close

Вход

Забыли?

вход по аккаунту

?

A Microwave Interferometer to Enhance Sensitivity ofMicrowave Measurement Systems

код для вставкиСкачать
UNIVERSITY OF CALGARY
A Microwave Interferometer to Enhance Sensitivity of
Microwave Measurement Systems
by
Dmitri Kagan
A THESIS
SUBMITTED TO THE FACULTY OF GRADUATE STUDIES
IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE
DEGREE OF MASTER OF SCIENCE
DEPARTMENT OF ELECTRICAL AND COMPUTER ENGINEERING
CALGARY, ALBERTA
April 2012
© Dmitri Kagan 2012
Library and Archives
Canada
Bibliothèque et
Archives Canada
Published Heritage
Branch
Direction du
Patrimoine de l'édition
395 Wellington Street
Ottawa ON K1A 0N4
Canada
395, rue Wellington
Ottawa ON K1A 0N4
Canada
Your file Votre référence
ISBN:
978-0-494-88240-5
Our file Notre référence
ISBN:
NOTICE:
978-0-494-88240-5
AVIS:
The author has granted a nonexclusive license allowing Library and
Archives Canada to reproduce,
publish, archive, preserve, conserve,
communicate to the public by
telecommunication or on the Internet,
loan, distrbute and sell theses
worldwide, for commercial or noncommercial purposes, in microform,
paper, electronic and/or any other
formats.
L'auteur a accordé une licence non exclusive
permettant à la Bibliothèque et Archives
Canada de reproduire, publier, archiver,
sauvegarder, conserver, transmettre au public
par télécommunication ou par l'Internet, prêter,
distribuer et vendre des thèses partout dans le
monde, à des fins commerciales ou autres, sur
support microforme, papier, électronique et/ou
autres formats.
The author retains copyright
ownership and moral rights in this
thesis. Neither the thesis nor
substantial extracts from it may be
printed or otherwise reproduced
without the author's permission.
L'auteur conserve la propriété du droit d'auteur
et des droits moraux qui protege cette thèse. Ni
la thèse ni des extraits substantiels de celle-ci
ne doivent être imprimés ou autrement
reproduits sans son autorisation.
In compliance with the Canadian
Privacy Act some supporting forms
may have been removed from this
thesis.
Conformément à la loi canadienne sur la
protection de la vie privée, quelques
formulaires secondaires ont été enlevés de
cette thèse.
While these forms may be included
in the document page count, their
removal does not represent any loss
of content from the thesis.
Bien que ces formulaires aient inclus dans
la pagination, il n'y aura aucun contenu
manquant.
Library and Archives
Canada
Bibliothèque et
Archives Canada
Published Heritage
Branch
Direction du
Patrimoine de l'édition
395 Wellington Street
Ottawa ON K1A 0N4
Canada
395, rue Wellington
Ottawa ON K1A 0N4
Canada
Your file Votre référence
ISBN:
978-0-494-88240-5
Our file Notre référence
ISBN:
NOTICE:
978-0-494-88240-5
AVIS:
The author has granted a nonexclusive license allowing Library and
Archives Canada to reproduce,
publish, archive, preserve, conserve,
communicate to the public by
telecommunication or on the Internet,
loan, distrbute and sell theses
worldwide, for commercial or noncommercial purposes, in microform,
paper, electronic and/or any other
formats.
L'auteur a accordé une licence non exclusive
permettant à la Bibliothèque et Archives
Canada de reproduire, publier, archiver,
sauvegarder, conserver, transmettre au public
par télécommunication ou par l'Internet, prêter,
distribuer et vendre des thèses partout dans le
monde, à des fins commerciales ou autres, sur
support microforme, papier, électronique et/ou
autres formats.
The author retains copyright
ownership and moral rights in this
thesis. Neither the thesis nor
substantial extracts from it may be
printed or otherwise reproduced
without the author's permission.
L'auteur conserve la propriété du droit d'auteur
et des droits moraux qui protege cette thèse. Ni
la thèse ni des extraits substantiels de celle-ci
ne doivent être imprimés ou autrement
reproduits sans son autorisation.
In compliance with the Canadian
Privacy Act some supporting forms
may have been removed from this
thesis.
Conformément à la loi canadienne sur la
protection de la vie privée, quelques
formulaires secondaires ont été enlevés de
cette thèse.
While these forms may be included
in the document page count, their
removal does not represent any loss
of content from the thesis.
Bien que ces formulaires aient inclus dans
la pagination, il n'y aura aucun contenu
manquant.
UNIVERSITY OF CALGARY
FACULTY OF GRADUATE STUDIES
The undersigned certify that they have read, and recommend to the Faculty of Graduate
Studies for acceptance, a thesis entitled "A Microwave Interferometer to Enhance
Sensitivity of Microwave Measurement Systems" submitted by Dmitri Kagan in partial
fulfilment of the requirements of the degree of MASTER OF SCIENCE.
Supervisor, Dr. Elise C. Fear
Department of Electrical and Computer Engineering
Dr. Leonid Belostotski
Department of Electrical and Computer Engineering
Dr. Fadhel Ghannouchi
Department of Electrical and Computer Engineering
Dr. David Knudsen
Department of Physics and Astronomy
Date
Abstract
A single source microwave interferometer is proposed to enhance the sensitivity of a
microwave measurement system. Theory of operation is provided to gain understanding
of the device’s generated null and the concept of a virtual match. The design and steps
leading to its realization are discussed and developed through a comprehensive analysis
involving simulations and prototyping. The performance of the interferometer is
demonstrated through multiple applications, involving low contrast and high impedance
measurements. Due to the ability of the interferometer to easily integrate with multiple
measurement systems, it has potential benefiting applications that require extreme
sensitivity and ones relying on differential information.
ii
Acknowledgements
First and foremost, I would like to thank my supervisor Dr. Elise Fear for her guidance
throughout my studies. I feel privileged to have had an opportunity to learn from her
knowledge and experience.
I would like to express my gratitude to Erwin Siegel, who provided much valued
mentoring throughout the project. Special thanks are extended to Jeremie Bourqui,
Hassan Tanbakuchi and Tom Le for their advice and expertise in many technical aspects
of my work.
I am grateful to Roger Stancliff and Agilent Technologies for their support in this
project. As well as everyone at Agilent who took time out of their day to help answer
questions.
To all members of TSAR and AEG thank you for a great source of inspiration and
insight. It has been a pleasure sharing ideas and laughs.
I would also like to thank my parents, Ana and Pavel Kagan, and my brother Tal
for their patience and support, which allowed me to focus on my work.
Finally, I would like to thank Irina Mikenina for her understanding and
encouragement.
iii
Table of Contents
Abstract .......................................................................................................................... ii
Acknowledgements ....................................................................................................... iii
Table of Contents ...........................................................................................................iv
List of Tables.................................................................................................................. vi
List of Figures and Illustrations .....................................................................................vii
List of Symbols, Abbreviations and Nomenclature ......................................................... xi
CHAPTER ONE: INTRODUCTION .............................................................................. 1
1.1 Basic description of interferometry ........................................................................ 3
1.2 Research goals ....................................................................................................... 4
1.3 Summary of contributions ...................................................................................... 5
1.4 Thesis outline......................................................................................................... 6
CHAPTER TWO: BACKGROUND LITERATURE REVIEW ....................................... 8
2.1 Network analyzer measurements ............................................................................ 9
2.1.1 Network analyzer basic operation .................................................................. 9
2.1.2 Limitations .................................................................................................. 12
2.1.3 Large impedance limitations to capacitive measurements ............................. 15
2.1.4 Techniques overcoming network analyzer limitations .................................. 16
2.2 Applications of interferometry in microwave measurements ................................ 17
2.2.1 Characterization of single nanoscale magnetic structures using
transmission interferometry .......................................................................... 18
2.2.2 Varactor characterization with reflection interferometry............................... 19
2.2.3 Single cell detection with reflection interferometry ...................................... 20
2.3 Relative measure Microwave Interferometry ........................................................ 21
2.4 Noise floor applied to Interferometry ................................................................... 23
2.5 Microwave Interferometery to address Tissue Sensing Adaptive Radar (TSAR)
sensitivity limitations ......................................................................................... 25
2.6 Preliminary study ................................................................................................. 26
2.7 Summary ............................................................................................................. 29
CHAPTER THREE: TWO SOURCE INTERFEROMETRY......................................... 31
3.1 Theory ................................................................................................................. 32
3.2 Prototype ............................................................................................................. 34
3.2.1 Test bed ....................................................................................................... 35
3.3 Interferometer Results .......................................................................................... 36
3.4 Data-minus-memory measurement procedure ...................................................... 39
3.5 Performance Metrics ............................................................................................ 40
3.5.1 Comparison experiment 1 ............................................................................ 41
3.5.2 Comparison experiment 2 ............................................................................ 43
3.6 Summary ............................................................................................................. 46
iv
CHAPTER FOUR: SIMULATIONS OF A SINGLE-SOURCE
INTERFEROMETER ........................................................................................... 48
4.1 Simple simulation model...................................................................................... 50
4.2 Component modelling .......................................................................................... 52
4.2.1 Input/output Amplifier – TC271................................................................... 54
4.2.2 Attenuator .................................................................................................... 58
4.2.2.1 Attenuator design ............................................................................... 58
4.2.2.2 Bond wire ........................................................................................... 61
4.2.2.3 PIN diode model ................................................................................. 62
4.2.2.4 Attenuator response ............................................................................ 63
4.3 Upgraded model .................................................................................................. 65
4.3.1 Signal propagation ....................................................................................... 68
4.3.2 Signal- to-Noise ratio ................................................................................... 71
4.4 Summary ............................................................................................................. 73
CHAPTER FIVE: HARDWARE IMPLEMENTATION ............................................... 74
5.1 PIN diode attenuator ............................................................................................ 75
5.1.1 DC supply sensitivity ................................................................................... 79
5.1.1.1 Low pass filter .................................................................................... 80
5.1.1.2 Voltage desensitizing (resolution enhancement).................................. 81
5.1.1.3 Current source implementation ........................................................... 82
5.2 Microwave interferometer hardware build ............................................................ 83
5.2.1 Hardware deviations from simulations ......................................................... 83
5.2.2 Design corrections ....................................................................................... 84
5.2.3 Final output ................................................................................................. 85
5.3 Comparison to simulation build ........................................................................... 88
5.4 Understanding the null ......................................................................................... 90
5.5 Summary ............................................................................................................. 94
CHAPTER SIX: EXPERIMENTAL VALIDATIONS AND APPLICATIONS ............. 96
6.1 Mini anechoic chamber ........................................................................................ 97
6.1.1 Interferometer measurements of low contrast glycerin ................................. 99
6.1.2 Comparison to current technique ................................................................ 102
6.2 Oil tank test bed – Low relative permittivity contrast comparison ...................... 103
6.2.1 Interferometer sensitivity response to low contrast objects ......................... 104
6.2.2 Comparison of techniques .......................................................................... 106
6.3 MEMS ............................................................................................................... 107
6.3.1 Capacitance sensitivity ............................................................................... 108
6.3.2 Capacitance quantification ......................................................................... 110
6.4 Conclusion ......................................................................................................... 112
CHAPTER SEVEN: CONCLUSION .......................................................................... 114
7.1 Future Work ...................................................................................................... 116
BIBLIOGRAPHY ....................................................................................................... 118
APPENDIX A: APLAC MODEL ................................................................................ 123
v
List of Tables
Table 2.1: Reflection coefficient sensitivity to impedance changes in specified
regions. .................................................................................................................. 14
Table 2.2: Comparison between two methods of reflection interferometry. .................... 22
Table 3.1: Result summary for comparison test of current and two source nulling
technique. The performance of both systems is defined by the delta between
reference and object measurement. ......................................................................... 42
Table 4.1: TC271 Behaviour summary. ......................................................................... 55
Table 4.2: Signal propagation through the interferometer. .............................................. 69
Table 4.3: Signal flow through interferometer at -60 dBm input..................................... 70
Table 4.4: Noise flow through the interferometer stages................................................. 72
Table 4.5: SNR calculated for the various stages from noise and signal data. ................. 73
Table 5.1: PIN diode attenuator performance test measurements with MPN-7300. ......... 78
Table 5.2: Interferometer signal propagation using 1 KHz BW signal simulation. .......... 91
Table 5.3: Signal propagation through the output stage of the MI................................... 94
Table A.1: APLAC model parameter breakdown. ........................................................ 123
vi
List of Figures and Illustrations
Figure 2.1: Simplified network analyzer diagram. .......................................................... 11
Figure 2.2: S-Parameter measurement of a DUT. ........................................................... 12
Figure 2.3: Reflection coefficient sensitivity for the simple case of a purely resistive
load change (ZL) with constant characteristic impedance (Zo = 50). Calculation
done following equation (2-5). ............................................................................... 14
Figure 2.4: BAVA antenna response [30]....................................................................... 17
Figure 2.5: Transmission Interferometery for common-mode rejection of noise
(unwanted reflections). ........................................................................................... 19
Figure 2.6: Interferometery reflection cancellation setup. ............................................... 21
Figure 2.7: Simplified block diagram of MI. .................................................................. 23
Figure 2.8: Test setup for performance comparison of current TSAR procedure and
Interferometery. ..................................................................................................... 27
Figure 2.9: Measured reflection of Eccostock rod (R ~ 6.35mm & εr ~ 15) with
interferometer nulling at 4.055 GHz using Agilent’s PNA-L and BAVA antenna.
