close

Вход

Забыли?

вход по аккаунту

?

Investigation of microwave antennas with improved performances

код для вставкиСкачать
INVESTIGATION OF MICROWAVE ANTENNAS
WITH IMPROVED PERFORMANCES
by
Rongguo Zhou
______________________
A Dissertation Submitted to the Faculty of the
DEPARTMENT OF PHYSICS
In Partial Fulfillment of the Requirements for the Degree of
DOCTOR OF PHILOSOPHY
at the
UNIVERSITY OF ARIZONA
2010
UMI Number: 3412593
All rights reserved
INFORMATION TO ALL USERS
The quality of this reproduction is dependent upon the quality of the copy submitted.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.
UMI 3412593
Copyright 2010 by ProQuest LLC.
All rights reserved. This edition of the work is protected against
unauthorized copying under Title 17, United States Code.
ProQuest LLC
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106-1346
2
THE UNIVERSITY OF ARIZONA
GRADUATE COLLEGE
As members of the Dissertation Committee, we certify that we have read the dissertation
prepared by Rongguo Zhou
entitled Investigation of Microwave Antenna with Improved Performances
and recommend that it be accepted as fulfilling the dissertation requirement for the
Degree of Doctor of Philosophy.
_______________________________________________________________________
Date: 06/01/2010
Hao Xin
_______________________________________________________________________
Date: 06/01/2010
Richard W. Ziolkowski
_______________________________________________________________________
Date: 06/01/2010
Lizhi Fang
_______________________________________________________________________
Date: 06/01/2010
Brian Leroy
_______________________________________________________________________
Date: 06/01/2010
Michael Shupe
Final approval and acceptance of this dissertation is contingent upon the candidate’s
submission of the final copies of the dissertation to the Graduate College.
I hereby certify that I have read this dissertation prepared under my direction and
recommend that it be accepted as fulfilling the dissertation requirement.
________________________________________________ Date: 06/01/2010
Dissertation Director: Hao Xin
3
STATEMENT BY AUTHOR
This dissertation has been submitted in partial fulfillment of requirements for an
advanced degree at The University of Arizona and is deposited in the University Library
to be made available to borrowers under rules of the Library.
Brief quotations from this dissertation are allowable without special permission,
provided that accurate acknowledgment of source is made. Requests for permission for
extended quotation from or reproduction of this manuscript in whole or in part may be
granted by the head of the major department or the Dean of the Graduate College when in
his or her judgment the proposed use of the material is in the interests of scholarship. In
all other instances, however, permission must be obtained from the author.
SIGNED: _____Rongguo Zhou_______
4
ACKNOWLEDGEMENT
I would like to thank my advisor Dr. Hao Xin for his guidance, advice and full
support throughout this dissertation work, which have been crucial to the success of the
work. I am very grateful to have the opportunity working with him. I would also like to
thank all the dissertation committee members for reading the dissertation and all the
support they offered.
I owe many thanks to Dr. Richard W. Ziolkowski for his great help on the
discussion about metamaterial, to Dr. Zhang Hualiang in Department of Electrical
Engineering, University of North Texas for his technique support on the antenna
measurement and circuit fabrication, my colleagues Dr. Wang Lu, Dr. Wu Ziran and TC
Chen for their helpful advice and support, and Dr. Liu Duixian in IBM research center in
Yorktown Heights for his help on the MMW antenna fabrication and measurement. I
also wish to thank all the members in the mmW Antennas and Circuits group.
I am forever indebted to my father Youping Zhou, my mother Guilan Yao, and
the rest of my family for their constant love. Without their support, this work would not
have been possible. I am also deeply grateful for the understanding and encouragement
from all my lovely friends.
5
TABLE OF CONTENTS
LIST OF FIGURES............................................................................................................. 7
LIST OF TABLES ............................................................................................................ 13
ABSTRACT...................................................................................................................... 14
CHAPTER 1.
INTRODUCTION ............................................................................ 16
1.1. METAMATERIAL WITH NEAR-ZERO INDEX OF REFRACTION .................................. 16
1.1.1. Basic Background of Metamaterials............................................................ 16
1.1.2. Applications of Metamaterials with Near-Zero Index of Refraction........... 18
1.1.3. Metallic Wire Array as Metamaterials with Near-Zero Index of Refraction19
1.2. MMWAVE ANTENNA FOR WIRELESS COMMUNICATIONS ....................................... 22
1.2.1. Potential Applications at 60 GHz Band ....................................................... 22
1.2.1.1. Wireless Local Area Networks (WLAN) ............................................. 23
1.2.1.2. Wireless Personal Area Networks ...................................................... 24
1.2.1.3. Multimedia Streaming......................................................................... 24
1.2.2. Challenges and Antenna Requirements for 60 GHz Communications........ 24
1.2.3. Motivation.................................................................................................... 25
1.3. NOVEL DIRECTION OF ARRIVAL ESTIMATION TECHNIQUE INSPIRED BY HUMAN
EARS 27
1.4. DISSERTATION ORGANIZATION .............................................................................. 32
CHAPTER 2.
EFFECTIVE MEDIUM PARAMETERS EXTRACTION .............. 35
2.1. SCATTERING PARAMETERS .................................................................................... 36
2.2. NICOLSON-ROSS-WEIR METHOD FOR PARAMETER EXTRACTION .......................... 37
2.3. S-SHAPE METAMATERIAL ...................................................................................... 40
2.4. CONCLUSION.......................................................................................................... 46
CHAPTER 3.
METALLIC WIRE ARRAY AS A LOW-EFFECTIVE INDIXE OF
REFRACTION MEDIUM [35] ........................................................................................ 49
3.1. METALLIC WIRE ARRAY AS EFFECTIVE MEDIUM .................................................. 49
3.2. A 2-D EMXT IN A LOW INDEX WIRE ARRAY HOST .............................................. 53
3.3. DIRECTIVE MONOPOLE ANTENNA EMBEDDED IN A WIRE ARRAY MEDIUM .......... 58
3.3.1. Principles of Antenna Operation.................................................................. 58
3.3.2. An X-band Directive Monopole/Wire Array Antenna Design .................... 60
3.3.3. Parametric Study of the Monopole Antenna System................................... 62
3.3.3.1. Monopole Length ................................................................................ 63
3.3.3.2. Wire Array Size ................................................................................... 64
3.3.3.3. Wire Array Height............................................................................... 65
6
TABLE OF CONTENTS - Continued
3.3.4. Experimental Verification............................................................................ 65
3.3.5. Discussion .................................................................................................... 69
3.4. CONCLUSION.......................................................................................................... 71
CHAPTER 4.
A WIDEBAND CIRCULARLY POLARIZED PATCH ANTENNA
FOR 60GHZ WIRELESS COMMUNICATIONS [90]..................................................... 72
4.1. INTRODUCTION ...................................................................................................... 72
4.1.1. Techniques for Improving Antenna Performances ...................................... 74
4.1.2. Motivation.................................................................................................... 75
4.2. ANTENNA STRUCTURE/PACKAGING ....................................................................... 76
4.3. DETAILED ANTENNA DESIGN................................................................................. 78
4.4. COMPARISON OF MEASUREMENT AND SIMULATION .............................................. 84
4.5. CONCLUSION.......................................................................................................... 89
CHAPTER 5.
IMPROVED TWO-ANTENNA DIRECTION FINDING INSPIRED
BY HUMAN EARS [114]................................................................................................. 91
5.1. MOTIVATION.......................................................................................................... 91
5.2. ANALOGY BETWEEN HUMAN SOUND LOCALIZATION AND MICROWAVE DIRECTION
FINDING .......................................................................................................................... 93
5.3. NUMERICAL SIMULATIONS .................................................................................... 97
5.4. EXPERIMENTAL RESULTS ..................................................................................... 105
5.5. CONCLUSION........................................................................................................ 111
CHAPTER 6.
A MICROWAVE DIRECTION OF ARRIVAL ESTIMATION
TECHNIQUE USING A SINGLE UWB ANTENNA [115] .......................................... 112
6.1. INTRODUCTION .................................................................................................... 112
6.2. SIMULATION RESULTS OF THE SINGLE ANTENNA DOA ESTIMATION .................. 114
6.2.1. Elliptical UWB Antenna and DOA Estimation ......................................... 115
6.2.2. D-shaped UWB Antenna and DOA Estimation......................................... 119
6.3. EXPERIMENTAL VERIFICATION OF THE SINGLE ANTENNA DOA ESTIMATION ..... 122
6.4. CONCLUSION........................................................................................................ 131
CHAPTER 7.
CONCLUSION AND FUTURE WORK........................................ 132
APPENDIX: MUSIC ALGORITHM ............................................................................. 145
REFERENCES ............................................................................................................... 148
7
LIST OF FIGURES
Figure 1-1. Categorization diagram of metamaterials based on the real parts of their
permittivity (ε′) and permeability (μ′)............................................................................... 18
Figure 1-2. Top view of a 2-D square lattice of wires with radius r and periodicity a. ... 20
Figure 1-3. International unlicensed spectrum around 60 GHz [37]. .............................. 22
Figure 1-4. Comparison of microwave direction finding system (left) and the human
auditory system (right)...................................................................................................... 28
Figure 1-5. A simple microwave direction finding scheme with two antennas separated
by a low-pass filter (LPF) like object mimicking the human head................................... 30
Figure 1-6. A simplified illustration of monaural direction finding mechanism: the
received spectrum has an incident angle dependent notch response. ............................... 31
Figure 2-1. The incident and reflected waves of a two-port network............................... 36
Figure 2-2. (a) S-shaped metamaterial (b) Top view of one unit cell............................... 41
Figure 2-3. The simulated S-parameters (S11 and S21) of the S-shaped structure (a)
magnitude, (b) phases. ...................................................................................................... 42
Figure 2-4. (a) The magnitude of the two solutions of the transmission term T1 and T2, (b)
The real part of their corresponding normalized impedances 1 (z1) and 2 (z2)............ 42
Figure 2-5. Calculated real part of n of the S-shaped slab for different m values. ........... 43
Figure 2-6. Extracted effective parameters of the S-shaped slab, (a) relative permittivity ε
(solid line: real part ε', dashed line: imaginary part ε''), (b) relative permeability μ (solid
line: real part μ', dashed line: imaginary part μ'').............................................................. 44
Figure 2-7. The retrieved index of refraction as a function of frequency: real part n' (solid
line) and imaginary part n'' (dashed line). The dash-dotted line in n' indicates the upper
edge of the first Brillouin zone, nedge =π/(k0*g1). ............................................................. 45
Figure 3-1. Top view of a 2-D square lattice of wires with radius r and periodicity a. It
can also be thought of as two independent square lattices (solid and empty circles) with a
periodicity of 2a , embedded within each other............................................................. 50
Figure 3-2. Calculated effective permittivity of a 2-D wire array (a = 4 mm, r = 1 µm)
using plasma theory, treating the array as either a single square lattice (circles) or two
square lattices (squares) embedded within each other...................................................... 51
Figure 3-3. Top view of the HFSS model of a wire array structure. A normal-incident
plane wave propagates in the x direction and the array is infinite in both the z and y
directions........................................................................................................................... 52
Figure 3-4. A two-dimensional dielectric rod EMXT structure embedded in a low index
of refraction wire array (top view).................................................................................... 53
8
LIST OF FIGURES - Continued
Figure 3-5. Calculated real part of n of the wire array with a =5 mm and r = 0.25 mm for
different branches (m values)............................................................................................ 54
Figure 3-6. Extracted real components of ε and μ compared with the ε calculated from Eq.
(1.1) (a = 5 mm, r = 0.25 mm).......................................................................................... 55
Figure 3-7. Simulated transmissions of the actual composite of the dielectric rods and the
metallic wire array (squares) and the dielectric rods embedded within a uniform slab with
the extracted effective medium parameters of the wire array (circles)............................. 57
Figure 3-8. Geometrical illustration of the beam narrowing effect for a source embedded
in an n2 < 1 medium. ......................................................................................................... 59
Figure 3-9. Top view of a monopole embedded within a wire array that is infinite in the
z-direction. ........................................................................................................................ 60
Figure 3-10. The extracted n' and n" (n = n' + i n") of a small refractive index wire array
medium for X-band applications (a = 9.0 mm and r = 0.25 mm). ................................... 61
Figure 3-11. (a) Simulated return losses of a 17.4-mm monopole in the actual wire array
(solid line) and in a uniform medium with extracted effective permittivity and
permeability representing the wire array (dashed line). (b) Simulated gain (in dB) of the
monopole in the wire array (solid line) comparing with the case without the wire array
(dashed line). The wire array has the following dimensions: a = 9 mm and r = 0.25 mm;
and the 11 x 11 wires are terminated by two conducting ground planes separated by 50
mm. ................................................................................................................................... 62
Figure 3-12. The simulated return losses of the monopole / wire array system for various
monopole lengths. ............................................................................................................. 64
Figure 3-13. A photo of the fabricated antenna prototype. The monopole and wire lengths
are 17.4 mm and 50.0 mm, respectively. .......................................................................... 66
Figure 3-14. Comparison of the measured (circles) and simulated (solid line) return losses
(S11) of the 17.4 mm monopole antenna embedded in the wire array............................... 66
Figure 3-15. Measured x-y plane radiation patterns (normalized to 0 dB) of the antenna
system (solid line) compared with the simulated radiation patterns of the monopole in the
actual wire array (dashed line) and in the effective medium (dashed-dotted line) at: (a)
8.5 GHz, (b) 9.5 GHz, (c) 10.3 GHz and (d) 11.3 GHz.................................................... 68
Figure 3-16. The simulated and measured radiation patterns (normalized to 0 dB) of the
antenna in the x-z plane at 9.5 GHz.................................................................................. 69
Figure 4-1. (a) The side view of the entire antenna structure. (b) The copper cavity...... 76
Figure 4-2. Schematic view of the antenna packaged with an integrated chipset. ........... 78
Figure 4-3. The microstrip-fed square patch antenna with a diagonal slot....................... 79
9
LIST OF FIGURES - Continued
Figure 4-4. Simulated reflection coefficients (S11 in dB) of the square patch antenna (L =
1400 µm) with different slot lengths C (slot width d = C/10). ......................................... 80
Figure 4-5. Simulated microstrip-fed square patch antenna performance: (a) Reflection
coefficients (S11 in dB). (b) Axial ratio............................................................................ 81
Figure 4-6. The top view of the antenna layer including (a) the CPW feed with straight
open-stubs; (b) the CPW feed with bent open-stubs......................................................... 82
Figure 4-7. Comparison of antenna performances with different feedings (a) axial ratio
(b) reflection coefficient (S11 in dB) ................................................................................. 84
Figure 4-8. A photo showing the top view of the fabricated and packaged 60 GHz lefthand circular polarized patch antenna............................................................................... 85
Figure 4-9. The simulated and measured reflection coefficients (S11 in dB) of the fully
packaged antenna. ............................................................................................................. 87
Figure 4-10. The simulated and measured axial ratio of the fully packaged antenna. ..... 87
Figure 4-11. The simulated and measured antenna gain (a) and antenna efficiency (b) of
the fully packaged antenna................................................................................................ 88
Figure 4-12. The simulated and measured radiation patterns of the fully packaged antenna
at 61 GHz in: (a) y-z plane (b) x-z plane. ......................................................................... 88
Figure 4-13. The measured co-polarized and cross-polarized radiation patterns of the
fully packaged antenna at 61GHz in: (a) y-z plane (b) x-z plane ..................................... 89
Figure 5-1. Comparison of a passive microwave direction finding system (left) and the
human auditory system (right). ......................................................................................... 94
Figure 5-2. Utilizing the HRTF, the human auditory system can achieve unambiguous
direction finding for high frequency signals..................................................................... 96
Figure 5-3. A finite-element model illustrating the geometry of the two-monopole and
scatterer configuration with an incoming signal from an azimuth angle.......................... 98
Figure 5-4. Simulated phase (a) and magnitude (b) differences of two monopole antennas
without a scatterer in between using two different methods: horn antenna illumination –
individual markers and plane wave excitation - lines....................................................... 99
Figure 5-5. Simulated phase (a) and magnitude (b) differences at 10 GHz and phase (c)
and magnitude (d) differences at 12 GHz with and without a lossy scatterer in between.
......................................................................................................................................... 100
Figure 5-6. Simulated MUSIC output of the three scenarios with a signal incident from
the 80° direction: without a scatterer (dotted dashed line), with a symmetrically
positioned rectangular scatterer (dashed line) and with an asymmetrically positioned
rectangular scatterer (solid line)...................................................................................... 102
10
LIST OF FIGURES - Continued
Figure 5-7. Simulated averaged DOA estimation errors assuming ±0.25 dB magnitude
difference and ±2° phase difference errors for rectangular shaped scatterers: (a) versus
frequency (averaged over all incident angles from 0° to 360° with 2° step) and (b) versus
incident angle (averaged over all frequencies from 8 to 12 GHz with 0.5 GHz step).... 103
Figure 5-8. Simulated averaged DOA estimation errors assuming ±0.25 dB magnitude
difference and ±2° phase difference errors for cylindrical shaped scatterers: (a) versus
frequency (averaged over all incident angles from 0° to 360° with 2° step) and (b) versus
incident angle (averaged over all frequencies from 8 to 12 GHz with 0.5 GHz step).... 105
Figure 5-9. A photograph of the X-band two-monopole and symmetric scatterer prototype.
......................................................................................................................................... 106
Figure 5-10. Comparison of measured (solid lines) and simulated (dashed lines)
magnitude (left) and phase (right) differences at 10 GHz for the three cases: (a) No
scatterer; (b) Symmetric scatterer; (c) Un-symmetric scatterer...................................... 108
Figure 5-11. MUSIC output of a 12 GHz signal incident from 90o for the two-antenna
configurations without scatterer (dotted dashed line), with the symmetric scatterer
(dashed line), and with the asymmetric scatterer (solid line). ........................................ 109
Figure 5-12. Measured averaged estimation errors for all three two-antenna
configurations (a) versus frequency (averaged over all incident angles from 0° to 360°
with 15° step) and (b) versus incident angle (averaged over all frequencies from 8 to 12
GHz with 0.5 GHz step).................................................................................................. 110
Figure 6-1. (a) The schematics (left: side view; right: top view) of the symmetric UWB
antenna incorporating two slots. (b) The simulated antenna return loss......................... 115
Figure 6-2. (a) The received spectra of the symmetric UWB antenna with the incident
waves (H-field in the –z direction) at -180º, -120º, -60º, 0º, 60º and 120º in the x-y plane.
(b) The correlation coefficients between the spectra with added noise (SNR = 15 dB) and
the pre-determined spectra at the incident angle θ (-180º, -120º, -60º, 0º, 60º and 120º) in
the x-y plane. (c) The RMS of the DOA estimation error with different SNR............... 117
Figure 6-3. The received spectra of the UWB antenna at -180º, -120º, -60º, 0º, 60º and
120º for the incident waves in (a) the x-z plane (E-field in the y direction), (b) the x-z
plane (H-field in the y direction), and (c) the y-z plane (E-field in the x direction). ...... 118
Figure 6-4. The RMS of the estimation error with different SNRs for the incident waves
in the x-z plane with the (a) E-field in the y direction, (b) H-field in the y direction , and (c)
in the y-z plane with the E-field in the x direction. ......................................................... 119
Figure 6-5. (a) The schematics (left: side view; right: top view) of the proposed UWB
monopole antenna with unsymmetrical shape. (b) Reflection coefficient...................... 120
11
LIST OF FIGURES - Continued
Figure 6-6. DOA performance of the D-shaped UWB antenna in the x-y plane (H-field in
the z direction): (a) The correlation coefficients between spectra with added noise (SNR =
15 dB) and the pre-determined spectra at the incident angle θ (-180º, -120º, -60º, 0º, 60º
and 120º); (b) The RMS of the estimation errors for different SNR. ............................. 121
Figure 6-7. The RMS of the estimation errors of the D-shaped antenna for different SNR
when the incident waves are in (a) the x-z plane (E-field in the y direction), (b) the x-z
plane (H-field in the y direction), and (c) the y-z plane (E-field in the x direction). ...... 122
Figure 6-8. The simulated and measured return losses of (a) the elliptical UWB antenna
and (b) the D-shaped UWB antenna. .............................................................................. 123
Figure 6-9. The simulated and measured received-patterns of the elliptical UWB antenna
at 9 GHz with the incident wave in (a) the x-y plane (H-field in the z direction), (b) the x-z
plane (E-field in the y direction), (c) the x-z plane (H-field in the y direction), and (d) the
y-z plane (E-field in the x direction). .............................................................................. 124
Figure 6-10. The simulated and measured received-patterns of the D-shaped UWB
antenna at 9 GHz with the incident wave in (a) the x-y plane (with H-field in the z
direction), (b) the x-z plane (with E-field in the y direction), (c) the x-z plane (with H-field
in the y direction), and (d) the y-z plane (with E-field in the x direction)....................... 125
Figure 6-11. The estimation errors of the elliptical UWB antenna with the incident wave
in (a) the x-y plane (with the incident H-field in the z direction), (b) the x-z plane (with Efield in the y direction), (c) the x-z plane (with H-field in the y direction), and (d) the y-z
plan (with E-field in the x direction)............................................................................... 127
Figure 6-12. The estimation errors of the D-shaped UWB antenna with the incident
waves in (a) the x-y plane (with H-field in the –z direction), (b) the x-z plane (with E-field
in the +y direction), (c) the x-z plane (with H-field in the +y direction), and (d) the y-z
plane (with E-field in the +x direction). ......................................................................... 129
Figure 6-13. The estimation errors of the D-shaped UWB antenna with increased SNR
when the incident waves are in (a) the x-z plane (with H-field in the y direction). (b) the
y-z plane (with E-field in the x direction). ...................................................................... 131
Figure 7-1. T ................................................................................................................... 137
Figure 7-2. The schematic of the receiver in ADS (a) front side. (b) back side. ............ 138
Figure 7-3.The received IF signals with the input signals to be 0.01*exp(iωt) and
0.02*exp(iωt+i*π/4). (a) the magnitude of the output signal at receiver 1, (b) the
magnitude of the output signal at receiver 2, (c) magnitude difference between the two
output signals, (d) phase difference between the two output signals.............................. 139
Figure 7-4. A photo of the fabricated receiver (a) front side. (b) back side. .................. 140
12
LIST OF FIGURES - Continued
Figure 7-5. Measured waveforms of the four branches of the receiver. ......................... 141
Figure 7-6. The ADS schematic for emulating the MLD effect. .................................... 142
Figure 7-7. The ADS schematic inside channel 1 and channel 2.. ................................. 143
13
LIST OF TABLES
Table 3-1. The 2-D Dielectric Rods Band Gap Frequency and Bandwidth for Different
Backgrounds ..................................................................................................................... 56
Table 4-1. Configuration of the Antenna Structure and Packaging.................................. 77
Table 4-2. The Dimension of the Finalized Patch Antenna Design ................................ 80
Table 4-3. The Geometry of the CPW to Microstrip Transition...................................... 84
14
ABSTRACT
This dissertation presents the investigation of antennas with improved
performances at microwave frequencies. It covers the following three topics: the study of
the metamaterial with near-zero index of refraction and its application in directive
antenna design, the design technique of a wideband circularly polarized patch antenna for
60GHz wireless application and the investigation of a novel direction of arrival (DOA)
estimation technique inspired by human auditory system.
First, the metamaterial
composed of two-dimensional (2-D) metallic wire arrays is investigated as an effective
medium with an effective index of refraction less than unity (neff < 1). The effective
medium parameters (permittivity εeff, permeability μeff and neff) of a wire array are
extracted from the finite-element simulated scattering parameters and verified through a
2-D electromagnetic band gap (EBG) structure case study. A simple design methodology
for directive monopole antennas is introduced by embedding a monopole within a
metallic wire array with neff < 1 at the antenna operating frequencies. The narrow beam
effect of the monopole antenna is demonstrated in both simulation and experiment at Xband (8 – 12 GHz). The measured antenna properties including return loss and radiation
patterns are in good agreement with simulation results. Parametric studies of the antenna
system are performed.
The physical principles and interpretations of the directive
monopole antenna embedded in the wire array medium are also discussed. Second, a
fully packaged wideband circularly polarized patch antenna is designed for 60GHz
15
wireless communication. The patch antenna incorporates a diagonal slot at the center and
features a superstrate and an air cavity backing to achieve desired performances including
wide bandwidth, high efficiency and low axial ratio. The detailed design procedure of the
circularly polarized antenna, including the design of the microstrip-fed patch antenna and
the comparison of the performances of the antenna with different feeding interfaces, is
described. The experimental results of the final packaged antenna agree reasonably with
the simulation results. Third, an improved two-antenna direction of arrival (DOA)
estimation technique is explored, which is inspired by the human auditory system. The
idea of this work is to utilize a lossy scatter, which emulates the low-pass filtering
function of the human head at high frequency, to achieve more accurate DOA estimation.
A simple 2-monopole example is studied and the multiple signal classification (MUSIC)
algorithm is applied to calculate the DOA. The improved estimation accuracy is
demonstrated in both simulation and experiment. Furthermore, inspired by the sound
localization capability of human using just a single ear, a novel direction of arrival
estimation technique using a single UWB antenna is proposed and studied. The DOA
estimation accuracies of the single UWB antenna are studied in the x-y, x-z and y-z planes
with different Signal to Noise Ratios (SNR). The proposed single antenna DOA
technique is demonstrated in both simulation and experiment, although with reduced
accuracy comparing with the case of two antennas with a scatter in between. At the end,
the conclusions of this dissertation are drawn and possible future works are discussed.
16
CHAPTER 1.
INTRODUCTION
In recent years, metamaterials have attracted much attention because of their
special electromagnetic properties and potential applications in microwave, infrared and
optical frequencies. A brief review of metamaterials, including double-negative (DNG)
media [1], epsilon-negative (ENG) media [2], mu-negative (MNG) media [3], near-zero
index-of-refraction media and their applications, will be presented, followed by the
motivation of our work in the metallic wire array as a medium with a near-zero index of
refraction for directive antenna design. In addition, demands of modern communication
and sensor systems for more bandwidth, higher resolution and compactness lead to an
operating frequency up to the millimeter wave (mmWave, f > 30 GHz) or even submmWave regime. The applications and challenges of a 60 GHz wireless system, as well
as the mmWave antenna design requirements for the 60 GHz wireless communications,
will be explored. Moreover, this chapter also provides the background knowledge of the
sound-localization mechanisms of the human auditory system and their inspirations on
microwave passive direction finding applications. Finally, the dissertation organization
will be described.
1.1. Metamaterial with Near-Zero Index of Refraction
1.1.1. Basic Background of Metamaterials
Metamaterials (MTMs) have emerged in the last few years. They are artificially
17
constructed composite materials, which have engineered electromagnetic responses that
are not readily available in nature [1-27]. Metamaterials can be categorized by the
diagram given in Figure. 1-1 according to the signs of the real parts of the material
permittivity (ε′) and permeability (μ′). If both ε′ and μ′ are positive, as most of the
materials in nature are, they are named double-positive (DPS) media. On the other hand,
if both of these quantities are negative, as shown in the third quadrant in Figure. 1-1, the
materials are called double-negative (DNG) media. They have only been demonstrated
with artificial constructions [6-20]. Because of their anomalous properties, the DNG
metamaterials have been the subject of great interest, and have been widely proposed for
future applications, such as perfect lens [8], sub-wavelength cavity resonators [14], etc.
In the second quadrant, the media with a negative ε′ and positive μ′ are named epsilonnegative (ENG) materials. Examples of ENG metamaterials are the plasma and
plasmonic materials (e.g., noble metals at optical frequencies) below their plasma
frequencies [2]. Similarly, in the fourth quadrant, the mu-negative (MNG) media are
those with negative μ′ and positive ε′, which may be realized with ferromagnetic
materials or synthesized with suitable inclusions in host materials [3]. Near the two axes
of Figure. 1-1, where ε′ or μ′ is near zero, the media are named epsilon-near-zero (ENZ)
and mu-near-zero (MNZ) materials, respectively. A medium with both epsilon and mu
equal to zero, which fall at the origin of Figure 1-1, are called matched zero-index
metamaterials [21]. A summary of the various proposed and demonstrated metamaterials
and their applications can be found in [22-23].
18
Re(μ)
ENG
DPS
(ε′< 0, μ′> 0)
(ε′> 0, μ′> 0)
Re(ε)
DNG
MNG
(ε′< 0, μ′< 0)
(ε′> 0, μ′< 0)
ENZ
Figure 1-1. Categorization diagram of metamaterials based on the real parts of their
permittivity (ε′) and permeability (μ′).
1.1.2. Applications of Metamaterials with Near-Zero Index of Refraction
In virtue of their unique electromagnetic properties, metamaterials with near-zero
index of refraction offer exciting potential applications in microwave, infrared and optical
frequencies and can be utilized to overcome some conventional physical limits. In [24],
an ultra-compact resonator has been demonstrated using bi-layers of metamaterials with
near-zero index of refraction. The ENZ metamaterials can also be used as covers to
cancel the scattering from dielectrics or even conductors, dramatically reducing their total
scattering cross sections and making the covered objects transparent to external observers
[25]. In addition, it is possible to squeeze electromagnetic energy through ENZ
subwavelength narrow channels with reduced reflection coefficient at the junctions
leading to improved transmission [26]. Moreover, zero phase variation at various points
in the zero-index medium is demonstrated, making the possibility of realizing delay lines
19
with no phase differences between their inputs and outputs [21]. Additionally, with the
use of the ENZ materials, we can manipulate the impinging wave front and transform its
phase distribution into a desired shape by properly tailoring the exit side of the ENZ slab
[27], which may have important implications in imaging and communications technology.
Furthermore, antennas with enhanced directivity and gain have been proposed and
demonstrated by embedding the radiating source within a metamaterial with near-zero
index of refraction [28-35].
1.1.3. Metallic Wire Array as Metamaterials with Near-Zero Index of Refraction
Because of their wide range of applications as described in Section 1.1.2,
metamaterials that exhibit near-zero index of refraction are highly desirable. It has been
demonstrated that the propagation constant of a dispersive DNG metamaterial
continuously passes through zero in its transition from a DNG region to a DPS region
[36], giving a zero-index at its transition frequency. Another simple structure to realize an
effective medium with near-zero index of refraction (neff < 1) is the 2D periodic metallic
wire array, as shown in Figure 1-2.
20
2r
a
2a
Wires
Figure 1-2. Top view of a 2-D square lattice of wires with radius r and periodicity a.
The effective permittivity εeff of this kind of wire arrays can be calculated using
plasma theory with a reduced electron density [2], as shown in Eq. (1.1):
 eff  1   p2  2  1  2c 2 /[ a 2 ln(a / r ) 2 ]
(1.1)
where a is the wire array periodicity, r is the wire radius, ω is the angular frequency, c is
the speed of light in free space and ωp is the plasma frequency, at which the effective
permittivity εeff, and thus the effective index of refraction neff, is zero. An important
assumption of Eq. (1.1) is that the wires are very thin (r << a) so that the plasma
frequency corresponds to a free space wavelength much greater than the lattice spacing
and the Bragg diffraction effect can be ignored [2].
As shown in Eq. (1.1), the periodic 2-D metallic wire arrays, which can be
analyzed using a simple plasma theory with reduced electron density, are an easily
realizable structure with εeff < 1 and neff < 1 at frequencies around its plasma frequency
ωp. Therefore, a directive antenna with increased gain can be achieved by embedding an
antenna within a periodic wire array with neff <1. The proposed metamaterial for directive
21
radiation here is different from the structures in [29, 30, 32], which were based on a
ground-plane backed metamaterial slab composed of rows or meshes of metallic wires
with finite length and their effective medium parameters could not be directly estimated
by the simple plasma theory. In addition, most of the previous work did not directly apply
the effective medium parameters in the antenna designs. In this work, two-dimensional
(2-D) metallic wire arrays terminated by two ground planes are studied as media with low
effective index of refraction (neff < 1). A simple methodology is then utilized to design
directive antennas based on the effective medium parameters of a wire array that are
extracted from finite-element simulation results. First, the geometry of the wire array is
optimized to achieve a low neff at the desired frequencies. The extraction results confirm
the low neff properties of the wire array at frequencies just above the theory-predicted
plasma frequency. A 2-D electromagnetic crystal (EMXT) structure (a square lattice
made of dielectric rods) embedded in the wire medium is then investigated. It is found
that the first band gap of the 2-D EMXT structure shifts to a higher frequency as expected
when the hosting free space region (n = 1) is replaced by the neff < 1 wire array medium.
Moreover, the simulated transmission response of the actual composite of the EMXT and
metallic wires agrees with that of the EMXT embedded in a background assigned to have
the extracted effective medium parameters, confirming the validity of treating the wire
array as an effective medium. Directive monopole antennas are then realized by
embedding a monopole within a metallic wire medium with neff < 1 at the antenna
operating frequency. The monopole length, wire array size and height effects on the
antenna properties are studied. A prototype antenna operates within the X-band is
22
designed, fabricated and characterized. The measured antenna properties are in good
agreement with the simulation results, confirming the expected narrow beam radiation
and the design methodology. The narrow beam effect achieved by the monopole / wire
array antenna system is examined in the contexts of Snell’s law and effective aperture
size in order to gain more physical insight. The detailed work will be presented in
Chapter 3.
1.2. mmWave Antenna for Wireless Communications
Demands of modern communication and sensor systems for more bandwidth,
higher resolution and compactness lead to operating frequencies up to the millimeter
wave (mmWave, f > 30 GHz) or even sub-mmWave regime. In this section, a brief
review of the applications and challenges of 60 GHz wireless systems, as well as the
mmWave antenna design requirements for the 60 GHz wireless communications, are
described.
1.2.1. Potential Applications at 60 GHz Band
Figure 1-3. International unlicensed spectrum around 60 GHz [37].
23
As illustrated in Figure 1-3, a wideband of unlicensed spectrum around the 60
GHz ISM (the industrial, scientific and medical) band is available in many countries. As
a result of the large bandwidth available and its high data throughput rate, the 60 GHz
band is very attractive for a number of applications, including wireless personal-area
networks (WPAN), local-area networks (WLAN), multimedia streaming, and file transfer
between personal-computers and portable devices, etc [37-38]. Moreover, as the battery
lifetime is a major bottleneck for many portable devices, the 60 GHz band has a distinct
benefit because the large bandwidth can be utilized to tradeoff bandwidth efficiency for
low power consumption, while still maintaining high data rate [38]. In addition, due to
the atmospheric absorption peak around 60 GHz (mainly oxygen molecules), a
communication link at this band is inherently secure and has less interference among
users, which is ideal for indoor applications.
1.2.1.1. Wireless Local Area Networks (WLAN)
The WLAN for computer networking and internet traffic has been the most
popular wireless applications of the unlicensed spectrum. The key feature of 60 GHz
technology for WLANs is its ability to provide gigabits per second of throughput, making
it possible for the next generation of Ethernet. The increased attenuation of 60 GHz
signals will require 60 GHz repeaters for typical WLAN applications. Some venders have
proposed the hybrid 2.5/5/60 GHz WLAN solutions using lower frequencies for normal
operation and 60 GHz for high speed operation with short-range line-of-sight path [37].
24
1.2.1.2. Wireless Personal Area Networks
Presently in the United States, ultra-wide band (UWB) systems are is able to
provide hundreds of megabits per second date rate for the WPAN applications in the
frequency range from 3.1 to 10.6 GHz [39]. However, there are two major problems for
the UWB communications [37]. First, the UWB spectrum is not the same worldwide. For
example, the UWB spectrum in Europe covers only 3.1 ~ 4.8 GHz and 6 ~ 9 GHz.
Therefore, the UWB system does not support the worldwide market for the WPAN
applications. Second, it does not provide high enough data rates (480 Mb/s at its highest
rate). In order to achieve multi-gigabit per second wireless communication world-widely,
it seems that the 60 GHz band is the only probable solution in the near future.
1.2.1.3. Multimedia Streaming
The 60 GHz technology also provides an important application for high data rate
multimedia streams such as high definition multimedia. Currently, the wireless high
definition multimedia interface (HDMI) operates in the 2.5 GHz and 5 GHz spectra with
limited bandwidth and limited data rate. As a result, these systems implement either lossy
or lossless compression, leading to an increased number of components, design costs,
system complexities and product sizes. A 60 GHz communication system will likely
provide a cost-effective wireless HDMI solution with compact size [37].
1.2.2. Challenges and Antenna Requirements for 60 GHz Communications
As described in Section 1.2.1, there are a wide range of applications of the 60
25
GHz system. However, many technological challenges arise with the 60 GHz system as
well. First of all, the small wavelength at 60 GHz (5 mm in free space) requires highprecision machining, accurate alignment, or high-resolution photolithography. In addition,
mm-Wave circuits are usually assembled using expensive and bulky waveguides with
low level of integration [40-43]. The conventional low-cost packing technology is limited
and can only be applied at low frequencies [44]. With recent progress in the
semiconductor technologies such as BiCMOS (Bipolar Complementary Metal–Oxide–
Semiconductor) SiGe and CMOS (Complementary Metal–Oxide–Semiconductor), the
IBM group reports a cost-effective chip-scale packaging solution for 60-GHz wireless
chipsets capable of multi-gigabit per second wireless communication [45]. To further
improve the integration capability, the antenna, serving as the interface between the
external world and the integrated circuits, should also be integrated into the package. As
an important front-end component, the antenna for a 60 GHz system needs to be
wideband (to cover the entire 60 GHz ISM band), compatible to integrated circuits with
low interconnect loss, highly efficient, as well as low cost and a compact size. Moreover,
previous wireless channel propagation studies have shown that circular polarization can
effectively suppress multi-path fading and inter-symbol interference (ISI) and would be
desired [46-48].
1.2.3. Motivation
As part of this dissertation, we will present the design of a 60 GHz left-hand
circularly polarized antenna by combining various techniques to satisfy all the
26
requirements for the 60 GHz wireless communications. This patch antenna incorporates a
diagonal slot at the center and features a superstrate and an air cavity backing to achieve
desired performances including wide bandwidth, high efficiency and low axial ratio. The
metal frame underneath the antenna layer serves as the cavity backing, useful for antenna
bandwidth enhancement, as well as mechanical support for the antenna, making the
antenna much more stable than using a supporting pin (only feasible at low frequencies)
[49]. The microstrip-fed patch antenna is packaged with a flip-chip CPW interface that is
fully compatible with semiconductor integrated circuits (ICs). A prototype antenna is
fabricated and characterized using a probe-based measurement setup as described in [51],
demonstrating a 6-dB axial ratio bandwidth of 22.7%, which is much greater than the
1.1% axial ratio bandwidth in [50]. Moreover, this antenna achieves a wide bandwidth of
more than 26% and an efficiency of greater than 75% over the entire operating band. The
detailed design procedure of the circularly polarized antenna, including the design of the
microstrip-fed patch antenna and the comparison of the performances of the antenna with
different feeding interfaces, will be described in Chapter 4.
27
1.3. Novel Direction of Arrival Estimation Technique Inspired by Human Ears
In recent years, there has been increasing interest in microwave direction finding
systems due to their wide applications in military and commercial areas, such as
electronic warfare [52], wireless communications [53], etc. A wide variety of techniques
have been developed to estimate the direction of arrival (DOA) of the input signals,
including the estimation of signal parameters via rotational invariance techniques
(ESPRIT) [54] and multiple signal classification (MUSIC) [55], etc. All of these
techniques are based on a large number of antenna elements to achieve a high degree of
accuracy. However, as the number of antenna elements increases, power consumption,
and the size and cost of the RF circuits associated with the antenna arrays increase as well.
Therefore, an accurate DOA estimation technique with reduced number of antenna
elements is highly desirable.
A very interesting biological system that is capable of direction finding for
acoustic waves is the human auditory system, which has many intriguing abilities related
to direction finding, for example, estimating arrival angle with accuracy up to 1º under
binaural (utilizing two ears) conditions, sound source localization with a single ear
(mono-aural), etc [56]. As shown in Figure 1-4, the passive direction finding for
microwave signals is very analogous to the direction finding of acoustic waves by the
human auditory system. The microwave antennas are as analogous to the pinnae which
are natural directional antennas for acoustic waves. The band pass filters, mixers,
amplifiers and detectors provide similar functions as the guiding and detecting parts of
human ears. The signal processing component can be thought of as the human brain.
28
Even the wavelengths of audible sound waves are comparable to the microwave
wavelengths, for example, 3 KHz sound has a wavelength of 113 mm, corresponding to
the wavelength of a microwave signal around 3 GHz. This section provides a basic
background on the direction finding mechanisms of the human auditory system and their
inspirations for microwave passive direction finding applications.
Antenna
Pinnae
BPF
BPF
Detection
LO
Guiding, detecting
and processing
Signal Processing
and Decision
Figure 1-4. Comparison of microwave direction finding system (left) and the human
auditory system (right).
1.3.1. Binaural Direction Finding Mechanism
The remarkable direction finding abilities of the human auditory system have
been widely explored and experimentally demonstrated [56–61]. Within the audible
frequency range (20 Hz ~ 20 KHz), human ears are able to estimate sound directions with
an accuracy up to 1º, which is very remarkable considering there are only two pinnaes
(antennas).
For low frequency sound (f < 1.5 KHz), the wavelength of the signal is long
29
compared with the size of the human head, and thus the signal goes through/around the
human head and the phase difference between the signals received by the two ears serves
as the most important cue for sound localization. The typical distance between two
human ears is around 23 cm, which equals the wavelength of the 1.5 KHz sound waves.
The physical reason for this low frequency limit is to avoid the 2π phase ambiguity,
similar to the element spacing limit of antenna arrays. The front-back ambiguity is
eliminated by the directivity of human ears [56]. On the other hand, for higher frequency
sound (f > 1.5 KHz), where the wavelength is short compared to the head size, the
received signal of one ear is not influenced by the head while that of the other ear is
attenuated. The interaural differences in intensity are incident angle dependent and can
have an attenuation as much as 20 dB [58]. The combination of the phase and amplitude
information enables the human auditory system to have great localization abilities in both
low and high frequency ranges.
Both of the two binaural sound localization mechanisms can be directly applied to
microwave direction findings. The phase difference method for the low-frequency
scheme has been widely applied for RF direction finding. However, the method for the
high-frequency case using an effective low pass filter has not yet been reported for
microwave direction finding, at least to our knowledge. In this work, we utilize a lossy
scatterer between two antennas, which emulates the low-pass filtering function of the
human head at high frequency, to achieve accurate DOA estimation without phase
ambiguity for high frequencies signals, as shown in Figure 1-5. A simple 2-monopole
example is studied and the multiple signal classification (MUSIC) algorithm [55] is
30
applied to calculate the DOA. The improved estimation accuracy of the novel DOA
technique is demonstrated in both simulation and experiment. The detailed work will be
presented in Chapter 5.

