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 MillimeterWave
Remote Sensing
for Security Applications
Jeffrey A. Nanzer
artechhouse.com
Library of Congress CataloginginPublication Data
A catalog record for this book is available from the U.S. Library of Congress.
ISBN13: 9781608071722
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 MillimeterWave
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 MillimeterWave Remote Sensing
in Related Fields 5
1.3 The Microwave and MillimeterWave Spectrum 7
1.3.1 Frequency Designations 7
1.3.2 Propagation of Microwave and MillimeterWave 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 ThroughWall 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 TimeHarmonic 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
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 FarField 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
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 MillimeterWave 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
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 BandLimited 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 OneDimensional Scanning 223
6.6.3.2 TwoDimensional Scanning 225
References 226
chapter 7
Radar 229
7.1 Radar Fundamentals 230
7.1.1 Configurations and Measurements 231
Contents ix
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 TwoDimensional 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
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 TimeFrequency Domain 341
9.3 Interferometric Measurement of Moving Humans 344
9.3.1 NarrowBeamwidth Response to a Moving Human 344
9.3.2 WideBeamwidth Response to a Moving Human 346
References 349
Index 359
Preface
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 microDoppler. 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 securityrelated 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
Microwave and MillimeterWave Remote Sensing for Security Applications
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.
1.2.1 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 millimeterwave regions can be approximated by a linear
Microwave and MillimeterWave 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 highgain
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.
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 frequencysteered 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 beamsteering
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.
carriers. The result has been component and system technology with increased ca
pabilities and reduced cost, making millimeterwave remote security sensor systems
more realizable.
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 millimeterwave bands.
Figure 1.3 (a) 350GHz active imaging system with dualreflector beam steering. (b) Image of a
concealed mock explosive beneath clothing. (© IEEE 2009 [8].)
Figure 1.4 (a) 72–80GHz 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].)
Figure 1.5 Images of concealed objects taken from a 77GHz videorate passive radiometric imag
ing system. (© 2010 SPIE [30].)
Figure 1.6 A 22GHz videorate 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 MillimeterWave Remote Sensing for Security Applications
Figure 1.7 (a) A 90GHz 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 36GHz continuouswave scanningbeam 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 Dopplershifted 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 36GHz continuouswave 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 MillimeterWave Remote Sensing for Security Applications
3.5º, and the system had a Doppler frequency resolution of 16 Hz. Figure 1.9(a)
shows a 27.4GHz 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 27.4GHz 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 microDoppler signature of a walking human from a 2.4GHz radar. (c) Simulated microDoppler signature of a walk
17
ing human with a 12GHz radar. (© IEEE 2008 [53], courtesy of H. Ling.)
18 Microwave and MillimeterWave Remote Sensing for Security Applications
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 Ultrawideband throughwall 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
Figure 1.12 (a) Block diagram of a Kaband continuouswave 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 MillimeterWave Remote Sensing for Security Applications
with frequency, and thus most throughwall 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
highresolution imaging techniques. However, the bulk movement of people can be
detected, and microDoppler signatures can also be measured. Figure 1.11 shows
a measured crossrange and downrange plot of a 5GHz ultrawideband 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.
References
[1] “IEEE Standard Letter Designations for RadarFrequency Bands,” IEEE Std 5212002
(Revision of IEEE Std 5211984), 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
MillimeterWave 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, MillimeterWave Radar,” in Advances in Sensing
with Security Applications, J. Burnes, Ed., Dordrecht, the Netherlands: Springer, 2006.
[6] Appleby, R., and R. N. Anderton, “MillimeterWave and SubmillimeterWave 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., “MillimeterWave
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 MillimeterWave
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 WBand
MillimeterWave Sensor,” Microwave Symposium Digest (MTT), 2010 IEEE MTTS 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 MillimeterWave Imaging System for Concealed Objects Detection,” Infrared Millime
ter and Terahertz Waves (IRMMWTHz), 2010 35th International Conference on, 2010,
pp. 1–2.
[11] Ahmed, S. S., A. Schiessl, and L. P. Schmidt, “A Novel Fully Electronic Active RealTime
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 MillimeterWave Arrays,” Lightwave Technology, Journal of,
Vol. 17, 1999, pp. 418–425.
[13] Lettington, A. H., M. R. Yallop, and D. Dunn, “Review of SuperResolution Techniques
for Passive MillimeterWave 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 MillimeterWave Imaging for the Detection of Concealed Weapons,”
Air Force Research Laboratory Technical Report AFRLIFRSTR200537, 2005.
[16] Luthi, T., and C. Matzler, “Stereoscopic Passive MillimeterWave 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, LowCost, Passive MMW Security Scan
ner,” Proceedings of the SPIE, Orlando, FL, 2005, pp. 109–116.
[18] Nohmi, H., S. Ohnishi, and O. Kujubu, “Passive MillimeterWave Camera with Interfero
metric Processing,” Proceedings of the SPIE, Vol. 6211, 2006, pp. 621104–621108.
[19] Lovberg, J. A., C. Martin, and V. Kolinko, “VideoRate Passive MillimeterWave Imaging
Using Phased Arrays,” Microwave Symposium, 2007, IEEE/MTTS 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 TShaped MillimeterWave Imaging System,” Microwave Theory and Techniques, IEEE
Transactions on, Vol. 55, 2007, pp. 1246–1254.
[21] Nohmi, H., S. Ohnishi, and O. Kujubu, “Passive MillimeterWave Camera with Interfero
metric Processing,” Proceedings of the SPIE, Vol. 6548, 2007, p. 65480C8.
[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 MillimeterWave 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 MillimeterWave 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 MillimeterWave Imaging System Using Optical Upconversion,”
Proceedings of the SPIE, Vol. 7309, 2009, p. 730908.