The x-axis represents the location of a data point on the trace at the chosen
frequency on the NA. Initially the null (green line) is created with no object
present in the oil tank. Then the Eccostock rod is inserted and its response is
recorded (orange line). The rod is removed and the response of the null is
recorded (green line shifted). This process is repeated until four measurements of
null and object are recorded. .................................................................................. 28
Figure 3.1: Two sources nulling technique block diagram on a network analyzer
platform. ................................................................................................................ 32
Figure 3.2: Phase retrieval procedure for two source nulling. ......................................... 34
Figure 3.3: Canola oil tank measurement setup with test objects. ................................... 35
Figure 3.4: Two source nulling repeatability with a null set to 6.5 GHz and a BAVA
antenna in a tank filled with Canola oil without a rod present using PNA-X NA
as platform. Measurements are done consecutively in 5 second intervals with
null conditions applied once at the start and no re-adjustments throughout the
test. ........................................................................................................................ 37
Figure 3.5: Two source interferometer object detection validation results. The mean
of the null is ~-65 dBm with a standard deviation (σ) of 3. For Eccostock rod the
vii
mean is -60 dBm with σ = 1.6, the Delrin rod mean is -64 dBm with σ = 5.5 and
the Acrylic rod mean is -61 with σ = 4.3. ............................................................... 38
Figure 3.6: Antenna response elimination experiment MI test setup. .............................. 44
Figure 3.7: Amplitude spectrum of excitation signal used to synthesize the desired
time-domain signal prior to the chirp-z transform. .................................................. 45
Figure 3.8: DMM technique (left) and Interferometer technique (right) antenna
response elimination. Antenna only is a result of the antenna’s interface with free
space while object only is the interface between free space and the copper plate. ... 46
Figure 4.1: Basic block diagram of a single source interferometer. ................................. 49
Figure 4.2: Simple ADS model of a single source MI. ................................................... 51
Figure 4.3: Simple ADS model response without null optimization shows multiple
nulls with a null of interest at 1.8 GHz with amplitude of -7.244 dBm. ................... 51
Figure 4.4: Simple ADS model response showing an optimized null at 1.79 GHz with
amplitude of -129.412 dBm. ................................................................................... 53
Figure 4.5: Updated interferometer block diagram. ........................................................ 55
Figure 4.6: TC271 Amplifier possible amplifier setups. ................................................. 56
Figure 4.7: TC271 Gain simulations with various setup scenarios. For SE-SE gain is
26 dB, DE-DE gain is 31 dB and DE-SE or SE-DE gain is 28.5 dB........................ 57
Figure 4.8: TC271 NF simulations with various setup scenarios. For SE-SE and SEDE NF is ~15 dB while DE-DE and DE-SE NF is ~12.5 dB. ................................. 57
Figure 4.9: PIN diode attenuator with dual DC control modeled in ADS. ....................... 62
Figure 4.10: PIN diode ADS APLAC model. The main parameters of interest in this
model are Rmin and Rmax representing the minimum and maximum RF resistance
as well as the SDD1P, which is a two port model for a current controlled
variable resistor. Inductor L and capacitor C represent the connection inductance
and diode capacitance respectively. ........................................................................ 63
Figure 4.11: PIN diode attenuator insertion loss V1 = 0 V, V2 = 0 V. ............................. 64
Figure 4.12: PIN diode attenuator response V1 = 20 V, V2 = 20 V. ................................ 64
Figure 4.13: Nulling interferometer complete design. Signal flow through
interferometer at various stages is shown. Stages are labelled according to their
respective purpose a) System input b) Input confirmation c) Excess noise
verification d) Null formation e) Null amplification f) Second amplification g)
Third amplification and output. .............................................................................. 66
viii
Figure 5.1: PIN diode attenuator using AEROFLEX diodes. .......................................... 77
Figure 5.2: MPN-7300 attenuator response at 0V bias with 560 ohm external on bias
lines. ...................................................................................................................... 77
Figure 5.3: MADP-000165-01340W attenuator response at 0V bias. ............................. 79
Figure 5.4: Low pass filter implemented at attenuator to improve sensitivity. ................ 81
Figure 5.5: Modified interferometer block diagram. ....................................................... 85
Figure 5.6: Output of MI with dc biasing working properly. .......................................... 86
Figure 5.7: Interferometer connections to a network analyzer. ....................................... 86
Figure 5.8: Microwave interferometer modified topology connection. ........................... 88
Figure 5.9: ADS schematic matching hardware implementation. ................................... 89
Figure 5.10: Interferometer hardware implementation output null response.................... 91
Figure 5.11: Output stage of MI. .................................................................................... 93
Figure 5.12: Output of cascaded amplifier output stage of interferometer with 50 Ohm
matched loads on differential lines. ........................................................................ 94
Figure 6.1: Mini anechoic chamber setup using the MI with a PNA-X (N5242A
26.5GHz) as a platform. Top view made available by Jeremie Bourqui
(University of Calgary) shows dimensions in mm for Horn antenna (outside
chamber), tube insertion location and chamber measurements. ............................... 98
Figure 6.2: Interferometer output when the system is dormant. .................................... 100
Figure 6.3: System response to the addition of glycerin using an interferometry. ......... 101
Figure 6.4: Dielectric constant change as water is diluted with glycerin in room
temperature at 4.8 GHz. ....................................................................................... 101
Figure 6.5: Interferometer and DMM performance comparison by looking at relative
changes for anechoic chamber results. .................................................................. 103
Figure 6.6: Oil tank test bed for revisited low contrast comparison. ............................. 104
Figure 6.7: Interferometer response to Acrylic rod insertion in an oil test bed. ............. 105
Figure 6.8: Interferometer and DMM performance comparison by looking at relative
changes for oil tank experiment. ........................................................................... 107
Figure 6.9: MEMS capacitor measurement setup. ........................................................ 108
ix
Figure 6.10: Interferometer output at nulled state (red) and shifted null (green) in
response to a change. ........................................................................................... 110
Figure 6.11: Linear relationship between null frequency shifts and capacitance values. 111
x
List of Symbols, Abbreviations and Nomenclature
Abbreviations
Definition
ADC
ADS
BAVA
BPF
BW
CPLR
DC
DE-DE
DE-SE
DMM
DUT
FMR
IF
LHS
LNA
LPF
MEMS
MI
NA
NF
PNA
REF
RF
RHS
SE-DE
SE-SE
SNR
S-parameters
SRC
TSAR
VNA
Analog to Digital Converter
Advanced Design System
Balanced Antipodal Vivaldi Antenna
Band Pass Filter
Bandwidth
Directional Coupler
Direct Current
Differential to Differential Topology
Differential to Single Ended Topology
Data Minus Memory
Device Under Test
Ferromagnetic Resonance
Intermediate Frequency
Left Hand Side
Low Noise Amplifier
Low Pass Filter
Micro Electro Mechanical Systems
Microwave Interferometer
Network Analyzer
Noise Figure
Performance Network Analyzer
Reference
Radio Frequency
Right Hand Side
Single Ended to Differential Topology
Single Ended to Single Ended Topology
Signal to Noise Ratio
Scattering Parameters
Source
Tissue Sensing Adaptive Sensor
Vector Network Analyzer
xi
Symbol
a
B
b
C
Γ
H
L
ω
R
S11
S12
S21
S22
t
Y
Z
ZL
Zo
Modified Matching Amplitude
Incident Wave
Frequency Bandwidth
Reflected Wave
Capacitance
Relative Permittivity
Frequency
Modified Matching Phase
Phase of Sinusoid
Reflection Coefficient
Hybrid Parameters
Boltzmann’s Constant
Inductance
Angular Frequency
Thermal Noise Power
Radius
Maximum RF Resistance
Minimum RF Resistance
Node Series Resistance
Forward Reflection Coefficient
Reverse Transmission Coefficient
Forward Transmission Coefficient
Reverse Reflection Coefficient
Modified Matching Signal
Signal to Receiver
Reflected Signal
Electrical Conductivity
Temperature
Time
DC Voltage
Admittance Parameters
Impedance Parameters
Capacitor Impedance
Load Impedance
Characteristic Impedance
xii
1
Chapter One: Introduction
Over the course of the past century, applications based on microwave frequencies have
made their way into our daily lives. This is most noticeable in the wireless
communication field, where we use items such as cell phones, navigation tools and
satellite television among many others. Aside from the general population, major
consumers of microwave devices, as well as test and measurement tools are the military
and aerospace for applications such as remote sensing. It is obvious that microwave
frequencies are of great interest for various applications with new ones emerging from
research centers and universities. In recent years there has been a great deal of work done
in the medical field with microwaves. For example, localized heating [1] for therapeutic
purposes, thermal ablation [2] of tissue and malignant tissue detection via sensing and
imaging have been explored.