l
LPF
d
Figure 1-5. A simple microwave direction finding scheme with two antennas
separated by a low-pass filter (LPF) like object mimicking the human head.
1.3.2. Monaural Direction Finding Mechanism and Inspiration
Another exciting and intriguing feature of the human auditory system is the
capability of monaural direction finding for broadband signals, for which the pinnae /
head and shoulder are natural directional antennas for acoustic waves [62-63], although
with reduced estimation accuracy compared to that of the binaural case. The main
mechanism for monaural direction finding is that the pinnae and head function as a combline filter with its frequency response depending on the incident angles. As a simplified
illustration shown in Figure 1-6, the received spectrum of a broadband signal by a single
ear is incident angle dependent, having a notch response depending on the incident angle.
31
θ
Detector
Spectrum Amp
θ dependent
notch response
Frequency
Figure 1-6. A simplified illustration of monaural direction finding mechanism: the
received spectrum has an incident angle dependent notch response.
In a similar way, the DOA of the microwave signal can also be estimated using a
single UWB antenna with direction-dependent spectra. Because the received spectra of
the UWB signals are different at different DOA, the DOA of the incident signal is then
estimated from the cross-correlation coefficients of the received spectra of the antenna
with the pre-determined incident-angle-dependent spectra. The estimation accuracy of the
DOA technique is investigated in all three planes with different Signal-to-Noise Ratios
(SNRs). Another modified UWB antenna with unsymmetrical shape is also studied for
the direction finding applications, demonstrating improved estimation accuracy due to the
break in symmetry. In addition, both the symmetric and unsymmetrical UWB antennas
are fabricated and tested. The measured results confirm the feasibility of the proposed
single antenna DOA technique, although with reduced accuracy in comaprison with the
32
two antennas and a scatterer case. Chapter 6 presents the detailed work of the direction of
arrival estimation technique using a single UWB antenna.
1.4. Dissertation Organization
The work presented in this dissertation includes the study of a metallic wire array
as an effective medium with near-zero index of refraction and its application in directive
antennas, the design technique and experimental results of a wideband circularly
polarized patch antenna for 60 GHz communications, and the investigation of the novel
direction of arrival (DOA) estimation technique inspired by human ears. The dissertation
is organized as the followings.
Chapter 2 summarizes the difficulties and issues with the effective parameters (εeff,
μeff and neff) extraction method based on the NRW approach. As we know, the extraction
process may fail in some instances, for example, when the reflection coefficient S11 is
close to 1 and the transmission coefficient S21 is close to 0, or when S11 is close to 0 and
S21 is close to 1, etc. The parameters extraction under these circumstances will be
discussed. The selection of the sign of the roots when solving for the medium impedance
Z, the determination of the branch of the real part of the refractive index n, and so on, will
also be described. Moreover, the reason for some of the unexpected effects during the
effective medium parameters retrieving process will be revealed using an S-shaped
metamaterial example.
33
Chapter 3 presents the study of the metallic wire array as low-effective index of
refraction medium and its application for directive antenna design. An algorithm is
developed based on the Nicolson-Ross-Weir (NRW) approach to extract the material’s
complex permittivity and permeability from the Ansoft HFSS simulated scattering
parameters (S-parameters), as described in Chapter 2. A simple design methodology for
directive monopole antennas is introduced by embedding a monopole within a metallic
wire array with neff < 1 at the antenna operating frequencies. The narrow beam effect of
the monopole antenna is demonstrated in both simulation and experiment at X-band. The
physical principles and interpretations of the directive monopole antenna embedded in
the wire array medium are also discussed. The manuscript of this work is first authored
by the dissertation author.
Chapter 4 describes the design of a wideband circularly polarized antenna for 60
GHz wireless communications. Various techniques are combined to improve the antenna
performances, such as wide bandwidth, high efficiency and low axial ratio. The detailed
design procedure of the circularly polarized antenna, including the design of the
microstrip-fed patch antenna and the comparison of the performances of the antenna with
different feeding interfaces, will be presented. This is a collaborative project with Dr. Liu
Duixian at IBM. The manuscript submitted to the IEEE Transactions on Antenna and
Propagation is first authored by the dissertation author.
Chapter 5 proposes a novel microwave direction finding technique inspired by the
human auditory system. The idea of this work is to utilize a lossy scatterer between two
antennas, which emulates the low-pass filtering function of the human head at high
34
frequencies, to achieve accurate DOA estimation without phase ambiguity for high
frequency signals. The MUSIC algorithm is applied to calculate the DOA based on a simple
2-monopole example. The effectiveness of the proposed technique is demonstrated in both
simulation and experiment. The manuscript is first authored by the dissertation author
and has been submitted to the IEEE Transactions on Antenna and Propagation.
Chapter 6 presents the direction of arrival estimation technique using a single
UWB antenna, inspired by the monaural sound localization of the human ear. Because
the received spectra of the UWB antenna are different for different DOA, the DOA of the
incident signal is estimated from the cross-correlation coefficients of the received spectra
of the antenna and the pre-determined incident-angle-dependent spectra. The estimation
accuracy of the DOA technique is investigated in all three planes with different Signal-toNoise Ratios (SNRs). The feasibility of the single antenna DOA technique is
demonstrated in both simulation and experiment, although with reduced accuracy,
compared to the two antennas with a scatterer case. The manuscript is first authored by
the dissertation author and will be submitted to the IEEE Antenna Wireless Propagation
Letters.
Finally, the conclusions and future work are discussed in Chapter 7.
35
CHAPTER 2.
EFFECTIVE MEDIUM PARAMETERS
EXTRACTION
Metamaterials have been realized and widely applied in microwave, infrared and
optical frequencies. There is also an increasing amount of numerical work on extracting
the effective medium parameters (i.e., relative permittivity ε = ε' - iε", relative
permeability μ = μ' - iμ", etc) of the metamaterial. For a finite slab of metamaterial, the
complex transmission and reflection coefficients are directly related to the refractive
index n and impedance z associated with the slab, which can in turn be expressed in terms
of the permittivity ε and permeability μ. A retrieval procedure can then be applied to find
the material parameters using the scattering parameters (S-parameters) based on the
NRW (Nicolson-Ross-Weir) approach. However, this retrieval process may fail in some
instances, such as when the reflection (S11) is very small and the transmission (S21) is
close to 1, or when the reflection (S11) is close to 1 and the transmission (S21) is very
small. In addition, the retrieval process uncovers some unexpected effects, for example,
whenever there is a resonance in ε', we simultaneously observe an anti-resonant behavior
in μ' and vice versa. The anti-resonant behavior in the real part of a constitutive parameter
is also accompanied by a negative imaginary part. In this chapter, an improved parameter
extraction method based on the NRW approach will be presented, including the selection
of the sign of the roots when solving for the impedance z and the determination of the
branch of the real part of refractive index n. Moreover, the reason for the resonant / antiresoant coupling will be revealed using an S-shaped metamaterial example.
36
2.1. Scattering Parameters
Scattering parameters, also known as S-parameters or S-matrix, are widely used at
microwave frequencies to describe an arbitrary electrical network. For a 2-port network
in Figure 2-1, the S-parameters are in relation to the incident and reflected voltage waves
by [64-65]
V
S  |
V

i
ij

Vk  0 , k  j
(i, j =1, 2)
(2.1)
j
where the superscript “+” denotes the incident voltage wave at port 1 or 2, and “-”
denotes the scattered voltage wave at port 1 or 2. From the definition, one can see that S11
and S22 are essentially the reflection coefficients of the network at port 1 and port 2,
respectively. S21 is essentially the transmission coefficient from port 1 to port 2; and S12 is
the transmission coefficient from port 2 to port 1. These four parameters form a matrix,
which is referred to as the S-matrix. For a passive network, the S-matrix is symmetric and
S21 = S12. For a symmetric network, S11 = S22 are equal to each other.
V1