[27] Wikner, D., and E. Grossman, “Demonstration of a Passive, LowNoise, MillimeterWave
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 MillimeterWave 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. AflakiBeni, et al., “Rapid
Rotary Scanner and Portable Coherent Wideband QBand Transceiver for HighResolution
22 Microwave and MillimeterWave Remote Sensing for Security Applications
[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 MicroDoppler 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 MicroDoppler 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
MicroDoppler 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 MicroDoppler
Signatures Using an Artificial Neural Network,” Antennas and Propagation Society Inter
national Symposium, 2008, APS 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
MicroDoppler Signatures,” MicroNanoelectronics, 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
MicroDoppler Signals from a ScanningBeam 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 MicroDoppler 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 &
MicroDoppler 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, “MicroDoppler Phenomenology of
Humans at UHF and KuBand for Biometric Characterization,” Proceedings of the SPIE,
Orlando, FL, 2009, pp. 73080X9.
[60] Ram, S. S., C. Christianson, Y. Kim, and H. Ling, “Simulation and Analysis of Human
MicroDopplers in ThroughWall 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, “MicroDoppler Radar Sig
natures of Human Activity,” in Proceedings of the SPIE, 2010, p. 78370L.
[62] Chen, V. C., The MicroDoppler 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 ChaosBased HighResolution Imaging
Technique and Its Application to ThroughtheWall Imaging,” Signal Processing Letters,
IEEE, Vol. 12, 2005, pp. 528–531.
[65] Ram, S. S., and H. Ling, “ThroughWall 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, “ThroughtheWall 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 MillimeterWave Remote Sensing for Security Applications
[67] ChiWei, W., and H. ZiYu, “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., “ThroughWall 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, “ThroughWall Sensing with Multifrequency
Microwave Radiometry: A ProofofConcept Demonstration,” IEEE Transactions on Geo
science and Remote Sensing, Vol. 47, 2009, pp. 1281–1288.
[71] GonzalezPartida, J. T., P. AlmoroxGonzalez, M. BurgosGarcia, B. P. DortaNaranjo,
and J. I. Alonso, “ThroughtheWall Surveillance with MillimeterWave 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
theWall Imaging,” IEEE Transactions on Image Processing, Vol. 18, 2009, pp. 1255–
1265.
[73] Hong, W., R. M. Narayanan, and Z. Zheng Ou, “ThroughWall 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 UltraWideband Radar System for ThroughtheWall
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, “ThreeDimensional ThroughWall 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 RealTime
ThroughWall Imaging Under Unknown Wall Characteristics,” IEEE Transactions on Geo
science and Remote Sensing, Vol. 48, 2010, pp. 423–431.
[77] ChiehPing, L., and R. M. Narayanan, “Ultrawideband Random Noise Radar Design for
ThroughWall 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 ThroughWall Imaging Based
on Compressive Sensing,” IEEE Transactions on Geoscience and Remote Sensing, Vol. 48,
2010, pp. 1408–1415.
[79] Yanming, X., L. Jenshan, O. BoricLubecke, and V. M. Lubecke, “A KaBand 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 LowPower KaBand Non
Contact Heartbeat Detector Measured from Four Sides of a Human Body,” Microwave
Symposium Digest, 2006, IEEE MTTS International, 2006, pp. 1576–1579.
[81] Jenshan, L., and L. Changzhi, “Wireless NonContact Detection of Heartbeat and Respira
tion Using LowPower Microwave Radar Sensor,” Microwave Conference, 2007, APMC
2007, AsiaPacific, 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 MillimeterWave/Terahertz Radars,” Proceedings of the SPIE,
2008, pp. 71170I.
[83] Droitcour, A. D., O. BoricLubecke, and G. T. A. Kovacs, “SignaltoNoise 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. HostMadsen, and O. BoricLubecke, “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
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 millimeterwave electromagnetic radiation, the energy is transported via
timevarying 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 millimeterwave 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 millimeterwave 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].
27
28 Microwave and MillimeterWave 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.
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)
¶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 AmpereMaxwell 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
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 MillimeterWave 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.
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
µ0 = 4π ´ 107 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 TimeHarmonic Electromagnetic Fields
31
¶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.
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 timevarying or timeharmonic fields, as these are the types of fields
that propagate through a medium. In general, a timeharmonic field is given by
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 timederivative of (2.22) is
¶
χ (r,t ) = jωχ (r)e jωt = jωχ (r, t) (2.23)
¶t
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 MillimeterWave Remote Sensing for Security Applications
Ñ ´ E =  jωµH (2.28)
Ñ ´ H = jωε E (2.29)
Ñ × E = 0 (2.30)
Ñ × H = 0 (2.31)
Ñ ´ (Ñ ´ E ) =  jωµÑ ´ H (2.32)
Substituting (2.29) for the righthand 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 lefthand side is zero due to (2.30), and thus
Ñ 2 E + ω 2µε E = 0 (2.35)
Ñ2 H + ω 2µε H = 0 (2.36)
k = ω µε (2.37)
Ñ 2 E + k2 E = 0 (2.38)
Ñ 2 H + k2 H = 0 (2.39)
2.2 TimeHarmonic 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 Ex
Ñ2 E = (2.42)
¶z 2
and the scalar Helmholtz equation becomes
¶ 2 Ex
+ k2 Ex = 0 (2.43)
¶z 2
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
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 MillimeterWave 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
2π
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
2π
k= (2.47)
λ
or
kλ = 2π (2.49)
which is the same as (2.47).
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 TimeHarmonic Electromagnetic Fields
35
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
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
1 æ ¶H y ¶H x ö
Ez = ç  ÷ = 0 (2.60)
jω ε è ¶x ¶y ø
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 MillimeterWave 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)
k k
H = xˆ Ey + yˆ Ex (2.63)
ωµ ωµ
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
E (r ) = E0 e ± jk×r (2.68)
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 TimeHarmonic Electromagnetic Fields
37
E = ωµk ´ H (2.72)
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 MillimeterWave Remote Sensing for Security Applications
yields
¶u
+ Ñ × (E ´ H) = E × J (2.81)
¶t
S = E ´ H (2.83)
is the Poynting vector, which represents the power carried by the electromag
netic wave. For timeharmonic 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.
