-Wall Imaging of Moving Tar-
gets Using UWB Random Noise Radar,” Antennas and Wireless Propagation Letters,
IEEE, Vol. 8, 2009, pp. 802–805.
[74] Braga, A. J., and C. Gentile, “An Ultra-Wideband Radar System for Through-the-Wall
­Imaging Using a Mobile Robot,” IEEE International Conference on Communications,
2009, ICC ‘09, 2009, pp. 1–6.
[75] Solimene, R., F. Soldovieri, G. Prisco, and R. Pierri, “Three-Dimensional Through-Wall Im-
aging Under Ambiguous Wall Parameters,” IEEE Transactions on Geoscience and Remote
Sensing, Vol. 47, 2009, pp. 1310–1317.
[76] Lianlin, L., Z. Wenji, and L. Fang, “A Novel Autofocusing Approach for Real-Time
Through-Wall Imaging Under Unknown Wall Characteristics,” IEEE Transactions on Geo-
science and Remote Sensing, Vol. 48, 2010, pp. 423–431.
[77] Chieh-Ping, L., and R. M. Narayanan, “Ultrawideband Random Noise Radar Design for
Through-Wall Surveillance,” IEEE Transactions on Aerospace and Electronic Systems,
Vol. 46, 2010, pp. 1716–1730.
[78] Qiong, H., Q. Lele, W. Bingheng, and F. Guangyou, “UWB Through-Wall Imaging Based
on Compressive Sensing,” IEEE Transactions on Geoscience and Remote Sensing, Vol. 48,
2010, pp. 1408–1415.
[79] Yanming, X., L. Jenshan, O. Boric-Lubecke, and V. M. Lubecke, “A Ka-Band Low Power
Doppler Radar System for Remote Detection of Cardiopulmonary Motion,” 27th Annual
International Conference of the Engineering in Medicine and Biology Society, 2005, IEEE-
EMBS 2005, 2005, pp. 7151–7154.
[80] Yanming, X., L. Changzhi, and L. Jenshan, “Accuracy of a Low-Power Ka-Band Non-
Contact Heartbeat Detector Measured from Four Sides of a Human Body,” Microwave
Symposium Digest, 2006, IEEE MTT-S International, 2006, pp. 1576–1579.
[81] Jenshan, L., and L. Changzhi, “Wireless Non-Contact Detection of Heartbeat and Respira-
tion Using Low-Power Microwave Radar Sensor,” Microwave Conference, 2007, APMC
2007, Asia-Pacific, 2007, pp. 1–4.
[82] Petkie, D. T., E. Bryan, C. Benton, C. Phelps, J. Yoakum, et al., “Remote Respiration and
Heart Rate Monitoring with Millimeter-Wave/Terahertz Radars,” Proceedings of the SPIE,
2008, pp. 71170I.
[83] Droitcour, A. D., O. Boric-Lubecke, and G. T. A. Kovacs, “Signal-to-Noise Ratio in ­Doppler
Radar System for Heart and Respiratory Rate Measurements,” IEEE Transactions on Mi-
crowave Theory and Techniques, Vol. 57, 2009, pp. 2498–2507.
1.4  Examples of Remote Security Sensors 25

[84] Changzhi, L., J. Cummings, J. Lam, E. Graves, and W. Wenhsing, “Radar Remote Monitor-
ing of Vital Signs,” Microwave Magazine, IEEE, Vol. 10, 2009, pp. 47–56.
[85] Massagram, W., V. M. Lubecke, A. Host-Madsen, and O. Boric-Lubecke, “Assessment
of Heart Rate Variability and Respiratory Sinus Arrhythmia via Doppler Radar,” IEEE
Transactions on Microwave Theory and Techniques, Vol. 57, 2009, pp. 2542–2549.
Chapter 2

Electromagnetic Plane Wave


Fundamentals

Remote sensing, whether for security sensing, satellite sensing, or radio astronomy, in-
volves the transport of energy between a sensor and a remote location or object for the
purpose of determining some information about the object. When the sensing involves
microwave or millimeter-wave electromagnetic radiation, the energy is transported via
time-varying electromagnetic fields. Remote sensing of an object is generally facilitated
using one of two methods: measuring the intrinsic signals emanating from the remote
location or transmitting a predetermined signal and measuring the signal that reflects
back from the remote location. As will be shown, these signals take the form of electro-
magnetic waves. If the distance between the sensor and the remote object is sufficiently
large compared to the wavelength of the radiation, the electromagnetic waves can be
considered to be planar in that they change only in time and along the direction of
propagation; they remain constant in amplitude and phase in a plane orthogonal to
the direction of propagation. Such waves are called plane waves, and although they
are a theoretical construct in that by definition they extend infinitely in the directions
perpendicular to the direction of propagation, they are approximated well in practice
by a wave that has travelled a distance from the source that is large relative to the
wavelength and the size of the aperture. For example, a spherical wavefront emanating
from a point radiator becomes approximately planar at great distances, as illustrated in
Figure 2.1, and a receiver far from the source measures a wavefront in the z direction
that is very nearly constant in the x and y directions. This specific range is called the far
field, and is discussed further in Chapter 4. In practice, the far field is near enough to
the sensor for microwave and millimeter-wave frequencies that many remote security
sensors can be analyzed using the plane wave approximation. The understanding of the
characteristics of plane waves is therefore an important foundation for the understand-
ing of microwave and millimeter-wave remote sensors and their application to security
sensing.
In this chapter, the equations describing plane electromagnetic waves are de-
rived from Maxwell’s equations. General aspects are covered which describe con-
cepts of plane waves and their characteristics; these are important throughout the
rest of the book. More involved treatments of plane waves can be found in most
textbooks on electromagnetic waves [1–4].

2.1  Maxwell’s Equations

Electromagnetic phenomena are in general described in terms of six basic physical


quantities:

27
28 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Planar wavefronts
Spherical wavefronts

. . .

Radiating
point source

Figure 2.1  A spherical wavefront emanating from a point radiator is approximately planar after
traveling a sufficiently long distance.

· E (V·m–1), the electric field;


· H (A·m–1), the magnetic field;
· D (C·m–2), the electric flux density;
· B (Wb·m–2), the magnetic flux density;
· J (A·m–2), the electric current density;
· ρ (C·m–3), the electric charge density.

The first four terms are electromagnetic field quantities, while the current and
charge densities are source quantities. Electromagnetic fields are generated by cur-
rent and charge, and hence the latter are referred to as current and charge sources;
however, the fields are also characterized in the absence of sources, where the fields
are assumed to have been generated by sources outside the region of interest. This
is the case when considering the propagation of electromagnetic waves in free
space. Each quantity is defined in terms of its spatial and temporal dependencies;
that is,

χ = χ (r, t) (2.1)

where χ represents one of the vector or scalar quantities, r is the three-dimensional


position vector, and t is the time variable. For simplicity, the dependence on r and t
will generally be implicitly assumed through the rest of the book and will be explic-
itly introduced when necessary.
The interactions of the field and source quantities are defined by the four Max-
well equations:

¶B
Ñ´E = - (2.2)
¶t

¶D
Ñ´H = + J (2.3)
¶t
2.1  Maxwell’s Equations 29

Ñ × D = ρ (2.4)

Ñ × B = 0 (2.5)
where (2.2) is Faraday’s law, (2.3) is the Ampere-Maxwell law, or simply Ampere’s
law, (2.4) is Gauss’s law, and (2.5) is the magnetic Gauss’s law, or simply the ab-
sence of magnetic point sources. Taken together, (2.2)–(2.5) are called the Maxwell
equations or Maxwell’s equations.
The source quantities are interrelated through

¶ρ
Ñ× J+ =0 (2.6)
¶t
which is the equation of continuity. This equation is an expression of conservation
of charge; it states that the current density diverging from a closed surface must
be equal to the negative of the time rate of change of the charge density within the
surface. That is, electric charge is neither created nor destroyed.
Maxwell’s equations as described by (2.2)–(2.5) are referred to as the Minkowski
form of Maxwell’s equations. In this form, however, the equations appear to be
imbalanced; the electric field and electric flux density each have magnetic coun-
terparts; however, there are no magnetic complements to the electric current and
charge densities. To address this, the following quantities

· Jm (V·m–2), the magnetic current density,


· ρm (Wb·m–3), the magnetic charge density,

were introduced as purely theoretical constructs, since such quantities have not been
shown to exist in nature. The magnetic charge density would be a magnetic monopole;
however, only magnetic dipoles exist: if a magnetic dipole is split, the result is two smaller
magnetic dipoles. Despite their apparently fictitious nature, the magnetic source quantities
are often mathematically useful when describing certain problems. For instance, the mag-
netic current density is used in the discussion of aperture antennas in Chapter 4. Using the
magnetic current and charge densities results in:

¶B
Ñ´E = - - J m (2.7)
¶t

¶D
Ñ´H = + J (2.8)
¶t
Ñ × D = ρ (2.9)

Ñ × B = ρ m (2.10)

Discussions on other forms of Maxwell’s equations can be found in [5]. The analy-
ses that proceed throughout this book will generally use the Minkowski form of
Maxwell’s equations, with magnetic sources introduced where necessary. Deriva-
tions and analyses including the electric sources are easily extended to include
30 Microwave and Millimeter-Wave Remote Sensing for Security Applications

the magnetic sources due to the balanced nature of Maxwell’s equations; deriva-
tions generally proceed identically with the electric and magnetic field quantities
i­nterchanged.

2.1.1  The Constitutive Parameters


To describe the behavior and characteristics of electromagnetic phenomena within
media, some parameters describing relevant aspects of the media and how they
interact with the field quantities are required. In general, the electromagnetic flux
densities are related to the electromagnetic fields through

D = ε E (2.11)

1
H= B (2.12)
µ
where ε (F·m ) and μ (H·m ) are the permittivity and permeability of the medium, re-
–1 –1

spectively. In addition, the electric current density is related to the electric field through

J = σ E (2.13)
where σ (S·m–1) is the conductivity of the medium. Equations (2.11)–(2.13) are
called the constitutive relations and ε, μ, and σ are called the constitutive parame-
ters. In general, the constitutive parameters are complex vector quantities; however,
in linear, homogeneous, nondispersive media, they are simple constants.
In free space, the conductivity is zero and the permittivity and permeability are

ε 0 = 8.854 ´ 10-12 F × m -1 (2.14)

µ0 = 4π ´ 10-7 H × m -1 (2.15)

The permittivity and permeability of media other than free space are generally char-
acterized in terms of the free space values through

ε = ε r ε 0 (2.16)

µ = µ r µ0 (2.17)

where εr and μr are called the relative permittivity and relative permeability, respec-
tively. The permittivity of a medium is never less than that of free space; thus, the
relative permittivity is always greater or equal to one, with equality only in the case
of free space. Most media have relative permeability values very close to one; fer-
romagnetic and ferrimagnetic materials are exceptions and can have permeability
values much different than that of free space.
Relative conductivity values are often used to classify media; materials with
very high conductivity are conductors, while those with very low conductivity are
dielectrics or insulators. Good conductors, such as metals, are often approximated
by σ = ∞, whereas good dielectrics are often approximated by σ = 0; such media are
considered in Chapter 3.
2.2  Time-Harmonic Electromagnetic Fields 31

Use of the constitutive relations allow Maxwell’s equations to be cast in terms


of the electromagnetic field quantities E and H. Specifically,
¶H
Ñ ´ E = -µ (2.18)
¶t

¶E
Ñ´H = ε + σ E (2.19)
¶t
ρ
Ñ×E = (2.20)
ε

Ñ × H = 0 (2.21)

In this form, the characteristics of the medium are explicitly included through the
constitutive parameters.

2.2  Time-Harmonic Electromagnetic Fields

Maxwell’s equations as described in the previous section are fairly general in the
sense that no assumptions were made about the nature of the field quantities. For
remote security sensing, time varying fields are of primary interest—in particular,
sinusoidally time-varying or time-harmonic fields, as these are the types of fields
that propagate through a medium. In general, a time-harmonic field is given by

χ (r, t) = χ (r)e jωt (2.22)

where χ(r) describes the field in the spatial dimensions and ω = 2πf is the angular
frequency of the temporal oscillation of the field. Temporal derivatives of time-
harmonic quantities are straightforward; the time-derivative of (2.22) is


χ (r,t ) = jωχ (r)e jωt = jωχ (r, t) (2.23)
¶t

In terms of the representations described by (2.22), the time-harmonic electric


and magnetic fields can be given by

E(r, t) = E(r)e jωt (2.24)

H(r, t) = H(r)e jωt (2.25)

Using the property (2.23), the time derivatives in Faraday’s and Ampere’s laws
(2.18) and (2.19) can be simplified, resulting in

Ñ ´ E = - jωµH (2.26)

Ñ ´ H = jωε E + σ E (2.27)
32 Microwave and Millimeter-Wave Remote Sensing for Security Applications

2.2.1  The Wave Equation


Time-harmonic fields are governed by wave equations, and therefore time-harmonic
electromagnetic fields can be considered to be waves. In this section, the wave equa-
tion is derived from Maxwell’s equations; the plane wave solution to the wave
equation is considered in the following section. Consider an electromagnetic field
in free space, with no sources present; that is, J = 0 and ρ = 0 and the divergence of
both the electric and magnetic field quantities are zero. In the absence of sources,
Maxwell’s equations thus are given by

Ñ ´ E = - jωµH (2.28)

Ñ ´ H = jωε E (2.29)

Ñ × E = 0 (2.30)

Ñ × H = 0 (2.31)

Taking the curl of (2.28) results in

Ñ ´ (Ñ ´ E ) = - jωµÑ ´ H (2.32)

Substituting (2.29) for the right-hand side and using the vector identity

Ñ ´ Ñ ´ A = Ñ(Ñ × A) - Ñ 2 A (2.33)

yields

Ñ(Ñ × E) - Ñ 2 E = ω 2 µε E (2.34)

The first term on the left-hand side is zero due to (2.30), and thus

Ñ 2 E + ω 2µε E = 0 (2.35)

A similar derivation follows for Ampere’s law, which results in

Ñ2 H + ω 2µε H = 0 (2.36)

Defining the wavenumber k m–1 to be

k = ω µε (2.37)

equations (2.35) and (2.36) become

Ñ 2 E + k2 E = 0 (2.38)

Ñ 2 H + k2 H = 0 (2.39)
2.2  Time-Harmonic Electromagnetic Fields 33

Equations (2.38) and (2.39) are vector Helmholtz wave equations for E and H. If
the vector fields are separated into their rectangular coordinates, the result is six
scalar Helmhotz wave equations of the form
¶χ n
+ k2 χ n = 0 (2.40)
¶n
where χ = E, H, and n = x, y, z.

2.2.2  Plane Waves


The simplest solution to the scalar Helmholtz equation can be found by considering
a straightforward case where a given field consists of only one coordinate compo-
nent. Consider an electric field with a component only in the x direction given by

E(r, t) = xˆ Ex e jωt (2.41)

The Laplacian operator is then

¶ 2 Ex
Ñ2 E = (2.42)
¶z 2
and the scalar Helmholtz equation becomes

¶ 2 Ex
+ k2 Ex = 0 (2.43)
¶z 2

The solution of (2.43) is of the general form

Ex (z) = E1e - jkz + E2 e jkz (2.44)

which is the superposition of two waves, one traveling in the +z direction, the other
traveling the –z direction.
Consider a solution consisting of one wave traveling in the +z direction; that is,
E2 = 0. The electric field in such a case is

E(r, t) = E1e j(ωt -kz) (2.45)

The phase of this wave is constant over the surface ωt–kz; that is, the wave
extends infinitely in both the x and y dimensions and furthermore is constant in
both dimensions. The wave therefore changes only in the z direction and in time.
Such a wave defines a geometric plane and is thus termed a plane wave. Plane
waves are therefore electromagnetic waves that are constant in phase and ampli-
tude in plane orthogonal to the direction of propagation, and, as discussed in the
introduction to this chapter, a wave that has travelled a sufficient distance from
the source generating the wave can be considered planar. The analysis of plane
waves is generally simpler than that of nonplanar waves; thus, approximating a
wave as a plane wave eases the computational complexity of wave propagation
problems.
34 Microwave and Millimeter-Wave Remote Sensing for Security Applications

The angular frequency ω describes the dependence of the plane wave in the tem-
poral domain, and in a similar way the wavenumber describes the dependence of the
wave in the spatial domain along the direction of propagation. The angular frequency
can be defined in terms of the temporal oscillation of the field by the relation


ω= (2.46)
T

where T is the period of the temporal oscillation, or the time duration between
adjacent maxima (or minima) of the sinusoidal temporal oscillation. Similarly, the
wavenumber can be defined in terms of the distance between adjacent maxima of
the sinusoidal oscillation of the wave in the spatial domain by the relation

k= (2.47)
λ

where λ is the wavelength, or the spatial distance between maxima.


The relationship between wavenumber and wavelength can also be derived by
considering the distance between a point z0 and a point a distance Dz away in the z
direction given by z0 + Dz. Since the wavelength is the distance between maxima of
the spatial sinusoidal oscillation, setting Dz = λ,
(ωt - kz) - [ωt - k(z + λ )] = 2 π (2.48)

or
kλ = 2π (2.49)
which is the same as (2.47).

2.2.2.1  Phase Velocity


Electromagnetic waves propagate through a medium at a finite velocity. For a plane wave,
the phase is constant in the dimensions perpendicular to the direction of propagation, and
thus the velocity of the phase is equal to the velocity of propagation of the wave. The
phase velocity is found by the time rate of change of the phase in the spatial domain:

dz (2.50)
vp =
dt
By definition of a plane wave, the phase is constant:

ωt - kz = const (2.51)
or

ωt - const (2.52)
z=
k
and thus

d æ ωt - const ö ω (2.53)
vp = ç ÷=
dt è k ø k
2.2  Time-Harmonic Electromagnetic Fields 35

Using the definition of the wavenumber (2.37) results in

1 (2.54)
vp =
µε
The phase velocity is therefore determined solely by the constitutive parameters of
the medium through which the wave travels. In free space, the phase velocity is
1
c= = 2.9979 × 108 m ⋅ s−1 (2.55)
µ 0ε 0

which is the velocity of light.

2.2.2.2  Relationship Between E and H


Consider an electric field propagating in the z direction and consisting of only a
component in the x direction given by

E = xˆ Ex e j(ωt -kz) (2.56)

Using Faraday’s law (2.28), the magnetic field is given by

1 k ε
H=j Ñ ´ E = yˆ Ex e j(ωt -kz) = yˆ Ex e j(ωt- kz) (2.57)
ωµ ωµ µ

Thus, the resulting magnetic field consists of only a component in the y direction.
As discussed earlier, a plane wave at a fixed point z is constant in both the x and y
directions. Therefore,

¶E ¶E
= = 0 (2.58)
¶x ¶y

and

¶H ¶H
= = 0 (2.59)
¶x ¶y

From Ampere’s law, the z component of the electric field is given by

1 æ ¶H y ¶H x ö
Ez = ç - ÷ = 0 (2.60)
jω ε è ¶x ¶y ø

and similarly from Faraday’s law,

H z = 0 (2.61)

Therefore, the longitudinal components of the electric and magnetic fields are zero;
the fields of a plane wave consist of only components perpendicular to the direction
of propagation.
36 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Additionally, both the electric and magnetic fields are perpendicular to each
other. This can be seen by considering the electric field

E = xˆ Ex + yˆ Ey (2.62)

The magnetic field is then, from (2.57), given by

k k
H = -xˆ Ey + yˆ Ex (2.63)
ωµ ωµ

The vector product of the electric and magnetic fields is therefore

k k
E×H = - Ex Ey + E E = 0 (2.64)
ωµ ωµ x y

The electric and magnetic fields are thus perpendicular to one another and perpen-
dicular to the direction of propagation. Such a wave is referred to as a transverse
electromagnetic (TEM) wave.
The magnitude of the magnetic field as given by (2.57) is simply the magnitude
of the electric field scaled by ε µ . The relationship between the electric and mag-
netic field components can therefore be given by
1
H = E (2.65)
η
where
µ E
η= = (2.66)
ε H

is the intrinsic impedance of the medium, measured in W. Substituting the free-space


permittivity and permeability yields the intrinsic impedance of free space
µ0
η0 = » 377 W (2.67)
ε0
Up to this point, only simple fields where the electric or magnetic fields con-
sisted of one component have been considered; however, the analyses are easily
extended to fields consisting of components in multiple dimensions. In general, the
solution to the vector Helmholtz wave equation (2.38) is given by

E (r ) = E0 e ± jk×r (2.68)

where the wavenumber vector is

k = xˆ kx + yˆ ky + zˆ kz (2.69)
and the position vector is

r = xˆ x + yˆ y + zˆ z (2.70)
2.2  Time-Harmonic Electromagnetic Fields 37

The magnetic field is given in general by


1
H= k ´ E (2.71)
ωµ
and similarly the electric field is given in general by

E = -ωµk ´ H (2.72)

2.2.3  Energy and Power


The electric energy density of an electromagnetic wave is
1
ue = E × D (2.73)
2
while the magnetic energy density is
1
um = H × B (2.74)
2
The total electromagnetic energy density is the sum of the individual electric and
magnetic energy densities. If the electromagnetic wave is propagating through a lin-
ear, homogenous, lossless medium, the constitutive parameters are time-­invariant,
and the total energy density is

u = ue + um =
1
2 ( 2 2
ε E + µH ) (2.75)

In order to determine the power carried by the electromagnetic wave, the time
rate of change of the energy density is considered. This is given by
¶u ¶E ¶H
= εE × + µH × (2.76)
¶t ¶t ¶t
Using Faraday’s and Ampere’s laws, the time derivatives of the electric and mag-
netic fields are
¶E 1
= (Ñ ´ H - J) (2.77)
¶t ε
and
¶H 1
= - Ñ ´ E (2.78)
¶t µ
and the time rate of change of the energy density is therefore

¶u
= E × (Ñ ´ H) - H × (Ñ ´ E) - E × J (2.79)
¶t
Now using the vector identity

B × (Ñ ´ A) - A × (Ñ ´ B) = Ñ × (A ´ B) (2.80)
38 Microwave and Millimeter-Wave Remote Sensing for Security Applications

yields
¶u
+ Ñ × (E ´ H) = -E × J (2.81)
¶t

Equation (2.81) describes the balance of energy in an electromagnetic system


and is referred to as Poynting’s theorem after the physicist J. H. Poynting. The
first term on the left-hand side represents the differential change in energy
density, while the second term represents the flow of power. The term on the
right-hand side, the product of the electric field and the electric current density,
represents the work done by the fields on the sources. In the absence of sources,
the rate of change of energy density is exactly balanced by the outward flow of
power:
¶u
Ñ×S + = 0 (2.82)
¶t
where

S = E ´ H (2.83)

is the Poynting vector, which represents the power carried by the electromag-
netic wave. For time-harmonic electromagnetic fields the Poynting vector is given
by
1
S= E ´ H* (2.84)
2
where * indicates the complex conjugate. Equation (2.82) is of the same form as
the continuity equation (2.6), and the Poynting theorem is similarly a statement of
conservation of energy. The net flow of power through a closed surface must be
equal to the rate of change of energy density within the surface.

2.3  Wave Polarization

As a time-harmonic wave propagates through time and space, the electric and mag-
netic field vectors change. The orientation of the electric field vector as a function of
time and space is called the polarization of the wave. The orientation may be static
or dynamic in time and space. The polarization of the receiving antenna must be
matched to that of the incident radiation for maximum signal conversion into the
receiver, as discussed in Chapter 4: if the polarization of the antenna and the inci-
dent radiation are orthogonal, no signal is transferred to the receiving hardware.
In general, the polarization of the incident wave will be different from that of the
antenna. In passive radiometric remote sensing, the incident radiation is unpolar-
ized while the antenna can respond to one direction of polarization; thus, only half
of the power of the incident radiation can be received by the receiver system. Active
systems receive signals transmitted by the system and reflected off an object. The
polarization is thus well matched; however, some depolarization, however minimal,
generally occurs on the reflected signal in practice.
2.3  Wave Polarization 39

|E|
Ey

φ
z Ex x

Figure 2.2  Linearly polarized electric field vector.

2.3.1  Linear Polarization


In the simplest case, the electric field vector is oriented in a single direction in the x-y
plane, which is called linear polarization. A linearly polarized wave can be described by

E = (xˆ Ex + yˆ Ey )e j(ωt -kz) (2.85)


The magnitude of this field traces out a line in the x-y plane with magnitude

E = Ex2 + Ey2 (2.86)


as indicated by Figure 2.2. The angle separating the field vector and the x-axis is
given by
æ Ey ö
φ = tan-1 ç ÷ (2.87)
è Ex ø
The dynamic nature of the field vector can be seen by considering the real part of
(2.85) at an arbitrary fixed location along the z axis; for simplicity, z = 0 is chosen.
Then the real part of the field is

Re{E} = (xˆ Ex + yˆ Ey )cos(ωt) (2.88)


Figure 2.3 shows the change in the electric field vector as a function of time. When
ωt = 0, the vector has positive magnitude |E| at an angle f. When ωt = π ∕ 2, the

ωt = 0 ωt = π/2 ωt = π

|E|
Ey

-Ey
Ex |E|

-Ex
|E|
Figure 2.3  Linearly polarized electric field vector orientation over time.
40 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Figure 2.4  Over time (or space) a linearly polarized electric field traces out a sinusoid.

cosine term is zero and the vector has zero magnitude. When ωt = π, the vector has
magnitude –|E| at an angle f + π, in the opposite direction. The vector thus traces
out a sinusoid over time, as described in Figure 2.4.
If the y component of the electric field is zero (Ey = 0), the magnitude of the
vector is Ex and its angle is f = 0; the field is oriented along the x-axis. Such a wave
is linearly polarized along the x direction. Similarly, if Ex = 0, the wave is linearly
polarized along the y direction. Note that the same linear behavior arises in both
the temporal and spatial domains; the earlier analysis could have alternatively fixed
the time variable (conveniently at t = 0) and varied the spatial variable kz with the
same sinusoidal response in space.

2.3.2.  Elliptical Polarization


Whereas linear polarization represents the simplest form of wave polarization, the
most general form of polarization of a monochromatic plane wave is elliptical po-
larization, of which other forms of polarization are special cases. Thus, any mono-
chromatic wave can be described in terms of elliptical polarization. Consider an
electromagnetic wave given by

E = (xˆ Ex + yˆ Ey ) (2.89)

where

Ex = E1 cos(ω t - kz) (2.90)

Ey = E2 cos(ωt - kz + δ ) (2.91)

The phase δ represents the phase difference between the two waves making up the
total wave. In order to more simply analyze the dynamic nature of the wave, take a
fixed location in space z = 0. The wave components are then

Ex = E1 cos ω t (2.92)

Ey = E2 cos(ωt + δ ) = E2 (cos ωt cos δ - sin ωt sin δ ) (2.93)


2.3  Wave Polarization 41

or
Ex
= cos ω t (2.94)
E1

Ey
= cosω t cos δ - sinω t sin δ (2.95)
E2
Squaring and adding (2.94) and (2.95) results in

2 2
æ Ex ö æ Ey ö 2 2 2 2 2
ç E ÷ + ç E ÷ = cos ωt + cos ω t cos δ + sin ω t sin δ
è 1ø è 2ø (2.96)

-2 cos ωt sin ωt cos δ sin δ

Using the squares of (2.94) and (2.95), the last term on the right-hand side can be
written

æ E ö æ Ey ö
2 cosω t sinω t cosδ sin δ = 2 cos2 ω t cos2 δ - 2 ç x ÷ ç ÷ cos δ (2.97)
è E1 øè E2 ø

Substituting this into (2.96) yields


2 2
æ Ex ö æ Ey ö æ Ex ö æ Ey ö 2
ç E ÷ + ç E ÷ - 2ç E ÷ç E ÷ cos δ = sin δ (2.98)
è 1ø è 2ø è 1 øè 2 ø

This is the equation of an ellipse, and thus the electric field vector traces out an el-
liptical shape over time. The wave is therefore said to be elliptically polarized.
Equation (2.98) represents the most general form of polarization that a mono-
chromatic wave can take. Other forms of wave polarization are given as special
cases of (2.98). When the phase difference between the components is δ = 0 or π,
the polarization reduces to linear polarization. In such a case, with δ = 0 and E2 = 0,
the resulting wave is linearly polarized in the x-direction.

y
ωt = π/2

ωt = 3π/4 ωt = π/4

ωt = π ωt = 0
x

ωt = 5π/4 ωt = 7π/4

ωt = 3π/2

Figure 2.5  Circularly polarized wave vectors.


42 Microwave and Millimeter-Wave Remote Sensing for Security Applications

A significant special case is found when δ = ±π ∕ 2 and E1 = E2 = E0. The polar-


ization given by (2.98) then becomes

Ex2 + Ey2 = E02 = const (2.99)


which is the equation of a circle, and the wave is said to be circularly polarized.
Consider, for example, the case where δ = π ∕ 2 and E1 = E2 = E0. Substituting (2.90)
and (2.91) into (2.89) then results in

E = xˆ E0 cos ωt + yˆ E0 sin ω t (2.100)


where the wave is analyzed at the point z = 0. As the wave propagates in time the
electric field vector traces out a circle, as shown in Figure 2.5. When ωt = 0, the elec-
tric field is directed along the x-axis; when ωt = π ∕ 4 it is at an angle of 45˚ to the
x-axis; and when ωt = π ∕ 2 the electric field is directed along the y-axis. Thus, when
δ = π ∕ 2 the field traces out a circle in the counter-clockwise direction; this is called
right-hand circular polarization because it conforms to the right-hand rule along
the direction of propagation. When δ = –π ∕ 2, the electric field vector traces out a
circle in the clockwise direction and is known as left-hand circular polarization.
It should be noted that in general, there is no unique way of defining the ori-
entation of the electric field vectors in a wave. That is, the axes of the ellipse do
not necessarily conform to the axes of the predefined coordinate system, and the
ellipse coordinates will be shifted away from the defined coordinates by an angle
given by (2.87). This is particularly true for passive systems, which do not transmit
a signal; the electromagnetic waves emitted by a source may be of any orientation.
However, the precise angle of the wave coordinate system is not generally useful
in passive radiometric systems; what is important is the relative angle or relative
phase difference between the field components, as this is what determines the type
of polarization. In active systems, the polarization of the reflective wave also may
not match that of the transmitted wave, as the source may alter the polarization of
the incident wave upon reflecting it.

References

[1] Jackson, J. D., Classical Electrodynamics, 3rd ed., Hoboken, NJ: John Wiley & Sons,
1999.
[2] Stratton, J. A., Electromagnetic Theory, New York, NY: McGraw-Hill, 1941.
[3] Harrington, R. F., Time-Harmonic Electromagnetic Fields, New York, NY: McGraw-Hill,
1961.
[4] Balanis, C. A., Advanced Engineering Electromagnetics, Hoboken, NJ: John Wiley & Sons,
1989.
[5] Rothwell, E. J., and M. J. Cloud, Electromagnetics. Boca Raton, FL: CRC Press, 2001.
Chapter 3

Electromagnetic Waves in Media

In the previous chapter, fundamental concepts of electromagnetic wave propaga-


tion were introduced for a general lossless media; in this chapter, the focus is on the
propagation of electromagnetic waves through media that is lossy due to conduc-
tivity, the effect of boundaries between different media, and properties of specific
media of interest in remote security sensing.
The media encountered by a propagating wave in remote sensing depends on
the kind of measurement being performed, as indicated in Figure 3.1. In a radar ap-
plication, the wave propagates from the radar antenna through the atmosphere and
reflects off the object of interest. The reflected wave then propagates back through
the atmosphere before impinging on the receiving antenna. In the case of human
presence detection, the transmitted wave encounters the clothing being worn by
the person and, if a fraction of the wave energy transmits through the clothing, the
wave encounters the human skin. If contraband detection is the goal, the wave may
also encounter the contraband beneath the clothing. The wave may also transmit
through walls between the antenna and the clothing. In a passive, radiometric ap-
plication, the wave is generated by natural thermal radiation in the object, which
then propagates through the atmosphere to the antenna. If a human is the object of
interest, the wave propagates through a clothing layer and perhaps through walls
before being incident on the antenna.
The specific types of media encountered in remote security sensing, as well as
their physical makeup such as thickness, depend on the application. In most cases,
however, the propagating wave travels through the atmosphere, and in many situa-
tions in security sensing a human is present in the measurement as well, either as the
object to be detected or as the background to concealed contraband. Propagating
waves thus travel through multiple media types, each with different physical and
electromagnetic characteristics that affect the amplitude and phase of the propagat-
ing waves. Upon encountering a boundary between two media, part of the wave
energy will reflect off the new medium, and part of the energy will transmit into it. It
is therefore important to understand the effects of lossy media and media boundaries
in order to properly design a remote sensor. For example, if contraband detection
through walls using radar is the specified application, the transmitted signal must
have enough power to transmit through various wall and clothing materials, reflect
off the contraband, and then pass again through the clothing and wall materials and
still retain enough power to ensure a reasonable signal-to-noise ratio for detection.
This chapter begins by deriving the effects on a propagating electromagnetic
wave through a general lossy medium. Following this, the interaction of propa-
gating waves and boundaries between different media are discussed. Finally, the
chapter concludes by presenting parameters of specific media that are of interest
in microwave and millimeter-wave remote security sensing, including atmospheric

43
44 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Radar

Atmosphere Clothing Air Skin

(a)

Radiometer

Atmosphere Clothing Air Skin

(b)
Figure 3.1  (a) Diagram of the media encountered in the application of radar to contraband detec-
tion. (b) Diagram of the media encountered in the application of radiometry to human presence
detection.

effects, attenuation through wall and clothing materials, and parameters of differ-
ent types of human tissue.

3.1  Plane Wave Propagation in Unbounded Media

In a general medium, the time-harmonic Faraday’s law and Ampere’s law were
shown in Chapter 2 to be

Ñ ´ E = - jωµH (3.1)

Ñ ´ H = (σ + jω ε)E (3.2)

With nonzero conductivity, the resulting wave equation is of the form


æ σ ö
Ñ 2 E - ω 2µε ç 1 - j ÷ E = 0 (3.3)
è ω εø

In Chapter 2, it was also shown that a plane wave propagating in the z direction has
components only in the x and y directions. Consider a plane wave with an electric
field component directed along the x-axis. The solution of (3.3) is then
σ
- jω µε 1- j z
Ex (z) = E0 e ωε = E0 e - jkz (3.4)
where
σ σ
k = ω µε 1 - j = kr 1 - j (3.5)
ωε ωε
3.1  Plane Wave Propagation in Unbounded Media 45

is the complex wave number and

kr = ω µε (3.6)

is the noncomplex wavenumber introduced in Chapter 2. For a lossless media,


s = 0, and (3.5) reduces to kr.
The effects of the complex wavenumber on the propagation of the wave can be
more easily determined by separating the argument of the exponential into real and
imaginary parts. That is, let

Ex (z) = E0 e - jkz = E0 e -α z e - j β z (3.7)


where, from (3.5),

jk = α + jβ = -ω 2 µε + jωµσ (3.8)

The quantity (3.8) is also called the complex phase constant γ = α + jβ, with which
the electric field component can be written

Ex (z) = E0 e -γ z (3.9)

Squaring (3.8) and equating the real and imaginary parts yields the relations

α 2 - β 2 = -ω 2µε (3.10)

2αβ = ωµσ (3.11)


from which is derived
1
é æ 2 öù 2
1 æ σ ö
α =ω µε ê ç 1+ ç ÷ - 1 ÷ú (3.12)
ê2 ç è ωε ø ÷ú
ë è øû

1
é æ 2 öù 2
ê 1ç æ σ ö
β = ω µε 1 + ç ω ε ÷ + 1÷ú (3.13)
ê2 ç è ø ÷ú
ë è øû

The first term α is called the attenuation coefficient, or the absorption coef-
ficient, which is measured in Np·m−1, and β is the phase coefficient, mea-
sured in rad·m−1. In (3.7), the wave includes an exponential of the attenuation
c­oefficient multiplied by the distance, which results in a reduction of the am-
plitude of the wave as it propagates further into the medium; the lost energy
is converted to thermal energy in the medium. The phase coefficient alters
the phase of the wave as it propagates. Note that for a medium with zero
conductivity,

a = 0 (3.14)

β = ω µε = kr (3.15)
46 Microwave and Millimeter-Wave Remote Sensing for Security Applications

and the wave is given by

Ex (z) = E0 e - jkr z (3.16)

which is the same form that was derived in Chapter 2. Such a wave does not experi-
ence attenuation; thus, nonconducting media are lossless.
As the wave propagates in a conducting medium, the amplitude is decreased.
The distance at which the wave amplitude is decreased to a value of e−1 of the origi-
nal amplitude is call the skin depth δ, which is found by

e -α z e - j β z = e - z (3.17)
which yields α = 1, or
1
δ = m (3.18)
α
Certain media can be categorized by the relative value of the conductivity to the
product of the angular frequency and permittivity. In particular, the loss tangent
tanδ can be defined by
σ
tan δ = (3.19)
µε
which is present in the definitions of both the attenuation and phase coefficients.
Note that the loss tangent tanδ is not the tangent of the skin depth δ. Materials with
large loss tangents have high conductivity and high loss, whereas a low loss tangent
indicates a low-loss medium.
The phase velocity of the wave can be given in terms of the phase constant
through
ω
v= (3.20)
β

and the wavelength can be given by



λ= (3.21)
β

3.1.1  Good Conducting Media


A medium that has high conductivity relative to the product of the permittivity and
permeability is considered to be a good conductor. Such a medium also has high loss,
as its attenuation will be significant. A good conducting medium is characterized by
σ
>> 1 (tan δ >> 1) (3.22)
µε

The argument in the square root of (3.12) and (3.13) can then be approximated by
2
æ σ ö σ σ
1+ ç ÷ +1 » +1 » = tan δ (3.23)
è ωε ø ωε ωε
3.1  Plane Wave Propagation in Unbounded Media 47

and the resulting attenuation and phase coefficients are therefore

1 1
α » ω µε tan δ = ωµσ (3.24)
2 2

1 1
β » ω µε tanδ = ωµσ (3.25)
2 2

and the wave is given by


ω µσ
-(1+ j) z
Ex (z) = E0 e 2 (3.26)

From (3.18) and (3.24), the skin depth of a good conducting medium is given by

2
δ = (3.27)
ω µσ

Because σ is large, the skin depth is small. The skin depth for good conductors,
such as gold, copper, and aluminum, is on the order of 8 ´ 10−10 m at 10 GHz.
Therefore, the majority of the current in a conductor is confined to the surface; the
energy dissipates rapidly as the wave propagates into the medium. A perfect con-
ductor has infinite conductivity, and therefore its skin depth is 0, and a propagating
wave cannot penetrate into the medium. Thus, no fields can exist within a perfectly
conducting medium.

3.1.2  Good Dielectric Media


In contrast to a good conductor, a good dielectric medium is characterized by a rela-
tively low conductivity relative to the product of the permittivity and permeability,
and thus relatively low loss, by

σ
<< 1 (tan δ << 1) (3.28)
µε

By using the binomial theorem, the term inside the square root in (3.12) can be
written

2 2 4 2
æ σ ö 1æ σ ö 1æ σ ö 1æ σ ö
1+ ç ÷ = 1+ ç ÷ - ç ÷ � » 1+ ç ÷ (3.29)
è ωε ø 2 è ωε ø 8 è ω εø 2 è ωε ø

where the higher order terms are neglected due to (3.28), and the first two terms
have been retained. The attenuation coefficient can then be written

2
1æ σ ö 1 σ µ
α » ω µε ç ÷ = kr tan δ = (3.30)
4 è ωε ø 2 2 ε
48 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Similarly, the phase constant is

β » ω µ ε = kr (3.31)

where only the first term of the binomial expansion has been retained. The wave is
then given by
1
- kr tan δ z
Ex (z) = E0 e 2 e - jkr z (3.32)

The skin depth of a good dielectric is therefore, from (3.18) and (3.30),
2
δ = (3.33)
kr tan δ

3.1.3  Wave Impedance in Media


The electric and magnetic vector components of a transverse wave are orthogonal,
as discussed in Chapter 2, where it was shown in particular that the magnetic field
component associated with an electric field component directed along the x-axis is
given by
1 ¶Ex
Hy = j (3.34)
ωµ ¶z

In a lossy medium, this results in

α + jβ γ
Hy = - j Ex = - j Ex (3.35)
ωµ ωµ

The wave impedance is the ratio of the electric and magnetic field components,
which is thus
Ex ωµ jω µ
η= =j = (3.36)
Hy γ σ + jωε

For a good dielectric, the resulting wave impedance is therefore

µ
η» (3.37)
ε

whereas the impedance for a good conductor is given by


jωµ ωµ
η» = (1 + j) (3.38)
σ σ

3.1.4  Complex Permittivity and Dispersion


Considering Ampere’s law with time-harmonic fields

Ñ ´ H = (σ + jωε )E (3.39)
3.1  Plane Wave Propagation in Unbounded Media 49

the permittivity of the medium can be defined as

σ
εc = ε - j = ε¢- jε ¢¢ (3.40)
ω

which is the complex permittivity such that Ampere’s law can be written

Ñ ´ H = jωεc E (3.41)

Similarly, the complex permeability can be written

µc = µ¢ - j µ¢¢ (3.42)

Most media have noncomplex permeability, with the imaginary part of the perme-
ability equal to zero. The permeability is then simply mc = m. The real part of the
permittivity can also be given by ε¢ = ε, while the complex part of the permittivity
is defined through (3.40) in terms of the conductivity as

σ
ε ¢¢ = (3.43)
ω

In terms of (3.40) and (3.42), the attenuation and phase constants can be given by

1
ì é ü2
ï1 æ ε¢¢ öæ
2
µ¢¢ ö æ
2
µ¢¢ε ¢¢ ö ùú ï
α = ω µ¢ε¢¢ í ê çç 1 + ÷ç 1 + -
÷ ç 1 - ÷ ý (3.44)
ïî 2 êë è ε ¢2 ÷ç
øè µ¢2 ÷ø è µ ¢ε ¢ ø ú ï
ûþ

1
ì é ü2
ï1 æ ε ¢¢ öæ
2
µ¢¢ ö æ
2
µ¢¢ε ¢¢ ö ùú ï
β = ω µ ¢ε ¢¢ í ê çç 1 + ÷ç 1 + +
÷ ç 1 - ÷ ý (3.45)
ïî 2 êë è ε ¢2 ÷ç
øè µ ¢2 ÷ø è µ ¢ε ¢ ø ú ï
ûþ

which, for noncomplex permeability, reduces to (3.12) and (3.13). Throughout the
rest of the book, ε and μ will be used to denote the permittivity and permeability,
whether real or complex, for brevity.
In general, the permittivity and permeability are not constant with frequency;
that is,

ε ® ε (ω ), µ ® µ (ω) (3.46)

Generally, however, the permeability does not vary appreciably with frequency for
most media. The frequency-dependent relative permittivity is dependent on various
properties of the medium and is given by [1]

ε (ω) ωp2,i
= 1+ å 2 2
(3.47)
ε0 i ωi - ω + jω 2Gi

50 Microwave and Millimeter-Wave Remote Sensing for Security Applications

where ωp,i is a plasma frequency of the medium, ωi is a dipole moment resonance


frequency, and Gi is a damping coefficient. In the limit as the frequency approaches
zero, the permittivity is

ε (ω) ωp2,i
lim = 1+ å 2 (3.48)
ω ®0 ε 0 i ωi

which is the static, or dc, relative permittivity of the medium. When the frequency
approaches infinity, the real and imaginary components of the relative permittivity
are given by

ε ¢(ω )
lim = 1 - ω -2 åω p2,i (3.49)
ω ®¥ ε 0
i

ε ¢¢(ω )
lim = -2ω -3 åω p2,i Gi (3.50)
ω ®¥ ε0 i

which are the real and imaginary components of the relative permittivity of a
plasma. The real and imaginary components of the complex permittivity are not
independent and are related through the Kramers-Kronig relations [2, 3]
¥
2 ω ¢ε ¢¢
ε ¢(ω ) = 1 + Pò 2 dω ¢ (3.51)
π 0 ω¢ - ω2

¥
2ω 1 - ε¢
ε ¢¢(ω ) = Pò 2 dω ¢ (3.52)
π 0 ω¢ - ω2

where σ0 is the static (dc) conductivity of the medium, and Pò indicates the principal
value, which excludes 0 and ω from the integration.
The effect of the frequency dependence of the permittivity and permeability on
propagating waves arises in the phase velocity of the wave, which is

1
vp = (3.53)
µ(ω )ε (ω )

The phase velocity is thus a function of frequency, and different frequency compo-
nents of nonmonochromatic waves will thus experience different phase velocities,
a phenomenon called dispersion. The implications of dispersion can be seen by
considering a rectangular electromagnetic pulse, which contains multiple frequency
components. Because each frequency component experiences a slightly different
phase velocity, the pulse shape begins to broaden and spread as some components
propagate faster and some slower. As illustrated in Figure 3.2, the distortion of
the pulse increases as the distance travelled through the medium increases. Waves
that are sufficiently narrowband can be approximated as monochromatic, and the
permittivity and permeability can be considered to be constant, in which case dis-
persion may be neglected. In practice, however, all waves experience dispersion. A
3.2  Plane Wave Propagation in Bounded Media 51

Distance

Initial pulse Distorted pulse


Figure 3.2  Dispersion of a rectangular pulse in a medium.

wave can be considered a superposition of multiple monochromatic waves, each


having a different constant phase velocity.

3.2  Plane Wave Propagation in Bounded Media

The previous analyses focused on the propagation of plane waves in homogeneous,


unbounded media. When two different media are joined, there exists at the inter-
face a discontinuity of the constitutive parameters, which thus affects the electro-
magnetic waves present in both media. The propagation characteristics derived for
a single unbounded medium do not hold when the plane wave propagates across
the boundary; however, if the fields immediately on either side of the boundary can
be specified, the solutions to the fields can be carried over from one medium to the
next.
The conditions that specify the behavior of the fields at the boundary are called
the boundary conditions or jump conditions, and are postulated from the integral
form of Maxwell’s equations [3, 4]. Two sets of boundary conditions specify the be-
havior across the boundary of the tangential and normal components of the fields.
Consider the situation described in Figure 3.3, where two media area separated
by a boundary containing electric and magnetic sources. The tangential boundary
conditions are

nˆ ´ (E1 - E 2 ) = - Jm (3.54)

nˆ ´ (H1 - H 2 ) = J (3.55)

Thus, the difference between the tangential components of the electric fields must be
equal to the magnetic current density at the boundary, and the difference between
the magnetic fields must be equal to the electric current density at the boundary. The
normal boundary conditions are

nˆ × (D1 - D2 ) = ρ (3.56)

Media 1 ^
n
E1 , H 1 , D 1 , B 1

J, Jm, ρ, ρm ^
-n

Media 2
E2 , H 2 , D 2 , B 2

Figure 3.3  Electromagnetic fields and sources at the boundary between two media.
52 Microwave and Millimeter-Wave Remote Sensing for Security Applications

nˆ × (B1 - B2 ) = ρm (3.57)

The normal components of the fields are thus discontinuous by the charge densities
present on the boundary surface.
In the absence of sources, the boundary conditions are given by

nˆ ´ (E1 - E 2 ) = 0 (3.58)

nˆ ´ (H1 - H 2 ) = 0 (3.59)

nˆ × (D1 - D2 ) = 0 (3.60)

nˆ × (B1 - B2 ) = 0 (3.61)

Thus, the tangential and normal components of the fields are continuous across the
boundary when no sources are present on the boundary surface.

3.2.1  Reflection and Transmission of Normally Incident Waves


When a plane wave is incident on a boundary between two media, a portion of the
amplitude of the wave is reflected off the surface back into medium 1 and a portion
is transmitted into medium 2. Consider a plane wave propagating in the z direction
that is normally incident on a boundary with no sources located at z = 0, as shown
in Figure 3.4. The incident electric field is specified by

Ei = xˆ E0e - jk1z (3.62)

From Ampere’s law, the incident magnetic field is then given by


E0 - jk1z
Hi = yˆ e (3.63)
η1

The fields reflected by the boundary back into medium 1 are given by

E r = xˆ RE0e - jk1z (3.64)

E0 - jk1z
H r = -yˆ R e (3.65)
η1

Ei
Et
Er

Media 1 Media 2
ε1, µ1 ε2, µ2

Figure 3.4  A normally incident wave reflected and transmitted by a boundary between two media.
3.2  Plane Wave Propagation in Bounded Media 53

where R is the reflection coefficient, which specifies the fraction of the wave am-
plitude that is reflected by the boundary. The fields transmitted into the second
medium are given by

Et = xˆ TE0 e - jk2 z (3.66)

E0 - jk2 z
Ht = yˆ T e (3.67)
η2

where T is the transmission coefficient, which specifies the fraction of the amplitude
that passes into medium 2.
The boundary conditions are used to determine the reflection and transmission
coefficients, which are defined at the boundary between the two media. Because
the wave is normally incident and the electric and magnetic field components of a
plane wave are perpendicular to the direction of propagation, the fields given by
(3.62)–(3.67) are tangential to the boundary surface. Because the boundary is free
of sources, the boundary conditions require that the tangential fields on either side
of the boundary must be equal; that is,

Ei z =0
+ Er z =0
= Et z =0
(3.68)

Hi z =0
+ Hr z =0
= Ht z =0
(3.69)

Substituting (3.62)–(3.67) results in

1 + R = T (3.70)

1 1
(1 - R) = T (3.71)
η1 η2

Solving (3.70) and (3.71) for R and T yields

η2 - η1
R= (3.72)
η 2 + η1

2η2
T= (3.73)
η2 + η1

Thus, the reflection and transmission coefficients for a normally incident plane
wave can be determined in terms of only the intrinsic impedances of the media. In
the case that η1 = η2, R = 0 and T = 1; thus, the wave is completely transmitted into
the second medium with no reflections.
The reflection coefficient can be determined at any arbitrary point z0 in medium
1 by deriving R from (3.62) and (3.64), which results in

E r (z = z0 )
R z =z = = Re j 2k1z0 (3.74)

0 Ei (z = z0 )
54 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Similarly, the transmission coefficient away from the boundary can be specified by
the transmitted wave at point z2 in medium to and the incident wave at point z1 in
medium 1 by
Et (z = z2 )
T z =z = = Te - j(k2 z2 -k1z1) (3.75)
1 , z2 Ei (z = z1)

3.2.2  Reflection and Transmission of Arbitrarily Incident Waves


A wave incident on a boundary at an arbitrary angle can be categorized into one of
two cases depending on the orientation of the electric field in relation to the plane
of incidence, which is the plane formed by the direction vector of the wave and the
unit vector normal to the boundary surface. When the electric field is perpendicular
to the plane of incidence, it is called transverse electric; when the electric field is
parallel to the plane of incidence, it is called transverse magnetic. If the boundary is
assumed to extend infinitely in one dimension, the axes may be arbitrarily defined
such that the plane of incidence conforms to one of these cases. Additionally, an
arbitrary wave may be considered to be the superposition of a TE and TM wave, in
which case the fields can be decomposed and analyzed separately.

3.2.2.1  Transverse Electric (Perpendicular) Incidence


The incident, reflected, and transmitted fields of a transverse electric wave incident
on a surface as depicted in Figure 3.5 are given by

Ei = yˆ E0e - jk1(sin θi x + cos θi z) (3.76)

E r = yˆ R^ E0 e - jk1(sin θr x -cos θr z) (3.77)

Et = yˆ T^ E0 e - jk2 (sin θt x + cos θt z) (3.78)

x
θr
Hr
z
Er
Et
θr θt
θi Ht
θt

Ei

Hi
θi
Media 1 Media 2
ε1, µ1 ε2, µ2

Figure 3.5  Transverse electric wave arbitrarily incident on a boundary.


3.2  Plane Wave Propagation in Bounded Media 55

E0 - jk1(sinθi x + cos θi z)
Hi = (-xˆ cosθi + zˆ sin θi ) e (3.79)
η1

E0 - jk1(sinθ r x -cos θr z)
H r = (xˆ cos θr + zˆ sin θr)R^ e (3.80)
η1

E0 - jk2 (sin θt x + cos θt z)


Ht = (-xˆ cosθt + zˆ sin θt )T^ e (3.81)
η2

where R^ and T^ are the perpendicular, or TE, reflection and transmission


coefficients.
Two notable relationships between the angles of the incident, reflected, and
transmitted waves can be derived according to the boundary conditions, which
state that the tangential fields must be continuous at z = 0. The phases of the fields
at the boundary must therefore be equal, in which case

k1 sin θi = k1 sin θr = k2 sin θt (3.82)

From the first equality,

θi = θr (3.83)

which is known as Snell’s law of reflection. The reflected wave thus propagates
away from the boundary at an angle equal to that of the incident wave. Therefore,
from the second equality,
k1 sin θt
= (3.84)
k2 sin θi

which is Snell’s law of refraction, from which an expression for the angle of the
transmitted wave is found in terms of the incident wave angle and the wavenumbers
in the two media:
æk ö
θt = sin-1 ç 1 sin θi ÷ (3.85)
k
è 2 ø

In the same fashion as was done for the case of normal incidence, the TE reflection
and transmission coefficients are derived by equating the tangential components of
the fields at the boundary. The result of the electric fields at z = 0 is

1 + R^ = T^ (3.86)

while that of the magnetic fields is

1 1
(1 - R^ )cosθi = T^ cos θt (3.87)
η1 η2

The first expression is the same as that derived for normal incidence; this is true
because the electric field vector is still parallel to the boundary. The second expres-
sion reduces to that derived for normal incidence when the angle θi = 0, where,
56 Microwave and Millimeter-Wave Remote Sensing for Security Applications

a­ccording to (3.85), θt is also zero. Solving (3.86) and (3.87) for the TE reflection
and transmission coefficients results in

η2 cos θi - η1 cos θt (3.88)


R^ =
η2 cos θi + η1 cos θt

2η2 cos θi (3.89)


T^ =
η2 cos θi + η1 cos θt

The magnitude of the TE reflection and transmission coefficients are plotted in


Figure 3.6 as a function of the incidence angle θi for nonmagnetic media (μ1 = μ2 =
μ0) for various ratios of the permittivities of the two media. As the incidence angle
approaches 90˚ the magnitude of the reflection coefficient approaches unity, as the
wave reaches an angle where it is no longer incident on the boundary. Conversely,
the transmission coefficient approaches zero as the incidence angle increases for the
same reason.

0.8
ε2/ε1 = 20
0.6 ε2/ε1 = 10
|R⊥|
ε2/ε1 = 5
0.4
ε2/ε1 = 2.5
0.2

0
0 20 40 60 80 90
θi (degrees)
(a)
1
ε2/ε1 = 2.5
0.8
ε2/ε1 = 5
0.6
|T⊥| ε2/ε1 = 10
0.4 ε2/ε1 = 20

0.2

0
0 20 40 60 80 90
θi (degrees)
(b)
Figure 3.6  Magnitudes of (a) the reflection coefficient and (b) the transmission coefficient for a TE
wave as a function of incidence angle.
3.2  Plane Wave Propagation in Bounded Media 57

3.2.2.2  Transverse Magnetic (Parallel) Incidence


The electric and magnetic field components for an incident TM wave as shown in
Figure 3.7 are given by

Ei = (xˆ cosθi - zˆ sin θi )E0 e - jk1(sin θi x + cos θi z) (3.90)

E r = (xˆ cosθr + zˆ sin θr )R||E0e - jk1(sin θr x -cos θr z) (3.91)

Et = (xˆ cosθt - zˆ sin θt )T||E0e - jk2 (sin θt x + cos θt z) (3.92)

E0 - jk1(sin θi x + cosθi z)
Hi = yˆ e (3.93)
η1

E0 - jk1(sin θr x -cos θr z)
H r = -yˆ R|| e (3.94)
η1

E0 - jk2 (sin θt x + cosθt z)


Ht = yˆ T|| e (3.95)
η2

By equating the phases at the boundary, the same expressions (3.83) and (3.84) are
found. Equating the tangential fields at the boundary yields

(1 + R|| )cosθi = T|| cosθt (3.96)

1 1
(1 - R|| ) = T|| (3.97)
η1 η2

Solving (3.96) and (3.97) simultaneously yields the TM reflection and transmission
coefficients

x
θr
Er
z
θt
Hr Et

θr θt Ht
θi

θi
Ei

Hi

Media 1 Media 2
ε1, µ1 ε2, µ2

Figure 3.7  Transverse magnetic wave arbitrarily incident on a boundary.