1
V
Port 1
Port 2
V2
V 2
Figure 2-1. The incident and reflected waves of a two-port network.
37
As described in [66-67], the S-parameters of a plane wave normally incident on a
slab of homogeneous material are related to the permittivity ε and the permeability μ of
the material. Therefore, the Nicolson-Ross-Weir (NRW) approach has been implemented
to extract the constitutive parameters (the permittivity ε and the permeability μ) of the
material from the simulated or measured S-parameters [15, 66-73]. The details of the
NRW approach will be described in Section 2.2.
2.2. Nicolson-Ross-Weir Method for Parameter Extraction
Consider a plane wave incident normally on a homogeneous material with relative
permittivity ε, relative permeability μ, and thickness d embedded in free space. The
attenuation and phase delay introduced by a single pass through the material are
described by the transmission term
T  e  ikd  e  ik 0
 d
(2.2)
where ε = ε' - i ε" is the relative permittivity, μ = μ' - i μ" is the relative permeability, k0 is
the free space wave number, and k is the wave number in the material. The reflection on
the interface of the material and free space is given by

where  
  1
 1

 1
  1
(2.3)

is the normalized wave impedance of the material with relative

permittivity ε and relative permeability μ.
38
With the defined transmission term T and wave impedance  , the scattering
parameters can be derived as follows [15, 66-68]:
S11 
( 2  1)(1  T 2 )
(1   ) 2  (1   ) 2 T 2
(2.4)
S 21 
4T
(1   )  (1   ) 2 T 2
(2.5)
2
The Nicolson-Ross-Weir (NRW) approach introduces two composite terms [15, 66-68]
V1  S 21  S11
(2.6)
V2  S 21  S11
(2.7)
The appropriate combination of Eqs. (2.4), (2.5), (2.6) and (2.7) leads to a function only
in terms of T
X
1  V1V2 1  T 2

V1  V2
2T
(2.8)
Therefore, with the S-parameters, the transmission term T can be solved as
T  X  X 2 1
(2.9)
The proper sign in Eq. (2.9) is selected by applying the criteria that | T | 1 , which is the
physically meaningful result for any passive structure. As long as T is determined, one
can obtain the interface reflection coefficient Г and normalized impedance η as [15]:

T  V2
1  TV2
(2.10)
 1  T  1  V2 


 1  T  1  V2 
    
(2.11)
39
One thing to notice is that the two solutions for the transmission coefficient in Eq. (2.9)
are related as: T1  1 T2 . Therefore, according to Eq. (2.3) and Eq. (2.10), their
corresponding normalized impedances 1 and  2 are related as:
1   2
(2.12)
Hence, for the special case of the transmission coefficient | T1 || T2 | 1 , the correct
solution can be selected by applying the criteria of Re() > 0.
In principle, one should be able to obtain ε and μ from equation Eqs. (2.2), (2.9)
and (2.11) as long as T is determined. However, there is still an ambiguity to be clarified,
which is caused by the multi-valued function ln(T). Since T is a complex number, the
solution of Eq. (2.2) can be written as
n   
i
1
ln(T )  
[Im(ln(T ))  2m  i | ln(T ) |]
k0 d
k0 d
(2.13)
where Im(ln(T)) is the imaginary part of ln(T), and m can be any integer. Evidently, the
imaginary part of n is uniquely determined and the real part of n has infinite solutions due
to the phase ambiguity of the transmission term T. An effective way to resolve this
ambiguity is to start the extraction at a low frequency such that the sample thickness d is
much smaller compared to a wavelength such that it guarantees m = 0. For higher
frequencies, the continuity of n as a function of frequency is utilized to determine the
correct solution for n [68]. As an effective homogeneous medium, the extracted effective
parameters of the material slab should not depend on the number of unit cells.
40
2.3. S-shape Metamaterial
As was demonstrated in Section 2.2, the two solutions of the transmission term T
are the reciprocal of each other, and their corresponding impedances are opposite of each
other. Therefore, the NRW method to retrieve the effective parameters of passive
metamaterials can easily resolve the problems associated with the selection of the
solution for the transmission term T, the determination of the correct impedance η, and
the selection of the branch of the real part of refractive index n, etc. The effectiveness and
robustness of this method has been demonstrated in various metamaterials [34-35, 68, 7378]. However, the retrieval process reveals some unexpected effects, for instance, the
resonance in ε' is accompanied by the anti-resonance in μ' and negative μ", and vice versa.
To have an insightful understanding of these anomalous effects, a metamaterial
composed of S-shape is going to be constructed as an explicit example, and its effective
medium parameters will be extracted using the NRW approach described in Section 2.2.
The configuration of the S-shaped structure (perfect conductor) is shown in
Figure 2-2, where a = 260 um, b = 140 um, h = 20 um, d = 25 um, and the relative
dielectric constant of the substrate is εr = 3.8. The unit cell dimensions are g1 = 200 um,
g2 = 50 um, and g3 = 280 um. In order to extract its effective medium parameters, the
electromagnetic responses of the S-shaped metamaterial are simulated using the finiteelement electromagnetic solver HFSS. The normal-incident plane wave is propagating
along the x-axis, with the E field polarized along the z-axis and the H field polarized
along the y-axis. The S-shaped metamaterial is infinite in both z and y directions by
using appropriate perfect electric and magnetic conducting boundary conditions. Figure
41
2-3 plots the simulated S-parameters of this S-shaped metamaterial. Following the NRW
approach described in Section 2.2, the magnitude of the two solutions of the transmission
term T are shown in Figure 2-4(a), confirming the reciprocity of T1 and T2. The real parts
of their corresponding normalized impedances 1 and  2 are opposite of each other, as
plotted in Figure 2-4(b). Therefore, the correct solutions for T and  can be selected as T1
and 1 from 100 GHz to 271 GHz, and T2 and 2 from 272 GHz to 350 GHz, by applying
the criteria that Re() > 0 and abs(T) ≤ 1.
d
h
g3
d
a
vacuum
g2 ε = 3.8
vacuum
z
b
g1
y
x
(a)
(b)
Figure 2-2. (a) S-shaped metamaterial (b) Top view of one unit cell.
42
1
200
Phase (degree)
Magnitude
0.8
0.6
0.4
0.2
0
100
S11
S21
150
200
250
300
Frequency (GHz)
100
0
-100
-200
100
350
S11
S21
150
200
250
300
Frequency (GHz)
350
(a)
(b)
Figure 2-3. The simulated S-parameters (S11 and S21) of the S-shaped
structure (a) magnitude, (b) phases.
2
Real Part of impedance
Magnitude
1.5
1
T1
T2
0.5
0
100
150
200
250
300
Frequency (GHz)
350
z1
z2
1
0
-1
-2
100
150
200
250
300
Frequency (GHz)
350
(a)
(b)
Figure 2-4. (a) The magnitude of the two solutions of the transmission term T1 and
T2, (b) The real part of their corresponding normalized impedances 1 (z1) and 2
(z2).
The resulting real part of the refractive index n' for different m values in Eq. (2.13)
are plotted in Figure 2-5. Since the length of the S-shaped slab g1 = 200 um, which is
much smaller than the wavelength at the lowest frequency 100 GHz (3mm), it is
reasonable to assume that m=0 at 100 GHz. At higher frequencies, the m values are
selected to ensure the continuity of n' as a function of frequency. With the correct
43
solutions for the index of refraction n’ and normalized impedance , one can easily
obtain the effective permittivity
ε = n/
(2.14)
μ = n*.
(2.15)
and permeability
40
m=1
m=0
m=-1
m=-2
30
n'
20
10
0
-10
-20
100
150
200
250
300
Frequency (GHz)
350
Figure 2-5. Calculated real part of n of the S-shaped slab for different m values.
44
Permittivity
30
4
Real
Imaginary
Permeability
40
20
10
0
-10
100
150
200
250
300
Frequency (GHz)
350
2
Real
Imaginary
0
-2
-4
100
150
200
250
300
Frequency (GHz)
350
(a)
(b)
Figure 2-6. Extracted effective parameters of the S-shaped slab, (a) relative
permittivity ε (solid line: real part ε', dashed line: imaginary part ε''), (b) relative
permeability μ (solid line: real part μ', dashed line: imaginary part μ'').
The extracted ε and μ of the S-shaped structure as functions of the frequency are plotted
in Figure 2-6 (a) and Figure 2-6 (b), respectively. The resonant behavior of the real part
of the effective permittivity ε' at frequency f0 = 231 GHz is clearly visible, and the
imaginary part of the permittivity ε'' is greater or equal to zero as expected. On the other
hand, the frequency dependence of the effective permeability μ is anti-resonant. Its real
part μ' decreases to zero at f0 = 231 GHz and its imaginary part μ'' is negative from 231
GHz ~ 271 GHz (the shaded area in Figure 2-6).
45
Index of Refraction
4
3
Real
2
Imaginary
Brillourin zone edge
1
0
100
150
200
250
300
Frequency (GHz)
350
Figure 2-7. The retrieved index of refraction as a function of frequency: real part n'
(solid line) and imaginary part n'' (dashed line). The dash-dotted line in n' indicates
the upper edge of the first Brillouin zone, nedge =π/(k0*g1).
The anti-resonant behavior of the effective parameters has its origin in the finite
lattice periodicity of the metamaterial structure [77-78]. When the wavelength of the
electromagnetic wave is comparable with the lattice periodicity, the effects of the
periodicity of the structure set in and the metamaterial cannot be simply approximated
using the homogeneous effective medium model. As plotted in Figure 2-7, the real part of
the effective index of refraction n' is confined at the edge of the first Brillouin zone nedge
=π/(k0*g1), corresponding to the appearance of additional band gaps (as shown in
Figure 2-4 (a)) originating from the periodicity rather than from the underlying material
properties. With the index of refraction confined to the edge of the first Brillouin zone,
the permeability μ' has to go to zero when the permittivity ε' exposes resonant behavior.
In addition, the anti-resonance in μ' is also accompanied by the negative value in the
46
imaginary part of permittivity μ'', which seems to be in contradiction to our physical
intuition. However, according to Eq. (2-14) and (2-15), we can obtain that
 '' ''
1
[(n' ' ' ) 2  (n' ' ' ) 2 ]
2
| |
(2-16)
One sees that the sign of ε''μ'' is fully determined by the right hand side of Eq. (2-16). In
fact, the plot in Figure 2-4 (b) shows that ' ≈ 0 and |n'''| < | n'''| within the shaded
frequency range. As a result, the product of ε'' and μ'' in Eq. (2-16) is negative, leading to
the negative μ'' as shown in the shaded area in Figure 2-6 (b). Another explanation for the
negative μ'' is to treat the metamaterial as a lossless system without conduction loss or
dielectric loss, which is indeed the case in the simulation setup. Therefore, for a planar
electromagnetic wave normally incident on the metamaterial, the total dissipated energy
must be zero, given as
W
1
d[ ' ' ( ) | E ( ) |2   ' '| H ( ) |2 ]  0 .
4 
(2.17)
As a result, the imaginary part of permeability μ'' must be negative with a positive ε'', and
vice versa.
2.4. Conclusion
In this chapter, we presented an efficient method for extracting the effective
constitutive parameters (permittivity ε and permeability μ) of a slab of metamaterial using
its S-parameters. The effectiveness of this method has been demonstrated using a
47
metamaterial composed of an S-shaped structure. The NRW based parameter retrieval
method is summarized as follows:
(1) The two solutions of the transmission term T are reciprocal to each other (T1 =
1/T2). The real parts of their corresponding normalized impedances are opposite of each
other (1   2 ). Therefore, when ' is close to 0, the requirement that ' ≥ 0 can not be
directly applied for the retrieval, because the numerical or measurement errors may flip
the sign of ', making the results unreliable. In this case, the correct solutions of  can be
easily selected by the value of its corresponding T so that |T| <1. On the other hand, when
|T| is close to 1, the requirement of |T| ≤1 can not be directly applied for the retrieval
and the correct solutions for T can be selected by the value of its corresponding  so that
' > 0.
(2) The real part of n has an infinite number of solutions due to the phase
ambiguity of the transmission term T. An effective way to resolve this ambiguity is to
start the extraction at a low frequency such that the sample thickness d is much smaller
comparing to wavelength such that it guarantees m = 0. For higher frequencies, the
continuity of n as a function of frequency is utilized to determine the correct solution for
n.
(3) When the effective wavelength of the electromagnetic wave is comparable
with the lattice periodicity, the metamaterial cannot be simply approximated using the
homogeneous effective medium model. The extraction results show that the resonance in
ε' is accompanied by an anti-resonance in μ', and vice versa. The anti-resonant behavior
of the effective parameters is due to the finite lattice periodicity of a metamaterial
48
structure. Because of the finite periodicity of a metamaterial, it cannot support arbitrarily
large wave numbers and the real part of the effective index of refraction n' is confined at
the edges of the Brillouin zones, corresponding to the appearance of additional band gaps
originating from the periodicity rather than from the underlying material properties. With
the confined index of refraction, the permeability μ' has to go to zero when the
permittivity ε' exposes resonant behavior, and vice versa.
(4) The fact that the imaginary part of either the permittivity or permeability is
negative seems to contradict our physical intuition. However, according to Eq. (2-16), the
product of ε'' and μ'' must be negative when |n'''| < | n'''|, indicating that either ε'' or μ'' is
negative. In fact, the negative ε'' or μ'' does not violate the passivity of the material as
long as the total dissipated power W in Eq. (2-17) is greater or equal to zero. Actually, the
zero dissipation energy for a lossless system requires that the μ'' be in an opposite sign
with ε'', which agrees with the negative μ'' or ε''.
49
CHAPTER 3.
METALLIC WIRE ARRAY AS A LOW-EFFECTIVE
INDIXE OF REFRACTION MEDIUM [35]
In this chapter, two-dimensional (2-D) metallic wire arrays are studied as
effective media with an index of refraction less than unity (neff < 1). The effective
medium parameters (permittivity εeff, permeability μeff and index of refraction neff) of a
wire array are extracted from the finite-element simulated scattering parameters and
verified through a 2-D electromagnetic crystal (EMXT) structure case study. A simple
design methodology for directive monopole antennas is then introduced by embedding a
monopole within a metallic wire array with neff < 1 at the antenna operating frequencies.
Parametric studies of the antenna system are performed. The narrow beam effect of the
monopole antenna is demonstrated in both simulation and experiment at X-band (8 – 12
GHz). The physical principles and interpretations of the directive monopole antenna
embedded in the wire array medium are also discussed.
3.1. Metallic Wire Array as Effective Medium
The effective permittivity εeff of the 2-D periodic metallic wire arrays in Figure 12 can be calculated using plasma theory with a reduced electron density [2], as shown in
Eq. (1.1):  eff  1   p2  2  1  2c 2 /[ a 2 ln(a / r ) 2 ] , where a is the wire array
periodicity, r is the wire radius, ω is the angular frequency, c is the speed of light in free
space and ωp is the plasma frequency, at which the effective permittivity εeff, and thus the
effective index of refraction neff, is zero.
50
An important assumption of Eq. (1.1) is that the wires are very thin (r << a) so
that the plasma frequency corresponds to a free space wavelength much greater than the
lattice spacing and the Bragg diffraction effect can be ignored [2]. A useful metamaterial
should keep its effective medium properties when used as a constituent of a composite
structure. The metallic wire array itself in Figure 1-2 can alternatively be decomposed
into two identical but independent square lattices with lattice constants of
2a ,
embedded within each other, as shown in Figure 3-1. Therefore, the effective permittivity
can be calculated by assuming that the second lattice (represented by solid circles) is
embedded in a medium with the effective permittivity ε1 of the first lattice (represented
by empty circles).
2r
a
2a
Wires
Figure 3-1. Top view of a 2-D square lattice of wires with radius r and periodicity a.
It can also be thought of as two independent square lattices (solid and empty circles)
with a periodicity of 2a , embedded within each other.
The overall effective permittivity is then:
51
 eff  1 (1   p2 2 /  2 )  1 (1 
2 c 2 / 1
2 c 2