The investigation of malignant tissue detection [3] resulted in the development of
various prototype systems that detect malignancies located in the breast in an attempt to
provide means of early stage detection. These include tomographic [4], holographic [5]
and radar based [6] imaging methods. Both the tomographic prototype developed at
Dartmouth College, as well as the radar based prototype developed at the University of
Bristol, has been tested on numerous patients [7]. Aside from the development of
prototypes, a great deal of work has been done in holographic methods [8]. Although
holographic methods feature benefits such as real-time performance, they are currently in
experimental stages utilizing phantoms as imaging objects. Even though several
prototypes have made it to clinical trials, microwave based detection systems have many
2
challenges to overcome before they become readily available. These challenges are
discussed by looking at the prototype system developed at the University of Calgary.
At the University of Calgary research is underway developing a radar based
microwave measurement system for breast cancer detection. The challenge is to
distinguish signals reflecting from low contrast areas [9], while overcoming large
unwanted reflections [10]. Low contrast areas are due to the difference in properties
between malignant and glandular tissue being only 10% [11]. The skin layer response
dominates the signal due to the significant differences in properties of the skin and
underlying fat. This problem requires a highly sensitive microwave measurement system.
In addition, techniques such as the use of contrast agents [12-14] are under investigation
in hope to bring out the tumorous region through contrast enhancement. These contrast
agents also require sensitivity to changes in properties resulting from accumulation of the
agents.
The sensitivity of a system dictates the lowest possible signal power level that can
be recorded. Various contributions such as receiver noise and component mismatch can
degrade sensitivity and limit the ability to detect minute powers of interest. In microwave
measurement systems, a major contributor to the degradation of sensitivity is the match
of the system to the device under test (DUT). When a device is not well matched to the
characteristic impedance of the system (50 ohms), a large reflection occurs. This
reflection poses an obstacle while measuring devices of large impedances, as well as
when attempting to isolate the response of an object located in an area of low contrast
levels. The purpose of this of study is to develop an improved methodology to tackle
these issues.
3
1.1 Basic description of interferometry
Interferometry involves the destructive and constructive interference of signals in order to
obtain practical information by looking at differences from a known state. Unique
information can be extracted from the output of such combination because the output has
dependency on the phase difference between the signals. Two types of detection are
possible with an interferometer, heterodyne and homodyne.
Heterodyne detection [15] involves non-linear mixing of a signal with a reference.
This mixing results in an output of multiple components, one of which is a beat
component at the difference frequency of the original signals and carries useful
information such as amplitude and phase as shown:
(
)
(
)
((
)
(1-1)
(
))
The second type, the one which this thesis examines, is homodyne detection [15].
In this case, the combining occurs between signals having the same frequency, since the
reference signal and the one of interest are both generated by the same source. The
amplitude of the output is dependent on the phase difference between the signals (1-2).
Increased sensitivity is achieved by setting the amplitudes to be equal and tuning the
phase difference to 180 degrees, resulting in noise. The Microwave Interferometer (MI)
implements this concept, by providing a system the ability to cancel unwanted
reflections, which otherwise clutter the receiver and keep the system operating at a state
that is greater than the noise level.
4
(
)
(
)
(
)
(1-2)
1.2 Research goals
The purpose of this study is to develop a Microwave Interferometery tool that has the
ability to integrate with multiple microwave measurement systems in order to provide
enhanced sensitivity over a wide bandwidth (BW). The approach explored in this study is
characterized by the use of relative differential information. The various specific aims of
this project are as follows:

Gain insight into the performance of the first attempted prototype provided by
Agilent Technologies.

Implement a two source interferometric setup to perform proof-of-concept
validation experiments. Experiments must be done over a wide BW.

Design a comprehensive simulation model containing components with realistic
behaviour. Model performance should be in close agreement to expected
hardware implementation.

Develop a MI hardware prototype and validate its performance with the
simulation model.

Integrate the MI with multiple measurement systems to show improved sensitivity
as well as other possible utilizations.
5
1.3 Summary of contributions
The research work done in this study is successful in accomplishing its defined goals.
These accomplishments can be seen through the comprehensive insight that is offered
throughout the chapters with regards to the intricacies of the MI and its performance. The
success of this research work yields three valuable contributions/accomplishments:
simulation model, hardware implementation and identification of possible applications.
The simulation model developed in this study was designed to represent a realistic
model of the MI and to push the limits of the readily available block diagram models
available in literature. This allowed performing a detailed analysis of the signal
propagation through the device, providing access to nodes that are otherwise difficult to
measure. From this, valuable insight was extracted to understand the relationship of the
null to the noise floor of the system. Experiments showed agreement between the
simulation model and the hardware prototype, validating that concepts learned from
simulations can be applied to hardware.
The outcome of the project was a fully operational hardware prototype of the MI.
The MI is constructed from readily available components, making it relatively
inexpensive. In comparison to currently researched MIs, the design of this device allows
for extreme flexibility. The device is able to integrate with ease to a measurement system
and perform its nulling operation regardless of sensor type.
Lastly, by exploring applications with high demand for sensitivity, the MI is
shown to perform in par with or better than current techniques. In order to outperform the
current techniques the MI had to be pushed to its limits. To do so an in-depth analysis of
6
the null was launched and provided great knowledge on the possible relative
measurements both in magnitude and frequency that can be done using this device.
1.4 Thesis outline
Chapter 2 begins by initially discussing the obstacles that stand in the way of
making accurate microwave measurements when a system is facing large reflections. A
review is provided of selected techniques that have been previously proposed to tackle
this, as well as current research directed at the possibility of sensitive measurements
under the difficult conditions.
Chapter 3 provides a simplified implementation of microwave interferometry by
using two sources for a proof-of-concept purpose. This generates basic understanding of
performance expectations, as well as design obstacles.
Next, Chapter 4 covers the first step towards a single source implementation of the
interferometer. The discussion covers the step-by-step development of a RF simulation
model representing the topology of the interferometer and providing the behaviour of its
components.
With a model in place, Chapter 5 explores prototyping the interferometer by
following the previously developed schematic and the necessary additions required to
obtain a properly working device. A comparison between simulations and hardware is
also provided.
Using the prototype, Chapter 6 describes the various experiments validating that
interferometer is able to take measurements, provides improved sensitivity and is
straight-forward to apply to multiple systems.
7
Finally, Chapter 7 concludes the results of this research and provides a brief
overview of the possible future work, which can be done in order to improve the
performance of this device.
8
Chapter Two: Background Literature Review
Utilizing interferometry to obtain meaningful information is not a new concept. In fact
interferometers have been key investigative tools in multiple fields of study. In
astronomy, an interferometer can be used to precisely measure the position of a source
and its equivalent diameter [16]. In optics, a phase shifting interferometer can be used to
extract the phase of an electric field [17]. Physics applications of interferometry include
studying noise phenomena in microwave components [18] and applications in biology
can involve measurements of physiological movements by non-contact means [19],
among many others [20-22]. The basic principle of this methodology is to allow two
similar signals (e.g. equal in frequency), to combine constructively (in phase) or
destructively (out of phase) to gain useful information. For example, consider two
signals, where one is used as reference and the other undergoes a change. The signals are
combined to produce an output. The shape of the generated output is highly dependent on
the resulting phase offset, which produces a particular pattern of constructive and
destructive interference between the two signals. This output signal can provide useful
information regarding the obstacles encountered along the signal path, as various objects
will result in unique phase shifts.
The focus of this study is to generate an output as described above, resulting from
two signals with the same frequency having a phase offset. Instead of observing the
complete output, only a selected region where destructive interference takes place is of
concern. The idea is that by operating a system around this region, it is possible to obtain
greater sensitivity to changes between the two signals. We refer to this approach as
9
Relative Measure Differential Interferometry. To provide motivation for pursuing such an
approach, the standard approach to microwave measurement using a Network Analyzer
(NA) is described and limitations in terms of sensitivity are identified.
Next,
interferometry in active microwave measurement is reviewed with a focus on
applications that demand sensitivity. This leads to a description of the objective of this
work, the development of a differential interferometry approach to provide extreme
sensitivity to changes in target signals. Preliminary results are provided, illustrating the
potential for this approach and motivation for further exploration of this technique.
2.1 Network analyzer measurements
A NA is a measurement tool, which allows the characterization of devices in the RF and
Microwave frequency ranges. Since the interferometer requires three external
components (source, coupler and receiver) to operate, the NA is an effective tool for
initial testing as it possesses all three. In this study the interferometer device will be
integrated together with the NA. Therefore there is a need to have a basic understanding
of how a NA operates. More importantly the main focus is on the sensitivity limitation of
the NA, which is needed to fully understand what the interferometer offers and how it
will achieve this. These insights will be further developed in this chapter and will be
solidified throughout this study.
2.1.1 Network analyzer basic operation
Devices and components that operate at high frequencies cannot be simply characterized
by measuring voltage and current levels at the input and output ports. The reasons for this
10
are due to the impedance of the probe which will interfere with the characterization of the
device, the difficulty of placing the probe and the fact that by connecting shorts and
opens to active devices there is a risk of oscillation as well as breaking down of parts.
This resulted in the development of the Scattering (S) Parameters that are able to
characterize high frequency networks accurately and safely [23].
S-Parameters offer great benefits such as having a direct relation to commonly
known measures such as gain and loss. They are obtained by measuring voltage travelling
waves which is relatively simple through the use of a VNA. There is no need to connect a
short or an open to the device for characterization. S-Parameter matrices provide the
relationship between the reflected and incident waves at the ports of a device. SParameters can be also be converted to other parameters, such as Z, Y and H parameters
[24]. Finally, S-Parameters also are very easy to import into modern simulation tools
such as Advanced Design System (ADS, Agilent Technologies), which will be apparent
in later chapters as it is the tool of choice.
Figure 2.1 depicts a basic block diagram showing some of the key components of
a NA as applied to measurement of a 2-port device [24]. SRC 1 and SRC 2 are two
sources, which are used to send high frequency signals into a DUT. A portion of the
source’s signal gets coupled using a directional coupler to be used as a reference in order
to obtain phase information about the DUT. The input and output of the DUT are
connected to ports 1 and 2 on the NA with no particular orientation. Energy is allowed to
propagate from the ports into the DUT, where it undergoes reflection and transmission.
These reflected and transmitted signals are picked up through another set of directional
couplers. All of the above coupled signals pass through a mixer to mix down to an
11
Figure 2.1: Simplified network analyzer diagram.
intermediate frequency (IF), which is sampled and processed to obtain the S-Parameters
of the DUT. The S-Parameters provide the characteristics of the DUT over the measured
frequency range.
Figure 2.2 shows the incident (a1 and a2) and reflected (b1 and b2) voltage signals
for a 2-port DUT [24]. These voltages are the necessary quantities of a network from
which the S-Parameters are derived. When measuring these quantities a condition is
imposed on the system where the incident signal from the opposite port is set to zero.
This is done by terminating the port with a perfect match, which is equal to Zo
(characteristic impedance) of the test system. Equations (2-1) – (2-4) show the
relationship between voltages and S-Parameters. It is important to note that S-Parameters
are complex values generally expressed in units of dB and therefore both magnitude and
phase of the voltage measurements are needed for completeness.
12
a1
DUT
b1
b2
a2
Figure 2.2: S-Parameter measurement of a DUT.