58 Microwave and Millimeter-Wave Remote Sensing for Security Applications

η2 cos θt - η1 cos θi
R|| = (3.98)
η2 cos θt + η1 cos θi

2η2 cos θi
T|| = (3.99)
η2 cos θt + η1 cos θi

The magnitude of the TM reflection and transmission coefficients are plotted in Fig-
ure 3.8 as a function of the incidence angle θi for nonmagnetic media (μ1 = μ2 = μ0)
for various ratios of the permittivities of the two media. The magnitude of the trans-
mission behaves similarly to the TE case, trending toward zero with increasing angle
of incidence. The magnitude of the reflection coefficient, however, displays different
behavior; there is an angle where the reflection coefficient is zero, and all the power
in the direction of propagation is transmitted into the second medium. This angle is
called the Brewster angle and is discussed in Section 3.2.4 in more detail.

3.2.3  Power Reflection and Transmission


The power carried by a propagating wave is given by the Poynting vector, and the
power that is reflected and transmitted by the boundary is given in terms of the

0.8
ε2/ε1 = 20
0.6 ε /ε = 10
2 1
|R|||
ε2/ε1 = 5
0.4
ε2/ε1 = 2.5
0.2

0
0 20 40 60 80 90
θi (degrees)
(a)
1
ε2/ε1 = 2.5
0.8
ε2/ε1 = 5
0.6
ε2/ε1 = 10
|T|||
0.4 2/ε1 = 20
ε

0.2

0
0 20 40 60 80 90
θi (degrees)
(b)
Figure 3.8  Magnitudes of (a) the reflection coefficient and (b) the transmission coefficient for a TM
wave as a function of incidence angle.
3.2  Plane Wave Propagation in Bounded Media 59

component of the Poynting vector that is normal to the boundary. The normal com-
ponents of the reflected and transmitted Poyning vectors are thus given in terms of
the normal component of the incident Poynting vector by

zˆ × Sr = -zˆ × GSi (3.100)

zˆ × St = zˆ × ¡ Si (3.101)

where G and ¡ are the power reflection and transmission coefficients, called the
reflectivity and transmissivity, respectively. Due to conservation of energy, the re-
flectivity and transmissivity are related through

G + ¡ = 1 (3.102)

The reflectivity and transmissivity can be derived in terms of the normal compo-
nents of the Poynting vectors in both cases of TE or TM waves. For an incident TE
wave, the Poynting vector components normal to the boundary are
2
1 E0
zˆ × Si = zˆ × (Ei ´ H*i ) = zˆ cosθi (3.103)
2 2η1

2
1 R^ E0 2
zˆ × Sr = zˆ × (E r ´ H*r ) = -zˆ cosθi = -zˆ R^ Si (3.104)
2 2η1

2
1 T^ E0 2 η2 cosθi
zˆ × St = zˆ × (Et ´ H*t ) = zˆ cosθt = zˆ T^ Si (3.105)
2 2η2 η1 cosθ t

Equating (3.100) with (3.104) and also equating (3.101) with (3.105) yields the
expressions for the perpendicular reflectivity and transmissivity

2
G ^ = R^ (3.106)

2 η2 cos θi
¡ ^ = T^ (3.107)
η1 cos θt

A similar derivation for an incident TM wave yields the parallel reflectivity and
transmissivity
2
G|| = R|| (3.108)

2 η2cos θt
¡|| = T|| (3.109)
η1 cos θi

Thus, the power reflection coefficients for the two cases of polarization are simply
the squared magnitude of the reflection coefficients, and the power transmissions
coefficient are the squared magnitude of the transmission coefficients scaled by ra-
tios of the impedances and the cosines of the incidence and transmitted angles. For
60 Microwave and Millimeter-Wave Remote Sensing for Security Applications

a normally incident wave, the transmissivity reduces to the squared magnitude of


the transmission coefficient.

3.2.4  Total Transmission and Total Reflection


As shown in Section 3.2.2.2, there exists an incidence angle, called the Brewster
angle qB, at which all the incident power is transferred into the second medium with
no reflection. In addition to the Brewster angle there exists an angle at which the
incident power is completely reflected by the boundary that is called the critical
angle θc. The Brewster angle can be derived by setting R^ = R|| = 0 and solving
(3.88) and (3.98) for the angle of incidence. The result is [1]

µ2 µ1 ε2 - µ2 ε1
θB^ = sin-1 × (3.110)
ε1 µ12 - µ22

ε 2 µ2 ε1 - µ1 ε2
θB|| = sin-1 × (3.111)
µ1 ε12 - ε22

For nonmagnetic materials, the permeabilities of the two media are nearly equal. In
such a case, (3.110) results in a nonphysical angle, and thus, for perpendicularly polar-
ized waves, there does not exist an angle where the reflection coefficient is zero and all
energy is transmitted into the medium. Such an angle does exist in the case of parallel
polarization. When the permeabilities of the two media are equal, (3.111) reduces to

ε2
θB|| = tan-1 (3.112)
ε1

When θi = θB||, there is no reflected wave. Although the transmission coefficient


does not equal unity at this angle, the incident power carried by the wave in the z
direction is completely transmitted into the second medium. This can be seen from
(3.102) and (3.108); since the reflection coefficient is zero, the reflectivity is likewise
zero, and thus the transmissivity is unity.
The critical angle is the angle at which the incident wave is completely reflected
and can be derived from Snell’s law of refraction (3.84). The transmitted wave no
longer exists in the second medium when θt ³ 90˚, thus, the critical angle is found by
Snell’s law when the angle of the transmitted wave is equal to 90˚, which yields

æk ö µ2 ε2
θc = sin-1 ç 2 ÷ = sin -1 (3.113)
è k1 ø µ1ε1

From this and the definitions of reflectivity, it can be shown [5] that when θi ³ θc,

G ^ = G|| = 1 (3.114)

and thus

R^ = R|| = 1 (3.115)

and therefore the wave is completely reflected by the boundary.


3.2  Plane Wave Propagation in Bounded Media 61

3.2.5  Layered Media


Propagating plane waves often encounter multiple layers of different media in secu-
rity sensing applications. For instance, in concealed object detection a transmitted
wave encounters one or more layers of clothing before impinging on the concealed
object, or on the human skin, which itself consists of multiple layers. Such layered
media have multiple interfaces, each of which may transmit and reflect some por-
tion of the wave energy. Analysis of the interaction of the incident wave with the
separate media layers follows from the discussions on a single boundary in previous
sections; however, the effect of multiple reflections within the inner layers can make
the analysis of many layers complicated, particularly when the reflection c­oefficients
are not small. The bulk properties of the layered media as a whole are generally
derived recursively by starting with one of the outer layers and propagating the
solution toward the other side. This can be accomplished through a matrix formu-
lation [6] or by a recursive formulation defining the properties of the nth layer in
terms of layers n – 1 and n + 1 (see Figure 3.9).
The bulk reflection and transmission coefficients Rb and Tb, those of the layered
media as a whole, can be found in terms of the amplitudes of the fields within the me-
dia, illustrated in Figure 3.10. Each layer (except for the final semi-infinite layer) con-
tains an incident, or forward traveling, wave, and a reflected, or backward traveling,
wave. For layer n the TE incident and reflected electric fields can be given by [1]

(n )
x + kz(n)Dzn )
E(in) = yˆ an +1e - j(kx (3.116)

(n )
x -kz(n)Dzn )
E(rn) = yˆ bn +1e - j(kx (3.117)

where an+1 and bn+1 are the amplitudes of the incident and reflected wave in layer
n, and

D zn = zn +1 - zn (3.118)

is the thickness of layer n. In layer n = 0, the incident wave is the original incident
wave before impinging on the first media layer. In the final layer n = N, there is no
reflected wave.
Derivation of the field amplitudes follows from the solution of the boundary
conditions at the interface; the result is [1]

(1 + Rn )e - jkzδ zn
an +1 = an (3.119)
1 + Rn Rnb+1e - jkzδ zn

n-1 n n+1
Tn Tn+1
Rn Rn+1

Figure 3.9  Reflection and transmission of layered media.


62 Microwave and Millimeter-Wave Remote Sensing for Security Applications

0 1 n N-1 N
a1 a2 an+1 aN aN+1
b1 b2 ... bn+1 ... bN

δz1 δzn δzN

Figure 3.10  Amplitudes of the incident and reflected waves in layered media.

Rn + Rnb+1e - jkzδ zn
bn = an (3.120)
1 + Rn Rnb+1e - jkzδ zn

where the reflection coefficient between layers n and n – 1 is
ηn - ηn -1
Rn = (3.121)
ηn + ηn -1
The bulk reflection coefficient at layer n is thus

bn Rn + Rnb+1e - jkzδ zn
Rnb = = (3.122)
an 1 + Rn Rnb+1e - jkzδ zn

Equation (3.122) represents the reflection coefficient of the layers n to N, and is defined
recursively in terms of those layers following n. The bulk transmission coefficient is

an +1 (1 + Rn )e - jkzδ zn
Tnb+1 = = (3.123)
an 1 + Rn Rnb+1e - jkzδ zn

which is similarly defined in terms of the following layers.
Due to the dependence of the bulk reflection and transmission coefficients on
the layers following n, the recursion begins with the final layer, the semi-infinite
layer N. Because the layer is semi-infinite, RN+1 = 0, and the bulk reflection coef-
ficient from (3.122) is
b
RN = RN (3.124)

The transmission coefficient at the final boundary is evaluated likewise, with Dzn =
0 in the semi-infinite layer, yielding

TNb = 1 + RN = TN (3.125)

The bulk reflection and transmission coefficients at the initial boundary can then be
found recursively through (3.122) and (3.123).
If the reflection coefficients of all layers are small such that the bulk reflection
b
coefficients Rn+1 << 1 , the bulk reflection and transmission coefficients can be ap-
proximated by

Rnb » Rn + Rnb+1e - jkzδ zn (3.126)


Tnb+1 » (1 + Rn )e - jkzδ zn (3.127)


3.3  Electromagnetic Propagation in Specific Media 63

3.3  Electromagnetic Propagation in Specific Media

The previous sections presented a general formulation for the interaction of propa-
gating electromagnetic waves with general media. In this section, characteristics
of specific media are discussed, emphasizing those media that are encountered in
remote security sensing. Referring back to the general propagation scenario at the
beginning of this chapter, a propagating wave travels through the atmosphere and
interacts with clothing and garment material, human tissue, contraband material,
or building material such as those included in walls.
The atmosphere consists of the gases and water vapor present in dry air, as well
as hydrometeors such as fog, rain, or snow. Additional inclusions that may be pres-
ent are dust and smoke. Clothing and garment materials are generally thin layers
of specific media with little moisture content. At low microwave frequencies, these
materials are generally transparent. Building material may include drywall, wood,
cinder blocks, or other common wall-type materials.
In many remote sensing applications in security—whether remote detection of
intruders, through-wall measurements of moving people, imaging for contraband
detection, or other applications—a prominent factor is the presence of one or more
humans. As such, the dielectric properties of human tissue will be discussed in
some detail. The dielectric properties of other materials of interest, contraband
such as concealed weapons or explosives, for example, depend on the specific type
of material, which can vary significantly depending on the application, and will be
discussed in general terms. It is often the case, particularly in contraband detec-
tion, that the specific media properties are not known, and that changes against the
known profile of the human body are detected. Atmosphere, building material, and
garment material are generally not the media to be detected, and as such the char-
acteristic of primary concern is the attenuation suffered by the propagating wave as
it passes through the media. Thus, the discussions of these media will focus on the
attenuation effects over frequency.

3.3.1  Atmospheric Propagation Effects


Electromagnetic waves propagating between the sensor and the object of interest
undergo attenuation due to various resonances of molecules in the atmosphere.
The dry atmosphere is made up of ~78% nitrogen, ~21% oxygen, ~1% argon, and
~0.03% various other gases [7]. Water vapor is generally present in the atmosphere
and is quantified in terms of the density in g·m−3; a typical value used for general at-
mospheric calculations is 7.5 g·m−3. In the microwave and millimeter-wave regions
of the electromagnetic spectrum, atmospheric attenuation effects are dominated by
the molecular absorption of water vapor and oxygen. Water absorption is due to the
dipole moment of the water molecule and has absorption resonances at approximately
22 GHz, 183 GHz, 323 GHz, and 380 GHz. Oxygen has no electric dipole moment;
absorption is due to magnetic dipole transfers and has resonances in the millimeter-
wave region at approximately 60 GHz and 118 GHz. Higher frequency resonances
also occur.
Modeling of atmospheric absorption is complicated due to the various differ-
ent absorption effects that occur. In addition to water vapor, hydrometeors such
64 Microwave and Millimeter-Wave Remote Sensing for Security Applications

fog and rain must also be accounted for; snow incurs a different attenuation as
well. A model of the attenuation and temporal dispersion of the atmosphere was
developed by Liebe in 1993 [8], which includes the effects of temperature, pressure,
water vapor, and rain. The Liebe model calculates the absorption based on the
complex refractive index of the various atmospheric components, and is accurate
to approximately 0.2 dB in the frequency range 0–1000 GHz [9]. The model was
packaged into an openly available program called simply the millimeter-wave prop-
agation model (MPM). Figure 3.11(a) shows the attenuation of the atmosphere in
the frequency range 0–1000 GHz at sea level with 50% relative humidity and 15˚C
temperature. The resonances due to various molecules are apparent. Figure 3.11(b)
shows the attenuation of the atmosphere up to millimeter-wave frequencies with no
relative humidity for various temperatures, where it can be seen that higher tem-
peratures result in greater attenuation. Figure 3.11(c) shows the delay imparted on
a propagating wave due to dispersion.
There are locations in the attenuation spectrum where the propagation loss has
minima, at approximately 35 GHz, 94 GHz, 140 GHz, 215 GHz, and 342 GHz.
These bandwidths between the various resonances in the atmospheric attenuation
spectrum are referred to as atmospheric “windows” because they represent band-
width where signals can propagate for longer distances due to the lower attenu-
ation. Long-range remote sensing applications generally operate in one of these
windows to achieve maximum signal power reception; however, some clandestine
sensors operate at the resonance frequencies, such as 60 GHz, where signals are
difficult to intercept due to the high path loss. As seen in Figure 3.11(b), the attenu-
ation is generally below 10 dB·km–1 outside of the resonances, and thus short-range
remote sensors will not be significantly affected by propagation loss due to the
atmosphere and water vapor.
Water vapor and rain each cause additional attenuation and dispersion. Figures
12(a) and 12(b) show the attenuation and dispersion due to various levels of relative
humidity; Figures 13(c) and 13(d) show the attenuation and dispersion due to vari-
ous rates of rainfall. Significant rainfall can cause high levels of attenuation, as seen in
Figure 3.12, which will affect short-range sensors. An extensive review of the effects
of rain and snow on the propagation of electromagnetic waves is given in [10].
Fog (considered haze that renders the visibility less than 1 km) can be char-
acterized by dry air with suspended hydrometeors, either water or ice. Inclusions
of water droplets in fog or haze have radii on the order of 1 μm to approximately
100 μm or larger. The density of a medium fog is approximately 0.05 g·m−3, while
that of a thick fog can be 0.5 g·m−3. The International Telecommunication Union
has recommended a model of the attenuation due to clouds or fog that is valid up
to 200 GHz [11], given by

α = KM (3.128)
where α is measured in dB·km−1, K is the specific attenuation coefficient in
dB·m3·km−1·g−1, and M is the water density in g·m–3. The specific attenuation coef-
ficient is approximated by a Debye model of the permittivity of water:
0.819f
K= (3.129)
ε ¢¢(1 + ξ 2 )
3.3  Electromagnetic Propagation in Specific Media 65

104

103

102

α (dB/km)
101

100

10-1

10-2
0 200 400 600 800 1000
f (GHz)
(a)
103

102

40 oC
101
20 oC
α (dB/km)

0 oC
100
-20 oC
-40 oC
10-1 H 2O
O2 H 2O H 2O
O2
10-2
H 2O
10-3
0 50 100 150 200 250 300 350 400
f (GHz)
(b)
60

50

40
40 oC
τ (ps/km)

30

20
20 oC
10
0 oC
0
-20 oC -40 oC
−10
0 50 100 150 200 250 300 350 400
f (GHz)
(c)
Figure 3.11  (a) Atmospheric attenuation at sea level with 50% relative humidity and a temperature
of 15˚C. (b) Atmospheric attenuation and (c) propagation delay due to dispersion for various tem-
peratures with no relative humidity. The molecular absorption regions are highlighted.
66 Microwave and Millimeter-Wave Remote Sensing for Security Applications

103

102

101

α (dB/km)
100
90%
10-1 70%
50%
10-2 30%
10%

10-3
0 50 100 150 200 250 300 350 400
f (GHz)
(a)
30

25

20 90%
τ (ps/km)

70%
15
50%
10 30%
10%
5

−5
0 50 100 150 200 250 300 350 400
f (GHz)
(b)
102
50 mm/hr
40 mm/hr
30 mm/hr

101 20 mm/hr
α (dB/km)

10 mm/hr
5 mm/hr
2.5 mm/hr
100 1 mm/hr

10-1
0 100 200 300 400
f (GHz)
(c)
Figure 3.12  (a) Attenuation and (b) dispersion due only to water vapor density. (c)–(d) Attenuation
and (e) dispersion due to rainfall.
3.3  Electromagnetic Propagation in Specific Media 67

2
10
50 mm/hr
40 mm/hr
1
30 mm/hr
10 20 mm/hr
10 mm/hr

α (dB/km)
0
10

−1
10 5 mm/hr
2.5 mm/hr
1 mm/hr
−2
10
0 10 20 30 40 50
f (GHz)
(d)

1 mm/hr 2.5 mm/hr


0
5 mm/hr
10 mm/hr
−5
20 mm/hr
τ (ps/km)

30 mm/hr
−10
40 mm/hr

50 mm/hr
−15

−20
0 100 200 300 400
f (GHz)
(e)
Figure 3.12  (continued )

where f is the frequency in GHz,

2 +ε¢
ξ= (3.130)
ε ¢¢

and the permittivity is given by the Debye models

ε 0 - 5.48 1.97
ε ¢(f ) = 2
+ + 3.51 (3.131)
1 + (f f1) 1 + (f f2 )2

(ε 0 - 5.48)f 1.97 f (3.132)


ε ¢¢(f ) = +
f1 é1 + (f f1) ù f2 é1 + (f f2 )2 ù
2
ë û ë û
68 Microwave and Millimeter-Wave Remote Sensing for Security Applications

where

ε 0 = 77.6 + 103.3(300T -1 - 1) (3.133)

and T is the temperature in Kelvin. The first and second relaxation frequencies in
(3.131) and (3.132) are given by

f1 = 20.09 - 142(300T -1 - 1) + 294(300T -1 - 1)2 GHz (3.134)

f2 = 590 - 1500(300T -1 - 1) GHz (3.135)

The attenuation due to fog is graphed in Figure 3.13.


Attenuation by dry snow is approximately an order of magnitude lower than
that of rain at the same precipitation rate. A model of propagation through dry
snow was developed by Gunn and East [12], where the attenuation in dB·km−1 is
given in terms of wavelength by

r1.6 r
α = 0.00349 4
+ 0.00224 (3.136)
λ λ

where r is the precipitation rate in mm·hr−1. Equation (3.136) is a valid approxi-


mation for wavelengths greater than 1.5 cm (20 GHz) and precipitation rates less
than 10 mm·hr−1, and is plotted in Figure 3.14. The inclusion of water in dry snow
increases the attenuation, in rough proportion with the amount of water, which can
result in attenuation values up to 40% greater than that of equivalent rainfall [10].
Dust and smoke are atmospheric inclusions that may also be present in some
situations. Multiple studies, the results of which are summarized in [7], have shown,
however, that there is negligible attenuation of waves propagating through these par-
ticulates at frequencies up to 140 GHz. Measureable attenuation was only observed
in cases of explosives dispersing large dirt particulates into the air and only for a short
duration on the order of a few seconds. Thus, dust and smoke can be considered to be
transparent to microwave and millimeter-wave frequencies up to 140 GHz.

11
10
9
K (dB/km)/(g/m3)

8
7
6
5
4
20 oC
3 10 oC
2 0 oC
1 -10 oC
0
0 50 100 150 200
f (GHz)
Figure 3.13.  Theoretical attenuation of fog calculated from (3.129).
3.3  Electromagnetic Propagation in Specific Media 69

102

101

α (dB/km)
100

10-1
10 mm/hr
10-2 5 mm/hr
2.5 mm/hr
1 mm/hr
10-3
20 40 60 80 100
f (GHz)
Figure 3.14  Theoretical attenuation of dry snow calculated from (3.136).

3.3.2.  Propagation Through Building Materials


In applications such as through-wall surveillance or search and rescue, the elec-
tromagnetic waves must travel through various wall materials before reaching the
receiver. In passive systems, the wave passes through the wall once, while in active
systems the wave travels through the wall on transmit as well as before reception,
after the wave reflects off the object of interest. While the attenuation of wall mate-
rials in the millimeter-wave region has been measured, the results must be taken as
general approximations because the specific wall material and physical construction
will significantly affect the attenuation.
Frazier [13] performed measurements of the attenuation of a number of stan-
dard building materials over the frequency range 4 GHz–140 GHz at five discrete
frequencies, the results of which are shown in Figure 3.15. Standard thicknesses
for each media type were measured. Table 3.1 details measurements with paral-
lel and perpendicular incident polarization at 94 GHz, 326 GHz, 584 GHz, and
1042 GHz performed by Gatesman, et al. [14], for a large number of materials,
some with varying thickness. With the exception of spruce-pine-fir (SPF) board,
the difference in attenuation between the two polarization types is very small. It
may also be noted that the attenuation scales roughly with the density of the mate-
rial; concrete blocks significantly attenuate waves, while the attenuation of drywall
is relatively low until the frequency approaches 100 GHz. These drastically dif-
ferent attenuation curves between materials highlight the difficulty in developing
compact, high-resolution through-wall sensors, since only very low frequencies
will penetrate through most walls without significant attenuation. The inclusion
of any metal within the wall will also significantly decrease the amplitude of the
transmitted wave due to additional conductive losses. Analyses of the propagation
effects through specific wall types are generally carried out by developing a model
of the wall based on a layered media formulation as described in the previous
s­ection [15–17].
70 Microwave and Millimeter-Wave Remote Sensing for Security Applications

35
Concrete Block Painted 2X6 Board

Clay Brick
30
Total One Way Attenuation (dB)
25

3/4” Plywood
20

15
3/4”Pine Board

10
Wet Paper Towel
Glass
Drywall
5
Asphalt Shingle
Kevlar Sheet
Polyethylene
Paper Towel (Dry)
0 Fiberglass Insul.
3 5 8 10 20 30 50 80 100 200
Frequency (GHz)
Figure 3.15  Attenuation of common building materials. (© 1997 SPIE [13].)

3.3.3  Propagation Through Clothing and Garment Materials


There has been increased interest in recent years in the transmission properties of
clothing and garment materials due to the increased threat from concealed contra-
band. A number of studies have been performed that measured the attenuation of
radiation propagating through various garment materials, the results of which have
shown that such media generally affect propagating electromagnetic waves at tera-
hertz frequencies and above; microwave and millimeter-wave attenuation of most
garment materials is generally very low or negligible. Bjarnason [18] measured the
attenuation of eight common garment materials over the frequency range 100 GHz
to 1200 GHz, as well as 40 THz to 100 THz. The point at which the transmitted
power was reduced by 3 dB did not occur in any sample until 350 GHz, and in some
thinner samples until 1 THz; the results are summarized in Table 3.2. Figure 3.16
shows the measured attenuation versus frequency for the eight materials measured.
The results show that many clothing materials can be considered effectively trans-
parent up to at least 350 GHz.
Gatesman [14] additionally measured six different garment materials at the
discrete frequencies 94 GHz, 326 GHz, 584 GHz, and 1042 GHz; the attenuation
for parallel and perpendicular incident polarization is shown in Table 3.3. The
materials again did not show a 3 dB reduction in transmitted power until about
350 GHz. The thickest material, 0.084-inch sweater, showed the highest attenuation
of 14.5 dB at 584 GHz, whereas all the material except for the cotton shirt showed
attenuation of 10 dB or more at 1042 GHz. As is the case with building m­aterials,
there is little difference in the attenuation between the two polarizations.
3.3  Electromagnetic Propagation in Specific Media 71

Table 3.1  Attenuation in dB of Common Building Materials (© 2006 SPIE [14]); (n/t = no transmission,
n/m = no measurement)
Material Thickness (in) 94 GHz 326 GHz 584 GHz 1024 GHz
Pol: || ^ || ^ || ^ || ^
Cardboard 0.155 1.2 1.3 2.8 3.2 4.4 5.0 9.0 9.4
Maple 1 0.125 2.6 1.8 8.4 5.3 16.2 11.0 n/m n/m
Maple 2 0.25 5.9 4.0 20.1 16.0 31.4 22.7 65.7 52.4
Maple 3 0.5 10.8 7.1 32.4 22.5 2.6 45.9 n/t n/t
Maple 4 0.762 16.9 10.4 48.0 33.1 n/t 68.4 n/t n/t
Maple 5 0.762 16.6 9.5 46.5 31.0 n/t 62.7 n/t n/t
OSB 1 0.25 6.4 7.1 33.2 33.4 47.8 48.2 n/t n/t
OSB 2 0.5 20.6 18.9 59.4 55.8 n/t n/t n/t n/t
Plywood 1 0.25 5.3 4.5 18.2 16.7 31.3 30.2 n/t 61.9
Plywood 2 0.5 8.7 10.8 30.3 30.4 31.3 30.2 n/t n/t
SPF1 0.125 3.0 1.7 7.2 4.9 14.3 9.3 29.2 20.7
SPF2 0.25 4.8 2.8 14.0 8.5 24.5 15.9 56.5 38.4
SPF3 0.5 10.8 5.8 28.8 19.2 55.9 38.4 n/t n/t
SPF4 0.74 15.8 9.6 42.6 27.3 53.3 72.0 n/t n/t
SPF5 0.985 21.0 12.3 55.3 35.8 n/t 74.0 n/t n/t
SPF6 1.43 30.2 18.7 70.4 54.4 n/t n/t n/t n/t
Concrete 0.438 9.8 10.5 47.7 49.2 n/t n/t n/t n/t
Drywall 0.375 1.6 1.7 10.7 10.5 35.2 35.0 n/t n/t
Drywall 0.5 2.2 2.8 12.8 13.1 49.1 50.4 n/t n/t
Glass 0.087 4.5 4.3 10.8 11.0 25.3 25.4 n/t n/t
Plastic blind 0.033 0.5 0.5 1.3 1.4 3.6 3.4 8.2 8.1
Vinyl siding 0.042 0.7 0.8 2.5 2.3 5.4 5.4 12.6 11.4
Vin. sid. & ½’’ CDX 0.512 10.8 12.6 32.5 32.5 68.1 68.2 n/t n/t
Wall section 4.768 17.9 22.1 69.0 70.6 n/t n/t n/t n/t
Brick 1 0.452 8.7 8.9 62.7 64.3 n/t n/t n/t n/t
Brick 2 0.595 15.1 16.3 n/t n/t n/t n/t n/t n/t
Brick 3 0.252 5.4 6.1 39.9 40.3 n/t n/t n/t n/t
Brick 4 0.206 5.9 5.7 n/m n/m n/t n/t n/t n/t
Brick 9 0.338 7.1 6.7 46.9 50.0 n/t n/t n/t n/t
Cinder block 1 1.091 45.9 48.3 n/m n/m n/t n/t n/t n/t
Cinder block 5 0.331 16.5 16.5 53.1 53.1 52.8 56.8 n/t n/t
Cinder block 7 0.385 17.7 17.4 54.1 52.7 62.3 60.6 n/t n/t
Cinder block 9 0.67 26.6 26.9 74.7 75.6 n/t n/t n/t n/t

3.3.4  Dielectric Properties of Explosives, Plastics, and Metals


Detection of explosive material concealed on a person can be facilitated by measur-
ing the difference between the signature of the person and that of the contraband.
An image formed by a passive radiometric system would show a radiometric profile
of the person with the outline of the contraband overlaid on top. Most explosive
materials have relative permittivity values of 2.70–3.14 and loss tangents of ap-
proximately 0.001 [19]. Detecting specific explosive materials can be done by mea-
suring the spectral characteristics of the media (e.g., interrogating the material at a
known resonance frequency and measuring the amplitude of the return); however,
the absorption frequencies of most explosives occur in the terahertz region, gener-
ally above 700 GHz [19]. Plastic materials also have low relative p­ermittivity values
72 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Table 3.2  Measured Attenuation (in dB) of Common Garment Materials


Fabric Thickness Density −3 dB point Attenuation at εr (100 kHz)
(mm) (kg/m3) (THz) 1 THz (dB)
Wool 2.2 214 0.35 11.0 1.6
Linen 1.1 509 0.35 8.0 2.9
Leather 0.75 813 0.40 10.0 5.1
Denim 0.96 490 0.50 6.5 2.3
Naugahyde 0.65 800 0.70 5.5 2.6
Silk 0.36 256 1.0 3.0 1.4
Nylon 0.19 379 1.0 3.0 1.9
Rayon 0.15 733 >1.0 2.5 1.9
(Reprinted with permission from [18], Copyright 2004, American Institute of Physics.)

in the range of 2.08–5.04 [20]. Table 3.4 is a compilation of the relative dielectric
constants of some explosive and plastic materials.
Metals, because of their high conductivity (1×106 S·m−1 or higher for most met-
als), are lossy and have low skin depths. The imaginary part of the complex per-
mittivity is proportional to the conductivity and is also a large quantity for metals;
thus, when coupled with air or other media with low permittivity, metal materials
will have a very different impedance and will thus be highly reflective. Detection
of metal objects can thus be accomplished by measuring the reflection of incident
waves off the object; the reflection from a human concealing a metal object will
generally be very low relative to the return from the metal object.

3.3.5  Dielectric Properties of Human Tissue


Many applications of remote security sensing involve the interaction of electro-
magnetic waves with human tissues, either from the standpoint of the detection

0
Transmission (dB)

-10

rayon
nylon
silk
-20 naugahyde
denim
leather
linen
wool
-30
0 0.6 1.2 40 70 100
f (THz)
Figure 3.16  Millimeter-wave and terahertz attenuation through the materials in Table 3.2 versus
frequency. (Reprinted with permission from [18], Copyright 2004, American Institute of Physics.)
3.3  Electromagnetic Propagation in Specific Media 73

Table 3.3  Measured Attenuation (in dB) of Common Garment Materials (ã 2006 SPIE [14])
Material Thickness 94 GHz 326 GHz 584 GHz 1024 GHz
(in)
Pol: || ^ || ^ || ^ || ^
Cotton Shirt 0.012 0.2 0.1 0.3 0.5 1.0 1.1 3.1 3.2
Denim 0.025 0.7 0.7 1.3 1.4 3.4 2.9 10.0 7.9
Drapery 0.035 0.3 0.5 3.0 1.7 7.5 7.6 12.3 11.4
Leather 0.051 0.7 0.6 2.3 2.1 6.0 5.2 17.9 15.3
Sweater 0.084 0.4 0.4 3.8 4.0 14.5 13.7 19.1 21.4
Sweatshirt 0.082 0.3 0.2 0.8 1.1 4.3 3.8 14.3 13.9

of human presence or the detection of contraband hidden on a person’s body. It is


thus important to understand the dielectric properties of human tissues in order
to predict the operation of various security sensors. The human body consists of
multiple layers of different tissues, some with very different dielectric properties.
An incident electromagnetic wave first contacts the skin, beneath which are vari-
ous layers of fat, cartilage, muscle, and bone, among other tissues. As will be seen
later in this section, the skin depth, or penetration depth, of human skin becomes
smaller than the average human skin thickness at millimeter-wave frequencies and
above. Thus, for sensors operating at higher frequencies, the skin absorbs most
of the incident energy, with very little penetrating into the deeper layers of tis-
sue. A model of the human body at higher millimeter-wave frequencies can then be
approximated in terms of the properties of the skin only, whereas lower millimeter-
­wave frequencies and microwave frequencies must consider human tissue as a layered
media.
A simple model of human tissue layers, illustrated in Figure 3.17, consists of a
layer of skin, comprised of the epidermis and dermis, followed by the hypodermis,
which is a layer of subcutaneous fat, underneath which may be muscle, bone, or
cartilage, depending on the specific location on the body [21]. The thickness of
the epidermis varies from 0.5–1.5 mm, the thickness of the dermis varies from

Table 3.4  Relative Permittivity of Some Materials (Compiled from [19, 20])
Material Frequency εr
Ceramic 3 GHz 5.60
Comp B (explosive) 1 GHz 2.90
Comp C-4 (explosive) 1 GHz 3.14
Glass (Pyrex) 3 GHz 4.82
Lucite 10 GHz 2.56
PETN (explosive) 1 GHz 2.72
Plexiglass 3 GHz 2.60
Polyethylene 10 GHz 2.25
Polystyrene 10 GHz 2.54
RDX 1 GHz 3.14
Styrofoam 3 GHz 1.03
Teflon 10 GHz 2.08
TNT (explosive) 1 GHz 2.70
74 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Epidermis (0.5-1.5 mm)


Skin
Dermis (0.3-3 mm)

Hypodermis (3-10 mm) Subcutaneous fat

Other tissues
Figure 3.17  Model of human tissue layers.

0.3–3mm, and the thickness of the hypodermis can vary from 3–10 mm or more
[22]. A layered model for a specific location on the body is dependent on the other
subcutaneous layers (such as muscle, bone, and cartilage) that are present at the
location, as well as the thicknesses of each layer.
As seen in the previous sections, the majority of the dielectric media parameters
of interest, such as skin depth, impedance, and attenuation, are dependent on the
constitutive parameters. Human tissues, like most materials, are nonmagnetic and
have permeabilities approximately equal to that of free space; thus, the constitutive
parameter of interest is the complex permittivity [recall that the conductivity can
be derived from the imaginary part of the permittivity by (3.43)]. Gabriel, et al.
[23–25], developed a model for the complex permittivity of various tissues over the
frequency range of 10 Hz to 100 GHz based on measured data from 23 data sets.
The model was developed for electromagnetic dosimetry research in tissues, and
thus includes models for a large number of tissues found within the human body.
Based on a general Debye dispersion model, the permittivity was fit to the measured
data sets using a number of parameters. The focus of the study by Gabriel, et al.,
was human tissue, although some measurements from animal tissue were consid-
ered in the development of the model; however, various measurements of the rela-
tive permittivity and derived parameters of human tissue have verified the model
since its publication [26–31]. As noted in [32], some variability of measured results
from the model resulted from differing measurement techniques and measurements
on different parts of the body.
The expression for the permittivity of tissues in the model developed by Gabriel
is given as a summation of Debye dispersion regions by
N
Dε n σ
ε (ω ) = ε ¥ + å (1- an )
+ 0 , (3.137)
n =1 1 + (jωτn ) jωε0

where ε¥ is the permittivity as the frequency approaches infinity, Dεn describes the
magnitude of dispersion region n, N is the total number of dispersion regions, τn
is a time constant related to the dispersion region, an is a measure of the broaden-
ing of the dispersion, and σ0 is the dc conductivity. The parameters for the Debye
model for blood, bone, cartilage, fat, muscle, dry skin, and wet skin are given
Table 3.5  Parameters for the Debye Model of the Permittivity of Various Tissues (© 1996 IOP [25])
Tissue ε¥ Dε1 τ1 (ps) a1 Dε2 τ2 (ns) a2 Dε3 τ3 (μs) a3 Dε4 τ4 (ms) a4 σ0
Blood 4.0 56.0 8.38 0.10 5200 132.63 0.10 0.0 - - 0.0 - - 0.7000
Bone 2.5 18.0 13.26 0.22 300 79.58 0.25 2.0×104 159.15 0.20 2×107 15.915 0.00 0.0700
3.3  Electromagnetic Propagation in Specific Media

Cartilage 2.5 41.1 12.8 0.27 0.0 - - 0.0 - - 0.0 - - 0.5800


Fat 2.5 3.0 7.96 0.20 15 15.92 0.10 3.3×104 159.15 0.05 1×107 7.958 0.01 0.0100
Muscle 4.0 50.0 7.23 0.10 7000 353.68 0.10 1.2×106 318.31 0.10 2.5×107 2.274 0.00 0.2000
Skin (dry) 4.0 32.0 7.23 0.00 1100 32.48 0.20 0.0 - - 0.0 - - 0.0002
Skin (wet) 4.0 39.0 7.96 0.10 280 79.58 0.00 3×104 1.59 0.16 3×104 1.592 0.20 0.0004
75
76 Microwave and Millimeter-Wave Remote Sensing for Security Applications

60
bone cartilage
fat muscle
50 dry skin wet skin

40

30
r
ε
20

10

0
0 20 40 60 80 100
f (GHz)
Figure 3.18  Relative permittivity of human tissues.

in Table 3.5. Gabriel, et al., include parameters for other tissues as well, such as
various internal organ tissues; however, those listed here are of primary interest in
security sensing as the waves incident on the human body primarily interact with
the top-most layers and not specific internal organs. Equation (3.137) has been
implemented in an online program by the Italian National Research Council that
generates the permittivity and a few derived parameters for the tissues outlined by
Gabriel, et al. [33].
Figure 3.18 shows the real part of the relative permittivity of cortical bone, car-
tilage, fat, muscle, dry skin, and wet skin over the frequency range 1 GHz to 100
GHz, generated from the model in [25]. While most of the tissues exhibit similar
dispersive behavior, fat and bone tissues have overall much lower relative permittivi-
ties. The conductivity, derived from the imaginary part of the permittivity, is shown

70
bone cartilage
60 fat muscle
dry skin wet skin
50

40
σ (S/m)

30

20

10

0
0 20 40 60 80 100
f (GHz)
Figure 3.19  Conductivity of human tissues.
3.3  Electromagnetic Propagation in Specific Media 77

3500
bone cartilage
3000 fat muscle
dry skin wet skin
2500

α (Np/m)
2000

1500

1000

500

0
0 20 40 60 80 100
f (GHz)
Figure 3.20  Attenuation constant of human tissues.

in F­igure 3.19. Muscle shows the highest conductivity, while fat and bone have the
lowest. The conductivity of skin increases when the skin is wet, as may be expected
from the additional conductivity of water.
With the relative permittivity and the conductivity (with μr = 1), many other
dielectric quantities of interest can be derived directly. Figures 3.20–3.26 show the
attenuation constant calculated from (3.12), the phase constant from (3.13), the
skin depth from (3.18), the loss tangent from (3.19), the phase velocity from (3.20),
the wavelength from (3.21), and the impedance from (3.36), respectively. Figures
3.27 and 3.28 show the reflection and transmission coefficients, calculated from
(3.72) and (3.73) for normal incidence with respect to free space.

8000
bone cartilage
7000 fat muscle
dry skin wet skin
6000
5000
β (rad/m)

4000
3000
2000
1000
0
0 20 40 60 80 100
f (GHz)
Figure 3.21  Phase constant of human tissues.
78 Microwave and Millimeter-Wave Remote Sensing for Security Applications

50
bone cartilage
fat muscle
40 dry skin wet skin

30

δ (mm) 20

10

0
0 20 40 60 80 100
f (GHz)
Figure 3.22  Skin depth of human tissues.

bone cartilage
fat muscle
1.5 dry skin wet skin

1
tan δ

0.5

0
0 20 40 60 80 100
f (GHz)
Figure 3.23  Loss tangent of human tissues.
7
x 10
18
16
14
12
vp (m/s)

10
8
6
4 bone cartilage
fat muscle
2
dry skin wet skin
0
0 20 40 60 80 100
f (GHz)
Figure 3.24  Phase velocity of waves in human tissues.
3.3  Electromagnetic Propagation in Specific Media 79

140
bone cartilage
120 fat muscle
dry skin wet skin
100

80

λ (mm)
60

40

20

0
0 20 40 60 80 100
f (GHz)
(a)
10
bone cartilage
fat muscle
8 dry skin wet skin

6
λ (mm)

0
0 20 40 60 80 100
f (GHz)
(b)
Figure 3.25  (a) Wavelength of waves in human tissues. (b) Close view of the wavelength.

250

200

150
η (Ω)

100

50 bone cartilage
fat muscle
dry skin wet skin
0
0 20 40 60 80 100
f (GHz)
Figure 3.26  Impedance of human tissues.
80 Microwave and Millimeter-Wave Remote Sensing for Security Applications

1
bone cartilage
fat muscle
0.8 dry skin wet skin

0.6

|R| 0.4

0.2

0
0 20 40 60 80 100
f (GHz)
Figure 3.27  Reflection coefficient of human tissues with respect to free space.

A few general observations can be made from the graphs of the various parame­
ters. Most tissues exhibit similar dispersion properties, with fat and bone tissue
being the notable outliers; because of the low relative permittivity and conductivity
relative to the other tissues, fat and cortical bone are better dielectrics than the other
tissues and as a result have higher impedances. The impedance of skin is signifi-
cantly different from both air and fat; thus, the reflection coefficients at the air-skin
boundary and the skin-fat boundary will be relatively high. Much of the energy in
an incident wave will therefore reflect within the skin layer, thereby absorbing much
of the power carried by the wave.
It is also notable that the skin depth of both dry and wet skin is less than the
average maximum skin thickness of 4.5 mm above approximately 9 GHz; Figure
3.29 shows the skin depth focused on the higher frequencies. Because of the low

0.8

0.6
|T |

0.4

0.2 bone cartilage


fat muscle
dry skin wet skin
0
0 20 40 60 80 100
f (GHz)
Figure 3.28  Transmission coefficient of human tissues with respect to free space.
3.3  Electromagnetic Propagation in Specific Media 81

0.8

0.6

δ (mm)
0.4

0.2
cartilage muscle
dry skin wet skin
0
0 20 40 60 80 100
f (GHz)
Figure 3.29  Skin depth at millimeter-wave frequencies. Note that fat and bone tissues have skin
depths greater than 1 mm and are not pictured.

skin depth, the energy of a wave incident on the skin at frequencies above 9 GHz
is mostly attenuated before the wave encounters the skin-fat boundary. At high mi-
crowave and millimeter-wave frequencies, a reasonable approximation of layered
tissue would consist primarily of the skin only, or perhaps with the subcutaneous
fat layer.

References

[1] Rothwell, E. J., and M. J. Cloud, Electromagnetics, Boca Raton, FL: CRC Press, 2001.
[2] Jackson, J. D., Classical Electrodynamics, 3rd ed., Hoboken, NJ: John Wiley & Sons,
1999.
[3] Balanis, C. A., Advanced Engineering Electromagnetics, Hoboken, NJ: John Wiley & Sons,
1989.
[4] Ulaby, F. T., Fundamentals of Applied Electromagnetics, Upper Saddle River, NJ: Prentice
Hall, 2001.
[5] Ulaby, F. T., R. K. Moore, and A. K. Fung, Microwave Remote Sensing, Vol. I: Microwave
Remote Sensing Fundamentals and Radiometry, Reading, MA: Addison-Wesley, 1981.
[6] Kong, J. A., Electromagnetic Wave Theory, New York, NY: John Wiley & Sons, 1986.
[7] Currie, N. C., and C. E. Brown, Principles and Applications of Millimeter-Wave Radar,
Norwood, MA: Artech House, 1987.
[8] Liebe, H. J., G. A. Hufford, and M. G. Cotton, “Propagation Modeling of Moist Air and
Suspended Water/Ice Particles at Frequencies Below 1000 GHz,” in Proc. NATO/AGARD
Wave Propagation Panel, 52nd Meeting, 1993, pp. 1–10.
[9] McMillan, R. W., “Terahertz Imaging, Millimeter-Wave Radar,” in Advances in Sensing
with Security Applications, J. Burnes, Ed., Cordrecht, the Netherlands: Springer, 2006.
[10] Oguchi, T., “Electromagnetic Wave Propagation and Scattering in Rain and Other Hydro-
meteors,” Proceedings of the IEEE, Vol. 71, 1983, pp. 1029–1078.
82 Microwave and Millimeter-Wave Remote Sensing for Security Applications

[11] “Attenuation due to Clouds and Fog,” International Telecommunication Union Radiocom-
munication Assembly Recommendation ITU-R P.840-3, 1999.
[12] Gunn, K. L. S., and T. W. R. East, “The Microwave Properties of Precipitation Particles,”
Quarterly Journal of the Royal Meteorological Society, Vol. 80, 1954, pp. 522–545.
[13] Frazier, L., “Radar Surveillance Through Solid Materials,” Proc. SPIE, Vol. 2938, 1997,
p. 139.
[14] Gatesman, A., “Terahertz Behavior of Optical Components and Common Materials,”
Proc. SPIE, Vol. 6212, 2006, p. 62120E.
[15] Soldovieri, F., R. Solimene, A. Brancaccio, and R. Pierri, “Localization of the Interfaces of
a Slab Hidden Behind a Wall,” Geoscience and Remote Sensing, IEEE Transactions on,
Vol. 45, 2007, pp. 2471–2482.
[16] Solimene, R., F. Soldovieri, G. Prisco, and R. Pierri, “Three-Dimensional Through-Wall
Imaging Under Ambiguous Wall Parameters,” Geoscience and Remote Sensing, IEEE
Transactions on, Vol. 47, 2009, pp. 1310–1317.
[17] Dehmollaian, M., and K. Sarabandi, “Refocusing Through Building Walls Using Synthetic
Aperture Radar,” Geoscience and Remote Sensing, IEEE Transactions on, Vol. 46, 2008,
pp. 1589–1599.
[18] Bjarnason, J. E., T. L. J. Chan, A. W. M. Lee, M. A. Celis, and E. R. Brown, “Millimeter-
Wave, Terahertz, and Mid-Infrared Transmission Through Common Clothing,” Applied
Physics Letters, Vol. 85, 2004, pp. 519–521.
[19] Daniels, D. J., EM Detection of Concealed Targets, Hoboken, NJ: John Wiley & Sons,
2009.
[20] Pozar, D. M., Microwave Engineering, 3rd ed., New York, NY: John Wiley & Sons,
2005.
[21] McGrath, J. A., R. A. J. Eady, and F. M. Pope, “Anatomy and Organization of Human
Skin,” in Rook’s Textbook of Dermatology, Vol. 1, D. A. Burns, S. M. Breathnach, N. H.
Cox, and C. E. M. Griffiths, Eds., 7 ed., Malden, MA: Blackwell, 2010.
[22] Black, D., J. Vora, M. Hayward, and R. Marks, “Measurement of Subcutaneous Fat
Thickness with High Frequency Pulsed Ultrasound: Comparisons with a Caliper and a
Radiographic Technique,” Clinical Physics and Physiological Measurement, Vol. 9, 1988,
pp. 57–64.
[23] Gabriel, C., S. Gabriel, and E. Corthout, “The Dielectric Properties of Biological Tissues:
I. Literature Survey,” Physics in Medicine and Biology, Vol. 41, 1996, p. 2231.
[24] Gabriel, S., R. W. Lau, and C. Gabriel, “The Dielectric Properties of Biological Tissues: II.
Measurements in the Frequency Range 10 Hz to 20 GHz,” Physics in Medicine and Biol-
ogy, Vol. 41, 1996, p. 2251.
[25] Gabriel, S., R. W. Lau, and C. Gabriel, “The Dielectric Properties of Biological Tissues: III.
Parametric Models for the Dielectric Spectrum of Tissues,” Physics in Medicine and Biol-
ogy, Vol. 41, 199, p. 22716.
[26] Gustrau, F., and A. Bahr, “Biological Effects in the cm/mm Wave Range Part II/III Determi-
nation of Material Parameters and Analysis of Field Strengths in Human Tissue,” Institute
of Mobile and Satellite Communication Techniques, Germany, 1998.
[27] Alabaster, C. M., “Permittivity of Human Skin In Millimetre Wave Band,” Electronics
L­etters, Vol. 39, 2003, pp. 1521–1522.
[28] Alekseev, S. I., A. A. Radzievsky, M. K. Logani, and M. C. Ziskin, “MillimeterWave
D­osimetry of Human Skin,” Bioelectromagnetics, Vol. 29, 2008, pp. 65–70.
[29] Ito, H., and H. Yamamoto, “Millimeter-/Terahertz-Wave Measurements for Biological Ma-
terials Using Photonically Generated Continuous Waves,” in Proceedings of the XXXth
URSI General Assembly, Istanbul, 2011.
[30] Dallinger, A., S. Schelkshorn, and J. Detlefsen, “Short Distance Related Security Millimeter-
Wave Imaging Systems,” in German Microwave Conference, Ulm, Germany, 2005.
3.3  Electromagnetic Propagation in Specific Media 83

[31] Hyeonseok, H., Y. Jounghwa, C. Jei-Won, C. Changyul, and K. Youngwoo, “110 GHz
Broadband Measurement of Permittivity on Human Epidermis Using 1 mm Coaxial
Probe,” in Microwave Symposium Digest, 2003 IEEE MTT-S International, Vol. 1 2003,
pp. 399–402.
[32] Chahat, N., M. Zhadobov, R. Augustine, and R. Sauleau, “Human Skin Permittivity Mod-
els for Millimetre-Wave Range,” Electronics Letters, Vol. 47, 2011, pp. 427–428.
[33] “An Internet Resource for the Calculation of the Dielectric Properties of Human Tissues
in the Frequency Range 10Hz–100 GHz,” the Italian National Research Council, http://­
niremf.ifac.cnr.it/tissprop/. Last accessed March 2012.
Chapter 4

Antennas

An antenna may be defined as the region of transition between a wave propagating


in free space and a wave propagating in a guiding medium such as a transmission
line. When used in a receiver, the antenna collects the incident electromagnetic
energy and couples it into the receiving system, where it can be amplified and pro-
cessed. In a transmitter, the antenna serves to radiate energy, often focused in a spec-
ified direction. Radiometer systems use antennas to receive thermal electromagnetic
radiation emitted by a person, scene, or object of interest. Antennas in radar sys-
tems are used for both transmission of energy and reception of the energy scattered
off the person or object back toward the sensor. A single antenna may be used for
both the transmission and reception (monostatic) or multiple antennas may be used
independently, each for transmission or reception only (bistatic or multistatic).
The performance of an antenna can be characterized in terms of the spatial distri-
bution of the power radiated by an antenna in the transmit mode; this distribution is
referred to as the antenna pattern, the radiation pattern, or the power pattern, as shown
in Figure 4.1. Antennas are in general reciprocal devices: the antenna pattern describes
both the radiation pattern for the antenna when it transmits and the power reception
pattern of the antenna when it is used to receive incident radiation; the radiation pat-
tern and reception pattern are identical. Exceptions to the reciprocal rule exist and are
antennas that generally exhibit nonlinear behavior. Such antennas are not typically used
in remote sensing and will not be considered in this book. Thus, the antennas described
will be considered to be reciprocal, and no distinction will be made between transmit-
ting and receiving antennas as the results for one case apply equally to the other.
The analysis of the radiation properties of an antenna derives from the currents
on the antenna aperture. Time-varying currents give rise to electric and magnetic
fields, which can be used to calculate the power density at a point in space away
from the antenna itself. The currents on the antenna can be induced by a driving
signal at the antenna terminals in the case of a transmitting antenna or they may
be induced on the antenna by impinging waves in the case of a receiving antenna.
The currents on the antenna may be either current densities on the physical antenna
itself such as a linear wire dipole or they may be equivalent current densities in the
opening of an aperture antenna such as a horn antenna.
This chapter begins by deriving a standard method for calculating the electro-
magnetic fields generated by time-varying sources on the antenna through the use of
the electromagnetic potentials. This method calculates the potentials from the current
density, and then calculates the fields from the potentials. Although incorporating
an additional step, use of the potentials generally results in simpler calculations than
deriving the fields directly from the currents. Following this, general properties of an-
tennas are described, most of which are derived from the fields that are distant from
the antenna. An example of radiation from a current density is given in a discussion

85
86 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Figure 4.1  Antenna pattern of a rectangular aperture antenna.

of a long linear dipole antenna. Discussion of linear wire antennas is necessary for
the understanding of antenna systems; however, in microwave and millimeter-wave
remote sensing, aperture antennas are most often used because greater directivity can
be achieved using a two-dimensional aperture such as a horn or a metal patch than
can be achieved with a linear wire. Such apertures are also of a more manageable size
at higher frequencies where the wavelengths are shorter. Thus, this chapter covers the
theory of equivalent currents in an aperture and the analysis of the resulting fields.
Following this, antenna array theory is covered. Finally, the chapter concludes with a
description of typical antennas used in remote sensing applications.

4.1  Electromagnetic Potentials

Propagating electromagnetic waves are generated by time-varying sources. In the pre-


vious chapters, it was assumed that the waves had travelled a distance far from the
sources and thus were analyzed in free space, with no charge or current present. In the
description of the radiation from an antenna, it is the currents on the antenna structure
that generate the radiated waves, and thus the solution of the electric and magnetic fields
must be found with sources present. The fields can be derived directly from the sources
using Maxwell’s equations with nonzero current density; however, such calculations
are generally very difficult except in the most simplistic cases. An easier, if less direct,
way of calculating the fields due to the currents is through the use of the electromagnetic
potentials. The potentials are mathematical constructs that serve as an intermediate
quantity in the calculation of the fields and are calculated through integration of the
current density. The fields are then found through differentiation of the potentials.

4.1.1  Electromagnetic Potentials due to Electric Current Density J


The potentials due to the electric sources (with no magnetic sources present) are
found through Maxwell’s equations with nonzero sources, described in Chapter 2
and repeated here.

Ñ ´ E = -jωµH (4.1)

Ñ ´ H = jωε E + J (4.2)

Ñ × D = ρ (4.3)

∇⋅B = 0 (4.4)
4.1  Electromagnetic Potentials 87

In terms of the constitutive relations for linear, homogeneous, isotropic media,


the magnetic Gauss’s law can also be written

Ñ × H = 0 (4.5)

Because of the vector identity Ñ × ( Ñ ´ Y ) = 0 , where Y is an arbitrary vector, the


magnetic field can be defined by
H = Ñ ´ A (4.6)

such that (4.5) is still satisfied. A (Wb·m–1) is called the magnetic vector potential, and
so far it is not uniquely specified because only its curl is defined, and its divergence is
undefined. Using this definition of the magnetic vector potential, Faraday’s law is

Ñ ´ E = - jωµÑ ´ A (4.7)
or
Ñ ´ (E + jωµA) = 0 (4.8)

Now, using the identity Ñ ´ Ñψ = 0, where y is an arbitrary scalar, the quantity in


the brackets in (4.8) can be defined as

E + jωµA = -Ñφ (4.9)


such that (4.8) is still satisfied. The scalar quantity f (V) is the electric scalar poten­
tial. Using (4.6) and (4.9), Ampere’s law can then be written

Ñ ´ Ñ ´ A = ω 2 µεA - jωεÑφ + J (4.10)

or, using the vector identity Ñ ´ Ñ ´ Y = Ñ(Ñ × Y) - Ñ 2 Y

Ñ 2 A + ω 2µε A = - J + jωεÑ φ+ Ñ(Ñ × A) (4.11)

The left-hand side of (4.11) is the same as that of the homogeneous vector
Helmholtz equation that was solved in Chapter 2; the right-hand side now contains
the electric current density as well as the magnetic vector potential and electric sca-
lar potential. To this point, only the curl of the magnetic vector potential has been
defined. Because it is otherwise unspecified, the divergence of A can be arbitrarily
chosen, which then results in a unique definition of A. A common definition of the
divergence of the potential is
Ñ × A = - jωεφ (4.12)

which is called the Lorenz gauge. Other definitions, or gauges, have been proposed;
however, the Lorenz gauge is the most useful in the present discussion. Under this
gauge, (4.11) becomes

 Some authors define the curl of the magnetic vector potential as B = Ñ ´ A. This differs from the definition
in this book by a factor of m.
 This is commonly referred to as the Lorentz gauge; however, as noted in [1], this definition of the divergence
of the magnetic vector potential was first proposed by the physicist L. V. Lorenz, not the well-known physi-
cist H. A. Lorentz.
88 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Ñ 2 A + ω 2µε A = - J (4.13)
which is an inhomogeneous vector Helmholtz equation and is very similar to the
vector Helmholtz equation analyzed in Chapter 2, though the current density is
now explicitly included.
The solutions of (4.13) are called Green’s functions; the particular solution in
free space is given by [1]
- jk r - r¢
e
g(r, r¢) = (4.14)
4π r - r¢
The free space Green’s function can be described as the spatial impulse response at a
position in space r to a unit point source at r¢. In terms of (4.14), the magnetic vector
potential is given by integrating over the volume occupied by the current density:
- jk r - r¢
e
A(r) = ò J(r¢)g(r, r¢)dV = ò J(r¢) dV (4.15)
V V
4π r - r¢
The magnetic vector potential is thus found at a point in space in terms of the su-
perposition of current density point sources over the total volume occupied by the
current density.
As noted earlier, the potentials are mathematical constructs and do not by
themselves describe the properties of the radiation. The second part of the calcula-
tion is then to find the electric and magnetic fields, from which the radiated power
density can be calculated. The electric field can be found in terms of the magnetic
vector potential by substituting (4.12) into (4.9), which yields
1
E = - jωµA + Ñ(Ñ × A) (4.16)
jωε
The magnetic field is found through (4.6).