As a timeharmonic 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
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 MillimeterWave 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 xaxis. 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.
E = (xˆ Ex + yˆ Ey ) (2.89)
where
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)
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)
Using the squares of (2.94) and (2.95), the last term on the righthand 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
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 xdirection.
y
ωt = π/2
ωt = 3π/4 ωt = π/4
ωt = π ωt = 0
x
ωt = 5π/4 ωt = 7π/4
ωt = 3π/2
References
[1] Jackson, J. D., Classical Electrodynamics, 3rd ed., Hoboken, NJ: John Wiley & Sons,
1999.
[2] Stratton, J. A., Electromagnetic Theory, New York, NY: McGrawHill, 1941.
[3] Harrington, R. F., TimeHarmonic Electromagnetic Fields, New York, NY: McGrawHill,
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
43
44 Microwave and MillimeterWave Remote Sensing for Security Applications
Radar
(a)
Radiometer
(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.
In a general medium, the timeharmonic 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)
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 xaxis. 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
kr = ω µε (3.6)
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)
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 MillimeterWave Remote Sensing for Security Applications
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 lowloss medium.
The phase velocity of the wave can be given in terms of the phase constant
through
ω
v= (3.20)
β
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
1 1
α » ω µε tan δ = ωµσ (3.24)
2 2
1 1
β » ω µε tanδ = ωµσ (3.25)
2 2
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.
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 MillimeterWave Remote Sensing for Security Applications
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
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
µ
η» (3.37)
ε
Ñ ´ H = (σ + jωε )E (3.39)
3.1 Plane Wave Propagation in Unbounded Media 49
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)
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 frequencydependent 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 MillimeterWave Remote Sensing for Security Applications
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 KramersKronig 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
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 MillimeterWave 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.
The fields reflected by the boundary back into medium 1 are given by
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
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
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 MillimeterWave 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)
x
θr
Hr
z
Er
Et
θr θt
θi Ht
θt
Ei
Hi
θi
Media 1 Media 2
ε1, µ1 ε2, µ2
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
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 = sin1 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)
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 MillimeterWave 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
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
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
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 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
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.
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 × 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 MillimeterWave Remote Sensing for Security Applications
µ2 µ1 ε2  µ2 ε1
θB^ = sin1 × (3.110)
ε1 µ12  µ22
ε 2 µ2 ε1  µ1 ε2
θB = sin1 × (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 = tan1 (3.112)
ε1
æk ö µ2 ε2
θc = sin1 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)
(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
n1 n n+1
Tn Tn+1
Rn Rn+1
0 1 n N1 N
a1 a2 an+1 aN aN+1
b1 b2 ... bn+1 ... bN
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 semiinfinite
layer N. Because the layer is semiinfinite, 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 semiinfinite 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
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 walltype materials.
In many remote sensing applications in security—whether remote detection of
intruders, throughwall 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.
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 millimeterwave 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 millimeterwave 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. Longrange 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 shortrange
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 shortrange 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
101
102
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
101 H 2O
O2 H 2O H 2O
O2
102
H 2O
103
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 MillimeterWave Remote Sensing for Security Applications
103
102
101
α (dB/km)
100
90%
101 70%
50%
102 30%
10%
103
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
101
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)
30 mm/hr
−10
40 mm/hr
50 mm/hr
−15
−20
0 100 200 300 400
f (GHz)
(e)
Figure 3.12 (continued )
2 +ε¢
ξ= (3.130)
ε ¢¢
ε 0  5.48 1.97
ε ¢(f ) = 2
+ + 3.51 (3.131)
1 + (f f1) 1 + (f f2 )2
where
and T is the temperature in Kelvin. The first and second relaxation frequencies in
(3.131) and (3.132) are given by
r1.6 r
α = 0.00349 4
+ 0.00224 (3.136)
λ λ
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
101
10 mm/hr
102 5 mm/hr
2.5 mm/hr
1 mm/hr
103
20 40 60 80 100
f (GHz)
Figure 3.14 Theoretical attenuation of dry snow calculated from (3.136).
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].)
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
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.
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 Millimeterwave 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
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 C4 (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 MillimeterWave Remote Sensing for Security Applications
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
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 topmost 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 MillimeterWave 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 MillimeterWave 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 airskin
boundary and the skinfat 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
0.8
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 millimeterwave 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 skinfat boundary. At high mi
crowave and millimeterwave 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: AddisonWesley, 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 MillimeterWave 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, MillimeterWave 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 MillimeterWave Remote Sensing for Security Applications
[11] “Attenuation due to Clouds and Fog,” International Telecommunication Union Radiocom
munication Assembly Recommendation ITUR P.8403, 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, “ThreeDimensional ThroughWall
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 MidInfrared 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/TerahertzWave 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. JeiWon, 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 MTTS International, Vol. 1 2003,
pp. 399–402.
[32] Chahat, N., M. Zhadobov, R. Augustine, and R. Sauleau, “Human Skin Permittivity Mod
els for MillimetreWave 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
85
86 Microwave and MillimeterWave Remote Sensing for Security Applications
of a long linear dipole antenna. Discussion of linear wire antennas is necessary for
the understanding of antenna systems; however, in microwave and millimeterwave
remote sensing, aperture antennas are most often used because greater directivity can
be achieved using a twodimensional 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.
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
Ñ × H = 0 (4.5)
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)
The lefthand side of (4.11) is the same as that of the homogeneous vector
Helmholtz equation that was solved in Chapter 2; the righthand 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 wellknown physi
cist H. A. Lorentz.