(2-1)
(2-2)
(2-3)
(2-4)
2.1.2 Limitations
The reflection coefficient, generally expressed as Γ (linear scale) is a network analysis
measure from which the impedance of an unknown DUT can be calculated [24]. For the
simple case of a purely resistive load, the reflection coefficient is expressed by equation
(2-5), where ZL is the impedance of the load and Z o is the impedance of the measurement
system (50 Ω). From Γ it is also possible to calculate the S 11 parameter expressed in dB
for a circuit as shown in equation (2-6).
(2-5)
13
( )
(2-6)
Figure 2.3 is a plot showing the variation in value of the reflection coefficient as the load
impedance changes [25]. Three significant regions are noted on the plot with two outer
regions having low resolution and a central region having the highest resolution. In terms
of system sensitivity, this means that when the load is well matched to the system the
resolution is at its best and it is possible to distinguish reflections resulting from closely
spaced impedances. Table 2.1 confirms that the change in the reflection coefficient for
the same percent change in impedance is not the same in the three regions. The case
presented in the table (Zo=50 Ω) shows that a 20% change in impedance around 50 ohm
region can mean a reflection coefficient change that is 30 times greater than that for a 20
% impedance change in the low resolution regions.
The significance of the change that occurs in the reflection coefficient when a
load is present comes in play when taking into consideration the resolution of the receiver
(i.e. ADC). In the low resolution regions the receiver will not distinguish the load change
and will output the same reflection coefficient. This means that if one were to calculate
the load resistance from the reflection coefficient the value of the load will be the same,
even though the reflection came from two different loads. By operating the system at the
high sensitivity region using a load that is well matched, the receiver will be able to pick
up changes in the load that were previously not possible in the low resolution region and
provide an accurate measurement. This resolution problem is what the interferometer is
supposed to tackle and improve upon.
14
Figure 2.3: Reflection coefficient sensitivity for the simple case of a purely resistive
load change (ZL) with constant characteristic impedance (Zo = 50). Calculation done
following equation (2-5).
Table 2.1: Reflection coefficient sensitivity to impedance changes in specified
regions.
Region
Z1 (Ω)
Z2 (Ω)
Γ1
Γ2
1.2
Impedance
Change
20%
-0.961
-0.953
Reflection
Change
0.008
Small Impedance
1
Match
Impedance
Large Impedance
50
60
20%
0
0.091
0.091
5000
6000
20%
0.98
0.983
0.003
15
2.1.3 Large impedance limitations to capacitive measurements
Sensitivity limitation poses a major problem when there is a need to accurately
characterize objects of large impedances [26]. There is a great need and benefits to the
development of Nano-scale components and devices due to their miniature size allowing
building IC’s that are capable of multiple operations. There is also a great deal of need to
measure characteristics of biological cells for medical advances whose characteristics are
obtained by looking at minute capacitance changes.
Nano-scale objects have tiny capacitances. From equation (2-7), it can be seen
that for small capacitance value the impedance is very large unless the frequency of
operation is pushed extremely high, since they are inversely proportional.
(2-7)
Where
For example, let c take a realistic value of 1 pF and assume a common frequency of
operation of 1 GHz
(2-8)
(
)(
)
16
2.1.4 Techniques overcoming network analyzer limitations
To improve the measurement system, averaging as well as reduction in the IF BW is
added using manual settings on a NA, thus further enhancing the dynamic range [27].
Both averaging and IF reduction achieves this task by reducing the amount of noise in the
receiver, thus reducing the noise floor of the test equipment. When vector averaging is
applied to a trace on the NA containing coherent and uncorrelated noise, the result is
reduction in the noise component, allowing noise floor reduction. The downside to this is
the resulting speed reduction in the measurement speed. While the IF BW reduces the
noise floor by having a narrower digital filter, this filters noise outside of the BW. Further
details regarding the noise floor reduction with the use of narrow BW are explored
throughout this study. Again the downside of this is speed reduction in data acquisition
time.
Outside of the above methods, matching techniques are used that specifically
build sensors to provide impedance matching to enhance sensitivity [28, 29]. One
example of a sensor is the Balanced Antipodal Vivaldi Antenna (BAVA) sensor [30]
used for microwave imaging, which is specifically designed for breast cancer detection.
This antenna is well matched to canola oil, which is the immersion medium in which a
breast is suspended and has a beamwidth of 34mm by 44mm at a distance of 20mm away
from the antenna aperture. Figure 2.4 shows the S11 response of the BAVA antenna, the
lower the S11 the closer the system to the high resolution region.
17
0
Measurement
Simulation
S11 [dB]
-10
-20
-30
-40
2
4
6
8
Frequency [GHz]
10
12
Figure 2.4: BAVA antenna response [30].
Such techniques have proved to be highly effective and yield good results when
attempting to measure responses from relatively small objects as well as objects of
interest that are located in low contrast areas making them difficult to detect. The
downside is that well matched antennas can get expensive and require a significant
amount of time to develop while post processing techniques have acquisition time
implications. The purpose of this study is to analyze, develop and test a method to
provide further sensitivity improvement to the detection of minute changes in a test
object.
2.2 Applications of interferometry in microwave measurements
Several measurement applications that suffer from sensitivity limitations of standard
microwave measurement approaches are explored in this section. For these applications
interferometry approaches have been proposed to solve the sensitivity problems. Two
18
different interferometry topologies are discussed. The first involves transmission, where a
signal propagates through a reference channel and a DUT to recombine and provide a
difference output. The other involves reflection, where a reference signal is mixed with a
reflected signal from a DUT to result in a null. Both systems improve sensitivity by
eliminating unwanted reflections.
2.2.1 Characterization of single nanoscale magnetic structures using transmission
interferometry
Ferromagnetic resonance (FMR) is a spectroscopic technique used in the investigation of
the magnetization of magnetic structures. This is one of many applications where the
sensitivity of conventional transmission line approaches such as lumped element
modelling [31] are lacking for the characterization of single nanoscale magnetic
structures due to generated excessive noise (unwanted reflections). Therefore the
traditional approaches can only be used for bulk materials. In order to address this issue,
an interferometry technique has been proposed [32] which claims an improvement of 20
dB over traditional methods as a result of the ability to cancel common mode noise. The
proposed technique shown in Figure 2.5 involves transmission interferometry by passing
a signal through a reference circuit and a DUT. After the two responses recombine, S21
is measured and the common-mode noise is rejected. This creates a null in the S 21
response at 4.3 GHz when the FMR frequency of the Permalloy nanowire is away from
the working frequency. Through the use of an external magnetic field, the FMR
frequency is shifted to 4.3 GHz, resulting in a change in the null. From this change,
properties of interest such as damping parameters can be derived.
19
Common-Mode Rejection
Reference
Device
Input
Splitter
Output
Combiner
DUT
Figure 2.5: Transmission Interferometery for common-mode rejection of noise
(unwanted reflections).
2.2.2 Varactor characterization with reflection interferometry
Sub fF varactors [26] are used in applications that operate at high frequencies such as
THz. The challenge in developing such varactors is the difficulty to accurately
characterize them in order to understand their behaviour and integrate them into designs.
The difficulty is due to their low capacitance resulting in very large impedance, thus
having significant sensitivity problems. A study has been conducted which proposes a
method to improve varactor characterization of aF capacitance values. This is achieved
through the use of an interferometer. The interferometer consists of a single source that is
divided between mixing and reference paths. The reflection from a DUT is mixed with a
modified reference signal to generate a null that is fed to an NA. The test setup used is
shown in Figure 2.6. The interest is to record signal variations around this null and with
the use of a calibration kit to extract capacitance values at various voltage biases. The
result is the capacitance characteristic of the varactor as a function of voltage. The
advantage of this technique is the cancellation of unwanted reflections and the ability to
20
tune a system to the high resolution region. [33] Reports successful characterization of a
300 aF varactor utilizing this method with a sensitivity of 90 dB.
2.2.3 Single cell detection with reflection interferometry
Medical applications generally involve the need to characterize and detect biological
cells. Cells are very small, which makes them difficult to measure especially if single
cells are being analyzed. Aside from the need to detect cells it is rather important to do so
with minimally invasive and non-destructive techniques. A study has been done to show
that by using Interferometery in a transmission setup [34] it is possible to obtain high
resolution (0.65 aF sensitivity) in the detection of a capacitance changes as single cells
flow past a sensing apparatus. The interferometer consists of a single source that is split
to generate reference voltage and to flow through a resonator after it undergoes phase
changes. The split signals are recombined using a mixer and the output is fed to a LockIn Amplifier. The purpose of the amplifier is to lock onto the mean of the interferometer
output and with the use of the phase delay component the output is tuned to produce a
null. The idea is that as cells flow through the resonator the output of the interferometer
will no longer be a null. Again, interferometry is being considered as a sensitivity
enhancing technique to obtain improved performance over traditional methods. The setup
for this measurement utilizes the general idea shown in the above Figure 2.5.
21
Interferometer
Sensor
DUT
NA
Figure 2.6: Interferometery reflection cancellation setup.
2.3 Relative measure Microwave Interferometry
It has been shown above that interferometry is a technique which is being investigated as
a method to greatly enhance the sensitivity of microwave measurement systems. Two
different setups of interferometry have been discussed. One setup involves transmission
Interferometery while the other uses reflection Interferometery. Studies into these
techniques are showing success and promise for sensitivity enhancement over traditional
methods.
The purpose of this study is to explore a different approach to reflection based
Interferometry. Unlike all of the studies presented above that are concerned with the
absolute measure of the objects under investigation, the idea here is to obtain relative
information. This means that the interest of this method is to measure the change from a
reference. Table 2.2 provides a summary of the key different features between the two
methods of performing reflection Interferometery. Throughout the rest of the chapters,
evidence will be provided to support the claims that the proposed Interferometry method
and setup is able to achieve extremely sensitive measurements and is able to meet all of
the mentioned advantages.
22
Table 2.2: Comparison between two methods of reflection interferometry.
Advantages
-
-
Disadvantages
Absolute Measure
Interferometry
Information is gained as for the
properties of the objects of
interest.
Enhanced sensitivity over
traditional methods.
-
Calibration procedure is
required.
-
Application dependent.
-
Relative Measure
Interferometry
Independent of application.
-
Same setup can be used
with multiple sensors.
-
No need for calibration
procedure.
-
Enhanced sensitivity over
traditional methods.
No information provided
into the absolute properties
of objects of interest.
-
The Microwave interferometer proposed in this study is a tool that is able to
cancel (null) unwanted reflections in order to detect small changes of interest. This is
achieved through the device being able to create a virtual match (50 ohm termination) to
the environment, which means that the system is sitting ideally at noise floor, enabling
detection of a small change. The better the virtual match, the closer the system is to being
at noise floor and the more sensitive the system is to any changes.
Figure 2.7 shows a basic block diagram of the MI. A source is used to generate an
input to feed a sensor, which will radiate towards an object and detect the reflection.
Directional couplers are used to couple portions of signals which will then be canceled in
the combiner by setting the attenuator to provide magnitude match while a 180 degree
phase shift results from the cables acting as transmission lines.
23
Source
Directional
Coupler
Directional
Coupler
Sensor
Attenuator
Combiner
Receiver
Figure 2.7: Simplified block diagram of MI.