4.1.2  Electromagnetic Potentials due to Magnetic Current Density Jm


Radiation from physical current sources on antennas is found in terms of the electric
current density. Despite the lack of physical evidence supporting the existence of mag-
netic currents, they prove useful in the determination of radiation from aperture, as will
be discussed in a later section. In particular, the electric fields within an aperture can be
replaced by an equivalent current density that manifests as magnetic current density.
In an analogous way as was done in defining the magnetic vector potential from
the magnetic Gauss’s law, the electric Gauss’s law in the absence of electric charge
can be used to define a similar vector potential. When only magnetic sources are
present, Gauss’s law is, using the constitutive relations,

Ñ × E = 0 (4.17)
The electric field can then be defined in terms of the electric vector potential
F (C·m–1) by

 Some authors define the curl of the electric vector potential as D = –Ñ ´ F. This differs from the definition
in this book by a factor of e.
4.1  Electromagnetic Potentials 89

E = -Ñ ´ F (4.18)
Similarly as was done for the electric current density, the resulting inhomoge-
neous vector Helmholtz equation is

Ñ 2 F + ω 2µε F = - Jm (4.19)

where the divergence of the electric vector potential has been defined as

Ñ × F = - jωµφm (4.20)

where fm (A) is the magnetic scalar potential. The solution of (4.19) is given in
terms of the free space Green’s function by
- jk r - r¢
e
F(r) = ò Jm (r¢) dV (4.21)
V
4π r - r¢

The electric field is then found from (4.18), and the magnetic field is found by

1
H = - jωε F + Ñ(Ñ × F) (4.22)
jωµ
If both electric and magnetic current densities are present, the total fields may
be found by superposition of the fields due to the electric and magnetic sources
separately. The total fields are thus

1
Etotal = E electric + E magnetic = -Ñ ´ F - jωµ A + Ñ(Ñ × A) (4.23)
jωε

1
Htotal = H electric + H magnetic = Ñ ´ A - jωε F + Ñ(Ñ × F) (4.24)
jωµ

4.1.3  Infinitesimal Dipole Radiation


A useful example of the results of the previous section is the infinitesimal dipole
antenna, shown in Figure 4.2, which is an antenna with length dl much less than the
wavelength and assumed to have a diameter of zero. The current on an infinitesimal
dipole located at the origin is

I = zˆ Iδ (x¢)δ (y¢)δ (z¢) (4.25)


where d(x) is the Dirac delta function. The current density is the current integrated
over the antenna,

J = zˆ Idlδ (x¢)δ (y¢)δ (z¢) (4.26)


The magnetic vector potential is thus

- jk r - r¢
e e - jkr
A(r) = zˆ ò J(r¢) dV ¢ = zˆ Idl (4.27)

4π r - r¢ 4π r
90 Microwave and Millimeter-Wave Remote Sensing for Security Applications

x
(a)

dl
2

y
- dl
2

x
(b)
Figure 4.2  (a) Coordinate system. (b) Infinitesimal dipole antenna.

The magnetic field can then be calculated from (4.6). In spherical coordinates, the
result is [2]
e - jkr æ 1 ö
H = φˆ jkIdl ç1 + ÷ sinθ (4.28)
4π r è jkr ø

while the electric field, calculated by (4.16), is

e - jkr æ 1 ö ˆ e - jkr æ 1 1 ö
E = rˆ η Idl ç 1 + ÷ cos θ + θ jkηIdl ç1 + - 2 2 ÷ sinθ (4.29)
2π r 2 è jkr ø 4π r è jkr k r ø
Note that the electric and magnetic fields are orthogonal, as was the case of
plane wave propagation.

4.1.4  Far Field Radiation


The radiation fields generated by the infinitesimal dipole antenna, (4.28) and
(4.29), include terms with varying degrees of dependence on the distance from the
4.1  Electromagnetic Potentials 91

antenna r. All components are inversely proportional to some power of the distance.
Such a result is rather intuitive; as the radiated waves propagate away from the an-
tenna, the wavefronts spread over a larger area. In a passive medium that does not
inject power into the propagating wave, the field strength per unit area must therefore
decrease because of conservation of energy. The fields include three distinct compo-
nents. The first is dependent on r –3 and is present only in the electric field; this compo-
nent of the field is referred to as the static dipole field. Both the electric and magnetic
fields include components that are dependent on r –2, which are the induction field
components. Finally, the electric and magnetic fields contain components dependent
on r –1, called the radiation fields. In remote sensing applications, the radiated waves
travel over distances that are long compared to the wavelength. At such distances, the
static dipole and induction fields decay in amplitude more quickly than the radiation
fields, rapidly decreasing to negligible levels. What are left are the radiation fields,
which travel the farthest distance; these are the fields of interest in remote sensing.
There are three general regions of distance from an antenna, defined in terms
of the wavelength and the maximum antenna dimension d, which roughly define
which of the three field components is dominant over the others in that region [2].
The reactive near field region is defined by the range

d3
r < 0.62 (4.30)
λ
and is the range where the static dipole fields dominate. The near field region, also
called the Fresnel region, is defined by
d3 2d 2
0.62 <r< (4.31)
λ λ
where the induction fields dominate. The far field region, also called the Fraunhofer
region, is defined as
2d 2
r> (4.32)
λ
where the radiation fields dominate.
The far field region is the region of primary interest in remote security sensing
applications, particularly at microwave and millimeter-wave frequencies, where the
wavelengths and antenna apertures are small such that the distance 2d2/l is small.
Whether detecting hidden contraband or moving intruders, the object of interest is
typically in the far field. Thus, the rest of this chapter will focus on the analysis of
radiated fields in the far field.
Consider the geometry of Figure 4.3, where the observation point p1 is signifi-
cantly farther from the origin than the antenna location p2. Thus, r >> r' and

r - r¢ » r - r¢ × r
ˆ (4.33)

The free space Green’s function can then be written


e - jkr e jkr¢×rˆ
g(r, r¢) » (4.34)
4π (r - r¢ × rˆ )
92 Microwave and Millimeter-Wave Remote Sensing for Security Applications

z p1
|r-r`|
p2
r
r`

x
Figure 4.3  Geometry for the calculation of the far fields.

The term in the denominator affects only the amplitude and can further be ap-
proximated by
1 1 r¢ × rˆ 1 (4.35)
» (1 + )»
r - r¢ × rˆ r r r

since r is large. The Green’s function is then given by

g(r, r¢) » g(r)e jkr¢×rˆ (4.36)

where
e - jkr
g(r) = (4.37)
4π r
The magnetic vector potential can then be given by

e - jkr
J(r¢)e jkr¢×rˆ dV ¢ = g(r) ò J(r¢)e jkr¢×rˆ dV ¢
4π r Vò¢
A(r) = (4.38)

The term in the exponential can be rewritten using

r¢ = x¢xˆ + y¢yˆ + z¢z


ˆ (4.39)
and

rˆ = sin θ cosφ xˆ + sin θ sin φ yˆ + cosθ z


ˆ (4.40)

which yields

kr¢ × rˆ = k sin θ cosφ x¢ + k sin θ sinφ y¢ + k cosθ z¢ = kx x¢ + ky y¢ + kz z¢ (4.41)

The magnetic vector potential can then be written

j(kx x¢+ ky y¢+kz z¢)


A(r) = g(r) ò ò ò J(x¢, y¢, z¢)e dx¢dy¢dz¢ (4.42)
x¢ y¢ z¢
4.1  Electromagnetic Potentials 93

The integral is simply the three-dimensional Fourier transform of the current


density,

�J(kx , ky , kz ) = j(kx x¢+ ky y¢+kz z¢)


ò ò ò J(x¢, y¢, z¢)e dx¢dy¢dz¢ (4.43)
x¢ y¢ z¢

and thus
A(r) = g(r)�J(kx , ky , kz ) (4.44)

The magnetic vector potential is thus found directly in terms of the Fourier trans-
form of the current density.
Although the Fourier transform of the current density is a function of the three
wavenumbers kx, ky, and kz, the wavenumbers, as defined in (4.41), are functions
only of the two spherical coordinates q and f, while the Green’s function accounts
for the radial dependence. Thus, the vector potential can be described in terms of
the Green’s function and a directional function a that is only a function of the tan-
gential spherical coordinates:

A(r) = g(r)a(θ , φ) (4.45)

where the directional function is defined, in rectangular coordinates, as

a = �J x xˆ + �J y yˆ + �J z zˆ (4.46)

In spherical coordinates, the components of a can be found by

aθ = θˆ × a = cos θ cos φ �J x + cos θ sin φ �J y - sin θ �J z (4.47)

aφ = φˆ × a = - sinφ �J x + cos φ �Jy (4.48)

The electric and magnetic fields in the far field can then be found using (4.6), (4.16),
and (4.45), retaining only the terms that are proportional to r–1. The result is

E = - jkη g(r)(θˆaθ + φˆaφ) (4.49)

H = jkg(r)(θˆaφ - φˆ aθ ) (4.50)

The far field electric and magnetic fields therefore contain components only in
the q and f directions, which are orthogonal to the radial direction of propagation.
This corresponds precisely to the orientation of the fields in a plane wave, as derived
in Chapter 2. Thus, the waves in the far field generated by current density can be
considered to be planar.
A similar analysis can be carried out for the fields generated by magnetic cur-
rent densities, yielding

F(r) = g(r)f (θ , φ) (4.51)


94 Microwave and Millimeter-Wave Remote Sensing for Security Applications

where f is a directional function associated with the electric vector potential, the
components of which are given in terms of the Fourier transform of the magnetic
current density by

fθ = cosθ cos φ �Jmx + cos θ sin φ �Jmy - sin θ �Jmz (4.52)

fφ = - sin φ �Jmx + cos φ �Jmy (4.53)


The fields are then given by

E = - jkg(r)(θˆfφ - φˆfθ ) (4.54)

1
H=- jkg(r)(θˆfθ + φˆ fφ ) (4.55)
η

If both electric and magnetic current densities are present, the resulting fields are
the superposition of the fields generated by the two current densities separately:

E = - jkg(r) éëθˆ(ηaθ + fφ ) + φˆ(ηa φ - fθ )ùû (4.56)

é 1 1 ù
H = jkg(r) êθˆ(aφ - fθ ) - φˆ(a θ + fφ)ú (4.57)
ë η η û

4.1.5  Infinitesimal Dipole Far Field Radiation


Returning to the example of the infinitesimal dipole antenna with electric current
density given by (4.26), the Fourier transform of the current is given by

�J(kx , ky , kz ) = zˆ j(k x¢+ k y¢+k z¢)


ò ò ò Idlδ (x¢)δ (y¢)δ (z¢)e x y z dx¢dy¢dz¢ = zˆ Idl (4.58)
x¢ y¢ z¢

Then, from (4.47) and (4.48), the components of the directional function are

aθ = - sin θ Idl (4.59)

aφ = 0 (4.60)

because the current density has only a component in the z direction. The radiated
fields are then, from (4.49) and (4.50),

- jkr
e
E = θˆ jkηIdl sinθ (4.61)
4π r
- jkr
e
H = φˆ jkIdl sin θ (4.62)
4π r
4.2  Antenna Parameters 95

Referring back to the earlier example of the infinitesimal dipole in Section 4.1.3, it
can be seen that (4.61) and (4.62) correspond exactly to the components of (4.28) and
(4.29) that are dependent on r –1. This formulation is valid for the far field only, while
that of the earlier example is valid for all space. Note that, again, the electric and
magnetic fields are orthogonal to each other and to the direction of propagation.

4.2  Antenna Parameters

A number of parameters are used to describe various aspects of the performance


of an antenna, and this section summarizes the more common parameters used to
describe microwave and millimeter-wave antennas in remote sensing applications.
Many of the parameters derive from the electromagnetic fields in the far field region
and are thus interrelated. The IEEE standard definitions of these and other param-
eters can be found in [3].

4.2.1  Radiated Power Density and Total Radiated Power


The radiated power density from an antenna is found from the Poynting vector
(power density) calculated from the electric and magnetic fields radiated by the
antenna


1
S = Re E ´ H*
2
{ } (4.63)

In the far field, the electric and magnetic fields radiated by the antenna are functions
only of q and f,
E = θˆEθ + φˆ Eφ (4.64)

1 1
H= (rˆ ´ E) = (φˆ Eθ - θˆEφ ) (4.65)
η η

and the magnitude of the power density radiated by the antenna is


1 2 2
S= ( Eθ + E φ ) (4.66)

where S is measured in W·m–2·str–1. The power density can alternatively be written
in terms of the directional function a (or f) as
η k2 2 2
S= (aθ + a φ ) (4.67)
32π 2r 2
The total radiated power Pr of an antenna is found by integrating the power
density over a closed surface encompassing the antenna:

òò Sds = òò Sr
2
Pr = sinθ dθ d φ (4.68)
4π 4π

where Pr has units of W.


96 Microwave and Millimeter-Wave Remote Sensing for Security Applications

4.2.2  Antenna Pattern


The antenna pattern, or the radiation intensity, is the radiated power per unit solid
angle (W·str –1), and is given by
r2 2 2 η k2 2 2
A(θ , φ) = r 2 S = ( Eθ + Eφ ) = 2
(aθ + aφ ) (4.69)
2η 32π
Note that because the definitions of the electric fields include the Green’s func-
tion, they are inversely proportional to r, and the antenna pattern is thus indepen-
dent of the radial dimension and depends only on q and f. Due to the reciprocity
theorem [2], the antenna pattern describes the spatial distribution of both the ra-
diation pattern of a transmitting antenna and the reception pattern of a receiving
antenna.
A common practice is to normalize the antenna pattern to its maximum value
Amax and define the normalized antenna pattern as
A(θ , φ)
AN (θ , φ) = (4.70)
Amax
which is unitless. The normalized antenna pattern is useful in describing various
aspects of the antenna pattern in terms of relative levels. Figure 4.4 shows the three-
dimensional antenna pattern of a rectangular aperture.
A number of prominent characteristics of the antenna pattern will affect the
performance of a system in which it is used, and other antenna parameters are de-
rived from the antenna pattern; the antenna pattern is thus an important character-
istic of the antenna. The main beam of the antenna is the largest peak in the antenna
pattern, and it characterizes the direction in which most of the power is radiated. In

x
y
Figure 4.4  Antenna pattern of a square aperture antenna of length 3l per side.
4.2  Antenna Parameters 97

θHPBW
θNNBW
SLL

θNNBW
θHPBW

Figure 4.5  Definitions of antenna beamwidths and sidelobe levels.

addition to the direction of the peak, an important characteristic of the main beam
is its angular width, or beamwidth, which describes the angle over which the power
in the main beam is radiated.
The antenna pattern also includes a number of secondary peaks, called side­
lobes, through which additional power is radiated. Sidelobes are a major concern
in antenna design for remote sensing and antenna applications in general, as they
represent undesired directions where signals will also be transmitted or received.
In an active radar system, the transmitted signal in a sidelobe may reflect off the
ground or other nearby objects, causing large reflected signals that may mask small
returns from the more distant object in the main beam. The metric describing sid-
elobes most often used is the sidelobe level, or the difference between the power
radiated by the main beam to the power radiated by the sidelobe, typically ex-
pressed in decibels. The highest sidelobes often occur directly adjacent to the main
beam and are called the first sidelobes, primary sidelobes, or major sidelobes; the
other, smaller sidelobes are referred to as secondary sidelobes or minor sidelobes.
Figure 4.5 depicts a typical antenna pattern and shows the definition of the sidelobe
level (SLL).

4.2.3  Antenna Pattern Beamwidth


The angular width in a given plane between the points where the radiation intensity
is half of the maximum radiation intensity is called the half-power beamwidth, or
simply the beamwidth and is shown in Figure 4.5. For a symmetric beam, on the
f = 0 plane, the half-power beamwidth in the q direction qHPBW is found by

θ HPBW = 2θ ¢ (4.71)
where
Amax
A(θ ¢,0) = (4.72)
2
or
1
AN (θ ¢,0) = (4.73)
2
98 Microwave and Millimeter-Wave Remote Sensing for Security Applications

The angle between the first two nulls adjacent to the main beam is called the
null-to-null beamwidth. For a symmetric beam, on the f = 0 plane, the null-to-null
beamwidth in the q direction qNNBW is found by

θ NNBW = 2θ ¢¢ (4.74)
where
A(θ ¢¢,0) = 0 (4.75)

is the first null in the pattern.


The half-power beamwidth of a line of current with uniform amplitude is given by

λ λ
θ HPBW » 0.886 = 50.8° (4.76)
d d
where d is the length of the antenna. The null-to-null beamwidth of a line of cur-
rent is
λ λ
θ NNBW = 2 » 114.6° (4.77)
d d

These two values will be derived later when discussing radiation from a rectangular
aperture. The sidelobe level of the primary sidelobe for the uniform linear current
distribution is approximately 13.2 dB below the main beam peak.
If the current distribution is sinusoidal, with a maximum in the center of the
antenna and tapering to zero at the ends, the half-power beamwidth is given by

λ
θ HPBW » 68.8° (4.78)
d

while the null-to-null beamwidth is


λ
θ HPBW » 171.8°
d (4.79)

The sidelobe level for the sinusoidal distribution is 23 dB below the main beam
peak. Thus, by tapering the current distribution across the aperture of the antenna,
the sidelobes can be reduced; however, the drawback is an increase in the beam-
width. It is a rather general construct in antenna engineering that the reduction of
sidelobes results in a wider main beam.
Sinusoidal current distributions are more representative of the currents present
on a real linear antenna; thus, (4.78) and (4.79) can be used as approximations for
the beamwidth of a linear antenna. In a rectangular aperture antenna, the distribu-
tion can more closely approximate a uniform distribution in a given dimension, in
which case (4.76) and (4.77) are more accurate. For instance, the directivity of a
horn antenna can be approximated along either principle dimension by considering
the current distribution as a line current.
The half-power beamwidth and null-to-null beamwidth of a circular aperture
with diameter d and uniform current distribution are given by [4] as
4.2  Antenna Parameters 99

λ
θ HPBW » 58.9° (4.80)
d
λ
θ NNBW » 139.6° (4.81)
d
These relationships demonstrate an important characteristic of the performance
of antennas in general: antennas of large size compared to wavelength yield narrow,
highly directive beams. Antennas with small dimensions compared to wavelength
yield wide, less directive beams.

4.2.4  Antenna Solid Angles


In remote sensing applications, it is often useful to describe some characteris-
tics of the antenna pattern in terms of the solid angles into which various frac-
tions of the transmitted power are radiated. Such parameters will be useful in
the derivation of received power densities when radiometry is considered in
Chapter 6.
The pattern solid angle, or beam solid angle, is the solid angle through which
all the radiated power would stream if the power per unit solid angle were constant
throughout this solid angle and at the maximum value of the radiation intensity. It
is defined as the spatial integral of the normalized antenna pattern,

WA = òò AN dW (4.82)

and has units of steradians. The main beam solid angle is the spatial integral of the
normalized antenna pattern over the extent of the main beam,
WM = òò AN dW (4.83)
main
beam

The extent of the main beam is generally defined as the angle between the first nulls
in the pattern. The minor lobe solid angle is the spatial integration of the antenna
pattern over the space excluding the main beam; thus,

W m = W A - W M (4.84)
The ratio of the main beam solid angle to the pattern solid angle is called the
main beam efficiency,
WM
εM = (4.85)
WA
and is a measure of the amount of power radiated through the main lobe compared
to power directed in other, usually undesired, directions.

4.2.5  Directivity
The directivity of an antenna is the ratio of the radiation intensity in a given direc-
tion to the radiation intensity averaged over all directions,
100 Microwave and Millimeter-Wave Remote Sensing for Security Applications

4π A(θ , φ) 4π A(θ, φ)
D(θ , φ) = = (4.86)
òò A(θ, φ)dW Pr

If no direction is specified, the directivity is typically taken along the direction
of maximum radiation intensity. The maximum directivity is given in terms of the
maximum antenna pattern by
4π Amax
Dmax = (4.87)
Pr
In terms of the normalized antenna pattern,

4π Amax 4π
Dmax = = (4.88)
òò A(θ, φ)dW òò AN (θ, φ)dW
4π 4π

Thus, from (4.83), the maximum directivity can be written



Dmax = (4.89)
WA
Any real antenna is directive; that is, the antenna pattern is not isotropic, or
constant over all space. Isotropic antennas are often useful as mathematical con-
structs in the analysis and design of antennas, such as in the definition of antenna
gain; however, they do not exist in practice [5].
For highly directive antennas, most of the radiated power is in the main beam,
and the contribution outside the main beam may be considered negligible in some
instances. In such a case, the pattern solid angle can be represented by

W A » θBW φ BW (4.90)

where qBW and fBW are the half-power beamwidths of the main beam in the q and f
dimensions. The maximum directivity can then be approximated by


Dmax » (4.91)
θBW φ BW

which is given in degrees by

41, 253
Dmax » (4.92)
θBW φ BW

This expression is an approximation for a single antenna with uniform illumi-


nation; for a planar array, considered later in this chapter, a more accurate approxi-
mation given by [6] is
32, 400
Dmax » (4.93)
θBW φBW
4.2  Antenna Parameters 101

4.2.6  Gain
The gain of an antenna is the ratio of the radiated power density in a given direction
to the power density of an isotropic antenna with the same input power Pin. The
gain is a standard metric of antenna performance and is often given simply as the
maximum gain along the direction of maximum radiation intensity. The gain can
be related to the directivity through the radiation efficiency of the antenna, which is
the ratio of the radiated power and the power input to the antenna,
Pr
εr = (4.94)
Pin
The lost power Pl­­ is converted into heat in the antenna material. Because

Pin = Pr + Pl (4.95)
the radiated and lost powers can be given in terms of the radiation efficiency by

Pr = ε r Pin (4.96)

Pl = (1 - ε r )Pin (4.97)

The gain is defined in terms of the radiated power density of an antenna with
input power Pin as
S(θ , φ)
G(θ , φ) = (4.98)
Si

where Si is the radiated power density of a lossless isotropic antenna with equal
input power Pi. Because the isotropic antenna is lossless, the input power is equal
to the isotropic radiated power Pri. Thus, from (4.68),

Pri P
Si = 2
= in 2 (4.99)
4π r 4π r
Using (4.96), the isotropic radiated power can be related to the radiated power
of the nonideal antenna under consideration. Thus,
Pr 1
Si =
4πεr r 2
=
4πε r òò S(θ, φ)dW (4.100)

The gain is therefore


S(θ , φ)
G(θ , φ) = 4πε r = ε r D(θ , φ) (4.101)
òò S(θ, φ)dW

and the maximum gain is


Gmax = ε r Dmax = ε r (4.102)
WA
102 Microwave and Millimeter-Wave Remote Sensing for Security Applications

4.2.7  Aperture Area and Pattern Solid Angle


A relationship that will be useful in the discussion of radiometry is that between
the pattern solid angle and the effective area of the antenna aperture. The effective
aperture Ae is the ratio of the available power PT at the terminals of a receiving
antenna to the power flux density of a plane wave incident on the antenna from
a given direction, assuming that the polarization of the antenna is matched to the
polarization of the incident wave, as discussed in the next section, and is given by
PT
Ae = (4.103)
Sinc
The effective area is given in terms of the physical area of the antenna Ap by

Ae = ε A Ap (4.104)
where 0 £ eA £ 1 is the aperture efficiency, which accounts for conduction losses in
the antenna and impedance mismatches.
The relationship between the effective aperture area and the pattern solid angle
can be found by considering the power due to the fields in the aperture
2
Eap Ae
P1 = (4.105)

and the power at a distance r
2
Er 2
P2 = r W A (4.106)

The electric field at r is related to that in the aperture by [7]
Eap Ae
Er = (4.107)

which holds true for a uniformly illuminated aperture. From (4.106) and (4.107),
2
Eap
P2 = Ae2 W A (4.108)
2η λ2

If the medium through which the waves propagate is lossless, the radiated power
P2 must be equal to the aperture power P1. Thus, equating (4.105) and (4.108) yields
λ2
Ae = (4.109)
WA
Thus, the product of the effective aperture area and the pattern solid angle is equal
to the square of the wavelength.
The maximum directivity can be given in terms of the effective area by substi-
tuting (4.109) into (4.89), yielding

Dmax = Ae (4.110)
λ2
4.2  Antenna Parameters 103

From (4.101), the gain is thus



Gmax = ε r Ae (4.111)
λ2

A useful approximation of the directivity of an antenna is the assumption that


for a given antenna, the effective area is equal to the physical area of the aperture.
Then, (4.110) can be used to directly calculate the directivity of an antenna at a
given wavelength. If the radiation efficiency is known, the gain may then be found
by multiplying the directivity by the efficiency. That is,

Gmax » ε r Ap (4.112)
λ2
for antennas with good efficiency.
A useful example is that of an isotropic antenna with unity gain. For such a
radiator, AN = 1, and the antenna solid angle (4.82) is WA = 4p. The effective area
of an isotropic antenna is therefore, from (4.109),
λ2
Ae,isotropic = (4.113)

The effective area of an isotropic radiator therefore scales with the square of the
wavelength.

4.2.8  Antenna Temperature and Noise Power


Due to movement of electrons within materials, all objects at a nonzero temperature
radiate energy, referred to as thermal radiation and discussed in detail in Chapter
6. Antennas thus intrinsically radiate energy that manifests as noise power at the
antenna terminals. At microwave and millimeter-wave frequencies, the power of
the radiated energy is proportional to the noise temperature of the antenna, called
the antenna noise temperature TA, and is given by

PA = kTA Df (4.114)
where k = 1.38 × 10–23 J·K–1 is Boltzmann’s constant and Df is the bandwidth of
the signal. Due to the random nature of thermal fluctuations, the noise power has a
Gaussian distribution. In a radiometer, the antenna noise temperature is primarily
the result of incident thermal radiation, and the antenna noise power represents the
signal of interest; this is discussed further in Chapter 6.

4.2.9  Polarization
The polarization of an antenna is given by the polarization of a wave transmitted by
the antenna in a given direction. While generally dependent on angle, the polariza-
tion of an antenna is typically considered as that of a wave transmitted along the
direction of maximum radiation intensity.
A measure of how well an antenna is matched to the polarization of an incident
wave is given by the polarization loss factor. Given an incident wave
104 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Einc = pˆ inc Einc (4.115)


where p̂inc is the unit vector of the incident wave polarization, and the wave in the
receiving antenna
E a = pˆ a Ea (4.116)

where p̂a is the antenna polarization, the loss factor is defined as

ε p = pˆ inc × pˆ a = cos Y (4.117)

where Y is the angle between the two waves.

4.3  Properties of Wire Antennas

In this section, the concepts covered in the previous sections will be utilized in
examples of radiation from an electric current source. As described in the intro-
duction, radiation sources may be either electric current densities on an antenna
or equivalent current densities in an aperture. Linear wire antennas fall under the
first category, whereas antennas such as horn and microstrip antennas fall under
the latter.
Earlier, the radiation fields of an infinitesimal dipole antenna were derived.
Due to its infinitesimal linear nature, such an antenna may be used as a differ-
ential element to calculate the fields of a finite length linear wire antenna. First,
some properties of the infinitesimal dipole antenna will be derived. Following
this, the radiation fields and the antenna pattern of a long wire dipole antenna
will be considered. For a detailed analysis of linear wire antennas, the reader is
referred to [8].

4.3.1  Infinitesimal Dipole


As shown in Section 4.1, the directional function of an infinitesimal dipole is given
by
aθ = - sin θ Idl (4.118)

As defined by (4.67), the radiated power density is therefore


2
ηk2 Idl
S= 2 2
sin2 θ (4.119)
32π r
and the antenna pattern is
2
2 ηk2 Idl
A(θ) = r S = sin2 θ (4.120)
32π 2
The maximum point in the antenna pattern is found when sin2q = 1, yielding
2
ηk2 Idl
Amax = (4.121)
32π 2
4.3  Properties of Wire Antennas 105

θ
0
30 30

60 60

1
0.5
90 90

120 120

150 150
180
Figure 4.6  Normalized antenna pattern of an infinitesimal dipole.

The normalized antenna pattern is therefore

AN (θ) = sin2 θ (4.122)


The antenna pattern is plotted in Figure 4.6. It is toroidally shaped, has nulls at
q = 0° and 180, and is maximum at q = 90°. The peak in the pattern corresponds to
the broadside direction, whereas the nulls correspond to the axis along the length
of the antenna. Additionally, due to the symmetrical shape of the antenna in the f
direction, the antenna pattern is constant in f and depends only on q.
The total power radiated by an infinitesimal dipole is found from (4.68); carry-
ing out the integration results in
2
ηk2 Idl
òò Sr sin θ dθ dφ =
2
Pr = (4.123)

12π 2
The directivity is therefore

3 2
D(θ , φ) = sin θ (4.124)
2

and the maximum directivity is


4π Amax 3
Dmax = = (4.125)
Pr 2

4.3.2  Long Dipole


A finite-length dipole antenna can be considered to be a superposition of infini-
tesimal dipoles. Figure 4.7 shows a long (compared to wavelength) dipole antenna
of length l oriented along the z axis. The current along a center-fed, linear dipole
antenna has been shown experimentally to be approximately sinusoidal along the
106 Microwave and Millimeter-Wave Remote Sensing for Security Applications

length of the antenna, with nulls at the ends [8, 9]. The current along the length of
the antenna can thus be described by
é æl öù
I = zˆ I0δ (x ¢)δ (y ¢)sin êk ç - z ¢ ÷ ú (4.126)
ë 2è øû

where the current is zero elsewhere. The Fourier transform of the current density is
found by integrating over the length of the antenna,
�J = zˆ é æl öù j(kx x¢ + ky y ¢ + kz z ¢)
ò ò ò I0δ (x¢)δ (y¢)sin êëk çè 2 - z ¢ ÷ø úû e dx ¢dy ¢dz ¢
x¢ y ¢ z ¢

l
2
é æl öù
ò I0 sin êëk çè 2 - z ¢ ÷ø úû e
jkz z ¢
= zˆ dz ¢
l
-
2 (4.127)
The wavenumber is kz = kcosq; thus,
l
2
�J = zˆ é æl öù
ò I0 sin êëk çè 2 - z ¢ ÷ø úû e
jkz ¢ cos θ
dz ¢ (4.128)
l
-
2

The directional function is given by substituting (4.128) into (4.47); carrying


out the integration yields
é æ1 ö æ1 öù
ê cos ç kl cos θ ÷ - cos ç kl ÷ ú
2I è 2 ø è2 ø
aθ = - 0 ê ú
k ê sin θ ú (4.129)
êë úû

The far field electric and magnetic fields are then


é æ1 ö æ1 öù
cos ç kl cos θ ÷ - cos ç kl ÷ ú
e - jkr ê è2 ø è2 ø
Eθ = jηI0 ê ú
2π r ê sin θ ú (4.130)
êë úû
z

dz

Figure 4.7  Long dipole antenna.


4.4  Aperture Antennas 107

l = 1λ l = 2λ
0 0

90 90 90 90

180 180
l = 3λ l = 4λ
0 0

90 90 90 90

180 180
Figure 4.8  Normalized antenna patterns of a long linear dipole with l = 1l, 2l, 3l, 4l.

é æ1 ö æ1 öù
ê cos ç kl cos θ ÷ - cos ç kl ÷ ú
- jkr
e è2 ø è2 ø
Hφ = jI0 ê ú
2π r ê sinθ ú (4.131)
êë úû

The radiated power density from the long dipole can be calculated from either
the directional function (4.129) or the electric field (4.130); the result is the same
in either case,
2
é æ1 ö æ1 öù
cos ç kl cos θ ÷ - cos ç kl ÷ ú
2
η I0 ê è2 ø è2 ø
S= 2 2ê ú
8π r ê sin θ ú (4.132)
êë úû

The antenna pattern is found by A = r2S, yielding


2
é æ1 ö æ1 öù
η I0
2 ê cos çè 2 kl cos θ ÷ø - cos çè 2 kl ÷ø ú
A(θ) = ê ú
8π 2 ê sin θ ú (4.133)
ëê ûú
Due to the symmetry of the antenna, the antenna pattern is only a function of q. The
normalized antenna pattern is plotted in Figure 4.8 for lengths of l = 1l , 2l , 3l , and 4l.

4.4  Aperture Antennas

The simplicity of the method of calculating the radiated fields introduced in Section
4.1 was predicated on the fact that the potentials could be calculated from current
densities existing on surface of the antenna. In an aperture antenna, such as a horn
antenna, the radiated fields are also ultimately generated by current sources; how-
ever, the antenna, where the wave transitions from a guided wave to a propagating
108 Microwave and Millimeter-Wave Remote Sensing for Security Applications

wave, is no longer described as a metal region on which current densities reside. The
relevant part of the antenna is the aperture, occupied by free space or a dielectric,
where electromagnetic fields exist but no physical current densities do. Thus, in
order to use the relatively straightforward method of calculating the radiated fields
from the electromagnetic potentials, a relation between the potentials and the fields
in the aperture must be devised.
Such a relation can be made by using the equivalence principle, with which elec-
tromagnetic fields can be replaced by equivalent sources that satisfy the boundary
conditions. The equivalence principle was originally devised by Schelkunoff [10, 11]
as an extension of Huygen’s principle, which states that the points comprising a wave-
front can be considered to be individual point sources generating new wavefronts [2].
Using this principle, the fields in the aperture of an antenna can be replaced by equiva-
lent sources and the potentials can then be used to calculate the radiated fields.

4.4.1  Image Theory


The formulation of the equivalence principle for use in aperture antenna analysis
requires a discussion of image theory, which can be derived from the evaluation of
the fields generated by sources in the presence of a conducting plane [12]. The prob-
lem of primary interest will be the boundary value problem, where the conductor
is a perfect electric conductor; however, the dual problem with a perfect magnetic
conductor is analogous.
Consider a current source density in the presence of a perfect electric conduct-
ing plane of infinite extent. The boundary conditions state that the tangential com-
ponents of the electric field on the conducting boundary must be zero. According to
image theory, the boundary conditions can be satisfied by replacing the conductor
with free space in which an appropriately oriented image source is placed such
that the tangential components of the electric field at the location of the boundary
are zero. Figure 4.9 shows the required orientation of the image current, given a
real current density in the cases of both a perfect electric conductor and a perfect
magnetic conductor. In the case of a perfect electric conductor, a tangential current
source has associated with it an image that is in the opposite direction, whereas the

dd dd
Perfect Perfect
electric magnetic
J J
conductor conductor

J J

Jm Jm

Jm Jm

Images Sources Images Sources


Figure 4.9  Real and image sources in the presence of a conducting plane.
4.4  Aperture Antennas 109

dd
Perfect
electric
J 0
conductor lim
d 0

Jm 2Jm

Figure 4.10  Resulting tangential current densities as the spacing between the source and conduct-
ing plane approaches zero.

image is in the same direction for a magnetic source. In the case of a perfect mag-
netic conductor, the orientations of the images are reversed.
An important aspect of the equivalence principle is derived from the result ob-
tained when the distance between the plane and the tangential sources approaches
zero. In the limit, the current densities become surface current densities; replacing the
conductor with free space and image currents results in the superposition of the real
and image sources, since they are both on the surface of the boundary. With a perfect
electric conductor, the tangential electric currents cancel one another, as shown in
Figure 4.10, resulting in a null current on the surface; thus, electric currents do not
exist on a perfect electric conductor. The tangential magnetic currents, however,
combine to produce a current density that is twice the real current density.

4.4.2  The Equivalence Principle


Consider the fields E1 and H1, generated by the sources J and Jm, which are in free
space. As shown in Figure 4.11, a surface s can be defined that completely encom-
passes the current densities in the volume V0. The fields outside the surface, in the
volume V1, are unchanged. The equivalence principle states that the sources and
fields within V0 can be replaced by new fields E0 and H0 with current densities Js
and Jms on the surface s, which satisfy the boundary conditions

J s = nˆ ´ (H1 - H0 ) (4.134)

E1, H1 E 1, H 1 E 1, H 1

V1 V1

Js

E 0, H 0
J J Jms
Jm Jm

s s
Figure 4.11  Equivalence principle.
110 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Jms = - nˆ ´ (E1 - E0 ) (4.135)

With the new fields and surface currents defined such that (4.134) and (4.135)
are satisfied, the fields E1 and H1 in the volume V1 will remain unchanged from the
original problem.
In replacing the volume V0, the implicit assumption is that it is the fields within
V1 that are of interest, rather than the fields within V0. Thus, the problem can be
simplified by setting the fields within the surface to be zero. The boundary condi-
tions are then
J s = nˆ ´ H1 (4.136)

Jms = - nˆ ´ E1 (4.137)

where E1 and H1 are specified on the surface. The equivalent current densities are
now specified in terms of only the fields at the surface s. If the surface is defined to
enclose an antenna, such as a horn, and is coincident with the aperture, the fields
E1 and H1 are the fields present in the antenna aperture. Thus, equivalent currents
can be formed in the aperture using (4.136) and (4.137), and the electromagnetic
potentials can be used to calculate the radiated fields.
Consider an open-ended waveguide in an infinite ground plane, as shown in
Figure 4.12. The aperture contains the electric field Ea and no magnetic field. Us-
ing the equivalence principle, a surface s can be defined to be coincident with the
ground plane and the aperture, and the field in the aperture can be replaced by an
equivalent magnetic current density. The tangential electric fields on the ground
plane are zero; thus, no equivalent magnetic current exists outside of the aperture.
The magnetic fields outside of the aperture are undefined; therefore, an equivalent
electric current can exist on the surface s.
Because the fields in the volume of the ground plane (left of the surface) are not
of interest, the volume can be replaced by an arbitrary medium. For the analysis
of aperture antennas, it is convenient to replace the volume with a perfect electric
conductor. Image currents are then generated by the equivalent magnetic current
density in the aperture, as well as by the equivalent electric current density on s.

PEC

^
n Js=0

^
Jms=-nxE Jms Jms ^
Jms=-2nxE
Ea a a

Figure 4.12  Equivalence principle for an open-ended waveguide in an infinite ground plane.
4.4  Aperture Antennas 111

Because these are all surface currents, the image of the equivalent electric current
density is in the opposite direction, nulling the equivalent current. The image cur-
rent of the equivalent magnetic current density in the aperture is in the same direc-
tion, adding to the equivalent current density. Thus, the result is that the equivalent
electric current densities are zero everywhere, and the equivalent magnetic current
density is zero outside of the aperture, and in the aperture is

Jms = -2nˆ ´ E a (4.138)

The resulting equivalent problem is that of a magnetic current density (rep-


resenting the aperture) in free space. From Jms the electric vector potential F can
be calculated using (4.51), from which the radiated fields can be calculated using
(4.18) and (4.22), or they can be calculated directly using (4.52)–(4.55).
In many cases, the antenna under consideration will not be flush with a ground
plane, and may be surrounded by free space. The typical approach in such a case
is to define the surface s in the plane of the aperture (as if there were a ground
plane) and simply assume that the electric and magnetic fields are zero outside of
the aperture. Thus, the equivalent currents will only be nonzero within the antenna
aperture, and the analysis proceeds as described in the previous example. Although
an approximation, this approach provides results that closely match measured data
and is highly accurate for the antenna main beam [2].

4.4.3  Radiation from a Rectangular Aperture


Rectangular apertures are commonly implemented in microwave and millimeter-
wave remote sensing in the form of horn antennas, waveguide slot antennas, and
microstrip antennas, among others. The results from analyzing a general rectan-
gular aperture can be applied as an approximation to most rectangular antenna
apertures. Consider a rectangular aperture of dimensions a × b on a ground plane,
as shown in Figure 4.13. In the aperture is an electric field given by

E a = yˆ E 0 (4.139)

The aperture contains no magnetic field, and the fields outside the aperture are
also zero. By the equivalence principle, the field in the aperture can be replace by an
equivalent magnetic current density given by

Jm = -2nˆ ´ E a = 2xˆ E0 (4.140)


The current exists only within the region –a/2 £ x £ a/2, –b/2 £ y £ b/2. Taking the
Fourier transform thus yields

�Jmx = 2abE0sinc æç akx ö÷ sinc çæ bky ÷ö (4.141)


è 2 ø è 2 ø
where
sin α
sinc(α ) = (4.142)
α
112 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Ea

PEC PEC
y
a

b
x

Figure 4.13  Rectangular aperture in a ground plane.

The components of the magnetic directional function f are found using (4.52) and
(4.53), and are given by

æ ak ö æ bky ö
fθ = 2abE0cosθ cosφsinc ç x ÷ sinc ç (4.143)
è 2 ø è 2 ÷ø

æ ak ö æ bky ö
fφ = -2abE0 sinφ sinc ç x ÷ sinc ç (4.144)
è 2 ø è 2 ÷ø

From (4.54), the electric field components are then given by

æ ak ö æ bky ö
Eθ = jk2abE0 g(r)sinφ sinc ç x ÷ sinc ç (4.145)
è 2 ø è 2 ÷ø

æ ak ö æ bky ö
Eφ = jk2abE0 g(r)cos θ cosφ sinc ç x ÷ sinc ç (4.146)
è 2 ø è 2 ÷ø

The antenna pattern is calculated in terms of the electric field components by (4.69),
which yields
1
A(θ , φ) = 2
(abk E0 )2
8ηπ

é æ ak ö æ bk ö
´ êsin2 φsinc 2 ç sin θ cos φ ÷ sinc 2 ç sinθ sinφ ÷ (4.147)
ë è 2 ø è 2 ø

æ ak ö æ bk öù
+ cos2 θ cos2 φ sinc 2 ç sinθ cosφ ÷ sinc 2 ç sinθ sinφ ÷ ú
è 2 ø è 2 øû

where the wavenumbers have been expanded. The maximum value of the term in
brackets in (4.147) is 1; thus, the normalized antenna pattern is
4.4  Aperture Antennas 113

æ ak ö æ bk ö
AN (θ , φ) = sin2 φ sinc 2 ç sinθ cosφ ÷ sinc 2 ç sinθ sinφ ÷ (4.148)
è 2 ø è 2 ø

æ ak ö æ bk ö
+ cos2 θ cos2 φ sinc 2 ç sinθ cosφ ÷ sinc 2 ç sinθ sinφ ÷
è 2 ø è 2 ø

and is plotted in Figure 4.14(a) for the parameters a = 4l, b = 3l.


It is often convenient to examine the antenna pattern as a two-dimensional
function along a plane through the origin. The logical planes to consider are the x-z



x
y
(a)

0
E−plane
H−plane
−5

−10
dB

−15

−20

−25

−30
−1.5 −1 −0.5 0 0.5 1 1.5
θ (rad)
(b)
Figure 4.14  (a) Normalized antenna pattern and (b) E-plane and H-plane patterns of a rectangular
aperture with a = 4l, b = 3l.
114 Microwave and Millimeter-Wave Remote Sensing for Security Applications

plane and the y-z plane, as they align with the geometry of the aperture. The y-z
plane is oriented along the direction of the electric field, and thus the antenna pat-
tern on this plane is called the E-plane pattern. Similarly, the pattern along the x-z
plane aligns with the magnetic field and is called the H-plane pattern.
The E-plane pattern is found by evaluating the normalized antenna pattern at
the angle f = p/2, which yields
æ bk ö
AE (θ) = AN (θ ,φ = π /2) = sinc 2 ç sinθ ÷ (4.149)
è 2 ø
The H-plane pattern is found by setting f = 0, which results in
æ ak ö
AH (θ) = AN (θ ,φ = 0) = cos2 θ sinc 2 ç sinθ ÷ (4.150)
è 2 ø
The E-plane and H-plane antenna patterns are plotted in Figure 4.14(b).
The half-power beamwidth of the E-plane pattern is found in terms of the angle
where the normalized antenna pattern in the main beam is equal to 0.5,
æ bk ö
sinc 2 ç sin θ H ÷ = 0.5 (4.151)
è 2 ø
This occurs when the argument of the sinc function is equal to approximately
1.391. Thus,
bk
sin θ H = 1.391 (4.152)
2
or
æ 2.782 ö æ λö
θ H = sin -1 ç = sin -1 ç 0.443 ÷ (4.153)
è kb ÷ø è bø
The half-power beamwidth is thus
æ λö
θ HPBW = 2sin -1 ç 0.443 ÷ (4.154)
è bø
For large apertures where b >> l, the argument of (4.154) is small, and thus
λ
θ HPBW » 0.886 (4.155)
b
The null-to-null beamwidth of the E-plane pattern is found in terms of the angle
where the pattern reaches its first zero, which occurs when
bk
sin θ N = π (4.156)
2
Thus,
æ 2π ö æ λö
θ N = sin -1 ç ÷ = sin -1 ç ÷ (4.157)
è kb ø è bø
and the null-to-null beamwidth is
æ λö
θ NNBW = 2sin -1 ç ÷ (4.158)
è bø
4.4  Aperture Antennas 115

For large apertures where b >> l, the beamwidth is approximated by


λ
θ NNBW » 2 (4.159)
b

4.4.4  Radiation from a Circular Aperture


Although rectangular apertures are the most commonly implemented antenna
shapes, circular apertures are also commonly used, for instance, in cylindrical horn
antennas with circular apertures, and reflector and lens antenna systems. As with a
rectangular aperture, the results from the analysis of a general circular aperture may
be applied as an approximation to an antenna or antenna system with a circular
aperture. Consider the circular aperture of radius a in an infinite ground plane in
Figure 4.15(a). The field in the aperture is given by

E = yˆ E0 (4.160)
and the current is thus

Jm = xˆ 2E0 (4.161)
for r £ a, and is zero elsewhere. The calculation of the radiated fields follows the same
procedure as before; however, for the circular aperture the analysis is more easily ac-
complished using cylindrical coordinates (e.g., see [2]). The resulting electric fields are

é J (ak sin θ) ù
Eθ = jk4π a2 E0 g(r)sin φ ê2 1 (4.162)
ë ak sin θ úû

é J (ak sin θ) ù
Eφ = jk4π a2 E0 g(r)cos θ cos φ ê2 1 (4.163)
ë ak sin θ úû
where J1(x) is the first-order Bessel function of the first kind (see, e.g., [13]). The
normalized antenna pattern is
2
é J (ak sin θ) ù
AN (θ , φ) = (sin2 φ + cos2 θ cos2 φ ) ê2 1 (4.164)
ë ak sin θ úû
and is plotted in Figure 4.15(b) for a = 2l.
The E-plane pattern of the circular aperture is given by
2
é J (ak sin θ) ù
AE (θ) = AN (θ , φ = π 2) = ê2 1 (4.165)
ë ak sin θ úû
and the H-plane pattern is
2
é J (ak sin θ) ù
AH (θ) = AN (θ , φ = 0) = cos2θ ê2 1 (4.166)
ë ak sin θ úû

The E-plane and H-plane antenna patterns are plotted in Figure 4.15(c) for a = 3l.
The half-power beamwidth of the E-plane pattern is found in terms of the angle
where (4.165) is equal to 0.5, or
116 Microwave and Millimeter-Wave Remote Sensing for Security Applications

PEC a PEC
y
Ea

x (a)
z


x
y

(b)

0
E−plane
−5 H−plane

−10

−15
dB

−20

−25

−30
−1.5 −1 −0.5 0 0.5 1 1.5
θ (rad)
(c)
Figure 4.15  (a) Circular aperture in a ground plane. (b) Normalized antenna pattern and (c) E-
plane and H-plane patterns of a circular aperture with a = 2l.
4.5  Antenna Arrays 117

J1(ak sin θ)
= 0.3535 (4.167)
ak sin θ
which is satisfied when aksinq » 1.6, and thus
æ 1.6 ö æ λö
θ H = sin -1 ç = sin -1 ç 0.25 ÷ (4.168)
è ak ÷ø è aø
The half-power beamwidth is therefore
æ λö
θ HPBW = 2θH = 2sin -1 ç 0.25 ÷ (4.169)
è aø
which for large apertures (a >> l) reduces to
λ
θ HPBW » (4.170)
2a
The null-to-null beamwidth is found when
J1(ak sin θ)
= 0 (4.171)
ak sin θ
which is satisfied when ak sin q » 3.825. Therefore,

æ 3.825 ö æ λö
θ N = sin -1 ç ÷ = sin -1 ç 0.61 ÷ (4.172)
è ak ø è aø
and
æ λö
θ NNBW = 2sin -1 ç 0.61 ÷ (4.173)
è aø
For large apertures, this simplifies to
λ
θ NNBW » 1.22 (4.174)
a

4.5  Antenna Arrays

An antenna array may be described as an antenna comprised of a number of ra-


diating elements whose inputs or outputs are combined to obtain one or more
prescribed radiation patterns. The individual elements of an array need not be iden-
tical; however, it simplifies the analysis of arrays to consider identical elements. As
seen in the previous sections, the directional characteristics of antennas, such as the
beamwidth, are dependent on the antenna dimensions; larger antenna dimensions
result in narrower beams. Arrays are widely used due to the ability to increase the
gain of an antenna and simulate a large aperture using smaller, cheaper discrete
elements. However, due to the increase in hardware required in an array, the cost
can be prohibitive for some applications, particularly at millimeter-wave frequen-
cies where component costs are greater. Additionally, the discrete nature of such an
antenna causes aliasing effects that manifest as additional highly directive beams,
called grating lobes, which must be addressed to remove spatial ambiguities. Array
elements can also be individually addressed to steer the main beam for scanning.
118 Microwave and Millimeter-Wave Remote Sensing for Security Applications

4.5.1  Linear Array Theory


An antenna array can be considered to be a single large antenna comprised of
a number of smaller antennas that are physically separated from one another.
Figure 4.16 shows a linear array comprised of N infinitesimal dipoles separated by
a distance d, which can be seen to be a discretely sampled version of the long linear
dipole antenna considered earlier. The current density along the array is given by
N -1
2
J = zˆ å In dlδ (x ¢)δ (y ¢)δ (z ¢ - nd) (4.175)
N -1
n =-
2

The current is thus a summation of delta functions spaced along the z axis in
increments of length d. The Fourier transform of the current density is then
N -1
2
�J = zˆ å ò ò ò In dlδ (x ¢)δ(y ¢)δ(z ¢ - nd)e j(k x¢ +k y¢ +k z ¢)dx ¢dy ¢dz ¢
x y z

N -1 x¢ y¢ z ¢
n =-
2
(4.176)
N -1 N -1
2 2
= zˆ å In dle jkz nd = zˆ å In dle jknd cos θ
N -1 N -1
n =- n =-
2 2

If the current on each antenna element is identical, the Fourier transform of the
current density is
N -1
2
�J = zˆ Idl å e jknd cos θ = zˆ IdlAF(θ) (4.177)
N -1
n =-
2

where AF(q) is called the array factor.


A more compact form of the array factor is found by first letting y = kdcosq;
the array factor is then

Figure 4.16  Linear array consisting of infinitesimal dipole antennas.


4.5  Antenna Arrays 119

N -1
2
AF = å e jn ψ
N -1
n =-
2
(4.178)

æ N -1ö æ N -3ö æ N -3ö æ N -1ö


- jç ψ - jç ψ jç ψ jç ψ
è 2 ø÷ è 2 ø÷ - jψ jψ è 2 ø÷ è 2 ø÷
=e +e +…+ e +1+ e +…+ e +e

Multiplying the array factor by ejy yields


æ N -3ö æ N -1ö æ N +1ö
- jç ψ jç ψ jç ψ
jψ è 2 ÷ø -jψ jψ è 2 ÷ø è 2 ÷ø
AFe =e +…+ e +1+ e + …+ e +e (4.179)

Subtracting (4.178) from (4.179) gives


æ N +1ö æ N -1ö
jç ψ - jç ψ
è 2 ÷ø è 2 ÷ø
AFe jψ - AF = AF(e j ψ - 1) = e -e (4.180)

or
æ N +1 ö æ N -1ö 1 æ j 1 Nψ 1 ö
AF =
e

è 2 ÷ø
ψ
-e
- jç
è 2 ÷ø
ψ

=
e
j
2
ψ
çe 2
- j Nψ
-e 2
1
÷ sin 2 Nψ ( )
÷ = sin 1 ψ (4.181)

e -1 1
j ψ
e2
ç j1ψ
çè
e 2 -e 2
1
-j ψ ÷ø 2 ( )
The current (4.177) is therefore given by
æ1 ö
sin ç Nkd cos θ ÷
�Jz = Idl è 2 ø
æ1 ö
sin ç kd cos θ ÷ (4.182)
è2 ø

The electric field is then derived from the directional function a using (4.47), result-
ing in
æ1 ö
sin ç Nkd cos θ ÷
è2 ø
Eθ = jkηIdlg(r)sinθ
æ1 ö
sin ç kd cos θ ÷ (4.183)
è2 ø

The electric field of the uniform linear array (4.183) is similar to that derived
for the long dipole, however with an additional function that is the ration of two
sine functions. As given by (4.183), the maximum value of the array factor is N, the
number of elements. It is customary to normalize the array factor by N so that its
maximum value is unity; thus,
æ1 ö
sin ç Nkd cos θ ÷
è2 ø
AF(θ) =
æ1 ö
N sin ç kd cos θ ÷ (4.184)
è2 ø
and the electric field is given by

Eθ = jkηIdlg(r)N sinθ AF(θ ) = N ´ Ee ´ AF (4.185)


120 Microwave and Millimeter-Wave Remote Sensing for Security Applications

where Ee is the electric field pattern due to one of the identical elements comprising
the array. Equation (4.185) is an expression of pattern multiplication for arrays of
identical elements, which states that the total field pattern of an array is given by
the multiplication of the pattern of the individual antenna elements and the array
factor, scaled by the number of elements N. Thus, the pattern due to the array can
be evaluated separately from the individual elements. The factor of N accounts for
the additional energy due to the total number of elements in the array. The normal-
ized antenna pattern of the array is therefore given by
1 
sin2  N kd cosθ
AN (θ ) = sin2θ 2  = A (θ ) × A F 2(θ )
e
2 21  (4.186)
N sin  kd cosθ
2 
where Ae is the antenna pattern of the element.
The nulls in the array factor are found when the numerator of (4.184) is zero, which
corresponds to
1
Nkd cosθn = ±nπ , n = 1, 2,3,… (4.187)
2
Thus, the angles of the nulls are
æ nλ ö
θn = cos -1 ç
è Nd ÷ø (4.188)

The locations of the maxima of the array factor are found when the denominator
of (4.184) is zero, which is true when
1
kd cos θm = ± mπ, m = 0,1, 2,… (4.189)
2

and thus the maxima are located at the angles

æ mλ ö
θm = cos -1 ç (4.190)
è d ÷ø

Multiple maxima can therefore be present in the array factor. This arises from
the discrete nature of the current density, which results in spatial aliasing in the
form of grating lobes in the pattern. Equation (4.190) produces real angles when
ml /d < 1; thus, in order that only the m = 0 maxima is present in the pattern, the
element spacing should be
d
< 1 (4.191)
λ
which ensures that no grating lobes will be in the pattern. This holds for broadside
arrays; for scanning arrays, as discussed later, the element spacing must be shorter
to ensure that no grating lobes are present in the full scan volume.
Figure 4.17 shows the array factor of a linear array as the element spacing
increases. The grating lobes appearing at wider element spacings are due to the
discrete nature of the current distribution, as shown in Figure 4.18. In contrast to a
4.5  Antenna Arrays 121

0
−10 d = 0.5λ

dB
−20
−30
−1.5 −1 −0.5 0 0.5 1 1.5
0
d = 1λ
−10

dB
−20
−30
−1.5 −1 −0.5 0 0.5 1 1.5
0
d = 2λ
−10
dB

−20
−30
−1.5 −1 −0.5 0 0.5 1 1.5
θ (rad)
Figure 4.17  Array factor for element spacing d = 0.5l, d = l, and d = 2l, showing the presence of
grating lobes at longer element spacing.

continuous current distribution, the Fourier transform of a discrete current distri-


bution results in special aliasing that manifests as grating lobes.