88 Microwave and MillimeterWave 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).
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
 jk r  r¢
e e  jkr
A(r) = zˆ ò J(r¢) dV ¢ = zˆ Idl (4.27)
V¢
4π r  r¢ 4π r
90 Microwave and MillimeterWave 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
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.
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 millimeterwave 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)
z p1
rr`
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
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)
V¢
which yields
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 = �J x xˆ + �J y yˆ + �J z zˆ (4.46)
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
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
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
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:
é 1 1 ù
H = jkg(r) êθˆ(aφ  fθ )  φˆ(a θ + fφ)ú (4.57)
ë η η û
Then, from (4.47) and (4.48), the components of the directional function are
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.
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
òò Sds = òò Sr
2
Pr = sinθ dθ d φ (4.68)
4π 4π
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
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).
θ HPBW = 2θ ¢ (4.71)
where
Amax
A(θ ¢,0) = (4.72)
2
or
1
AN (θ ¢,0) = (4.73)
2
98 Microwave and MillimeterWave Remote Sensing for Security Applications
The angle between the first two nulls adjacent to the main beam is called the
nulltonull beamwidth. For a symmetric beam, on the f = 0 plane, the nulltonull
beamwidth in the q direction qNNBW is found by
θ NNBW = 2θ ¢¢ (4.74)
where
A(θ ¢¢,0) = 0 (4.75)
λ λ
θ HPBW » 0.886 = 50.8° (4.76)
d d
where d is the length of the antenna. The nulltonull 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 halfpower beamwidth is given by
λ
θ HPBW » 68.8° (4.78)
d
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 halfpower beamwidth and nulltonull 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.
WA = òò AN dW (4.82)
4π
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 MillimeterWave Remote Sensing for Security Applications
4π A(θ , φ) 4π A(θ, φ)
D(θ , φ) = = (4.86)
òò A(θ, φ)dW Pr
4π
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
W A » θBW φ BW (4.90)
where qBW and fBW are the halfpower beamwidths of the main beam in the q and f
dimensions. The maximum directivity can then be approximated by
4π
Dmax » (4.91)
θBW φ BW
41, 253
Dmax » (4.92)
θBW φ BW
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)
4π
4π
Gmax = ε r Dmax = ε r (4.102)
WA
102 Microwave and MillimeterWave Remote Sensing for Security Applications
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)
2η
and the power at a distance r
2
Er 2
P2 = r W A (4.106)
2η
The electric field at r is related to that in the aperture by [7]
Eap Ae
Er = (4.107)
rλ
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
4π
Dmax = Ae (4.110)
λ2
4.2 Antenna Parameters
103
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 MillimeterWave Remote Sensing for Security Applications
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].
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.
3 2
D(θ , φ) = sin θ (4.124)
2
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
dz
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 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 MillimeterWave 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.
dd dd
Perfect Perfect
electric magnetic
J J
conductor conductor
J J
Jm Jm
Jm Jm
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.
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 MillimeterWave Remote Sensing for Security Applications
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 openended 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 openended 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
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
Ea
PEC PEC
y
a
b
x
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
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
4λ
3λ
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) Eplane and Hplane patterns of a rectangular
aperture with a = 4l, b = 3l.
114 Microwave and MillimeterWave Remote Sensing for Security Applications
plane and the yz plane, as they align with the geometry of the aperture. The yz
plane is oriented along the direction of the electric field, and thus the antenna pat
tern on this plane is called the Eplane pattern. Similarly, the pattern along the xz
plane aligns with the magnetic field and is called the Hplane pattern.
The Eplane 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 Hplane 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 Eplane and Hplane antenna patterns are plotted in Figure 4.14(b).
The halfpower beamwidth of the Eplane 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 halfpower 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 nulltonull beamwidth of the Eplane 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 nulltonull beamwidth is
æ λö
θ NNBW = 2sin 1 ç ÷ (4.158)
è bø
4.4 Aperture Antennas
115
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 firstorder 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 Eplane 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 Hplane 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 Eplane and Hplane antenna patterns are plotted in Figure 4.15(c) for a = 3l.
The halfpower beamwidth of the Eplane pattern is found in terms of the angle
where (4.165) is equal to 0.5, or
116 Microwave and MillimeterWave Remote Sensing for Security Applications
PEC a PEC
y
Ea
x (a)
z
2λ
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 Hplane 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 halfpower 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 nulltonull 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
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
N 1
2
AF = å e jn ψ
N 1
n =
2
(4.178)
or
æ N +1 ö æ N 1ö 1 æ j 1 Nψ 1 ö
AF =
e
jç
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)
jψ
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
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 21 (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
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.
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 MillimeterWave 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 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.
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
Wavefront
3dcosθ
2dcosθ θ
dcosθ
d d d
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
halfpower 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
mλ
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
(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.
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
millimeterwave 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.
This section describes the radiation characteristics of a few of the most common
antennas used in microwave and millimeterwave 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.
and is designated by the direction of the field parallel to the direction of the flare as
either an Eplane or Hplane 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 halfpower beamwidth and nulltonull 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
a
E
a
(a)
(b)
(c)
(d)
Figure 4.22 Horn antenna: (a) Eplane, (b) Hplane, (c) pyramidal, (d) conical.
130 Microwave and MillimeterWave Remote Sensing for Security Applications
{
´ [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 MillimeterWave Antennas
131
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
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
L h
(a)
Antenna
Signal Microstrip Coupled feed
Antenna
Aperturecoupled feed
Signal
Microstrip
Antenna
Coaxial feed
Coaxial
Signal
(b)
Figure 4.24 (a) Microstrip antenna. (b) Feeding methods.
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 MillimeterWave 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.