)
1
( 2a) 2 ln( 2a / r )   2
a 2 ln(a / r )   2  a 2 ln 2   2
(3.1).
Equations (1.1) and (3.1) are very close to each other under the thin wire condition, in
which case ln(a / r )  ln( 2)  0.35 . As an explicit example, Figure 3-2 plots the
calculated effective permittivity of a wire array with a = 4 mm and r = 1 m (ln(a/r) =
8.3) using Eqs (1.1) and (3.1). The almost exact agreement of the two methods indicates
that the plasma theory is self-consistent and a metallic wire array may be used as an
effective medium with special properties such as very small or near-zero permittivity.
Furthermore, the near-zero permittivity of a wire array medium can be realized in any
microwave frequency by adjusting the wire array spacing and the wire radius according
to Eq. (1.1).
1
0
eff
-1
One - lattice
Two - lattice
-2
-3
-4
5
10
15
20
25
Frequency (GHz)
Figure 3-2. Calculated effective permittivity of a 2-D wire array (a = 4 mm, r = 1
µm) using plasma theory, treating the array as either a single square lattice (circles)
or two square lattices (squares) embedded within each other.
52
The above analysis is simple and effective in terms of predicting the plasma
frequency. However, it neglects magnetic response and loss of the wires, which may be
important in practical applications. In order to extract its effective medium parameters
accurately, the exact electromagnetic responses of a wire array medium are simulated
using finite-element electromagnetic solvers (i.e., Ansoft’s HFSS). The finite-element
model consists of five unit cells along the propagation direction (x-axis), as shown in
Figure 3-3. The normal-incident plane wave is polarized along the wires (z-axis). The
wire array is infinite in both the z and y directions by using appropriate perfect electric
and magnetic conducting boundary conditions. The Nicolson-Ross-Weir (NRW)
approach [15, 66-68] is then implemented to extract the effective medium parameters (neff,
εeff and µeff) from the simulated scattering parameters (S-parameters). As an effective
homogeneous medium, the extracted effective parameters of the wire array slab should
not depend on the number of unit cells, as will be shown in section 3.3.2.
z
x
2r
a
a
y
Figure 3-3. Top view of the HFSS model of a wire array structure. A normalincident plane wave propagates in the x direction and the array is infinite in both
the z and y directions.
53
3.2. A 2-D EMXT in a Low Index Wire Array Host
As it is predicted in Section 3.1, the effective index of refraction of a 2-D metallic
wire array is close to zero around its plasma frequency, which can be realized at any
microwave frequency by adjusting the wire array spacing and the wire radius. In this
section, a 2-D square lattice of dielectric rods embedded within a background of such
wire array is investigated to validate the description of the wire array as an effectively
small index of refraction medium. The top view of the dielectric rod / wire array structure
is shown in Figure 3-4. In this case, the host medium of the 2-D EMXT is not the typical
free space but a low refractive index medium.
Dielectric Rod:
Wire :
2r=0.5mm
E
k
a=5m
m
a=5mm
4mm
H
10mm
Figure 3-4. A two-dimensional dielectric rod EMXT structure embedded in a low
index of refraction wire array (top view).
The 2-D metallic wire array used here has a periodicity a = 5 mm and a wire
radius r = 0.25 mm. Following the NRW approach [15, 66-68], the finite-element
simulated S-parameters of the wire array model as illustrated in Figure 3-3 are utilized to
54
extract the effective parameters of the wire array. The resulting real part of the refractive
index
n'
for
different
branches
(indexed
by
the
m
values)
( n    i ln(T )   1 {Im[ln(T )]  2m  i Re[ln(T )]} , where k0 is the free space wave
k0 d
k0 d
number, T  e ink 0 d is the transmission term and d is the thickness of the sample under
study along the wave propagation direction) from 1 GHz to 30 GHz is plotted in Figure
3-5 (beyond 30 GHz, the Bragg effect sets in and the extracted effective medium
parameters are not physically meaningful). Since the length of the wire medium slab
containing five unit cells is d = 5a = 25 mm in the wave propagation direction, which is
much smaller than the wavelength at the lowest frequency 1 GHz (300 mm), it is
reasonable to assume that m = 0 at 1 GHz. At higher frequencies, the m values are
selected to ensure the continuity of n´ as a function of frequency. After the m values are
determined, one can easily obtain the effective ε and μ of the wire array, as plotted (the
real components) in Figure 3-6 along with the calculated εeff using Eq. (1.1).
5.0
m=2
m=1
m=0
m=-1
m=-2
m=-3
n
2.5
0
-2.5
-5.0
0
10
20
30
Frequency (GHz)
Figure 3-5. Calculated real part of n of the wire array with a =5 mm and r = 0.25
mm for different branches (m values).
55
Relative  and 
10
0
-10
Calculated 
Extracted 
Extracted 
-20
-30
0
10
20
30
Frequency (GHz)
Figure 3-6. Extracted real components of ε and μ compared with the ε calculated
from Eq. (1.1) (a = 5 mm, r = 0.25 mm).
As shown in Figure 3-6, the calculated and extracted ε’ have similar frequency
trends, both going from negative values at low frequencies to positive values at high
frequencies. The extracted values from the finite-element simulation have more
complicated features above 30 GHz (not plotted) and do not approach 1 as the simple
plasma theory predicts. The breakdown of the simple plasma picture is expected, since
above 30 GHz, the wire spacing becomes greater than the half free space wavelength and
the Bragg diffraction effect sets in. In addition, the extracted μ’ are not exactly equal to 1
due to the finite radius of the wires. Instead, μ’ is slightly greater than 1 below 30 GHz
and smaller than 1 (can even be negative) at higher frequencies, again due to the Bragg
diffraction. Moreover, for certain frequency range (i.e., 18 - 30 GHz), the extracted ε has
small positive values (ranges from 0.005 to 0.34) such that n’ is less than 1.
56
Embedded in the wire array discussed above, the dielectric rods shown in Figure
3-4 have a dielectric constant of 2.53, a periodicity of 10 mm and a radius of 2 mm. For
an incident plane wave polarized along the length of the rods (see Figure 3-4), frequency
band gaps exist, within which no transmission is allowed. The first band gap center
frequencies and bandwidths of this dielectric rod array are calculated using the MIT
Photonic-Bands (MPB) package [79] and are listed in Table 3-1 for backgrounds with
different permittivity ε. It can be seen that with the background permittivity decreasing
from  = 1 (free space) to  = 0.1, both the first band gap center frequency and its
bandwidth increase significantly. The physical reason for the increased band gap
frequency is that for a smaller background permittivity, the effective wavelength becomes
longer and the electrical length of the dielectric rod spacing becomes smaller.
Table 3-1. The 2-D Dielectric Rods Band Gap Frequency and Bandwidth for
Different Backgrounds
Background
permittivity ε
ε = 0.1
ε = 0.2
ε = 0.3
ε = 0.4
Band gap center
frequency
24 GHz
23 GHz
21 GHz
20 GHz
Bandwidth
12.2 GHz
9.2 GHz
6.2 GHz
4.1 GHz
Therefore, when the dielectric rods are embedded within a wire array with εeff < 1
as shown in Figure 3-4, it is expected that the first band gap would occur at a higher
frequency although the exact frequency and bandwidth are hard to predict due to the
frequency dependency of the wire array properties. To verify the effective medium
parameters of the wire array, the results of two HFSS simulations are compared in Figure
57
3-7: the curve of squares is the transmission response of the actual composite of dielectric
rods and metallic wires while the curve of circles is that of the dielectric rods embedded
in a uniform background assigned to have the extracted frequency dependent complex
permittivity and permeability of the wire array. Two important observations can be made
from Figure 3-7. First, the two simulation results agree quite well for frequencies below
30 GHz, confirming the properness of treating the wire array as a homogeneous effective
medium. Second, both of the transmission responses show the first band gap at around 20
GHz with a bandwidth around 6 GHz, which agrees with our expectations, indicating a
smaller than 1 index of refraction of the wire medium. In addition, the two simulated
results are quite different for frequencies above 30 GHz because the Bragg diffraction
effect sets in and the effective medium approach is no longer valid.
S21 (dB)
0
-50
Actual
Eff. Med.
-100
-150
0
10
20
30
40
50
Frequency (GHZ)
Figure 3-7. Simulated transmissions of the actual composite of the dielectric
rods and the metallic wire array (squares) and the dielectric rods embedded
within a uniform slab with the extracted effective medium parameters of the
wire array (circles).
58
3.3. Directive Monopole Antenna Embedded in a Wire Array Medium
A very interesting application of low-index of refraction media is highly directive
antennas. Several studies concerning directive antenna design using metallic wires or
other EMXT structures have been reported. It is shown that the radiation of a source can
be confined within a small angular region by embedding the source within two combined
EMXT structures (one in reflection and the other in transmission) [80-82]. Another
technique for antenna gain enhancement is to use a superstrate composed of EMXT
structures [83]-[85]. In addition, EMXT based Fabry-Perot cavities can be used to realize
directive antennas [86]-[88]. In this section, a simple design methodology for narrow
beam antennas is introduced by embedding a monopole antenna within a wire array
medium with neff < 1. The narrow beam effect due to the low refractive index of the wire
medium is demonstrated in both simulation and experiment.
3.3.1. Principles of Antenna Operation
When a source is placed in a medium with a refractive index n2 < 1, as shown in
Figure 3-8, at the boundary between this medium and free space, assuming simple
geometric optics applies, Snell’s law (n1 sinθ1 = n2 sinθ2, n1 = 1) requires the refractive
angle θ1 to be less than θ2, indicating that the radiated beam will be refracted towards the
normal direction of the interface and a beam narrowing effect can be achieved. Therefore,
a directive antenna should be obtained by embedding an omni-directional antenna within
a metallic wire medium with neff < 1 at the antenna operating frequencies.
59
Figure 3-8. Geometrical illustration of the beam narrowing effect for a source
embedded in an n2 < 1 medium.
The top view of a monopole / wire array system is shown in Figure 3-9, where
the monopole antenna is surrounded by a finite number of wires (11 columns and 11 rows
in this case). Both the wires and the monopole are along the z direction. By placing two
perfect conducting ground planes at the ends of the wire array, it can be considered
infinitely long in the z-direction. To minimize the near field interaction between the
monopole and the surrounding wires, the center element of the wire array is removed and
replaced by the monopole.
60
Monopole
antenna
y
x
Figure 3-9. Top view of a monopole embedded within a wire array that is
infinite in the z-direction.
3.3.2. An X-band Directive Monopole/Wire Array Antenna Design
A monopole / wire array antenna as shown in Figure 3-9 is designed to achieve
narrow-beam radiation within the X-band (8 to 12 GHz). The 11-column x 11-row copper
wire array has a periodicity a = 9 mm and a wire radius r = 0.25 mm. The wires have a
length of 50 mm and are terminated at the ends by two copper ground planes. Following
the procedures described in Section II, the extracted n' and n" (n = n' + i n") as a function
of frequency at X-band from finite-element simulations of a 5-unit cell slab are shown in
Figure 3-10. The extraction results show that the plasma frequency of the wire array is
around 8.7 GHz, which is not too far from the calculated plasma frequency (7.02 GHz)
using Eq. (1.1). The effective n' of this wire array is less than 0.75 for the entire X-band.
The effective n" is close to zero from 8.7 to 12 GHz, indicating that this wire array can be
thought of as a lossless medium in that frequency range. Similar wire arrays with
different number of unit cells (from 2 to 6) are also simulated and their effective material
parameters are extracted, yielding the same values as the 5-unit cell case. This confirms
61
that there is no interaction between the unit cells and the wire array may be treated
effectively as a standalone uniform slab of material.
0.75
n & n
0.50
n
n
0.25
0
-0.25
-0.50
8
9
10
11
12
Frequency (GHz)
Figure 3-10. The extracted n' and n" (n = n' + i n") of a small refractive index wire
array medium for X-band applications (a = 9.0 mm and r = 0.25 mm).
A coax-fed monopole with a length L of 17.4 mm is placed at the center of the
wire array as indicated in Figure 3-9. If the wire array can be treated as an ideal lowindex medium, the resonance frequency f of the monopole would satisfy L = 17.4 mm ~
λeff /4 = c / (4 f n'), where c is the speed of light, and n' is the extracted frequency
dependent index of refraction as shown in Figure 3-10, in which case it would be 9.5
GHz (see dashed line in Figure 3-11 (a)). However, the simulated antenna return loss
shows multiple resonances within the X-band (solid line in Figure 3-11 (a)) which are
due to the interactions between the monopole and the finite wire array. The simulated
radiation patterns confirm the expected narrow beam effect due to the wire array’s less
than unity index of refraction. The maximum gain of the antenna in the x-y plane occurs
at 9.5 GHz with a magnitude of 10.4 dB (see Figure 3-11 (b)). The 10-dB beam width
62
(with four-fold symmetry as expected) of the antenna at 9.5 GHz is 37◦. At this frequency,
the extracted refractive index neff of the wire medium is 0.42. Thus, Snell’s law predicts a
beam width of 2θ1 = 2sin-1[(n2 sin θ2)/ n1] = 2sin-1 [(0.42  sin 45◦)/1] = 34.6◦, which is
consistent with the simulated 10-dB beam width (+/- 45◦ are the limits of 2). In addition,
the simulated radiation efficiency is higher than 95%.
90
0
15
120
60
With wire array
Without wire array
10
5
0
-5
30
150
S11 (dB)
-5
(dB)
-10
-15 180
-10
-5
Actual Wires
Eff Med.
0
5
-15
0
-10
330
210
10
8
9
10
11
12
13
15
240
300
270
Frequency (GHz)
(a)
(b)
Figure 3-11. (a) Simulated return losses of a 17.4-mm monopole in the actual wire
array (solid line) and in a uniform medium with extracted effective permittivity and
permeability representing the wire array (dashed line). (b) Simulated gain (in dB) of
the monopole in the wire array (solid line) comparing with the case without the wire
array (dashed line). The wire array has the following dimensions: a = 9 mm and r =
0.25 mm; and the 11 x 11 wires are terminated by two conducting ground planes
separated by 50 mm.
3.3.3. Parametric Study of the Monopole Antenna System
There are several important parameters that may influence the behavior of the
monopole / wire array system such as the monopole length, the wire array size and the
wire array height (distance between the top and bottom metal plates). Parametric studies
63
are performed to provide insights into how these parameters affect the radiation
characteristics of the antenna system.
3.3.3.1. Monopole Length
The resonance frequency of the monopole antenna system is expected to be f = c /
(4n'L) if the wire array surrounding it can be treated as a standalone homogeneous
effective medium. If this is true, it should vary slower as a function of the monopole
length L than a linear dependence because n' increases with frequency. However, as
shown in Figure 3-11 (a), for L = 17.4 mm, there are multiple resonances instead of just
the predicted one at 9.5 GHz. To evaluate the length effect, monopoles with lengths of
7.5 mm, 10.0 mm, 12.5 mm, 15.0 mm and 17.4 mm embedded in the same 11-column x
11-row wire array as described in the previous section are studied. The simulated return
losses are shown in Figure 3-12 (for clarity, only 3 curves are plotted). It can be seen that
there are multiple resonance frequencies for all of the monopole lengths due to
interactions between the monopole and the finite wire array, as mentioned previously. In
addition, the resonance frequencies are quite similar for different monopole lengths from
7.5 to 17.4 mm although the detailed levels of return losses are not exactly the same.
Furthermore, the radiation patterns across the entire X-band are almost identical for all
the monopoles with different lengths, confirming that the wire array medium determines
the radiation pattern. The maximum gain in the x-y plane always occurs at 9.5 GHz with a
magnitude of about 10.4 dB.
64
0
S11 (dB)
-5
-10
-15
-20
-25
L = 7.5mm
L = 12.5mm
L = 17.4mm
8
9
10
11
12
Frequency (GHz)
Figure 3-12. The simulated return losses of the monopole / wire array system for
various monopole lengths.
3.3.3.2. Wire Array Size
As it is mentioned above, the radiation patterns are the same for different
monopole lengths when embedded within a fixed wire array size. However, for different
wire array sizes, the antenna radiation patterns are different due to the finite sizes of the
wire array and the interactions between the wires and the monopole. The simulation
results of the same monopole as described in Section 3.3.2 embedded within 11 x 11, 9 x
9 and 7 x 7 square wire arrays show that the maximum gain of the monopole in the x-y
plane increases when the wire array size increases (from 5 dB to 10.4 dB), which is
related to the antenna aperture size (see discussion in Section 3.3.5). In addition, the
resonance frequencies of the antenna are close to each other for different sizes of wire
arrays.
65
3.3.3.3. Wire Array Height
The wires are terminated by two ground planes to emulate an infinitely long wire
medium. The impact of the wire array height (or the separation of the ground planes) is
also studied. Simulations show that for a 10-mm long monopole embedded in the 11 x
11 wire array with a height of 30, 40 and 50 mm, the resonance frequencies shift up
slightly (less than 1 GHz) with the decrease of the wire array height. The maximum gain
of the antenna also decreases slightly (about 2 dB) as the wire array height decreases.
From these observations, it can be concluded that even though the upper ground plane
interacts with the monopole, the interaction is quite small when the distance between the
ground planes is significantly larger than the monopole length.
3.3.4. Experimental Verification
To verify the theoretical prediction of the narrow beam effect, a prototype of the
proposed antenna system operating within the X-band is designed, fabricated and
measured. The schematic top view of the antenna is the same as that shown in Figure 3-9
(a monopole embedded at the center of an 11 x 11 wire array) and a photo of the
fabricated prototype is shown in Figure 3-13. The monopole has a length of 17.4 mm and
is fed by a 50 Ω coaxial connector. Two copper plates of size 110 mm x 110 mm are used
as ground planes. Copper wires are then soldered onto the ground planes. The separation
between the ground planes is 50 mm. This antenna configuration should lead to directive
radiation in the x-y plane instead of the omni-directional radiation pattern of a regular
monopole.
66
Figure 3-13. A photo of the fabricated antenna prototype. The monopole and wire
lengths are 17.4 mm and 50.0 mm, respectively.
The prototype antenna in Figure 3-13 is characterized using an Agilent E8361A
vector network analyzer in an anechoic chamber. The measured and simulated return
losses agree quantitatively well, both showing multiple resonances, as plotted in Figure 314.
0
S11 (dB)
-5
-10
-15
Simulation
Measurement
-20
-25
8
9
10
11
12
Frequency (GHz)
Figure 3-14. Comparison of the measured (circles) and simulated (solid line) return
losses (S11) of the 17.4 mm monopole antenna embedded in the wire array.
67
The measured radiation patterns of the antenna system at various resonance
frequencies (8.5, 9.5, 10.3 and 11.3 GHz) are plotted in Figure 3-15, together with the
HFSS simulated radiation patterns of the actual monopole / wire array system and the
monopole embedded in an effective medium assigned to have the extracted permittivity
and permeability of the wire array. It is very interesting that the two simulation results
agree excellently (except for some small deviations at 11.3 GHz), which verifies the
legitimacy of treating the wire array as an effective medium in this case. The agreement
between measurements and simulations is also quite good as shown in Figure 3-15. The
narrow beam effect of the antenna system is apparent in both simulations and
measurements at all the frequencies. The simulated antenna gains are 7.8 dB, 10.4 dB, 7.4
dB and 5dB at 8.5 GHz, 9.5 GHz, 10.3 GHz and 11.3 GHz, respectively. The simulated
and measured 10-dB beam width of the monopole at 9.5 GHz is 37◦ and 34◦, respectively,
being consistent with Snell’s law estimation as discussed before.
The four main beams in Figure 3-15 are due to the symmetry of the antenna
system. If a single main beam is desired, an effective medium with anisotropic property,
for instance, a wire array with different periodicities in the four directions in x-y plane,
can be applied. Another way is to add metallic reflectors at the three boundaries of the
wire arrays as reported in [87]. Simulation and measurement results show that with three
metallic reflectors installed on three sides of the antenna system shown in Figure 3-13, a
single main beam can be achieved with a gain of about 6 dB higher compared to the case
without the metallic reflectors, as expected.
68
90
0
120
Simulation (Effective Medium)
Simulation (Actual)
Measurement
60
90
0
-5
-5
-10
-10
-15
30
150
-25
-30 180
0
-25
-30 180
0
-25
-20
-10
30
150
-20
-25
-15
120
-15
-20
-20
-15
330
210
240
0
330
210
-10
-5
-5
300
240
0
270
300
270
(a)
90
0
120
(b)
Simulation (Effective Medium)
Simulation (Actual)
Measurement
60
-5
-10
-15
90
0
-10
30
150
-15
-25
-30 180
0
-30 180
-25
-25
-20
-20
330
210
-15
-10
0
330
210
-5
-5
0
30
150
-20
-25
-10
120
Simulation (Effective Medium)
Simulation (Actual)
Measurement
60
-5
-20
-15
Simulation (Effective Medium)
Simulation (Actual)
Measurement
60
240
300
270
0
240
300
270
(c)
(d)
Figure 3-15. Measured x-y plane radiation patterns (normalized to 0 dB) of the
antenna system (solid line) compared with the simulated radiation patterns of the
monopole in the actual wire array (dashed line) and in the effective medium
(dashed-dotted line) at: (a) 8.5 GHz, (b) 9.5 GHz, (c) 10.3 GHz and (d) 11.3 GHz.
69
The radiation in the x-z plane is significantly shielded by the two ground planes,
as shown in Figure 3-16. The ideal ∞-shaped radiation is truncated near the z-direction.
However, if the x-z plane radiation pattern is desired, the upper ground plane may be
eliminated. In this case the radiation in the x-z plane will be increased significantly while
the narrow beams in the x-y plane still remain, although with slightly lower gain (plot not
shown).
Measurement
Simulation
90
0
120
60
-5
-10
-15
30
150
-20
-25
-30 180
0
-25
-20
-15
-10
330
210
-5
0
240
300
270
Figure 3-16. The simulated and measured radiation patterns (normalized to 0 dB) of
the antenna in the x-z plane at 9.5 GHz.
3.3.5.
Discussion
As it is pointed out in Section 3.3.1, the narrow beam effect of the antenna system
can be explained using Snell’s law by treating the wire array as an effective medium with
neff < 1. Our studies on the band gaps of the 2-D dielectric rod electromagnetic crystal
and the radiation patterns of the monopole embedded in the wire array show that the
effective medium approach is quite accurate and provides a very useful design
methodology, at least when the unit cell dimension is much smaller than the operating
70
wavelength. However, it is important to point out an approximation used here that the
longitudinal component of the incident waves is omitted (reasonable approximation for
the monopole antenna configuration in this work) so that the spatial dispersion of the wire
medium is not accounted for [89].
Another important thing worth mentioning is that the Snell’s law estimation of the
narrow beam effect is also an approximation and care should be taken when applying it.
For example, one may expect that the beam width decreases with neff and becomes equal
to zero when neff is zero, which is certainly not the case from Figure 3-15. The reason is
that Snell’s law assumes plane wave incidence (or far field from its source) while the
wire array has a finite size (110 mm x 110 mm) which becomes electrically smaller for
smaller neff so that the far field assumption is no longer valid. Another physical
explanation of the directive radiation is to treat the antenna system as four identical
aperture antennas with a size of 110 mm x 50 mm. Thus, the first null beam width of the
aperture antenna is the minimum when the E-field over the whole aperture is constant,
and can be calculated as: 2θmin = 2sin-1(λ/l), where λ is the wavelength and l = 110 mm is
the aperture length. As an example, the calculated minimum first null beam width at 9.5
GHz is 33.4◦, which is very close to the Snell’s law prediction and the simulated and
measured 10-dB beam width. Therefore, it can be concluded that uniform excitation of
the aperture is achieved by this monopole / wire array system.
71
3.4. Conclusion
2-D metallic wire arrays as low-index (neff < 1) effective media are investigated
for directive antenna application. The self-consistency of the plasma model describing the
wire array is demonstrated. Effective medium parameters (εeff , μeff and neff) of 2-D wire
arrays are extracted based on finite-element simulations and the results are consistent
with the plasma theory prediction. An EMXT structure made of 2-D dielectric rods
embedded in a wire array is studied to validate the low-index property of the wire array.
A simple design methodology for a directive monopole antenna is then introduced by
embedding a monopole within a wire array with neff < 1 at the antenna working
frequencies. A monopole / wire array antenna (working within the X-band) is designed,
fabricated and characterized. Parametric studies of this antenna system show that the
antenna resonance frequencies and radiation patterns are not sensitive to the monopole
length L, and the antenna gain increases when the wire array size increases. Experimental
results of the fabricated antenna prototype agree well with simulation results, confirming
the narrow beam effect of the antenna system. The design procedure of this antenna
system is flexible and can be applied at all microwave frequencies by adjusting the wire
array spacing and wire radius.
72
CHAPTER 4.
A WIDEBAND CIRCULARLY POLARIZED PATCH
ANTENNA FOR 60GHz WIRELESS COMMUNICATIONS [90]
Demands of modern communication and sensor systems for more bandwidth,
higher resolution and compactness lead to operating frequencies up to the millimeter
wave (mmWave, f > 30 GHz) or even sub-mmWave regime. However, at mmWave
frequencies, higher material losses, fabrication tolerances and packaging issues often
hinder the performance of wireless front-ends. Therefore, as an important front-end
component, mmWave antennas that are wideband, efficient and packaged to be
compatible with integrated circuits are highly desirable. In this chapter, we will present
the design of a fully packaged wideband circularly polarized patch antenna for 60 GHz
wireless communications. This patch antenna incorporates a diagonal slot at the center
and features a superstrate and an air cavity backing to achieve desired performances
including wide bandwidth, high efficiency and low axial ratio. The detailed design
procedures of the circularly polarized antenna are explored. The performances of the
antennas with different feeding interfaces are compared. The experimental results of the
final packaged antenna are also presented.
4.1. Introduction
As it is discussed in Chapter 1, the wideband unlicensed spectrum around the 60
GHz ISM band is very attractive for a number of applications, including wireless
73
personal-area networks (WPAN), local-area networks (WLAN), multimedia streaming,
and file transfer between personal-computers and portable devices [37-38]. Moreover, as
the battery lifetime is a major bottleneck for many portable devices, the 60 GHz band has
a distinct benefit because the large bandwidth can be utilized to tradeoff bandwidth
efficiency for low power consumption, while still maintaining high data rates [38]. In
addition, due to the atmospheric absorption peak around 60 GHz (mainly oxygen
molecules), communication links at this band are inherently secure and has less
interference among users, which are ideal for indoor applications. However, there are a
lot of technological challenges associated with a 60 GHz system. First, the small
wavelength at 60 GHz requires high-precision machining and accurate alignment. In
addition, mm-Wave circuits are usually assembled using expensive and bulky