4.5.2  Planar Arrays


Two-dimensional arrays, called planar arrays due to the conformity of the elements
on a geometric plane, are formulated in the same manner as the linear arrays, with
the discrete current density now extended along two dimensions. The current den-
sity of an array on the x-y plane is given by
M -1 N -1
2 2
J = zˆ å å Imn dlδ (x ¢ - mdx )δ (y ¢ - ndy )δ (z ¢) (4.192)
M -1 N -1
m =- n =-
2 2

Current distribution

FT

FT

Figure 4.18  (top) Continuous current density distribution and its Fourier transform. (bottom)
D­iscrete current density and its Fourier transform. The discrete nature of the current density causes
aliasing in the form of grating lobes.
122 Microwave and Millimeter-Wave Remote Sensing for Security Applications

dy dy

dx dx

Figure 4.19  Planar array with elements arranged in a rectangular and a triangular grid.

where dx and dy are the element spacings along the x and y dimensions. If the current is
identical on all elements, the Fourier transform of the current density can be written
M -1 N -1
2 2
�J z = å å Idle
jk(mdx sin θ cosφ + ndy sin θ sin φ)
(4.193)
M -1 N -1
m =- n =-
2 2

The array factor is


sin( 12 Mkdx sin θ cos φ) sin( 12 Nkdy sinθ sin φ)
AF(θ , φ) = ´ (4.194)
M sin( 12 kdx sin θ cos φ ) N sin( 12 kdy sinθ sinφ )

The arrangement of the elements in the array can be varied to change the per-
formance. For instance, the interelement spacing of a triangular grid as shown in
Figure 4.19 results in grating lobes at wider angles than rectangular arrays of simi-
lar size [14]. Thinned arrays reduce the total number of elements by selectively
or randomly removing elements; an array with resulting nonuniform interelement
spacing results in lower primary sidelobes in the array factor.

4.5.3  Array Beamwidth


The half-power and null-to-null beamwidths of an antenna array pattern can be eval-
uated along any plane of the pattern, although typically they are considered along the
E- and H-planes. The half-power beamwidth of the array pattern is found in terms of
the angle where (4.186) is equal to 1/2. From tables of the function sin(Nx)/Nsin(x),
found in many textbooks on array theory (e.g., [2]), this occurs when
1
Nkd sin θh = ±1.391 (4.195)
2
and thus
æ λ ö
θh = sin -1 ç 0.44 ÷ (4.196)
è Nd ø
For a symmetric beam, the half-power beamwidth is thus
æ λ ö
θ HPBW = 2sin -1 ç 0.44 ÷ (4.197)
è Nd ø
Similarly, the null-to-null beamwidth is given in terms of the location of the first
null, given by (4.188) with n = 1, resulting in
4.5  Antenna Arrays 123

æ λ ö
θ NNBW = 2sin -1 ç (4.198)
è Nd ÷ø

If the array is large such that N is a large number, the arguments of both (4.197) and
(4.198) will be small, and the beamwidths can be approximated by
λ
θ HPBW » 0.88 (4.199)
Nd

λ
θ NNBW » 2 (4.200)
Nd

4.5.4  Phased Arrays


The array architectures considered up to this point have all been cophasal for waves
incident broadside to the array; that is, a wavefront impinging on the array from
the broadside direction is incident on each element at the same time, and thus each
element receives the wave with the same phase. This is, in essence, the process that
causes the variation in the beam pattern, as the wavefronts received by the elements
become increasingly out of phase as the angle off broadside increases. If the ele-
ment spacing is wide enough, there are additional angles where the wavefront is
cophasal, giving rise to grating lobes.
The direction of the main beam can be steered by altering the phase of the indi-
vidual elements to make a wavefront cophasal at a specified direction q0. The phase
at each element is set to offset the phase shift due to the additional length the wave-
front travels to the elements at oblique angles, as shown in Figure 4.20. The current
at element n after phase shifting is then given by

In = Ie - jknd sin θ0 (4.201)

Wavefront

3dcosθ
2dcosθ θ
dcosθ

d d d

Figure 4.20  Linear phased array.


124 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Following as before, it can be shown that the array factor of a linear array
changes to
é1 ù
sin ê Nkd(sin θ - sinθ 0 )ú
ë 2 û
AF(θ) =
é1 ù
N sin ê kd(sin θ - sinθ 0 )ú (4.202)
ë2 û
where the array broadside direction is now along the z axis, and the cosines in
(4.184) are thus changed to sines. The locations of the pattern nulls, maxima, and
half-power point are also shifted:
æ nλ ö
θn = sin -1 ç sinθ 0 ± ÷ (4.203)
è Nd ø
æ mλ ö
θm = sin -1 ç sinθ 0 ± ÷ (4.204)
è d ø
æ λ ö
θh = sin -1 ç sinθ 0 ± 0.443 ÷ (4.205)
è Nd ø
Because the maxima of the array pattern are shifted by the sine of the point-
ing angle, grating lobes may be present even if the element spacing corresponds to
(4.191); in particular, maxima are now located at physical angles when

sin θ 0 ± < 1 (4.206)
d
for an array steering electrically over the whole hemisphere, –90° £ q0 £ 90°. To
restrict (4.206) to only the m = 0 solution, thereby avoiding grating lobes, the ele-
ment spacing should be defined such that

d 1 (4.207)
<
λ 2
If the array is to steer over –45° £ q0 £ 45°, the spacing can be increased to
d
< 0.59 (4.208)
λ
This formulation for a linear array extends directly to a planar array. The array
factor for a planar array is given in terms of the linear array factors determined by
the element spacings in the x and y directions:
é1 ù
sin ê Mkdx (sin θ cos φ - sinθ 0 cosφ 0 )ú
AFx (θ , φ) = ë2 û
é1 ù
M sin ê kdx (sin θ cos φ - sinθ 0 cosφ 0 )ú (4.209)
ë2 û
é1 ù
sin ê Nkdy (sin θ sin φ - sinθ 0 sinφ 0 )ú
AFy (θ , φ) = ë2 û
é1 ù
N sin ê kdy (sin θ sin φ - sinθ 0 sinφ 0 )ú (4.210)
ë2 û

where q0, f0 denote the direction of the main beam.


4.5  Antenna Arrays 125

4.5.5  Array Architectures

4.5.5.1  Signal Feeds


Transmission of a wavefront in a given direction requires a signal to be applied
to the inputs of the antennas comprising the array in such a way that the signal is
phase coherent across the array elements. Reception of a wavefront from a given
direction similarly requires a phase coherent combination of the signals received by
all the elements. The method of feeding a phase coherent signal to an array depends
in general on the formation of the signal by the transmitter hardware.
Traditionally, a single transmitter generates an analog signal that is sent through
a feed network, as shown in Figure 4.21(a), called a parallel feed network. The sig-
nal is divided equally among all the elements using a series of power dividers, and
the path to each element is typically designed to be approximately equal in length.
A circulator or switch placed at the feed network port can be used to incorporate
both transmission and reception functionality with the same array. Parallel feed net-
works have the benefit of the simplicity of a single input/output port, and thus only
one transmitter/receiver is required. However, such networks are lossy due to the
necessary power dividers and due to losses in the transmission lines or waveguides
used to implement the network, and thus the input power must generally be very
high for a large number of elements to transmit reasonable power. These losses can
be overcome somewhat by including additional amplifiers at the antennas; receiv-
ing applications typically require low-noise amplifiers at the antenna terminals to
ensure a large enough signal-to-noise ratio at the network output. Moreover, the
ability to steer the beam requires additional circuitry at each element to alter the
phase relationship between the elements.
A similar method to parallel feeding is a space feed, as shown in Figure 4.21(b),
which utilizes a single antenna that transmits a broad beam to an array of elements
formed into what can be described as an electronic lens or a reflect array [14].
The lens consists of an array of elements that receives signals from the transmit
antenna, phase shifts the signals, and transmits the signals on the other side of the
lens. In an electronic reflector, the antennas facing away from the feed antenna are
replaced by a short circuit so that the received signals are reflected and transmit-
ted back out the antenna array. The benefit of the space fed array is that lossy feed
networks are avoided, and the feed antenna can be moved, or multiple feeds can be
implemented to increase the bandwidth capability by utilizing true time delay beam
steering.
Series feed networks, as shown in Figure 4.21(c), align the elements in series
along a transmission line and couple varying amounts of power into each successive
element. The transmission line is then terminated in a matched load to eliminate
reflections. For an array with uniform amplitude at each element, the coupling
ratio is smallest at the first element and greatest at the last element. As the input
power is gradually reduced at each element, a larger coupling ratio ensures an equal
amount of power to each element. Series feed networks are narrowband, since the
distance traveled by the input signal between each element is predefined, and there-
fore changing the wavelength alters the phase relationship between the elements,
causing the beam to steer to a different direction. Due to this phase relationship, the
beam of a series-fed array can be steered by altering the input frequency.
126 Microwave and Millimeter-Wave Remote Sensing for Security Applications

(a)

(b)

Load
Couplers

(c)
Figure 4.21  (a) Parallel feed network. (b) Space feed network. (c) Series feed network. (d) Digital
architecture. (e) Digital subarray architecture.

As digital technology progresses, the ability to address array elements individu-


ally using a distributed series of direct digital synthesizers (DDSs) and analog-to-
digital converters (ADCs) is improving the steering capabilities of arrays. An array
that includes a DDS and ADC behind each element is referred to as a digital array,
and it feeds the antenna elements by individually generating phase-coherent signals
to the elements, as shown in Figure 4.21(d). Steering is thus accomplished through
phase shifting, or by adjusting the timing of each signal to match the additional
length ndsinq0/c such that the wavefronts are cophasal in a given direction. Digital
4.5  Antenna Arrays 127

TX TX TX TX
RX RX RX RX

(d)

TX TX
RX RX

(e)
Figure 4.21  (continued).

arrays can transmit and receive wideband signals using true time delay since the
delay between the elements is independent of the wavelength. There are also a
number of benefits to digitally processing the received signals from each array ele-
ment, including improved direction of arrival estimation [15, 16] and cancellation
of interfering signals [17]. Digital array architectures require a significant amount
of hardware to be implemented at each element of the antenna, since a digital trans-
mitter and receiver are required. Thus, the small element spacings at microwave and
millimeter-wave frequencies make the inclusion of such hardware in a large array
a significant challenge.
Subarray architectures include some benefits of digital architectures with par-
allel or series feed networks. The array is subdivided into subsets of elements that
are fed using parallel or series networks, which are each addressed using a separate
transmitter and receiver, as shown in Figure 4.21(e). True time delay beam steer-
ing can be used between the subarrays, while steering within a subarray is accom-
plished with phase or frequency steering.

4.5.5.2  Beam Steering


Steering of the beam can be accomplished by adjusting the phase in a parallel feed
network, the frequency in a series feed network, or the time delay between ele-
ments in a digital array. Phase scanning requires the use of phase shifters to control
the relative phase between the elements. Typical phase shifters are diode or fer-
rite based. Diode phase shifters use PIN diodes to switch the input signal between
128 Microwave and Millimeter-Wave Remote Sensing for Security Applications

different lengths of transmission lines, each length designed to impart a different


phase shift. Ferrite phase shifters switch the input signal between different lengths
of ferrite material that adjust the phase of the signal. Unlike diode phase shifters,
ferrite phase shifters are not reciprocal devices; thus, an array using the same ele-
ments for both transmit and receive must employ separate ferrite phase shifters for
both transmit and receive. Ferrite shifters can generally support higher power than
diode shifters.
Frequency steering techniques in series feed arrays are often implemented at
millimeter-wave frequencies where rectangular slots are cut into the sides of a wave-
guide to create an array of radiators. The benefit is the simplicity of the architecture;
a single input to a waveguide can be used to steer a beam by simply altering the
frequency. Due to the nature of the steering method, frequency steering is nar-
rowband and the information of interest must be frequency independent across the
bandwidth of interest. Frequency steering also works only in one dimension; thus,
for two-dimensional steering, multiple series-feed arrays must be implemented and
phase shifted in the orthogonal direction.
True time delay steering is implemented using digital synthesizers at each ele-
ment, or by using true time delay units at each element. Digital architectures require
more hardware at the element, making them impractical at higher frequencies. De-
lay line modules require long transmission lines to match the length between ele-
ments and are thus also generally bulky.

4.6  Common Microwave and Millimeter-Wave Antennas

This section describes the radiation characteristics of a few of the most common
antennas used in microwave and millimeter-wave remote sensing applications: the
horn, slot, microstrip, reflector, and lens antennas. The horn, slot, and microstrip
antennas can be examined using the equivalence principle described earlier. Deriva-
tions of the far field radiation for such antennas are generally straightforward and
are not covered here; the interested reader is referred to the references included in
the following sections for more detailed analyses. The reflector and lens antennas
may be considered to be antenna systems that collimate the wide beam of a feed
antenna into a narrower, more directive beam.

4.6.1  Horn Antennas


The horn antenna is one of the most ubiquitous antennas in microwave and
millimeter-wave engineering. Essentially a flared waveguide, the antenna is formed
by varying the angle of the walls of a waveguide to create a wider aperture. Due to
its simplicity and ease of construction, the horn antenna has been in extensive use
since the early twentieth century, and its radiation characteristics are well under-
stood. It is used as a universal calibration standard, due to the ability to accurately
predict its measured radiation characteristics, and as stand-alone antennas, feed
elements for reflectors and lenses, and array elements [18].
Rectangular horn antennas are either sectoral or pyramidal horns, as shown in
Figures 4.22(a)–4.22(c). A sectoral horn antenna is flared along only one dimension
4.6  Common Microwave and Millimeter-Wave Antennas 129

and is designated by the direction of the field parallel to the direction of the flare as
either an E-plane or H-plane horn. The pyramidal horn is flared in both directions.
Circular horn antennas are flared conically and are thus referred to as conical horn
antennas [see Figure 4.22(d)]. The radiated electromagnetic fields can be calculated
using the equivalence principle outlined in Section 4.4 [11]. The fields in the ap-
erture of the horn may be approximated by a rectangular or circular aperture and
derived as shown previously, where the radiation efficiency of horn antennas de-
pends on the construction material but is on the order of 80% for good rectangular
horn antennas. The half-power beamwidth and null-to-null beamwidth are given
approximately as [18]
λ
θ HPBW » 58.5° (4.211)
d
λ
θ NNBW » 101° (4.212)
d
Derivations for the expressions for the radiated fields from the sectoral and pyra-
midal horn antennas are given by Balanis [2]; the reader is referred there for more
details. The radiated electric field components of the pyramidal horn are

Eθ = jkE0 g(r)[ sin φ(1 + cosθ ) J1 J2 ] (4.213)

a
E

a
(a)

(b)

(c)

(d)
Figure 4.22  Horn antenna: (a) E-plane, (b) H-plane, (c) pyramidal, (d) conical.
130 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Eφ = jkE0 g(r)[ cos φ(1 + cosθ ) J1 J2] (4.214)


where
1 π L2 æ ì é L2 ùü
J1 = çè exp í j ê 2k (k sin θ cos φ + π a)ú ý
2 l î ë ûþ

{
´ [C(t2¢ ) - C(t1¢ )] - j [ S(t2¢ ) - S(t1¢ )] }
(4.215)
ì éL ùü
+ exp í j ê 2 (k sin θ cos φ + π a)ú ý
î ë 2k ûþ

{ })
´ [C(t2¢¢) - C (t1¢¢)] - j [ S (t2¢¢) - S (t1¢¢)]

π L1 ì æL öü
J2 = exp í j ç 2 k sin θ sin φ÷ ý (4.216)
l î è 2k øþ

{
´ [C(t2 ) - C(t1)] - j [ S(t2 ) - S(t1)] }

1 æ kb ö
t1 = - çè + k sin θ sin φ L1 ÷ (4.217)
π kL1 2 ø

1 æ kb ö
t2 = çè - k sin θ sin φ L1 ÷ (4.218)
π kL1 2 ø

1 é ka ù
t1¢ = - ê + (k sin θ sinφ + π a)L2 ú (4.219)
π kL2 ë 2 û

1 é ka ù
t2¢ = - (k sin θ sin φ + π a)L2 ú (4.220)
ê
π kL2 ë 2 û

1 é ka ù
t1¢¢ = - ê + (k sin θ sinφ - π a)L2 ú (4.221)
π kL2 ë 2 û

1 é ka ù
t2¢¢ = - (k sin θ sinφ - π a)L2 ú (4.222)
π kL2 êë 2 û
and C and S are the cosine and sine Fresnel integrals [13], given by
x
æπ ö
C(x) = ò cos ç t 2 ÷ dt (4.223)
è2 ø
0

x
æπ ö
S(x) = ò sin ç t 2 ÷ dt (4.224)
è2 ø
0
The antenna pattern is then found by (4.69).
4.6  Common Microwave and Millimeter-Wave Antennas 131

4.6.2  Slot Antennas


Slot antennas are commonly employed in frequency scanned, series-fed arrays by
cutting rectangular openings into the sidewalls of a waveguide, as shown in Figure
4.23, and they are able to handle very large power levels (on the order of 1 MW at
10 GHz [19]), making waveguide slot arrays useful in high-power radar applica-
tions. The slot openings are generally not perfectly rectangular; due to manufactur-
ing, the edges are often rounded. However, the radiation can be accurately modeled
by a rectangular aperture [20]. The radiation for a narrow slot can be evaluated
through Babinet’s Principle, which states that an aperture in an infinite ground
plane can be replaced by a conductor of equal size to the aperture placed in free
space [12]. The fields are then approximately those of a small dipole antenna. In
such a case, the antenna pattern is omnidirectional in azimuth.
The radiated fields of the slot antenna are given by [20]

Eθ = - jkabE0 g(r)cos φh(θ , φ) (4.225)

Eφ = jkabE0 g(r)cos θ sinφ h(θ , φ) (4.226)

where
cos(bk sin θ sin φ)
h(θ , φ) = sinc(ak sinθ cos φ) (4.227)
(bk sinθ sin φ)2 - (π 2)2

d d

Input signal

Input signal

d θ

Figure 4.23  Slotted waveguide arrays.


132 Microwave and Millimeter-Wave Remote Sensing for Security Applications

If the slot is narrow such that ak << 1 and the length is b = l /2, (4.227) reduces to
æπ ö
cos ç sin θ sin φ ÷
è2 ø
h(θ , φ) » (4.228)
1 - sin2 θ sin2φ

4.6.3  Microstrip Antennas


A microstrip antenna is formed by a thin metal plate on a conductor-backed dielec-
tric substrate. Printed microstrip antennas, sometimes called patch antennas, are
simple and cheap to manufacture and can be formed in arbitrary shapes, making
them a very versatile solution to many remote sensing problems. Because the metal
of the antenna is thin, they can be made to conform to curved surfaces; the simplic-
ity of construction also lends itself to the creation of antenna arrays. The antenna
elements can be manufactured to tight tolerances, benefiting the construction of
antennas and arrays in the millimeter-wave band. For these reasons, microstrip
antennas are being used more often in remote sensing. Despite their simplicity,
however, challenges to their implementation exist, primarily in the integration of
the antenna with front-end electronics such as amplifiers, phase shifters, and circu-
lators for arrays. At millimeter-wave frequencies, these electronics must fit into an
area on the order of l2/4, along with the feed network, which can be a significant
challenge.
Microstrip antennas can be individually fed in a number of ways, as shown in
Figure 4.24. The signal can be input directly from a microstrip transmission line
on the surface, by a coaxial connector whose conductor is vertically connected to
the underside of the antenna, or via coupling from an underlying microstrip line or
aperture. The physical implementation of an array often helps determine the most
practical and beneficial methods of feeding the elements. Figure 4.25(a) shows a se-
ries feed network using a coupled feeding structure where the underlying microstrip
line feeds progressively reduce the power levels to each element. Figure 4.25(b)
shows a parallel feed network with a surface microstrip feed network.
One result of the discussion of image theory was that electric currents close to
a conducting plane will be cancelled by their image currents. Thus, the fact that a
microstrip antenna can radiate may seem paradoxical. The radiation generated by
the antenna is a result of the antenna patch forming a resonant cavity between the
metal and the conducting plane; the characteristics of the dielectric and the size of
the patch affect the resonance. In the analysis of the antenna as a resonant cavity,
it is found that induced fields are formed on the walls of the cavity, which can be
converted to current densities using the equivalence principle [21]. The radiation
efficiency of microstrip antennas depends primarily on the permittivity of the dielec-
tric between the patch antenna and the conductor backing, and can vary from 70%
to 90%, depending on the dielectric.
Microstrip antennas can be made in virtually any planar shape. The most com-
mon shapes are rectangular and circular. A derivation of the radiated fields requires
an analysis of the cavity model, which will not be covered here. The radiated elec-
tric fields from a rectangular patch antenna of width W, length L, and height above
the conducting ground plane h are given by Jackson in [21] as
4.6  Common Microwave and Millimeter-Wave Antennas 133

L h

(a)

Signal Microstrip Antenna


Microstrip feed

Antenna
Signal Microstrip Coupled feed

Antenna
Aperture-coupled feed
Signal
Microstrip
Antenna
Coaxial feed
Coaxial
Signal

(b)
Figure 4.24  (a) Microstrip antenna. (b) Feeding methods.

Eθ = jkηWhE0 cos φRTM (θ ) (4.229)

æ1 ö
ø
æ1
è2
ö
(
´ cos ç Lk sin θ cos φ ÷ sinc ç Wk sin θ sin φ ÷ tanc hk ε r µ r - sin2 θ
è2 ø )
Eφ = - jkηWhE0 cos θ sin φRTE (θ)

æ1 ö
ø
æ1
è2
ö
(
´ cos ç Lk sin θ cos φ÷ sinc ç Wk sin θ sin φWk sinθ sin φ ÷ tanc hk εr µr - sin 2 θ
è2 ø )
(4.230)
where
-1


é æ
RTM (θ) = 2 ê1 + j ç ε r µ r - sin2 θ
è
sec θ ö
εr ÷
ø
( ù
tan hk ε r µ r - sin2 θ ú ) (4.231)
ë û
134 Microwave and Millimeter-Wave Remote Sensing for Security Applications

(a)

(b)
Figure 4.25  (a) Series fed microstrip array. (b) Parallel fed microstrip array.

-1
é ù
RTE (θ) = 2 ê1 + j
ê
æ
ç
µ r cos θ ö
2 ÷
(
tan hk εr µ r - sin2θ ) ú
ú
(4.232)
ë è ε rµ r - sin θ ø û

and er, mr are the relative permittivity and permeability of the dielectric.

4.6.4  Reflector Antenna Systems


Reflector antenna systems consist of a feed antenna, commonly a horn antenna, di-
rected at one or more reflecting surfaces [11, 22–25]. The reflecting surface typically
has very low reflection loss and may be of various shape; however, planar, spherical,
and particularly parabolic reflectors are the most common. The reflector serves to
collimate the relatively wide beam of the small feed antenna to generate a narrow
beam in the direction of reflection by creating a planar wave. Large reflecting sur-
faces thus effectively increase the aperture size of the antenna system, resulting in
higher directivity. Reflector antenna systems have been in use since the early part
of the twentieth century for UHF frequencies and above; in remote security sens-
ing applications, reflectors are commonly used at millimeter-wave frequencies. In
imaging applications, discussed further in Chapter 8, the reflecting surface can be
made to be physically separated from the feed antenna and placed on a mechani-
cal rotator. The position of the reflector is then adjusted to steer the beam over the
imaging area. This allows the sensitive millimeter-wave hardware to be placed in
a fixed location, separate from the mechanical vibrations due to the rotator. When
the feed antenna is physically connected to the reflector, the entire antenna system
4.6  Common Microwave and Millimeter-Wave Antennas 135

is rotated to support imaging. Alternatively, the feed antenna may be rotated and
the reflector fixed. If a parabolic reflector is used, the scan volume is limited, on the
order of a few degrees in any direction. Any scanning away from the primary focus
of the parabola results in a loss of directivity, which can be on the order of 3–8 dB
over a scan volume of 10 beamwidths [24]. The scan areas for planar and spherical
reflectors do not include this limitation; however, the directivity is less than that of
a parabolic reflector. Multiple feed elements may also be implemented to support si-
multaneous beams. Common reflector antenna systems are shown in Figure 4.26.
Reflector antenna systems can be divided generally into two categories: those
with a single reflector, and those with multiple reflectors. Single reflector systems
utilize a feed antenna directed at the primary reflector, which collimates the beam.
This has the least amount of hardware but suffers from potential problems of locat-
ing the feed antenna and the transmitter or receiver in a place that does not signifi-
cantly affect the radiation characteristics of the antenna system. Center-fed reflector
systems position the feed antenna at the focal point of the reflector, where, for a
parabolic reflector, the surface is formed on the vertex of the parabola; the signals
must then be transported to the feed antenna over a transmission line, which is lossy
at millimeter-wave frequencies. Additionally, passive millimeter-wave systems suffer
increased system noise if the initial amplifier is not close to the receiving antenna,
while millimeter-wave radar systems benefit from including both the transmit and
receive hardware near the feed antenna. Thus, a single reflector system must po-
sition the feed antenna and hardware in an offset configuration or suffer from

Primary
Reflector
Secondary
Reflector Sensor
Reflector

(a) (c)

Primary
Reflector
Reflector
Secondary
Reflector

Sensor
(b) (d)
Figure 4.26  Single reflector antenna systems: (a) center-fed reflector, and (b) offset-fed reflector.
Double reflector antenna systems: (c) Cassegrain, and (d) Gregorian.
136 Microwave and Millimeter-Wave Remote Sensing for Security Applications

increased loss and noise to transfer the signals between the receiver and the feed ele-
ment in a center-fed configuration. Offset-fed reflectors also alleviate the problems
of blockage due to the feed antenna and reflections from the vertex of the parabola
back into the feed antenna. Double reflector antenna systems alleviate the problems
of feed antenna placement by including a smaller, secondary reflector. The feed
antenna and associated hardware can then be placed behind the primary reflector
where space is less constrained. Less weight is then located in front of the reflector,
easing mechanical steering constraints. Typical double reflector systems include the
Cassegrain and the Gregorian reflector systems. The Cassegrain system consists of
a parabolic primary reflector and a hyperbolic secondary reflector, whereas the Gre-
gorian system includes a parabolic primary and an ellipsoidal secondary reflector.
Reflector antenna systems suffer from a number of efficiency problems due to
the location and beam shape of the feed antenna, and the maximum efficiency for a
standard reflector is on the order of 81%. The directivity of center-fed antennas us-
ing either a feed antenna or a secondary reflector is reduced due to the blockage of
the feed antenna or secondary reflector. Energy transmitted by the feed is reflected
off the vertex of the reflector and is incident back on the feed. The effect can be
characterized approximately by reducing the aperture area of the reflector by the
blockage area. The beam shape of the feed antenna can also cause a reduction in
efficiency: generally, the beam pattern of the feed element is wider than the angle
subtended by the reflector in order to create the proper amplitude distribution on
the reflecting surface. The energy in the beam pattern that is not incident on the
reflector is called spillover loss and reduces the efficiency of the antenna system.
Diffraction of waves from the edge of the primary and secondary reflector creates
increased energy in certain direction, resulting in increased sidelobes. Additionally,
inaccuracies in the surface, due to mechanical tolerances or damage, result in phase
variations across the surface, resulting in reduced directivity.

4.6.5  Lens Antenna Systems


To create a uniform phase front, the fields radiated by an antenna can also be col-
limated using a lens [11, 23, 26, 27], as shown in Figure 4.27. The function of the
lens in a microwave or millimeter-wave antenna system is analogous to that of an
optical system in that it collimates transmitted waves and focuses received waves
onto the focal point of the lens, which is located in the feed antenna. Microwave
and millimeter-wave lenses are commonly made of dielectric, although lenses can
also be manufactured out of different lengths of waveguide sections. A lens system

(a) (b)
Figure 4.27  (a) Lens antenna. (b) Lens antenna with planar outer surface.
4.6  Common Microwave and Millimeter-Wave Antennas 137

has no blockage problems that are prevalent in reflector systems, but the loss in
the dielectric is generally higher than the loss incurred by a reflecting surface. Lens
antennas have additional degrees of freedom in the engineering process, includ-
ing multiple lens surface and the dielectric composition, which can be designed to
achieve higher performance.

References

  [1] Jackson, J. D., Classical Electrodynamics, 3rd ed., Hoboken, NJ: John Wiley & Sons,
1999.
  [2] Balanis, C. A., Antenna Theory: Analysis and Design, 3rd ed., Hoboken, NJ: Wiley-
Interscience, 2005.
  [3] “IEEE Standard Definitions of Terms for Antennas,” IEEE Std 145-1983, 1983, p. 0_1.
  [4] Volakis, J. L., “Fundamentals of Antennas, Arrays, and Mobile Communications,” in An­
tenna Engineering Handbook, J. L. Volakis, Ed., 4th ed., New York: McGraw-Hill, 2007.
  [5] Mathis, H. F., “A Short Proof that an Isotropic Antenna Is Impossible,” Proceedings of the
IRE, Vol. 39, 1951, p. 970.
  [6] Mailloux, R. J., F. K. Schwering, A. A. Oliner, and J. W. Mink, “Antennas III: Array, Mil-
limeter Wave, and Integrated Antennas,” in Handbook of Microwave and Optical Compo­
nents, K. Chang, Ed., New York: John Wiley & Sons, 1989.
  [7] Kraus, J. D., Antennas, New York: McGraw-Hill, 1950.
  [8] King, R. W. P., The Theory of Linear Antennas: With Charts and Tables for Practical Ap­
plications, Cambridge, MA: Harvard University Press, 1956.
  [9] Aharoni, J., Antennae, Oxford: Oxford University Press, 1946.
[10] Schelkunoff, S. A., “Some Equivalence Theorems of Electromagnetics and Their Applica-
tion to Radiation Problems,” Bell System Tech. Journal, Vol. 15, 1936, pp. 92–112.
[11] Schelkunoff, S. A., Antennas: Theory and Practice, New York: John Wiley & Sons,
1952.
[12] Harrington, R. F., Time-Harmonic Electromagnetic Fields, New York: McGraw-Hill,
1961.
[13] Abramowitz, M., and I. A. Stegun, Handbook of Mathematical Functions, New York:
Dover, 1965.
[14] Frank, J., and J. D. Richards, “Phased Array Radar Antennas,” in Radar Handbook, M. I.
Skolnik, Ed., New York: McGraw-Hill, 2008.
[15] Schmidt, R., “Multiple Emitter Location and Signal Parameter Estimation,” Antennas and
Propagation, IEEE Transactions on, Vol. 34, 1986, pp. 276–280.
[16] Paulraj, A., R. Roy, and T. Kailath, “Estimation of Signal Parameters Via Rotational In-
variance Techniques—Esprit” in Nineteeth Asilomar Conference on Circuits, Systems and
Computers Nov. 6–8 , 1985, pp. 83–89
[17] Manolakis, D. G., V. K. Ingle, and S. M. Kogon, Statistical and Adaptive Signal Processing,
Norwood, MA: Artech House, 2005.
[18] Bird, T. S., and A. W. Love, “Horn Antennas,” in Antenna Engineering Handbook, J. L.
Volakis, Ed., 4th ed., New York: McGraw-Hill, 2007.
[19] Gilbert, R. A., “Waveguide Slot Antenna Arrays,” in Antenna Engineering Handbook,
J. L. Volakis, Ed., 4th ed., New York: McGraw-Hill, 2007.
[20] Compton, R. T., and R. E. Collin, “Slot Antennas,” in Antenna Theory Part 1, R. E. Collin
and F. J. Zucker, Eds., New York: McGraw-Hill, 1969.
[21] Jackson, D. R., “Microstrip Antennas,” in Antenna Engineering Handbook, J. L. Volakis,
Ed., 4th ed., New York: McGraw-Hill, 2007.
138 Microwave and Millimeter-Wave Remote Sensing for Security Applications

[22] Sletten, C. J., “Reflector Antennas,” in Antenna Theory Part 2, R. E. Collin and F. J.
Zucker, Eds., New York: McGraw-Hill, 1969.
[23] Bodnar, D. G., J. J. Lee, G. L. James, F. K. Schwering, and J. W. Mink, “Antennas II: Re-
flector, Lens, Horn, and Other Microwave Antennas of Conventional Configuration,” in
Handbook of Microwave and Optical Components, Vol. 1, K. Chang, Ed., New York: John
Wiley & Sons, 1989.
[24] Rahmat-Samii, Y., “Reflector Antennas,” in Antenna Engineering Handbook, J. L. Volakis,
Ed., 4th ed., New York: McGraw-Hill, 2007.
[25] Cooley, M. E., and D. Davis, “Reflector Antennas,” in Radar Handbook, M. I. Skolnik,
Ed., 4th ed., New York: McGraw-Hill, 2008.
[26] Brown, J., “Lens Antennas,” in Antenna Theory Part 2, R. E. Collin and F. J. Zucker, Eds.,
New York: McGraw-Hill, 1969.
[27] Bodnar, D. G., “Lens Antennas,” in Antenna Engineering Handbook, J. L. Volakis, Ed., 4th
ed., New York: McGraw-Hill, 2007.
Chapter 5

Receivers

Security sensing encompasses a broad range of applications, from human pres-


ence detection to concealed object detection to moving object classification, among
others. The sensors used for these various applications differ in many ways; some
may transmit a coherent signal whereas others detect thermal radiation; some may
use antenna arrays whereas others may use single aperture antennas. A common
feature of all sensors is the use of a receiver, which detects the signal and converts
it to a form suitable for signal processing or data recording. In theory, a receiver
in its simplest form consists of an antenna followed by a data recorder, such as an
analog-to-digital converter (ADC), however in practice the signal must be “condi-
tioned” before the ADC. At the least, this conditioning includes impedance match-
ing and amplifying the signal to appropriate levels for the input of the ADC. At low
enough frequencies, direct digital conversion of the signal at the carrier frequency
can feasibly be achieved, as long as the signal frequency does not exceed half the
input frequency range of the ADC for Nyquist sampling. However, currently avail-
able ADCs have input bandwidths which do not exceed a few GHz at the most,
thus for sensors operating at microwave frequencies and millimeter-wave frequen-
cies the signal must be converted to a lower carrier frequency which can be digitally
converted. This increases the complexity of the system, often necessitating the use
of a non-linear frequency conversion mixer, and also requiring other components
such as filters and amplifiers.
Increased system complexity from the additional components in the receiver has
the obvious disadvantage of increased physical size and increased cost, but in addi-
tion, the integrity of the input signal degrades as the number of components increases.
In designing a security sensor it is necessary to understand how the hardware affects
the received signal and the limitations of each component in the receiver chain. Com-
ponents will introduce noise which can mask small signals, and distortions which can
alter the signal. Additionally, spurious signals will be generated, some of which may
fall within the bandwidth of the system. All of these factors will affect the perfor-
mance of the system and its ability to reliably detect the input signal.
This chapter discusses design considerations for receiver systems, covering as-
pects relevant to microwave and millimeter-wave receiver architectures. The topics
covered are relevant to receivers in general, as specific receiver architectures per-
taining to radiometry, radar, and imaging are considered in the following chapters.
General receiver system operation is first outlined, giving a brief overview of the
functionality of a receiver. Sources of internal noise are characterized in the follow-
ing section, including thermal, shot, and flicker noise. The discussion of thermal
noise is particularly relevant to the proceeding section on the overall system noise
of a receiver, which includes all internal noise sources and which is characterized

139
140 Microwave and Millimeter-Wave Remote Sensing for Security Applications

in terms of an equivalent thermal noise. Following this, a section on the linearity of


receivers covers aspects arising from non-linear components within the system, such
as gain compression, intermodulation products, and dynamic range.

5.1  General Operation of Receivers

The function of a receiver is to detect an electromagnetic signal within a specified


frequency bandwidth and provide a measure of one or more of its properties. Re-
ceivers are used in both active and passive systems; in the former, the detected signal
is transmitted by the sensor or a system related to it and the receiver measures the
return signal which is reflected off the object of interest; in the latter the signals may
be of thermal origin and thus generated and transmitted by the source itself, or the
signal may be transmitted by a system separated from the receiver, and its reflected
signal measured. In any case, the purpose of the receiver is to provide a reliable
reconstruction of some property of the input signal, and it may include circuitry to
detect the signal amplitude, phase, or perform correlations between separate receiver
channels or separate frequencies. The reconstructed signal is, however, inevitably
altered by intrinsic properties of the receiver hardware including bandwidth, conver-
sion gain, intrinsic noise, and other distortion processes such as signal compression
and intermodulation distortion. Therefore, an accurate description of these intrinsic
processes is necessary to design the receiver to reliably detect the signal. That is,
when the signal passes through the receiver and is detected, it must be known which
properties of the detected signal are the results of processes within the receiver, and
which properties are intrinsic to the signal itself. Considering these processes, the
system can be designed in order to minimize their effects as much as possible.
An ideal receiver can be described in general as a two-port network, as given in
Figure 5.1, the input of which is the received signal Si(f ) and the output of which is
a modified reconstruction of the input signal, given by

So (f ) = H(f )Si (f ) (5.1)


The output signal is modified by the transfer function of the receiver H(f ) which
alters the signal bandwidth and adds power gain or loss to the signal. The receiver
may also shift the center frequency of the signal, typically reducing the frequency
as in the case of a downconverting receiver. If the receiver is ideal and downconver-
sion does not take place, the output signal power is simply the input signal power
modified by the conversion gain of the receiver. In real systems, however, the out-
put signal is corrupted by factors which can generally be described as either noise
or distortion. The following sections will discuss the sources of noise in a receiver
system, signal distortions introduced by the receiver, and the effects of these factors

Si(f) So(f)
H(f)

Receiver
Figure 5.1  General receiver modeled as a two-port network.
5.1  General Operation of Receivers 141

on the receiver’s ability to accurately reconstruct the input signal at the output of
the network.
The receiver system consists of a number of different individual components,
and is defined in general as the cascade of components between the antenna and
the data recorder. Although the antenna is crucial to the operation of the receiver,
its analysis is distinct enough from that of the rest of the receiver that it was char-
acterized separately in Chapter 4. Receiver components can be grouped into two
categories: passive components, such as filters, attenuators, transmission lines, cou-
plers, etc . . . and active components, such as amplifiers, oscillators, etc . . . There
are numerous references covering each of these components in detail [1-6]; the
focus in this section is the use of components within a system and how they affect
the system response.
The most common receiver architecture found in modern sensors is the super-
heterodyne receiver. A block diagram of a typical superheterodyne architecture is
shown in Figure 5.2. The primary distinguishing feature of the heterodyne re-
ceiver is the downconversion of the signal of interest to a frequency which can be
more easily handled by the hardware. Components operating at microwave and
millimeter-wave frequencies are typically less efficient and more expensive than
those operating below a few GHz. By downconverting, the more efficient low fre-
quency components can be leveraged. In addition, components which have the
performance required may not exist at the desired microwave or millimeter-wave
frequency. For example, the signal in a radiometer operating at 30 GHz can be
downconverted to 500 MHz where amplifiers and filters are significantly cheaper.
In this way, more gain can be added to the signal at a lower cost than having a
high-gain amplifier at the upper frequency. This scheme still requires mixers and
oscillators operating at or near the upper frequency, which will tend to be more
expensive. In addition, as will be seen in the section on the noise performance of a
cascaded network, a high gain low-noise amplifier may be required at the output of
the antenna to reduce the overall system noise.
The front-end portion of the receiver is referred to as the radio frequency (RF)
section, and the center frequency of interest is fRF, which is detected by the antenna.
The signal is amplified after the antenna using a low-noise amplifier (LNA), and the
amplifier frequency response filters the received RF signal; a bandpass filter (BPF) is
also often used at the RF frequency. The signal then passes to a mixer which is also
fed with a local oscillator (LO) signal f­LO. The mixer is a non-linear device, and
generates the frequencies mfRF ± nfLO, where m = 0, 1, 2, …, n = 0, 1, 2, … The LO
and RF signals are passed by the device, as are sums and differences of multiples

Antenna
fRF fIF fBB
LNA BPF Mixer BPF LNA Detector LPF

fLO
LO

Figure 5.2  A typical superheterodyne receiver architecture.


142 Microwave and Millimeter-Wave Remote Sensing for Security Applications

ADC

RF Section IF Section Detection


Figure 5.3  A superheterodyne architecture employing digital detection.

of each signal. The filter after the mixer selects the frequency of interest from these
various frequencies. For a downconverting receiver, the desired frequency is one or
both of fRF ± f­LO; this is called the intermediate frequency (IF) fIF. The IF signal may
be amplified prior to detection. The detector, usually a square-law device, squares
the voltage signal generating a voltage signal proportional to the signal power. The
output voltage of the detector is at the baseband frequency, and a low-pass filter
may be used to integrate the signal and reduce noise fluctuations.
A more typical heterodyne architecture used in modern sensors is shown in
Figure 5.3 where the IF signal is digitized using an ADC and the detection process
is performed in software on the digitized signal. Digitizing the IF signal allows
significant versatility; multiple detection schemes can be used in conjunction or
the detection scheme can be dynamically altered. An ideal digitized receiver could
consist simply of an antenna connected to an ADC, where amplification, filtering,
and detection would all occur in software. However, current state-of-the-art high-
performance ADCs can only convert signals up to a few GHz at the maximum, and
typical frequencies for more general and widely available ADCs are in the hundreds
of MHz. Thus, microwave and millimeter-wave sensors will still require hardware
to downconvert the RF signal to a frequency range that can be detected by the
ADC.
The receiver hardware produces a signal output of the general form

so (t) = A(t)cos [2π ft + φ (t)] (5.2)


where A(t) represents a time-dependent amplitude variation and f(t) represents a
time-dependent phase, or frequency, variation. The output signal thus represents
only the real component of a complex received signal. In order to reproduce a com-
plex input signal, the imaginary component must also be detected. This is accom-
plished with a complex demodulator (shown in Figure 5.4) that mixes the received

sI(t)

s(t)

90
o
sQ(t)

Figure 5.4  Complex demodulator


5.2  Receiver Noise 143

signal with both an in-phase local oscillator signal and a local oscillator signal
in quadrature with the in-phase oscillator signal, or shifted by 90˚ in phase. The
received signal is split into two output paths, and the local oscillator is also split,
with one channel shifted by a 90˚ hybrid. The output of the two mixers are referred
to as the in-phase and quadrature signals, given by

sI (t) = A(t)cos [2π ft + φ (t)] (5.3)

sQ (t) = A(t)sin [2π ft + φ (t)] (5.4)


The in-phase and quadrature outputs can then be combined to recover the
complex received signal

so (t) = sI (t) + jsQ (t) = A(t)e - j 2π ft +φ (t) (5.5)

5.2  Receiver Noise

The ability of a receiver to detect small amplitude signals is limited by the presence
of noise. The antenna collects noise signals which are not of interest and the receiver
itself generates noise; the noise is additive and its sum is referred to as system noise.
Noise signals collected by the antenna include galactic noise, atmospheric noise, and
thermal emission from the surrounding environment within the antenna beam. The
hardware within the receiver generates noise due to thermal agitation of electrons
in conductors, shot noise of electrons passing through semiconductor junctions, hot
electron noise due to current fluctuations from electrons in strong electric fields,
flicker noise due to device imperfections, and a number of others. The internally
generated noise can be such that the change in output power between when the
desired signal is present and absent is indistinguishable from the noise itself. Thus it
is important to evaluate the system noise because it affects the minimum detectable
signal level of the receiver.
The antenna noise power was discussed in Chapter 4 and is given by

PA = kTA Df (5.6)

where k = 1.38 ´ 10–23 J·K–1 is Boltzmann’s constant, TA is the antenna noise tem-
perature, and Df is the frequency bandwidth. PA is band-limited Gaussian white
noise; in radiometric observations the signal of interest is itself broadband thermal
noise and it has the same statistical characteristics as the unwanted noise signals.
Figure 5.5 shows a general diagram of a noisy receiver. The noise power present
at the antenna terminals is modified by the conversion gain of the receiver, and the
output power of the receiver, without including the noise generated in the receiver
hardware, is given by

Po, signal = GPA (5.7)


144 Microwave and Millimeter-Wave Remote Sensing for Security Applications

PA
G, Pr Po

Receiver
Figure 5.5  Diagram of a general receiver.

where G is the conversion gain of the receiver and may be either gain or loss. The
components of the receiver also introduce noise power Pr which is additive to the
antenna noise power, and the total receiver output power is given by

Po = G(PA + Pr ) (5.8)
In a similar fashion to the antenna noise power, the receiver noise power Pr can
be defined in terms of its noise temperature Tr by

Pr = kTr Df (5.9)
The receiver noise temperature is often characterized by assuming that the receiver
is noiseless and that a noisy resistor is placed at the input. The resulting equivalent
temperature of the resistor which produces a noise power equal to the receiver noise
is then obtained; this is called the equivalent noise temperature Te.
The system output power due to the antenna temperature at the input to the
receiver and the noise power generated within the receiver is then given by

Po = Gk(TA + Tr )Df = GkTsys Df (5.10)


where Tsys is the system equivalent noise temperature arising from the signal noise
power and the undesired noise powers. Tsys is not a measure of the physical tem-
perature of the receiver; rather, it is a representation of the noise defined in terms of
the thermal noise generated by a system in thermal equilibrium at temperature Tsys,
as will be seen in Section 5.3.

5.2.1  Sources of Receiver Noise


Whereas the noise present on the antenna is due to external sources, the noise power
generated by the components in the receiver is the result of a number of different
physical processes. The most common types of noise powers encountered in micro-
wave and millimeter-wave receivers are thermal noise, shot noise, and flicker noise.

5.2.1.1  Thermal Noise


Electrons present in a conducting material in thermal equilibrium at temperature T >
0 are in constant motion due to thermal agitation. The fluctuations of the electrons
induce a variation in voltage which is random in time and normally distributed in
amplitude. Figure 5.6 shows the time-varying voltage across a resistor in thermal
equilibrium. Because the kinetic energy of the electrons is proportional to the physical
5.2  Receiver Noise 145

Figure 5.6  Voltage fluctuations over time in a resistor due to thermal agitation.

temperature and because of the random, noise-like nature of the voltage over time,
the resulting signal is referred to as thermal noise. This was first observed by Johnson
in 1928 [7], and is therefore also referred to as Johnson noise. The RMS value of the
voltage fluctuations is also proportional to the temperature. Nyquist [8] showed that
the RMS voltage across a resistor of resistance R in thermal equilibrium is

2
Vrms = 4RkT Df (5.11)
When the resistance is terminated in a matched load, the thermal noise power
is
2
Vrms
Nthermal == kT Df (5.12)
4R
Thus components present in a receiver will introduce noise power proportional to
their physical temperature and bandwidth. Equation (5.12) is valid for any resis-
tance terminated with a matched load, and thus the resistor may be replaced by
an antenna with resistance R, which then results in the antenna noise power (5.6).
Note also that (5.11) implies that the material must have resistance to produce a
thermal noise voltage: as R®0 the voltage fluctuations cease. Thus a purely reactive
material or component produces no thermal noise.

5.2.1.2  Shot Noise


Shot noise arises from the discrete nature of the electrons which traverse a semicon-
ductor diode junction; the name arises from the similarity of the physical process
to shot pellets hitting a flat surface. If the junction length is short compared to the
wavelength, the current spikes due to the travelling electrons are approximately
impulses. Although the time-average value of the current is equal to the dc current,
the instantaneous impulses fluctuate randomly, following a Poisson distribution.
The RMS fluctuation of the current is [3]

2
Irms = 2qI Df (5.13)
where q is the electron charge, I is the current in the resistive part of the diode, and
Df is the bandwidth. The noise power of (5.13) is
η
N shot =
kT Df (5.14)
2
where h is the diode ideality factor, which is typically close to unity, and T is the
diode temperature. Thus the noise power due to shot noise is also thermally depen-
dent, and an equivalent shot noise temperature can be defined as
146 Microwave and Millimeter-Wave Remote Sensing for Security Applications

η
Tshot = T (5.15)
2
and thus

N shot = kTshot Df (5.16)


Because shot noise and thermal noise are both dependent on the temperature,
they are often combined into a single thermal noise process. In such a case, the two
noise processes are not separated when characterizing or measuring the overall
thermal noise of a receiver.

5.2.1.3  Flicker Noise


Flicker noise arises from imperfections on semiconductor contacts, impurities in the
junction area, electromagnetic radiation, and quantum noise, among other sources
[3]. Flicker noise is proportional to f –1 and is also called 1/f noise. Because of this
frequency dependence, flicker noise is particularly important at lower frequencies,
but beyond 1 MHz thermal noise usually dominates.

5.2.2  Equivalent Noise Bandwidth


The frequency passband of a receiver is often assumed to be ideal, with a perfectly
rectangular spectral shape, when discussing receiver characteristics from a theoreti-
cal standpoint. However, the passband of a real system will not be ideal and will
spread into the out-of-band regions, and thus a rectangular passband cannot always
be assumed. Because thermal noise is independent of frequency, assuming a perfectly
square passband results in an integration of less noise power than is present in a real
system: the actual frequency response will roll off gradually at the desired frequency
cutoff values, and the noise power will extend beyond the desired passband, resulting
in a wider noise bandwidth and more noise power present in the system than would
be seen for a rectangular passband. This can be seen graphically in Figure 5.7 which
shows ideal and non-ideal passbands. For low-order filters which result in a gradual
roll-off of 20 or 40 dB/Decade, the difference in noise power is more significant than
for filters with a steep frequency roll-off. The equivalent noise bandwidth is the ideal
square passband whose integrated noise power is equal to the integrated noise power
of the non-ideal filter passband and is defined for a general passband H(f ) by [9]
2
é ¥ H(f )df ù
ê ò0 úû
Dfe = ë ¥ (5.17)
ò
2
H (f )df
0

Ideal passpand

Actual passpand
f
Figure 5.7  Ideal and non-ideal system passbands.
5.2  Receiver Noise 147

|H(f)|2
3 dB
Ho 2

f
fc ∆fe

Figure 5.8  The equivalent noise bandwidth is the ideal passband whose integrated spectral power
is equal to that of the non-ideal system passband.

For a low-pass filter, the equivalent noise bandwidth is defined by

¥ 2
H(f )
Dfe = ò Ho
df (5.18)
0

where Ho is the maximum value of the filter response H(f ). The equivalent noise
bandwidth is shown graphically in Figure 5.8. The square passband produces a
total noise power equal to that of the non-ideal passband.
As an example, consider a single-pole RC low-pass filter, the transfer function
of which is given by

1 (5.19)
H(f ) =
f
1+ j
fc

where

1
fc = (5.20)
2π RC

is the filter cutoff frequency. The equivalent noise bandwidth is then


¥
fc2
Dfe = ò fc2 + f 2
df
0

¥
-1 æ f ö
= fc tan çè f ÷ø
c 0 (5.21)

π
= fc
2
» 1.57 fc

This states that for a single-pole RC filter, the bandwidth used to characterize the noise
power in the system should be 57% greater than the filter cutoff frequency. Table 5.1
gives the equivalent noise bandwidths for higher order filters. As the number of poles
148 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Table 5.1  E quivalent noise bandwidths Dfe/fc


and noise power error DPn
Number of filter poles Dfe/fc DPn (dB)
1 1.57 1.96
2 1.22 0.86
3 1.15 0.61
4 1.13 0.53
5 1.11 0.45

increases, Dfe /fc®1, and for filters with 4 or more poles, the noise power calculated
using fc­­ is a good estimate of the actual noise power generated in the system.
The difference in noise powers obtained when using the equivalent noise band-
width Dfe and the filter bandwidth can be found by

10log10 (DPn ) = 10log10 (kT Dfe - kTfc ) = 10log10 (Dfe fc ) (5.22)


The difference in noise power for a single-pole filter is 10log10(1.57) = 1.96 dB. The
noise power differences DPn for higher-order filters are also given in Table 5.1. For fil-
ters of order greater than 4, the difference in power between using the 3 dB bandwidth
of a non-ideal passband and the equivalent noise bandwidth is less than 0.5 dB.

5.2.3  Thermal Noise at Millimeter-Wave Frequencies


The previous discussions of thermal noise power assumed that the frequency of
interest was low enough and the temperatures high enough that the thermal noise
power was directly proportional to the temperature. In reality, this assumption
is an approximation, although one that holds well at frequencies below 1 THz.
However, when temperatures are close to 290 K and the frequency is above 1 THz,
this approximation is no longer valid. Returning to Nyquist’s derivation of the
noise power of a resistor, the noise power delivered over a transmission line to a
matched load can be given in general by

N n = ε Df (5.23)
where e is the energy associated with the mode in the transmission line of frequency
f. Comparing (5.23) to (5.12) it can be seen that e = kT in Nyquist’s derivation,
which holds true for low millimeter-wave frequencies. Using statistical thermody-
namic arguments and the Boltzmann distribution, the energy can be found in gen-
eral to be [10]

hf
ε= hf kT
(5.24)
e -1
where h = 6.626 ´ 10–34 J·s is Planck’s constant. When the temperature is on the
order of 300 K, approximately room temperature, the relation hf << kT holds for
millimeter-wave frequencies and below. Under this assumption, (5.24) simplifies to
e ~ kT, and thus (5.23) results in (5.12). The region where hf << kT holds is called
the Rayleigh-Jeans region which will be discussed in detail in Chapter 6; (5.12) is
the Rayleigh-Jeans approximation of the thermal noise power. In general, the noise
power of a conductor is described by combining (5.23) and (5.24), yielding
5.2  Receiver Noise 149

−100

−120
Rayleigh-Jeans form

−140

Nn (dBm)
−160
Plankian form

−180

−200 8 10 12 14
10 10 10 10
f (Hz)
Figure 5.9  Rayleigh-Jeans form and Planck form of the thermal noise power in a conductor versus
frequency.

hf Df
Nn = (5.25)
e hf kT - 1

This is referred to the Planck form due to its relation to Planck’s blackbody radia-
tion law, which is discussed in detail in Chapter 6.
Figure 5.9 shows the noise power using the Rayleigh-Jeans approximation
(5.12) and the Planck form (5.25) as a function of frequency for T = 310 K and Df =
1 MHz. The noise power given by the Planck form approaches zero as the frequency
increases, and converges to kTDf as the frequency decreases. This is consistent with
(5.24), which states that the average total energy ε ¾¾¾ f ®0
® kT and ε ¾¾¾®
f ®¥
0.
For T = 310 K and f = 1 THz, the deviation of the approximate form kTDf from the

−120

−130

−140 Rayleigh-Jeans form


Nn (dBm)

−150

−160 Plankian form, f = 1 THz


Plankian form, f = 100 GHz
−170

−180 1 2 3 4
10 10 10 10
T (K)
Figure 5.10  Rayleigh-Jeans form and Planck form of the thermal noise power in a conductor versus
temperature.
150 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Planck form is only 0.34 dB. Figure 5.10 shows (5.12) and (5.25) as a function of
T for f = 100 GHz and f = 1 THz with Df = 1 MHz. At low temperatures the Planck
form of the noise power decreases more rapidly than the Rayleigh-Jeans form. For
f = 100 GHz, the deviation of kTDf at T = 10 K is 1.08 dB, and for T = 100 K is
only 0.10 dB. For f = 1 THz and T = 100 K, the deviation is also 1.08 dB. Thus,
for characterizing the noise power of components or systems operating below 100
GHz, the Rayleigh-Jeans form is sufficient, and is reasonably accurate up to 1 THz.
Characterizing cryogenically cooled systems or evaluating the noise power emitted
by an object with low emissivity, which results in a low radiometric temperature,
may also be done with the Rayleigh-Jeans form if the frequency does not exceed a
couple hundred GHz. The Planck form becomes necessary when operating at fre-
quencies near 1 THz and above.

5.3  Noise Figure and Noise Temperature

The purpose of a two-port network may be to add gain to a signal in the case that
the network represents an amplifier, or induce losses in the case of an attenuator.
If the two-port represents a transmission line, it ideally does not affect the signal
but in reality will induce some losses and can thus be characterized as an attenuator.
As seen in the previous sections, any component in the network whose temperature
is above 0 K will generate noise due to internal non-idealities of the components,
and the signal at the output of the network will thus be corrupted by the induced
noise in the network. A number of different physical processes give rise to the in-
ternal noise, however it is convenient to model the output noise in terms of a single
equivalent noise temperature which encompasses the noise due to all the internal
noise sources rather than to characterize each noise source individually. Defining
the system noise in this way allows a component to be described as a “black box”
with a single specified noise response. The noise introduced by multiple components
can then be more easily taken into account when characterizing a cascade of mul-
tiple components.