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. Centerfed 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 millimeterwave frequencies. Additionally, passive millimeterwave systems suffer
increased system noise if the initial amplifier is not close to the receiving antenna,
while millimeterwave 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) centerfed reflector, and (b) offsetfed reflector.
Double reflector antenna systems: (c) Cassegrain, and (d) Gregorian.
136 Microwave and MillimeterWave Remote Sensing for Security Applications
increased loss and noise to transfer the signals between the receiver and the feed ele
ment in a centerfed configuration. Offsetfed 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 centerfed 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.
(a) (b)
Figure 4.27 (a) Lens antenna. (b) Lens antenna with planar outer surface.
4.6 Common Microwave and MillimeterWave 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 1451983, 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: McGrawHill, 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: McGrawHill, 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., TimeHarmonic Electromagnetic Fields, New York: McGrawHill,
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: McGrawHill, 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: McGrawHill, 2007.
[19] Gilbert, R. A., “Waveguide Slot Antenna Arrays,” in Antenna Engineering Handbook,
J. L. Volakis, Ed., 4th ed., New York: McGrawHill, 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: McGrawHill, 1969.
[21] Jackson, D. R., “Microstrip Antennas,” in Antenna Engineering Handbook, J. L. Volakis,
Ed., 4th ed., New York: McGrawHill, 2007.
138 Microwave and MillimeterWave 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: McGrawHill, 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] RahmatSamii, Y., “Reflector Antennas,” in Antenna Engineering Handbook, J. L. Volakis,
Ed., 4th ed., New York: McGrawHill, 2007.
[25] Cooley, M. E., and D. Davis, “Reflector Antennas,” in Radar Handbook, M. I. Skolnik,
Ed., 4th ed., New York: McGrawHill, 2008.
[26] Brown, J., “Lens Antennas,” in Antenna Theory Part 2, R. E. Collin and F. J. Zucker, Eds.,
New York: McGrawHill, 1969.
[27] Bodnar, D. G., “Lens Antennas,” in Antenna Engineering Handbook, J. L. Volakis, Ed., 4th
ed., New York: McGrawHill, 2007.
Chapter 5
Receivers
139
140 Microwave and MillimeterWave Remote Sensing for Security Applications
Si(f) So(f)
H(f)
Receiver
Figure 5.1 General receiver modeled as a twoport 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 [16]; 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
millimeterwave 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 millimeterwave
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
highgain 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 lownoise amplifier may be required at the output of
the antenna to reduce the overall system noise.
The frontend 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 lownoise 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 nonlinear 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
ADC
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 squarelaw 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 lowpass 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 stateoftheart 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 millimeterwave 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
sI(t)
s(t)
90
o
sQ(t)
signal with both an inphase local oscillator signal and a local oscillator signal
in quadrature with the inphase 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 inphase and quadrature signals, given by
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 bandlimited 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
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
Figure 5.6 Voltage fluctuations over time in a resistor due to thermal agitation.
temperature and because of the random, noiselike 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.
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 MillimeterWave Remote Sensing for Security Applications
η
Tshot = T (5.15)
2
and thus
Ideal passpand
Actual passpand
f
Figure 5.7 Ideal and nonideal 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 nonideal system passband.
¥ 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 nonideal passband.
As an example, consider a singlepole RC lowpass 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
¥
1 æ f ö
= fc tan çè f ÷ø
c 0 (5.21)
π
= fc
2
» 1.57 fc
This states that for a singlepole 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 MillimeterWave Remote Sensing for Security Applications
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
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 millimeterwave 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
millimeterwave 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 RayleighJeans region which will be discussed in detail in Chapter 6; (5.12) is
the RayleighJeans 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
RayleighJeans form
−140
Nn (dBm)
−160
Plankian form
−180
−200 8 10 12 14
10 10 10 10
f (Hz)
Figure 5.9 RayleighJeans 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 RayleighJeans 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
−150
−180 1 2 3 4
10 10 10 10
T (K)
Figure 5.10 RayleighJeans form and Planck form of the thermal noise power in a conductor versus
temperature.
150 Microwave and MillimeterWave 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 RayleighJeans 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 RayleighJeans 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 RayleighJeans 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.
The purpose of a twoport 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 twoport 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 nonidealities 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.
Si , N i G So, No
Twoport
network
Figure 5.11 Generalized twoport 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 signaltonoise 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:
Si S
=F o (5.28)
Ni No
A nonideal twoport 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
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 twoport, the noise generated by the twoport, and the input noise power.
152 Microwave and MillimeterWave 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 twoport 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.
Nn = GkTE Df (5.36)
Note that TE is defined at the input terminals of the twoport and represents all
the noise sources within the twoport; 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.
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 twoport network, an equivalent input noise tem
perature can be defined which describes the output noise power of a noisefree at
tenuator with an input noise temperature TE:
1
Na = kTE D f (5.42)
L
154 Microwave and MillimeterWave 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.
æ T ö
No = kTE Df = G1G2k ç Ti + TE1 + E2 ÷ Df (5.46)
è G1 ø
(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
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
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 lowpass
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 MillimeterWave Remote Sensing for Security Applications
The equivalent noise temperature of the cascade is given by (5.49), and recalling
that G = 1/L results in
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
» 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 lownoise
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.
X dB FS
FSX
ADC DR DR
10log(fs/2)
The noise figure is then the normalized noise power minus the thermal noise in a 1
Hz bandwidth:
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
1
fco = (5.55)
2π C
The cutoff frequency of the sampleandhold 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 peakto
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
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 peaktopeak is less than the fullscale peak to peak, and is
given by
where the righthand 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 MillimeterWave 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
sampleandhold circuit [13]. Time jitter can be characterized in general by consid
ering an input voltage sinusoid
dvin (t)
= 2π fV0 cos(2π ft) (5.65)
dt
the rms voltage of which is
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.
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 nonlinearities 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
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.