waveguides with a low level of integration [40-43]. With the recent progress in
semiconductor technologies like BiCMOS (Bipolar Complementary Metal–Oxide–
Semiconductor), SiGe and CMOS (Complementary Metal–Oxide–Semiconductor), the
IBM group reports a cost-effective chip-scale packaging solution for 60-GHz wireless
chipsets capable of multi-gigabit per second wireless communication [45]. As an
important front-end component of the 60 GHz communication system, the mmWave
antenna that are wide band (to cover the whole 60 GHz ISM band), high efficiency, low
cost, compact size and compatible with integrated circuits with low interconnect losses
are highly desirable. Moreover, previous wireless channel propagation studies have
shown that circular polarization can effectively suppress multi-path fading and intersymbol interference (ISI) [46-48].
74
4.1.1. Techniques for Improving Antenna Performances
Various techniques have been reported in the literature to improve antenna
performances such as antenna bandwidth and efficiency. As shown in [91], cavity
backing can isolate an antenna from its surroundings, suppress its surface waves and
increase the antenna bandwidth. The detailed characteristics of various cavity backed
antennas are summarized in [92]. Another approach to improve the bandwidth and
efficiency of a conventional patch antenna is to minimize the substrate dielectric constant,
as demonstrated in [93] and [94]. It is also known that the antenna gain can be
considerably increased by covering the antenna with a high permittivity superstrate [9596]. The fundamental effects of the substrate-superstrate structure on printed-circuit
antennas are explored in [97], demonstrating that the antenna radiation efficiency can be
optimized to approximately 100% by selecting the proper materials and dimensions of the
substrate and superstrate.
All of the above mentioned techniques for improving the antenna performances
can be applied in a mmWave antenna design. Several cavity backed wide-band antennas
have been demonstrated in the V-band in [98] – [100]. In addition, metamaterial / EBG
(electromagnetic band gap structures) based antennas with increased antenna gain have
been reported in [101] – [103]. A good review of various mmWave integrated circuit
antennas can be found in [104]. Furthermore, several topologies of circularly polarized
antenna have been proposed for the 60 GHz wireless communications [105] – [109].
75
4.1.2. Motivation
In this chapter, we present the design of a 60 GHz left-hand circular polarized
antenna that has a good overall performance by combining various techniques reported in
[45], [51], [91-100] and [110]. This patch antenna incorporates a diagonal slot at the
center and features a superstrate and an air cavity backing to achieve the desired
performances including wide bandwidth, high efficiency and low axial ratio. The metal
frame underneath the antenna layer serves as the cavity backing, useful for antenna
bandwidth enhancement, as well as mechanical support for the antenna, making the
antenna much more stable than using a supporting pin (only feasible at low frequencies)
[93]. The microstrip-fed patch antenna is packaged with a flip-chip CPW interface that is
fully compatible with semiconductor integrated circuits (ICs). In [45], [51], [110], linear
polarized folded dipole antennas for 60 GHz applications using this superstrate / substrate
topology and packaging scheme have been demonstrated. In this work, by optimizing the
slot length of the patch antenna, a 6-dB axial ratio bandwidth of 22.7% is achieved,
which is much greater than the 1.1% axial ratio bandwidth in [50]. Moreover, this
antenna achieves a wide bandwidth of more than 26%.
A prototype antenna is fabricated and characterized using a probe-based
measurement setup as described in [51]. The demonstrated patch has a dimension of
1050-m x 1050-m, while the final packaged antenna including feed lines has a
dimension of 4849-m x 5555-m. The detailed design procedure of the circularly
polarized antenna, including the design of the microstrip-fed patch antenna and a
comparison of the performances of the antenna with different feeding interfaces, is
76
described. The measured antenna efficiency is also obtained, demonstrating greater than
75% efficiency over the entire measured frequency range. Discussions on the
measurement uncertainties that may explain the simulation / measurement discrepancies
are included.
Top Ring
Antenna and
feed line Layer
Fused Silica
Top Ring
EN
C
4
4.2. Antenna Structure/Packaging
500 µm
Cavity
ENC4
Air
Copper Cavity
38
9
48
9
0
µm
0
µm
49
38
µm
49
48
µm
(a)
(b)
Figure 4-1. (a) The side view of the entire antenna structure. (b) The copper cavity.
Since packaging at mmWave frequencies is challenging and critical to the system
performance [45], mmWave antenna designs should incorporate the packaging aspects as
early as possible. The geometry of the entire antenna structure, including its packaging,
is shown in Figure 4-1(a). It consists of a top fused silica substrate (superstrate), a bottom
air cavity made of copper and dielectric sealant encapsulant with a dielectric constant of 4
(ENC4) for further integration with flip-chipped ICs [45]. Similar package structures
have been applied previously for folded dipole antennas. More detailed package design
information can be found in [45], [110]. The applied cavity size is optimized according
77
to the antenna impedance bandwidth.
The superstrate thickness T = 300 µm is a
compromise since thinner fused silica samples would be difficult to manufacture and a
thicker superstrate leads to a reduced efficiency due to more energy staying in the
dielectric layer as surface waves. All of the metallization (2-m thick gold) of the patch
antenna and its feed lines is on the lower side of the fused silica substrate, which has a
negligible loss tangent at 60 GHz and a relatively low dielectric constant of 3.8. To
minimize the surface wave effects that can affect the radiation in the antenna plane, a
gold ring is placed on the upper side of the fused silica. The fused silica substrate is
mounted on top of an air cavity with copper walls, as shown in Figure 4-1(b). The cavity
has an internal size of 3890-m x 3849-m and an external size of 4890-m x 4849-m.
The cavity height is 500-m, being a trade-off between the package size and the resulting
antenna bandwidth. The detailed parameters of the antenna structure and packaging are
listed in Table 4-1.
Table 4-1. Configuration of the Antenna Structure and Packaging
Fused Silica (εr = 3.8, tan δ =0.0)
ENC4 (εr = 4, tan δ =0.02)
Air (εr = 1, tan δ =0.0)
Top Ring (gold) (width = 500 m)
Size (m2)
4849 x 5555
4849 x 665
3849 x 3890
4849 x 4890
Thickness (m)
300
600
500
2
The antenna structure shown in Figure 4-1 has several advantages. First, a patch
antenna at the lower side of the fused silica can be thought as having an air substrate with
a relative dielectric constant of 1.0 and a superstrate with a dielectric constant of 3.8.
78
These will lead to significant bandwidth and efficiency improvements compared with a
normal one substrate patch antenna. In [110], a broadband dipole antenna for 60 GHz
applications using a similar cavity and superstrate was reported. Second, this packaging
scheme is fully compatible with flip-chip mounted monolithic integrated circuits such as
an entire SiGe transceiver [45]. A schematic of the antenna flip-chip integrated with an
IC chipset is shown in Figure 4-2.
60 GHz Chip
Feed Line
Opening for
Feed Line
Low Frequency
Wire bond
Antenna Structures
Fused Silica
Air Cavity
PCB
Flip-chip
Attachment
Sealant
GND
Metal Cavity Walls
Figure 4-2. Schematic view of the antenna packaged with an integrated chipset.
4.3. Detailed Antenna Design
For the initial design, a microstrip-fed square patch antenna with a diagonal slot is
applied to achieve left-hand circular polarization [50], [111], as shown in Figure 4-3.
The patch size is first selected to obtain a resonant frequency around 61 GHz and the
microstrip line width is tuned to match the antenna impedance. The slot length mainly
controls the circular polarization performance. It is observed that the frequency at which
the axial ratio is minimum increases when the slot length C increases (10 GHz when C =
79
L/2, 51 GHz when C = L/1.1 and 75 GHz when C = L/0.9), with the fixed patch size of
1400 µm and the slot width d kept at C/10 [112]. In addition, it is found that the slot
length C also influences the antenna resonance frequency: the resonance frequency
decreases when the slot length increases (with the fixed patch size and the slot width kept
at C/10). Figure 4-4 compares the simulated antenna reflection coefficients (S11 in dB)
for different slot lengths with the square patch size L = 1400 µm. The antenna resonance
frequency shifts from 62 GHz to 57 GHz when the slot length C increases from L/2 to
L/0.9. The final dimensions of the patch antenna are optimized to make both the resonant
frequency and the frequency with the minimum axial ratio at 61GHz. The finalized
design parameters are listed in Table 4-2.
Fused Silica
300 µm
C
L
d
L
Ground
plane
Air gap
500 µm
Figure 4-3. The microstrip-fed square patch antenna with a diagonal slot.
80
S11 (dB)
0
-10
C=L/0.9
C=L/1.1
C=L/2
C=L/2.72
-20
-30
55
60
65
70
75
Frequency (GHz)
Figure 4-4. Simulated reflection coefficients (S11 in dB) of the square patch antenna
(L = 1400 µm) with different slot lengths C (slot width d = C/10).
Table 4-2. The Dimensions of the Finalized Patch Antenna Design
Antenna length
Antenna width
Slot length
Slot width
Microstrip line width
L = 1050 µm
L = 1050 µm
C = 1141 µm
d = 114 µm
W0 = 400 µm
The simulated reflection coefficients (S11 in dB) and the axial ratio of the
optimized antenna are plotted in Figure 4-5. It can be seen that S11 is less than -10 dB
from 59 GHz to 75 GHz, with a bandwidth of about 24%. The axial ratio is less than 6
dB from 58 GHz to 72 GHz, with a bandwidth of 21.5%. Comparing with conventional
patch antennas, the bandwidth of the optimized antenna is much greater due to the
utilization of the air cavity and the fused silica superstrate. The simulated antenna
efficiency is higher than 90% for the entire frequency range from 50 to 75 GHz.
81
0
S11 (dB)
-5
-10
-15
-20
-25
50
55
60
65
70
75
Frequency (GHz)
(a)
Axial Ratio (dB)
40
30
20
10
6 dB
0
50
55
60
65
70
75
Frequency (GHz)
(b)
Figure 4-5. Simulated microstrip-fed square patch antenna performance: (a)
Reflection coefficients (S11 in dB). (b) Axial ratio.
A CPW to microstrip transition is then designed to satisfy the CPW interface of
the antenna packaging and enable convenient measurements using a CPW probe. The
transition section is composed of a tapered microstrip line and a CPW line (labeled
CPW4) with two approximately λ/4 open-stubs to achieve the desired grounding at the
transition [113] around the center frequency of 61 GHz, as shown in Figure 4-6(a) with
82
straight open stubs and Figure 4-6(b) with bent open stubs. The CPW4 section is
designed to have the same impedance (102 Ω) as the microstrip. Two additional CPW
lines (labeled CPW3 and CPW2) are used to transform the 102 Ω CPW line (CPW4) to
the input 50 Ω CPW line (labeled CPW1).
L
CPW1
CPW2 CPW3
Taper
Patch antenna
1
CPW4
(a)
L2
CPW1
CPW2
CPW3
CPW4
Taper
Patch antenna
L3
(b)
Figure 4-6. The top view of the antenna layer including (a) the CPW feed with
straight open-stubs; (b) the CPW feed with bent open-stubs.
To evaluate the impact of the CPW packaging, the HFSS simulated performances
of the antennas with direct microstrip feed and CPW feeds with straight open-stubs and
bended open-stubs are compared in Figure 4-7. The simulated antenna efficiencies are
greater than 90% for all the cases. The reflection coefficients and axial ratios of the three
83
cases are similar at frequencies around 61 GHz (58 GHz ~ 64 GHz), as expected. With
the CPW feed sections, the antenna has a wider bandwidth and a narrower axial ratio
bandwidth compared to those of the direct microstrip feed. The S11 of the antennas with
the CPW feeds are less than 10 dB from 50 GHz to 70 GHz with a bandwidth of 33%,
which are larger than the 17% bandwidth of the antenna with the microstrip feed. The 6dB axial ratio bandwidths of the antennas with the microstrip feed, the CPW feeds with
straight open-stubs and bended open-stubs are 18.7%, 9%, and 9.9%, respectively. In
combination, the bandwidth with both S11 less than -10 dB and the axial ratio less than 6
dB drops from 17% to 9% and 9.9% when the CPW feed sections are added to match the
CPW interface, which is still sufficient for the applications at 60 GHz. In addition, the
axial ratios of the antenna with bended open-stubs are around 1 dB lower than the case
with straight open-stubs at frequencies around 61 GHz. Therefore, to reduce the impact
of the λ/4 open-stubs on the radiation characteristics (especially the axial ratio) of the
antenna, the bent λ/4 open-stubs are used for the final design. The final dimensions of the
feed sections are listed in Table 4-3, where L is the transmission line length, W is the
transmission line width, g is the CPW gap, and W1 and W2 are the widths of the two
ends of the tapered microstrip. The widths of the two CPW grounds are 150 µm each.
84
0
Microstrip feeding
CPW feeding (straight stub)
CPW feeding (bended stub)
30
20
10
Microstrip feeding
CPW feeding (straight stub)
CPW feeding (bended stub)
-5
S11 (dB)
Axial Ratio (dB)
40
-10
-15
-20
0
50
55
60
65
70
-25
50
Frequency (GHz)
55
60
65
70
Frequency (GHz)
(a)
(b)
Figure 4-7. Comparison of antenna performances with different feedings (a) axial
ratio (b) reflection coefficient (S11 in dB)
Table 4-3. The Geometry of the CPW to Microstrip Transition
CPW1
CPW2
CPW3
CPW4
λ /4 open-stub
Taper
Microstrip
L = 665 µm, W = 128 µm, g = 31 µm
L = 500 µm, W = 124 µm, g = 33 µm
L = 631 µm, W = 56µm, g = 67 µm
L = 800 µm, W = 50 µm, g = 70 µm
L1 = 700 µm, L2 = 500 µm, L3 = 350 µm
L = 340 µm, W1 = 50 µm, W2 = 400 µm
L = 100 µm, W = 400 µm
4.4. Comparison of Measurement and Simulation
To validate the performances of the designed 60 GHz left-hand circular polarized
antenna, a prototype is fabricated by a thin-film resolution metal patterning process. The
detailed fabrication procedure can be found in [45], [110]. Figure 4-8 shows a photo of
the fabricated and packaged antenna. Note that the antenna layer on the bottom side of
the fused silica can be clearly seen because of the transparency of the fused silica
superstrate.
85
Figure 4-8. A photo showing the top view of the fabricated and packaged 60 GHz
left-hand circular polarized patch antenna.
The antenna performances, including return loss, axial ratio, gain and radiation
pattern, are characterized from 50 GHz to 65 GHz using a CPW probe-based
measurement setup in an anechoic chamber at IBM. To achieve an accurate measurement
of the antenna parameters, the measurement setup is calibrated in two steps. First, the
gain is calibrated by replacing the device under test (DUT) by a known standard gain
horn antenna. The second calibration is the SOL (Short, Open, Load) calibration to obtain
the calibrated return loss using the on-wafer short, open and load standards. The gain
calibration is done first because it requires more changes in the setup (probe has to be
replaced by the standard gain horn) and the gain calibration stays valid for a much longer
time than the SOL calibration. The detailed measurement procedures are described in
[51]. To measure the axial ratio of the fabricated antenna, both a straight and a twisted
waveguide adapter are used to feed the receiving standard gain horn. The challenge of the
axial ratio measurement at 60 GHz is that it is very sensitive to the phases of the two
86
polarizations. For example, when the straight waveguide adapter is replaced by the
twisted one or vice versa, small misalignment would lead to the inaccuracy of the
measured axial ratio.
Figure 4-9 plots the simulated and measured S11 of the fully packaged antenna.
The measured result indicates a 10-dB bandwidth greater than 26% (limited by the
experimental frequency range), agreeing well with the simulation result. The measured
axial ratio is plotted in Figure 4-10, together with the simulated result. The simulated 6dB axial ratio bandwidth is 9.9%, with a minimum value of 2.7 dB at 61 GHz. The
measured minimum axial ratio (3.0 dB) also occurs around 61 GHz and the 6-dB axial
ratio bandwidth is 22.7%. The measurement and simulation results agree well above 60
GHz but the measured values are significantly smaller at the lower frequencies. The
discrepancy between the simulated and measured axial ratios is likely due to the
measurement uncertainties discussed previously.
The radiation characteristics of the antenna are measured in both the x-z and y-z
planes (see Figure 4-8). The simulated and measured antenna gains (x-z plane) as a
function of frequency are shown in Figure 4-11(a). They match reasonably well with
each other near the center frequency (at 61 GHz, the measured and simulated gains are
6.4 dB and 7.3 dB, respectively). The measured gain in the y-z plane (not shown) also has
a similar shape and range but is distorted a bit due to the interference of the antenna feeds
in the measurement. Our simulation shows that the antenna efficiency is greater than
90% within the simulated frequency range, as the dashed curve plotted in Figure 4-10(b).
The measured antenna efficiency ηmea is then estimated as:  mea   sim * Gmea Gsim (with the
87
assumption that the measured and simulated directivities are consistent), where  sim is the
simulated antenna efficiency, Gsim and Gmea are the simulated and measured antenna
gains, respectively. It can be seen from Figure 4-10(b) that the measured efficiency
(solid curve) varies from 75% to 92% from 57 to 64 GHz.
-10
-12
-14
S11 (dB)
-16
-18
-20
Simulation
Measurement
-22
-24
-26
50
55
60
Frequency (GHz)
65
70
Figure 4-9. The simulated and measured reflection coefficients (S11 in dB) of the
fully packaged antenna.
25
Axial ratio (dB)
20
Simulation
Measurement
15
10
5
0
50
55
60
Frequency (GHz)
65
70
Figure 4-10. The simulated and measured axial ratio of the fully packaged antenna.
88
8
100
7
Antenna efficiency (%)
Gain (dB)
6
5
4
3
Simulation
Measurement
2
50
Simulation
Measurement
1
0
50
55
60
Frequency (GHz)
65
0
50
70
55
60
Frequency (GHz)
65
70
10
10
5
5
0
0
Radiation pattern (dB)
Radiation pattern (dB)
(a)
(b)
Figure 4-11. The simulated and measured antenna gain (a) and antenna efficiency
(b) of the fully packaged antenna.
-5
-10
-15
Simulation
Measurement
-20
-25
-15
Simulation
Measurement
-20
-25
-30
-30
-35
-200
-5
-10
-100
0
Theta (degree)
100
200
-35
-200
-100
0
Theta (degree)
100
200
(a)
(b)
Figure 4-12. The simulated and measured radiation patterns of the fully packaged
antenna at 61 GHz in: (a) y-z plane (b) x-z plane.
89
10
10
0
5
Co-polarized
Cross-polarized
Radiation pattern (dB)
Radiation pattern (dB)
5
-5
-10
-15
-20
-25
-30
-35
-200
0
-5
Co-polarized
Cross-polarized
-10
-15
-20
-25
-30
-150
-100
Theta (degree)
-50
0
-35
-200
-150
-100
Theta (degree)
-50
0
(a)
(b)
Figure 4-13. The measured co-polarized and cross-polarized radiation patterns of
the fully packaged antenna at 61GHz in: (a) y-z plane (b) x-z plane
Finally, reasonable agreements between the simulated and measured radiation
patterns in both the x-z and y-z planes are observed, as shown in Figure 4-12. The
measured patterns away from the broadside direction in the x-z plane show some
discrepancy, probably due to measurement inaccuracies. The measured co-polarized and
cross-polarized radiation patterns at 61 GHz are compared in Figure 4-13. One thing to
notice is that the broadside gains in the y-z and x-z planes are different by about 1.4 dB.
This discrepancy is also believed to be caused by experimental uncertainties.
4.5. Conclusion
A fully packaged left-hand circularly polarized antenna for 60 GHz wireless
communications is demonstrated. Wide bandwidth and high efficiency are achieved by
utilizing an air cavity and a fused-silica superstrate. The antenna packaging with a CPW
interface is compatible with semiconductor integrated circuits. The measured antenna
properties including return loss, axial ratio, gain and radiation patterns agree reasonably
90
well with the simulation results. The demonstrated antenna achieved an impedance
bandwidth greater 26%, a 6-dB axial ratio bandwidth of 22.7% and efficiency above 75%
for the entire frequency range.
91
CHAPTER 5.
IMPROVED TWO-ANTENNA DIRECTION FINDING
INSPIRED BY HUMAN EARS [114]
A microwave antenna is also an important front-end component for microwave
direction finding systems, which has been of increasing interest in recent years due to its
wide applications in the military and commercial areas, such as electronic warfare [52],
wireless communications [53]. This chapter presents an improved two-antenna
microwave passive direction finding system inspired by the human auditory system. By
incorporating a head-like scatterer between two monopole antennas, both phase and
magnitude information can be utilized to estimate the direction of arrival (DOA) of a
microwave signal, thus eliminating ambiguities associated with phase wrapping at high
frequency, just like human ears. In addition, better DOA sensitivity is demonstrated with
the incorporation of the head-like scatterer. Both numerical model and experimental
results of a simple X-band two-monopole configuration with symmetric and asymmetric
scatterers are presented.
5.1. Motivation
Microwave passive direction finding is a very important technology that has many
military and commercial applications including electronic warfare [52], mobile
communications [53]. A typical microwave direction finding system may use an array
consisting of a large number of antenna elements and sophisticated algorithms to achieve
a high degree of accuracy [52]. However, the size, weight, speed and cost associated
92
with the large number of hardware components and the complicated signal processing
can be impractical, especially for portable and commercial applications. For example, a
hand-held direction finding gadget that is capable of identifying and locating the source
of various RF signals is highly desirable for a soldier in a battlefield. In addition,
accurate and efficient direction finding will be very useful in next generation wireless
systems for location based services and applications.
A very interesting biological system that is capable of direction finding for
acoustic waves is the human auditory system, which includes a pair of pinnae located at
each side of the human head to collect acoustic signals, ear canals, eardrums, and cochlea
to guide and detect incoming acoustic signals, and auditory nerves and neurons in the
brain to process the detected signals. In fact, human ears have many amazing and
intriguing abilities related to direction finding, among them, estimating arrival angle with
accuracy up to 1° without ambiguity under binaural (utilizing two ears) conditions, the
famous “cocktail party phenomenon”, the ability of “learning the surrounding
environment within seconds”, and sound source localization with a single ear (monoaural), to name a few [56].
In this chapter, we propose a novel DOA estimation technique using only two
antennas, which is inspired by the human auditory system. The idea is to utilize a lossy
scatterer between the two antennas, which emulates the low-pass filtering function of the
human head at high frequency [56], to achieve a more accurate DOA estimation. A
simple 2-monopole example at X-band frequency (8 to 12 GHz) is studied and the
multiple signal classification (MUSIC) algorithm [55] is applied to calculate the DOA.
93
Our theoretical and experimental studies have shown that by incorporating a head-like
scatterer between the two antennas, not only high frequency ambiguity associated with
phase wrapping is eliminated, but also the DOA sensitivity can be improved significantly.
5.2. Analogy between Human Sound Localization and Microwave Direction
Finding
Passive direction finding (or sometimes referred to as direction of arrival, DOA,
estimation) for microwave signal is very analogous to the direction finding of acoustic
wave by human ears. Even the wavelengths of audible sound waves are comparable to
that of microwave frequencies, for example, 3 KHz sound has a wavelength of 113 mm
which is about the wavelength of a 2.65 GHz RF signal. Figure 5-1 illustrates the
analogy between a microwave direction finding system and the human auditory system.
The microwave antennas are similar to the pinnae, which are natural directional antennas
for acoustic waves; the band-pass filters, amplifiers, mixers and detectors provide similar
functions as the guiding and detecting parts of the human auditory system; and the signal
processing component can be thought as the brain.
94
Antenna
Pinnae
BPF
BPF
Detection
LO
Guiding, detecting
and processing
Signal Processing
and Decision
Figure 5-1. Comparison of a passive microwave direction finding system (left) and
the human auditory system (right).
The remarkable localization (mainly in the azimuth plane) capabilities of human
ears for both continuous waves (CW) and transient signals have long been recognized
and studied quite extensively [56-61].
Many intriguing facts and phenomena were
experimentally observed and underlying mechanisms were proposed and proved. As
early as 1936, Stevens and Newman reported free space experimental data on localization
of sound sources by human ears which revealed the two main mechanisms of binaural
sound localization, one operating best at high frequency and the other at low frequency
[59]. Later on, more studies in anechoic chambers confirmed the earlier results [60-61].
For most of the audible frequency range (20 Hz – 20 KHz), human ears are able to
estimate sound sources with up to 1° of accuracy (this is rather impressive considering
there are only two “antennas” - pinnae).
For low frequency sound (f < 1.5 – 3.0 KHz), the phase difference between the
acoustic signals received by the two ears (referred to as the binaural case) serves as the
most important cue. The typical distance between two human ears is about 23 cm which
95
is equal to the wavelength of a 1.5 KHz sound wave (speed of sound in air is 340 m/s).
The physical reason of this low frequency limit is similar to the antenna element spacing
limitation for RF direction finding or phased array antennas: to avoid the phase ambiguity
of multiples of 2π, the antenna elements should be spaced less than half a wavelength,
/2. The front-back ambiguity is eliminated by the directivity of human ears (analogous
to an antenna radiation pattern) [56]. For higher frequency sound (f > 3.0 KHz), the
human hearing system incorporates a simple but elegant solution: using the head as a
low-pass filter. For most incident angles, one ear receives without the influence of the
head while the other receives after the incident signal goes through (or around) the lowpass filter - human head, whose response function is incident angle dependent and can
have an attenuation as much as 20 dB [58]. This effect is often referred to as the headrelated-transfer function (HRTF), which provides important cues for sound source
localization for high frequencies. Figure 5-2 illustrates the HRTF effect, which leads to
both a phase and a magnitude difference between the received signals at two ears. The
combination of the phase (or time for transient signals) and amplitude information
enables the human auditory system to have great localization capabilities for both low
and high frequency ranges.
96
Figure 5-2. Utilizing the HRTF, the human auditory system can achieve
unambiguous direction finding for high frequency signals
Both of the binaural mechanisms mentioned above have analogies or may be
directly applied to microwave systems. In this chapter, we confine our discussions in the
azimuth plane for simplicity without losing generality.
The low-frequency phase
difference method is widely used in microwave direction finding [52]. However, the
high-frequency scheme utilizing an effective low-pass filter (the shadowing effect of
human head) has not been reported in the literature, at least to our knowledge. By
introducing a carefully designed scatterer in between two adjacent antennas, accurate
direction finding without phase ambiguity for high frequency signals may be achieved.
Furthermore, because of the spacing between the adjacent antenna elements can now be
much larger than /2, the mutual coupling issue that is common to antenna array systems
can be greatly reduced.
97
5.3. Numerical Simulations
To evaluate the feasibility of applying some of the human sound localization
mechanisms in microwave direction finding, a simple two-antenna (omni-directional
monopoles are used here for simplicity) configuration is considered.
Take an
electromagnetic (EM) signal coming from an azimuth direction  that is impinging on
two monopole antennas separated by a distance d, as shown in Figure 5-3. The phase
difference  between the received signals at these two antennas is,  =
d sin 