5.3.1  Noise Figure


A general two-port network is shown in Figure 5.11 with input and output signal
and noise powers. The quantity used to characterize the degradation of the input
signal by the noise in the network is called the noise factor, and is defined by [11]
Si Ni
F= (5.26)
So No

Si , N i G So, No

Two-port
network
Figure 5.11  Generalized two-port network.
5.3  Noise Figure and Noise Temperature 151

where Si and So are the input and output signal powers, and Ni = kT0Df and No are
the input and output noise powers and S ⁄ N is the signal-to-noise ratio at the input
or output. Note that the definition of F has been standardized by defining the input
noise power to be that of a matched resistance at temperature T0 = 290 K which is
approximately room temperature. In most situations it is typical to see values for
the noise figure, which is simply the noise factor expressed in decibels:

F(dB) = 10log10 (F) (5.27)


According to (5.26) the noise figure gives a measure of the reduction in signal-to-
noise ratio through the network:

Si S
=F o (5.28)
Ni No
A non-ideal two-port network will alter the amplitude of the signal, either add-
ing gain or inducing losses. The output signal power can thus be given by

So = GSi (5.29)
where G is the conversion factor which may represent either the gain or the loss.
The network injects its generated noise power Nn which is additive to the input
noise power. Thus the output noise power can be given by

No = GNi + Nn (5.30)
Note that Nn is not modified by G. That is, the gain or loss of the network only
affects the input signal and input noise powers, and not the intrinsic noise generated
by the network. Using (5.29) and (5.30) the noise figure can be written

Nn
F = 1+ (5.31)
GNi

The noise power generated by the two-port network is thus

Nn = (F - 1)GNi (5.32)
and the output noise power is

No = FGNi (5.33)
In terms of the temperature,

Nn = (F - 1)GkT D f (5.34)
and

No = FGkT D f (5.35)
The noise figure defined by (5.31) is given only in terms of the conversion gain
of the two-port, the noise generated by the two-port, and the input noise power.
152 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Ni = 0 No = GkTE∆f
Noisy
T=0K network
G, TE, ∆f

Ni = kTE∆f No = GkTE∆f
Noiseless
T = TE network
G
Figure 5.12  Representation of equivalent input noise temperature for a two-port network.

That is, the output signal power is affected only by the conversion gain, while the
output noise power is modified by the noise power generated in the network. Thus
the choice of input signal power is arbitrary since it does not factor into the cal-
culation of F. The input noise power, however, is present in (5.31), and thus for a
meaningful definition of noise figure it must be standardized; for this reason it has
been defined to be T0 = 290 K.

5.3.2  Noise Temperature


Because the spectral noise power of a conductor with resistance matched to the
system can be given in terms of its temperature through (5.12), it is beneficial to
characterize the two-port in terms of an equivalent input noise temperature. In this
definition the two-port is assumed to be noise-free, and the output power is ob-
tained which would be produced by an input temperature TE. Figure 5.12 illustrates
the definition of equivalent input noise temperature; one two-port is noisy and is
at temperature TE, while the other is noiseless. The input to the noisy two-port is
a matched resistance at T = 0 K, which therefore contributes no input noise. The
input to the noiseless two-port is a matched resistance at temperature TE which
thus generates noise power kTEDf. The noise power of each two-port is therefore
the same, and given by

Nn = GkTE Df (5.36)
Note that TE­ is defined at the input terminals of the two-port and represents all
the noise sources within the two-port; the conversion gain affects the input noise
power, which is thus Ni = kTEDf.
An expression for TE­ in terms of the noise figure is found by equating (5.34)
and (5.36), which results in

TE = (F - 1)T0 (5.37)
Alternatively, the noise figure can be given in terms of the equivalent noise tem­
perature by
TE
F = 1+ (5.38)
T0
5.3  Noise Figure and Noise Temperature 153

N = kTp∆f No N = kTp∆f
Attenuator
Tp Tp
Loss = L

Figure 5.13  Conceptual setup for calculating the noise power of an attenuator. Matched resis-
tances are placed at the input and output ports with all components in thermal equilibrium.

The choice of using either F or TE when characterizing the noise of a two-port


is essentially arbitrary since they ultimately describe the same characteristics of the
two-port. However, noise figure is most often used when describing the perfor-
mance of systems and amplifiers whereas the equivalent noise temperature is typi-
cally used in the analysis of mixers and other components [6].

5.3.3  Noise Figure of an Attenuator


Aside from amplifiers, most components of a receiving system impart signal loss.
Filters, transmission lines, and adapters all have losses which introduce noise into
the system. Diode mixers also introduce conversion loss and therefore generate
noise. To characterize the noise contributions of a general attenuator, it is analyzed
by placing matched resistive loads at the input and output terminals and assum-
ing that all components are in thermal equilibrium at temperature Tp, as shown in
Figure 5.13.The attenuator introduces a noise power Na and has conversion loss
L. Note that the conversion gain may also be used, where G = 1/L. The noise input
to the attenuator is that generated by the matched resistor at temperature Tp, and
the output noise power is the sum of the input noise power modified by L and the
attenuator noise power Na:

1
No = kTp Df + N a (5.39)
L
The load resistor produces noise power kTpDf, and because the load resistor and
the attenuator are in thermal equilibrium, the noise power of the load resistor must
equal the noise power output of the attenuator. Thus,

1
kTp Df + N a = kTp Df (5.40)
L
The noise power generated by the attenuator is thus given by

1
N a = (1 - )kTp D f (5.41)
L
As was done for the general two-port network, an equivalent input noise tem-
perature can be defined which describes the output noise power of a noise-free at-
tenuator with an input noise temperature TE:

1
Na = kTE D f (5.42)
L
154 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Equating (5.41) and (5.42) gives the equivalent input noise temperature of an at-
tenuator with conversion loss L at a physical temperature Tp:

TE = (L - 1)Tp (5.43)

The noise figure of the attenuator can then be found by substituting (5.43) into
(5.38), which results in
Tp
F = 1 + (L - 1) (5.44)
T0
For an attenuator at a physical temperature Tp = T0 the noise figure is simply
F = L.

5.3.4  Noise in Cascaded Systems


Microwave and millimeter-wave receivers consist of a number of individual ele-
ments in a cascaded network, and each component will generate its own noise. The
noise performance of multiple cascaded stages can be analyzed by first considering
a cascade of two stages as seen in Figure 5.14. The network has input signal and
noise powers Si­­ and Ni = kTiDf, where Ti is the input noise temperature. The two
stages have gains G1,2 and generate noise powers ­N1,2. The output noise power is
then given by

No = G1G2Ni + G2N1 + N2 (5.45)

Each subsystem of the cascade can be replaced by a noise-free system with


equivalent input noise temperatures, and the noise powers of each of the stages are
then N1,2 = G1,2kTE1,E2Df. The output noise power is then

æ T ö
No = kTE Df = G1G2k ç Ti + TE1 + E2 ÷ Df (5.46)
è G1 ø

where TE is the equivalent input noise temperature of the cascaded network.

Si , N i G1, F1 G2, F2 So, No

(a)

Si , N i G1 G2 So, No

TE1 TE2
(b)
Figure 5.14  Cascaded network of two subsystems. (a) Noisy subsystems. (b) Cascade with noise-
less subsystems and equivalent input noise temperatures.
5.3  Noise Figure and Noise Temperature 155

Si , N i G1G2 So, No

TE

Figure 5.15  Equivalent representation of a cascade as a single two-port.

To analyze the noise performance of the entire cascade as a whole, the cascaded
network can be represented as a single subsystem with one equivalent input noise
temperature, as seen in Figure 5.15. The output noise power of the single subsystem
in this representation is

No = Gk(Ti + TE ) D f (5.47)

where G = G1G2 such that the gain of the subsystem is equal to that of the cascade.
Equating (5.47) and (5.46), the equivalent input noise temperature of the cascaded
network can be found in terms of the equivalent input noise temperatures of the
individual subsystems. In particular,

TE2
TE = TE1 + (5.48)
G1

This can be directly extended to a general number of subsystems N, the result


of which is
N
TE2 T TEN TEn
TE = TE1 + + E3 + � = TE1 + å n -1
(5.49)
G1 G1G2 G1G2 �GN -1 n=2 Õ G
m =1 m

The noise figure of a cascaded system can be found by substituting (5.37) into
(5.49), which gives
N
F2 - 1 F3 - 1 FN - 1 Fn - 1
F = F1 + + � = F1 + å n -1
(5.50)
G1 G1G2 G1G2 �GN - 1 n=2 Õ m=1Gm
As an example consider the cascaded network shown in Figure 5.16a, which
is a simple homodyne receiver consisting of an antenna, amplifier, and low-pass
filter, with transmission lines between each element. The physical temperature and
conversion loss are given for the passive components and the conversion gain and
noise figure are given for the amplifier. Noise figure is not given for the passive
components since the conversion loss is more likely to be provided to the system
designer; the noise figure can be calculated using (5.44) or, as will be done below,
the equivalent temperature can be calculated using (5.43). The equivalent noise
temperature of the receiver is referred to the input of the cascade, or at the antenna
terminals, and thus the first component of the cascade is the transmission line be-
tween the antenna and the amplifier. The equivalent noise temperatures of each
component can be found using (5.37) and (5.43), and are
156 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Transmission LNA Transmission LPF Transmission


line line line

Tp = 300 K G2 = 27 dB Tp = 300 K Tp = 300 K Tp = 300 K


L1 = 0.5 dB F2 = 4 dB L3 = 3 dB L4 = 1 dB L5 = 1 dB
(a)

LNA Transmission LPF Transmission


line line

G2 = 27 dB Tp = 300 K Tp = 300 K Tp = 300 K


F2 = 4 dB L3 = 3 dB L4 = 1 dB L5 = 1 dB
(b)
Figure 5.16  Homodyne receiver (a) with and (b) without a lossy transmission line between the
antenna and the amplifier.

TE1 = (L1 - 1)Tp = (1.122 - 1)300 » 37 K

TE2 = (F2 - 1)T0 = (2.5 - 1)290 » 438 K


TE3 = (L3 - 1)Tp = (2 - 1)300 » 300 K

TE4 = (L4 - 1)Tp = (1.26 - 1)300 » 78 K

TE5 = (L5 - 1)Tp = (1.26 - 1)300 » 78 K

The equivalent noise temperature of the cascade is given by (5.49), and recalling
that G = 1/L results in

L1TE3 L1L3TE4 L1L3L4TE5


TE = TE1 + L1TE2 + + +
G2 G2 G2
1.122 ´ 300 1.122 ´ 2 ´ 78 1.122 ´ 2 ´ 1.26 ´ 78
= 37 + 1.122 ´ 438 + + +
501 501 501
» 530 K

Now consider the receiver in Figure 5.16b, where the initial transmission line
has been removed, and the amplifier is connected directly to the antenna. In this
case, the equivalent noise temperature of the receiver is
5.3  Noise Figure and Noise Temperature 157

TE3 L3TE4 L3L4TE5


TE = TE2 + + +
G2 G2 G2
300 2 ´ 78 2 ´ 1.26 ´ 78
= 438 + + +
501 501 501

» 439 K

Thus the receiver noise temperature is essentially equivalent to the noise tem-
perature of the amplifier. This demonstrates the importance of the first element in
the receiver chain; in calculating the receiver noise temperature, the noise tempera-
ture of all elements of the receiver chain following the initial element is divided
by the conversion gain of the initial element, as seen in (5.49). For a low-noise
receiver it is thus important that the initial element have high gain and low noise
temperature. As demonstrated above, even the inclusion of a transmission line
with a relatively low loss of 0.5 dB resulted in an increase in noise temperature of
91 K.

5.3.5  ADC Noise


The ADC is the last hardware component in the analog RF receiver chain. Because
most receivers include gain, and the noise contribution of the ADC is reduced by
the gain of the components prior, thus the noise contributed by the ADC is gen­
erally small compared to the rest of the receiver circuit. As the bandwidth of the
baseband signals increases, however, the noise contribution of the ADC becomes
more important. The primary factors contributing to noise in the ADC are thermal
noise, quantization noise, and sampling time jitter. When specifying the noise gen­
erated by the ADC, it is typical to define the dynamic range of the ADC in terms
of the noise power and the maximum signal the ADC can accommodate, called the
full-scale (FS) signal. Additionally, it is typical to refer to the dynamic range as the
ADC signal-­to-noise ratio (SNR), where the assumption is that the input signal is
a specified level below the full-scale signal, typically 0.5 dB to 1 dB below FS. For
clarity, this level is referred to as FSX where X is the level below FS in dB. In essence,
this definition of dynamic range characterizes the noise contribution of the ADC in
terms of the worst-case situation. Here, the range between FSX and the noise floor
is referred to as the ADC dynamic range (ADC DR) for consistency.
For a given ADC FSX and DR, the ADC noise figure can be found in terms of the
DR, FSX, sampling frequency fs­, and the input thermal noise power. The sampling
frequency is generally taken to be twice the signal frequency in order to satisfy the
Nyquist sampling criteria, with the frequency range between dc and fs ∕ 2 referred
to the first Nyquist zone. Typically, the ADC DR is given in terms of the integrated
noise power in the first Nyquist zone. This noise power is given by

N1 = FSX - DR dBm (5.51)

The normalized noise power, or the power in a 1 Hz bandwidth, is thus


158 Microwave and Millimeter-Wave Remote Sensing for Security Applications

X dB FS
FSX
ADC DR DR

Signal power (dBm) Integrated noise power


in Nyquist zone

10log(fs/2)

Normalized noise power


in 1 Hz bandwidth
F
Normalized thermal noise
power in 1 Hz bandwidth

Figure 5.17  Summary of ADC signal and noise levels.

N2 = N1 - 10log10 (fs 2) dBm (5.52)

The noise figure is then the normalized noise power minus the thermal noise in a 1
Hz bandwidth:

FADC = N2 - kT = FSX - DR - 10log10 (fs 2) - kT dBm (5.53)

This formulation relies on the measured SNR of the ADC and includes all noise
contributions, including thermal, quantization, jitter, and any other noise sources.
A summary of the signal and noise levels in an ADC is given in Figure 5.17.
In some cases, the noise generated by the ADC can be characterized directly.
The thermal noise of a wideband sampling ADC can become large simply due to the

Vpp qv

Input analog signal Quantizer Output digitized signal


Figure 5.18  Quantization of an input analog signal.
5.3  Noise Figure and Noise Temperature 159

wide bandwidth of the sample-and-hold circuit. This circuit consists in essence of a


resistance R and capacitance C which form a low-pass filter with cutoff frequency
1
fco = (5.54)
2π RC
For an input resistance that is matched to the source resistance, the cutoff frequency
can be written

1
fco = (5.55)
2π C
The cutoff frequency of the sample-and-hold circuit is generally greater than the
sampling frequency fs, and thus more thermal noise power is generated in the ADC.
The thermal noise power generated is thus given by

kT
Nthermal = (5.56)
2π C
Quantization noise is caused in the digitization process when an analog volt-
age is converted and represented by a discrete voltage value. The ADC peak-to-
peak input voltage range Vpp is divided into n voltage regions, separated by the
quantization voltage qv, as illustrated in Figure 5.18. An input voltage can then be
represented by

Vi = kqv + ε , k = 0,1,..., n (5.57)

where kqv << (k+1)qv. The voltage e falls between two quantization levels, and
the output voltage is either kqv or (k + 1)qv, depending on the logic of the decision
circuit. Thus, the fraction of the input signal represented by e is lost and cannot be
recovered; noise is introduced as a result.
The input voltage range Vpp can be given in terms of the quantization voltage
and the number of levels by

Vpp = 2n qv (5.58)

For a sinusoid, the rms peak-to-peak is less than the full-scale peak to peak, and is
given by

(2n - 1)qv £ Vpp £ 2n qv (5.59)

where the right-hand side holds when n ³ 5. The rms quantization error can be
given by [12]
qv
Nq = (5.60)
12
In the first Nyquist zone, the quantization noise power spectral density is given by
Nq qv
Nq (f ) = = (5.61)
fs 2 6 fs
160 Microwave and Millimeter-Wave Remote Sensing for Security Applications

The full scale signal voltage is 2nqv, and the rms full scale voltage is thus 2n qv (2 2).
The dynamic range due to the quantization error is thus

2n qv 6 fs
DR q = = 2n -1 3fs (5.62)
2 2qv

For a signal bandwidth which is less than the Nyquist frequency, Df < fs ∕ 2, the
dynamic range is

3fs
DR q = 2n -1 (5.63)
Df
Jitter induced noise is caused by random variations in the sampling time of the
sample-and-hold circuit [13]. Time jitter can be characterized in general by consid-
ering an input voltage sinusoid

vi (t) = V0 sin(2π ft) (5.64)

The time-derivative of the input signal is

dvin (t)
= 2π fV0 cos(2π ft) (5.65)
dt
the rms voltage of which is

dvrms 2π fV0 (5.66)


=
dt 2

Defining dvrms as the rms voltage error due to the jitter and dt as the rms jitter tj,
the voltage error can be given by
2π fV0t j
dvrms = (5.67)
2
The rms input signal power is

V0
vin,rms = (5.68)
2
and thus the dynamic range due to the jitter noise is given by

vin,rms 1
DR j = = (5.69)
dvrms 2π ft j
Jitter noise thus follows a decreasing linear curve with slope f –1.

5.4  Receiver Linearity

Receivers are generally designed to operate such that the output signal power is lin-
early proportional, through the conversion gain or loss, to the input signal power.
5.4  Receiver Linearity 161

Such a linear relation of input power to output power allows for simpler analyses
of the input signal since the output signal power is simply a scaled version of the
input signal power and non-linearities of the receiver need not be accounted for in
the reconstruction of the input signal. However, in reality there is a finite range of
input powers for which the output of a receiver will be linearly proportional to the
input. For small input signals, the noise power, as discussed in the previous sec-
tions, constitutes a large portion of the output power, and the change in the output
due to presence or lack of the input signal may be indistinguishable from the noise
itself. That is, the output power appears to be only due to the noise generated by
the receiver, and the incremental power due to the signal is indiscernible above the
noise. As the signal power increases, the output power gradually increases until the
output signal power is greater than the noise floor. Once the output signal power is
twice that of the receiver noise power, the response becomes approximately linear.
There is also a limiting region for high power signals, where the output power is
constrained by the maximum power that can be physically delivered by the receiver.
At high input signal levels the output signal power is no longer linearly proportional
to the input signal power such that increasing the input power further results in
progressively smaller increases in output power, which is referred to as gain com-
pression. Increasing the input signal power beyond the region of compression, the
output signal saturates such that increasing the input signal further results in no
increase in the output power. If the input signal is increased beyond the point of
saturating the receiver, physical damage may occur to the components.
The range of power levels between where the output signal is just discernible
from the noise and where it begins to saturate is called the linear region, and is
where most receivers are operated. Figure 5.19 shows the output signal power for
a typical receiver versus its input signal power. For very low input signals, the noise

Saturation
Output power (dB)

Compression

Linear region

Noise floor

Input power (dB)


Figure 5.19  Output signal power of a typical receiver.
162 Microwave and Millimeter-Wave Remote Sensing for Security Applications

power dominates. Once the input signal is greater than the noise power, the output
is linearly related to the input power, and the plot of the output power versus the
input power has a slope of unity. For higher input power the output begins to roll
off and eventually converges to the output saturation power of the receiver.

5.4.1  Gain Compression


The non-linearity of receivers at low input powers is the result of the receiver noise,
where the output power is approximately equal to the noise power until the sig-
nal power is within an order of magnitude of the noise power. This low-power
non-linear region continues until the output signal is approximately 3dB above the
noise power. At high input powers, the non-linearity in the output signal is due to
the inherent non-linearities of semiconductor devices within the receiver, such as
amplifiers or diode mixers. The voltage response of such a non-linear device can be
represented by the Taylor series
¥
vo = a0 + a1vi + a2 vi2 + a3vi3 + � = å an vin (5.70)
n =0

where vo is the device output voltage, vi is the input voltage signal, and the Taylor
coefficients are given by

d n vo
an = (5.71)
dvin vi = 0

The coefficient a0 represents a dc voltage signal, while a1 is the coefficient of the


linear output of the signal.
For a monochromatic input signal given by

vi = A cos(ω 0t) (5.72)

where A is the signal amplitude and w0 = 2pf is the angular frequency, the output
voltage of the non-linear device, neglecting terms above third order, is

vo = a0 + a1A cos(ω 0t) + a2 A2 cos2 (ω 0t) + a3 A3 cos3 (ω 0t) (5.73)

which expands to

æ 1 ö æ 3 ö 1
vo = ç a0 + a2 A2 ÷ + ç a1A + a3 A3 ÷ cos(ω0t) + a2 A2 cos(2ω 0t)
è 2 ø è 4 ø 2
1
+ a3 A3 cos(3ω0t) (5.74)
4

The amplitude of the output signal at the fundamental frequency w0 is thus

3
a1A + a3 A3 (5.75)
4
5.4  Receiver Linearity 163

The first term of (5.75) is the output signal due to the linear gain of the signal
at the fundamental frequency while the second term is the amplitude of a secondary
signal at w0 due to the third order response of the system. The conversion gain of
the output signal at w0 is thus

æ 3 3ö
çè a1A + a3 A ÷ø cos(ω0t) 3
4
g= = a1 + a3 A2 (5.76)
A cos(ω0t) 4

In active components such as amplifiers, a3 is most often negative. Because the


third order term of (5.74) is proportional to A3, for small input signals this term is
also much smaller than the fundamental term, however it increases in power at a
rate three times faster than the fundamental signal power. Thus, as the amplitude
A increases, the linear gain of the signal is reduced due to the increasing amplitude
of the third harmonic. Note that (5.76) includes only terms up to third order, and
that higher order harmonics also contribute to the signal non-linearity. Physically,
the device can only output a finite amount of power, and the reduction in the power
of the signal at the fundamental frequency is due to the increasing power of the sig-
nals at the harmonic frequencies, which draw power from the fundamental signal.
As the input signal power increases, the output power converges to a static value,
which is the saturation power of the device. The point at which the output power
reaches this static power is therefore called the saturation point. Beyond the satura-
tion point, an increase in the input signal results in no change in the output power.
Increasing the signal power beyond the saturation point may also result in physical
damage of components within the receiver.
In the linear region, as the input power is increased the output power increases
with a slope of unity. That is, for every 1dB increase in input power there is a 1dB
increase in output power of the fundamental. For the second order terms, the slope
is 2:1, and for the third order terms it is 3:1. Thus, the harmonic signals increase in
power at a much faster rate than the fundamental signal.
The reduction in linear gain due to the higher order harmonic signals can be
quantified by characterizing the difference between the expected output power of
the ideal, linear output and the actual output power, which will be somewhat lower,
as shown in Figure 5.20. This reduction is called gain compression and the point

X dB
OPXdB

IPXdB

Figure 5.20  Gain compression point.


164 Microwave and Millimeter-Wave Remote Sensing for Security Applications

at which the output power is compressed by XdB is called the PXdB compression
point. For example, the linear output signal power is given by

Pl = GPi (5.77)

where Pi is the input signal power and G is the gain. If the measured output power
is

Po = Pl - 1 dB = GPi - 1 dB (5.78)

the signal is being compressed by 1dB, and the compression point is referred to as
P1dB. The compression point is often written either IP1dB or OP1dB, referring to the
power level at the input or output, respectively, where the output signal is in 1dB
compression. Typical compression points are P1dB and P3dB; for high sensitivity
systems P0.1dB is sometimes used.
The input signal voltage which equals PXdB on the output signal can be found
analytically in terms of the Taylor coefficients in (5.70) by equating the coefficient of
the signal at frequency w0 in (5.73) with the linear voltage gain reduced by X dB:

3
a1AXdB - 3
a3 AXdB = 10- X 20 a1AXdB (5.79)
4

where it has been assumed that a3 < 0. The compression point is thus found to be

2 4a1(1 - 10- X 20 )
AXdB = (5.80)
3a3

The input power compression point is simply

2
IPXdB = AXdB (5.81)

The output power compression point is found by squaring the fundamental signal
in (5.73) at w0, which results in

1 2 2 2a3 (1 - 10- X 20 )
OPXdB = a1 AXdB = 1 (5.82)
2 3a3

5.4.2  Intermodulation Products


When multiple signals at separate frequencies are input to a non-linear device, the
signals beat with one another producing additional harmonic signals in addition to
the original single tones. If the input signal is the superposition of two monochro-
matic signals

vi = A1 cos(ω1t) + A2 cos(ω 2 t) (5.83)


5.4  Receiver Linearity 165

where w = 2wf, the output voltage of the non-linear device, neglecting terms above
third order, is given by substituting (5.83) into (5.70) which results in

vo = a0 + a1 [A1 cos(ω1t) + A2 cos(ω 2 t)]+ a2 [A1 cos(ω 1t) + A2 cos(ω 2 t)]


2

é 1 ù
+ a3 [A1 cos(ω1t) + A2 cos(ω 2 t)] = ê a0 + a2 (A12 + A22 )ú
3
ë 2 û

é æ3 3 öù
+ ê a1A1 + a3 ç A13 + A1A22 ÷ ú cos(ω1t)
ë è4 2 øû

é æ3 3 öù 1
+ ê a1A2 + a3 ç A23 + A12 A2 ÷ ú cos(ω 2t) + a2 A12 cos(2ω 1t)
ë è 4 2 ø û 2 (5.84)

1 3 3
+ a2 A22 cos(2ω 2 t) + a3 A13 cos(3ω1t) + a3 A23 cos(3ω 2t)
2 4 4
+ a2 A1A2 {cos[(ω1 + ω 2 )t ] + cos[(ω1 - ω 2 )t ]}

3 2
+ a3 A A2 {cos[(2ω1 + ω 2 )t ] + cos[(2ω 1 - ω 2 )t ]}
4 1
3
+ a3 A1A22 {cos[(ω1 + 2ω 2 )t ] + cos[(ω 1 - 2ω 2 )t ]}
4
The output signal consists of components at the input frequencies w1 and w2, inte-
ger multiple of w1 and w2, and sums and differences of integer multiple of the two
frequencies. Including higher order terms results in frequency components of the
form

mω1 ± nω 2 , m = 0,1, 2,…, n = 0,1, 2,… (5.85)

The output signals at frequencies w1 and w2 are the input signals; the signals at
all other frequencies in (5.85) are called intermodulation products. Even order in-
termodulation products result in terms at dc and at even harmonics while odd order
products result in terms at the fundamental frequencies and at odd harmonics. The
second and third order products result in the frequencies

a2 vi2 ® 2ω1 , 2ω 2 , ω1 + ω 2 , ω 1 - ω 2
(5.86)
a3vi3 ® ω1 , ω 2 , 3ω 1 , 3ω 2 , 2ω 1 + ω 2 , 2ω 1 - ω 2 , ω 1 + 2ω 2 , ω 1 - 2ω 2

For signals whose frequency separation is much smaller than the average of
the two frequencies, the integer multiples of the two frequencies are sufficiently far
away that they can be neglected unless the system operates with greater than an
octave of bandwidth. The output spectrum is illustrated in Figure 5.21.
For a mixer, it is the sum and difference frequencies arising from the second
order intermodulation product that are the most useful. For an upconverter, the de-
sired product is that at frequency w1 + w2, while for a downconverter the product
166 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Amplitude
ω2-ω1 ω1 ω2 2ω1 2ω2
2ω1-ω2 2ω2-ω1 ω2+ω1
Figure 5.21  Output spectrum of a non-linear device showing the two fundamental signals and the
intermodulation products.

at frequency w1 – w2 is that which is desirable. The other intermodulation products


are typically at frequencies which are separated enough from the desired frequen-
cies that they can be easily filtered.
For an amplifier the fundamental frequencies w1 and w2 are the desired fre-
quency components while the other terms represent potential sources of signal dis-
tortion. If the two frequencies are close, the third order products 2w1 – w2 and 2w2
– w1 fall close to the two fundamental signals, whereas other products are separated
enough in frequency that they may easily be filtered. The two third order difference
frequencies often fall very close to or within the passband of the amplifier and thus
pose the greatest risk for distortion, which is called third order intermodulation
distortion. In addition to gain compression as the third order products increase in
amplitude, the desired signals can be corrupted by the presence of two signals at
separate frequencies which can mix in a detector.

5.4.3  Third Order Intercept Point


The linearity of a receiver or amplifier is often characterized by the power, referred
to either the input or output, where the power levels of the fundamental signal and
the third order harmonic are equal. This point is called the third order intercept
point and the higher this point is, the greater the suppression of the intermodulation
products, and thus the larger the region of linearity of the receiver. The third order
intercept point is denoted by IP3 or TOI, and can be found analytically by equating
the two terms of (5.75), which results in

2 4a1
AIP 3 = (5.87)
3a3
The output power at the intercept point is equal to the ideal output power of
the fundamental signal, and thus the output power at the intercept point can be
calculated using the amplitude of the fundamental from (5.82), which gives

2a13 (5.88)
OPIP3 =
3a3
2
whereas the intercept point referred to the input is IPIP3 = AIP 3
5.4  Receiver Linearity 167

The third order intercept point and XdB compression point can be related to
one another using (5.82) and (5.88). Specifically, the ratio of the third order inter-
cept point and the XdB compression point is

PIP3 1 (5.89)
=
PXdB 1 - 10- X 20

Note that (5.89) holds true for the ratio of either the input or output powers.
Relating PTOI to P1dB results in

PIP3 1
= = 9.195 = 9.636 dB (5.90)
P1dB 1 - 10-1 20

Thus, the third order intercept point of a non-linear device is, in general, 9.6
dB above the 1dB compression point. The third order intercept point is therefore a
theoretical point only, since the signals are well into the region of compression at
the input power level required at the intercept point. This interesting result implies
that the intercept point cannot, in general, be directly measured. It can be calcu-
lated, however, using the fact that the slopes of the fundamental and third order
harmonic signal powers are 1:1 and 3:1. The power of the third order harmonic
can be given by

P3 = PIP3 - 3(PIP3 - P1) = 3P1 - 2PIP3 (5.91)

where Pn is the output power of the nth harmonic expressed in decibels. Equation
(5.91) can be seen graphically on the curve of the output powers in Figure 5.22.
Expressed linearly, the power of the third order harmonic is

P13
P3 = (5.92)
2
PIP 3

OPIP3

IPIP3

Figure 5.22  Signal powers of the fundamental and third order harmonic signals.
168 Microwave and Millimeter-Wave Remote Sensing for Security Applications

The output power of the intercept point can then be found in terms of the powers
of the fundamental and third order harmonic signals by

P13
PIP3 = (5.93)
P3

Note that the intercept point can be calculated by measuring P1 and P3 at any point
in the curve providing the signals are not in the region of compression.
An increase in the ratio of the third order intercept point to the 1 dB compres-
sion point over the theoretical 9.636 dB can be achieved using techniques such as
feedback, feed-forward, or predistortion [5]; these techniques generally work to
suppress the harmonics in order to increase the third order intercept point. Ampli-
fiers with PIP3 ⁄ P1dB > 9.636 dB are referred to as high-linearity amplifiers. Com-
mercially available microwave and millimeter-wave amplifiers often employ these
techniques.

5.4.4  Intercept Point of a Cascade


The third order intercept point of a system consisting of multiple components is re-
duced by the coherent summation of the voltages of the multiple harmonics. Similar
to calculating the noise figure of a cascaded network, the third order intercept point
of a cascaded network can be found in terms of the gains and intercept points of its
components. The system intercept point is generally reduced due to the cascading
of multiple components, and is given by [1]

-1
æ N 1 ö
PIP3, sys = çå ÷ (5.94)
è n =1 PIP3,n ø

where PIP3,n is the third order intercept point of component n in the cascade referred
to the input of the network. That is, for each individual component, the gains and
losses prior are subtracted or added to the intercept point to refer it to the input to
the entire network. It is also important to note that the powers in (5.94) are linear,
and not in decibel scale.

5.4.5  Dynamic Range


The linear response where the output fundamental signal power has a slope of unity
is limited by system noise at low input power and by gain compression and output
saturation at high input power, as discussed in the previous sections. The differ-
ence in power between these two regions is the usable extent of the linear region of
operation of the receiver and is referred to as the dynamic range. Figure 5.23 shows
a curve of the output power of a receiver, denoting the dynamic range. Because the
output fundamental power has a slope of unity with increasing input power, the
dynamic range can refer to a range of input or output power over which the char-
acteristics of the signal response are within acceptable error bounds. The definitions
5.4  Receiver Linearity 169

DR

MDS
3 dB

IP1dB

Figure 5.23  Receiver dynamic range.

of the upper and lower limits of the dynamic range varies, but are often specified to
be the 1 dB compression point and the minimum detectable signal (MDS)

DR = P1dB - MDS (5.95)

where the minimum detectable signal is defined to be 3 dB greater than the noise
floor:

MDS = Gk(T0 + Tsys ) D f + 3 dB (5.96)

If a specific signal-to-noise ratio (SNR) is required for the accurate recovery of the input
signal, the lower limit of the dynamic range is then the noise floor plus the SNR:

DR = P1dB - Gk(T0 + Tsys ) D f - SNR (5.97)

The lower limit is sometimes defined to be the noise floor itself. In addition, the defi-
nition of the upper limit varies, and may be P0.1dB­ for highly linear systems or P3dB
for high power systems where linearity is not as important as output power.
As an example, consider a receiver with a 1 dB compression point of 15 dBm,
a gain of 40 dB, a system temperature of 500 K, and a bandwidth of 100 MHz.
The minimum detectable signal given by (5.96) is then MDS = –46.6 dBm and the
dynamic range as defined by (5.95) is DR = 61.6 dB. If the system requires SNR =
10 dB, the dynamic range given by (5.97) is DR = 54.6 dB, or 7 dB greater than the
dynamic range referred to the MDS.
The noise floor of a receiver can be calculated in a straight-forward manner by
using the noise figure of the receiver and units of decibels. Note that, using (5.37),
the output noise of a receiver is

No = Gk(T0 + Tsys ) Df = FGkT0 D f (5.98)


170 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Equation (5.98) is defined on a linear scale; it can be defined in dB in terms of the


noise present in a 1 Hz bandwidth plus the noise figure, gain, and bandwidth by

No = -174 + F + G + D f (dBHz) dBm (5.99)

where kT0 = –174 dBm and Df(dBHz) = 10log10(Df ) is the bandwidth in decibels
referred to 1 Hz. For a receiver with a noise figure F = 4 dB, gain G = 20 dB, and
bandwidth Df = 100 MHz = 80 dBHz, the output noise power is No = –174 + 4 +
20 + 80 = –70 dBm.

5.4.6  Spurious Free Dynamic Range


In addition to characterizing the range over which the fundamental signal is linear, it
is also often useful to define the range over which the power level of spurious signals
does not exceed that of the noise floor. In this range only the fundamental signal is
present above the noise. This is called the spurious free dynamic range (SFDRn) and
is the difference in power between the fundamental signal and the spurious signal
of interest when the power of the spurious signal is equal to the noise floor. The
index n refers to the harmonic for which the dynamic range is defined. Because the
second order harmonic often falls out of the band of interest and the third order
harmonic is the next largest in amplitude, SFDR is commonly used to simply refer
to the spurious free dynamic range relative to the third order harmonic.
Due to the linear nature of the power of the fundamental signal and each har-
monic with the input power, SFDR3 can be found in terms of the intercept point
PIP3 and the noise floor by

2
SFDR 3 = (PIP3 - No ) (5.100)
3

SFDR

IPIP3

Figure 5.24  Spurious free dynamic range.


5.4  Receiver Linearity 171

where the power levels are expressed in decibels. This result can be obtained by
analyzing the coefficients of the Taylor series expansion (5.70), however it is more
easily arrived at by simply analyzing a graph of the fundamental and third order
harmonic power versus the input power, as seen in Figure 5.24, and using the fact
that the slope of the fundamental is unity while that of the third order harmonic is
3. SFDRn can be found in a similar fashion for other harmonics, and it is easily seen
by graphical means that

n -1
SFDR n = (PIPn - No ) (5.101)
n

References

[1] D. M. Pozar, Microwave Engineering, 3rd ed. New York: John Wiley & Sons, 2005.
[2] K. Chang, Ed., Handbook of Microwave and Optical Components Vol 1: Microwave Pas-
sive and Antenna Components. New York: John Wiley & Sons, 1990.
[3] K. Chang, Ed., Handbook of Microwave and Optical Components Vol 2: Microwave Solid-
State Components. New York: John Wiley & Sons, 1990.
[4] P. L. D. Abrie, Design of RF and Microwave Amplifiers and Oscillators. Boston: Artech
House, 1999.
[5] P. B. Kenington, High-Linearity RF Amplifier Design. Boston: Artech House, 2000.
[6] S. A. Maas, Microwave Mixers, 2nd ed. Maas: Artech House, 1993.
[7] J. B. Johnson, “Thermal Agitation of Electricity in Conductors,” Physical Review, vol. 32,
p. 97, 1928.
[8] H. Nyquist, “Thermal Agitation of Electric Charge in Conductors,” Physical Review, vol.
32, p. 110, 1928.
[9] J. D. Kraus, Radio Astronomy. New York: McGraw-Hill, 1966.
[10] F. N. H. Robinson, Noise and Fluctuations in Electronic Devices and Circuits. Oxford:
Clarendon Press, 1974.
[11] W. B. Davenport and W. L. Root, An Introduction to the Theory of Random Signals and
Noise. New York: McGraw-Hill, 1958.
[12] P. E. Pace, Advanced Techniques for Digital Receivers. Norwood: Artech House, 2000.
[13] W. Kester, Ed., The Data Conversion Handbook. Oxford: Elsevier, 2005.
Chapter 6

Radiometry

Remote security sensors provide a measure of some property of the object of inter-
est, such as temperature, range profile, or velocity. The methods available to mea-
sure the desired quantity can be grouped into active sensors and passive sensors.
The primary distinction is that active sensors transmit energy and measure that is
reflected off the object, while passive sensors measure energy that is intrinsically
radiated by the object or generated elsewhere and reflected off the object. At the
system level, active sensors include both a transmitter and a receiver, whereas pas-
sive sensors include only a receiver.
A radiometer is a passive sensor that measures the electromagnetic power
radiating from or reflected off an object or scene of interest, specifically thermal
electromagnetic radiation. In the microwave and millimeter-wave regions of the
electromagnetic spectrum, the thermally radiated power from a terrestrial object is
linearly proportional to its physical temperature. Thus, by measuring the radiated
power, a radiometer provides a measure of the object’s temperature. Microwave
radiometers have been used extensively in radio astronomy [1–4] and satellite re-
mote sensing [5, 6], and there has been increasing interest in the last few years in
applying radiometry to security sensing, where it can be used to detect temperature
differences to locate concealed objects beneath clothing, create images of scenes
and people, and detect the presence of a human in clutter or through walls, whether
moving or stationary.
Microwave and millimeter-wave thermal radiation emanating from objects found
in terrestrial environments, which have temperatures on the order of 300 K, is very
low in power, on the order of 10–10 W or lower. Microwave and millimeter-wave ra-
diometers must therefore have very high gain to respond to the extremely low power
levels of the signals of interest. In addition, the power level of the thermal radiation
from the background can be very close to the temperature of the object of interest,
and thus radiometers must also have very high sensitivity in order to resolve small
temperature differences.
Microwave and millimeter-wave radiometry has been applied to a number of
applications in security sensing. There has been significant interest in implementing
radiometric arrays as imagers to detect concealed objects; imaging techniques are
the topic of Chapter 8. Such imagers exploit the negligible attenuation that micro-
wave and millimeter-wave radiation experiences when passing through clothing
materials to provide a temperature profile in order to detect concealed objects such
as weapons. Metals are highly reflective at millimeter-wave frequencies and below,
and when placed in front of a human, reflect the temperature of the surroundings,
which is generally colder than the human itself. Thus the object stands out as a
cold source in front of the warm human. Recent research has applied radiometry
to the detection of human presence in outdoor environments for the detection of

173
174 Microwave and Millimeter-Wave Remote Sensing for Security Applications

stationary humans hiding amongst clutter [7–12], and moving humans [13, 14].
When used in such applications, a radiometer operating a total power or correla-
tion detection mode passively scans a field of view, and a temperature difference is
measured if the person is present in the antenna beam. In the detection of moving
humans, the radiometer may operate in a fixed staring mode, where a temperature
difference is measured when the human enters the antenna beam. Microwave and
millimeter-wave radiation transmits favorably through some wall materials, such
as drywall, and thus there has been interest in applying radiometry to through-wall
detection [15, 16]. Radiometry has also been applied to the detection of landmines.
Millimeter-wave frequencies have been shown to be applicable for surface mine
detection [17, 18]; however, the detection of subsurface landmines is significantly
more challenging due to the dielectric layers covering the mine. In either case, mines
or other objects having low emissivities and placed below the ground surface will
reflect radiation from the sky, which has microwave temperatures on the order
of 10–20 K and can thus be distinguished from the warmer soil surrounding the
object. Microwave radiation can penetrate soil to various depths, depending on
several factors such as salinity and density; radiometers operating in the frequency
ranges of 1–10 GHz have been developed that can reliably detect buried objects at
depths of only a few inches [19, 20].
This chapter provides a theoretical overview of microwave and millimeter-wave
radiometry, and focuses on the essential material that is needed to understand and
design radiometer systems. The first section covers fundamental radiometry con-
cepts, including brightness and flux density. Following this is a detailed develop-
ment of thermal blackbody radiation theory and its application to terrestrial objects
at microwave and millimeter-wave frequencies. The next section covers applied
radiometry concepts including radiometric temperature and radiometer sensitivity.
Radiometer receivers are covered in the following section, focusing on the total
power and correlation radiometers, their responses, and their sensitivities. Finally,
the chapter concludes with a section discussing practical considerations for design-
ing radiometers.

6.1  Radiometry Fundamentals

6.1.1  Brightness
The function of a radiometer is to measure the electromagnetic energy radiated by
a source of interest. For security sensors the electromagnetic energy is thermally
generated, and in the microwave and millimeter-wave region the radiated power is
proportional to the physical temperature of the source for most sources of interest.
Millimeter-wave radiometers thus provide a measure of the physical temperature
of the source. The quantity that is measured by the radiometer is the brightness of
the source, which has units of W·m–2·str–1. The power received by the radiometer
is a function of the brightness as well as the spatial filtering properties of the receiv-
ing aperture, which is dependent on the antenna pattern and the bandwidth. As

 The term brightness is generally used in microwave and millimeter-wave remote sensing terminology. In
optics, this quantity is called radiance.
6.1  Radiometry Fundamentals 175

discussed in Chapter 5, the receiver hardware imparts various nonidealities onto


the received signal, such as noise and distortion. The noise power generated by the
receiver is in general much larger than the detected power of the source in security
sensing.
The thermal power radiated by a source and received by a radiometer can be
described by considering the radiation incident on a planar surface. The geometry
is described graphically in Figure 6.1. The infinitesimal power at the surface is given
by

dP = Bf (f , θ , φ)cosθ d WdAdf (6.1)

where Bf is the spectral brightness of the source with units W·m–2·Hz–1·str–1. The
cosθ term accounts for the projected area of the surface at angles away from broad-
side. The power collected by the aperture is found by integrating (6.1) over angle,
the aperture area Ar, and frequency:

P=ò ò òò Bf (θ, φ)cosθ dWdAdf (6.2)



f Ar W

where Ar defines the area bounding the receiving surface. The total power can be
given in terms of the integral over frequency of the spectral power, which is the
frequency-dependent radiation power of the source, in units of W·Hz–1. That is,

P = ò Pf (f )df (6.3)
f
where


Pf = ò òò Bf (θ, φ)cosθ dWdA (6.4)
Ar W

is the spectral power. The spectral power can be used in characterizing systems
where the spectral brightness, and therefore the spectral power, can be considered
constant over the bandwidth of interest. For a rectangular passband, the total power
is then simply the spectral power multiplied by the bandwidth.

At

R
Ωt

θ Ar

Figure 6.1  Geometry describing the radiation power incident on a receiving aperture from a gen-
eral source.
176 Microwave and Millimeter-Wave Remote Sensing for Security Applications

The brightness is in general a function of frequency; integrating over a finite band-


width gives the total brightness (W·m–2·str–1), also called simply the brightness,
within that bandwidth:
f +Df
B(θ , φ ) = ò Bf (f ,θ , φ)df (6.5)
f

The total band-limited power incident on the surface is found by integrating


(6.1) and substituting (6.5):

P= ò òò B(θ,φ)cosθ dWdA (6.6)


Ar W

If the source is significantly far from the surface, the incident radiation is nearly
constant across the surface, and the power can then be given by

P = Ar òò B(θ , φ )cosθ d W (6.7)


W
If the brightness is constant over the hemisphere, the power may be written
2π π 2
P = Ar B ò ò cos θ sinθ dθd φ (6.8)
0 0

The integral in (6.8) is equal to π, and thus the power incident on the surface from
a spatially constant brightness is

P = π Ar B (6.9)

6.1.2  Brightness and Distance


It is important to note that the brightness is a measure of the power radiated by the
source per unit area per unit solid angle. Because it is dependent on the differential
solid angle, the brightness describes the radiation carried by an individual ray of
radiation. Therefore, the brightness does not change with distance. To see this,
consider an infinitesimal ray of radiation propagating from the source to the receiv-
ing surface, as depicted in Figure 6.2. For simplicity, it is assumed that the source
is located broadside to the receiving surface so that cosθ = 1. Because the beam is
infinitesimal in angle, it does not spread in angle as it propagates over the distance
R. If the radiation is propagating in free space, and therefore does not encounter
absorption, scattering, or diffraction, the differential power emitted by the source is
equal to that received by the receiving surface by conservation of energy:

dPs = dPr (6.10)

 Integrating the power over both hemispheres with constant brightness results in òcosqdW = 0; the power
incident on the surface is equal to the power exiting the other side.
6.1  Radiometry Fundamentals 177

dAs dAr

Ωr Ωs

R
Figure 6.2  An infinitesimal ray of radiation propagates from the source to the receiving surface over
a distance R. The source is located broadside to the receiving surface so that the Lambertian cosine
term can be neglected.

If the differential solid angles of the source and receiving surface are given by
dWs and dWr, respectively, the differential powers emitted by the source and inter-
cepted by the receiving surface are

dPs = Bs dAs d W s (6.11)

and

dPr = Br dAr d Wr (6.12)

where Bs and Br are the brightness of the source and receiving surface, respectively,
and As and Ar are the areas of the source and receiving surface, respectively.
The differential area of the source surface can be given in terms of the solid
angle of the receiving surface by

dAs = R2d Wr (6.13)

and likewise

dAr = R2d W s (6.14)

The differential powers can then be given by

dAr
dPs = Bs dAs (6.15)
R2

and

dAs
dPr = Br dAr (6.16)
R2

Substituting (6.15) and (6.16) into (6.10) results in

Br = Bs (6.17)

Thus, the brightness at the source is equal to that received by the surface: the
brightness is independent of the distance travelled. This is a very important property
because, as will be seen in Section 6.2, the brightness is proportional to the physical
temperature of the source. Therefore, the temperature of a source can be measured
by a radiometer regardless of the distance separating them, so long as the power
178 Microwave and Millimeter-Wave Remote Sensing for Security Applications

incident on the receiving aperture is large enough to be detected by the receiver. In


radio astronomy the brightness of stellar objects thousands of light years away has
been measured, giving insight into the physical processes of galactic nebulae, black
holes, and numerous other objects. This is possible because the representation of
the object’s physical temperature does not change as a function of distance.

6.1.3  Flux Density and Source Distribution


Sources of radiation that subtend an infinitesimally small angle are referred to as
point sources. While true point sources are not encountered in nature, they are
useful analytic tools that can aid in the development of theories, design of systems,
and evaluation of sensor properties. Point sources can be approximated in practice,
for calibrating scanning radiometers for instance, by using a signal generator or
noise source and antenna placed in the far-field of the radiometer antenna. Sources
encountered in practice subtend a finite angle; however, for sources of small angu-
lar extent, such as stars for instance, they may be approximated as point sources.
Most sources encountered in terrestrial security sensing, such as humans, animals,
or various environmental clutter such as vegetation, subtend an appreciable angle
and are called extended sources.
While brightness is the primary quantity of interest in measuring the radia-
tion of a source, another useful quantity is the flux density of a source in units
of W·m–2, which is the integral of the brightness over the spatial extent of the
source:

S= òò B(θ, φ)dW (6.18)


Ws

where Ws is the solid angle subtended by the source. Here the cosθ dependence is
not included in order to make the flux density a function of the source only; the
flux density observed at the receiver would include the cosθ term. The brightness
distribution of a point source can be represented simply by a delta function at a
specific angle (q0,f0), scaled by the brightness. The flux density of a point source is
therefore

Sp = òò Bp (θ, φ)δ(θ - θ 0 )δ(φ - φ0 )dW = Bp (θ 0 ,φ 0 ) (6.19)


Ws

or simply the brightness of the point source.


The flux density collected by a receiving aperture, due to the cosθ dependence,
is referred to as the observed flux density, and is the source flux density modified by
the spatial filtering function of the aperture. It is given by

So = òò B(θ, φ)cosθ dW (6.20)


Ws

for a general receiving aperture. The spatial filtering of the aperture causes the ob-
served flux density to be less than or equal to the actual source flux density. For a
point source at angle (q0,f0) the observed flux density is
6.1  Radiometry Fundamentals 179

Sp = Bp (θ 0 ,φ 0 )cosθ 0 (6.21)

Although the brightness is independent of distance, the flux density, being the in-
tegral of the brightness over the angular extent of the source, is proportional to
the inverse square of the distance. To see this, consider a sphere of radius r with a
uniform brightness B, depicted in Figure 6.3. Using (6.18), the flux density observed
by an aperture at a distance R is given by

2π θ s
So = òò B cosθ dW = B ò ò cosθ sinθ dθdφ (6.22)
Ws 0 0

where θs is half of the angle subtended by the source. The integral is equal to
πsin2θs, and since

r
sin θ s = (6.23)
R

the flux density is

r2
So = π B (6.24)
R2

Thus, the observed flux density is proportional to the inverse square of the distance,
as opposed to the brightness, which does not change with distance.

6.1.4  Effect of the Antenna


If the receiving aperture represents that of an aperture antenna, the cosθ depen-
dence is replaced by the antenna pattern A(q,f). If it is assumed that the source is
significantly far from the antenna that the incident radiation is constant over the
aperture, the received power is given by

1
2 òf òò
P= Ae Bf (θ , φ)A(θ , φ) d Wdf (6.25)
W

where Ae is the effective aperture of the antenna, defined in Section 4.2.7. The ½
term is included in (6.25) to represent the effect of the polarization of the antenna:
thermal sources emit incoherent radiation that is unpolarized, and therefore the

Ar
r
θs
R

Figure 6.3  Flux density from a sphere with uniform brightness B.


180 Microwave and Millimeter-Wave Remote Sensing for Security Applications

polarized antenna can only respond to one of the two components of polarization.
The spectral power is thus
1
2 òò
Pf =
Ae Bf (θ , φ)A(θ , φ) d W (6.26)
W
The observed flux density of the source as seen by the receiver, including the an-
tenna pattern, is
So = òò B(θ , φ)A(θ , φ)d W (6.27)
W

6.2  Blackbody Radiation

Thermal radiation is the intrinsic electromagnetic energy emitted by an object due


to thermal agitation of electrons present in the material composing the object. All
objects with a physical temperature above absolute zero emit thermal radiation. A
blackbody is a hypothetical object in thermal equilibrium that absorbs all energy in-
cident upon it, without reflecting any energy. By Kirchhoff’s radiation law, the power
absorbed by an object in thermal equilibrium is equal to the power emitted, and a
blackbody is therefore also a perfect radiator of electromagnetic energy. To better un-
derstand the concept of a blackbody and its properties, it is typically approximated by
an enclosure in thermal equilibrium at temperature T. The enclosure is made to have
a small opening through which radiation may pass, as seen in Figure 6.4. The radia-
tion entering the opening is reflected about the inside of the enclosure and eventually
absorbed by the walls; if the hole is much smaller than the area on the inside of the
cavity, a negligible amount of radiation will have escaped back out through the open-
ing. Thus, all energy incident upon the opening is completely absorbed. Additionally,
the thermal radiation generated by the walls inside the enclosure is emitted through
the opening; since the opening is a perfect absorber, it must also be a perfect radiator,
and thus the hole has the same properties as the surface of a blackbody.

6.2.1  Planck’s Blackbody Radiation Law


The spectral brightness of the radiation emitted by a blackbody is given by Planck’s
law
2hf 3 1
Bf = 2
× hf kT (6.28)
c e -1

Figure 6.4  Approximation of a blackbody. All radiation incident on the opening is absorbed within
the enclosure, and all radiation emitted by the opening is thermally generated within.
6.2  Blackbody Radiation 181

where Bf is the spectral brightness of the blackbody, and h is Planck’s constant.


Planck’s equation gives the radiation emitted by a perfect emitter in terms of fre-
quency and the physical temperature of the object or medium. Figure 6.5 shows
the brightness of a blackbody over frequency at a few representative temperatures,
including the temperature of the sun, a “red hot” object (1,000 k), a human body,
and the cosmic microwave background (3k). Because (6.28) is not dependent on
direction, a blackbody is a Lambertian surface. This means that the power per unit
solid angle (the radiant intensity) is dependent on the cosine of the angle between
the normal direction of the surface of the blackbody and the observations point.
Because the projected area also decreases as the cosine of the angle, the brightness
remains constant.
Planck’s law may also be cast in terms of unit wavelength interval, rather than
unit bandwidth, by noting that

Bf df = Bλ d λ (6.29)

and using
c
f = (6.30)
λ

and
df c
= - 2 dλ (6.31)
dλ λ

The spectral brightness in terms of unit wavelength interval is then

2hc 2 1
Bλ = 5
× hc λkT (6.32)
λ e - 1

10-5

6000 K
10-10 (Sun)
B (W m-2 Hz-1 str-1)

1000 K
10-15 310 K
(human)
100 K
10-20
3K

10-25

10-30 5
10 1010 1015 1020
f (Hz)
Figure 6.5  Radiation curves for a blackbody at various temperatures over frequency.
182 Microwave and Millimeter-Wave Remote Sensing for Security Applications

which has units of W·m–3·str–1. Note that although the curves resulting from (6.28)
and (6.32) are of the same shape, the peaks occur at different locations; that is,
the peak of a given curve derived from (6.28) does not align with the peak for a
blackbody of the same temperature derived from (6.32) by simply using (6.30).
Figure 6.6 shows the brightness curves of Figure 6.5 over wavelength.
Some properties of Planck’s law may be discerned by noting the differences
in the curves as the temperature is changed. As the temperature of the blackbody
increases, the location of maximum radiation occurs at increasing frequency; this
is known as the Wien displacement law and can be found by differentiating (6.28)
with respect to f and equating the result to zero:

dBf 2hfm2 æ hfm kT hfm hfm kT ö


= çè e - e - 1÷ = 0 (6.33)
df c2 kT ø

where fm is the frequency at the maximum point of the curve. This reduces to

hfm )
= 3 (1 - e -hfm kT
(6.34)
kT

Substituting x = hfm / kT in (6.34) gives x = 3(1 – e–­x), the numerical solution of


which is approximately x = 2.82; thus,

hfm = 2.82kT (6.35)

or, equivalently,

fm = 5.87 ´ 1010 T (6.36)

For example, the typical human body temperature is T = 310.15 K, placing the
peak in the blackbody spectrum at fm = 1.82×1013 Hz, at which the wavelength is

1020

1010 6000 K
(Sun)
B (W m-3 str-1)

1000 K
100 310 K
(human)
10-10 100 K
3K

10-20

10-30
10-10 10-5 100
λ (m)
Figure 6.6  Radiation curves for a blackbody at various temperatures over wavelength.
6.2  Blackbody Radiation 183

λm = 1.65 μm (18.2 THz), which is in the mid-infrared band. Human skin behaves
nearly identical to a blackbody in the infrared band, and this is the reason that in-
frared imaging of human thermal radiation has been applied with such success.
The Wien displacement law can also be found for Bλ by taking the derivative
of (6.32) with respect to wavelength and setting the result equal to zero, which
results in

2.9 ´ 10-3
λm = (6.37)
T
An example is the radiation emitted by the sun, the curve of which is shown in
Figure 6.6. The temperature at the surface of the sun is approximately 6000 K, and
using (6.37) gives λm = 480 nm, which corresponds to the center of the optical band
of the electromagnetic spectrum. Human eyes have thus evolved to respond to the
wavelengths corresponding to the strongest brightness of the sun.
From the plots of Planck’s function, it can also be seen on the graphs that the
brightness of a blackbody is monotonic with temperature. That is,

dBf
>0 (6.38)
dT
regardless of location on the curve. Thus, an increase in the temperature of a black-
body at a given frequency always results in an increase in brightness.
Planck’s law gives the spectral brightness emitted by a blackbody as a function
of temperature and frequency. The total brightness is found by integrating (6.28)
over all frequencies:
¥ ¥
2h f3
B = ò Bf df = ò df (6.39)
c 2 0 e hf kT - 1
0
This equation may be more easily solved by letting x = hf / kT, which gives


2h æ kT ö x3
B= ç
c2 è h ø
÷ ò e x - 1 dx (6.40)
0

The integral is a constant equal to π 4/15. Thus,

2π 4 k 4 4
B= T (6.41)
15c 2 h3

The radiant emittance M in units of W·m–2, is found by integrating (6.41) over


the hemisphere that the blackbody radiates into. Because a blackbody is a Lamber-
tian radiator, the cosine of the angle must be included:
π
2π 2
M= ò ò B cosθ sinθ dθdφ = π B (6.42)
0 0
184 Microwave and Millimeter-Wave Remote Sensing for Security Applications

or

M = σ T 4 W × m -2 (6.43)

where

2π 5k4
σ= = 5.67 ´ 10-8 (6.44)
15c 2 h3

Equation (6.43) is the Stefan-Boltzmann Law and (6.44) is the Stefan-Boltzmann


constant, (W·m–2·K–4). This law states that the radiant emittance of a blackbody is
proportional to the fourth power of its physical temperature.