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
where A is the signal amplitude and w0 = 2pf is the angular frequency, the output
voltage of the nonlinear device, neglecting terms above third order, is
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
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
X dB
OPXdB
IPXdB
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
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
where w = 2wf, the output voltage of the nonlinear device, neglecting terms above
third order, is given by substituting (5.83) into (5.70) which results in
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
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 MillimeterWave Remote Sensing for Security Applications
Amplitude
ω2ω1 ω1 ω2 2ω1 2ω2
2ω1ω2 2ω2ω1 ω2+ω1
Figure 5.21 Output spectrum of a nonlinear device showing the two fundamental signals and the
intermodulation products.
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  101 20
Thus, the third order intercept point of a nonlinear 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
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 MillimeterWave 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, feedforward, 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 highlinearity amplifiers. Com
mercially available microwave and millimeterwave amplifiers often employ these
techniques.
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.
DR
MDS
3 dB
IP1dB
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)
where the minimum detectable signal is defined to be 3 dB greater than the noise
floor:
If a specific signaltonoise 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:
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 straightforward 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
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.
2
SFDR 3 = (PIP3  No ) (5.100)
3
SFDR
IPIP3
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, HighLinearity 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: McGrawHill, 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: McGrawHill, 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 millimeterwave 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 millimeterwave 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 millimeterwave 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 millimeterwave 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 millimeterwave 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 millimeterwave 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 MillimeterWave 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
millimeterwave radiation transmits favorably through some wall materials, such
as drywall, and thus there has been interest in applying radiometry to throughwall
detection [15, 16]. Radiometry has also been applied to the detection of landmines.
Millimeterwave 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 millimeterwave
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 millimeterwave 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.
6.1.1 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 millimeterwave region the radiated power is
proportional to the physical temperature of the source for most sources of interest.
Millimeterwave 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 millimeterwave remote sensing terminology. In
optics, this quantity is called radiance.
6.1 Radiometry Fundamentals 175
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:
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
frequencydependent 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 MillimeterWave Remote Sensing for Security Applications
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
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)
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
and
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
and likewise
dAr
dPs = Bs dAs (6.15)
R2
and
dAs
dPr = Br dAr (6.16)
R2
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 MillimeterWave Remote Sensing for Security Applications
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
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
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.
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
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
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
Bf df = Bλ d λ (6.29)
and using
c
f = (6.30)
λ
and
df c
=  2 dλ (6.31)
dλ λ
2hc 2 1
Bλ = 5
× hc λkT (6.32)
λ e  1
105
6000 K
1010 (Sun)
B (W m2 Hz1 str1)
1000 K
1015 310 K
(human)
100 K
1020
3K
1025
1030 5
10 1010 1015 1020
f (Hz)
Figure 6.5 Radiation curves for a blackbody at various temperatures over frequency.
182 Microwave and MillimeterWave 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:
where fm is the frequency at the maximum point of the curve. This reduces to
hfm )
= 3 (1  e hfm kT
(6.34)
kT
or, equivalently,
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 m3 str1)
1000 K
100 310 K
(human)
1010 100 K
3K
1020
1030
1010 105 100
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 midinfrared 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 ´ 103
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
4¥
2h æ kT ö x3
B= ç
c2 è h ø
÷ ò e x  1 dx (6.40)
0
2π 4 k 4 4
B= T (6.41)
15c 2 h3
or
M = σ T 4 W × m 2 (6.43)
where
2π 5k4
σ= = 5.67 ´ 108 (6.44)
15c 2 h3
¥
x2 xn
ex = 1 + x + +…= å (6.45)
2! n =0
n!
hf
e hf kT
» 1+ (6.46)
kT
2 f 2kT
Bf = (6.47)
c2
This is the RayleighJeans 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 RayleighJeans 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 RayleighJeans
law to frequencies below the frequency of maximum brightness; closer to the peak
of the blackbody curve the deviation of the RayleighJeans 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 RayleighJeans
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)
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 millimeterwave sensors.
Planck’s law is compared to the RayleighJeans and Wien laws in Figure 6.7.
The RayleighJeans 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 RayleighJeans law is illustrated by its dependence only on the
square of the frequency.
108
RayleighJeans
109
B (W m2 Hz1 str1)
Planck
1010
1011
Wein
1012
1013
1014 11
10 1012 1013 1014 1015
f (Hz)
Figure 6.7 Brightness of a blackbody at 1000 K. The RayleighJeans 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 MillimeterWave Remote Sensing for Security Applications
¥
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 '
¥ æ x3 3x 2 6x 6 ö
2k 4 T 4
B( f ¢) =
c 2 h3
å enx 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
onesided integrals
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 millimeterwave 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 bandlimited brightness can be approximated
6.3 Applied Radiometry 187
using the RayleighJeans 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
2f 2
B= kT D f (6.57)
c2
Equation (6.57) provides a simple, approximate calculation of the bandlimited
brightness for systems operating at microwave and millimeterwave frequencies.
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)
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 MillimeterWave Remote Sensing for Security Applications
1
TA =
WA òò T (θ , φ)A(θ , φ) dW (6.63)
source
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.
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)
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.
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
Ws
TA = Tavg (6.74)
WA
2π
A
Pf = k 2e
λ
ò T (θ)A(θ  θ0 )dθ (6.75)
0
where the onedimensional 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 crosscorrelation of the
temperature distribution and the antenna pattern. If the antenna pattern is symmet
ric, A(θ) = A(–θ), and the power becomes
2π
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
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
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 MillimeterWave Remote Sensing for Security Applications
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 righthand 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:
and thus
G + ¡ + a = 1 (6.84)
a = e (6.85)
Equation (6.84) can then equivalently be written
G + ¡ + e = 1 (6.86)
e = 1  G (6.87)
or very nearly the physical temperature of the skin. At 30 GHz, the radiometric
temperature is
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
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
millimeterwave 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 twodimensional 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 millimeterwave 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.