2 ,
where λ is the wavelength of the signal. It can be derived that for both EM and acoustic
DOA, if d is greater than a half wavelength, there may be ambiguities in the estimated
DOA. To avoid this kind of ambiguity at high frequency, it is proposed here that a lossy
scatterer is placed between the two antennas, providing the similar low-pass filtering
function as the human head between two ears. Without the scatterer, the phase difference
 of the signals measured at the two antennas is the key information to estimate the
DOA. With the scatterer, the magnitude difference M provides an additional important
cue in the DOA estimation, eliminating the phase ambiguity issue.
98
Absorber
Figure 5-3. A finite-element model illustrating the geometry of the two-monopole
and scatterer configuration with an incoming signal from an azimuth angle.
The simple configuration in Figure 5-3 (with / out a lossy scatterer) working at Xband (8 – 12 GHz; monopole length is 7 mm with a center frequency near 10 GHz; d =
15 mm = /2 at 10 GHz) is modeled with the full-wave finite-element EM solver HighFrequency-Structure-Simulator (HFSS). The scatterer used here is a cubic material with
a geometry of 15 mm x 12.8 mm x 10 mm and assigned to have r = 2 and tan = 20. The
steering vectors (phase and magnitude differences) as a function of frequency and
incident angle with all the EM environments included are obtained by two different
methods: far-field illumination by a horn antenna and direct plane-wave excitation. The
phase and magnitude differences ( and M) received by the two monopoles without a
scatterer in between are plotted in Figure 5-4 for incident angles  = 30°, 60° and 90°. It
can be seen that the two methods are consistent. The plane-wave excitation method is
99
then employed to sweep the entire azimuth dimension from  = 0° to 360° since it is
much less computationally intensive.
(a)
(b)
Figure 5-4. Simulated phase (a) and magnitude (b) differences of two monopole
antennas without a scatterer in between using two different methods: horn antenna
illumination – individual markers and plane wave excitation - lines.
Figure 5-5 plots the simulated  and M at 10 and 12 GHz, with and without
the lossy scatterer. Several important points are worth noting. First, the magnitude
difference is much larger with the presence of the lossy scatterer, as expected. Second,
the phase difference versus incident angle curve with the scatterer is significantly steeper
than that without the scatterer, which should lead to higher sensitivity in the DOA
estimation.
For example, with the scatterer, a given error in the measured phase
difference will cause a smaller error in the calculated DOA compared to the case without
the scatterer. Third, although the phase ambiguity issue seems to be more severe for the
case with the scatterer, for example, the same phase difference may lead to four different
incident angles, the large magnitude differences should be sufficient to overcome the
100
ambiguities. In addition, both the phase and magnitude differences with and without the
scatterer are the same for supplementary incident angles (or, it can be thought of as the
front / back ambiguity), which is due to the symmetry respect to the y-axis (see Figure 53). A simple solution for this issue is to shift the monopoles positions so that the system
becomes asymmetric.This can be achieved by using directional antennas just as well just
as the human ears. In the following examples, the symmetry is broken by shifting the
two monopoles toward one side of the scatterer block by 7.5 mm.
10 GHz
400
3
Without scatterer
With scatterer
2
Magnitude Difference (dB)
350
Phase Difference (deg)
10 GHz
300
250
200
150
100
30
Without scatterer
With scatterer
20
1
10
0
0
-1
-10
-2
-20
Magnitude Difference (dB)
450
50
45
90 135 180 225 270
Incident Angle  ( deg )
315
-3
0
360
45
90 135 180 225 270
Incident Angle  ( deg )
(a)
Phase Difference (deg)
400
12 GHz
0.4
Without scatterer
With scatterer
350
300
250
200
150
100
12 GHz
30
Without scatterer
With scatterer
45
90 135 180 225 270
Incident Angle  ( deg )
315
360
20
0.2
10
0
0
-10
-0.2
-20
50
0
0
-30
360
(b)
M agnitude D ifference (dB )
450
315
-0.4
0
45
90
135 180 225 270 315
Incident Angle  ( deg )
M agnitude D ifference (dB )
0
0
-30
360
(c)
(d)
Figure 5-5. Simulated phase (a) and magnitude (b) differences at 10 GHz and phase
(c) and magnitude (d) differences at 12 GHz with and without a lossy scatterer in
between.
101
Several scatterer materials with various dielectric constants and loss tangents are
studied. Although different phase and magnitude differences ( and M) are obtained
for each case, all of the important features are the same. In the following examples, a
lossy foam absorber material, ARC-LS-10211 (made by ARC Technologies Inc.), with r
= 2.05 and tan = 1.05 and dimensions of 15 mm x 12.8 mm x 10 mm, is used. This
material is adequate to provide sufficient magnitude difference information even though
it has much lower loss than the scatterer used in the previous example.
To evaluate the DOA performance of the two-monopole / scatter system, phase
and magnitude differences between the two monopoles as a function of incident angle
from 0° to 360° are first simulated and saved as the calibration steering vectors. Then,
for an arbitrary incident angle , assuming a phase error (2°), the multiple signal
classification (MUSIC) algorithm (for the details about the MUSIC algorithm, please
refer to Appendix I) is used to estimate the DOA [55]. In Figure 5-6, the MUSIC output
for the simulated case with a 12 GHz signal coming from the 80° direction is plotted for
the three scenarios mentioned above (without scatterer, symmetric scatterer and
asymmetric scatterer). It is clear that the addition of the attenuating scatterer (similar to
the low-pass-filter function of the human head) in between the two antennas eliminates
the ambiguity caused by the 2 phase wrapping. Moreover, it can be observed from
Figure 5-6 that the asymmetrically placed scatterer breaks the front / back symmetry and
eliminates the corresponding ambiguity as well. For the no scatterer case, there are four
peaks near 80° (100°) and 222° (319°) that can be observed, for which the 80° and 222°
ambiguity is due to the greater than /2 spacing of the monopoles. With the scatterer,
102
this ambiguity is clearly eliminated by the added magnitude information, which can be
seen from both the symmetric and asymmetric cases in Figure 5-6. The 80° and 100°
ambiguity manifested in the no scatterer and symmetric scatterer cases is due to the front
/ back symmetry of the two-antenna system, which is absent in the asymmetric
configuration.
Spectral Magnitude (dB)
50
Without scatterer
With symmetric scatterer
With asymmetric scatterer
40
30
20
10
0
-10
0
45
90 135 180 225 270 315
Incident Angle  (deg)
360
Figure 5-6. Simulated MUSIC output of the three scenarios with a signal incident
from the 80° direction: without a scatterer (dotted dashed line), with a
symmetrically positioned rectangular scatterer (dashed line) and with an
asymmetrically positioned rectangular scatterer (solid line).
To verify the predicted accuracy enhancement by the addition of a head like
scatterer, average DOA estimation errors as functions of frequency and incident angle are
evaluated, assuming the ambiguities can be resolved. For example, assumed errors (+/2° in phase and +/- 0.25 dB in magnitude) are first introduced to the simulated steering
vectors; then the modified steering vectors are fed back to the MUSIC algorithm to find
the average incident angles which are compared to the original correct incident angles.
103
Numerical studies for signals from 8 to 12 GHz and coming from 0° to 360° are
performed. Figure 5-7 plots the average DOA estimation errors as functions of the
frequency (a) and incident angle (b).
Average Error (degree)
2.5
2
Without scatterer
With symmetric scatterer
With asymmetric scatterer
1.5
1
0.5
0
8
9
10
11
Frequency (GHz)
12
(a)
Average Error (degree)
8
Without scatterer
With symmetric scatterer
With asymmetric scatterer
6
4
2
0
0
45
90 135 180 225 270 315 360
Incident Angle  (deg)
(b)
Figure 5-7. Simulated averaged DOA estimation errors assuming ±0.25 dB
magnitude difference and ±2° phase difference errors for rectangular shaped
scatterers: (a) versus frequency (averaged over all incident angles from 0° to 360°
with 2° step) and (b) versus incident angle (averaged over all frequencies from 8 to
12 GHz with 0.5 GHz step).
104
As shown in Figure 5-7, with scatterers (both symmetric and asymmetric cases),
the DOA estimation errors are smaller compared to those without the scatter, indicating
accuracy improvement as expected.
In addition, the asymmetric scatterer case has
slightly smaller errors than those of the symmetric case.
As a comparison, a cylindrical shaped scatterer made of the same material with
the radius of 6.4 mm and height of 10mm is also studied for the direction finding
applications. Assuming errors of +/- 2° in phase and +/- 0.25 dB in magnitude, the
average DOA estimation errors with the cylindrical scatterer as functions of the
frequency and incident angle are plotted in Figure 5-8 (a) and Figure 5-8 (b) respectively.
They demonstrate very similar improvements in the direction finding ability comparison
to the rectangular scatterers in terms of the ambiguity elimination and sensitivity
enhancement. Thus, this biological inspired technique is not very sensitive to the exact
form factor of the scatterer and only the scatterer with a rectangular shape will be
evaluated in the experiments.
105
Average Error (degree)
2.5
Without scatterer
With symmetric scatterer
With asymmetric scatterer
2
1.5
1
0.5
0
8
9
10
11
Frequency (GHz)
12
(a)
Average Error (degree)
8
Without scatterer
With symmetric scatterer
With asymmetric scatterer
6
4
2
0
0
45
90 135 180 225 270 315 360
Incident Angle  (deg)
(b)
Figure 5-8. Simulated averaged DOA estimation errors assuming ±0.25 dB
magnitude difference and ±2° phase difference errors for cylindrical shaped
scatterers: (a) versus frequency (averaged over all incident angles from 0° to 360°
with 2° step) and (b) versus incident angle (averaged over all frequencies from 8 to
12 GHz with 0.5 GHz step).
5.4. Experimental Results
An X-band prototype of the same configuration as illustrated in Figure 5-3 is built
and tested to verify the proposed concept experimentally. Figure 5-9 shows a photograph
of the two-monopole antenna with an absorber material (ARC-LS-10211) placed
symmetrically in between the two antennas. As described previously, the monopoles
have a length of 7.0 mm, and are separated by a distance of 15 mm. The absorber
106
properties are: r = 2.05 and tan = 1.05, with dimensions of 15 mm x 12.8 mm x 10 mm.
The asymmetric configuration is again realized by shifting the absorber along the xdirection by 7.5 mm.
x
Figure 5-9. A photograph of the X-band two-monopole and symmetric scatterer
prototype.
The X-band two-antenna systems without and with symmetric and asymmetric
scatterers are tested using a vector network analyzer in an anechoic chamber. The
steering vectors (spanning 8 to 12 GHz with 0.5 GHz step, 0° to 360° with 1° step) were
measured with a transmitting horn antenna, one receiving monopole while the other one
is terminated with a 50- load and vice versa. Figure 5-10 compares the measured and
simulated M and  at 10 GHz, without the lossy scatterer (Figure 5-10(a)), with the
symmetric (Figure 5-10(b)) and with the asymmetric lossy scatterer (Figure 5-10(c)).
The measured phase and magnitude differences (M and  , or steering vectors) agree
very well with the simulation results, verifying the accuracy of our simulation procedure.
107
Moreover, it is confirmed that with the incorporation of the lossy scatterer, the magnitude
difference becomes larger, which provides useful information in estimating the DOA, and
the phase slope (d/d) is greater, which leads to better DOA sensitivity.
108
3
400
350
Phase Difference (degree)
Magnitude Difference (dB)
2
1
0
-1
-2
-3
0
300
250
200
150
100
50
45
90
135 180 225 270
Incident Angle  (deg)
315
0
0
360
45
90
135 180 225 270
Incident Angle  (deg)
315
360
45
90
135 180 225 270
Incident Angle  (deg)
315
360
45
90
135 180 225 270
Incident Angle  (deg)
315
360
8
400
6
350
4
Phase Difference (degree)
Magnitude Difference (dB)
(a)
2
0
-2
-4
-6
250
200
150
100
50
-8
-10
0
300
45
90
135 180 225 270
Incident Angle  (deg)
315
0
0
360
10
400
8
350
6
Phase Difference (degree)
Magnitude Difference (dB)
(b)
4
2
0
-2
-4
-6
250
200
150
100
50
-8
-10
0
300
45
90
135 180 225 270
Incident Angle  (deg)
315
360
0
0
(c)
Figure 5-10. Comparison of measured (solid lines) and simulated (dashed lines)
magnitude (left) and phase (right) differences at 10 GHz for the three cases: (a) No
scatterer; (b) Symmetric scatterer; (c) Un-symmetric scatterer.
109
The DOA performance of the experimental prototypes is then evaluated by
transmitting a signal using a horn antenna in the far field region from 0° to 360° with 15°
step. MUSIC algorithm is used to estimate the DOA based on the previously measured
steering vectors. As predicted by simulations, it is observed that ambiguities at all
frequencies and incident angles are completely eliminated with the incorporation of the
un-symmetric scatter. Figure 5-11 plots the MUSIC output for an incident 12 GHz signal
coming from 90°, for all three antenna / scatter configurations.
Spectral Magnitude (dB)
40
Without scatterer
With symmetric scatterer
With asymmetric scatterer
30
20
10
0
-10
0
45
90
135 180 225 270
Incident Angle  (deg)
315
360
Figure 5-11. MUSIC output of a 12 GHz signal incident from 90o for the twoantenna configurations without scatterer (dotted dashed line), with the symmetric
scatterer (dashed line), and with the asymmetric scatterer (solid line).
As predicted by the simulation results, there are ambiguities in the estimated
DOA for the two-antenna system without a scatterer in between because the antenna
distance is greater than a half wavelength.
The asymmetric scatterer configuration
performs the best, also as expected. The overall DOA estimation accuracy for all three
cases versus frequency and incident angles are shown in Figure 5-12. Compared to the
simulated DOA estimation accuracies in Figure 5-7, quite good agreement is observed. It
110
can be seen that for the entire X-band, the average DOA estimation error decreased from
more than 3° to less than 1° with the incorporation of the scatterer in between the two
antennas. The average errors plotted as a function of incident angle also agree well with
simulation results and show significant improvement with the incorporation of the
scatterer. In summary, the asymmetric configuration has slightly better accuracy than the
symmetric case, in addition to its ability of eliminating all ambiguities including the front
/ back ambiguity.
Average Error (degree)
6
Without scatterer
With symmetric scatterer
With asymmetric scatterer
4
2
0
8
9
10
11
Frequency (GHz)
12
(a)
Average Error (degree)
10
Without scatterer
With symmetric scatterer
With asymmetric scatterer
8
6
4
2
0
0
45
90 135 180 225 270 315 360
Incident Angle  (deg)
(b)
Figure 5-12. Measured averaged estimation errors for all three two-antenna
configurations (a) versus frequency (averaged over all incident angles from 0° to
360° with 15° step) and (b) versus incident angle (averaged over all frequencies from
8 to 12 GHz with 0.5 GHz step).
111
5.5. Conclusion
A human ears inspired two-antenna direction finding configuration is proposed.
By incorporating a head-like scatterer between two monopole antennas, both phase and
magnitude information can be utilized to estimate the direction of arrival (DOA) of a
microwave signal, thus eliminating ambiguities associated with phase wrapping at high
frequency and achieving a more accurate DOA estimation. In addition, this biological
inspired technique is not very sensitive to the exact form factor of the scatterer. Both the
rectangular and cylindrical scatterers demonstrate very similar improvements in the
direction finding ability in terms of ambiguity elimination and sensitivity enhancement.
Moreover, the experimental investigations of a two-antenna direction finding system at
X-band incorporating a human head-like scatterer are performed. The measured phase
and magnitude differences (M and  , or steering vectors) agree very well with their
simulated values. Furthermore, it is confirmed that with the incorporation of the lossy
scatter, the magnitude difference becomes larger, which provides useful information in
estimating the DOA, and the phase slope (d/d) is greater, which leads to better DOA
sensitivity. The results have shown that the incorporation of the head-like scatter not only
eliminates the phase ambiguity issue at higher frequencies, but also improves the general
sensitivity of the two-antenna direction finding system. This kind of biological inspired
RF technique may lead to future direction finding systems that are low-cost, compact and
light weight.
112
CHAPTER 6.
A MICROWAVE DIRECTION OF ARRIVAL
ESTIMATION TECHNIQUE USING A SINGLE UWB ANTENNA [115]
Inspired by the sound localization capability of the human auditory system using
single ear (monaural localization), a novel direction of arrival (DOA) estimation
technique is described in this chapter using a single Ultra-Wide-Band (UWB) antenna.
By exploiting the incident angle dependent frequency response of a wide-band antenna,
the DOA of a broadband microwave signal can be estimated. The DOA estimation
accuracies are evaluated for various antenna configurations and microwave signals with
different signal-to-noise ratios (SNR). The effectiveness of the proposed direction finding
technique is demonstrated in both simulation and experiment, although with reduced
estimation accuracy comparing with the case with two-antennas as described in Chapter 5.
6.1. Introduction
In recent years, there has been increasing interest in microwave direction finding
systems due to its wide applications in the military and commercial areas, such as
electronic warfare [52], wireless communications [53], wireless localization, etc. A wide
variety of techniques have been developed to estimate the direction of arrival (DOA) of
incoming signals, including the estimation of signal parameters via rotational invariance
techniques (ESPRIT) [54] and multiple signal classification (MUSIC) [55], etc. All of
these techniques require a large number of antenna elements to achieve a high degree of
113
accuracy. However, as the number of antenna elements increases, the power consumption,
the size and cost of the system increase as well. Therefore, an accurate DOA estimation
technique with reduced number of antenna elements is highly desirable.
Inspired by the human auditory system, an improved DOA technique using just
two antennas with a scatterer in between them to emulate the function of the human head
as a low pass filter, has been reported in Chapter 5. It turns out that another amazing
capability of the human auditory system is its monaural direction finding capability.
Although not as accurate as the binaural (utilizing both ears) case, monaural direction
finding is possible without head movement. This first seems to be intuitively unthinkable
for a single stationary antenna. However, the monaural localization only works for
broadband signals and the main mechanism has been identified to be the spectral
alteration by the pinnae and head that provides cue for directions (similar to an incident
angle sensitive “comb-line filter”) [62-63]. For example, as a simplified illustration
shown in Figure 1-6, the received broadband signal by a single ear may have a notch
response that is incident angle dependent. Therefore, monaural direction finding works
only for broadband signals. By designing the frequency dependences of the antenna
pattern and guiding structure, this unique feature of human ear may be utilized for
microwave applications such as direction finding for an ultra-wide-band (UWB) signal
[117].
In this chapter, a previously demonstrated UWB (3.1 to 10.6 GHz) antenna [116]
is used to investigate the single antenna direction finding technique, which is analogous
to the monaural sound localization of the human auditory system, for which the pinnae /
114
head and shoulder are natural directional antennas for acoustic waves and the received
spectrum of a broadband signal depends on its DOA [56]. Because the received spectra
are different for different DOA, the DOA of the incident signal is then estimated from the
cross-correlation coefficients of the received spectra of the antenna with a pre-determined
incident-angle-dependent spectra. The estimation accuracy of this single antenna DOA
technique is theoretically studied in all three planes with the assumption of different
signal-to-noise ratios (SNR). Another UWB antenna with unsymmetrical shape is also
designed and studied, demonstrating improved estimation accuracy due to the break in
symmetry. Both UWB antennas are fabricated and their DOA performance is tested in an
anechoic chamber using a horn antenna as the illuminator. The measured results confirm
the feasibility of the proposed single antenna DOA technique.
6.2. Simulation Results of the Single Antenna DOA Estimation
This work is inspired by human monaural sound localization, in which, a single
ear functions as a comb-line filter with its frequency response depending on the incident
angle. Here, a single UWB antenna is used to estimate the DOA of incident broadband
microwave signals.
Similar to human monaural sound localization, with different
incident angles, the antenna frequency response varies, that is, the received spectra of the
UWB antenna are correlated to the DOA. Therefore, the DOA of a broadband signal
received by the antenna can be estimated by evaluating the correlation coefficients of the
measured spectrum with pre-determined incident angle dependent spectra. To implement
this idea in microwave DOA estimation applications, two UWB antennas, one previously
115
demonstrated with elliptical configuration and another modified antenna with D-shaped
configuration for improved performance are investigated.
6.2.1. Elliptical UWB Antenna and DOA Estimation
The structure of the first UWB antenna evaluated is illustrated in Figure 6-1(a)
[116]. It consists of an elliptical patch with two symmetric slots on a Rogers’ RT / Duroid
5880 substrate with a thickness of 31 mil, a relative dielectric constant of 2.2 and a loss
tangent of 0.0009. The antenna parameters are: L0 = 15.0 mm, L1 = 14.0 mm, L2 = 11.6
mm, L3 = 10.4 mm, S1 = 2.0 mm, W0 = 23.0 mm, W1 = 19.6 mm, W2 = 16.24 mm, W3 =
14.56 mm, W4 = 1.0 mm, W5 = 2.4 mm and g1 = 0.3 mm. The total antenna foot print,
including the feed line and the ground plane, is 29.0 mm x 23.0 mm.
z
W1
antenna
y
W2
W3
x
W4
ground
plane
0