6.2.2  Approximations of Planck’s Law


Two approximations of Planck’s law are useful simplifications of the expression for
the spectral brightness of a blackbody. The first is applicable when hf << kT. The
exponential in (6.28) can be replaced by its power series

¥
x2 xn
ex = 1 + x + +…= å (6.45)
2! n =0
n!

Under the approximation that hf << kT,

hf
e hf kT
» 1+ (6.46)
kT

and Planck’s law reduces to

2 f 2kT
Bf = (6.47)
c2

This is the Rayleigh-Jeans Law, and it states that the spectral brightness of a
blackbody is proportional to its physical temperature and the square of the fre-
quency. This law was initially derived independently of Planck, based on classical
considerations of the electromagnetic fields within a blackbody cavity. Note the
limitation of the Rayleigh-Jeans approximation that as the frequency increases,
the spectral brightness diverges toward infinity; this is known as the ultraviolet
catastrophe.
The requirement that hf << kT essentially limits the use of the Rayleigh-Jeans
law to frequencies below the frequency of maximum brightness; closer to the peak
of the blackbody curve the deviation of the Rayleigh-Jeans law from Planck’s law
begins to increase significantly. Although the law does not hold for low frequencies
combined with very low temperatures, typical objects and backgrounds encoun-
tered in practical applications of security sensing and remote sensing have high
enough temperatures that the law is a very good approximation. For example, at
the temperature of a human T = 310.15 K, the frequency at the peak of the bright-
6.2  Blackbody Radiation 185

ness curve was found to be f = 18.2 THz. At this frequency, the Rayleigh-Jeans
law (6.47) predicts a brightness 447% greater than that given by Planck’s law
(6.28). At f = 180 GHz (two orders of magnitude below the peak) the error is only
1.41%.
At much higher frequencies, where hf >> kT,

e hf kT
- 1 » e hf kT
(6.48)

and Planck’s law reduces to

2hf 3 - hf kT
Bf = e (6.49)
c2
This is the Wien law, and it accurately approximates Planck’s law for frequen-
cies greater than the frequency of maximum brightness. This typically means infra-
red and above. Thus, the Wien law finds its use in infrared and optical applications;
it is of limited use for microwave and millimeter-wave sensors.
Planck’s law is compared to the Rayleigh-Jeans and Wien laws in Figure 6.7.
The Rayleigh-Jeans law coincides with Planck’s law at low frequencies and diverges
as the frequency increases, whereas the Wien law coincides with Planck’s law at
high frequencies. It can be seen that both approximations fail near the peak of
the blackbody radiation curve, where Planck’s law must be used in its exact form.
The simplicity of the Rayleigh-Jeans law is illustrated by its dependence only on the
square of the frequency.

6.2.3  Band-Limited Integration of Planck’s Law


The Stefan-Boltzmann law gives the total brightness emitted over all frequencies,
but in practice, radiometer systems will have limited operating bandwidth. In
order to properly estimate the power radiated by the object or scene of interest,
it is desirable to know the total brightness emitted in the limited bandwidth of

10-8
Rayleigh-Jeans
10-9
B (W m-2 Hz-1 str-1)

Planck
10-10

10-11
Wein
10-12

10-13

10-14 11
10 1012 1013 1014 1015
f (Hz)
Figure 6.7  Brightness of a blackbody at 1000 K. The Rayleigh-Jeans law accurately approximates
Planck’s law for frequencies below the frequency of maximum brightness, while the Wien law ap-
proximates Planck’s law for frequencies above.
186 Microwave and Millimeter-Wave Remote Sensing for Security Applications

the receiver. An exact band-limited integration of (6.28) requires the evaluation of


incomplete polylogarithm functions. However, a more practical numerical solution
was found by Widger and Woodall [21], which was derived from the Planck func-
tion in terms of unit wavenumber; the following derives the solution in terms of
unit bandwidth.
The one-sided integral of (6.28) is

¥
2hf 3 1
B(f ¢) = ò c2 × ehf kT - 1 df (6.50)
f'

Letting x = hf / kT gives

¥
2k4T 4 x3
B(x ¢) = 2 3
c h
ò e x - 1 dx (6.51)
x'

Now using
¥
1
= å e -nx
e x - 1 n =1
(6.52)

results in

¥ ¥
2k4T 4
B (x ') = å ò x3e -nx dx (6.53)
c 2 h3 n =1 x '

which can be integrated by parts, resulting in

¥ æ x3 3x 2 6x 6 ö
2k 4 T 4
B( f ¢) =
c 2 h3
å e-nx çè n
+ 2 + 3 + 4÷
n n n ø
(6.54)
n =1

The brightness over a finite bandwidth is then found by taking the difference of two
one-sided integrals

B(Df ) = B(f1) - B(f2 ) (6.55)

For hf >> kT, convergence of (6.54) to within 10 significant digits of (6.28) is found
with n £ 3, where n is the number of terms in the summation in (6.54). For hf <<
kT, n ³ 100 is required for similar performance [21].
Microwave and millimeter-wave radiometers used for security purposes will
typically be viewing objects whose temperature is on the order of 290 K, such
as humans, building walls, and vegetation, among other typical indoor or out-
door clutter. In such a case, the band-limited brightness can be approximated
6.3  Applied Radiometry 187

using the Rayleigh-Jeans law, which is linear with frequency. Integrating (6.47)
over bandwidth Df gives

2kT æ 2 1 ö
B= ç f Df + Df 3 ÷ (6.56)
c2 è 12 ø

For narrowband systems, Df << f and (6.56) reduces to

2f 2
B= kT D f (6.57)
c2
Equation (6.57) provides a simple, approximate calculation of the band-limited
brightness for systems operating at microwave and millimeter-wave frequencies.

6.3  Applied Radiometry

In microwave and millimeter-wave radiometers used in security sensing, the fre-


quencies and source temperatures are such that the Rayleigh-Jeans approximation
to Planck’s radiation law is accurate for nearly all situations. This approximation
provides a useful simplification of Planck’s law in that the brightness of a source can
be linearly related to its physical temperature. Substituting the Rayleigh-Jeans law
(6.47) into the general expression for the spectral power received by a radiometer
antenna (6.26) yields

Ae
Pf = k
λ2
òò T (θ , φ)A(θ ,φ) d W (6.58)
W

where T(θ,f) is the source temperature distribution. In Section 4.2.7 it was shown
that the antenna effective area and the pattern solid angle are related through

Ae W A = λ 2 (6.59)

and thus (6.58) can be written

k
Pf =
WA òò T (θ , φ)A(θ , φ)dW (6.60)
W

The spectral power given by (6.58) represents the power collected by the antenna
aperture. From Chapter 4, the total power at the antenna terminals can be given by

PA = kò TAdf (6.61)

where TA is the antenna temperature expressed a function of frequency. Thus the
spectral power at the antenna terminals is

Pf = kTA (6.62)
188 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Equating (6.60) and (6.62) yields

1
TA =
WA òò T (θ , φ)A(θ , φ) dW (6.63)
source

Because the spectral brightness is proportional to the temperature through


(6.47), the brightness at the antenna terminals, called the apparent spectral bright-
ness, is

1 Sof
BAf =
WA òò Bf (θ, φ)A(θ, φ)d W = WA (6.64)
Ws

where Sof is the observed spectral flux density. The apparent spectral brightness is thus
the observed spectral flux density normalized by the antenna pattern solid angle.

6.3.1  Source Resolution


The spectral power collected by the radiometer antenna is dependent on the ratio
of the angular extent of the source and the angular extent of the antenna pattern.
If the antenna main beam solid angle WM is narrower in angular extent than the
source solid angle Ws, or WM < Ws as shown in Figure 6.8(a), the source is said to be
resolved in angle. When the source is narrower in angular extent than the antenna
main beam, or WM > Ws as shown in Figure 6.8(b), the source is said to be unre-
solved in angle.

6.3.1.1  Resolved Source


For a resolved source with a temperature distribution that is constant over the spa-
tial extent of the source, the spectral power is given by

k
Pf = T òò A(θ , φ) dW (6.65)
W A source

ΩM

(a)

ΩM

(b)
Figure 6.8  (a) Observation of a resolved source. (b) Observation of an unresolved source.
6.3  Applied Radiometry 189

This can be written in terms of the antenna pattern solid angle extending over the
source, given by

Wp = òò A(θ , φ) d W (6.66)
source

Note that Wp includes the contribution from the main lobe and any sidelobes that
cover the solid angle subtended by the source. The spectral power can then be given
by

Wp
Pf = k T (6.67)
WA

For a temperature distribution that is constant over the entire hemisphere, the
source solid angle is replaced by the hemisphere limits. The integral of the antenna
pattern results in the antenna pattern solid angle WA, and thus

Pf = kT (6.68)

The spectral noise power is thus simply the expression for the spectral power
of a noisy resistor: the source physical temperature multiplied by Boltzmann’s con-
stant. A temperature distribution that is constant over the viewing hemisphere of
the antenna would be encountered in blackbody enclosures, which are used for
receiver calibration. If the temperature is constant over the bandwidth, the total
power is

P = kT D f (6.69)

which is equivalent to the noise power generated by a resistive material at tempera-


ture T.
If the source solid angle coincides with the antenna main beam solid angle, the
spectral power is

k W
Pf = T òò A(θ , φ)d W = k M T (6.70)
W A source WA

where
WM = òò A(θ , φ) dW (6.71)
main
beam
is the main beam solid angle.

6.3.1.2  Unresolved Source


If the angular extent of the source is not greater than that of the antenna main
beam, the source is unresolved in angle. If the source is small and such that the
190 Microwave and Millimeter-Wave Remote Sensing for Security Applications

antenna pattern is nearly constant over the angular extent of the source, the spectral
power is

k W
Pf = A(θ0 ,φ 0 ) òò T (θ, φ) d W = k s Tavg A(θ0 , φ0 ) (6.72)
WA WA
source

where (θ0,f0) is the angular location of the center of the source, A(θ0,f0) is the an-
tenna directivity in the direction of the source, and Tavg is the average temperature
of the source, obtained by integrating the temperature angular extent of the source.
If the radiating source is located at the position of the antenna’s maximum directiv-
ity, A(θ0,f0) » 1 and

Ws
Pf = k Tavg (6.73)
WA

The antenna temperature is then found through (6.62) to be

Ws
TA = Tavg (6.74)
WA

6.3.2  Received Power as a Convolution


When the antenna is viewing a finite temperature distribution and the antenna
is not directed toward the source center, such as when an antenna scans over a
temperature distribution, the received power is given in terms of the center of the
distribution θ and the pointing angle of the antenna θ­ – θ0 by


A
Pf = k 2e
λ
ò T (θ)A(θ - θ0 )dθ (6.75)
0
where the one-dimensional case is considered for simplicity. The analysis that fol-
lows extends similarly in the orthogonal spatial dimension. The integral in (6.75)
is a correlation integral, and thus the power received is the cross-correlation of the
temperature distribution and the antenna pattern. If the antenna pattern is symmet-
ric, A(θ­) = A(–θ­), and the power becomes

A
Pf = k 2e
λ
ò T (θ)A(θ0 - θ)dθ (6.76)
0

The integral in this equation is a convolution integral. It is often the case that
the antenna pattern is symmetric, and thus the received power from a distributed
source can be given in terms of the convolution of the temperature distribution and
the antenna pattern:
Ae
Pf = k T *A (6.77)
λ2
6.3  Applied Radiometry 191

The reconstructed temperature distribution is thus widened in the spatial domain


by the convolution with the antenna pattern.

6.3.3  Emissivity and Radiometric Temperature


The previous formulations have been derived based on the idea that the radiating
source is a blackbody, which is a perfect radiator. The concept of a blackbody is a
powerful analytical tool; however, blackbodies do not exist in the natural world.
Although materials are manufactured to be used as blackbodies for calibration
purposes, these materials are generally only good approximations of blackbodies
over narrow frequency ranges, and absorb as much as 98% of the incident radia-
tion over these frequencies. Thus, all materials in the natural world are greybodies,
as bodies do not perfectly absorb all of the incident radiation.
Kirchhoff showed that a perfect absorber is also a perfect radiator. A greybody,
being an imperfect radiator, will therefore emit less radiation than a blackbody (a
perfect radiator) at the same physical temperature. Referring to (6.47), the spectral
brightness of a blackbody under the Rayleigh-Jeans approximation is

2k
B f , bb(θ , φ) = T (θ, φ) (6.78)
λ2
where the subscript bb has been included to indicate the brightness of a blackbody.
The spectral brightness of a greybody can thus be given by

2k
Bf (θ , φ) = TR (θ , φ) (6.79)
λ2

where TR­is the radiometric temperature or brightness temperature of the greybody.


The radiometric temperature, if there is no prior knowledge of the object, is the ap-
parent physical temperature of the object assuming it is a blackbody. Its apparent
temperature differs from its physical temperature due to absorption of a fraction
of the incident radiation. Thus, without prior knowledge of the object, its physical
temperature cannot be exactly known.
Taking the ratio of (6.79) to (6.78),

Bf ,bb TR
= = ef (θ , φ) (6.80)
Bf T
where ef is the emissivity of the greybody, which is in general a function of fre-
quency and position on the object. If the object is homogeneous, the emissivity is
only a function of frequency. From (6.80), the radiometric temperature at a given
frequency can be given by

TR = eT (6.81)

The emissivity thus characterizes how well a source radiates compared to a black-
body radiator at the same physical temperature. A greybody can only radiate as
much as, and not more than, a blackbody, and therefore 0 £ e £ 1.
192 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Emissivity can also be defined in terms of the reflectance and transmittance of


the object. Consider an object in thermal equilibrium that is illuminated by an in-
cident flux density Sinc. The incident flux must be equal to the flux that is reflected,
transmitted, and absorbed by the object. That is,

Sinc = Sr + St + Sa (6.82)

where Sr is the flux reflected by the object, St is that transmitted through the object,
and Sa is that absorbed by the object. Each component on the right-hand side of
(6.82) is dependent on the physical makeup of the object, and each term can be
given by the fractional amount of the incident flux that it represents:

Sinc = GSinc + ¡Sinc + aSinc = (G + ¡ + a)Sinc (6.83)

and thus

G + ¡ + a = 1 (6.84)

where G, ¡, and a are the reflectivity, transmissivity, and absorptivity, respectively,


of the object. By Kirchhoff’s radiation law, the power absorbed by an object in
thermal equilibrium is equal to the power emitted, and thus

a = e (6.85)
Equation (6.84) can then equivalently be written

G + ¡ + e = 1 (6.86)

For an opaque object, such as a human, the transmissivity is generally negligible


at microwave and millimeter-wave frequencies due to the thickness of the object,
and thus

e = 1 - G (6.87)

6.3.3.1  Emissivities of Human Skin and Common Materials


The emissivity is in general a function of frequency. Consider, for example, the
emissivity of human skin. At infrared frequencies, the emissivity of human skin has
been measured at 0.996 [22], which indicates that skin is closely approximated by
a blackbody in the infrared band. This is the reason that infrared sensors have been
developed and applied with such success to the detection of humans, under certain
conditions. In contrast, the emissivity of a human at 30 GHz has been measured
to be 0.50–0.64 [23]. Thus, a material that approximates a blackbody well at one
frequency does not necessarily at another. The radiometric temperature of human
skin in the infrared region is, assuming that the skin is 98.6° F = 310 K,

TR = 0.997 ´ 310 = 309.22 K (6.88)


6.3  Applied Radiometry 193

or very nearly the physical temperature of the skin. At 30 GHz, the radiometric
temperature is

TR = 0.64 ´ 310 = 198.50 K (6.89)

which is considerably lower due to the lower emissivity, and thus less thermal power
is radiated than at infrared frequencies.
The reflectivity, transmissivity, and emissivity of human tissue calculated from
the model by Gabriel [24–26], discussed in Chapter 3, are given over the frequency
range of 10 Hz to 100 GHz in Figure 6.9. Note that the measured emissivity of
0.50–0.64 at 30 GHz includes contributions from the skin and the underlying layers
and is thus between the predicted emissivity values of skin and fat. The transmis-
sivity of individual tissues is not zero; however, due to the overall thickness of the

1 1
bone cartilage bone cartilage
fat muscle fat muscle
0.8 dry skin wet skin 0.8 dry skin wet skin

0.6 0.6
|Υ|
|Γ|

0.4 0.4

0.2 0.2

0 0
0 20 40 60 80 100 0 20 40 60 80 100
f (GHz) f (GHz)

(a) (b)

0.8

0.6
e

0.4

0.2 bone cartilage


fat muscle
dry skin wet skin
0
0 20 40 60 80 100
f (GHz)
(c)
Figure 6.9  (a) Reflectivity, (b) transmissivity, and (c) emissivity of human tissues.
194 Microwave and Millimeter-Wave Remote Sensing for Security Applications

human body, the total transmissivity is negligible in the microwave and millimeter-
wave frequency regions.
Table 6.1 gives the emissivities of a number of different materials at three
millimeter-wave frequencies [27]. As discussed in Chapter 3, the reflectivity of metals
is high due to their high conductivity, thus through (6.87) the corresponding emis-
sivity is low. Some materials, including ground materials such as gravel, asphalt,
concrete, and packed dirt, have emissivities close to unity.
Electromagnetic absorbing materials are commonly used for calibration of ra-
diometers (see Section 6.5.3) and closely approximate blackbody radiators. Such
materials, commonly referred to as microwave absorbers, are also used in anechoic
chambers to simulate free space and are manufactured out of lightweight polyure-
thane or urethane foam embedded with a lossy dielectric solution typically includ-
ing carbon. Often formed in two-dimensional arrays of pyramidal cones to reduce
backscattered reflections, planar absorbing panels are also used. Electromagnetic
absorbing materials are designed for high absorptivity; typical measured reflectivi-
ties for absorber cones are –50 dB (1×10–5) between 20 GHz and 95 GHz, increasing
to –30 dB (0.001) at lower millimeter-wave and microwave frequencies. At lower
frequencies, the pyramidal cones must be larger due to the increased wavelength,
which makes manufacturing low reflectivity geometries more difficult. Planar ab-
sorbing slabs typically have higher reflectivities of –30 dB (0.001) between 20 GHz
and 95 GHz. The extremely low reflectivities of absorber materials correspond to
emissivities very close to unity.

6.3.3.2  Radiometric Temperature in an Environment


The relationship between emissivity and reflectivity for an opaque object is conse-
quential in detection applications. Consider the case of a radiometer viewing an ob-
ject that is bounded by a constant temperature environment, as depicted in Figure
6.10. Both the object and the surrounding environment radiate thermal energy, and
the object fills the antenna main lobe while the detected radiation directly from the
environment in the sidelobes is considered negligible. The radiometric temperature
measured by the radiometer is of the form

TR = eoTo + G oTenv (6.90)

Table 6.1  Effective Emissivity at Normal Incidence of Common Materials at


Various Frequencies (© 2003 IEEE [27]).
Material surface 44 GHz 94 GHz 140 GHz
Bare metal 0.01 0.04 0.06
Painted metal 0.03 0.10 0.12
Painted metal under canvas 0.18 0.24 0.30
Painted metal under camouflage 0.22 0.39 0.46
Dry gravel 0.88 0.92 0.96
Dry asphalt 0.89 0.91 0.95
Dry concrete 0.86 0.91 0.95
Smooth water 0.47 0.59 0.66
Rough or hard-packed dirt 1.00 1.00 1.00
6.3  Applied Radiometry 195

Tenv

ΓoTenv
To, Γo, eo
eoTo
TR = eoTo + ΓoTenv

Figure 6.10  When a greybody object is surrounded by an environment at a nonzero temperature,


the measured temperature will be composed of the object’s temperature and the environmental
temperature reflecting off the object.

where To, eo, and Go are the physical temperature, emissivity, and reflectivity of the
object, respectively, and Tenv is the temperature of the surrounding environment.
Through (6.87), this can be written

TR = eoTo + (1 - eo )Tenv (6.91)

The radiometric temperature is a function of both the direct energy from the object
and energy generated from the surrounding environment that is reflected off the
object.
If the emissivity of the object is close to unity, the object is approximately a
blackbody and

TR » To (6.92)

Thus, all the energy generated by the environment that is incident on the object is
absorbed by the object. If the emissivity is close to zero,

TR » Tenv (6.93)

The object is thus a reflector: it reflects the energy from the surrounding environ-
ment and produces no intrinsic radiation of its own at the detection frequency
of the radiometer. This is the case for many metals, whose emissivity in the mi-
crowave and millimeter-wave region is 0.01–0.06. If the object’s emissivity is not
approximately zero or unity, the radiometric temperature is highly dependent
on the value of the emissivity and the ratio of temperatures of the object and the
environment.
Consider a human at 310 K surrounded by a constant temperature of 290 K,
which is a typical indoor environment temperature. At 30 GHz, the radiometric
temperature detected by a radiometer is

Thuman = 0.64 ´ 310 + 0.36 ´ 290 = 302.8 K (6.94)

whereas the radiometric temperature of a human in the absence of a constant tem-


perature surrounding environment is TR = 198.5 K. The temperature given in (6.94)
196 Microwave and Millimeter-Wave Remote Sensing for Security Applications

is indicative of that which would be measured when viewing a human in a closed,


windowless room at a constant temperature. In a contraband detection applica-
tion, the temperature differential between the human and an object on the body is
detected, where the object has a physical temperature close to that of the body. For
a metal object with e = 0.01, the measured radiometric temperature is

Tobject = 0.01 ´ 310 + 0.99 ´ 290 = 290.2 K (6.95)

The temperature differential between this and the human body is 302.8–290.2 =
12.6 K. While small, most radiometers have sensitivities below 1 K, thus this differ-
ence in temperature is easily distinguishable. The concealed object must therefore
have an emissivity very close to that of the human body in order to remain unde-
tected by the radiometer.
In an outdoor environment, much of the reflected energy originates from the
sky in the form of galactic and atmospheric radiation. The radiometric tempera-
ture of the sky at microwave and millimeter-wave frequencies is low: at 30 GHz, it
varies between approximately 15 and 30 K depending on humidity [28]. For a sky
temperature of 20 K, the measured radiometric temperature of a human is

Thuman = 0.64 ´ 310 + 0.36 ´ 20 = 205.6 K (6.96)

In contrast, the measured radiometric temperature of a concealed metal object


with e = 0.01 is

Tobject = 0.01 ´ 310 + 0.99 ´ 20 = 22.9 K (6.97)

yielding a temperature differential of 205.6–22.9 = 182.7 K. It is easier to detect the


temperature differential in environments with colder radiometric temperatures.

6.4  Radiometer Receivers

The function of a radiometer receiver is to provide a measure of the thermal elec-


tromagnetic radiation of an object or scene of interest. The application may be
object discrimination, such as the detection of a human within a radiometrically
diverse background, or it may be to distinguish temperature differences, such as
the detection of concealed objects. The physical temperatures of interest are of the
order of 290 K, which is room temperature. The detection of objects whose physical
temperature significantly exceeds 290 K is generally not of interest in security sens-
ing; solar reflections from metal surfaces can have high radiometric temperatures,
but these are typically considered a nuisance. Thus, the brightness of the sources
of interest is of the order of 10–18 W·m–2·Hz–1·str–1. Because metals have very low
emissivities, much lower brightness values are also of interest. For receivers with
bandwidths on the order of hundreds of megahertz, the resulting power collected by
the radiometer can be of the order of 10–10 W or –70 dBm and can be much lower.
The radiometer must thus have high gain, often greater than 100 dB total, in order
to detect the thermal radiation of interest. Additionally, the received power differ-
ences between many objects can be small, such as the difference between a human
6.4  Radiometer Receivers 197

and a brick wall. The radiometer thus has to have good sensitivity to resolve small
power differences.
Radiometers are, in essence, high-gain, high-sensitivity receivers, and their ar-
chitectures are similar in many ways to typical electronic receivers in communica-
tions or radar. However, the small signal power levels involved and the statistically
random nature of the signals of interest place additional demands on the receivers,
which must be able to accurately and repeatedly provide a measure of the low-
power signal. Radiometer receivers typically employ a superheterdyne configura-
tion; however, direct detection radiometers are also feasible and practical in many
cases. However, due to the high cost and low efficiency of millimeter-wave ampli-
fiers relative to RF amplifiers, it is often beneficial to implement a superheterdyne
receiver so that the needed gain can be achieved at the lower frequencies. A single
high-gain, low-noise amplifier can then be used at the millimeter-wave frequency,
which may have a gain of 30 dB or more, and at IF a chain of amplifiers can pro-
duce the rest of the needed gain, which may be up to 70 dB. The signals can then be
further amplified at baseband.
Because the power detected by the radiometer is proportional, through the
Rayleigh-Jeans approximation, to the radiometric temperature of the source, radi-
ometers are generally designed to provide an output voltage that is linearly propor-
tional to the input power, and therefore the source radiometric temperature. This
is accomplished by using a square-law detector, such as a detector diode, in a single
receiver, or by cross-multiplying the outputs of multiple receivers.
Radiometers are designed to detect broadband, thermal radiation that has
noise-like statistical properties. The measured temperature is the rms value of the
randomly fluctuating thermal noise and thus only provides an estimate of the ra-
diometric temperature. This estimate is improved by using a low-pass filter after
detection; an ideal low-pass filter with a rectangular bandwidth acts as an integra-
tor with integration time

1
τ= (6.98)
2 DfLPF

and thus the random fluctuations are averaged over a time inversely proportional
to the filter bandwidth.

6.4.1  Sensitivity
The precision with which a radiometer receiver can measure the radiometric tem-
perature is called the radiometric sensitivity, radiometric resolution, or simply sen-
sitivity. It is the minimum detectable temperature difference that the receiver can
resolve. The sensitivity of a radiometer can be qualitatively defined as the smallest
temperature difference that can be detected in the presence of the noise tempera-
ture generated by the receiver hardware. Measurements of radiometric temperature
manifest as noise power; thus, the sensitivity amounts to the smallest detectable
change in noise power, as shown in Figure 6.11. The system noise temperature Tsys
includes components of the antenna temperature TA, which is the signal of interest,
and the receiver noise temperature Trec. Because of the small power levels received
198 Microwave and Millimeter-Wave Remote Sensing for Security Applications

∆T

∆T

Figure 6.11  Definition of sensitivity in terms of the rms fluctuation of the system noise.

and small power differences that need to be detected, Tsys is often greater than TA by
an order of magnitude or more. The radiometric sensitivity will be derived for dif-
ferent radiometer configurations in the following sections; in general, it is given by
Tsys
DT = C (6.99)
D fRF τ

where DT is the sensitivity, DfRF­ is the bandwidth of the receiver, τ is the integra-
tion time provided by the low-pass filter, and C is a constant that depends on
the receiver configuration. Equation (6.99) gives the rms noise fluctuation Tsys,
reduced by the number of independent samples accumulated by the integrator
DfRF τ . Since the antenna temperature is typically much lower than the receiver
noise temperature,

Tsys = Trec + TA » Trec (6.100)

The antenna temperature TA represents the signal of interest, while the sensi-
tivity is the standard deviation of the noise. Thus, the signal-to-noise ratio for a
radiometer can be given by

TA TA
SNR = = DfRF τ (6.101)
DT CTrec

or, using (6.98),

TA DfRF
SNR = (6.102)
CTrec 2 DfLPF

The constant C in (6.99) and (6.102) depends on the receiver configuration, but
does not vary significantly and is on the order of unity.
The SNR is plotted in Figure 6.12 for various values of DfRF and τ with C = 1
and Trec = 10TA. It is apparent from this figure and (6.101) that a wide system
bandwidth and long integration time is beneficial for detecting low power signals.
This result is rather intuitive, since the signal power is proportional to the system
bandwidth through (6.69). Additionally, the rms noise fluctuations are reduced by
the square root of the product of the system bandwidth and the integration time.
Thus, increasing the system bandwidth has the effect of both increasing the signal
power and reducing the average noise power.
Typical system noise temperatures for millimeter-wave radiometers vary be-
tween a few hundred and a thousand Kelvin. Higher noise temperatures require
6.4  Radiometer Receivers 199

40
35
∆fRF = 1 GHz
30

SNR (dB)
25
10 MHz
20
15
100 kHz
10
5
0 -4
10 10-3 10-2 10-1 100
τ (s)
Figure 6.12  Signal-to-noise ratio as a function of the integration time and the system bandwidth
for Trec = 10TA and C = 1.

wider system bandwidths or longer integration times to reduce the sensitivity to


practical levels, and bandwidths of a few gigahertz can reasonably be achieved
with a millimeter-wave radiometer operating at K-band and above. The band-
width limitation arises in the correlator that, if done in analog hardware, must
have a multiplier that can support the predetection bandwidth. If the correla-
tion is accomplished digitally, the ADC must be able to sample the predetection
bandwidth; with current technology, analog correlators are generally required
for systems with predetection bandwidths of a few hundred megahertz and
higher.
The temperature differences required for intruder detection or detection of
nonmetallic concealed objects are on the order of a few Kelvin to less than one
Kelvin. Scanning radiometric imaging systems require short dwell times in order
to produce images quickly, thereby limiting the integration time, as discussed in
Section 6.6. Practical aspects limit the bandwidth, as described earlier, and thus
it is important that the radiometer be designed with a low system temperature.
As indicated by (6.99), the sensitivity is directly proportional to the system tem-
perature, and therefore lowering Tsys has the greatest effect toward lowering the
sensitivity. As discussed in Chapter 5, the system temperature is largely defined
by the first component of the receiver chain, and thus to achieve a low noise
radiometer, the first component should be a low-noise amplifier, preferably at-
tached directly to the antenna, that has the highest gain and lowest noise figure
possible.
Types of radiometer receivers can be divided into two general categories, de-
pending on whether they utilize a single antenna or two antennas. Dual-antenna
radiometer architectures can in principle be extended to include more antennas;
however, they are more practically implemented with two. Imaging systems are
composed of multiple receivers and are discussed in Chapter 8. The most common
single-antenna radiometers include the total power radiometer and Dicke radiome-
ter; the most widely used dual-antenna radiometer is the correlation interferometer.
Each will be discussed in the following sections.
200 Microwave and Millimeter-Wave Remote Sensing for Security Applications

6.4.2  Total Power Radiometer


The total power radiometer is so called because it measures the total noise power from
both the antenna and the receiver hardware, without providing an intrinsic method of
differentiating the two. That is, the system must be calibrated in order to characterize
the system noise power so that its contribution can be removed from the measurement
to recover the desired antenna temperature. A block diagram of a general total power
receiver is shown in Figure 6.13, and consists of an antenna, a receiver with frequency
response H(f ), a square-law detector, and a integrator or low-pass filter. The receiver
may or may not include a downconversion, but its frequency response provides the de-
sired gain G and bandwidth DfRF. The receiver circuitry prior to the detector is referred
to as the predetection circuitry and DfRF is the predetection bandwidth. The bandwidth
of the low-pass filter following the detector is the postdetection bandwidth. Because
the noise power is linearly proportional to the system temperature Tsys, and the output
voltage of the square-law detector is proportional to its input power, the output volt-
age of the square-law detector is linearly proportional to the system temperature.

6.4.2.1  Total Power Response


The spectral power at the antenna terminals of the total power radiometer is

PAf = kTA (6.103)

where the antenna temperature TA is proportional to the temperature of the source


through (6.63). The predetection circuitry introduces a noise power Trec, and has
gain G and bandwidth DfRF. For a rectangular passband, the power incident on the
detector diode is

PD = GkTsys DfRF = Gk(Trec + TA )DfRF (6.104)

The detector has a power sensitivity K (V·W–1), and thus the mean voltage on
the output of the detector is

Vd = KGkTsys DfRF (6.105)

The integrator reduces the rms fluctuations of Vd; its average output voltage is given
by

Vout = gLPFVd = gLPF KGkTsys DfRF = gLPF KGk(Trec + TA )DfRF (6.106)

where gLPF is the conversion gain (typically negative) of the filter.

∆fRF ∆fLPF
H(f) postdetection

Predetection Detection
Figure 6.13  Block diagram of a total power receiver.
6.4  Radiometer Receivers 201

The output voltage consists of components Vs and Vn due to the signal and the
system noise, respectively, and are given by

Vs = Gsys kTA D fRF (6.107)

Vn = Gsys kTrec DfRF (6.108)

where Gsys = gLPFKG is the total gain of the system. The noise voltage (6.108) is
constant, in theory, and can therefore be calibrated out, yielding

Vout = Gsys kTA D fRF (6.109)

In practice, as will be discussed later, the system noise power varies randomly due
to temperature and gain instabilities and thus cannot be exactly compensated for by a
constant voltage. Consequently, the calibrated output signal voltage (6.109) varies in
accordance with the receiver instabilities, reducing the accuracy of the measurement.
The antenna temperature is given in terms of the temperature distribution and
the antenna pattern by (6.63) which, when used with (6.59), give the voltage re-
sponse of the total power radiometer as

Ae
Vout = Gsys k DfRF
λ2
òò T(θ ,φ )A(θ , φ) dW (6.110)
source

For a point source located at θ = θ0, f = f0, the output voltage is

Ae
Vout = Gsys A(θ0 ,φ 0 )kT D fRF (6.111)
λ2

Including Ae ∕ λ2 and the antenna gain in the system gain yields

Vout = Gsys
¢ kT DfRF (6.112)

The total power voltage response is thus directly proportional to the temperature
of the source.
As an example of the total power response, Figure 6.14 shows experimental data
of a 27.4-GHz total power radiometer scanning across a stationary human. The band-
width of this sensor was 500 MHz and had a beamwidth of 3.5°. Figure 6.14(a) shows
the response from a human standing at close range, while Figure 6.14(b) shows the
response from a human in a cluttered outdoor environment. The responses from other
objects, including wooden crates and vegetation, are also prominent.

6.4.2.2  Sensitivity
If the predetection receiver has the frequency response H(f ), the power input to the
square-law detector is
¥
2
P = kTsys ò H(f ) df (6.113)

202 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Response
from human Responses from
background

Amplitude
Amplitude

Response
from human

Scan angle Scan angle


(a) (b)
Figure 6.14  (a) Output signal of a 27.4-GHz total power radiometer as the antenna beam scans
past a human at close range, and (b) output signal of a total power radiometer scanning across a
standing human in a cluttered outdoor environment. The response from the person is prominent;
however, other background objects also produce strong responses.

where, for simplicity, the gain is assumed to be unity. The voltage input to the detec-
tor is thus

V (t) = kTsys h(t) (6.114)

where h(t) is the impulse response of the predetection circuitry. The power is as-
sumed, for simplicity, and without loss of generality, to be formed across a 1 W
resistance. The output of the low-pass filter is given by
Vd (t) = V 2 (t) (6.115)

where × denotes the time averaging implemented by the filter. To determine the sen-
sitivity in terms of the SNR given by (6.102), the relative power levels of the signal
and noise components must be found, which is accomplished by analyzing their
power spectra. Because the random fluctuations of the noise signal are stationary
random processes, their Fourier transform does not exist. To calculate the power
spectra, the Wiener-Khinchin theorem is used [29], which states that the autocorre-
lation function R(τ) of a stationary random process is related to its power spectrum
P(f ) through a Fourier transform:
¥

ò R(τ )e
- j 2π f τ
P(f ) = dτ (6.116)

and
¥

ò P(f )e
j 2π f τ
R(τ ) = df (6.117)

6.4  Radiometer Receivers 203

The autocorrelation of the detector output (6.115) is given by

R(τ ) = Vd (t)Vd (t - τ ) = V 2 (t)V 2 (t - τ ) (6.118)



This expression can be expanded using [30]

z1z2 � z2n +1 = 0
(6.119)
z1z2 � z2n = å zi z j zk zl
All pairs

and thus

x1 x2 x3 x4 = x1x2 x3 x4 + x1x3 x2 x4 + x1x4 x2 x3 (6.120)

For x1 = x2 = xi and x3 = x4 = xj, which is the case in (6.118), expression (6.120) is

2 2 2
(xi x j )2 = xi xj + 2 xi x j (6.121)

and thus the autocorrelation can be given by
2
Rd (τ ) = V 2 (t) V 2 (t - τ ) + 2 éë V (t)V (t - τ ) ùû (6.122)

Since
V 2 (t) = V 2 (t - τ ) = R(0) (6.123)

the autocorrelation of the detector output can be given in terms of the autocorrela-
tion of the input voltage to the detector by

Rd (τ ) = R2 (0) + 2R2 (τ ) (6.124)

The power spectra can then be found using (6.116), which yields
¥
Pd (f ) = R2 (0)δ (f ) + 2 ò R2 (τ )e - j 2π f τ dτ (6.125)

Note that because τ = 0 in the first term on the right-hand side of (6.125), the result
is a constant, and thus its Fourier transform is simply the value multiplied by the
delta function δ(f ) at zero frequency.
Substituting (6.117) for one of the autocorrelation functions in the second term
on the right-hand side of (6.125) gives
¥
Pd (f ) = R2 (0)δ (f ) + 2 ò R(τ )R(τ )e - j 2π f τ dτ

(6.126)
¥ ¥
= R2 (0)δ (f ) + 2 ò ò P(f ')R(τ )e
- j 2π (f - f ')τ
dτ df '
-¥ -¥
204 Microwave and Millimeter-Wave Remote Sensing for Security Applications

where P(f ) here is the power spectra of the input signal to the detector, given by
(6.113), and again using (6.116) results in
¥
Pd (f ) = R2 (0)δ (f ) + 2 ò P(f )P(f ¢ - f )df (6.127)

where in the second term the substitution f ® f ¢, f ¢ ® f has been made for consis-
tency of the frequency variable between the two terms. The autocorrelation in the
first term can be replaced using (6.116) with τ = 0, which yields

2
é¥ ù ¥
Pd (f ) = ê ò P(f )df ú δ (f ) + 2 ò P(f )P(f ¢ - f )df (6.128)
êë -¥ úû -¥

Equation (6.128) represents the power spectra of the output of the detector in
terms of the power spectra of the input signals. The input and output power spectra
are shown in Figure 6.15. Substituting (6.113) into (6.128), the power spectra can
be given by
2
é¥ 2
ù ¥
2 2
Pd (f ) = k 2 2
Tsys ê ò H(f ) df ú δ (f ) + 2k2Tsys
2
ò H(f ) H(f ¢ - f ) df (6.129)
êë -¥ úû -¥

∆fRF

-fc 0 fc
(a)

k2Tsys2∆fRF2

k2Tsys2∆fRF∆fLPF
2∆fRF

-2fc -∆fRF ∆fRF 2fc


(b)

k2Tsys2∆fRF2

k2Tsys2∆fRF∆fLPF

2∆fLPF

(c)
Figure 6.15  Power spectra of (a) input signal, (b) detector output, and (c) integrator output.
6.4  Radiometer Receivers 205

The first term represents the dc output power while the second term, the convolu-
tion of the input power spectra with itself, represents the output noise fluctuations,
which are centered at dc and twice the center frequency of the input radiation, and
have magnitudes that linearly decrease out to twice the bandwidth. The low-pass
filter is used to reduce the output noise fluctuations; the first term of (6.129), a dc
term, is unaffected, while the second term is modified by the integrator. In order to
significantly reduce the output fluctuations, the postdetection bandwidth, that of
the low-pass filter, should be much less than the predetection bandwidth. In prac-
tice, the postdetection bandwidth DfLPF is typically at least an order of magnitude
less than DfRF. When this is the case, the output noise spectrum is significantly
narrowed to a small bandwidth around dc, such that the noise fluctuations are
approximately constant with frequency, and therefore f ¢ » 0 in (6.129). Assuming
rectangular passbands with unity gain, the resulting power is therefore

Pd = k2Tsys
2 2
DfRF + 2k2Tsys
2
DfRF DfLPF (6.130)

The first term of (6.130) is the power due to the system noise Tsys = Trec + TA, com-
posed of the receiver noise and the antenna noise, and represents the mean value of the
signal. The second term is the output fluctuation noise power, which represents the rms
noise fluctuations. The mean power due to the receiver noise power alone is thus

Prec = k2Trec
2 2
DfRF (6.131)

The antenna noise temperature represents the minimum temperature difference


to be characterized, thus TA = DT, and the signal power, due to the temperature
change, is

PA = k2 DT 2 DfRF
2
(6.132)

The rms noise fluctuations are given by the second term of (6.130):

Pn = k2Tsys
2
DfRF DfLPF (6.133)

The voltage SNR is therefore

Ps DT DfRF
SNR = = (6.134)
Pn Tsys 2 DfLPF

The minimum sensitivity is found when SNR = 1, which yields

2 DfLPF
DT = Tsys (6.135)
DfRF

Substituting the integration time (6.98) gives

Tsys
DT = (6.136)
DfRF τ

Note that this equation is (6.99) with C = 1.


206 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Consider a total power radiometer with a system noise temperature of 500 K,


a predetection bandwidth of 300 MHz, and a low-pass filter bandwidth of 1 kHz.
The resulting sensitivity is then DT = 1.3 K, which is temperature difference that
the radiometer can reliably resolve; temperature difference less than this will be
masked by the noise fluctuations. This low-pass filter bandwidth corresponds to an
integration time of 0.5 ms, which may be increased to further reduce the sensitivity
of the measurement. However, as discussed later, the application of the sensor may
limit the maximum integration time that can be used for a defined sensitivity. For
instance, a radiometer that scans an area would require its integration time to be
no greater than its dwell time to avoid a reduction in spatial resolution caused by
integration over multiple dwells.

6.4.3  Interferometric Correlation Radiometer


The basic interferometric correlation radiometer, often referred to simply as the
correlation radiometer, is shown in Figure 6.16 and consists of two antennas feed-
ing two receivers, the output voltages of which are correlated through a multiplica-
tion and integration. The integration is generally performed using a low-pass filter,
as in the total power radiometer architecture. Systems that use more than two
antenna elements, such as interferometric imagers, cross-correlate the outputs of
all the antenna pairs and produce multiple signal responses, each representative of
a two-element interferometer. The antenna separation, called the baseline, is typi-
cally multiple wavelengths of the detected radiation frequency. Increased antenna
separation leads to increased angular resolution, which for the two-element inter-
ferometer is that of a single antenna whose diameter is equal to the interferometer
baseline.
If the antennas of the correlation radiometer are viewing the same source, lo-
cated broadside to the radiometer, the voltage outputs of the two receivers can be
given by

V1 = s + n1 (6.137)

V2 = s + n2 (6.138)

where s is the voltage due to the antenna temperature (which is proportional to the
source temperature), and ni is the voltage due to the noise of receiver i = 1,2. After
multiplication and integration, the output voltage is

Vout = V1V2 = s2 + s1n2 + s2 n1 + n1n2 (6.139)



Because the signal voltage is uncorrelated with the noise components, and the noise
components are uncorrelated with each other, the last three terms of (6.139) will
tend to zero as the integration time increases. The output voltage is thus

‡ The term correlation radiometer may refer to either a two-element interferometric correlation radiometer
or a single-element radiometer that splits the antenna output to two receivers, the outputs of which are then
correlated. However, the term generally refers to the two-element interferometric architecture.
6.4  Radiometer Receivers 207

∆fRF,1

H1(f)
∆fLPF
postdetection
∆fRF,2

H2(f)

Predetection Detection
Figure 6.16  Correlation radiometer diagram.

Vout = s2 . (6.140)

Since s is a voltage signal proportional to the square root of the antenna tempera-
ture through
¥
vi = gi k(Trec + TA ) ò Hi (f )Hi* (f )df (6.141)

where gi is the receiver voltage gain and Hi(f) is the receiver frequency response, the
output voltage signal, with infinite integration time, is proportional to the signal
power:

Vout = GkTA DfRF (6.142)

Thus, the output voltage is proportional to the source radiometric tempera-


ture, just as is the case for the total power radiometer using a square-law detector
(6.105), however, without the dc contribution from the receiver noise power. Thus,
the correlation radiometer does not require a voltage offset calibration to generate
a response proportional only to the antenna temperature.
The correlation process thus provides an important benefit because, with suf-
ficient integration time, any signal or noise components that are uncorrelated be-
tween the receivers will be removed, as their time-average value will tend to zero.
One result of this is increased sensitivity, as described later. A second implication is
robustness to gain variations in the receiver hardware: any time-varying instabilities
in one receiver will be statistically independent from those generated in the other
receiver and thus will be removed in the correlation process.

6.4.3.1  Spatial Point Source Response


In contrast to single-antenna radiometers, whose spatial response is defined only by
the antenna pattern, the response of the correlation radiometer is dependent on the
antenna pattern, the antenna baseline, and the receiver bandwidth as well. If a point
source emitting monochromatic radiation is located at an angle θ off the broadside
208 Microwave and Millimeter-Wave Remote Sensing for Security Applications

θ RX1 V1(t)

D r

Dsinθ RX2 V2(t)

Figure 6.17  Geometry of a correlation radiometer viewing a distant point source.

direction of the antennas as shown in Figure 6.17, the normalized voltage signals at
the inputs to the multiplier are given by

V1(t) = a1 cos(2π ft) + n1(t) (6.143)

V2 (t) = a2 cos éë2π f (t - τg )ùû + n2 (t) (6.144)



where ai is the signal amplitude, and τg is the geometric time delay, which is the
time difference between the reception of a plane wavefront at the two antennas,
given by
D
τg = sin θ (6.145)
c
where D is the antenna baseline separation, and c is the speed of light.
The signal and noise components are statistically uncorrelated; thus, the nor-
malized response of the correlation radiometer at the output of the low-pass filter
is given by
r(τ g ) = V1(t)V2 (t) = a2 cos(2π ft)cos éë2π f (t - τg )ùû (6.146)

where the amplitudes have been assumed to be equal for simplicity. Using the rela-
tion cos A cos B = 12 cos(A - B) + 12 cos(A + B), the response can be given by

1 2 1
r(τ g ) = a cos éë2π f (2t - τ g )ùû + a1a2 cos(2π f τg ) (6.147)
2 2

The first term is a time-varying sinusoid at twice the radiation frequency, while
the second term is a dc component, assuming τg is time-invariant. The bandwidth
of the low-pass filter will be lower than the radiation frequency; therefore, the first
term will be filtered out, while the second term is unaffected. The response is thus

1 2 æ D ö
r(θ) = a cos ç 2π f sin θ÷ (6.148)
2 è c ø

where (6.145) has been substituted. Equation (6.148) is called the fringe function,
and it describes an oscillation induced on the output voltage as θ varies. The magni-

 If the geometric time delay is time-dependent, the response is no longer located at dc; Chapter 9 covers this
case specifically.
6.4  Radiometer Receivers 209

tude of the fringe function is called the fringe pattern and is a spatial pattern gener-
ated by the interferometer, akin to an antenna pattern. Figure 6.18 shows the fringe
pattern for varying values of Dl = Df/c. Due to the sinusoidal variation sin θ, the
frequency of the oscillation decreases as the angle diverges from broadside.
The fringe function (6.148) is the response of the correlation radiometer to a dif-
ferential frequency bandwidth df. If the source radiates broadband thermal radiation,
the response is integrated over the bandwidth of the receiver. If the radiated power is
constant over the rectangular passband Df centered on fc, the response is given by

DfRF
fc +
2
1 2 a2
r(τ g ) =
2
a ò cos(2π fτ g )df =
4πτ g
cos(2π fτ g )sin(πD fRF τg ) (6.149)
Df
fc - RF
2

where the relation sin(A − B) − sin(A + B) = 2cosA sinB has been used. Substituting
the definition of τg, and using sinc(x) = sin (x)/x,

1 2 æ D ö æ D ö
r(θ) = a DfRF cos ç 2π f sin θ÷ sinc çπ DfRF sin θ÷ (6.150)
4 è c ø è c ø

The sinc function in (6.150) is called the bandwidth pattern, or fringe wash-
ing function, and it spatially modulates the fringe function at a spatial frequency
inversely proportional to the bandwidth and antenna baseline, with decreasing am-
plitude as the angle diverges from broadside.
In addition to the bandwidth pattern, the response is spatially filtered by the
patterns of the two antennas. The signal voltages in each receiver are proportional
to the square root of the power, and thus if the voltage amplitudes are equal in the
two receivers, the square of the amplitudes is equal to the antenna power (6.62)

1λ Baseline 5λ Baseline
00 00

900 900 900 900


θ θ

10λ Baseline 15λ Baseline


00 00

900 900 900 900


θ θ
Figure 6.18  Fringe pattern of a correlation radiometer for various baselines. (© 2010 IEEE [13].)
210 Microwave and Millimeter-Wave Remote Sensing for Security Applications

multiplied by the gain of the receivers. If the antennas and the gain in each receiver
is identical, the correlation radiometer response is then

1 æ D ö æ D ö
r(θ) = Gsys kTA DfRF cos ç 2π f sin θ÷ sinc çπ DfRF sin θ÷ (6.151)
4 è c ø è c ø

The antenna temperature can be related to the source distribution and antenna
pattern through (6.63) and (6.59), which yields
Ae æ D ö æ D ö
r(θ) = Gsys kT DfRF A(θ )cos ç 2π f sin θ÷ sinc çπ DfRF sin θ÷ (6.152)
4λ 2 è c ø è c ø

Including Ae ∕ 4λ2 in the system gain yields

æ D ö æ D ö
r(θ) = Gsys
¢ kT DfRF A(θ )cos ç 2π f sin θ÷ sinc çπ DfRF sin θ÷ (6.153)
è c ø è c ø

The response is thus a spatial oscillation that is modulated by the antenna pattern and
the bandwidth pattern. Figure 6.19 shows the components comprising the broadband
response. The fringe pattern is spatially filtered by the product of the antenna pattern
and the bandwidth pattern, resulting in the final interferometer beam pattern.
An example of the correlation radiometer response is given in Figure 6.20,
which shows the measured responses of a correlation radiometer to a point source
and a stationary human. The point sources response of Figure 6.20(a) is angularly

00

Narrow-band
response
900 900
θ

00 x
Product of the
antenna and
bandwidth
patterns
900 900
θ

00 =

Broadband
response
900 900
θ
Figure 6.19  Broadband response of the correlation radiometer with D = 5λ and DfRFD ∕ c = ½.
6.4  Radiometer Receivers 211

Bandwidth pattern

Antenna
pattern

Amplitude
Measurement
Theory

Scan angle
(a)
Amplitude

Response
from human

Scan angle
(b)

Response
Responses from
from human
background
Amplitude

Scan angle
(c)
Figure 6.20  (a) Response of a 27.4-GHz correlation radiometer to a point source, where the an-
tenna pattern is significantly narrower than the bandwidth pattern. (© 2008 IEEE [9].) (b) Output
signal of a person walking through the fringe pattern of a correlation radiometer. (c) Output signal
of a 27.4-GHz correlation radiometer scanning across a standing human in an outdoor environment.
This scan was taken simultaneously with the total power scan in Figure 6.14.
212 Microwave and Millimeter-Wave Remote Sensing for Security Applications

limited by the antenna pattern, while the bandwidth pattern is significantly wider
in angle. The responses of Figures 6.20(b) and 6.20(c) are those of a 27.4-GHz cor-
relation radiometer scanning across a human at close and long ranges in outdoor
environments. These measurements were taken simultaneously with the total power
measurements of Figure 6.14.