The necessity of having 180 degrees phase shift dictates the frequencies where the nulls
occur, since different frequencies will experience a different phase shift through the same
transmission line. Prior to providing a more complex and comprehensive analysis of this
device it is important to understand the limitation. The null depth is an important measure
of the device and to understand it a brief discussion of the concept of noise floor is
needed and provided.
2.4 Noise floor applied to Interferometry
The noise floor of the system is the level of signal generated at the output of the system
when there is no input. There are many types of noise that contribute to the noise floor.
However the emphasis is on the thermal noise of the system for this study. Thermal noise
is the noise that is generated by the thermal agitation of electrons at steady state. Ideally
the interferometer would provide such a perfect virtual match as to bring the noise level
of the system to its thermal noise [35]. This provides the greatest possible dynamic range
24
and sensitivity because the system should then be able to detect signals that have power
greater than the noise floor. Thermal noise power is calculated as follows:
(2-9)
Where
k – Boltzmann’s constant (1.374e-23 J/K)
T – Temperature in units of Kelvin (K =
+ 273.16)
B – Frequency bandwidth (Hz)
For example, consider room temperature of 25
(T = 298 K) and assume a 1 kHz BW
signal. Then,
(
(
)
(2-10)
)
(2-11)
(2-12)
Assuming that the measurement system is dominated by thermal noise, the noise floor of
the system is at -144 dBm. If the system is able to detect signals up to 0 dBm the dynamic
range for the detection of signals is 144 dB.
It is important to note that by using this interferometry technique, the information
obtained is a relative measure. This implies that the interest is in detecting small changes
as they occur relative to a reference point rather than seeking the absolute measure of an
object. In general it is often very difficult to detect minute changes. Noise from the
environment or unwanted parasitics can also mask miniscule changes. These additional
factors can contribute to the noise floor of a measurement setup and reduce the dynamic
25
range, motivating practical testing of designs in order to assess the impact of these
sources of noise.
2.5 Microwave Interferometery to address Tissue Sensing Adaptive Radar (TSAR)
sensitivity limitations
At the University of Calgary the Electromagnetics group is currently investigating a
complementary technique for early breast cancer detection using microwaves. This
project is called TSAR [36]. As previously discussed the measurement procedure
currently utilized involves an ultra-wideband antenna. The data acquisition is performed
by subtracting the response of an object in a test bed with the response of only the test
bed, thus isolating the response of the object of interest. A more comprehensive
explanation is done in later chapters as this is the application that the interferometer is
targeted to improve. The current sensitivity of TSAR primarily relies on the match of the
antenna to the environment in which the breast is positioned, and can be calculated by
subtracting two consecutive sweeps on the NA with no object present. TSAR has shown
great promise in the detection of tumors located in the breast, although there is room for
improvement in measurement sensitivity. It is an intriguing application for the proposed
interferometer, as many of the benefits of the interferometer can be investigated through
this application. The sensor used in TSAR although optimized, produces unwanted
reflections. This is a problem that the interferometer can tackle by the creation of the
virtual null, which mimics a perfectly matched antenna.
The challenge for TSAR is the fact that malignancies are expected to have
electromagnetic properties 10% different than those of glandular tissue, and since many
26
tumors are located in glandular tissue the detection of these low contrast tumors at their
early stage is quite challenging. This is the perfect place to utilize the differential
properties of the interferometer to help distinguish malignant tissues. Also, the mismatch
between skin and fatty tissue results in large reflections that can mask any small internal
variations. Therefore the proposed interferometry technique which is able to null all of
these unwanted parasitics is believed to provide the needed sensitivity to detect small
tumors beyond the current ability of standard procedure. Although the application to
realistic TSAR data is of interest long-term, this thesis focuses on testing the
interferometer in known environments in order to better understand the potential and
performance limitations of this technology.
2.6 Preliminary study
To gain initial insight into the performance of the interferometer for the TSAR
application, a preliminary study is conducted in order to compare the two techniques.
Figure 2.8 shows the test setup utilized to perform the comparison. The BAVA antenna
currently used in TSAR is immersed in an oil tank and a rod is inserted with a relative
permittivity different than that of oil. The ability of both techniques to detect the change
is studied. The preliminary study was very simple and is discussed below. A more
comprehensive study and proof of improvement is presented in chapters to come.
27
Figure 2.8: Test setup for performance comparison of current TSAR procedure and
Interferometery.
The first pass at microwave interferometry was done by testing an interferometer
which utilized a Field Effect Transistor based attenuator that had been claimed to have
the ability to increase sensitivity and provide enhanced dynamic range through the
utilization of a differential measurement setup using a single source. This device had
been made available for testing and research by Agilent Technologies. Initial testing of
this system showed potential for detection of objects as well as changes, although the
device proved to be very unstable and to have poor repeatability. Figure 2.8 above
discusses the test setup that was used to investigate this interferometer build. Figure 2.9
shows data obtained by consecutively measuring the null depth and the reflection
detected from a test object. The results indicate a significant difference between scenarios
with and without the rod present.
28
Figure 2.9: Measured reflection of Eccostock rod (R ~ 6.35mm & εr ~ 15) with
interferometer nulling at 4.055 GHz using Agilent’s PNA-L and BAVA antenna.
The x-axis represents the location of a data point on the trace at the chosen
frequency on the NA. Initially the null (green line) is created with no object present
in the oil tank. Then the Eccostock rod is inserted and its response is recorded
(orange line). The rod is removed and the response of the null is recorded (green line
shifted). This process is repeated until four measurements of null and object are
recorded.
It also is clearly noticed that throughout the four measurements the null had degraded
from approximately being positioned at -90 dBm to -74 dBm. This meant that once the
null had been found it would quickly disappear due to drift in the system. Even though
the null is not stable and degrades greatly, this first pass at an interferometer yields a
29
device that is able to provide 90 dB of sensitivity while using the antenna that is currently
utilized in TSAR. This is an encouraging result as the sensitivity of TSAR measurements
is ~100 dB. This also shows that the interferometer is indeed able to cancel reflections
and can integrate to a measurement system with ease. This motivates further exploration
into the interferometer in order to understand its behaviour. This will help re-designing
and developing a new prototype, which will improve upon its potential benefits.
2.7 Summary
This chapter provided a brief overview of the need to perform network analysis on
unknown devices using S-Parameters. The limitations such as the ability to characterize
large impedances have been presented and several techniques were explored with
proposed specialized test setups to overcome these challenges. It was shown that
interferometry is an intriguing area of interest as it has been explored for multiple
applications. Although the method which this study is proposing of reflection
interferometry is currently being investigated, the direction of this study is completely
different. Previous studies attempt to perform absolute measures of objects, while here
the interest is in the differential information which has its own set of advantages and
improvements. The goal of this research is to develop from the ground up a MI to
compare in performance to current techniques used for breast imaging. The first
prototype supplied for this research proved to have promise, although it was unstable. In
order to do proof-of-concept testing, the system under investigation must be reasonably
stable, thus a two source nulling technique is explored in the next chapters, as it offers a
more stable and repeated measurement at the cost of the depth of the null (lower
30
sensitivity and dynamic range). This is followed by a single source implementation that is
the final goal of this study.
31
Chapter Three: Two Source Interferometry
A simplified block diagram of the two sources nulling Interferometery technique with the
use of a NA as a platform is shown in Figure 3.1. Two source nulling utilizes two ports of
a two source NA and a power combiner in order to perform the cancellation procedure.
Source 1 is chosen to drive an antenna or a sensor that is able to send and receive signals.
Source 2 is used as a matching source whose magnitude and phase are manipulated. The
reflected wave that is incident upon the antenna is coupled using a directional coupler
that is built into the NA and is fed to one of the input ports of the power combiner. The
signal coming out of source 2 is fed into the other input of the power combiner and the
output of the combiner is connected to a receiver. To obtain a null at particular
conditions, source 2 is varied in magnitude and phase in order to result in the combiner
being fed by two signals that have the same magnitude and are 180 degrees out of phase.
At these conditions, the output of the combiner will result in a complete cancellation and
the receiver will be driven ideally by noise.
Realization of this technique is done through the following steps. First the theory
of operation is discussed to gain a better understanding as to what actually happens to an
RF signal inside the interferometer device. Second, a test procedure is developed, which
in this case is done utilizing test automation software (VEE, Agilent Technologies, Santa
Clara) as it integrates well with the NA used as part of the system. Lastly the theory and
testing procedures are validated through a series of experiments targeted at gaining
insight to the limitations of the device.
32
Figure 3.1: Two sources nulling technique block diagram on a network analyzer
platform.
3.1 Theory
To describe the two-source interferometer in more detail, we examine the signals that the
system generates and the progression of these signals through its various stages. For
simplicity, we assume neither loss nor phase change through transmission lines and
coupler. The signals generated by the two sources propagate through the system as
follows:
33
Let S1 be the signal that is generated by source 1 and represented by:
(
)
(3-1)
(3-2)
Let S2 be the signal generated by source 2 and represented by:
(
)
(3-3)
(3-4)
Let Sref be the signal that undergone reflection and is fed to the combiner through the
coupler
(
)
(3-5)
Let Smatch be the modified signal that is obtained by varying the magnitude and phase of
source 2
(
) (
)
(3-6)
Sref and Smatch are added together using the power combiner and fed to a receiver
(
)
(
) (
)
(3-7)
An ideal null means the receiver is being fed with noise
(3-8)
This condition is met when |Sref| = |Smatch| and both are 180 degrees out of phase
(
(
)
)
(
(3-9)
)
(3-10)
34
3.2 Prototype
As the measurement system is built using a NA as a platform (N5230A 20 GHz, Agilent),
the sources used are internal to the device and are to be controlled using a program
written in VEE. A procedure in VEE was written that automates the calculation of the
magnitude that source 2 must be set to (
), which is necessary to force the RHS
and LHS of equation (3-9) to be equal. The phase is slightly more complicated and a
direct equation was not found. Attempts were made to find an algorithm which will
retrieve the phase to which source 2 must be set to (
) in equation (3-10) to obtain
a null. However, as this was done for proof-of-concept and will not be the final
implementation a simple robust procedure was developed instead, which although fairly
time consuming proved to work well. The following is a flow chart representation of the
procedure.
Figure 3.2: Phase retrieval procedure for two source nulling.
35
From the procedure in Figure 3.2 the magnitude and phase parameters are
obtained when there is no object in front of the antenna at the frequencies of interest.
These values are applied later when an actual measurement is taken. After a measurement
is taken the validity of the offsets can be verified by repeating the measurement with
initial conditions (magnitude and phase) and observing the nulls.
3.2.1 Test bed
Figure 3.3: Canola oil tank measurement setup with test objects.
Next, the system’s ability to detect objects is tested with rods of three types of materials
with varying relative permittivity and radius. The materials are as follows: Eccostock (εr
~ 15, R ~ 6.35mm), Delrin (εr ~ 3, R = 4.76mm) and Acrylic (εr ~ 2.6, R = 1.56mm).
Figure 3.3 shows the setup used for the measurements, involving a BAVA antenna
(length = 80mm) immersed in canola oil (εr ~ 2.5) and placed in a tank (R = ~165mm)
that is surrounded by absorbers to minimize the effect from the external environment
36
while measurements are taken. The insertion point (~101 mm away from antenna) is
located such that the rods inserted are aligned with the main beam of the antenna to
maximize detectability.