L3 L2
L1
g1 L
0
Return loss (dB)
substrate
-10
-20
-30
W5
S1
W0
(a)
-40
3
4
5
6
7
8
Frequency (GHz)
9
10
(b)
Figure 6-1. (a) The schematics (left: side view; right: top view) of the symmetric
UWB antenna incorporating two slots. (b) The simulated antenna return loss.
The performance of this antenna is modeled with the full-wave finite-element
electromagnetic solver HFSS. The simulated return loss of the antenna is less than -10 dB
116
within the frequency range from 3.4 to 10 GHz, as shown in Figure 6-1(b). The received
antenna spectra are obtained using plane-wave incidence excitation sweeping from -180º
to 180º with 10º steps in the x-y, x-z and y-z planes (see Figure 6-1(a)). Figure 6-2(a) plots
the received spectra of the antenna for several incident angles in the x-y plane (with the
incident H-field in the z direction). It can be observed that the received spectra are
incident angle dependent. To take into account noise, Gaussian White Noise (GWN) with
different power levels are added to the antenna spectra. The cross-correlation coefficients
between the spectra with added noises (emulating a measured signal) and the original
spectra are then calculated at all the incident angles. As shown in Figure 6-2(b), for the
case of 15 dB SNR (power ratio between the signal and the added GWN) in the x-y plane,
the spectra with noise at the angle θ (θ = -180º, -120º, -60º, 0º, 60º and 120º) are still
highly correlated with the original spectra at the incident angles around θ. The correlation
is also quite high around -θ, which is due to the symmetry of the antenna in the x-y plane.
The DOA of the signals with noise is then estimated by selecting the angle with the
largest correlation coefficient. Figure 6-2(c) shows the Root Mean Square (RMS) of the
estimation errors for different SNRs, which are calculated by running the simulation 1000
times. The estimation errors are less than 16º for most incident angles for higher SNR (≥
20 dB). With the decrease of the SNR, the RMS of the estimation error increases due to
the influence of the noise. Nevertheless, even for SNR = 10 dB, the estimated DOA for
many incident angles can be accurate to within 40º, which is quite similar to the human
monaural sound localization performance [62].
117
Spectrum (dB)
-40
-50
-60
-70
-80
3
4
-180
-120
-60
0
60
120
5
6
7
8
9
Frequency (GHz)
10
(a)
Correlation Coefficient
1
-180
-120
-60
0
60
120
0.8
0.6
0.4
0.2
0
-180 -120 -60
0
60 120 180
Incident angle (degree)
(b)
RMS Error (degree)
140
SNR = 25 dB
SNR = 20 dB
SNR = 15 dB
SNR = 10 dB
120
100
80
60
40
20
0
-180 -120 -60
0
60 120
Incident Angle (degree)
180
(c)
Figure 6-2. (a) The received spectra of the symmetric UWB antenna with the
incident waves (H-field in the –z direction) at -180º, -120º, -60º, 0º, 60º and 120º in
the x-y plane. (b) The correlation coefficients between the spectra with added noise
(SNR = 15 dB) and the pre-determined spectra at the incident angle θ (-180º, -120º,
-60º, 0º, 60º and 120º) in the x-y plane. (c) The RMS of the DOA estimation error
with different SNR.
118
Similarly, the received spectra of the UWB antenna in the x-z plane with both
polarizations (E-field and H-field in the y direction), and in the y-z plane with the E-field
in the x direction are also directional dependent, although to different degrees, as shown
in Figure 6-3(a), 6-3(b) and 6-3(c), respectively. The corresponding RMS of the
estimation errors for these cases are shown in Figure 6-4. For example, for the incident
wave in the x-z plane with the E-field in the y direction, the RMS estimation errors are
less than 16º for noise with SNR ≥ 20 dB (Figure 6-4(a)). On the other hand, for the other
two cases as shown in Figure 6-3(b) and Figure 6-3(c), the angle dependence of the
spectra is relatively small. Therefore, the RMS estimation errors are quite large (greater
than 50º for SNR < 20dB), as plotted in Figure 6-4 (b) and Figure 6-4(c).
-40
-40
-40
-60
-70
-180
-120
-60
0
60
120
-80
-90
-100
3
4
5
6
7
8
Frequency (GHz)
(a)
-50
-60
-180
-120
-60
0
60
120
-70
-80
-90
3
9
10
Spectrum (dB)
Spectrum (dB)
Spectrum (dB)
-50
4
5
6
7
8
Frequency (GHz)
(b)
-50
-180
-120
-60
0
60
120
-60
-70
9
10
-80
3
4
5
6
7
8
Frequency (GHz)
9
10
(c)
Figure 6-3. The received spectra of the UWB antenna at -180º, -120º, -60º, 0º, 60º
and 120º for the incident waves in (a) the x-z plane (E-field in the y direction), (b) the
x-z plane (H-field in the y direction), and (c) the y-z plane (E-field in the x direction).
119
40
0
-180 -120 -60
0
60 120
Incident Angle (degree)
150
250
SNR = 25 dB
SNR = 20 dB
SNR = 15 dB
SNR = 10 dB
RMS Error (degree)
80
200
SNR = 25 dB
SNR = 20 dB
SNR = 15 dB
SNR = 10 dB
RMS Error (degree)
RMS Error (degree)
120
100
180
(a)
50
0
-180 -120 -60
0
60 120
Incident Angle (degree)
180
200
150
SNR = 25 dB
SNR = 20 dB
SNR = 15 dB
SNR = 10 dB
100
50
0
-180 -120 -60
0
60 120
Incident Angle (degree)
(b)
180
(c)
Figure 6-4. The RMS of the estimation error with different SNRs for the incident
waves in the x-z plane with the (a) E-field in the y direction, (b) H-field in the y
direction , and (c) in the y-z plane with the E-field in the x direction.
6.2.2. D-shaped UWB Antenna and DOA Estimation
Due to its symmetry respect to the x-axis, the spectra of the elliptical UWB
antenna with added noise are also correlated to the original spectra at the symmetry
direction in the x-y plane (i.e.,  and -). To break the symmetry to increase the
estimation accuracy, another UWB antenna with a D-shaped configuration is also
designed, as shown in Figure 6-5(a). The antenna design parameters are selected to be: L0
= 16.0 mm, L1 = 14.0 mm, L2 = 11.6 mm, L3 = 10.4 mm, S1 = 2.0 mm, W0 = 23.0 mm, W1
= 19.6 mm, W2 = 16.72 mm, W3 = 15.28 mm, W4 = 1.0 mm, W5 = 2.4 mm and g1 = 0.3
mm. The simulated reflection coefficient of this antenna is less than -6 dB within 3 ~ 10
GHz, as shown in Figure 6-5(b).
120
W1
W2
W3
Substrate
y
W4
0
L3 L2
-5
Antenna
S11 (dB)
L1
Ground
plane
x
g
L0
W5
W0
(a)
S1
-10
-15
-20
3
4
5
6
7
8
9
10
Frequency (GHz)
(b)
Figure 6-5. (a) The schematics (left: side view; right: top view) of the proposed UWB
monopole antenna with unsymmetrical shape. (b) Reflection coefficient.
Following the same DOA estimation technique described previously, the
correlation coefficients of the spectra with/out added noise (SNR = 15 dB) and the RMS
of the estimation errors with different SNR are plotted in Figure 6-6 (a) and Figure 6-6
(b), respectively, for the case of incident waves in the x-y plane with the H-field in the z
direction. The spectra with added noises are still highly correlated to the original spectra
around the direction of the incident signal, as expected. In addition, the correlation
coefficients at the symmetry direction respect to the x-axis are reduced due to the break
of symmetry of the D-shaped antenna, leading to improved estimation accuracy. The
RMS of the estimation errors are less than 5° for SNR ≥ 20 dB, as shown in Figure 6-6
(b).
121
-180
-120
-60
0
60
120
0.8
0.6
0.4
0.2
0
-180 -120 -60
0
60
120
Incident Angle (degree)
180
(a)
120
RMS Error (degree)
Correlation Coefficent
1
100
80
SNR = 25 dB
SNR = 20 dB
SNR = 15 dB
SNR = 10 dB
60
40
20
0
-180
-120 -60
0
60
120
Incident Angle (degree)
180
(b)
Figure 6-6. DOA performance of the D-shaped UWB antenna in the x-y plane (Hfield in the z direction): (a) The correlation coefficients between spectra with added
noise (SNR = 15 dB) and the pre-determined spectra at the incident angle θ (-180º,
-120º, -60º, 0º, 60º and 120º); (b) The RMS of the estimation errors for different SNR.
The unsymmetrical shape of the antenna also affects the received spectra and
DOA estimation accuracy in the x-z and y-z planes (for the y-z plane, symmetry is also
broken). As shown in Figure 6-7 (a), the RMS of the estimation errors is less than 12° for
SNR ≥ 20 dB when the incident waves are in the x-z plane with the E-field in the y
direction. Moreover, in the x-z plane for the other polarization (H-field in the y direction),
the RMS errors are less than 20° at most incident angles for the SNR ≥ 20 dB and the
RMS errors for the y-z plane (E-field in the x direction) DOA is also improved, as shown
in Figure 6-7(b) and Figure 6-7(c), respectively.
122
80
60
40
20
0
-180 -120 -60
0
60 120 180
Incident Angle (degree)
(a)
200
150
SNR = 25 dB
SNR = 20 dB
SNR = 15 dB
SNR = 10 dB
100
50
0
-180 -120 -60
0
60 120 180
Incident Angle (degree)
250
RMS Error (degree)
SNR = 25 dB
SNR = 20 dB
SNR = 15 dB
SNR = 10 dB
RMS Error (degree)
RMS Error (degree)
100
200
150
SNR = 25 dB
SNR = 20 dB
SNR = 15 dB
SNR = 10 dB
100
50
0
-180 -120 -60
0
60 120 180
Incident Angle (degree)
(b)
(c)
Figure 6-7. The RMS of the estimation errors of the D-shaped antenna for different
SNR when the incident waves are in (a) the x-z plane (E-field in the y direction), (b)
the x-z plane (H-field in the y direction), and (c) the y-z plane (E-field in the x
direction).
6.3. Experimental Verification of the Single Antenna DOA Estimation
To verify the DOA estimation technique using a single UWB antenna
experimentally, we design, fabricate and test the two UWB antennas discussed in Section
6.2. The measured return losses of the elliptical and D-shaped antennas using an Agilent
E8361A vector network analyzer are plotted in Figure 6-8. Good agreement between the
measured and simulated results was achieved over the entire frequency band, except for
some small frequency shifts, which are due to the SMA connector effects [116].
123
0
0
Simulation
Measurement
-5
-10
S11 (dB)
S11 (dB)
Simulation
Measurement
-20
-10
-15
-20
-30
3
4
5
6
7
8
Frequency (GHz)
9
10
(a)
-25
3
4
5
6
7
8
Frequency (GHz)
9
10
(b)
Figure 6-8. The simulated and measured return losses of (a) the elliptical
UWB antenna and (b) the D-shaped UWB antenna.
The received spectra of the two UWB antennas in different planes for different
polarizations are measured in an anechoic chamber using an X-band horn antenna as the
transmitting antenna sweeping from -180º to 180º with 2º step (with an input power of
+10 dBm). The measured radiation patterns from 7 to 11 GHz agree well with
simulations for both the elliptical UWB and the D-shaped UWB, as shown in Figure 6-9
and Figure 6-10, respectively at 9 GHz.
124
-40
120
90
Measurement 1
Measurement 2
Simulation
60
-50
30
-60 150
-60
120
-70
0
-70
30
150
-80180
0
330
240
270
-60
-50
300
330
210
240
(a)
120
90
270
300
(b)
60
-50
Measurement 1
Measurement 2
Simulation
30
-60 150
-40
120
90
60
-50
Measurement 1
Measurement 2
Simulation
30
-60 150
-70
-70
-80180
0
-70
-80180
0
-70
-60 210
-50
-40
60
-70
-60 210
-50
-40
90
-70
-80180
-40
-50
Measurement 1
Measurement 2
Simulation
330
240
270
(c)
300
-60 210
-50
-40
330
240
270
300
(d)
Figure 6-9. The simulated and measured received-patterns of the elliptical UWB
antenna at 9 GHz with the incident wave in (a) the x-y plane (H-field in the z
direction), (b) the x-z plane (E-field in the y direction), (c) the x-z plane (H-field in
the y direction), and (d) the y-z plane (E-field in the x direction).
125
-40
120
90
60
-50
Measurement 1
Measurement 2
Simulation
30
-60 150
-60
120
-80180
0
-70
30
150
-80180
0
-70
-60 210
-50
330
240
270
300
-60
330
210
-50
240
(a)
-40
120
90
60
Measurement 1
Measurement 2
Simulation
30
-60 150
-70
-80180
0
-70
-60 210
-50
330
240
270
(c)
270
300
(b)
-50
-40
60
Measurement 1
Measurement 2
Simulation
-70
-70
-40
-50
90
300
-40
-50
-60 150
-70
-80
-90180
-80
-70
-60 210
-50
-40
120
90
60
Measurement 1
Measurement 2
Simulation
30
0
330
240
270
300
(d)
Figure 6-10. The simulated and measured received-patterns of the D-shaped UWB
antenna at 9 GHz with the incident wave in (a) the x-y plane (with H-field in the z
direction), (b) the x-z plane (with E-field in the y direction), (c) the x-z plane (with Hfield in the y direction), and (d) the y-z plane (with E-field in the x direction).
126
As shown in Figure 6-9 and Figure 6-10, two sets of received spectra are
measured separately for both of the two antennas. The measured spectra of the antennas
are different at different incident angles as expected. Using one set of the measured
spectra as the calibration spectra, the incident angle of a broadband signal from 7 to 11
GHz is then measured and estimated by calculating the correlation coefficients of the two
spectra. The estimation errors using the elliptical and the D-shaped UWB antenna are
plotted in Figure 6-11 and Figure 6-12, respectively, both agreeing reasonably well with
the simulation predications and confirming the feasibility of the single UWB DOA
technique. As shown in Figure 6-11 (a) for the incident wave in the x-y plane with the Hfield in the z direction, the demonstrated estimation errors are less than 20° at most
incident angles. The relatively large estimation errors around 0° is probably due to the
scattering from the cable and SMA near the antenna. Additionally, the estimation errors
are less than 8° within the angles of -26° ~ 146° when the incident waves are in the x-z
plane with the E-field in the y direction (Figure 11 (b)). For the other polarization (Hfield in the y direction) in the x-z plane (Figure 6-11 (c)), the estimation errors are quite
large (greater than 60°) as predicted in the simulation. Furthermore, the DOA estimation
error is less than 30° with the incident waves in the y-z plane (E-field in the x direction),
as shown in Figure 6-11 (d).
127
180
Estimation Error (degree)
Estimation Error (degree)
180
150
150
120
120
90
60
30
0
-180 -120 -60
0
60 120 180
Incident Angle (degree)
90
60
30
0
-180 -120 -60
0
60 120
Incident Angle (degree)
180
150
120
90
60
30
0
-180 -120 -60
0
60 120 180
Incident Angle (degree)
(c)
(b)
Estimation Error (degree)
Estimation Error (degree)
(a)
180
180
150
120
90
60
30
0
-180 -120 -60
0
60 120 180
Incident Angle (degree)
(d)
Figure 6-11. The estimation errors of the elliptical UWB antenna with the incident
wave in (a) the x-y plane (with the incident H-field in the z direction), (b) the x-z
plane (with E-field in the y direction), (c) the x-z plane (with H-field in the y
direction), and (d) the y-z plan (with E-field in the x direction)
Comparing with the elliptical UWB antenna, the improved estimation accuracies
are demonstrated for the D-shaped UWB antenna, as shown in Figure 6-12 (a), Figure 612 (c) and Figure 6-12 (d). The DOA estimation errors are less than 20° at most incident
angles for incident wave in the x-y plane with H-field in the z direction (Figure 12 (a)).
The large errors at incident angles around 0° are also due to the impact from the cable
and SMA near the antenna. Moreover, the estimation errors using the D-shaped UWB
128
antenna are less than 15° at most of the incident angles in the x-z plane (H-field in the y
direction) and the y-z plane (E-field in the x direction), as shown in Figure 6-12 (c) and
Figure 6-12 (d), respectively. The large estimation errors at incident angles around -10º
and -75º in Figure 6-12 (c) are probably a result of the scattering effect of the SMA and
cables. For the other polarization in the x-z plane (E-field in the y direction), as shown in
Figure 6-12 (b), the DOA estimation errors are quite large because the received spectra
magnitudes and the SNRs are much smaller than the other polarizations (around 10 dB
lower, as shown in Figure 6-10 (b)).
129
180
Estimation Error (degree)
Estimation Error (degree)
180
120
60
0
-180 -120 -60
0
60 120
Incident Angle (degree)
180
150
120
90
60
30
0
-180 -120 -60
0
60 120 180
Incident Angle (degree)
(b)
Estimation Error (degree)
180
150
120
90
60
30
0
-180 -120 -60
0
60 120 180
Incident Angle (degree)
(c)
Estimation Error (degree)
(a)
180
150
120
90
60
30
0
-180 -120 -60
0
60 120 180
Incident Angle (degree)
(d)
Figure 6-12. The estimation errors of the D-shaped UWB antenna with the incident
waves in (a) the x-y plane (with H-field in the –z direction), (b) the x-z plane (with Efield in the +y direction), (c) the x-z plane (with H-field in the +y direction), and (d)
the y-z plane (with E-field in the +x direction).
One important reason causing the estimation errors of the DOA technique using a
single UWB antenna is the noise in the received spectra. The estimated SNR in the
measurement of the D-shaped UWB antenna are around 15 dB ~ 35 dB, with large SNR
values at 0º and 180º. As a result, the estimation errors are relatively small at 180º when
the incident waves are in the x-z plane with H-field in the y direction and the y-z plane
with the E-field in the x direction, as shown in Figure 6-12 (c) and Figure 6-12 (d).
130
Finally, a higher degree of estimation accuracy could be achieved using this
single UWB DOA technique for improved SNR values, as explored in Section 6.2. To
demonstrate it experimentally, we increase the input power and SNR by using the
elliptical UWB antenna fed with a signal generator (Agilent E8257C, with output power
+16 dBm) as the transmitting antenna, sweeping from 3 GHz to 11 GHz. In addition, the
received spectra of the D-shaped UWB antenna is measured using a spectrum analyzer
(Agilent E4407B, able to measure the noise level below -100 dBm), sweeping from -180º
to 180º with 10º steps. Following the aforementioned DOA technique, the estimation
errors of the D-shaped antenna are plotted in Figure 6-13 (a) and Figure 6-13 (b) for the
incident waves in the x-z plane (H-field in the y direction) and the incident waves in the
y-z plane (E-field in the x direction), respectively. High degree of estimation accuracy has
been demonstrated as expected, with the estimation errors less than 10º for most of the
incident angles as shown in Figure 6-13 (a) and Figure 6-13 (b).
131
60
Estimation Error (degree)
Estimation Error (degree)
60
50
40
30
20
10
0
-180 -120 -60
0
60 120
Incident Angle (degree)
(a)
180
50
40
30
20
10
0
-180 -120 -60
0
60 120
Incident Angle (degree)
180
(b)
Figure 6-13. The estimation errors of the D-shaped UWB antenna with increased
SNR when the incident waves are in (a) the x-z plane (with H-field in the y direction).
(b) the y-z plane (with E-field in the x direction).
6.4. Conclusion
In this chapter, a novel DOA technique is investigated using a single UWB
antenna. Because the received spectra of the UWB antenna are incident angle dependent,
the DOA of a broadband signal can be estimated from the correlation coefficients of the
measured spectra and pre-determined calibration spectra. Both an elliptical UWB antenna
and an unsymmetrical D-shaped UWB antenna are studied for the DOA estimation
applications. The feasibility of the single UWB DOA technique is demonstrated in both
simulation and experiment. In addition, with increased SNRs, a higher degree of
estimation accuracy can be achieved. This kind of biological inspired RF technique may
lead to future novel direction finding systems that are low-cost, compact and light weight.
132
CHAPTER 7.
CONCLUSION AND FUTURE WORK
This dissertation covers the following three topics: the study of a metamaterial
with near-zero index of refraction and its application in directive antenna design, the
design technique of a wideband circularly polarized patch antenna for 60GHz wireless
applications, and the investigation of a novel direction of arrival (DOA) estimation
technique inspired by the human ear.
First, a metamaterial composed of two-dimensional (2-D) metallic wire arrays is
investigated as an effective medium with an effective index of refraction less than unity
(neff < 1). The effective permittivity εeff of this kind of wire arrays can be calculated using
plasma theory with a reduced electron density, which is close to zero at frequencies
around its plasma frequency. The self-consistency of the plasma model is demonstrated.
In addition, the effective medium parameters (permittivity εeff, permeability μeff and neff)
of a wire array are extracted from finite-element simulated scattering parameters,
verifying the low neff properties of the wire array at frequencies just above the
theoretically predicted plasma frequency. A 2-D EMXT structure (a square lattice made
of dielectric rods) embedded in the wire medium is then investigated. It is found that the
first band gap of the 2-D EMXT structure shifts to a higher frequency as expected when
the hosting free space region (n = 1) is replaced by the neff < 1 wire array medium.
Moreover, the simulated transmission response of the actual composite of the dielectric
rods and metallic wires agrees with that of dielectric rods embedded in a background
133
assigned to have the extracted effective medium parameters, confirming the validity of
treating the wire array as an effective medium. A simple design methodology for the
directive monopole antenna is then introduced by embedding a monopole within a wire
array with neff < 1 at the antennas working frequencies. Parametric studies of this antenna
system show that the antennas resonance frequencies and radiation patterns are not
sensitive to the monopole length L, and the antenna gain increases when the wire array
size increases. Furthermore, a prototype antenna operates within the X-band is designed,
fabricated and characterized. The measured antenna properties are in good agreement
with simulation results, demonstrating the expected narrow beam radiation and the design
methodology. The narrow beam effect achieved by the monopole / wire array antenna
system is examined in the context of Snell’s law and the effective aperture size in order to
gain more physical insight.
Second, a fully packaged wideband circularly polarized patch antenna is designed
for 60GHz wireless communication. This patch antenna incorporates a diagonal slot at
the center and features a superstrate and an air cavity backing to achieve the desired
performances, including wide bandwidth, high efficiency and low axial ratio. The metal
frame underneath the antenna layer serves as the cavity backing, which is useful for
antenna bandwidth enhancement, as well as mechanical support for the antenna. The
microstrip-fed patch antenna is packaged with a flip-chip CPW interface that is fully
compatible with semiconductor integrated circuits (ICs). In addition, a prototype antenna
is fabricated and characterized using a probe-based measurement setup. The measured
antenna properties including return loss, axial ratio, gain and radiation patterns agree
134
reasonably well with the simulation results. The demonstrated antenna is compatible with
integrated circuits, wide band (able to cover the whole 60 GHz ISM band), highly
efficient and circularly polarized, which would be an ideal front-end component for 60
GHz wireless communication applications.
Third, an improved two-antenna microwave passive direction finding system is
proposed, which is inspired by the human auditory system. The idea is to utilize a lossy
scatterer between two antennas, which emulates the low-pass filtering function of the
human head at high frequency, to achieve a more accurate DOA estimation. By
incorporating a head-like scatterer between two monopole antennas, both phase and
magnitude information can be utilized to estimate the direction of arrival (DOA) of a
microwave signal, thus eliminating the ambiguities associated with phase wrapping at
high frequency. To evaluate the feasibility of the human ear inspired DOA estimation
technique, a simple configuration with two-monopole and a lossy scatter in between them
is modeled using the full-wave finite-element EM solver HFSS. The simulations show
that, with the presence of the lossy scatterer, the magnitude difference is much larger and
the phase difference versus incident angle curve is significantly steeper, which should
lead to higher sensitivity in the DOA estimation. The MUSIC algorithm is then applied to
evaluate the DOA performance of the two-monopole / scatterer system. It is
demonstrated that the DOA estimation errors are indeed smaller for the cases with
scatterers (both symmetric and asymmetric cases) comparing to those without the
scatterer, indicating accuracy improvement as expected. In addition, the asymmetric
scatterer case has slightly smaller errors than those of the symmetric case. Moreover, an
135
X-band prototype of the proposed configuration is built and tested to experimentally
verify the novel direction finding system. As predicted by the simulation results, the
incorporation of the head-like scatterer not only eliminates the phase ambiguity issue at
higher frequencies, but also improves the general sensitivity of the two-antenna direction
finding system. This kind of biological inspired RF technique may lead to future direction
finding systems that are low-cost, compact and light weight.
Another exciting and intriguing feature of human auditory system is the capability
of monaural sound localization, where the pinnae / head and shoulder are natural
directional antennas for acoustic waves, although with reduced estimation accuracy
comparing with the binaural case. The main mechanism for monaural direction finding is
that the pinnae and head function as a comb-line filter with its frequency response
depending on the incident angles. Similarly, the DOA of the microwave signal can also
be estimated using a single UWB antenna with direction-dependent spectra. Because the
received spectra of the UWB signal are different at different DOA, the DOA of the
incident signal is then estimated from the cross-correlation coefficients of the received
spectra of the antenna with the pre-determined incident-angle-dependent spectra. The
estimation accuracy of the DOA technique is investigated with different SNR values.
Another modified UWB antenna with an unsymmetrical shape is also studied for the
direction finding applications, demonstrating improved estimation accuracy due to the
break in symmetry. In addition, both the symmetric elliptical UWB and the modified Dshaped UWB antennas are fabricated and tested. The experimental results confirm the
feasibility of the single UWB DOA technique, demonstrating an estimation accuracy up
136
to 16° for most of the incident angles in the x-z plane and y-z plane. With increased SNR,
the demonstrated estimation error is less than 10° for most of the incident angles.
As potential extension of the work described in this dissertation, several areas
should be very interesting for further study.
First, further improvement of the two-antenna direction finding system may be
realized based on the principles of the human auditory system. Directional antennas can
be applied for direction finding applications instead of the omni-directional monopoles.
In fact, since the wavelength of the KHz acoustic signal is comparable to the GHz RF
signal, one possibility is to create a directional antenna by direct scaling of the human ear
structure. Obviously, the antenna building materials need to be carefully considered,
because the ear skin, muscle and bone are not simply interchangeable with simple
metallic structures commonly used in building antennas. One possible way to resolve this
problem is to investigate the acoustic response of various parts of the ear and find the
corresponding
electromagnetic
materials
(metallic/dielectric
material,
or
even
metamaterials) to realize an antenna that is truly analogous to the ear. In addition, the
direction finding performances of traditional directive antennas will be investigated and
compared with the original monopole design.
Second, all our human ear inspired direction finding ideas are demonstrated in the
anechoic chamber using advanced testing equipment such as a vector network analyzer.
This network analyzer measurement approach is tedious with measurement uncertainties
caused by switching the cables and terminators during the measurement. Therefore, to
enable an easy, fast and accurate measurement in realistic environments, a portable
137
direction finding test bed is highly desirable. The test bed should be integrated on a
microwave printed circuit board, including antennas, head-like scatter and digital receiver.
Figure 7-1 shows the schematic of the standard I/Q receiver circuitry. Commercial
surface mount components are acquired and integrated with microstrip based passive
circuits. The schematic of the receivers designed for the portable direction finding system
at 5.8 GHz is shown in Figure 7-2. The spacing between the two receivers is halfwavelength at 5.8 GHz. Each antenna element is connected to a low noise amplifier
(LNA, HMC604LP3 from Hittite microwave corporation) and image rejection mixer
(SIM-73+ from Mini-Circuits), which converts the RF signal into a 10 MHz IF signal.
Each mixer requires 4 dBm of LO power, thus the VCO is selected to be HMC358MS8G
(from Hittite microwave corporation), providing +11 dBm of power to drive the four
mixers. To generate 90º phase shift in the I and Q channels, a 90º delay line is added to
the LO signal path of the quadrature channel. Both I & Q channels of each element are
connected to the IF circuit board on the back through vias.
Antenna 1
0
90
LO
ADC
Antenna 2
LNA
ADC
0
90
LO
ADC
Laptop - Digital Signal Processing
ADC
LNA
Figure 7-1. The schematic of standard I/Q receiver
138
Via
Mixer
LNA
Filter
IF Amp
Via
Filter
IF Amp
Via
Filter
IF Amp
Via
Filter
IF Amp
Via
Mixer
Ant 1
Via
VCO
Via
Ant 2
Mixer
LNA
Via
Mixer
(a)
(b)
Figure 7-1. The schematic of the receiver in ADS (a) front side. (b) back side.
The entire receiver performances are simulated using Agilent Advanced Design
Systems (ADS) circuit envelope simulator, demonstrating a dynamic range of – 70 dBm
to -24 dBm. As an explicit example, Figure 7-3 plots the received output signals (output
= I_channel +i*Q_channel) with the input signals to be 0.01*exp(iωt) and
0.02*exp(iωt+i*π/4) at receiver 1 and receiver 2, respectively. The magnitudes of the
output signals are constant when the sampling time is large enough as expected, as shown
in Figure 7-3 (a) and Figure 7-3 (b). In addition, the magnitude difference
(|output2|/|output1|) and phase difference between the two output signals are consistent
with their corresponding values of the input signals, which are 2 and 45º respectively.
139
m13
time=6.893usec
mag(Output3)=0.269
m14
time=6.679usec
mag(Output1)=0.534
0.8
m12
m13
0.3
mag(Output1)
mag(Output3)
0.4
m12
time=6.469usec
mag(Output3)=0.270
0.2
(a)
0.1
m15
time=6.857usec
mag(Output1)=0.536
m15
m14
0.6
0.4
(b)
0.2
0.0
0.0
0
1
2
3
4
5
6
0
7
time, usec
5
100
4
0
3
-100
m16
m17
PD1
MD1
m17
time=6.850usec
MD1=1.991
m16
time=6.783usec
MD1=1.977
2
(c)
1
2
m18
time=1.547usec
PD1=45.267
m18
3
4
5
6
7
time, usec
m19
time=1.560usec
PD1=-315.008
-200
(d)
m19
-300
0
1
-400
0
1
2
3
4
time, usec
5
6
7
1.5
1.7
time, usec
Figure 7-3.The received IF signals with the input signals to be 0.01*exp(iωt) and
0.02*exp(iωt+i*π/4). (a) the magnitude of the output signal at receiver 1, (b) the
magnitude of the output signal at receiver 2, (c) magnitude difference between the
two output signals, (d) phase difference between the two output signals.
140
The proposed receiver for the direction finding system is then fabricated and
measured, as shown in Figure 7-4. To make sure the IF frequencies are the same for all
the four branches, only one VCO is applied together with a one-to-four power divider.
The ground planes of the front and back sides of the receiver are soldered together. The
ferrite beads are added to the bias circuits to suppress the high frequency noise, making
the circuit to be much more stable and less sensitive to the outside impact, such as
touching of the bias wires, etc. The branch 1 and branch 2 are the in-phase (I) channel
and quadrature (Q) channel of receiver 1, respectively. The branch 3 and branch 4 are the
I channel and Q channel of receiver 2, respectively.
Branch 1
Input 1
Branch 2
Input 2
Branch 3
Branch 4
(a)
(b)
Figure 7-2. A photo of the fabricated receiver (a) front side. (b) back side.
141
Branch 1
Branch 2
Branch 3
Branch 4
Figure 7-3. Measured waveforms of the four branches of the receiver.
The performances of the receiver are tested by connecting the signal generator
with an equal phase power divider as the two input signals. For the equal phase input
signals with equal power of -35 dBm, the measured waveforms of the four branches of
the receiver are plotted in Figure 7-5. The peak to peak voltages are different at different
branches, which are 0.26 V, 0.228 V, 0.189 V and 0.121 V for branches 1, 2, 3 and 4,
respectively. In addition, the phase difference between the I channel of receiver 1 (branch
1) and I channel of receiver 2 (branch 3), the I channel and Q channel (branch 1 and
branch 2) of receiver 1, and the I channel and Q channel (branch 3 and branch 4) of
receiver 2 are 23º, 87º and 97º,respectively. The mismatches in the magnitudes and
phases of the four branches are due to the fact that the low noise amplifiers, the mixers,
the IF amplifiers and the filters perform differently (with different gains and phases) at
different branches. A number of techniques have been proposed for the quadrature
mismatch compensation [118-120]. In our work, the waveforms are acquired using the
Data Acquisition (DAQ) function of Labview. The algorithm for calibrating the
142
mismatches in magnitudes and phases will be developed with the acquired waveforms, as
well as the algorithm for estimating the incident angles from the measured waveforms.
Finally, human auditory system has many other interesting and intriguing abilities
related to direction finding. Among them, one of the most interesting and potentially
useful features is the masking level differences (MLD), which is closely related to the
well-known “cocktail party” phenomenon. The essential feature of the cocktail party
effect is that our hearing system can selectively suppress the background noise by as
much as 15 dB. It has been demonstrated in [56] that the MLD effect is the maximum
when the inter-aural signals are 180º out of phase and noises are in phase (N0Sπ). In
addition, the MLD effect is very small when the noises are uncorrelated, and increases
with the increase of noise correlations.
The ADS schematic in Figure 7-6 is applied to emulate the MLD effect for
microwave direction finding application. The RF signal (5.8 GHz) is divided into two
Noise1
TLIN
TL1
Z=50.0 Ohm
E=180
F=5.8 GHz
Signal
P_1Tone
PORT1
Num=1
Z=Rsource
P=polar(dbmtow(Power_RF),0)
Freq=RF_Freq
Noise=no
Temp=Temp_Celsius
Var
Eqn
VAR
VAR1
Power_LO=10 _dBm
Power_RF=-50 _dBm
LO_Freq=5.81 GHz
Filt_Freq=10 MHz
RF_Freq=5.8 GHz
Channel 1
V_Noise
noise1
V_Noise=1.52 uV
OSCwPhNoise
OSC2
I_Probe
Freq=LO_Freq
I_sum
P=dbmtow(Power_LO)
Rout=50 Ohm
180_system_sim_zhou2
DCC1
PwrSplit2
PWR1
S21=0.707
S31=0.707
V_sum
Noise2
Channel 2
Balun4Port
CMP1
Term
Term3
Num=3
Z=50 Ohm
I_Probe
I_sub
V_sub
Noise
Source
Correlation
NoiseCorr
SRC3
CorrCoeff=1
Source1="noise1"
Source2="noise2"
V_Noise
noise2
V_Noise=1.52 uV
180_system_sim_zhou2_2
DCC2
OSCwPhNoise
OSC3
Freq=LO_Freq
P=dbmtow(Power_LO)
Rout=50 Ohm
Figure 7-6. The ADS schematic for emulating the MLD effect.
Term
Term2
Num=2
Z=50 Ohm
143
signals, followed by a 180 phase shifter to emulate the Sπ and correlated noises to imitate
the N0 . The inside circuits in channel 1 and 2 are the typical receiver circuit, including
the LNA, the mixer, the filter and the IF amplifier, as shown in Figure 7-7. The
simulation shows that, with the N0Sπ configuration and input SNR of 11.6 dB, the final
output SNR at the voltage summation port is 38.9 dB higher than input SNR when the
noise bandwidth is 30 KHz. However, the output SNR decreases to 5.3 dB when the
noise bandwidth increases to 1 GHz (with the same input SNR of 11.6 dB). In addition,
the time domain simulation of the MLD effect shows that the output waveforms are the
same for the cases with and without adding correlated white noises. The possible reasons
for these issues are still unknown, which requires further investigations and studies.
I_Probe
I_RF_In1
I_Probe
I_Mix _In1
V_RF_In1
Port
RF_input
Num=1
Mixer
b2_MIX3
SideBand=LOWER
RF_Rej=30 dB
ConvGain=dbpolar(-6.37,0)
NF=5 dB
TOI=14.5
V_Mix_In1
Amplifier
LO_In1
b1_RF_AMP
S21=dbpolar(11.878,-154.353) Port
S11=polar(0.0522757,95.424) LO_Inphase
S22=polar(0.31578,85.0066)
Num=2
S12=0.0557738
NF=2.5 dB
TOI=10
I_Probe
I_Mix _Out1
V_Mix _Out1
I_Probe
I_Filt_Out1
V_Filt_Out1
BPF_Butterworth
BPF8
Fcenter=Filt_Freq
BWpass=2 MHz
Apass=1 dB
BWstop=5 MHz
Astop=55 dB
I_Probe
I_IF_Out1
V_IF_Out1
Amplifier
b4_IF_AMP2
S21=dbpolar(23.2,-178)
S11=polar(0.083176,-226)
S22=polar(0.309,-170)
S12=0.0327
NF=2.5 dB
TOI=36.8
Port
IF_Inphase
Num=3
Figure 7-7. The ADS schematic inside channel 1 and channel 2..
To date, the neural mechanism or the physiology of the MLD/cocktail part effect
is still not fully understood. For the future work, we will investigate the mechanisms of
the cocktail party effect with the goal of developing RF techniques to enhance wireless
signal detection and direction finding capabilities for a broad range of military and
commercial systems. In addition, various direction finding algorithms based on the
144
cocktail party effect and other amazing phenomena of the human auditory system will be
investigated.
145
APPENDIX: MUSIC ALGORITHM
MUltiple SIgnal Classification (MUSIC) is a highly accurate subspace-based
technique for determining the parameters of multiple sources from the signals received at
an antenna array [55]. This technique is very general and has wide applications. It can be
implemented to provide highly accurate estimations of the numbers of signals, direction
of arrival (DOA), strength and cross correlations among the sources, etc.
The signals received at M antennas in an array are linear combinations of the N
incident waves (M > N) plus noise. Therefore, the following model can be applied to
characterize the M received signals, represented by the vector Y
 X 1  W1 
Y1 
   