6.4.3.2  Sensitivity
The derivation of the sensitivity of the correlation radiometer follows that of the total
power radiometer; however, in this case there are two signals (the outputs of the receiv-
ers) being multiplied rather than a single signal voltage being squared. As in the previous
derivation, the signal and noise powers are found by analyzing their power spectra.
The output of the correlator is given by

r = V1(t)V2 (t) (6.154)

where it is assumed that the source is located broadside to the antennas such that
θ = 0. The autocorrelation of (6.154) is

Rc (τ ) = V1(t)V2 (t)V1(t - τ )V2 (t - τ ) (6.155)

Using (6.120), this can be written

Rc (τ ) = V1(t)V2 (t) V1(t - τ )V2 (t - τ )


+ V1(t)V1(t - τ ) V2 (t)V2 (t - τ ) (6.156)
+ V1(t)V2 (t - τ ) V2 (t)V1(t - τ )

Each term in the angled brackets can then be written in terms of the autocorrelation
or cross-correlation it represents:
2
Rc (τ ) = R12 (0) + R11(τ )R22 (τ ) + R12 (τ )R21(τ ) (6.157)

where Rij(τ) is the cross-correlation between the voltage signals i and j. The first and
third terms, the cross-correlations of the two voltages, contain only components
due to the signal power since the components from the uncorrelated noise powers
are removed by the correlation process. The second term contains autocorrelations
that include both the signal and noise powers. The first term, for example, is

2 2
R12 (0) = V1(t)V2 (t)
2
= [ s1(t) + n1(t)][ s2 (t) + n2 (t)]
(6.158)
= { s1(t)s2 (t) + s1(t)n2 (t) + s2 (t)n1(t) + n1(t)n2 (t) }
2

2
= s1(t)s2 (t)

Since the noise signals are uncorrelated, their time-average tends to zero; the signal
components are correlated.
6.4  Radiometer Receivers 213

The Wiener-Khinchin relations (6.116) and (6.117) are used to convert (6.157)
to power spectra. The spectrum of the correlator output is
2
é¥ ù
Pc (f ) = ê ò P12 (f )df ú δ (f )
ëê -¥ ûú
¥

+ ò P11(f )P22 (f ¢ - f )df (6.159)


¥
+ ò P12 (f )P21(f ¢ - f )df

In general, the power can be given by

¥
2
Pii = kTsys,i ò Hi (f ) df (6.160)

ò Hi (f )H j (f )df
*
Pij = k TA,i TA, j (6.161)

Equation (6.160) is the power for the case of a single receiver (6.113), as is the
case for the total power radiometer. In (6.161) the receiver noise temperatures are
uncorrelated and removed by the correlation process, leaving only the antenna tem-
peratures. The resulting power spectrum is thus

2
é¥ ù
Pc (f ) = k TA1TA2 ê ò H1(f )H2* (f )df ú δ (f )
2

ëê -¥ ûú
¥

ò H1(f )H1 (f )H2 (f ¢ - f )H2 (f ¢ - f )df


* *
+k2Tsys1Tsys 2 (6.162)

ò H1(f )H2 (f )H2 (f ¢ - f )H1 (f ¢ - f )df


* *
+k2TA1TA2

The first term of (6.162) results from the autocorrelation of the signal com-
ponents, represented by the antenna temperatures, and it represents the signal of
interest, located at dc. The second and third terms result from the cross-correlations
of the receiver noise powers and the antenna noise powers. Their power spectra
represent noise and are the convolution of the receiver passbands, located at dc and
± fc as shown in Figure 6.15(b).
If the low-pass filter bandwidth DfLPF is significantly lower than the RF band-
width DfRF, f ¢ » 0, and the noise spectrum of the second two terms of (6.162) are
214 Microwave and Millimeter-Wave Remote Sensing for Security Applications

located approximately at dc. If the passbands are perfectly rectangular and have
equal frequency responses H1(f ) = H2(f ), the integration over H(f ) results in the RF
bandwidth DfRF, and the signal power can be written

Pc = k2TA1TA2 DfRF
2
+ k2Tsys1Tsys2 DfRF DfLPF + k2TA1TA2 DfRF DfLPF (6.163)

The first term on the right-hand side of (6.163) represents the signal power Ps,
while the second and third terms represent the noise power within the low-pass
filter bandwidth Pn. The signal-to-noise ratio is thus

Ps TA1TA2 DfRF
SNR = = (6.164)
Pn (Tsys1Tsys2 + TA1TA2 )DfLPF

If the antennas are viewing the same source, TA1 = TA2 = DT; if the receiver noise
temperature is much greater than the antenna temperature, the sensitivity can be
given by

DfLPF
DT = Tsys1Tsys 2 (6.165)
DfRF

Substituting the relation between an ideal integrator bandwidth and integration


time (6.98),

Tsys1Tsys 2
DT = (6.166)
2 DfRF τ

If the receivers have the same noise temperatures, the sensitivity is given by

Tsys
DT = (6.167)
2 DfRF τ

The sensitivity of the correlation radiometer is thus better than that of the total
power radiometer (6.136) by a factor of 1 2.
Consider a millimeter-wave correlation radiometer where the two receivers have
the same system noise temperature of 500 K and system passbands of 500 MHz.
The correlator has a low-pass filter cutoff frequency of 500 Hz, corresponding to
an integration time of 1 ms. The resulting sensitivity is

500
DT = = 0.5 K
2 ´ 500 ´ 106 ´ 1 ´ 10-3

A total power radiometer with the same noise temperature, passband, and inte-
gration time yields a sensitivity of

500
DT = = 0.707 K
500 ´ 106 ´ 1 ´ 10-3
6.5  Practical Considerations 215

6.5  Practical Considerations

6.5.1  Receiver Instabilities


The previous sections assumed that the gain of the receivers were constant in time.
In general, this is not true for real receivers. Amplifier gain varies over long time
periods, producing low-frequency variations that may be accounted for with pe-
riodic calibration, and over short time periods, for which periodic calibration is
not possible. Long-term variations are generally temperature based, and the gain-
temperature coefficient of transistor amplifiers is on the order of 0.01 dB·°C–1 per
stage [31]. Gain-temperature variations can be primarily traced to the variation
in bias resistance in the amplifiers, which results in variations in the source cur-
rent, thereby affecting the gain. Short-term variations can be due to power supply
instabilities.
As shown earlier, the output voltage of the total power radiometer is propor-
tional to GsysTsys, and thus variations in the system gain will produce variations
in the output voltage, introducing measurement uncertainty. The rms value of the
temperature uncertainty is

DG
DTG = Tsys (6.168)
G

where DG represents the differential change in the system gain. Because the gain
variations are caused by processes that are separate from those that introduce the
system noise fluctuations, the variation in temperature due to the gain variations
(6.168) and that due to the system noise (6.136) are statistically independent. The
total measured temperature is therefore

2
1 æ DG ö
DT = Tsys +ç ÷ (6.169)
DfRF τ è G ø

Note that while gain fluctuations affect total power radiometers, because the
gain variations of one receiver are independent of the gain of another, they are un-
correlated and the correlation radiometer is unaffected by them.
Variations in the receiver gain can have a significant effect on the sensitivity of
the total power radiometer. For example, consider a total power radiometer with
Tsys = 500 K, DfRF = 100 MHz, and τ = 1×10–3. If DG/G = 0, no gain variations
are present, and the sensitivity is DT = 1.58 K. If the gain varies by 1%, or DG/G =
0.01, the sensitivity is DT = 5.24 K, an increase over the ideal sensitivity of more
than three times. Thus, even small gain variations can cause significant sensitivity
reduction in total power radiometers.

6.5.2  Dicke Radiometer


A method of reducing the effect of gain variations on the sensitivity of a total
power radiometer was introduced by Dicke in 1946 [32], and is still in use today.
The method consists of repeatedly switching the input of the radiometer between
216 Microwave and Millimeter-Wave Remote Sensing for Security Applications

the antenna terminals and a reference load at a known temperature. The switching
frequency must be fast enough that the gain is constant over the switching period.
The receiver produces the output voltage signal

Vd , A = kG(Trec + TA )DfRF , 0 £ t £ τs 2
(6.170)
Vd ,R = kG(Trec + TR )DfRF , τ s 2 £ t £ τs

where TR is the reference temperature and τs is the switching period. The switch
driver also modulates the voltage input to the detector diode between two unity
gain amplifiers of opposite polarity, as illustrated in Figure 6.21. The switching fre-
quency must be higher than the low-pass filter cutoff frequency, so that the output
is the time-average of the two voltages. After the unity gain amplifiers, the output
of the filter is
Vout = Vd , A - Vd ,R

= kG [(Trec + TA ) - (Trec + TR )] DfRF (6.171)

= kG(TA - TR )DfRF

The temperature uncertainty due to gain variations is

DG
DTG = (TA - TR ) (6.172)
G

If the reference temperature is set such that TR = TA, the radiometer is said to be
balanced, and the measured temperature uncertainty due to gain variations is zero.
The sensitivity of the balanced Dicke radiometer is
Tsys
DT = 2 (6.173)
DfRF τ

which is worse than that of the total power radiometer by a factor of 2. This is
because the receiver is only connected to the antenna for half of the measurement
time and connected to the reference load for the rest of the time. In the previous
example, the ideal total power sensitivity was 1.58 K, whereas a 1% gain variation

TA

+1
H(f)
-1

TR

Figure 6.21  Dicke radiometer.


6.5  Practical Considerations 217

degraded the sensitivity to 5.24 K. In that example, a Dicke radiometer would pro-
duce a sensitivity of 2.16 K. In a contraband detection application, the temperature
differences between a concealed object and the human body may be only a few K,
thus a reduction in sensitivity by a factor of 2 could significantly impact the detec-
tion capabilities of such a sensor. Note that setting Tr = Ta ideally results in a zero
voltage output as given by (6.171). However, this only zeros out temperatures that
are the same temperature as the reference. Changes in the observed temperature,
due to a human or other object, will result in non-zero voltage responses.

6.5.3  Radiometer Calibration


The output voltage of a radiometer can be given in general by

Vout = α (TA + β ) (6.174)

where α includes the gain, bandwidth, and Boltzmann’s constant, and β represents
a temperature offset, which is either the system noise temperature in a total power
radiometer, or the reference temperature in a Dicke radiometer. Because micro-
wave and millimeter-wave radiometers used for terrestrial security applications
operate in the Rayleigh-Jeans region of the blackbody curve, the output voltage is
a linear function of the antenna temperature. The coefficient α thus gives the slope
of the curve of the output voltage versus the antenna temperature, as illustrated in
Figure 6.22.
Because the output is linear, the variables can be calculated by measuring the
output due to two different temperature sources. This provides two points on the
voltage curve, which can then be used to extrapolate the variables. The calibration
sources are typically denoted by Tc for the colder of the two temperatures, and Th for
the hotter temperature. The output voltages due to the calibration sources are thus

Vc = α (Tc + β ) (6.175)

Vh = α (Th + β ) (6.176)

Vh
Output voltage

Vc

Tc Th
Input temperature
Figure 6.22  Radiometer output voltage versus antenna temperature.
218 Microwave and Millimeter-Wave Remote Sensing for Security Applications

The variables are thus found by


Vh - Vc
α= (6.177)
Th - Tc
VcTh - VhTc
β= (6.178)
Vh - Vc

The ratio of the output voltages is referred to as the Y-factor, given by


Vh Th + β
Y= = (6.179)
Vc Tc + β

The intercept point can thus be given by


Th - YTc
β= (6.180)
Y -1

Recall that the variable β represents the temperature offset, which is the system
noise temperature for a total power receiver. Thus,

Th - YTc
Tsys = (6.181)
Y -1

The Y-factor measurement technique is a standard method of measuring the noise


temperature of a system.
Calibration sources may be passive matched loads held at a constant temper-
ature, or, more commonly in microwave and millimeter-wave systems, an active
noise diode such as an avalanche diode. Calibration can also be achieved by placing
an electromagnetic absorber held at a constant temperature in the antenna field of
view. As discussed earlier, such materials closely approximate blackbodies; thus, the
radiometric temperature of an absorber is approximately its physical temperature.
For a passive source, the noise temperature is simply its physical temperature.
For accurate calibration, the temperature of the source must therefore be closely
regulated. Active noise diodes are characterized by the excess noise ratio, which
specifies the amount of additional noise power supplied when the active diode is
powered on Pon divided by the noise power PN due to its physical temperature:

Pon - Poff k(Ton - Toff )DfRF T


ENR = = = on - 1 (6.182)
Poff kToff DfRF Toff

The temperature Toff is the noise power supplied when the diode is not pow-
ered, or that due to its physical temperature Tp. When used in the Y-factor method,
Ton corresponds to Th, while Toff corresponds to Tc.

6.6  Scanning Radiometer Systems

Radiometers used in security sensing are operated in either a stationary “staring”


mode where the sensor statically views a specific area or in a scanning mode where
6.6  Scanning-Beam Radiometer Systems 219

the antenna beam is moved in a pattern over a designated field of view. While staring-
mode sensors may be single-receiver systems, they are generally implemented as
imagers with multiple parallel receivers. The goal of staring-mode sensors may be
the detection of a human moving into a field of view, or detecting the temperature
difference between a concealed object and a person’s body.
Scanning-mode radiometers focus a narrow beam over a field of view in a re-
peated pattern to generate the desired coverage. The sensor may scan in one dimen-
sion (typically azimuth) or in both azimuth and elevation directions to generate a
one- or two-dimensional radiometric representation of the scene. Scanning-mode
sensors can form images through a repeated raster scan of an area or can be used
for object detection as rotating sensors continually scanning a wide field of view.
Radiometers implemented on mobile platforms for intruder detection or perimeter
monitoring of a large area are operated in scanning mode, even if the sensor itself is
not rotating, due to the platform motion.

6.6.1  Spatial Resolution


Spatial resolution refers to the minimum physical area that a sensor system can
resolve. The one-dimensional spatial resolution of a single-antenna system, Dxs, is
found in terms of the antenna beamwidth θBW by

Dx s
tan θBW / 2 = (6.183)
2R

where R is the distance between the antenna and the location of interest, as shown
in Figure 6.23. For antenna beams that are narrow, approximately 15° or less,

tan θBW / 2 » θBW / 2 (6.184)

and

Dxs = RθBW (6.185)

For a system utilizing two antennas, such as a correlation radiometer, the spa-
tial resolution depends on the range at which the antenna beamwidths converge, as
seen in Figure 6.24. Assuming identical antennas pointing in the same direction, the
convergence range can be found by

D
Rc = cot(θBW 2) (6.186)
2

θBW ∆x

R
Figure 6.23  Spatial resolution at a distance R.
220 Microwave and Millimeter-Wave Remote Sensing for Security Applications

θBW
rc
θBW

D
Figure 6.24  Convergence range of two identical antennas.

Because the antenna beams do not converge prior to Rc, the spatial resolution of a
dual-antenna system is reduced compared to that of a single antenna system and is
given by

Dxd = (R - Rc )θ BW (6.187)

Note that if R is significantly greater than Rc, the spatial resolution of the dual-an-
tenna system is equal to that of the single-antenna system.
When correlation radiometers are operated at ranges close to the objects of in-
terest, the antenna beams may not always be parallel. Angling the antennas inward
by an angle θa reduces the convergence range, which is then given by

D æ θ ö
Rc = cot ç θ a - BW ÷ (6.188)
2 è 2 ø

Equation (6.186) is simply (6.188) with θa = 0. If the angle is greater than half of
the beamwidth, the beams diverge at a range

D æ θ ö
Rd = cot ç θ a + BW ÷ (6.189)
2 è 2 ø

The divergence range (6.189) defines the range beyond which sources will not be de-
tected. This effectively range-gates the sensor, which can be necessary in some appli-
cations such as monitoring of traffic in front of a cluttered, dynamic background.
When angling the antennas inward, there are three cases to consider, shown in
Figure 6.25, each of which can yield a different value of the spatial resolution [9].

1. θ
 a=0
In this case [Figure 6.25(a)], the antenna beams are parallel, and Rd = ∞. The
spatial resolution is given by

ì(R - Rc )θ BW , Rc £ R

Dxd = í (6.190)
î 0, R < Rc
6.6  Scanning-Beam Radiometer Systems 221

θa
R1
Rc
Rd
θBW θBW Rc θa R1
θBW Rc

D D D
Case 1 Case 2 Case 3
Figure 6.25  Geometries of the three cases encountered when angling the antennas of a dual-
antenna system. (© 2008 IEEE [9].)

2. 0 £ θa £ θBW/2
In this case [Figure 6.25(b)] the beams converge at Rc; however, Rd = ∞. The
beams do not diverge. The spatial resolution is given by

ì(R - Rc )(2θ a + θ BW ), Rc £ R < Rm


ï 2b sin(θ + θ R = Rm
ï a BW 2),
Dxd = í (6.191)
ïD - 2R(θ a - θ BW 2), R > Rm
ïî 0, R < Rc

where

Rm = Rc + b(θ a + θ BW 2) (6.192)

sin(θBW ) D2
b= Rc2 + (6.193)
sin(2θ a ) 4

3. θ  BW/2 < θa
In this case [Figure 6.25(c)] the beams diverge. The spatial resolution is given by

ì(R - Rc )(2θ a + θ BW ), Rc £ R < Rm


ï R (2θ - θ ), R = Rm
ï m a BW
Dxd = í (6.194)
ï(Rd - R)(2θ a - θ BW ), Rm < R < Rd
ïî 0, R < Rc or Rd < R

222 Microwave and Millimeter-Wave Remote Sensing for Security Applications

For a given antenna beamwidth, the spatial resolution of a correlation radiom-


eter is dependent on the range R, the antenna angle θa, and the baseline separa-
tion D. Figure 6.26 gives a plot of the spatial resolution as a function of both
the range and antenna angle for an antenna beamwidth θBW = 3.8° and a static
baseline D = 40λ. When the antenna angle is greater than θBW/2, the range-
gating effect can be observed because the spatial resolution is zero beyond a
certain range.

6.6.2  Dwell Time


The dwell time τd of a scanning sensor is the length of time that a single point is
present within the antenna beamwidth. The dwell time is thus defined by

θBW
τd = (6.195)
ω

where ω is the scan rate of the sensor in rad·s–1. Using (6.185), the dwell time can
be given in terms of the spatial resolution by

Dx
τd = (6.196)

where Dx refers to either the single- or double-antenna spatial resolution. Figure


6.27 shows the dwell time of a correlation radiometer with angled antennas as a
function of range and angle for θBW = 3.8°, D = 40λ and ω = 1 rad·s–1. The spatial
resolution can conversely be given in terms of the dwell time by

Dx = τ d Rω (6.197)

4 1
0.9
0.8
3
0.7
0.6
θaW (deg)

∆xd (m)

2 0.5
0.4
0.3
1
0.2
0.1
0 0
0 2 4 6 8 10 12 14 16
R (m)
Figure 6.26  Spatial resolution of a correlation radiometer as a function of antenna angle and range,
for θBW = 3.8° and D = 40λ. (© 2008 IEEE [9].)
6.6  Scanning-Beam Radiometer Systems 223

4 10
9
8
3
7
6

θaW (deg)

τd (ms)
2 5
4
3
1
2
1
0 0
0 2 4 6 8 10 12 14 16
R (m)
Figure 6.27  Dwell time of a correlation radiometer as a function of antenna angle and range.
(© 2008 IEEE [9].)

6.6.3  Measurement Uncertainty


There is, in general, a tradeoff between minimizing the radiometric temperature
resolution and the scan rate for scanning radiometer systems. The more quickly the
radiometer scans, the less time the antenna beam dwells on any one point. The inte-
gration time is increased to reduce the radiometric temperature resolution; however,
if it is longer in duration than the antenna beam spends on one point, the result is
that the response from that point is spread over a wider spatial angle, reducing the
effective spatial resolution. Radiometers that function in a stare mode where the
system does not use scanning can, in some instances, use much longer integration
times to reduce the temperature resolution. The limiting time frame for stare-mode
systems is usually the required update rate fs, which limits the integration time to
τ = 1/fs.
On the other hand, if the integration time is less than the dwell time, the radi-
ometer is not achieving the best possible sensitivity, as the signal from a given point
is within the beam for longer time than the signal is integrated, resulting in lower
radiometric sensitivity. In this context, the integration time of the correlator defines
an effective radiometric spatial resolution, given by

D xeff = τ Rω (6.198)

which derives from (6.197) with the integration time in place of the dwell time.

6.6.3.1  One-Dimensional Scanning


For a radiometer system scanning in one dimension, the dwell time is the duration
that a given point remains within the beamwidth of the antenna. The system need
not scan a continuous 360˚ and may scan back and forth over a sector. A system
of this type may have antenna beams that are narrow in azimuth and wide in
224 Microwave and Millimeter-Wave Remote Sensing for Security Applications

φHPBW θHPBW
ω

Figure 6.28  One-dimensional scanning radiometer with narrow azimuth and wide elevation
beamwidths.

elevation, which could be used to scan an area for intruders, as illustrated in Figure
6.28, where fBW is the azimuth beamwidth and qBW is the elevation beamwidth.
The tall beams provide instantaneous elevation coverage; although the response is
an average of the temperatures within the beam, a human present within the beam
would still produce a detectable temperature differential, particularly if most of the
beam is viewing the radiometrically cold sky. The tall beams allow for faster azi-
muth scanning without the need for imaging in the vertical direction.
Consider a radiometer scanning across a sharp temperature transition. If τ = 2τd,
the radiometer will take twice as long to register the temperature transition as it
will take the transition to pass through the antenna beam; the effective radiometric
resolution (6.198) is twice that of the spatial resolution (6.197). The measured
transition will thus be spread out over a wider spatial extent, resulting in a reduc-
tion of the effective spatial resolution of the radiometer. The measured temperature
will not equal the source temperature until after the transition has left the antenna
beam, since the signal will still be in process in the integrator.
If τ = τd/2, the radiometer will register the temperature transition while it is still
within the antenna beam, although the measured temperature will be an average
of the source temperature before and after the transition. Although this improves
the effective spatial resolution relative to a long integration time, the radiometric
sensitivity is reduced, and small temperature changes may go undetected.
It is generally desirable to set the integration and dwell times equal,

τ = τd (6.199)

in which case the spatial resolution and effective spatial resolution are equal. Sub-
stituting the integration time with the dwell time (6.196) in the expression for the
radiometric sensitivity (6.97) yields


DT Dx = CTsys (6.200)
DfRF

The right-hand side of (6.200) is a constant. This is the one-dimensional radio-


metric uncertainty equation, which states that neither the temperature resolution
nor the spatial resolution can be decreased without a corresponding increase in the
other.
6.6  Scanning-Beam Radiometer Systems 225

6.6.3.2  Two-Dimensional Scanning


Consider the system described in the one-dimensional scanning example, but with
a circular symmetric antenna beam such that

φBW = θBW (6.201)

The radiometer scans in azimuth with angular frequency wf and scans over a
limited field of view in elevation θFOV with angular frequency wf , as illustrated in
Figure 6.29.
The elevation scan time is thus

θFOV
t s,θ = (6.202)
ωθ

and the dwell time is

θBW t s,θ θBW


τ d ,θ = = (6.203)
ωθ θ FOV

Assuming that there is no latency between finishing one elevation scan line and
beginning the next, and that one scan over θFOV is completed as the system rotates
in azimuth over one beamwidth fBW, the azimuth scan time over one beamwidth is
equal to the elevation scan time. Thus,

Dx
t s,θ = (6.204)
Rωφ

The dwell time is thus


DxθBW
τd = (6.205)
RωφθFOV

Setting the integration time of the radiometer to be equal to the dwell time (6.205)
yields

R2ωφθ FOV
DT Dx = CTsys (6.206)
DfRF

The term on the right-hand-side of (6.206) is a constant.

ωφ
ωθ

Figure 6.29  Two-dimensional scanning system.


226 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Equations (6.206) is the two-dimensional radiometric uncertainty equation,


and similar to the one-dimensional radiometric uncertainty equation (6.200), it
states that the product of the temperature uncertainty and the spatial uncertainty is
a constant. Thus, the uncertainty of one measurement cannot be improved without
detrimentally affecting the uncertainty of another measurement.

References

[1] Kraus, J. D., Radio Astronomy, New York: McGraw-Hill, 1966.


[2] Thompson, A. R., J. M. Moran, and G. W. Swenson, Interferometry and Synthesis in Radio
Astronomy, New York: John Wiley & Sons, 2001.
[3] Christiansen, W. N., and J. A. Högbom, Radiotelescopes, Cambridge: Cambridge Univer-
sity Press, 1969.
[4] Rohlfs, K., Tools of Radio Astronomy, Berlin: Springer-Verlag, 1990.
[5] Ulaby, F. T., R. K. Moore, and A. K. Fung, Microwave Remote Sensing, Vol. I: Microwave
Remote Sensing Fundamentals and Radiometry, Reading, MA: Addison-Wesley, 1981.
[6] Ruf, C. S., C. T. Swift, A. B. Tanner, and D. M. Le Vine, “Interferometric Synthetic Aper-
ture Microwave Radiometry for the Remote Sensing of the Earth,” Geoscience and Remote
Sensing, IEEE Transactions on, Vol. 26, 1988, pp. 597–611.
[7] Nanzer, J. A., and R. L. Rogers, “Human Presence Detection Using Millimeter-Wave
Radiometry,” Microwave Theory and Techniques, IEEE Transactions on, Vol. 55, 2007,
pp. 2727–2733.
[8] Nanzer, J. A., and R. L. Rogers, “A Ka-Band Correlation Radiometer for Human Presence
Detection from a Moving Platform,” in Microwave Symposium, 2007, IEEE/MTT-S Inter-
national, 2007, pp. 385–388.
[9] Nanzer, J. A., and R. L. Rogers, “Applying Millimeter-Wave Correlation Radiometry to the
Detection of Self-Luminous Objects at Close Range,” Microwave Theory and Techniques,
IEEE Transactions on, Vol. 56, 2008, pp. 2054–2061.
[10] Nanzer, J. A., E. Popova, and R. L. Rogers, “Applying Correlation Radiometry to Human
Presence Detection in Multiple Outdoor Environments,” in 2010 CNC/USNC/URSI Na-
tional Radio Science Meeting, 2010.
[11] Nanzer, J. A., E. Popova, and R. L. Rogers, “Analysis of the Detection Modes of a Human
Presence Detection Millimeter-Wave Radiometer,” in Antennas and Propagation Society
International Symposium (APSURSI), 2010 IEEE, 2010, pp. 1–4.
[12] Nanzer, J. A., and R. L. Rogers, “Frequency Estimation of Human Presence Detection
Signals From a Scanning-Beam Millimeter-Wave Correlation Radiometer,” Geoscience and
Remote Sensing Letters, IEEE, Vol. 8, 2011, pp. 78–82.
[13] Nanzer, J. A., “Interferometric Detection of the Angular Velocity of Moving Objects,” in
Microwave Symposium, 2010. IEEE/MTT-S International, 2010, pp. 1628–1631.
[14] Nanzer, J. A., “Millimeter-Wave Interferometric Angular Velocity Detection,” Microwave
Theory and Techniques, IEEE Transactions on, Vol. 58, 2010, pp. 4128–4136.
[15] Johnson, J. T., M. A. Demir, and N. Majurec, “Through-Wall Sensing with Multifrequency
Microwave Radiometry: A Proof-of-Concept Demonstration,” Geoscience and Remote
Sensing, IEEE Transactions on, Vol. 47, 2009, pp. 1281–1288.
[16] Tasselli, G., F. Alimenti, S. Bonafoni, P. Basili, and L. Roselli, “Fire Detection by Micro-
wave Radiometric Sensors: Modeling a Scenario in the Presence of Obstacles,” Geoscience
and Remote Sensing, IEEE Transactions on, Vol. 48, 2010, pp. 314–324.
[17] Frost, R., R. Appleby, S. Price, F. Nivelle, M. Allin, et al., “The Detection of Mines Using
RF/Millimetric Radiometry,” in The Detection of Abandoned Land Mines: A Humanitarian
6.6  Scanning-Beam Radiometer Systems 227

Imperative Seeking a Technical Solution, EUREL International Conference on (Conf. Publ.


No. 431), 1996, pp. 92–96.
[18] Suess, H., M. Peichl, M. Zeiler, and S. Dill, “Investigations on Anti-Personnel Mine De-
tection Using Microwave Radiometers,” in Geoscience and Remote Sensing Symposium,
2001. IGARSS ‘01. IEEE 2001 International, Vol. 7, 2001, pp. 3178–3181.
[19] Tan, R. J., and R. Bender, “Near-Field Synthetic Aperture Interferometric Microwave Ra-
diometry for Remote Sensing of Mines,” in Microwave Symposium Digest, 1999 IEEE
MTT-S International, Vol. 4, 1999, pp. 1611–1614.
[20] Wiggins, D., H. Kim, Y. Cheon, and J. T. Johnson, “Sub-Surface Object Sensing with Multi-
Frequency Microwave Radiometry,” in Geoscience and Remote Sensing Symposium, 2002.
IGARSS ‘02. 2002 IEEE International, Vol. 1, 2002, pp. 331–333.
[21] Widger, W. K., and M. P. Woodall, “Integration of the Planck Blackbody Radiation Func-
tion,” Bulletin of the American Meteorological Society, Vol. 57, 1976, pp. 1217–1219.
[22] Mitchell, D., C. H. Wyndham, T. Hodgson, and F. R. N. Nabarro, “Measurement of the
Total Normal Emissivity of Skin Without The Need For Measuring Skin Temperature,”
Physics in Medicine and Biology, Vol. 12, 1967, p. 359.
[23] Govan, C. D., “A Wideband Frequency-Tunable Dicke Radiometer for Microwave Ra-
diometric Measurements,” M.S.E. Master’s thesis, Electrical and Computer Engineering,
University of Texas at Austin, 1994.
[24] Gabriel, C., S. Gabriel, and E. Corthout, “The Dielectric Properties of Biological Tissues:
I. Literature Survey,” Physics in Medicine and Biology, Vol. 41, 1996, p. 2231.
[25] Gabriel, S., R. W. Lau, and C. Gabriel, “The Dielectric Properties of Biological Tissues: II.
Measurements in the Frequency Range 10 Hz to 20 GHz,” Physics in Medicine and Biol-
ogy, Vol. 41, 1996, p. 2251.
[26] Gabriel, S., R. W. Lau, and C. Gabriel, “The dielectric properties of biological tissues: III.
Parametric models for the dielectric spectrum of tissues,” Physics in Medicine and Biology,
Vol. 41, 1996, p. 2271.
[27] Yujiri, L., M. Shoucri, and P. Moffa, “Passive Millimeter Wave Imaging,” Microwave Mag-
azine, IEEE, Vol. 4, 2003, pp. 39–50.
[28] Cruz-Pol, S. L., C. S. Ruf, and S. J. Keihm, “Improved 20–32 GHz Atmospheric Absorption
Model,” Radio Science, Vol. 33, 1998, pp. 1319–1333.
[29] Ziemer, R. E., and W. H. Tranter, Principles of Communications, 5th ed., New York: John
Wiley & Sons, 2002.
[30] Middleton, D., An Introduction to Statistical Communciation Theory, New York:
McGraw-Hill, 1960.
[31] Bahl, I. J., Fundamentals of RF and Microwave Transistor Amplifiers, Hoboken, NJ: Wiley-
Interscience, 2009.
[32] Dicke, R. H., “The Measurement of Thermal Radiation at Microwave Frequencies,” Re-
view of Scientific Instruments, Vol. 17, 1946, pp. 268–275.
Chapter 7

Radar

Radar systems, and active sensors in general, transmit a signal that illuminates the
object of interest, is reflected off the object, and is collected by a receiver. An active
sensor is thus comprised of both a transmitter and a receiver, which are not neces-
sarily collocated. The signal reflected off the object is altered by the characteristics
of the object, and thus certain properties of the object can be discerned by charac-
terizing the difference between the transmitted and received signal; for radar, the
basic properties that can be measured are range, radial velocity, reflectivity, and
angle. These are the only quantities that radar can measure, although a great deal
of information about the object can be derived from measuring these simple quan-
tities. Each of these quantities is relevant to security sensors of various types. For
example, range and reflectivity are important in imaging applications where mea-
surements are taken at multiple pixels to form an image, and velocity and angle are
important in intruder detection systems. Radar systems have been studied for de-
cades, and a number of excellent references exist covering fundamental and applied
aspects of radar in applications such as surface maritime and ground-based radar
and airborne radar [1–6]; the goal of this chapter is to cover the aspects relevant
to microwave and millimeter-wave security radar systems and the basic waveforms
and system configurations required for security applications.
Microwave and millimeter-wave radar systems are increasingly being applied
to security sensing for a number of applications, some of the most prominent of
which are human presence detection and classification, and contraband detection.
Human presence detection relies on detecting the Doppler frequency shift generated
when the transmitted radar signal is reflected off an object or person moving with
a nonzero radial velocity relative to the radar. Radial motion from a human can be
due either to the gross movements of a person walking or running, or to fine move-
ments such as the motions of the torso generated through respiration. The motions
generated in human movement from different parts of the body such as arm and leg
swing generate Doppler frequency sidebands that can be analyzed for the purpose
of classification of objects for discriminating between humans and nonhumans, and
classifying human activity. Contraband detection systems measure the range profile
of a person or object, which is obtained by measuring the time delay of the radar
signal reflected off the object or measuring the difference in reflectivity from one
image pixel to the next to detect hidden objects by their material properties.
Systems operating at microwave and millimeter-wave frequencies have benefits
deriving from both the high carrier frequency and the wider bandwidths that can
be supported at higher frequencies. The Doppler frequency shift is proportional to
the carrier frequency of the radar signal; thus, higher frequencies produce larger
Doppler frequency shifts, making the detection of very small movements possible.
Micro­wave, and particularly millimeter-wave, radar systems are thus sensitive to

229
230 Microwave and Millimeter-Wave Remote Sensing for Security Applications

very small movements, or small shifts in range of the object being viewed. This is the
basis underlying recent research developments in the measurement of biometric hu-
man signatures such as respiration and heart rate where the change in range due to
movements of the body is small, but can be resolved by using millimeter-wave car-
rier frequencies. As will be shown, the resolution with which the range of an object
can be measured is inversely proportional to the bandwidth of a radar signal, and
because systems operating at high frequencies can also support wider bandwidths
than lower frequency systems, finer range resolution can be achieved. Other benefits
of radar systems operating at higher frequencies are the near-lossless transmission
of radiation through clothing materials and the opacity of human skin. Active sys-
tems can therefore be used to detect objects hidden, for instance beneath a person’s
clothing or in a bag. These traits have led to a significant amount of research into
high-frequency radar imaging systems for the detection of contraband.
Microwave and millimeter-wave systems also have benefits deriving from the
physical properties of the hardware systems used at higher frequencies. Antennas
with aperture dimensions of many multiple wavelengths can generate small beam-
widths for excellent spatial resolution while also remaining physically small due to
the short wavelengths. Increased spatial resolution can improve the detection capa-
bilities of a human presence detection radar system and can provide fine resolution
images in imaging radar systems. The transmitter and receiver components can also
be made physically small, which benefits the implementation of mobile sensors.

7.1  Radar Fundamentals

In this section, a simple radar system is considered in order to develop the funda-
mental aspects of radar. The physical foundations of each measurement are derived
in simple terms using a general radar system as an example; more elaborate tech-
niques are discussed in the following sections.
Radar systems are composed of a transmitter and a receiver, as shown in Figure
7.1, which are not necessarily collocated. The transmitter generates and emits a
signal waveform that propagates to the object; a portion of the incident energy is
reflected or scattered back toward the receiver. The fraction of the incident energy
that is reflected toward the receiver depends on the reflectivity of the object and the
number and relative positions of the scatterers that make up the object.

Transmitter
st(t)

Receiver sr(t)

R
Figure 7.1  A general radar system consists of a transmitter and a receiver, and measures the signal
reflected off the object of interest.
7.1  Radar Fundamentals 231

7.1.1  Configurations and Measurements


Radar systems can be categorized based on the relative locations of their transmit-
ters and receivers. Radar systems in which the transmitter and receiver are col-
located and share an antenna are referred to as monostatic radar systems. The
transmitted and received signals are separated in the hardware through the use of a
circulator connected to the antenna or by a switch that is flipped based on whether
the radar is transmitting or receiving. Monostatic radar has the benefit of hardware
reuse in the antenna and typically uses the local oscillator for both upconverting
the transmitted signal and downconverting the received signal. For high-power sys-
tems, the circulator or switch must have good isolation such that the transmitted
power does not leak into the receiver; this is especially true for continuous-wave
systems that continuously and simultaneously transmit and receive a signal.
Radar systems that have physically separated transmitters and receivers are re-
ferred to as bistatic radar systems and can measure the reflected signal off the object
of interest from different angles. Reflectivities of even simple objects are generally
angle-dependent, and thus bistatic radar offers a method of gathering informa-
tion that can be beneficial for certain purposes, such as object classification. Radar
systems can also be multistatic, in which there are multiple transmitters, multiple
receivers, or both.
Consider a simple radar system that transmits a signal with carrier frequency
fc in free space, as in Figure 7.1. The transmitter and receiver are now assumed to
be collocated and share the antenna. The object is assumed to be a simple scatterer,
composed of one scattering point, such as a sphere whose diameter is much shorter
than the wavelength. The complex transmitted signal can be given by

st (t) = At e - j 2π fct (7.1)

where At is the amplitude of the transmitted signal. For a transmitted pulse, the
amplitude is

æ tö
At (t) = AP ç ÷ (7.2)
è τ ø

where τ is the temporal duration of the pulse and P is the rectangle function, de-
fined by

ì a
æ t ö ï1, t £
Pç ÷ = í 2 (7.3)
è aø ï
î0, otherwise

The signal incident on the object is

æ Rö
- j 2π fc ç t - ÷
è cø
si (t) = LAt e (7.4)

where L is the one-way loss in amplitude due to propagation from the antenna to
the object, which includes both the effects of the wavefront spreading and losses in
the medium, and R is the range between the transmitter and the object. Propagation
232 Microwave and Millimeter-Wave Remote Sensing for Security Applications

losses, as described in Chapter 3, may include atmospheric propagation loss or loss


in media such as building walls or clothing.
The signal reflects off the object and propagates back to the antenna through
the same propagation medium, encountering the same loss L. The object has reflec-
tivity G; thus, the signal collected at the receiver is
æ 2R ö
- j 2π fc ç t - ÷
2 è c ø
sr (t) = GL At e (7.5)
where the loss L is squared due to the two-way propagation. The received pulse
travels a distance 2R, and the time delay between transmission and reception is thus
td = 2R/c, as shown in Figure 7.2. The range to the object can therefore be calcu-
lated by measuring the time delay,
td c
R=
2 (7.6)
If the propagation loss is known, the reflectivity of the object can be measured by
noting that the received signal amplitude is

Ar = GL2 At (7.7)
and thus the object’s reflectivity is given by

Ar
G=
L2 At (7.8)
If the object is moving towards the radar with radial velocity vr relative to the
sensor, and the transmitter and receiver are collocated, the range is no longer constant
and is given by

R(t) = R − vr t (7.9)
The received signal is then
æ 2 R 2 vr t ö
- j 2π fc ç t - + ÷
2 è c c ø
sr (t) = GL At e (7.10)
The instantaneous frequency of (7.10) is given by

1 dφ 2v f
f = = fc + r c
2π dt c (7.11)
st(t)
sr(t)

tf = 2R/c

Figure 7.2  The received signal is delayed by a time equal to twice the range divided by the velocity
of propagation.
7.1  Radar Fundamentals 233

The second term of (7.11) represents a shift in the frequency of the received
signal, which is due to the radial motion of the object; this shift in frequency is the
Doppler effect, and the frequency shift is called the Doppler frequency fD:

2vr fc
fD =
c (7.12)
A positive frequency shift indicates motion toward the radar, while a negative
shift indicates motion away from the radar. A simple frequency analysis can thus be
used to measure the radial velocity and radial direction of the object based on the
Doppler frequency shift.
The measurement of the angle of the object relative to the broadside direction
of the radar antenna is determined primarily by the antenna configuration. For
narrowbeam antennas, it is required to know only the direction that the receiver is
pointing in order to measure the angle to the object. For widebeam systems utilizing
a single antenna, precise measurement of the angle is not directly possible; for wide-
beam systems utilizing multiple antennas, the angle can be estimated using various
direction-of-arrival techniques.
The four basic quantities of the object measured by a radar system are the
range, determined by the time delay through (7.6); reflectivity, given by (7.8); ve-
locity, determined by the Doppler frequency shift though (7.12); and angle. Each
quantity is useful in various security sensing applications. Range, radial velocity,
and angle are important for detecting moving humans and classifying human activ-
ity; contraband detection relies on range and angle or reflectivity and angle to form
images; and the detection of human biological phenomena, such as respiration and
heartbeat, rely on velocity measurements.

7.1.2  Range Equation


The radar range equation is a fundamental equation relating the power received by
the radar to parameters of the radar system, the object of interest, and the propaga-
tion medium. The power received by the radar receiver is characterized by first con-
sidering a system that includes a transmitter and a receiver with isotropic antennas
and in which transmission occurs directly between the transmit and receive anten-
nas with no object present, as shown in Figure 7.3. The radar transmitter generates
the transmit power Pt, which it radiates through the antenna. The power density
incident on the receiving antenna is
Pt
Sr = (7.13)
4π R2

Gt Gr
Pt Pr
Transmitter Receiver

R
Figure 7.3  Power transmission between a transmitter and receiver in free-space. Pt and Pr are the
transmit and receive powers; Gt and Gr are the transmit and receive antenna gains.
234 Microwave and Millimeter-Wave Remote Sensing for Security Applications

where the R2 term accounts for the spreading of the wavefront as a function of
distance such that the power integrated over all space is equal to Pt­, satisfying con-
servation of energy:

òò Sr dW = Pt (7.14)
W
The receiver power incident on the receiving antenna with effective aperture size
Ar is

Pr = Sr Ar (7.15)

For an isotropic antenna, the gain is unity and the effective aperture is given by
(4.113), and is repeated here:

λ2
Ar =
4π (7.16)
The received power is therefore
2
 l 
Pr = Pt  
 4p R  (7.17)
For real antennas, the gains of the transmit and receive antennas will be greater
than unity, and the received power is then
2
æ λ ö
Pr = PtGtGr ç
è 4π R ÷ø (7.18)
where Gt and Gr are the gains of the transmitting and receiving antennas, re-
spectively. Equation (7.18) is the Friis transmission formula, which describes
the received signal power from a one-way transmission between systems in free
space [7].
To derive the received power in a two-way radar measurement, consider a sys-
tem that illuminates an object with reflectivity G and receives the reflected energy,
as in Figure 7.1, where the system has separate antennas for transmit and receive.
The power density incident on the object is

Pt Gt
Si =
4π R2 (7.19)
The object reradiates a fraction of the incident power proportional to its reflectivity.
The reflected power can be characterized by taking the ratio of the reflected power
to the incident power density:

Ps
σ=
Si (7.20)
This is called the object’s radar cross section (RCS). The units of the RCS are
square meters, although the quantity is not directly related to the physical size of
7.1  Radar Fundamentals 235

the object. The RCS represents the area of a perfectly reflecting surface that reflects
an equivalent amount of power back to the receiver.
The reflected power that is backscattered to the receiver is

Pt Gt σ
Ps = σ Si =
4π R2 (7.21)
The backscattered power is essentially the power that is “transmitted” by the ob-
ject; the power density at the receiving antenna is thus
Ps
Sr =
4π R2 (7.22)
and the received power is

λ2
Pr = Sr Ar = SrGr (7.23)

or

Gt Grσ λ2
Pr = Pt (7.24)
(4π )3 R4

Including propagation losses through the atmosphere and any medium be-
tween the radar and the object, all represented by L, the received power is given
by

Gt Grσ λ2 L2
Pr = Pt (7.25)
(4π )3 R4

This is the radar equation, and an important result is that the power received by
the radar is proportional to the inverse of the fourth power of the range. This loss
in power is due to the two-way spreading loss incurred and is a significant factor in
designing radars.
The radar equation describes the power of the received signal at the output ter-
minals of the receiving antenna but does not describe the ability of the radar to de-
tect such a signal. The minimum detectable signal is defined in terms of the system
noise parameters as described in Chapter 5 and is generally 3 dB above the noise
floor. However, to achieve sufficient probabilities of detection, many radar applica-
tions require SNR levels greater than 3 dB. In such a case, the minimum detectable
signal power Pmin varies depending on the application, and the difference between
Pmin and the noise floor defines the minimum signal-to-noise ratio SNRmin. The
maximum range within which a radar system can detect a reflected signal can then
be found in terms of (7.25) and the minimum required received power Pmin by

1
é P G G σ λ2 L2 ù 4
Rmax =ê t t r 3 ú (7.26)
ë Pmin (4π ) û
236 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Recall that the noise figure of a receiver is given by

Pi Ni
F= (7.27)
Po No

where Pi and Po are the input and output signal powers, No is the output noise
power, and the input noise power is

Ni = kT Df (7.28)

For Pi = Pmin,

Pmin = FkT Df SNR min (7.29)


where SNRmin is the minimum required signal-to-noise ratio. The maximum range
is therefore
1
é Pt Gt Grσλ 2 L2 ù4
Rmax =ê 3 ú
ë (4π ) FkT Df SNR min û (7.30)
Equation (7.30) is the radar range equation, and it gives the maximum range at
which a signal can be detected based on parameters of the system, the object, the
propagation medium, and the minimum required SNR.

7.2  Transmitter Systems

7.2.1  Transmitter Functionality


An active system is composed of both a receiver and a transmitter, which in con-
junction form a transceiver. The transmitter generates the desired signal waveform
and broadcasts it through the antenna. The receiver collects and measures the signal
that is backscattered off the object of interest. A simple transmitter may consist of
only a monochromatic signal generator and an antenna; a more typical and more
complicated transmitter, shown in Figure 7.4, consists of an oscillator to generate
a monochromatic signal close to the desired carrier frequency, a baseband signal
generator to create the desired signal waveform, a mixer to upconvert the baseband
signal to the carrier frequency, and a high-power amplifier to increase the transmit-
ted signal power. A system transmitting coherent pulses includes a pulse-power

sbb(t) st(t)
DAC

sLO(t) HPA Antenna

LO

Figure 7.4  Upconverting transmitter consisting of a local oscillator generating a signal sLO(t), base-
band signal generator such as a digital-to-analog converter (DAC) generating a signal sbb(t), upcon-
verting mixer, and high-power amplifier.
7.2  Transmitter Systems 237

amplifier after the local oscillator, which essentially switches the local oscillator
signal tone on and off.
The mixer translates the baseband signal to the desired carrier frequency by
nonlinear multiplication with the local oscillator. The baseband signal is centered
on the baseband frequency fbb and is modulated either in amplitude, phase, or both;
a general transmitted waveform can be given by
sbb (t) = A(t)cos [ 2π fbbt + φ (t)]
(7.31)
where A(t) is the amplitude modulation and f(t) is the phase modulation. The local
oscillator generates a monochromatic tone given by

sLO (t) = ALO cos(2π fLOt + φLO ) (7.32)


The mixer is a nonlinear device, as discussed in Section 5.4.2, and thus generates
intermodulation products at frequencies

m2π fbb ± n2π fLO , m = 0,1, 2,…, n = 0,1, 2,… (7.33)

For an upconverting mixer the frequencies of interest are the resulting fLO ±
fbb frequencies. This is shown in the diagram of Figure 7.5. Typically, fbb << fLO,
and thus the carrier frequency fc is close to fLO. This is the case because it is easier
to generate baseband signals with the desired waveform characteristics than it is
to generate directly modulated microwave or millimeter-wave signals; digital-to-
analog converters can generate any desired baseband signal within the bandwidth
limitations of the generating device, which can then be upconverted to the millimeter-
wave carrier frequency. The difference frequencies thus emerge as sidebands to the
LO frequency. If the desired carrier frequency is the fLO + fbb frequency, the signal
is referred to as the upper sideband, while the fLO − fbb frequency is the lower side-
band. The undesired sideband is referred to as the image frequency; mixers designed
to filter out the undesired sideband are called image-rejection mixers. The transmit-
ted signal is given by
st (t) = At (t)cos éë2π (fLO ± fbb )t + φ (t) + φ sys ùû (7.34)

where At includes gain and loss contributions from amplifiers and the upconversion
and fsys is the total phase shift through the system, including the LO phase and the
mixer phase.

∆fRF ∆fRF ∆fRF

0 fbb fLO - fbb fLO fLO + fbb

Input spectrum Output spectrum


Figure 7.5  Input and output signal spectrum of a typical millimeter-wave upconverter.
238 Microwave and Millimeter-Wave Remote Sensing for Security Applications

The use of a high-power amplifier prior to the antenna is common in order to


increase the transmitted power. Amplifiers introduce additional distortions, how-
ever, including harmonics of the input frequencies and intermodulation of multiple
frequencies, as described in Chapter 5. The compression point of the amplifier must
be considered if linear operation of the transmitter is required.
Monostatic radar systems use the antenna to both transmit and receive. In or-
der to facilitate the reuse of the antenna, a component must be included that isolates
the receiver from the transmitter. Because of the losses incurred by the signal as it
propagates and reflects off the object, receivers are designed to receive signals of
much lower power than the transmitter emits, and thus isolation is necessary to
ensure that the receiver hardware is not damaged by the high-power transmit signal
leaking into the receiver; additionally, transmit signals entering the receiver may
affect detection algorithms in the receiver where high-power signals directly from
the transmitter could generate false alarms. Monostatic radar systems use either
a transmit/receive (T/R) switch, or a circulator, as shown in Figure 7.6. The T/R
switch can be an electrical single-pole double-throw dielectric switch, or for high-
power systems, a mechanical relay switch. A circulator is a passive three-port de-
vice that facilitates signal transmission in one direction around the ports. It allows
nearly lossless transmission in the direction from port 1 to port 2 and from port 2
to port 3, with very high attenuation in the other direction. If the transmitted signal
is connected to port 1, the antenna to port 2, and the receiver to port 3, the radar
system can simultaneously transmit and receive signals since any transmitted signals
will be directed out of the antenna and any received signals will be directed to the
receiver. Reflections of the transmitted signal from the antenna due to impedance
mismatch will, however, propagate directly to the receiver; thus, it is important that
the antenna is well matched to reduce reflections.

Transmitter Transmitter

Receiver Receiver

(a) (b)
2

L=0 L=∞ L=∞ L=0


1 3

L=∞

L=0
(c)
Figure 7.6  (a) Monostatic radar with T/R switch. (b) Monostatic radar with a circulator. (c) Ideal
circulator functionality.
7.2  Transmitter Systems 239

7.2.2  Transmitter Noise


Noise generated by the transmitter is located in a bandwidth around the carrier
frequency and can mask return signals with low Doppler frequency shifts when
stationary clutter is also present in the return signal. The ideal oscillator signal in
a transmitter is a monochromatic tone with a bandwidth of 0 Hz. Such a signal
is clearly not realizable, if only for the fact that a zero bandwidth signal requires
an infinitely long temporal duration. Additionally, the mechanisms used to gen-
erate the oscillator signal are nonideal and impose additional noise effects onto
the signal. This noise includes amplitude-modulated (AM) noise and frequency-
modulated (FM) noise, the latter of which is usually analyzed in the time domain,
where it is referred to as phase-modulated (PM) noise. The effects of AM and PM
noise are shown in Figures 7.7(a) and 7.7(b). A nonideal oscillator produces a signal
that has a strong peak at the carrier frequency and spreads out in frequency around
the carrier, as depicted in Figure 7.8. A noisy oscillator signal is thus given by (7.32)
with AM and PM modulation:

sLO (t) = ALO [1 + ε(t)]cos [2π fLOt + φ (t)] (7.35)


where ε(t) is the AM noise and f(t) is the phase noise; thus, the noise contributes as
a modulation to the ideal signal. Typically,

ε(t) << 1 and φ (t) << 1 (7.36)


Because the noise is contained within a bandwidth about the carrier frequency,
multiplication or upconversion of the noisy signal results in an increase in the noise.
For instance, consider a signal at carrier frequency fc with noise power of −50 dBc
at a frequency fc ± 100 Hz, where dBc is the power level in decibels relative to the
power of the signal at the carrier frequency. If this signal is frequency multiplied up
to 2fc, the bandwidth will likewise increase, and the noise power at 2fc ± 100 Hz
will be −44 dBc, or an increase in 6 dB per doubling of the frequency. Similarly,
a signal that is frequency divided will have a decrease in noise power of 6 dB per
halving of the frequency.

(a) (b)
Figure 7.7  (a) Amplitude noise in the time domain. (b) Phase noise in the time-domain.
240 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Spectral
power

f
fLO

(a)

Spectral
power

f
fLO

(b)
Figure 7.8  (a) Ideal oscillator tone. (b) Real oscillator tone with noise.

Amplitude noise is caused by random fluctuations of the signal amplitude and


generally induces less noise power on the signal than the phase noise. As such, it is
often disregarded to focus on the larger noise contributed by the phase noise. Little
research is published in the open literature on amplitude noise in oscillators; how-
ever, a power law model has been suggested to approximate the specific contribu-
tions to the spectral amplitude noise power Sε(f ) [8]:

0
Sε (f ) = å hn f n (7.37)
n =-2
The AM spectrum model consists of contributions from three primary mecha-
nisms, depicted in Figure 7.9: random walk of amplitude with slope f −2; amplitude
flicker with slope f −1; and white noise, generating a spectrum with slope f 0.

f -2 Random walk
Sε (t)

f -1 Flicker
f 0 White noise

f
Figure 7.9  Oscillator amplitude noise spectrum.
7.2  Transmitter Systems 241

Table 7.1  Mechanisms Generating Phase Noise (After: [9])


Random walk of frequency f −4
Flicker frequency noise f −3
Random walk of phase f −2
Flicker phase noise f −1
White phase noise f0

The phase noise of an oscillator is due to random fluctuations in the signal


phase and frequency; a detailed analysis of phase noise in oscillators can be found
in [9]. A widely used power law model for the spectral phase noise is
0
Sφ (f ) = å bn f n (7.38)
n =-4
where the mechanisms that give rise to each term are summarized in Table 7.1. The
phase noise contributions of an oscillator are shown in Figure 7.10, including a
thermal noise component with slope f −5 that is sometimes seen at frequencies close
to the carrier, but that is often disregarded.
Amplifier phase noise includes contributions from only flicker phase noise and
white phase noise, as shown in Figure 7.11. There is also a term with a slope of f −5
that is sometimes included and often seen in close-in noise measurements of ampli-
fiers, and that is thermal in origin [9].

7.2.3  Millimeter-Wave Oscillators


Microwave oscillators can be produced using solid-state techniques, and can be
made with low noise characteristics; however, low-noise millimeter-wave oscillators,
with frequencies extending into the hundreds of gigahertz, are more difficult to pro-
duce. Millimeter-wave oscillators are constructed using tube oscillators, solid-state
oscillators, or low-frequency oscillators in combination with frequency multipliers
[10, 11]. At very low frequencies, crystal oscillators generate signals with good
noise characteristics; however, their output frequencies are in the megahertz range
and thus require multiple multiplication stages in order to generate microwave
frequencies, which degrade the noise performance. Solid-state voltage-controlled
oscillators (VCOs) such as the Colpitts oscillator are based on transistor oscilla-
tors or dielectric resonator oscillators (DROs) and can generate frequencies up to

f -5 Thermal noise

f -4 Frequency random walk


Sφ (t)

f -3 Frequency flicker
f -2 Phase random Walk
f -1 Phase flicker
f 0 White noise
f
Figure 7.10  Oscillator phase noise spectrum.
242 Microwave and Millimeter-Wave Remote Sensing for Security Applications

f -5 Thermal noise

Sφ (t)
f -1 Flicker
f 0 White noise
f
Figure 7.11  Amplifier phase noise spectrum.

tens of gigahertz, which can then be frequency multiplied up to millimeter-wave


frequencies. Frequency multiplication is accomplished using varactor diodes whose
reactance is varied by changing the junction capacitance. Varactors generate har-
monics of an applied RF signal; the desired harmonic may then be selected with a
filter. Multiplier chains can produce frequencies above 300 GHz; however, the noise
of the signal increases by 6 dB with each doubling, which can result in significant
noise degradation.
Solid-state oscillators are available at millimeter-wave frequencies; however,
their efficiency is typically much less than lower-frequency VCOs. Tube oscilla-
tors are generally too large to be used in a practical security sensor; however, they
produce high output power and are often used in military and commercial radar
applications. For size reasons, solid-state oscillators are typically used in security
sensors. Millimeter-wave solid-state oscillator devices are commonly diode based
and are designed with the diode coupled to a microstrip circuit or waveguide; wave-
guide circuits can provide high output power and are often found at frequencies
in the gigahertz range where microstrip is considerably more lossy. Frequency-
tuned devices use either mechanical tuners or electrical tuning with either varactors
or magnetically tuned ferrite spheres such as Yttrium Iron Garnet (YIG). Varac-
tors generally have high tuning bandwidths (greater than 1 GHz·μs−1 compared
to ~ 1 MHz·μs−1 for YIG spheres); however, ferrite sphere tuners generally have
better tuning linearity.
The most common solid-state millimeter-wave oscillators are the IMPATT di-
ode and Gunn diode oscillators. IMPATT diode oscillators are based on avalanche
breakdown multiplication in the diode, which is used to induce a phase shift between
an applied RF voltage and the resulting external current signal. This phase shift,
if greater than 90°, results in a negative reactance, and the various regions of the
diode are designed such that the negative reactance occurs at the desired millimeter-
wave frequency. IMPATT diode oscillators made of Gallium Arsenide (GaAs) have
been shown to produce frequencies in excess of 250 GHz; however, due to the
statistical nature of avalanche breakdown, their phase noise is generally poor com-
pared to Gunn diode oscillators.
Gunn diode oscillators are based on the transferred electron effect, where an
electric field applied to a semiconductor induces a charge dipole near the cathode;
the dipole drifts to the anode, producing a current pulse, and the process repeats.
7.3  Radar Measurement Sensitivity 243

The frequency of the resulting oscillation is dependent on the drift velocity vs of the
semiconductor and the device length l:

vs
fc =
l (7.39)

where, for GaAs, vs » 107 cm·s−1, and thus the oscillation frequency is

fc = l -1107 Hz (7.40)
where l is measured in cm. Gunn diode oscillators have efficiencies of around
10% and produce lower output power than IMPATT diode oscillators; however,
their phase noise characteristics are often better and thus are most commonly
used in millimeter-wave security sensors. Commercially available Gunn diode
oscillators are available at frequencies above 110 GHz with good phase noise
characteristics.

7.3  Radar Measurement Sensitivity

The sensitivity of a radar measurement is defined in terms of the rms error of the
measurement. Similar to the sensitivity of a radiometer, the sensitivity of a radar
measurement describes the accuracy with which a measurement can be made.
In contrast, the resolution of a radar measurement is determined in terms of a
specific resolution element, such as pulse width for the measurement of range,
that defines the accuracy to which a measurement can be known given a specific
threshold value. For instance, in the measurement of frequency in determining
radial velocity, the 4 dB width of the spectral signal is often taken as the threshold
for the resolution. Any two signal returns whose frequency responses occupy a
bandwidth greater than the defined resolution are said to be resolved, whereas
two returns whose responses occupy a bandwidth less than the resolution are
unresolved. Thus, measurement resolution is a derived quantity, determined by
an existing threshold value that the detection operates on. The sensitivity of the
measurement, however, is determined by the rms error of the measurement, and
in particular, for the measurement of range, velocity, and angle, the rms errors
are determined by the noise. In this section, the theoretical rms error in measur-
ing range, frequency, and angle are reviewed; detailed analyses are provided by
Skolnik [1] and Barton [2].