Tenv
ΓoTenv
To, Γo, eo
eoTo
TR = eoTo + ΓoTenv
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
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 millimeterwave 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
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 millimeterwave 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
and a brick wall. The radiometer thus has to have good sensitivity to resolve small
power differences.
Radiometers are, in essence, highgain, highsensitivity 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 millimeterwave 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
highgain, lownoise amplifier can then be used at the millimeterwave 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
RayleighJeans 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 squarelaw detector, such as a detector diode, in a single
receiver, or by crossmultiplying the outputs of multiple receivers.
Radiometers are designed to detect broadband, thermal radiation that has
noiselike 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 lowpass filter after
detection; an ideal lowpass 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 MillimeterWave 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 lowpass 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,
The antenna temperature TA represents the signal of interest, while the sensi
tivity is the standard deviation of the noise. Thus, the signaltonoise ratio for a
radiometer can be given by
TA TA
SNR = = DfRF τ (6.101)
DT CTrec
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 millimeterwave 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
10 103 102 101 100
τ (s)
Figure 6.12 Signaltonoise ratio as a function of the integration time and the system bandwidth
for Trec = 10TA and C = 1.
The detector has a power sensitivity K (V·W–1), and thus the mean voltage on
the output of the detector is
The integrator reduces the rms fluctuations of Vd; its average output voltage is given
by
∆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
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
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
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 27.4GHz 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
squarelaw detector is
¥
2
P = kTsys ò H(f ) df (6.113)
¥
202 Microwave and MillimeterWave Remote Sensing for Security Applications
Response
from human Responses from
background
Amplitude
Amplitude
Response
from human
where, for simplicity, the gain is assumed to be unity. The voltage input to the detec
tor is thus
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 lowpass 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 WienerKhinchin 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
z1z2 � z2n +1 = 0
(6.119)
z1z2 � z2n = å zi z j zk zl
All pairs
and thus
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
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 righthand 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 righthand 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 MillimeterWave 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
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 lowpass
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 lowpass 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)
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)
Ps DT DfRF
SNR = = (6.134)
Pn Tsys 2 DfLPF
2 DfLPF
DT = Tsys (6.135)
DfRF
Tsys
DT = (6.136)
DfRF τ
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
‡ The term correlation radiometer may refer to either a twoelement interferometric correlation radiometer
or a singleelement radiometer that splits the antenna output to two receivers, the outputs of which are then
correlated. However, the term generally refers to the twoelement 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:
θ RX1 V1(t)
D r
direction of the antennas as shown in Figure 6.17, the normalized voltage signals at
the inputs to the multiplier are 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 timevarying sinusoid at twice the radiation frequency, while
the second term is a dc component, assuming τg is timeinvariant. The bandwidth
of the lowpass 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 timedependent, 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
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
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
Narrowband
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 27.4GHz 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 27.4GHz 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 MillimeterWave 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 27.4GHz 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
where it is assumed that the source is located broadside to the antennas such that
θ = 0. The autocorrelation of (6.154) is
Each term in the angled brackets can then be written in terms of the autocorrelation
or crosscorrelation it represents:
2
Rc (τ ) = R12 (0) + R11(τ )R22 (τ ) + R12 (τ )R21(τ ) (6.157)
where Rij(τ) is the crosscorrelation between the voltage signals i and j. The first and
third terms, the crosscorrelations 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 timeaverage tends to zero; the signal
components are correlated.
6.4 Radiometer Receivers 213
The WienerKhinchin 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 P12 (f )P21(f ¢  f )df
¥
¥
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
¥
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 crosscorrelations
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 lowpass 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 MillimeterWave 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 righthand side of (6.163) represents the signal power Ps,
while the second and third terms represent the noise power within the lowpass
filter bandwidth Pn. The signaltonoise 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
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 millimeterwave 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 lowpass 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 ´ 103
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 ´ 103
6.5 Practical Considerations 215
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.
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 lowpass filter cutoff frequency, so that the output
is the timeaverage of the two voltages. After the unity gain amplifiers, the output
of the filter is
Vout = Vd , A  Vd ,R
= kG(TA  TR )DfRF
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
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 nonzero voltage responses.
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 millimeterwave radiometers used for terrestrial security applications
operate in the RayleighJeans 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 MillimeterWave Remote Sensing for Security Applications
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 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 Yfactor method,
Ton corresponds to Th, while Toff corresponds to Tc.
the antenna beam is moved in a pattern over a designated field of view. While staring
mode sensors may be singlereceiver systems, they are generally implemented as
imagers with multiple parallel receivers. The goal of staringmode 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.
Scanningmode 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 twodimensional radiometric representation of the scene. Scanningmode
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.
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,
and
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 MillimeterWave 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
dualantenna 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 dualan
tenna system is equal to that of the singleantenna 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 rangegates 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 ScanningBeam 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
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
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)
Rω
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 ScanningBeam 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].)
D xeff = τ Rω (6.198)
which derives from (6.197) with the integration time in place of the dwell time.
φHPBW θHPBW
ω
Figure 6.28 Onedimensional 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
Rω
DT Dx = CTsys (6.200)
DfRF
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
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
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
ωφ
ωθ
References
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 groundbased radar
and airborne radar [1–6]; the goal of this chapter is to cover the aspects relevant
to microwave and millimeterwave security radar systems and the basic waveforms
and system configurations required for security applications.
Microwave and millimeterwave 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 millimeterwave 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 millimeterwave, radar systems are thus sensitive to
229
230 Microwave and MillimeterWave 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 millimeterwave 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 nearlossless 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
highfrequency radar imaging systems for the detection of contraband.
Microwave and millimeterwave 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.