Y  
 2   a ( ) a ( )  X 2   W2 
N 
   
   1
    
  
YM 
 X N  WM 
or
Y = AX +W
(1)
where Y is the complex vector of the received signal; A is an M x N matrix of steering
vectors, which are known functions of the signal arrival angles and the array element
locations; X is the complex vector of the arriving signals; and W is the noise vector.
Under the assumption that the noise is white noise (with a mean of 0 and a
variance of  2 ) and is uncorrelated with the incident signals, the covariance matrix of the
vector Y is then given by
146
R  E[YY H ]  A( ) Rs AH ( )   2 I
(2)
where E[] and H are the expectation and conjugate transpose operators, respectively, and
Rs = E[XXH].
When the number of incident signals N is less than the number of array elements
M, the matrix ARsAH is singular with the minimum eigenvalue equal to 0 because it is
nonnegative definite and has a rank less than M. Let λ1 ≥ λ2 ≥ … ≥ λM denote the
eigenvalues of the covariance matrix R. Then it follows that
λi >  2 for i = 1, …, N; and
λi =  2 for i = N+1, …, M.
Denote the unit-norm eigenvectors associated with λ1, …, λN by s1, …, sN, and those
corresponding to λN+1, …, λM by g1, …, gM-N. Also define
S = [s1, …, sN], which is a M x N matrix;
and G = [G1, …, GM-N], which is a M x (M- N) matrix.
The span of the N vectors S defines the so-called signal subspace, and the orthogonal
complement spanned by G defines the noise subspace. This terminology is a consequence
of the fact that the span{S} = span{A}  span{G}. Therefore, any array manifold vector
a(θ) in the subspace span{A} is orthogonal to the noise subspace,
GHa(θ)= 0,
or equivalently, aH(θ)GGHa(θ)= 0.
(3)
(4)
Since a(θi) is a column vector of A, we have
aH(θi)GGHa(θi)= 0, i =1, 2, …, N.
(5)
147
As a result, the DOA of the incident signals can be estimated by varying the array
manifold vector a(θ) over       and select the N minima as the estimate. In
practice, we define the MUSIC angular spectrum as
PMU ( ) 
1
.
a ( )GG H a( )
H
The angles of arrival are obtained as the peaks of the angular spectrum.
(6)
148
REFERENCES
1. J. C. Bose, Proc. Roy. Soc. 63, 146 (1898).
2. J. B. Pendry, A. J. Holden, W. J. Stewart and I. Youngs, Phys. Rev. Lett. 76, 4773
(1996).
3. J. B. Pendry, A. J. Holden, D. J. Robbins, and W. J. Stewart, IEEE Trans.
Microwave Theory and Tech. 47, 2075 (1999).
4. K. F. Lindman, Ofversigt of Finska Vetenskaps Soceitetens Forhandlinger LVII,
1 (1914).
5. W. E. Kock, Bell Sys. Tech. J. 27, 58 (1948).
6. V. G. Veselago, Sov. Phys. Uspekhi, 10, 509 (1968)
7. D. R. Smith, W. J. Padilla, D. C. Vier, S. C. Nemat-Nasser, and S. Schultz, Phys.
Rev. Lett. 84, 4184 (2000).
8. J. B. Pendry, Phys. Rev. Lett. 85, 3966 (2000).
9. R. A. Shelby,a) D. R. Smith, S. C. Nemat-Nasser, and S. Schultz, Appl. Phys. Lett.
78, 489 (2001)
10. R. A. Shelby, D. R. Smith, and S. Schultz, Science 292, 77 (2001).
11. C. Caloz, C.-C. Chang, and T. Itoh, J. Appl. Phys. 90, 5483 (2001).
12. R. W. Ziolkowski, and E. Heyman, Phys. Rev. E 64, 056625 (2001).
13. P. Markos, and C. M. Soukoulis, Phys. Rev. B 65, 033401 (2001).
14. N. Engheta, IEEE Antennas and Wireless Propag. Lett. 1, 10 (2002).
15. R. W. Ziolkowski, IEEE Trans. Antennas Propag. 51, 1516 (2003).
16. M. Kafesaki, Th. Koschny, R. S. Penciu, T. F. Gundogdu, E. N. Economou, and C.
M. Soukoulis, J. Opt. A: Pure Appl. Opt. 7, S12 (2005).
17. H. O. Moser, B. D. F. Casse, O. Wilhelmi, and B. T. Saw, Phys. Rev. Lett. 94,
063901 (2005).
149
18. M. Gokkavas, K. Guven, I. Bulu, K. Aydin, R. S. Penciu, M. Kafesaki, C. M.
Soukoulis, and E. Ozbay, Phys. Rev. B 73, 193193 (2006).
19. C. Monzon, D. W. Forester, D. Smith, and P. Loschialpo, J. Opt. Soc. Am. A 23,
339 (2006).
20. N. Wongkasem, A. Akyurtlu, K. A. Marx, Q. Dong, J. Li, and W. D. Goodhue,
IEEE Trans. Antennas Propag. 55, 3052 (2007).
21. R. W. Ziolkowski, Phys. Rev. E 70, 046608 (2004).
22. A. Alù and N. Engheta1, A. Erentok, and R. W. Ziolkowski, IEEE Antennas
Propag. Magazine 49, 23 (2007).
23. N. Engheta and R. W. Ziolkowski, Metamaterials Physics and Engineering
Explorations (John Wiley & Sons, Inc., New York, 2006).
24. A. Alù, F. Bilotti, N. Engheta, and L. Vegni, IEEE MTT-S Int. Microwave Symp.
Dig., (Long Beach, CA, 2005), pp. 1733-1736.
25. A. Alù and N. Engheta, Phys. Rev. E, 72 016623 (2005).
26. M. Silveirinha and N. Engheta, Phys. Rev. Lett. 97, 157403 (2006).
27. A. Alù, M. G. Silveirinha, A. Salandrino, and N. Engheta, Phys. Rev. B 75,
155410 (2007).
28. K. C. Gupta, Electronics Lett. 7, 16 (1971).
29. I. J. Bahl and K. C. Gupta, IEEE Trans. Antennas Propag. 22, 119(1974).
30. S. Enoch, G. Tayeb, P. Sabouroux, N. Guérin, and P. Vincent, Phys. Rev. Lett. 89,
213902 (2002).
31. S. Enoch, G. Tayeb, and B. Gralak, IEEE Trans. Antennas Propag. 51, 2659
(2003).
32. G. Lovat, P. Burghignoli, F. Capolino, D. R. Jackson and D. R. Wilton, IEEE
Trans. Antennas Propag. 54, 1017 (2006).
33. H. Xin, and R. Zhou, IEEE AP-S Intl Symp. Dig., (Honolulu, HI, 2007), pp.
2530-2533.
150
34. R. Zhou, H. Zhang, and H. Xin, Microw. Opt. Technol. Lett. 50, 2341 (2008).
35. R. Zhou, H. Zhang, and H. Xin, IEEE Trans. Antennas Propag. 58, 79 (2010).
36. R. W. Ziolkowski, and C.-Y. Cheng, Radio Sci. 39, RS2017 (2004).
37. R.C. Daniels and Jr. R.W. Heath, Vehicular Tech. Magazine 2, 41 (2007).
38. J. F. Buckwalter, A. Bahakhani, A. Komijani, and A. Hajimiri, IEEE Trans.
Microwave Theory Tech. 54, 4271 (2006).
39. J. P. Pavon, N. S. Shankar, V. Gaddam, K. Challapali, and C.-T. Chou, IEEE
Comm. Magazine 44, 128 (2006).
40. N. Jain, IEEE MTT-S Int. Microw. Symp. Dig., (Boston, MA, 2000), pp. 565-568.
41. K. Kitazawa, S. Koriyama, H. Minamiue, and M. Fujii, IEEE Trans. Microw.
Theory Tech. 48, 1488 (2000).
42. D. Parker, IEEE Trans. Microw. Theory Tech. 50, 1039 (2002).
43. J. Schepps and A. Rosen, IEEE Trans. Microw. Theory Tech. 50, 1044 (2002).
44. A. Bessemoulin, M. Parisot, and M. Camiade, IEEE MTT-S Int. Microw. Symp.
Dig., (Fort Worth, TX, 2004), pp. 473-476.
45. U. R. Pfeiffer, J. Grzyb, D. Liu, B. Gaucher, T. Beukema, B. A. Floyd, and S. K.
Reynolds, IEEE Trans. Microw. Theory Tech. 54, 3387 (2006).
46. T. S. Rappaport and D. A. Hawbaker, IEEE Trans. Commun. 40, 240 (1992).
47. T. Manabe, K. Sato, H. Masuzawa, K. Taira, T. Ihara, Y. Kasashima, and K.
Yamaki, IEEE Trans. Vehic. Tech. 44, 268 (1995).
48. C. Loyez, N. Rolland, P. A. Rolland, and O. Lafond, Electronics Lett. 37, 654
(2001).
49. N. Herscovici, IEEE Trans. Antennas Propagat. 46, 471 (1998).
50. P. C. Sharma, and K. C. Gupta, IEEE Trans. Antennas Propagat. 31, 949 (1983).
51. T. Zwick, C. Baks, U. Pfeiffer, D. Liu, and B. Gaucher, IEEE Int. Antenna and
Propag. Symp. Dig. (Monterey, CA, 2004), pp. 747 – 750.
151
52. S. E. Lipsky, Microwave Passive Direction Finding, (John Wiley & Sons, Inc.,
New York, 1987).
53. L.C. Godara, Proc. IEEE 85, 1195 (1997).
54. R. Roy and T. Kailath, IEEE Trans. Acoust., Speech, Signal Processing 37, 984
(1989).
55. R. O. Schmid, IEEE Trans. Antennas Propagat. 34, 276 (1986).
56. B. C. J. Moore, Introduction to the Psychology of Hearing, (University Park Press,
Baltimore, Maryland, 1977), pp. 169-208.
57. E. Villchur, Acoustics for Audiologists, (Singular Publishing Group, 2000).
58. W. A. Yost, Fundamentals of Hearing: An Introduction, 4th ed., (Academic Press,
2000).
59. S. S. Stevens, and E. B. Newman, Am. J. Psychol. 48, 297 (1936).
60. T. T. Sandel, D. C. Teas, W. E. Feddersen, and L. A. Jeffress, J. Acous. Soc. Am.
27, 842 (1955).
61. L. A. Jeffress, Audiology 10, 77 (1971).
62. R. A. Butler, Percept. Psychophys. 9, 99 (1971).
63. J. Blauert, Acoustica 22, 206 (1969).
64. S. J. Tans, M. H. Devoret, H. Dai, A. Thess, R.E. Smalley, L. J. Geerligs, and C.
Dekker, Nature 386, 474 (1997).
65. D. M. Polzar, Microwave Engineering, (John Wiley & Sons, 1998).
66. A. M. Nicolson, and G. F. Ross, IEEE Trans. Instrum. Meas. IM-19, 377 (1970).
67. W. B. Weir, Proc. IEEE 62, 33 (1974).
68. X. Chen, T. M. Grzegorczyk, B. Wu, J. Pacheco, Jr., and J. A. Kong, Phys. Rev. E
70, 016608 (2004).
69. P. K. Kadaba, IEEE Trans. Instrum. Meas. IM-33, 336 (1984).
152
70. D. K. Ghodgaonkar, V. V. Varadan, and V. K. Varadan, IEEE Trans. Instrum.
Meas. 39, 387 (1990).
71. J. Baker-Jarvis, E. J. Vanzura, and W. A. Kissick, IEEE Trans. Microwave
Theory and Tech. 38, 1096 (1990).
72. D. R. Smith, S. Schultz, P. Markos, and C. M. Soukoulis, Phys. Rev. B 65,
195104 (2002).
73. L. Wang, R. G. Zhou, and H. Xin, IEEE Trans. Microwave Theory Tech. 56, 499
(2008)
74. H. Chen, L. Ran, J. Huangfu, X. Zhang, K. Chen, T. M. Grzegorczyk and J. A.
Kong, J. Appl. Phys. 96, 5338 (2004)
75. H. Chen, L. Ran, J. Huangfu, X. Zhang, K. Chen, T. M. Grzegorczyk and J. A.
Kong, Phys. Rev. E 70, 057605 (2004).
76. D. Schurig, J. J. Mock, B. J. Justice, S. A. Cummer, J. B. Pendry, A. F. Starr, and
D. R. Smith, Science 314, 977 (2006).
77. Th. Koschny, P. Markos, D. R. Smith, and C. M. Soukoulis, Phys. Rev. E 68,
065061 (2003).
78. Th. Koschny, P. Markoš, E. N. Economou, D. R. Smith, D. C. Vier, and C. M.
Soukoulis, Phys. Rev. B 71, 245105 (2005).
79. S. G. Johnson, and J. D. Joannopoulos, Opt. Express 8, 173 (2001).
80. G. Poilasne, J. Lenormand, P. Pouliguen, K. Mahdjoubi, C. Terret and Ph. Gelin,
Microw. Opt. Technol. Lett. 15, 384 (1997).
81. G. Poilasne, P. Pouliguen, K. Mahdjoubi, C. Terret, Ph. Gelin, and L. Desclos, Microw.
Opt. Technol. Lett. 18, 407 (1998).
82. S. Enoch, B. Gralak, and G. Tayeb, App. Phys. Lett. 81, 1588 (2002).
83. M. Thévenot, C. Cheype, A. Reineix, and B. Jecko, IEEE Trans. Microwave
Theory and Tech. 47, 2115 (1999).
84. C. Cheype, C. Serier, M. Thèvenot, T. Monédière, A. Reineix, and B. Jecko,
IEEE Trans. Antennas Propag. 50, 1285 (2002).
153
85. P. M. T. Ikonen, E. Saenz, R. Gonzalo, and S. A. Tretyakov, IEEE Trans.
Antennas Propag. 55, 2692 (2007).
86. T. Akalin, J. Danglot, O. Vanbésien, and D. Lippens, IEEE Microw. Wireless
Comp. Lett. 12, 48 (2002).
87. F. Ghanem, G. Y. Delisle, T. A. Denidni, and K. Ghanem, IEEE Antennas and
Wireless Propag. Lett. 5, 384 (2006).
88. B. Temelkuran, E. Ozbay, J. P. Kavanaugh, G. Tuttle, and K. M. Ho, App. Phys.
Lett., 72, 2376 (1998).
89. P. A. Belov, R. Marques, S. I. Maslovski, I. S. Nefedov, M. Silveirinha, C. R.
Simovski, and S. A. Tretyakov, Phys. Rev. B 67, 113103 (2003).
90. R. Zhou, D. Liu, and H. Xin, Submitted to IEEE Trans. Antennas Propag., (2010).
91. M. Lye, R. B. Waterhouse, D. Novak, F. Zavosh, and J. T. Aberle, IEEE
Microwave Guided Wave Lett. 8, 432 (1998).
92. N. C. Karmakar, IEEE Trans. Antennas Propag. 50, 1706 (2002).
93. N. Herscovici, IEEE Trans. Antennas Propagat. 46, 471 (1998).
94. M. M. Faiz and P. F. Wahid, Proc. IEEE Int. URSI Conf. (Orlando, FL, 1999), pp.
272–275
95. W. Choi, C. Pyo, Y. H. Cho, J. Choi, and J. Chae, IEEE Antennas Propagat.
Symp. Dig. (Columbus, OH, 2003), pp. 292–295.
96. L. Bernard, R. Loison, R. Gillard, and T. Lucidarme, IEEE Antennas Propagat.
Symp. Dig. (San Antonio, TX, 2002), pp. 522–525.
97. N. G. Alexopoulos and D. R. Jackson, IEEE Trans. Antennas Propagat. 32, 807
(1984).
98. J. Lee, N. Kidera, G. DeJean, S. Pinel, J. Laskar, and M. M. Tentzeris, IEEE
Trans. Microwave Theory Tech. 54, 2925 (2006).
99. X. Tang, S. Xiao, B. Wang and J. Wang, Int. J. Infrared Milli. Waves 8, 275
(2007).
100. M. Sun, Y. P. Zhang, K. M. Chua, L. L. Wai, D. Liu, and B. P. Gaucher, IEEE
Trans. Antennas Propagat. 56, 2780 (2008).
154
101. R. Sauleau and P. Coquet, Microw. Opt. Technol. Lett. 41, 369 (2004).
102. S. J. Franson and R. W. Ziolkowski, IEEE Trans. Antennas Propagat. 57, 2913
(2009).
103. A. E. I. Lamminen, A. R. Vimpari, and J. Säily, IEEE Trans. Antennas Propagat.
57, 2904 (2009).
104. G. M. Rebeiz, Proc. IEEE 80, 1748 (1992).
105. D. Nesic, A. Nesic, and V. Brankovic, IEEE Int. Antenna and Propag. Symp. Dig.
(Columbus, OH, 2003), pp. 912-915.
106. II K. Kim, S. Pinel, S. Laskar, and J. Yook, in Proc. APMC (Suzhou, China,
2005).
107. H. Uchimura, N. Shino, and K.Miyazato, IEEE MTT-S Int. Microw. Symp. (Long
Beach, CA, 2005), pp. 1875–1878.
108. M. Barakat, C. Delaveaud, and F. Ndagijimana, 2nd EuCAP 2007 (Edinburgh, U.
K., 2007), pp. 1-6.
109. R. Zhou, D. Liu, and H. Xin, 3rd EuCAP 2009 (Berlin, Germany, 2009), pp.
3787-3789.
110. J. Grzyb, D. Liu, U. Pfeiffer, and B. Gaucher, IEEE Int. Antenna and Propag.
Symp. Dig. (Albuquerque, NM, 2006), pp. 3939-3942.
111. J. L. Kerr, in Proc. of Antenna App. Symp. (Urbana, Ill, 1978).
112. C. A. Balanis, Antenna Theory, Analysis and Design, (1st ed., John Wiley & Sons,
1982).
113. J. P. Raskin, G. Gauthier, L. P. B. Katehi, and G. M. Rebeiz, IEEE Trans.
Microwave Theory Tech. 48, 158 (2000).
114. R. Zhou, H. Zhang, and H. Xin, to be submitted to IEEE Trans. Antennas
Propagat. (2010).
115. R. Zhou, H. Zhang, and H. Xin, to be submitted to IEEE Antennas and Wireless
Propag. Lett. (2010).
155
116. H. Zhang, R. Zhou, Z. Wu, H. Xin, and R. W. Ziolkowski, Microw. Opt. Technol.
Lett. 52, 466 (2010).
117. J. Oppermann, M. Hamalainen, and J. Linatti, UWB Theory and Applications,
(John Wiley & Sons, Inc., 2004).
118. J. K. Cavers and M. W. Liao, IEEE Trans. Veh. Technol. 42, 581 (1993).
119. J. Crols and M. S. J. Steyaert, IEEE J. Solid-Sate Circuits 30, 1483 (1995).
120. M. Valkama, M. Renfors, and V. Koivunen, IEEE Trans. Signal Process. 49,
2335 (2007).
Документ
Категория
Без категории
Просмотров
0
Размер файла
3 719 Кб
Теги
sdewsdweddes
1/--страниц
Пожаловаться на содержимое документа