Two different measurement techniques were compared: the interferometer and a
standard data-minus-memory (DMM) technique described in the next section. A single
frequency of 6.5 GHz at 1 KHz IF with 401 data points and port power out of the NA of
0 dBm were chosen for both measurement techniques for consistency. This particular
frequency was chosen because the null was found to be deepest for the two source nulling
method at this frequency; depth of the null is approximately -60 to -75 dBm as shown in
Figure 3.4. The platform for this test was Agilent’s PNA-X series 26.5 GHz NA.
3.3 Interferometer Results
First, the ability to obtain a stable null was assessed. At first look this seems like a
simple task which would require varying source 2 to meet the above conditions for ideal
null. After performing many measurements it was found that the magnitude match was
fairly easy to obtain. On the other hand it was seen that a good phase offset is difficult to
achieve. This difficulty was attributed to the limited resolution of the phase control
parameter as well as phase noise. Each source has an individual phase lock loop, which
means that the sources will not necessarily lock to the same phase, resulting in a quantity
that cannot be subtracted. The phase noise levels for different operating frequencies can
be found in the manual of the NA. For example the phase noise level in this case is
approximately -95 dBc/Hz. Both of those parameters are attributed to the limit on the
depth of the null.
37
-50
Magnitude (dBm)
-55
-60
-65
Repeatability
-70
-75
-80
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Measurement number
Figure 3.4: Two source nulling repeatability with a null set to 6.5 GHz and a BAVA
antenna in a tank filled with Canola oil without a rod present using PNA-X NA as
platform. Measurements are done consecutively in 5 second intervals with null
conditions applied once at the start and no re-adjustments throughout the test.
Figure 3.4 clearly shows that the null does drift while taking multiple measurements over
time, during which the phase and magnitude offsets for source 2 remain constant.
Next, the ability to detect the three objects described above is investigated. The
same procedure as above is performed with the exception that this time the null is
achieved, its power level is recorded and one of the rods is inserted into the tank. The
magnitude by which the null changes, is recorded and compared to the value of the null
prior to the insertion. This is repeated multiple times for each rod to gain validation that
the change is consistent. For each experiment the magnitude and phase conditions for the
38
null are obtained once and used through without resetting. From Figure 3.5 we can
conclusively observe that this system was able to detect the Eccostock rod, confirming
that this technique although simplified does work. On the other hand, the two rods that
provide a much more challenging detection scenario are not reliably detected by this
method, as there is no significant difference in the null whether the rod is present or not.
This type of result is to be expected as this system is not able to bring the null close
enough to the noise floor where the reflection of the two rods is located. However, the
results of this experiment are satisfactory as now there is a baseline for the interferometry
technique.
Figure 3.5: Two source interferometer object detection validation results. The mean
of the null is ~-65 dBm with a standard deviation (σ) of 3. For Eccostock rod the
mean is -60 dBm with σ = 1.6, the Delrin rod mean is -64 dBm with σ = 5.5 and the
Acrylic rod mean is -61 with σ = 4.3.
39
3.4 Data-minus-memory measurement procedure
The DMM technique that is used for TSAR measurements involves calibrating
reflections from objects under test. First, measurements of an empty region of interest
are obtained and then the data from these reference measurements are placed into
memory. Next, the conditions are modified (e.g. test object added), resulting in new set
of data. To obtain the response to the change, the mathematical operation of data minus
memory subtraction is performed, which will yield only the change. This mathematical
operation involves vector math which is done post-processing by the NA. The following
is a simplified general representation of the involved operations.
Let Data be the measurements resulting from a single sweep of the NA that includes
responses from an antenna, the (empty) test system and environmental effects
(3-11)
The response recorded in data is stored into memory in the NA
(3-12)
Let Data Object correspond to the response recorded during another sweep of the NA
with an object present
(3-13)
Assuming, that nothing other than the insertion of an object had changed between data
object and data. As well as ignoring multiple reflections that occur from the presence of
the rod in the tank, which result in minor reflection that will not cancel as they are not
40
present in the memory data. The response of the object may be extracted by taking the
difference between Data Object and Memory, as all other terms will cancel
(3-14)
The current technique provides TSAR measurements with sufficient dynamic
range in order to detect the small signatures of the tumours of interest. The subtraction
operation brings the system close to its maximum dynamic range (~100 dB), which in
this case depends on the measurement setup and hardware. Throughout several
measurements, it was found that the dynamic range depended heavily on the quality of
the match the antenna provides. For this study, the same antenna was used throughout the
measurements and therefore extra attention was given to verifying the effects of the
antenna on both systems.
3.5 Performance Metrics
The comparison between the nulling technique and DMM is focused on looking at the
sensitivity of each system. Sensitivity is a broad term and can be viewed in two ways.
The first is the resolution between measurements. There is a finite difference between
signals that these systems can detect and if a signal happens to fall within this resolution
its signature will simply not be properly detected and could match the signature of
another object. The second view quantifies the sensitivity of these systems as the ability
to detect the smallest possible reflected signal. This study focuses primarily on the second
view, as the purpose of the interferometer is to offer the ability to detect miniscule power
changes in reflections relative to a reference point. For the interferometer, the depth of
41
the null has direct correlation to the lowest possible power that the technique can detect.
Since the null depth is limited by the noise floor of the measurement system, the lowest
possible power that the system can detect is directly related to its noise floor. The
comparison is performed by examining the lowest possible detected signal that each
system can detect from a reference point. In the interferometer case, the reference point is
the point at which the null is generated and for DMM, it is the point where a single trace
is taken and stored, then a second trace is taken under the same conditions, and both
traces are subtracted.
3.5.1 Comparison experiment 1
First, the ability of each of the measurement techniques to determine the signature of the
rod is explored. Results discussed in the previous sections are summarized in Table 3.1.
The two-source nulling technique was able to detect the Eccostock rod which is the
simplest case of the three, while the DMM technique was able to detect the Delrin rod as
well as the Eccostock. Both methods were not able to detect the very difficult case of the
Acrylic rod. This experiment has shown that the two-source nulling technique does work
but it has limitations due to its inability to detect the second case which the DMM
technique picked up.
42
Table 3.1: Result summary for comparison test of current and two source nulling
technique. The performance of both systems is defined by the delta between
reference and object measurement.
Method
Eccostock
Delrin
Acrylic
No object response
(Relative reference)
Nulling – PNA-X (6.5
GHz)
DMM – PNA-X (6.5
GHz)
-60 dBm
-
-
-65 dBm
-50 dB
-65 dB
-
-100 dB
Throughout the measurements it was found that the null obtained with the two source
nulling method did not result in an increase of dynamic range over the DMM technique.
This can be seen from the fact that the null level obtained resulted in only 65 dB of
dynamic range at optimal operation as opposed to 100 dB of dynamic range for the DMM
procedure at the same conditions. This means that the two source interferometry method
will not show improved capabilities unless the cancelation can be further improved and
result in an increase in dynamic range above 100 dB. This led to an investigation into the
operation of the two sources nulling technique in order to find out the reason for not
being able to obtain a better cancellation that is closer to the ideal noise floor of the
system.
The conclusion that the phase and magnitude drift contribute to the poor
performance of the null is validated by looking at previously conducted measurements.
Initially, a lower performance NA (PNA-L series) was used in order to try and obtain a
null. The best cancellation using this device was found to be approximately -50 dBm.
43
This is much worse than the -65 dBm that is achievable with the higher performance NA
(PNA-X series).
3.5.2 Comparison experiment 2
Microwave interferometry is said to be able to remove unwanted reflections. This
statement has been investigated by performing an antenna response elimination
experiment. Similar to the previous experiment, this test compared the response of an
object as seen by the DMM technique and interferometry. In this case, the interest is not
in the detection of an object, but rather in the comparison of the object response to the
antenna response.
The setup involved measurements over a frequency range from 2-6 GHz with 41
evenly spaced data points. The two source interferometer platform and DMM technique
used Agilent’s PNA-L model #N5230. A BAVA was used to send and receive a signal to
and from the object. Figure 3.6 shows the test setup for the MI. A signal is generated by
the NA and is transmitted by the antenna towards a test object, which in this case is a
copper plate positioned ~4.5 cm in front of the antenna. The copper plate was chosen
because there will be minimal absorption and most of the signal will be reflected back,
thus the system is assured to be recording a reflection from the test object and not picking
up signals from the environment that could mask the reflection if it was small. The signal
is then received by the antenna and fed into the coupler of the NA. The rest of the
processing of the reflected signal is done as described above by the operation of the
interferometer in section 3.1.
44
Figure 3.6: Antenna response elimination experiment MI test setup.
The case where data is collected with the interferometer is compared with DMM.
The DMM case involves taking a measurement with only the antenna present (response
of antenna with environment), as well as taking a measurement with the antenna and the
object present, then performing a mathematical subtraction of the data in order to isolate
the response of the object as explained in the previous sections. For both methods, the
data acquisition was done in the frequency domain and converted to time domain after
weighting with an appropriate pulse using the chirp-z transform [36]. Figure 3.7 shows
the frequency domain representation of the excitation signal applied prior to the
transform.
Figure 3.8 shows that the DMM technique yielded a result where the response of
the antenna (including minor reflections from the surrounding) is greater than the
response of the object. On the other hand, with the interferometry technique, the response
of the object is greater than the response of the antenna with minimal surrounding effects.
This happens because the response of the antenna is nulled during the setup procedure of
the measurement. This allows the system to cancel the unwanted reflections coming from
the interface of the antenna and air as well as reflections from the surroundings. Ideally if
45
the system is able to completely cancel all of the reflections, the null power will be at the
noise floor. When the object is inserted into the test system, the only recorded response is
from the object. With the interferometer technique, the response of the object and the
antenna appear at a later time (~8.5 ns) while with DMM they appear earlier (~2 ns). This
is due to the fact that both measurements have different reference planes. For DMM, the
signal has to propagate from the port to the end of the antenna and back to the port, while
with the nulling technique the signal propagates from the port and is recorded back at the
receiver which is a longer transmission length and thus the signal appears later in time.
Figure 3.7: Amplitude spectrum of excitation signal used to synthesize the desired
time-domain signal prior to the chirp-z transform.
46
Figure 3.8: DMM technique (left) and Interferometer technique (right) antenna
response elimination. Antenna only is a result of the antenna’s interface with free
space while object only is the interface between free space and the copper plate.
3.6 Summary
The work discussed in this chapter focused on demonstrating that microwave
interferometry can indeed be used to perform at least some of the operations that are done
with methods such as DMM. It was shown that, as far as dynamic range enhancement,
the MI lacks sensitivity that is currently available with standard procedures but it does
work on simple cases. On the other hand, interferometry was shown to be able to remove
unwanted reflections due to the nature of the measurement system because a null is
obtained by cancelling all unwanted reflections prior to taking measurements. The reason
for the poor dynamic range has been discussed and explained to be a result of limitation
with the two source nulling technique (drift). The interferometer as an idea definitely
shows promise if the null depth can be improved.
47
From the study of the two sources nulling technique, it was shown that the major
limiting factor is the coherence of the phase noise. This is also suggested by the fact that
the first generation of the MI has a much better null depth then the two source nulling.