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
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
æ Rö
 j 2π fc ç t  ÷
è cø
si (t) = LAt e (7.4)
where L is the oneway 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 MillimeterWave Remote Sensing for Security Applications
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
directionofarrival 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.
Gt Gr
Pt Pr
Transmitter Receiver
R
Figure 7.3 Power transmission between a transmitter and receiver in freespace. Pt and Pr are the
transmit and receive powers; Gt and Gr are the transmit and receive antenna gains.
234 Microwave and MillimeterWave 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 oneway transmission between systems in free
space [7].
To derive the received power in a twoway 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)
4π
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 twoway 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 signaltonoise 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 MillimeterWave Remote Sensing for Security Applications
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,
sbb(t) st(t)
DAC
LO
Figure 7.4 Upconverting transmitter consisting of a local oscillator generating a signal sLO(t), base
band signal generator such as a digitaltoanalog converter (DAC) generating a signal sbb(t), upcon
verting mixer, and highpower 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
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 millimeterwave signals; digitalto
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 imagerejection 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.
Transmitter Transmitter
Receiver Receiver
(a) (b)
2
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
(a) (b)
Figure 7.7 (a) Amplitude noise in the time domain. (b) Phase noise in the timedomain.
240 Microwave and MillimeterWave 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.
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
f 5 Thermal noise
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 MillimeterWave 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.
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 millimeterwave security sensors. Commercially available Gunn diode
oscillators are available at frequencies above 110 GHz with good phase noise
characteristics.
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].
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 1Hz bandwidth,
244 Microwave and MillimeterWave Remote Sensing for Security Applications
and E is the signal energy. The ratio E ∕ N0 is the maximum signaltonoise 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
signaltonoise ratio. The error can thus be made arbitrarily small by increasing the
signaltonoise 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 signaltonoise 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 signaltonoise 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.
1 N0
σt = (7.43)
β 2E
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
1 N0
σf = (7.45)
α 2E
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)
¥ ¥
1 N0
σθ = (7.48)
γ 2E
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 MillimeterWave 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
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 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 timedelay 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 signaltonoise ratio was assumed to
1
tr = 0 tr ≈
∆fr
t t
(a) (b)
Figure 7.12 Pulse signals with (a) infinite bandwidth and zero risetime, (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
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
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 MillimeterWave 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 largeargument
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
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 outofband 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 bandlimited pulse (7.61) and the rms bandwidth given by the largeargument
approximation (7.64), as a function of the ratio of the pulse width to the rise time
τ ∕tr. The largeargument 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 bandlimited pulse calculated using the exact formulation and
the largeargument approximation. (b) Percentage error between the calculations of the rms band
width using the exact formulation and the largeargument approximation.
calculated using the exact rms bandwidth, σt,N, and the largeargument 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 largeargument 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 MillimeterWave 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
/σ
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 bandlimited pulse calculated using the
exact formulation and the largeargument approximation. (b) Percentage error between the calcula
tions of the rms time delay error using the exact formulation and the largeargument 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.
N0
σ t σf = (7.74)
αβ2E
For a fixed signaltonoise ratio, the measurement error can be improved through a
larger rms timebandwidth product; this requires long pulses with wide bandwidth.
As will be shown in the following, the rms timebandwidth product is always greater
than π. This section follows the derivation of this relation given by Skolnik in [14].
The rms timebandwidth 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 MillimeterWave 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 timebandwidth 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 MicroDoppler 253
¥ ¥ ¥ 2
2 2
ò f (x) dx ò g(x) dx ³ ò f (x)g(x)dx (7.83)
¥ ¥ ¥
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 timebandwidth 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
7.4 MicroDoppler
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 MillimeterWave 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 micromotions. Micromotion 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 timedependent sideband frequencies around the bulk Doppler
frequency; these sideband frequencies are referred to as microDoppler. Because
micromotions tend to be time dependent, microDoppler frequencies are likewise
time dependent.
is then combined in series to form the timefrequency plot. The STFT is defined
by [15]
¥
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 MillimeterWave 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 timefrequency domain.
7.4 MicroDoppler 257
ωr p1
r1
vb
rp,1 x
α
Radar
R0
ωr
p2
Figure 7.16 Geometric setup describing microDoppler induced by rotating propeller blades.
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
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
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
timevarying range to p1 due to the rotation is shown in Figure 7.17.
R0 + rp
R0
R 0  rp
t
Figure 7.17 Timevarying range to point p1 due to the rotation.
258 Microwave and MillimeterWave 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
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
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 microDoppler 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 MicroDoppler 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 microDoppler 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 timedependent nature of the microDoppler
frequencies.
260 Microwave and MillimeterWave 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 angledependent
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
data has also been published on the microDoppler signatures of moving vehicles
and animals such as deer, goats, dogs, and birds [25]. Additionally, microDoppler
analysis is being applied to biometric measurements through the detection of respi
ration and heartbeat monitoring for various applications such as search and rescue,
throughwall intrusion detection, and medical monitoring [33–38].
Characterization of the microDoppler 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 microDoppler
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 timefrequency signature of a human walking toward a 30GHz 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 MillimeterWave Remote Sensing for Security Applications
basis behind microDoppler 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 micromotions 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 microDoppler 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 microDoppler 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 microDoppler signatures and the
microDoppler energy [25, 30].
To demonstrate aspects of the human microDoppler 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 microDoppler 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 30GHz radar. The oscillatory patterns in the microDoppler 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 micromotions toward a
30GHz 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 MicroDoppler 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 micromotions 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
In Figure 7.21, the full signature is separated into the microDoppler 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 microDoppler 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
micromotions of the object and thus is highly angledependent. 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 microDoppler signature tends to
decrease significantly as α approaches 90°. Some microDoppler 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
30GHz 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 microDoppler signatures are essentially absent. While the
microDoppler 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 microDoppler 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 MicroDoppler 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 MillimeterWave Remote Sensing for Security Applications