7.3.1  Measurement Error


The rms error in a given radar measurement is of the form

N0
σ =C (7.41)
2E

where C is a constant that depends on the specific resolution element of the given
measurement (described later), N0 = kTsys is the noise power in a 1-Hz bandwidth,
244 Microwave and Millimeter-Wave Remote Sensing for Security Applications

and E is the signal energy. The ratio E ∕ N0 is the maximum signal-to-noise ratio,
achieved with a matched filter (see Section 7.6.2), and thus [2]

E
SNR matched = (7.42)
filter N0

Equation (7.41) therefore states that the measurement error is reduced by the
signal-to-noise ratio. The error can thus be made arbitrarily small by increasing the
signal-to-noise ratio; in other words, the measurement error is fundamentally limited
by the system noise. The following equations for the rms measurement error assume
that the signal-to-noise ratio is high enough that the signal amplitude for a given
measurement is greater than the detection threshold for a given measurement [12].
For each measurement of range, velocity, and angle, there is a resolution ele-
ment that defines the measurement resolution; for each measurement there is also
an rms error element, which, along with the signal-to-noise ratio, defines the rms
error of the measurement. These will each be described in the following discussion.
Table 7.2 summarizes the measurements and resolution elements; note that the rms
error elements are equivalent to the resolution elements in their respective trans-
form coordinates.

7.3.1.1  Range Measurement Error


The measurement of range to an object is essentially the measurement of the time
delay of the return signal. For a signal s(t) with the Fourier transform S(f ), the rms
error in measuring the time delay is given by

1 N0
σt = (7.43)
β 2E

where the rms bandwidth β is defined by


¥
2
ò (2π f )
2
S(f ) df

β2 = ¥
(7.44)
2
ò S(f ) df

The numerator of (7.44) is the second moment of the energy spectrum, while the
denominator is the total energy of the signal.

Table 7.2  Radar Measurements and Resolution Elements (Adapted from [2])
Measurement Resolution element Rms error element
Range Time delay tr Rms bandwidth β
Frequency Signal bandwidth Df Rms pulse width α
Angle Beamwidth θ­BW Rms aperture width γ
7.3  Radar Measurement Sensitivity 245

7.3.1.2  Frequency Measurement Error


The measurement of radial velocity through the Doppler shift is the measurement of
signal frequency. The rms error in the measurement of the signal frequency is given by

1 N0
σf = (7.45)
α 2E

where the rms time duration of the signal is defined by


¥

ò (2π t)
2 2
s (t)dt

α2 = ¥
(7.46)
òs
2
(t)dt

where the numerator is the second moment of the signal distribution. The denomi-
nator of (7.46) is the total signal energy, as is that of (7.44); this can be seen by using
Parseval’s theorem, which states that the integral of the square of a function is equal
to the integral of the square of its Fourier transform. Thus,
¥ ¥
2
òs ò
2
(t)dt = S(f ) df = E (7.47)
-¥ -¥

7.3.1.3  Angle Measurement Error


The rms error in measuring the angle of the signal return is given by

1 N0
σθ = (7.48)
γ 2E

where the rms antenna aperture width γ is defined by


¥ 2
æ 2π x ö 2
ò çè
λ
÷ø A(x) dx
γ2 = -¥
¥
(7.49)
2
ò A(x) dx

where A(x) is the illumination function on the antenna aperture that describes the
distribution of current across the aperture in the x dimension, and which may be
uniform or tapered to reduce sidelobes.

7.3.1.4  Example
An example of the rms measurement errors is given in the following for the simple
case of a rectangular pulse and uniform aperture illumination. A simple rectangular
pulse is given by
246 Microwave and Millimeter-Wave Remote Sensing for Security Applications

æ tö
s(t) = P ç ÷ e - j 2π fct (7.50)
èτø
and its Fourier transform is given by
S(f ) = sinc [π (f - fc ) τ] (7.51)

The signal can be considered to be downconverted to baseband, without loss
of generality, such that

æ tö
sbb (t) = P ç ÷ (7.52)
è τ ø

and

Sbb (f ) = sinc(π f τ ) (7.53)

The subscript bb, indicating baseband, will be dropped for the sake of brevity in
the following.
The rms bandwidth is given in terms of the spectral signal (7.53) by
¥ ¥

ò (2π f )2 sinc 2 (π f τ)df 4π 2 ò sin


2
(π f τ)df
2 -¥ -¥
β = ¥
= ¥
(7.54)

ò sinc òf
2 -2
(π f τ )df sin2 (π fτ )df
-¥ -¥
The numerator of (7.54) is
¥ f =¥
éf 1 ù
4π ò sin (π f τ )df = 4π ê -
2 2 2
sin(2π f τ )ú =¥ (7.55)
-¥ ë 2 4π û f =-¥

Substituting this result into (7.43) yields σt = 0; that is, for a perfectly rectan-
gular pulse there is no theoretical error in the time-delay measurement. This is due
to the assumption of a perfectly rectangular pulse, the rise time of which is zero,
depicted in Figure 7.12(a). The amplitude change of the received pulse is instanta-
neous and not subject to noise. Recall that the signal-to-noise ratio was assumed to

1
tr = 0 tr ≈
∆fr

t t
(a) (b)
Figure 7.12  Pulse signals with (a) infinite bandwidth and zero rise-time, (b) finite bandwidth and
nonzero rise time. The bandwidth and rise time are approximately inversely related.
7.3  Radar Measurement Sensitivity 247

be high enough so that the amplitude level of the pulse is greater than the change
detection threshold. Thus, the amplitude of the pulse is not affected by the noise;
however, the transition region is affected by noise if the rise time is finite.
Essentially, the result given by (7.55) is due to the assumption of infinite band-
width of the signal. For a real pulse, the rise time will not be zero due to the
limited bandwidth of the pulse signal, as shown in Figure 7.12(b). For a band-
limited pulse, the bandwidth is approximately proportional to the inverse of the
rise time:

1
Dfr » (7.56)
tr

The rms bandwidth is then given by


Dfr 2

ò
2
4π sin2 (π f τ )df
-Dfr 2
β2 = Dfr 2 (7.57)
ò
-2 2
f sin (π f τ )df
-Dfr 2
The numerator of (7.57) is
Dfr 2
é Df 1 ù
4π 2 ò sin2 (π f τ )df = 4π 2 ê r -
ë 2 2πτ
sin(π Dfrτ )ú
û
(7.58)
-Dfr 2

and the denominator is found to be


Dfr 2
1
ò f -2 sin2 (π f τ )df = 2πτ Si(πDfrτ ) +
Df
[cos(π Dfrτ) - 1] (7.59)
-Dfr 2
where Si(x) is the sine integral, given by [13]
x
Si(x) = ò t -1 sin(t)dt (7.60)
0
Substituting (7.58) and (7.59) into (7.57) yields the rms bandwidth

1 ïì πDfr τ - sin(π Dfrτ ) ïü


β2 = í ý (7.61)

τ2 ïî Si(πDfr τ) + [cos(π Dfrτ ) - 1] π Dfrτ ïþ

This expression simplifies in the case that the rise time is significantly shorter
than the pulse width, or when τ ® ∞ for a constant Dfr. The large argument limit
of the sine integral is

π
lim Si(x) = (7.62)
x®¥ 2
248 Microwave and Millimeter-Wave Remote Sensing for Security Applications

and therefore

2 Dfr
lim β 2 (τ ) = (7.63)
τ ®¥ τ

Thus, for rise times significantly shorter than the pulse length, the large-argument
rms bandwidth can be given by

2 Dfr
βL = (7.64)
τ

The rms time delay error is then found by substituting (7.64) into (7.43), which
yields

τ N0 τ t r N0
σt = = (7.65)
4Dfr E 4E

The rms error in measuring the time delay of the return signal is thus dependent
on the square root of the product of the pulse width and the rise time. In practice,
the assumption that the rise time is significantly smaller than the pulse duration is
not always satisfied; assuming that the pulse width is twice the rise time τ = 2tr and
substituting into (7.61) yields

2.1
β» (7.66)
τ

The rms time delay error is then given by

τ N0
σt = (7.67)
2.1 4E

Radar waveforms are shaped in some applications to have gradual rise and fall
times to minimize the out-of-band spectral content that is generated with fast rise
times [1]. Waveforms of this type incur errors closer to the form of (7.67).
Figure 7.13(a) shows a plot of the rms bandwidth given by the exact formula
for a band-limited pulse (7.61) and the rms bandwidth given by the large-argument
approximation (7.64), as a function of the ratio of the pulse width to the rise time
τ ∕tr. The large-argument approximation follows the general trend of the exact for-
mulation; however, it does not account for the ripple produced by the sinusoid
in the numerator of (7.61). As τ  ∕tr increases, the numerator approaches πτ  ∕tr as
the first term of the numerator increasingly dominates over the second sinusoidal
term. Figure 7.13(b) shows the percentage error between (7.61) and (7.64), which
oscillates due to the sinusoidal variation of the numerator. The trend of the error
is decreasing; when τ  ∕tr > 25 the maximum error is less than 1%. A comparison of
the normalized measurement error, given by

2E
σ t,N = σ t = β -1 (7.68)
N0
7.3  Radar Measurement Sensitivity 249

1 β
β
L
0
0 2 4 6 8 10
τ/tr
(a)

25

20

15
β/β (%)
L

10

0
0 2 4 6 8 10
τ/tr
(b)
Figure 7.13  (a) Rms bandwidth of a band-limited pulse calculated using the exact formulation and
the large-argument approximation. (b) Percentage error between the calculations of the rms band-
width using the exact formulation and the large-argument approximation.

calculated using the exact rms bandwidth, σt,N, and the large-argument approximated
rms bandwidth, σt,N,L, is shown in Figure 7.14. The error is greater for lower values
of τ /tr than the error for the rms bandwidth, indicating a more severe impact on the
measurement error using the large-argument approximation for small values of τ/tr.
Given the perfectly square pulse signal (7.52), the rms time duration is given by

¥ τ 2
æ tö
ò
2
(2π t) P ç ÷ dt
èτø
4π 2 ò t 2dt
-τ 2 π 2τ 2
α2 = -¥
¥
= τ 2
= (7.69)
æ tö 3
ò P ç ÷ dt
èτø ò dt
-¥ -τ 2
250 Microwave and Millimeter-Wave Remote Sensing for Security Applications

2
σ
t,N
σt,N,L
1.5

0.5

0
0 2 4 6 8 10
τ/tr

(a)

25

20
(%)

15
t,N,L

10
t,N
σ

0
0 2 4 6 8 10
τ/tr
(b)
Figure 7.14  (a) Rms time delay measurement error of a band-limited pulse calculated using the
exact formulation and the large-argument approximation. (b) Percentage error between the calcula-
tions of the rms time delay error using the exact formulation and the large-argument approximation
of the rms bandwidth.

The rms frequency error is then given by substituting (7.69) into (7.45), which
yields

1 3N0
σf = (7.70)
πτ 2E

In this case, the resolution element is the time duration, and due to its finite length,
the rms measurement error can be calculated using the representation for a perfectly
rectangular pulse.
The rms error in measuring the angle of the return signal is dependent on the
resolution element of antenna beamwidth. For an antenna with uniform illumina-
tion across an aperture of length D, the aperture illumination can be given by
7.3  Radar Measurement Sensitivity 251

æ xö
A(x) = P ç ÷ (7.71)
è Dø

where x is the spatial dimension of the antenna. Although this example accounts for
only one dimension, the extension to two dimensions involves simply repeating the
process in the orthogonal dimension to determine the angular measurement error
in both spatial dimensions. Substituting (7.71) into (7.49) yields

D2

2 ò x 2dx
æ 2π ö -D 2 π 2D2
γ2 =ç ÷ = (7.72)
è λ ø D2
3λ 2
ò dx
-D 2
The rms angle measurement error is found by substituting (7.72) into (7.48),
which results in

λ 3N0
σθ = (7.73)
π D 2E

The angle error is therefore inversely dependent on the aperture size, which is ex-
pected, since the beamwidth of an antenna is inversely proportional to the antenna
size as described in Chapter 4, and an antenna with a narrower beamwidth will
have finer angle accuracy.

7.3.2  Impact of the Time-Bandwidth Product on Measurement Error


The product of the rms time and bandwidth expressions gives insight into the
relationship between the measurement of time delay and frequency. Because the
resolution elements of these two measurements are inversely related, the two mea-
surements and their errors are fundamentally linked. In particular, the product of
the time-delay and frequency errors is given by

N0
σ t σf = (7.74)
αβ2E

For a fixed signal-to-noise ratio, the measurement error can be improved through a
larger rms time-bandwidth product; this requires long pulses with wide bandwidth.
As will be shown in the following, the rms time-bandwidth product is always greater
than π. This section follows the derivation of this relation given by Skolnik in [14].
The rms time-bandwidth is given by multiplying (7.44) and (7.46), yielding

¥ ¥
2
ò (2π t) ò (2π f )
2 2 2
s (t)dt S(f ) df
-¥ -¥
αβ = ¥ ¥
(7.75)
2
òs ò
2
(t)dt S(f ) df
-¥ -¥
252 Microwave and Millimeter-Wave Remote Sensing for Security Applications

The two integrals in the denominator are equal, by Parseval’s theorem (7.47), and
thus

¥ ¥
2
ò (2π t) ò (2π f )
2 2 2
s (t)dt S(f ) df
-¥ -¥
αβ = ¥
(7.76)
òs
2
(t)dt

Using the Fourier transform relationship between the temporal and spectral
signals,
¥

ò S(f )e
j 2π ft
s(t) = df (7.77)

the second derivative of the temporal signal is
¥
d 2 s(t)
= - ò (2π f )2 S(f )e j 2π ft df (7.78)
dt 2 -¥

and thus the second integral in the numerator of (7.76) can be written
¥ ¥ ¥
2 d 2 s(t)
ò (2π f )2 S(f ) df = ò (2π f )2 S(f )S* (f )df = - ò
dt 2
s(t)dt (7.79)
-¥ -¥ -¥
This can be integrated by parts; if it is assumed that the signals are of finite
temporal duration,
lim s(t) = 0 (7.80)
t ®±¥
and the integral results in
¥ ¥ 2
d 2 s(t) é ds(t) ù
-ò 2
s(t)dt = ò êë dt úû dt (7.81)

dt -¥

The rms time-bandwidth product is then
¥ ¥ 2
é ds(t) ù
ò ò êë dt úû dt
2 2
(2π t) s (t)dt
-¥ -¥
αβ = ¥
(7.82)

òs
2
(t)dt

Because the signal s(t) has finite energy, it is square integrable. Thus the Cauchy-
Schwartz inequality may be applied, which states that two square integrable func-
tions f(x) and g(x) satisfy the relation
7.4  Micro-Doppler 253

¥ ¥ ¥ 2
2 2
ò f (x) dx ò g(x) dx ³ ò f (x)g(x)dx (7.83)
-¥ -¥ -¥

Thus, the numerator of (7.82) satisfies


2
¥ ¥
é ds(t) ù
2 é¥ ds(t) ù
ò ò êë dt úû ê ò (2πt)s(t)
2 2
(2π t ) s (t )dt dt ³ dt ú (7.84)
-¥ -¥ êë -¥ dt úû

Integrating this result by parts yields
2 2
é¥ ds(t) ù é ¥ ù
ê ò (2π t)s(t) dt ú = ê π ò s2 (t)dt ú (7.85)
dt
ëê -¥ ûú êë -¥ ûú

and thus
2
¥ ¥
é ds(t) ù
2 é ¥ ù
ò ò êë dt úû ê π ò s2 (t)dt ú
2 2
(2π t ) s (t )dt dt ³ (7.86)
-¥ -¥ êë -¥ ûú

Substituting (7.86) into (7.82) results in

αβ ³ π (7.87)

The rms time-bandwidth product is thus limited at the lower end by the constant π.
The product of the time delay and frequency measurement errors is therefore
N0
σ t σf £ (7.88)
π 2E

This expression is referred to as the radar uncertainty equation. For a given


signal-to-noise ratio the right-hand side is a constant; for a given waveform, the
measurement error of either the time delay or the frequency cannot be decreased
without a corresponding increase in the other. However, the choice of waveform
can increase either or both the duration or bandwidth, thereby improving either or
both measurements, which is why the product of the errors is not strictly equal to
the constant on the right-hand side.

7.4  Micro-Doppler

An object with a nonzero radial velocity produces a Doppler shift in the frequency
of the backscattered radar signal, the frequency of which is proportional to the
radial velocity of the object through (7.12). If the object is composed of multiple
scattering points, Doppler shifts will be produced on the signals scattered off each
individual scattering point. If the scattering points are moving with different radial
velocities, either due to rotation of a rigid object such as a cone or the motion of
254 Microwave and Millimeter-Wave Remote Sensing for Security Applications

multiple parts of a nonrigid object such as a human, each scattering point will pro-
duce a different Doppler shift and thus a different resulting frequency in the radar
return signal. The motion of the object as a whole is referred to as the bulk motion,
and the motions of the scattering points moving at different radial velocities relative
to the bulk motion are referred to as micro-motions. Micro-motion characteristics
are dynamic in time and are often periodic, depending on the object. The bulk
motion of the object gives rise to the bulk Doppler frequency, while the micro-
motions give rise to time-dependent sideband frequencies around the bulk Doppler
frequency; these sideband frequencies are referred to as micro-Doppler. Because
micro-motions tend to be time dependent, micro-Doppler frequencies are likewise
time dependent.

7.4.1  Micro-Doppler in Security Radar


Micro-Doppler frequencies are produced whenever a radar system is viewing a
moving or stationary object with micro-motions in the radial direction. A human
walking toward a radar generates micro-Doppler through the backscattered radar
signals from the torso, arms, and legs. The torso gives rise to the bulk Doppler fre-
quency, while the periodic arm swing and leg swing gives rise to the micro-Doppler
frequencies around the bulk Doppler frequency. The information contained in the
micro-Doppler frequencies can therefore be used to discriminate between objects.
An object that does not generate micro-Doppler frequencies can easily be discrimi-
nated as nonhuman. Vehicles, for instance, primarily produce only a bulk Doppler
frequency shift. Other objects of interest that induce micro-Doppler frequencies
include the rotating blades of a helicopter or the oscillatory rolling and coning
motions of an object entering the atmosphere. Micro-Doppler analysis in radar has
been pioneered by Chen for numerous applications including those listed earlier
[15–19]: in [19] a thorough description of the kinematic processes behind the micro-
motions of rigid and nonrigid objects and the resulting micro-Doppler frequencies
are described.
Analysis of the micro-Doppler signature can be done for object classification.
For instance, the periodicity of the frequency shifts due to arm swing and leg swing
of a human walking toward the radar will be different from the periodicity due to
the leg swing of an animal such as a horse simply due to the differences of bipedal
and quadrupedal motion. This information is contained in the frequency sidebands;
however, because the frequencies vary in time, a simple Fourier analysis will not
extract all the information. If the maximum micro-Doppler frequency produced
by both the human leg swing and the animal leg swing are the same, the Fourier
transform of the signal over multiple stride periods will simply spread over the same
bandwidth; the time-dependence of the micro-Doppler due to the periodicity of the
limb motion is lost.
To recover the time-dependent frequency information contained in the micro-
Doppler frequency, the radar signal is analyzed in the joint time-frequency domain
using a time-frequency transform. The most common time-frequency transform is
the short-time Fourier transform (STFT), which subdivides the temporal radar sig-
nal into shorter temporal windows using a windowing function w(t) and are then
processed through a standard Fourier transform; the resulting collection of spectra
7.4  Micro-Doppler 255

is then combined in series to form the time-frequency plot. The STFT is defined
by [15]
¥

ò s(t ¢)w(t ¢ - t)e


- j 2π ft ¢
STFT(t, f ) = dt ¢

(7.89)
¥

ò S(f ¢)W (f - f ¢)e


- j 2π ft - j 2 π f ¢t
=e df ¢

where the second form is cast in the frequency domain. The magnitude of (7.89)
is called the spectrogram. The window functions w(t) and W(f ) are Fourier trans-
form pairs and thus their width is inversely proportional: a short time window
corresponds to a wide frequency window and vice versa. The resolution of the
transform is proportional to the window length; responses shorter than the win-
dow length will tend to be averaged out. Thus, there exists an inverse relationship
between the time resolution and the frequency resolution of the STFT. A shorter
time window will result in finer temporal resolution but coarser spectral resolution
and vice versa.
As an illustration of the STFT, consider a signal composed of four tones in series
with frequencies of 1 GHz, 4 GHz, 8 GHz, and 2.5 GHz, as illustrated in Figure
7.15(a). In Figure 7.15(b) the Fourier transform of the signal is shown, where the
frequency spikes at the four frequencies can be seen. However, the timing of the four
frequencies is lost in the Fourier transform; that is, the order cannot be discerned.
The spectrogram of the signal is shown in Figure 7.15(c); the temporal change in
frequency is easily discernable in the STFT.

7.4.2  Micro-Doppler Theory


The primary applications for micro-Doppler analysis in microwave and millimeter-
wave security sensing are the detection and discrimination of humans and nonhu-
mans, and the classification of human activity. Both are accomplished by measuring
various aspects of the micro-Doppler signature generated by human motion, such
as arm swing rate, leg swing rate, or the presence or lack of signatures from specific
limbs. Human motion is very complex, and in practice it is analyzed through mea-
surements and measurement-based models in simulation. To illustrate the funda-
mentals of micro-motion-induced micro-Doppler, the following is a simple example
of the motion of rotating propeller blades, a detailed analysis of which can be found
in [19].
Consider the situation of Figure 7.16, where an object with rotating blades is
approaching the radar with a bulk velocity vector

vb = -vb xˆ (7.90)

For simplicity, the only scattering points considered are those at the tips of the
propeller blades; the backscattered signals from the propeller edges and the body
of the vehicle are disregarded. This will allow the resulting signature of the micro-
motions to be more easily characterized.
256 Microwave and Millimeter-Wave Remote Sensing for Security Applications

0 2 4 6 8
t (ns)
(a)

0 2 4 6 8 10
f (GHz)
(b)
10

6
f (GHz)

0
0 2 4 6 8
t (ns)
(c)
Figure 7.15  Stepped frequency signal consisting of four frequencies of 1 GHz, 4 GHz, 8 GHz, and
2.5 GHz in the (a) time domain; (b) frequency domain; (c) joint time-frequency domain.
7.4  Micro-Doppler 257

ωr p1
r1
vb
rp,1 x
α
Radar
R0

ωr
p2

Figure 7.16  Geometric setup describing micro-Doppler induced by rotating propeller blades.

The range to the platform center is

Rb (t) = (R0 + vb ) × xˆ = R0 - vb cos(α )t (7.91)

where α is the angle between the platform motion and the broadside direction of
the radar. The range to the point scatterer p1 is given by

r1(t) = (R0 + ω ´ rp,1) × xˆ = Rb (t) + ( ω ´ rp,1) × xˆ (7.92)

where

ω = - ωr zˆ (7.93)

is the angular rotation vector of the blade with angular velocity ωr and rp,1 is the
vector from the center of the blade (the bulk center point) and p1, given by

rp,1 = xr
ˆ p cos θ + yr
ˆ p sin θ (7.94)

where rp is the length of the blade from the center to the tip. Substituting (7.91) and
(7.94) into (7.92) yields

r1(t) = R0 - v p cos(α )t + rp sin(ω r t) (7.95)

where W(r)t = q
The resulting range to the tip of the blade is therefore oscillatory in time due to
the sin(ωrt) term. The complex radar return from the point scatterer at p1 is there-
fore given by

é 2 ù ì 2 ü
- j 2π fc êt + r1 (t )ú - j 2π fc ít + éëR0 - v p cos(α )t + rp sin(ω r t )ùû ý
ë c û î c þ
s1(t) = A1e = A1e (7.96)

where the reflectivity of the scatterer is included in the signal amplitude A1. The
time-varying range to p1 due to the rotation is shown in Figure 7.17.
R0 + rp

R0

R 0 - rp
t
Figure 7.17  Time-varying range to point p1 due to the rotation.
258 Microwave and Millimeter-Wave Remote Sensing for Security Applications

The range to the scattering point p2 can be found similarly; however, since p2 is
rotating 180° out of phase with p1, the angle term θ = ωrt in (7.96) can be replaced
by θ + 180° = ωrt + 180° for p2, which simply changes the sign of the sinusoidal
term, resulting in

r2 (t) = R0 - v p cos(α )t - rp sin(ω r t) (7.97)

and thus

ì 2 ü
- j 2π fc ít + éëR0 - v p cos(α )t - rp sin(ωr t )ùû ý
î c þ
s2 (t) = A2 e (7.98)

The total radar return signal in the general case is the summation of the signals
backscattered from each scattering point in the object. For the two point scatterers
present in this example, the total return signal is

2 é 2 ù
- j 2π fc êt + rn (t )ú
sr (t) = å An e ë c û (7.99)
n =1
The frequency of the return signal is given by the time derivative of the phase term.
The frequency of the return signal from p1 is thus

1 dφ d é 2 ù
f1 = = fc ê t + r1(t)ú
2π dt dt ë c û
(7.100)
ì 2 ü
= fc í1 - éëv p cos(α ) - rpω r cos(ωr t)ùû ý
î c þ

Similarly, the return signal from p2 is

ì 2 ü
f2 = fc í1 - éëv p cos(α ) + rpω r cos(ωr t)ùû ý (7.101)
î c þ

The frequency spectrum is thus comprised of two sinusoidal oscillations which are
180° out of phase and which form sidebands around a constant center frequency,
which is the bulk Doppler frequency. For a baseband system, the signal is mixed with
the carrier frequency fc, and the baseband bulk Doppler frequency is given by

2 fc vb
fD,b = cos(α ) (7.102)
c
The baseband micro-Doppler frequencies due to the blade tips are therefore given
by

2 fc rpω r
fD,1 = cos(ω r t)
c
(7.103)
2 fc rpω r
fD,2 =- cos(ω r t)
c
7.4  Micro-Doppler 259

The Fourier transform of (7.99) is shown in Figure 7.18(a) with no bulk mo-
tion (fD,b = 0), where the signal amplitudes are equal and normalized to unity. The
detailed characteristics of the micro-Doppler frequency signatures of (7.103) are
clearly lost, as the spectrum spreads continuously between fD and –fD, where fD is
the same for both points p1 and p2 and is given by the magnitude of (7.103); the
maximum Doppler shift may be measured, but the periodicity of the frequency
shift cannot. Analyzing the signal in the joint time frequency recovers the time de-
pendence of the frequency sidebands, as depicted in Figure 7.18(b). The oscillatory
nature of the frequencies is clearly evident. If the bulk motion was included in the
analysis, a strong signal return would also be present at fD,b, where the sidebands
would be located around.
A significant amount of information is present in the spectrogram of (7.99),
which is not present in the spectrum. For instance, it can be seen that there are
clearly two blades present due to the 180° phase shift between the two coincident
sinusoidal oscillations. The propeller rotation rate ωr can also be measured by the
period of one of the oscillations. Furthermore, since the oscillation is proportional
to rpωr in (7.103), once the rotation rate is measured the propeller blade length can
also be calculated by measuring the maximum Doppler shift. The type of vehicle
may be classified from the length of the propeller blade.
The previous example did not account for the reflectivities of each scatterer.
Many nonrigid objects can be modeled as a system of components, each repre-
sented by a point scatterer with reflectivity G(α), where the angle α accounts for the
orientation of the component, or the angle between the radar antenna broadside

-fD f=0 fD
(a)

p1
fD

-fD
p2
t
(b)
Figure 7.18  (a) Spectrum of the full radar return from the simplified rotating propeller model.
(b) Spectrogram of the full radar return, showing the time-dependent nature of the micro-Doppler
frequencies.
260 Microwave and Millimeter-Wave Remote Sensing for Security Applications

direction and the local normal vector of the component. The human body, for
example, can be represented by a system of ellipsoids of various angle-dependent
reflectivity [19]. For an object composed of N scatterers, each with different reflec-
tivity Gn, a model for the total radar return signal can be given in terms of the range
and reflectivity of each scatterer:

N é 2 ù
- j 2π fc êt + rn (t )ú
sr (t) = å An G n (α n )e ë c û (7.104)
n =1

where the amplitude A accounts for the transmitted signal power and the propaga-
tion loss. If R0 is much greater than the maximum dimension of the object (R0 >> rp),
the amplitude incident on each scatterer is approximately the same, and thus the
total return signal can be given by

N é 2 ù
- j 2π fc êt + rn (t )ú
sr (t) = A å G n (α n )e ë c û (7.105)
n =1

7.4.3  Human Micro-Doppler Signature


The micro-Doppler signature of a moving human is generated by the micro-mo-
tions of various parts of the body as it moves. Contributions to the backscattered
radar signal can be generally divided into three primary groups due to the move-
ments of the torso and head, the arms, and the legs. Contributions from second-
ary groups such as the hips, shoulders, forearms, and so on can be considered parts
of one of the primary groups. The movements of the primary groups are not in-
dependent, since the groups are physically connected; however, the movement
of one primary group is not necessarily dictated by that of another. For instance,
in a normal walking stride, the arms could be swinging in a normal fashion, or
they may be stationary with respect to the torso, as is the case of someone hold-
ing an object with both arms. The torso induces the bulk Doppler shift, which can
be considered constant; however, there are small oscillatory motions due to rota-
tion and translation incurred while walking. The arms and legs induce the micro-
Doppler frequencies, which for a normal walking or running stride are periodic.
Models of human movement are generally derived from measurements of moving
humans, often recorded through video measurements. A detailed kinematic model
of walking human motion was presented in [20] based on a large set of experimental
data. Chen developed simulation code based on this model, which provides simulated
radar returns based on a human walking at various velocities by emulating body parts
as ellipsoids [19]. In this section, simulations based on this code are used to analyze
the time-frequency signature of a walking human in detail.
The analysis of human micro-Doppler has been gaining interest in the last de-
cade, and a number of publications have presented measured and simulated time-
frequency micro-Doppler data of moving humans [18, 21–32]. The signature is in
general characterized by the periodicity of the moving limbs, as seen in Figure 1.10,
which shows the spectrogram of a human walking toward the radar. Measured
7.4  Micro-Doppler 261

data has also been published on the micro-Doppler signatures of moving vehicles
and animals such as deer, goats, dogs, and birds [25]. Additionally, micro-Doppler
analysis is being applied to biometric measurements through the detection of respi-
ration and heartbeat monitoring for various applications such as search and rescue,
through-wall intrusion detection, and medical monitoring [33–38].
Characterization of the micro-Doppler signatures of moving humans focuses
primarily on the periodicity of the arm swing and leg swing, as well as the micro-
Doppler frequency energy. The primary goals of the analysis of micro-Doppler
signatures of moving humans in security applications include discrimination and
classification of humans and nonhumans, such as animals or vehicles, and clas-
sification of human activity for determination of intent. Other applications include
biomedical analysis for physical therapy, sports medicine, and other medical related
applications. The ability to measure the signature of noncooperative humans is the

1500 0

1000 −10
fD (Hz)

500 −20

0 −30

−500 −40
0 0.5 1 1.5
t (s)
(a)

1500 0

1000 −10
fD (Hz)

500 −20

0 −30

−500
0 0.5 1 1.5 2 2.5
t (s)
(b)
Figure 7.19  (a) Simulated time-frequency signature of a human walking toward a 30-GHz pulse
radar at a velocity of 1.5 m·s–1. (b) Simulation of a human walking with a radial velocity of 0.5 m·s–1.
Simulations were performed using modified code from [19].
262 Microwave and Millimeter-Wave Remote Sensing for Security Applications

basis behind micro-Doppler for gate analysis; that is, the signature can be measured
from a remote distance without additional measurement equipment on the human.
This is clearly important in security sensing applications such as intruder detection
where the person’s signature would have to measured at a standoff distance. Video
analysis of noncooperative moving humans can also be accomplished, although it
requires image processing in order to derive a gait signature; the radar, in contrast,
provides a direct measure of the signature.
Discrimination of humans and nonhumans is directly accomplished if the
nonhuman objects of interest do not include micro-motions or, therefore, micro-
Doppler, such as vehicles [39]. In such cases, the inclusion of frequency sidebands
is a clear indication that the object is a human. However, the human does not nec-
essarily produce micro-Doppler at every instance in time, as seen in Figure 1.10.
Thus, measurements over a range of time must be performed to identify the object
as a human. Discrimination of humans and animals is a more difficult task due to
the presence of micro-Doppler frequencies in the signatures of both. Additionally,
quadrupedal leg swing can produce maximum Doppler shifts of approximately the
same frequency as that produced by human bipedal motion. Classification there-
fore is generally performed on the periodicity of micro-Doppler signatures and the
micro-Doppler energy [25, 30].
To demonstrate aspects of the human micro-Doppler signature, simulations of
a walking human will be presented in the following. The simulations were gener-
ated using a modified version of code by Chen [19] based on the model developed
by Boulic et al. [20]. A simulated micro-Doppler of a walking human is shown in
Figure 7.19(a), where the human is walking with radial velocity of 1.5 m·s–1 toward
a 30-GHz radar. The oscillatory patterns in the micro-Doppler frequencies due the
arm and leg swings are prominent, producing the largest Doppler frequency shifts.
The bulk Doppler shift due to the torso is present and centered at approximately
300 Hz; however, the bulk Doppler frequency shows some oscillation and a maxi-
mum frequency spread of 200 Hz. This oscillation is due to the secondary micro-

1500 0

1000 −10
fD (Hz)

500 −20

0 −30

−500 −40
0 0.5 1 1.5
t (s)
Figure 7.20  Simulated signature of a human walking with no secondary micro-motions toward a
30-GHz radar with a velocity of 1.5 m·s–1; only arm and leg swing and bulk torso radial velocity are
simulated. The primary difference is the lack of oscillation in the torso signature.
7.4  Micro-Doppler 263

motions produced by rotation of the hips, radial oscillation of the torso due to the
planting of the feet, and so on. The legs swing produces a maximum Doppler shift
of 1550 Hz, while the arm swing produces a maximum shift of 850 Hz. The Dop-
pler shift in the negative direction is lower in maximum frequency since the legs do
not swing backwards as the arms do. Figure 7.19(b) shows the full signature of the
human walking with a radial velocity of 0.5 m·s–1, demonstrating the reduced Dop-
pler frequency spread due to the slower micro-motions of the arms and legs.
In Figure 7.20, the secondary motions of the torso (rotation, oscillation due
to planting of the feet, and so on) have been removed from the signature of the
human walking with a radial velocity of 1.5 m·s–1; the torso and head move in
a smooth bulk trajectory, with the arms and legs swinging as normal. The bulk
Doppler frequency no longer displays the oscillations and is centered at 300 Hz,
which is the frequency calculated using (7.12) with fc = 30 GHz and v = 1.5 m·s–1.

1500
−10
1000 Lower arms
Upper arms −20
500
fD (Hz)

−30
0

−40
−500

−1000 −50
0 0.5 1 1.5
t (s)
(a)
1500
Feet Lower legs Upper legs
−10
1000

−20
500
fD (Hz)

−30
0

−500 −40

Feet Lower legs Upper legs


−1000
0 0.5 1 1.5
t (s)
(b)
Figure 7.21  (a) Arm swing signature. (b) Leg swing signature. The legs produce many of the dis-
tinctive features of the overall walking human signature.
264 Microwave and Millimeter-Wave Remote Sensing for Security Applications

In Figure 7.21, the full signature is separated into the micro-Doppler signatures of
the arm swing and leg swing in isolation. The arm swing signature is roughly equal
in the positive and negative frequencies due to the smooth oscillatory nature of the
arm swing. The leg swing, in contrast, displays asymmetric positive and negative
frequency signatures, due to the plating of the feet. In Figure 7.22, the walking hu-
man is simulated without arm swing, emulating a person holding an object. The
differences between this plot and 7.20, where arm swing is included, are subtle but
detectable, and lack the oscillation between the leg swing micro-Doppler frequency
and the bulk Doppler frequency due to the torso. Characterization of minor differ-
ences such as these is the basis for activity classification [25, 30, 40].
The Doppler signature depends on the radial velocity of the bulk motion and
micro-motions of the object and thus is highly angle-dependent. The radial velocity
scales as the cosine of the angle α between the radial direction and the direction
of motion of the scattering point, and thus the micro-Doppler signature tends to
decrease significantly as α approaches 90°. Some micro-Doppler signature is still
present due to the lateral motions of the body during the walking motion. Figure
7.23(a) shows a simulation of a walking human at a velocity of 1.5 m·s–1 with the
30-GHz radar positioned at an angle of 45° from the direction of motion. The
Doppler frequency spread is reduced; however, the periodicity of the arm and leg
swing is still clear. In Figure 7.23(b), the radar is positioned at 60° degrees; the
signature is significantly affected by the reduction in radial velocity. Figure 7.23(c)
shows the signature when the radar is positioned at 90° relative to the motion of
the person. The periodic micro-Doppler signatures are essentially absent. While the
micro-­Doppler signature is present in most of the plots of Figure 7.23(b), it should
be noted that system noise and multipath effects are not included in these simula-
tions, both of which will tend to further degrade the signature. Some studies have
suggested that the maximum usable angle with which micro-Doppler can be reli-
ably detected is around 60° [25, 28].

1500 0

1000 −10
fD (Hz)

500 −20

0 −30

−500 −40
0 0.5 1 1.5
t (s)
Figure 7.22  Simulation of a human walking with a radial velocity of 0.5 m·s–1 without arm swing,
emulating a person holding an object. Minor but detectable differences are seen between this figure
and Figure 7.20.
7.4  Micro-Doppler 265

1500 0

1000 −10

fD (Hz)
500 −20

0 −30

−500 −40
0 0.5 1 1.5
t (s)
(a)

1500 0

1000 −10
fD (Hz)

500 −20

0 −30

−500
0 0.5 1 1.5
t (s)
(b)

1500 0

1000 −10
fD (Hz)

500 −20

0 −30

−500
0 0.5 1 1.5
t (s)
(c)
Figure 7.23  Signature of a human with the radar viewing at an angle of (a) 45°, (b) 60°, (c) 90°.
Various oscillations are reduced more than others due to occlusion behind the body. The reduction in
maximum frequency over the duration of the signature is due to the slightly changing aspect angle:
because the person was simulated walking in a straight line, the starting aspect angle increases over
the length of the measurement.
266 Microwave and Millimeter-Wave Remote Sensing for Security Applications

7.5  Continuous-Wave Radar

The previous sections have introduced the fundamental concepts of radar sys-
tems and radar measurements. A significant characteristic differentiating types
of radar systems is the type of waveform used to perform the measurement. The
waveform may be continuous or pulsed and may transmit a single frequency or
a range of frequencies. In this and the following sections, the most prominent
radar waveforms, those with the most applicability to security sensing, will be
discussed.
The characteristics of the transmitted waveform have a significant effect on the
ability of the radar to accurately measure the properties of the object. The effect of
the waveform is significant enough that radar systems are often categorized by the
type of waveform they transmit. The simplest waveform is a continuous monochro-
matic tone, or continuous-wave (CW) signal. As will be shown, CW radar systems
do not inherently have the capability to measure range, but provide very good
Doppler resolution; such radar systems are sometimes referred to as CW Doppler
radar systems. The transmitter of the CW radar essentially consists of an oscil-
lator generating the desired carrier frequency that is transmitted by the antenna.
A monostatic CW radar utilizes a circulator prior to the antenna, as depicted in
Figure 7.6, such that the radar simultaneously transmits the signal and receives the
reflected signal. The received signal is mixed with the signal from the oscillator to
downconvert the received signal to baseband. Any Doppler frequency shift on the
signal will therefore be converted to a baseband frequency that can be measured
with a lower sampling rate.
The signal transmitted by a CW radar is given by

st (t) = At e - j 2π fct (7.106)

If the object being illuminated is stationary, the received signal is

æ 2R ö
- j 2π fc ç t - ÷
è c ø
sr (t) = Ar e (7.107)

The baseband signal after mixing the received signal with the transmitted signal is

sb (t) = Ar e j 4π fc R c (7.108)

The phase difference between the transmitted and received signals is the phase
of the baseband signal, which is given by

4π Rfc 4πR
δφ= = (7.109)
c λ

Note that


δφ = 2 π when R= , n = 1, 2,3,... (7.110)
2
7.5  Continuous-Wave Radar 267

Thus, the range can only be unambiguously measured up to one half wavelength
from the radar, which is not useful in most situations. That is, the radar cannot de-
termine whether the object is at a range of λ, 10λ, or 1×109λ, for example; the range
is highly ambiguous. The maximum range beyond which the radar cannot unam-
biguously determine the true range is called the maximum unambiguous range, and
for a CW radar is given by

λ
Rmax = (7.111)
2

The CW radar therefore does not provide a reasonable measure of the range to the
object. In millimeter-wave security applications the object is most often at a stand-
off distance of at least a meter, which is many multiples of the wavelength.

7.5.1  Continuous-Wave Doppler


The strength of the CW radar is in its ability to accurately measure the Doppler
frequency of a moving object. To see this, consider the frequency domain repre-
sentation of the baseband signal, which is the Fourier transform of (7.108) and is
given by

St (f ) = At δ (f ) (7.112)

which a delta function at dc. If the object is moving with velocity v­r, the baseband
signal is given by

sb (t ) = Ar e j 4π fc (R− v r t ) c (7.113)

the Fourier transform of which is

Sb (f ) = Ar δ (f - fD ) (7.114)

where fD is the Doppler frequency shift (7.12). Figure 7.24 shows the spectra of
the transmitted and received signals of a CW Doppler radar viewing a moving
object.
The frequency representation of the received signal is thus infinitely narrow,
and the radar can discern between any two objects that do not have identical radial
velocity. In practice, the radar does not have infinite integration time, and thus the
time domain signal will be truncated. For a finite integration time τ, the baseband
signal can be represented by

t  j 4π fc (R− v r t ) c
sb (t ) = Π   Ar e (7.115)
τ 
The frequency domain signal is therefore

Ar
Sb (f ) = sinc [τ (f - fD )] (7.116)
τ
268 Microwave and Millimeter-Wave Remote Sensing for Security Applications

st (t) t

(a)

St (f)

fc
(b)

Sr (f)

fc fc+fD
(c)

Sbb (f)

fD fc fc+fD
(d)
Figure 7.24  (a) Transmitted signal in the time domain. (b) Transmitted signal in the frequency
domain. (c) Received signal in the frequency domain, including a Doppler frequency shift. (d) Base-
band signal in the frequency domain, after downconversion by mixing with the transmitted carrier
fc, including a Doppler frequency shift.

where the Fourier transform


¥
ætö 1
ò P çè a ÷ø e
- j 2π ft (7.117)
dt = sinc(af )

a

has been used. The Rayleigh bandwidth of the baseband signal is

1
δf = (7.118)
τ

When two objects are present that have Doppler frequency shifts close in fre-
quency, the bandwidth of the signal determines whether the radar can correctly
distinguish the presence of two objects or incorrectly decide that only one is present.
7.5  Continuous-Wave Radar 269

The minimum separation in frequency is often taken to be the Rayleigh resolution,


which is the 4-dB bandwidth of the signal, which is that given in (7.118); there-
fore, the Doppler resolution of the CW radar is proportional to the inverse of the
integration time. CW radar can distinguish between two objects whose Doppler
frequency shift differs by no less than τ-1. Figure 7.25 shows the effect of integration

s(t) S(f)
τ
FT 1
τ

t f

(a)

δf ≈ τ-1

4 dB
Sr (f)

fD1 fD2

(b)

Sr (f)

fD1 fD2

(c)

Sr (f)

fD1 fD2

(d)
Figure 7.25  (a) Relationship between integration time and frequency bandwidth of a finite signal.
(b) Rayleigh resolution. (c) Two signals with different Doppler frequency shifts can be resolved with
infinite integration time. (d) Finite integration time increases the bandwidth of the signals.
270 Microwave and Millimeter-Wave Remote Sensing for Security Applications

time on the resolution of two objects with different Doppler shifts. The relationship
between the temporal duration of the signal and the frequency bandwidth (7.118)
is shown in Figure 7.25(a). The definition of resolution for two signals is shown in
Figure 7.25(b). In Figure 7.25(c), the frequency response of two objects is shown
with infinite integration time, where the frequency responses of the objects are in-
finitely narrow and therefore resolved. With a finite integration time, as in Figure
7.25(d), the frequency responses widen, causing the responses to overlap.
As an example, consider a 75-GHz radar designed for the analysis of human
micro-Doppler signatures. A typical walking velocity is on the order of 1.5 m·s–1;
using (7.12) yields a frequency shift of 750 Hz. The maximum leg velocity dur-
ing a walking cycle is approximately 6 m·s–1, yielding a frequency shift of 3 kHz,
while the maximum arm velocity is approximately 2.5 m·s–1, yielding a frequency
shift of 1.25 kHz. The difference between the maximum arm and leg velocities is
1.75 kHz. In order to resolve the responses from the arms and the legs, the frequency
resolution must therefore be less than the separation frequency; from (7.118) this
yields an integration time of τ > 570 μs. To differentiate the arm and torso, where
the frequency separation is 500 Hz, requires an integration time of τ > 2 ms. This
time is much less than the periodicity of human movement; thus, longer integration
times can be implemented to yield finer frequency resolutions.
For a scanning CW radar system, the Doppler resolution is affected by the dwell
time and the integration time. The dwell time of an antenna scanning in the azimuth
direction can be given in terms of the azimuth antenna beamwidth fBW and the
rotation rate of the sensor ω by

φBW
τd = (7.119)
ω

The resolution is inversely proportional to the amount of time the signal is in-
tegrated, given by δf = τ –1. If the dwell time τd < τ, the signal will be integrated over
multiple spatial points, which has the effect of spreading the spatial response from a
point over multiple spatial bins, resulting in coarser spatial resolution. On the other
hand, if τd > τ, the system is not utilizing the fully available signal since the point
is present in the beam over multiple integration times, resulting in coarser Doppler
resolution. Thus, the optimal integration time for a scanning CW system is

τ = τ d (7.120)

which provides the finest Doppler resolution available without a negative impact
on spatial resolution. This relationship is the same as that derived for a scanning
radiometer (6.199). The Doppler resolution can thus be given by

1 ω
δf = = (7.121)
τ d φBW

Note that for a system where τd > τ, the spatial resolution is decreased; however, the
Doppler resolution does not decrease since the object producing the Doppler return
is only within the antenna beam for the dwell time.
7.5  Continuous-Wave Radar 271

7.5.2  Frequency-Modulated CW
As discussed earlier, the unambiguous range of the CW radar is too short com-
pared to the operational situations encountered in security sensing to provide any
useful range information. To overcome this limitation, a time-varying modula-
tion must be imparted on the signal in some way. The modulation on the signal
must be applied at a period no less than the required round-trip delay time for
the specified application. For example, a system screening for contraband may be
required to operate at a distance of 5 m, for which the round-trip time delay is
33 ns. A modulation frequency of (33 ns)–1 = 30 MHz or lower is thus necessary
to achieve a 5m unambiguous range. The time variance can be implemented by
either amplitude or phase modulation of the signal. Pulse radar systems modulate
the amplitude of the signal to generate a short temporal pulse; the time delay
between transmission and reception can then be measured. However, the range
resolution is proportional to the pulse length; thus, shorter pulses are needed for
finer range resolution. The integration time of the pulse is thereby shorter, reduc-
ing the Doppler resolution.
A method to combine the fine Doppler resolution of the CW radar with an
increased unambiguous range is to continuously modulate the frequency of the
CW signal; this is referred to as frequency-modulated CW (FMCW). For a linear
frequency modulation (LFM), the frequency of the transmitted signal is

f (t) = fc + ηt (7.122)

where η Hz·s–1 is the slope of the modulation. The LFM waveform is also called a
chirp waveform due to the spectral resemblance of the signal to the acoustic spec-
trum of the chirp of a bird. The transmitted signal is therefore

st (t ) = At e − j 2π (fc +ηt )t (7.123)

and the received signal is

st (t) = Ar e - j 2π [fc +η(t -td )](t -td ) (7.124)

where td is the delay time. After heterodyning the received signal and the transmis-
sion signal, the baseband signal is

j 2π (fc td +ηtd2 +ηtd t ) j 2π (fc td +ηtd2 ) j 2πηtd t


st (t) = Ar e = Ar e e (7.125)

The phase of the first exponential term of (7.125) is a constant and thus repre-
sents a constant phase shift of the return signal. It can therefore be included in the
complex amplitude term, yielding

st (t) = Ar¢ e j 2π fr t (7.126)

where

2R
fr = ηtd = η (7.127)
c
272 Microwave and Millimeter-Wave Remote Sensing for Security Applications

st sr

τd
f
fr

t
Figure 7.26  The difference between the frequencies of the transmit and receive signal frequencies
is determined by the slope of the modulation and the time delay.

is the difference in frequency between the transmitted signal and the received signal
and is directly proportional to the range. The measurement of the frequency differ-
ence is shown in the diagram of Figure 7.26.
In (7.127) the modulation is infinite, and thus the unambiguous range is like-
wise infinite. In practical systems, the frequency modulation can only be imple-
mented across a finite bandwidth, and the modulation slope is given by

η = fm Df (7.128)

where fm = Tm–1 is the modulation frequency, determined by the temporal length


Tm of the modulation waveform, and Df is the modulation bandwidth, or the fre-
quency range over which the signal is swept. Triangular modulation, shown in
Figure 7.27, is often used in practice, as is saw-tooth modulation. The range is thus
found through (7.127) and (7.128), yielding

cfr
R= (7.129)
2 fm Df

The maximum difference frequency that can be encountered is Df, and thus the
maximum unambiguous range is

c
Rmax = (7.130)
2 fm

f
∆f
τd

st fr
sr
fc
t
f
fr

t
Figure 7.27  FMCW using triangular modulation.
7.5  Continuous-Wave Radar 273

This can similarly be seen by noting that the period of the modulation can be
written

2Rmax 1
Tm = = (7.131)
c fm

The unambiguous range for FMCW radar is thus dependent on the frequency of the
modulation placed on the signal.
If the object is moving, the Doppler shift sums with the difference frequency.
The baseband signal is then

sb (t) = Ar¢ e j 2π (fr + fD )t (7.132)

The measured difference frequency thus differs depending on whether the modula-
tion is on the upslope or the downslope for the triangle. Figure 7.28 shows the fre-
quencies of the transmit and receive signals. The upslope and downslope frequency
differences are

fu = fr - fD (7.133)

f d = f r + f D (7.134)

The frequency difference proportional to the range can be determined by averaging


the two frequency differences by

1
fr = (fu + fd )
2 (7.135)

and the Doppler frequency can likewise be calculated by

1
fD = (fd - fu ) (7.136)
2

f
∆f + fD
∆f

st sr

t
f
fr + fD

fr - fD
t
Figure 7.28  Doppler frequency shifts due to moving objects add to the difference frequency.
274 Microwave and Millimeter-Wave Remote Sensing for Security Applications

7.5.3  Multifrequency CW
The range can also be measured with CW radar by transmitting more than one
frequency. Multifrequency CW (MFCW) radar systems increase the unambiguous
range through the difference in frequency between two or more unmodulated CW
signals. Figure 7.29 shows the diagram of a general two-frequency system, trans-
mitting carrier frequencies f1 and f2 where f1 < f2. The received signals are

sr,1(t) = Ar,1e - j 2π f1(t -td ) (7.137)

sr,2 (t) = Ar,2 e - j 2π f2 (t -td ) (7.138)

After heterodyning, the baseband signals are given by

sb,1(t) = Ar,1e j 2π f1td (7.139)

sb,2 (t) = Ar,2 e j 2π f2td (7.140)

The phase difference between the two baseband signals is therefore

4π Rδ f
δφ = 2π td δ f = (7.141)
c

where

δ f = (f2 - f1) (7.142)

is the difference between the two frequencies. The range is therefore

cδ φ
R= (7.143)
4πδ f

The maximum unambiguous range is found when df = 2p, which gives

c
Rmax = (7.144)
2δ f

sLO1(t)
LO1
st(t)
Σ
HPA
LO2
sLO2(t)

sbb1(t) sr(t)

sbb2(t) LNA
Figure 7.29  MFCW radar.
7.5  Continuous-Wave Radar 275

The maximum unambiguous range for MFCW radar is identical in form to that
of FMCW radar, however with the FMCW modulation frequency replaced with the
MFCW separation frequency.

7.5.4  Moving Target Indication Radar


Radar systems that are used to detect moving objects are referred to as moving
target indication (MTI) radar systems. Such radar systems discriminate between
the static background and moving objects by measuring the Doppler frequency
shift. The background, called clutter in radar terminology, is primarily stationary
and thus produces no large Doppler shift; however, it does have a bandwidth equal
to the inverse of the integration time given by (7.118). Additionally, the clutter fre-
quency spread may be wider due to small motions of objects such as leaves or grass.
The spectrum of a return signal with a moving object and clutter is shown in Figure
7.30. The clutter return is typically much stronger than that of moving objects due
to the wide spatial extent of the background. To ease detection, the clutter signal
can be filtered out using either a high-pass filter or a band of Doppler filters, shown
in Figure 7.31.
The clutter return becomes a more significant problem when the radar is placed
on a moving platform, such as a vehicle patrolling an area. For a platform moving
with velocity vp, the background motion relative to the broadside direction of the
radar antenna is

vb = v p cos θ (7.145)

where θ is the angle between the platform direction and the antenna direction. As a
function of angle, the clutter return signal is thus sinusoidal, with center frequency

2vb fc 2v p fc
fb = = cos θ (7.146)
c c
Figure 7.32 shows a diagram of the radar measurement and the resulting clutter
frequency as a function of angle.
Due to the finite beamwidth of the antennas, the radar will detect the clutter
within a range of frequencies due to the background moving with different relative
velocities within the beam. For an antenna with beamwidth θBW pointing in the
direction θ = θr, the range of frequencies seen by the radar is bounded by

Clutter

Moving
Object

0 fD

Figure 7.30  Frequency responses of clutter centered at dc and a moving object centered at fD.
276 Microwave and Millimeter-Wave Remote Sensing for Security Applications

Clutter Filter passband Clutter Filter passbands

f f
(a) (b)
Figure 7.31  Clutter mitigation filters: (a) High-pass filter. (b) Doppler filter bank.

2v p fc æ θ ö
fb = cos ç θr ± BW ÷ (7.147)
c è 2 ø

Because the return signal within the beam is a superposition of returns from
discrete objects and scatterers at different ranges, the return signals will be uncorre-
lated and will sum incoherently. The Doppler bandwidth of the clutter, or the clutter
frequency spread, is therefore the difference between the maximum and minimum
Doppler frequencies detected by the radar within the antenna beam and is given by

δ fb = max fb - min fb (7.148)

within the angles

θBW θ
θr - £ θ £ θ r + BW (7.149)
2 2

When the radar is facing in the direction of the platform motion, θr = 0°, the
maximum and minimum Doppler frequencies of the background are

2v p fc 2v p fc æθ ö
max fb = , min fb = cos ç BW ÷ (7.150)
c c è 2 ø

Moving
vp
platform
θ

Rotator Radar

(a)

fb

-fb
-π/2 0 π/2 π 3π/2 2π 5π/2 3π
θ
(b)
Figure 7.32  (a) Overhead view of a radar system on a moving platform. (b) Clutter frequency as a
function of antenna pointing angle.
7.5  Continuous-Wave Radar 277

and the Doppler frequency spread is

2v p fc é æ θBW ö ù
δ fbf = ê1 - cos çè 2 ÷ø ú (7.151)
c ë û

This is referred to as the forward-looking clutter bandwidth δ fbf . When the beam
is pointed perpendicular to the motion of the platform, θr ± 90°, the clutter band-
width is at its widest. The maximum and minimum Doppler frequencies are

2v p fc æ θ ö 2v p fc æθ ö
max fb = cos ç 90° + BW ÷ = sin ç BW ÷
c è 2 ø c è 2 ø
(7.152)
2v p fc æ θ ö 2v p fc æθ ö
min fb = cos ç 90° - BW ÷ = - sin ç BW ÷
c è 2 ø c è 2 ø

and the clutter bandwidth is


4v p fc æθ ö
δ fbs = sin ç BW ÷ (7.153)
c è 2 ø

This is referred to as the side-looking clutter bandwidth δ fbs. Figure 7.33 shows the
calculation of the clutter spread.
A typical method of dealing with the clutter bandwidth is to simply high-pass
filter the signal with a cutoff frequency at the maximum clutter frequency. This
simplifies the detection of objects that produce Doppler frequencies greater than
the clutter bandwidth. However, this method will eliminate responses from objects
moving slower than the platform velocity. Additionally, the clutter bandwidth will
tend to mask objects moving at velocities close to that of the platform. Both the
forward-looking clutter bandwidth (7.151) and side-looking clutter bandwidth
(7.153) are proportional to the platform velocity and the antenna beamwidth, and

max fb
δfbf
min fb