This first generation interferometer has a single source, which means that the phase noise
and magnitude noise generated by the source affects both the reflected signal and the
matching signal. Because of this coherence, the cancellation of the two signals is much
better and results in a lower null. Although in its current state it greatly suffers from null
drift. Therefore, the next step in this work is to understand what is needed to build the
single source interferometer, which will improve on the issue faced with the first
generation implementation.
48
Chapter Four: Simulations of a single-source interferometer
The previous chapter clarified the feasibility of the interferometer and its limitations in
the current two sources build. There is a clear need to further study Microwave
Interferometry by pursuing the single source method, as this is expected to provide
further improvements and will result in a true test of the limits that this technique has to
offer.
Figure 4.1 shows the basic block diagram of the single source MI. As opposed to
the two source nulling method, as the name suggests, only a single source is used to both
drive the antenna and match the signal to obtain a cancellation. This is done by utilizing
two directional couplers. The first coupler couples a portion of the source that is then fed
to an attenuator which will be used for magnitude matching. The second coupler is used
as previously to couple the reflected wave that will be combined with the matched signal.
Aside from magnitude matching, there is also a requirement to perform a 180 degree
phase offset between the signals. In this design, phase matching is not explicitly
performed. The phase is determined based on the system parameters, which means that
the location in frequency of the null is dictated by the system itself. The control of phase
is currently achieved by using various lengths of cables between the output of the
interferometer and the input of the antenna. As this transmission length changes, the
phase delay changes as well, which results in a phase shift that translates to a null
frequency shift. Since the length of a cable is physically fixed it does pose a limitation
which requires constant length variation. This problem can be addressed in
49
Figure 4.1: Basic block diagram of a single source interferometer.
future work with the use of a phase shifter. In this study, the MI is used to enhance the
sensitivity of a measurement system and the focus is on the response at a single
frequency.
The first step to achieve an operational interferometer is to study and understand
its performance in a test set that allows for quick modifications and provides fairly
realistic performance assessment. Therefore, it was chosen to model the interferometer
with a simulation tool, namely ADS. The idea is to obtain insight into the performance of
the interferometer through simulations that include ideal components for initial testing
and then move to a more complicated design that includes components with realistic
behaviour. This in turn will help determine the critical components of the interferometer
that dictate the performance.
This chapter focuses on characterizing components such as the amplifier and
attenuator as well as others that are brought together to form the interferometer. These
components are analyzed in depth depending on their complexity and importance to the
performance of the interferometer. In the end, once a complete model is obtained it is
50
validated by monitoring the flow of a signal through the system and verifies that it is as
expected.
4.1 Simple simulation model
The simple block diagram of the MI as shown in Figure 4.1 is followed step by step from
input to output to generate the first pass of the interferometer ADS model as seen in
Figure 4.2. The basic building blocks of the interferometer include directional couplers,
an attenuator, phase shifter and a power combiner. The idea for this simplistic model is to
observe the signal at the receiver represented by terminal 2, in order to study the ability
of the interferometer to reduce this signal ideally to the noise floor (i.e. thermal noise
of
). Terminal 1 represents the input source to the system and R1
represents the load. The load is set to be purely resistive in this case to offer a magnitude
change in the reflected signal, or reactive to provide phase variation. For more complex
load models both resistive and reactive component configurations can be used. This
reflected signal is going to be cancelled by setting the attenuator and transmission line
acting as a phase shifter to their appropriate values.
Now that an ideal simulation model has been built, the next step is to set the
simulation parameters. The frequency is swept from 1.0 GHz to 5.0 GHz in steps of 10
MHz, load R1 is set to 20 Ohm and the transmission line length is set to an initial value.
Figure 4.3 shows the initial state of the simulation, where the transmission line is set to
have an electrical length of 200 degrees to provide several nulls in the simulation
frequency range and the attenuator is not providing any attenuation.
51
Figure 4.2: Simple ADS model of a single source MI.
S(21) (dBm)
0
-2
-4
-6
-8
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
Frequency (GHz)
Figure 4.3: Simple ADS model response without null optimization shows multiple
nulls with a null of interest at 1.8 GHz with amplitude of -7.244 dBm.
Several nulls can be observed to have a magnitude of -7.244 dBm. The next step
is to choose one of the null frequencies (in this case 1.8 GHz) and modify the attenuator
and transmission line in order to obtain a signal at the combiner that will better add
52
destructively with the coupled reflected wave and provide a deeper null. Figure 4.4 shows
that by changing the electrical length to 201.1176 degree and providing 8.2747 dB of
attenuation, a null at 1.79 GHz can be set to have a magnitude of -129.412 dBm. It can be
seen that with this simple model it is possible to manipulate design parameters in order to
obtain a null at a particular frequency. Also it is important to note that the null is very
sensitive to the length of the transmission line since an electrical length of 201.12 degrees
results in a null at -88 dBm, which is a degradation of 40 dB.
This test had verified that ADS can be used to study the interferometer. For this
exercise, the frequency at which the null occurred or the depth of the null were not
critical. This is a very idealistic model of the interferometer and does not offer the
insight that needs to be gained when realizing a null at a specific frequency or
investigating the depth of the null, although it did provide insight into the sensitivity to
the phase shifting properties of the transmission line. The rest of this chapter will discuss
the important steps that were taken in order to transform this simple model into a working
realistic design that in the future will be used as a reference design for a hardware
prototype.
4.2 Component modelling
The simple ADS model of the interferometer has shown that the null can be obtained
through the manipulation of design parameters. Next, realistic models of these
components are included into the simulation in order to observe more realistic behaviour
and performance of the device. Special care is taken in the design of the attenuator as
53
Figure 4.4: Simple ADS model response showing an optimized null at 1.79 GHz with
amplitude of -129.412 dBm.
the first interferometer prototype showed that this is a sensitive component which has
direct impact on the null. The attenuator is designed by modelling the diodes which are
biased to provide attenuation as well as the bond wires that connect them together. The
response of the attenuator is carefully analyzed at no bias and biased conditions to verify
that it will meet the requirements as it will have a direct effect on the magnitude of
reflections, which will be possible to null.
Also prior to re-designing the components of the interferometer, several
amplifiers were added at the output stage and a single amplifier at the input stage. The
input stage amplifier had been added to provide amplification of the input source power
to compensate for losses through the system before the signal is transmitted by the
antenna. The output stage amplifiers were added because the output of the interferometer
is ideally at noise level and it would be impossible to record a signal in the receiver if the
power of the null is too low for detection. Therefore, the output signal is amplified and
54
then fed into the receiver for recording. The significance of the amplifiers will be further
developed through this chapter. Figure 4.5 shows the integration of the new components
into the design model.
4.2.1 Input/output Amplifier – TC271
Several available differential amplifiers were considered (TC271, TC272 and TC931
provided by Agilent Technologies). The complexity of use, gain and Noise Figure (NF)
of each amplifier was examined in order to decide which will be used in the final design.
The TC271 was chosen due to its simple use as well as relatively high gain with a
reasonable NF as compared to the other choices. The TC271 only requires two voltages
to provide bias, VCC = +1V@34mA and VEE = -4.1V@135mA. Since these are
differential amplifiers, the gain of the amplifier and the NF depend on its connection at
the input and output. Four different possible connections can be made, as shown in Figure
4.6. The gain and NFs were obtained using an ADS model of the TC271. The ADS
model for this amplifier was supplied to the project by its designer and therefore it is very
comprehensive. The gain and NF results are validated by comparing the ADS results to
the data sheet for the part. The ADS behaviour is found to match closely with the
expected hardware performance. The gain and NF are shown in Figure 4.7 and Figure
4.8, and summarized in Table 4.1.
55
Figure 4.5: Updated interferometer block diagram.
Table 4.1: TC271 Behaviour summary.
Input Output
Connection
Gain
NF
Single Ended Single Ended
Single Ended Differential
Differential –
Single Ended
Differential Differential
~26dB
~28.5dB
~28.5dB
~32dB
~15dB
~15dB
~12dB
~12dB
The output of the interferometer is composed of three TC271 amplifiers. The first
is fed from the attenuator (used for a magnitude match), as well as from the directional
coupler which couples a portion of a reflected signal. The operation of this amplifier is to
produce the difference between the two signals that are fed into it. This difference is the
null that we are interested in. The next two TC271 are fed differentially in order to
amplify the null. The first amplifying TC271 is connected differential – differential
which provides ~32dB gain and the second TC271 is connected differential – single
ended which provides ~28.5dB gain. This three stage signal amplification is done in
order to be able to resolve the low power changes in the null by the ADC in the PNA.
56
Figure 4.6: TC271 Amplifier possible amplifier setups.
The profile of the receiver’s noise will dictate the lowest possible power that it is
able to detect. Since we are trying to generate an extremely low power level signal with
the null it is not possible to feed it directly to the receiver and hope that a reading other
than noise can be done without cooling the temperature of the receiver to reduce its
thermal noise. The PNA-X used as a platform for the interferometer in this study has a
typical noise floor of -129 dBm at the frequencies of interest. The amplification stage
overcomes this by supplying a null that is no longer at noise level but rather at a much
higher amplified state. This also helps distinguish changes around the null as they are
also now within the receivers detectable range.
57
Figure 4.7: TC271 Gain simulations with various setup scenarios. For SE-SE gain is
26 dB, DE-DE gain is 31 dB and DE-SE or SE-DE gain is 28.5 dB.
Figure 4.8: TC271 NF simulations with various setup scenarios. For SE-SE and SEDE NF is ~15 dB while DE-DE and DE-SE NF is ~12.5 dB.
58
4.2.2 Attenuator
To study the limitations of the interferometer design, a more realistic model of the
attenuator is required. This was needed because the resolution and stability of the
attenuator have direct impact on the depth and drift of the null. The attenuator was
initially modeled as an ideal component, which has a constant response across all
frequencies. This is not the case with real attenuators, as response typically varies across
frequencies. The interferometer incorporates a DC controlled variable attenuator in order
to be able to tune the null. Variable attenuators are built from active components such as
diodes and the attenuation is controlled by a current or voltage source. This introduces
further uncertainties into the design that are not covered with the ideal model. These
uncertainties must be taken into account in order to gain further insight into the
performance limiting factors of the interferometer. Figure 4.9 shows the desired topology
of the PIN diode attenuator as modeled in ADS.
4.2.2.1 Attenuator design
The implementation of this attenuator shown in Figure 4.9 will consist of two separate
controls for fine (V1) and coarse (V2) tuning which will allow tuning the attenuation with
greater precision in order to obtain a deeper null. This is achieved by connecting the
diodes to ground in two different configurations. In Figure 4.9 the coarse control diodes
are connected directly to ground at node 2, while the fine control diodes are connected
from node 1 to a 500 Ohm resistor. The purpose of this resistor is to desensitize the
resistance change at node 1 as biasing voltage varies. This point can be illustrated as
follows:
59
Let R1 and R2 be the series resistance of diodes from node 1 and 2 respectively
(4-1)
(4-2)
Let Rn1and Rn2 be the series resistance from node 1 and 2 respectively to ground
(4-3)
(4-4)
R1 and R2 vary with current. When the current is such that R1 and R2 are twice the initial
value the resulting node resistance is
(4-5)
(4-6)
Therefore the ratio of the change in resistance between the nodes is not equal
(4-7)
(4-8)
The calculations above clearly show that providing enough current to diodes in
node 1 to double their resistance value will not be sufficient to result in the same
resistance drop from the node to ground as in node 2. Since the attenuation of the
attenuator will depend on the biasing voltage, this implies that the same voltage change
60
across V1 and V2 will not result in the same loss through the device and in fact the V 1
contribution to the loss is smaller.
One of the requirements for the attenuator is to have minimal attenuation when it
is turned off. This will provide a wider range of reflected power values that will be able
to be nulled. When the reflected signal is of high power, the attenuator must be able to
provide the necessary minimal attenuation in order to match the modified signal
amplitude with the reflected signal at the input to the combiner. Therefore, the lowest
attenuation achieved by the attenuator will have direct effect on the maximum limit of
reflected power that can be nulled. On the other hand, when the reflected signal is of low
power, the attenuator must be able to provide large attenuation in order to match the
modified signal amplitude to the reflected. As such, the greatest attenuation that is
achieved by the attenuator will have direct effect on the lowest reflected power that can
be nulled. This consideration led to the inclusion of 6 diodes in the coarse node as it was
found from original design that 4 diodes did not result in sufficient attenuation.
A major contribution to the S21 of the attenuator is the match between the off
capacitance of the attenuator which is in shunt position and the series inductance that
connects the PIN diode together and to the input/output terminals of the device. In the
hardware implementation of the attenuator, the PIN diodes will not be used as a packaged
device. Rather they will be added as chips and connected using bond wires as shown in
Figure 4.9. These bond wires behave as inductors and are used to connect the device
together in the RF path. Therefore the bond wires must be designed to provide the proper
inductance values as to obtain the match for minimal S 21.
61
4.2.2.2 Bond wire
Two sets of bond wires are going to be used, the first connects the PIN diodes to the DC
blocking capacitors and will have a value of 0.05 nH and the second design is for the
bond wire that connects diodes together and will be of value 0.1 nH. These particular
values are chosen because PIN diodes with approximately 30 fF off state capacitance will
be used and this combination will result in a small reflection of the RF under the off state
of the device. In ADS, there is a feature that allows the design of bond wires by providing
parameters for diameter of the wire and the length of six segments. In general 1 mm of
bond wire has inductance of 1 nH meaning that the above bond wires are impractical.
Although for this first pass at an ADS design, there was no consideration of the
manufacturing limitations of such bond wires. Therefore, the radius and the length of the
six segments were chosen to provide the needed inductive values without any further
considerations.
Two shapes are built and the difference between them that provides the change in
inductance is the different length of wire segments. Simulations showed acceptable and
constant inductance values over 20 GHz. Specifically, shape 1 is of value 0.105 nH and
shape 2 is 0.063 nH.
62
Figure 4.9: PIN diode attenuator with dual DC control modeled in ADS.
4.2.2.3 PIN diode model
In order to meet the low off capacitance value of ~30 fF, the diode HPND – 4005
(AVAGO Technologies, San Jose) was chosen and modeled in ADS using an APLAC
model [37] as shown in Figure 4.10. The advantage of this model is that it shows good
agreement with the known “on” resistance in the forward bias region. The disadvantage is
that it does not properly predict the behavior of the diode in the reverse bias region. In
attenuators, PIN diodes are used as resistive components with variable resistance in order
to attenuate RF signals. Since this type of operation occurs in the forward bias region of
the diode, the APLAC model is a good candidate for simulations. Utilizing the data sheet
provided by AVAGO for this particular part, the APLAC model is populated to obtain a
simulated response to match actual measurements. After several trials varying various
model parameters, good agreement was found between the simulated data and data
sheets/measurements. The results of this are provided in appendix A.
63
Figure 4.10: PIN diode ADS APLAC model. The main parameters of interest in this
model are Rmin and Rmax representing the minimum and maximum RF resistance as
well as the SDD1P, which is a two port model for a current controlled variable
resistor. Inductor L and capacitor C represent the connection inductance and diode
capacitance respectively.
4.2.2.4 Attenuator response
Figure 4.11 shows the insertion loss of the designed PIN diode attenuator with all of its
final parameters and non-ideal components. This means that when the PIN diode
attenuator is turned off (V1 and V2 are 0 V), the loss through the device ranges from -1.2
dB to -3.2 dB from 1 GHz to 18 GHz. This is the lowest attenuation that can be attained
by the device and is well within the desired range. Similar analysis is presented in Figure
4.12 showing the maximum attenuation, which is found by increasing both bias voltages
until a negligible change in response is observed. The voltage setting to reach maximum
attenuation was found to be V1 = 20 V and V2 = 20 V. These two plots show that the best
region of operation is between 4 GHz and 6 GHz where minimum and maximum
attenuation are at their greatest.
64
Figure 4.11: PIN diode attenuator insertion loss V1 = 0 V, V2 = 0 V.
Figure 4.12: PIN diode attenuator response V1 = 20 V, V2 = 20 V.
This is sufficient as the operation of the interferometer is expected to be around 5 GHz.
As previously noted minimum attenuation will limit the ability to null large reflections
coming from the load and maximum attenuation will limit the ability to null small
reflections.
65
4.3 Upgraded model
With the key components of the interferometer properly modeled to exhibit a more
realistic behaviour than in the initial simulation, it is now possible to combine them into a
complete simulation of the interferometer shown in Figure 4.13 to allow a comprehensive
study of its behaviour. This involves adding the input and output amplifiers, attenuator
and reversal of the directional coupler connected to the transmission line. The analysis is
done by monitoring the propagation of a signal and noise through the system in order to
validate that the topology designed in ADS works as expected.
Several decisions were made throughout the design process. The power splitter is
replaced with the TC271, which has a differential output. The positive output of the
amplifier will be used as the input to the sensor through a directional coupler and the
negative output will be fed to the attenuator.
In order to maintain the integrity of the signal that the sensor will be radiating,
directional couplers are typically connected such that the input signal passes along the
thru path and the reflected signal is coupled. The disadvantage of this configuration is
that the small reflection that is of interest gets attenuated by approximately 10 dB which
is quite significant and can become too small to detect. To eliminate this, the directional
coupler is flipped and connected with the incoming signal passing through the coupled
section and the reflection fed to the thru. The magnitude of the incoming signal is not
compromised by the introduced loss, because the TC271 at the input amplifies the signal
to overcome this loss.
(a) & (b)
(c)
(g)
(d)
(e)
(f)
Figure 4.13: Nulling interferometer complete design. Signal flow through interferometer at various stages is shown. Stages are
labelled according to their respective purpose a) System input b) Input confirmation c) Excess noise verification d) Null
formation e) Null amplification f) Second amplification g) Third amplification and output.
66
67
Multiple simulations at varying input power levels had shown that the output
stage amplifiers could saturate due to harmonics. The cause of these harmonics had been
found to be the input amplifier. If the input signal driving a TC271 is greater than -25
dBm, it will cause the amplifier to saturate and the output signal will begin to look like a
square wave (multiple frequencies) instead of a sinusoid (single frequency). In order to
eliminate this saturation, two band pass filters were introduced at the outputs of the input
stage amplifier to filter out the unwanted harmonic components generated by the
amplifier. The filters are designed to pass signals in a BW of 0.5 GHz around the null
frequency, allowing only the fundamental frequency to propagate. The reason for driving
the input amplifier into saturation is to compress noise that could be contaminating the
input sinusoid.
Since the TC271 is a differential amplifier, it is beneficial to take complete
advantage of this and connect the amplifiers in differential mode which provides an extra
6 dB of amplification and a lower NF, as well as eliminating the need to add a DC offset
or a blocking capacitor at the output of each amplifier in order to remove the generated
DC bias, which can potentially saturate the next stage.
The last upgraded piece is the new two source bias PIN diode attenuator. With
this, there is no longer a need to use an ideal component which has a flat response across
all frequencies. Instead it is now possible to manipulate the attenuation of the matched
signal through the application of two DC biases. The fine and coarse control will be used
to tune the null to its maximum possible depth.
Now that the model of the interferometer has been upgraded with more realistic
components, the next step is analysis of the signal propagation through the system as well
68
as the Signal to Noise Ratio (SNR). These two quantities will be used to validate the
performance and operation of the MI.
4.3.1 Signal propagation
Analysing the signal propagation through the system is performed in order to verify that
the signal flow is correct and the components behave as expected. As seen in Table 4.2,
several nodes were labelled for voltage recordings and current probes were added for
current readings. Using Circuit Envelope [38] analysis available in ADS, it is possible to
take those voltage and current values and convert them to power readings in dBm. Circuit
Envelope allows the use of an RF input signal while the output of the simulation is
provided in the time domain. The input signal is sampled at periodic intervals for
amplitude and phase at the carrier frequency. This is beneficial since we are interested in
calculating and simulating noise, which is dependent on BW. Knowing the required noise
floor, the sampling rate can be calculated and set accordingly.
The sampled values of the carrier at each time step are then fed through the
circuit. This yields a series of spectral results at each time step containing the carrier
frequency and its harmonics. The final time domain response is obtained by taking
successive magnitude and phase values at a chosen frequency (for example, the carrier
frequency at each time step).
69
Table 4.2: Signal propagation through the interferometer.
This procedure was performed at all nodes of interest which includes the input
and output of the system as well as the input and output stages of the amplifiers in order
to analyze the response of the system and its components and is summarizes as follows:
A particular point is chosen where the input is at -60 dBm in order to look at the flow of
the signal. This point is chosen because this is going to be studied as the future operating
point of the interferometer. Table 4.3 shows the expected and simulated values of the
power in dBm at various stages A-G for an input of -60 dBm. Each of the stages is
described below:
•
Stage A is the input to the system that is set by a parameter in the simulation
setup.
•
Stage B is a verification that the input sent to the system is indeed working
properly.
70
•
Stage C is the output of an amplifier that is used to add excess noise to the system
to represent a real source. This amplifier only affects the noise and not the signal.
•
Stage D is where the null is generated by subtracting the reflected signal and the
matched signal. The null value is dependent on how well the magnitude and phase
are matched.
•
Stage E is the amplification stage of ~32 dB to signal and ~ 12 dB to noise.
•
Stage F is the amplification stage of ~32 dB.
•
Stage G is a result of an amplification of ~28 dB and it is the output of the system.
Table 4.3: Signal flow through interferometer at -60 dBm input.
Stage
A
B
C
D
E
F
G
(dBm)
(dBm)
(dBm)
(dBm)
(dBm)
(dBm)
(dBm)
Expected
-60
-60
-60
-144
-100
-68
-40
Simulated
-60
-60
-60
-142
-102
-70
-41
From Table 4.3 it can be seen that the signal does indeed propagate as expected
through the topology that has been developed in ADS. Slight discrepancies can be
ignored as the simulation values of noise can vary by approximately 1 dB between runs.
This variation of noise depends on the length of time of the simulation and longer
simulation time will result in better noise approximations between runs as more noise is
averaged together. This is evident at D, which indicates the signal propagating with a
noise floor of -142 dBm.
71
4.3.2 Signal- to-Noise ratio
As the signal propagates through the system so does the noise. In order to be able to view
the output and make sure that what is recorded is an actual signal and not noise, the SNR
has to be positive throughout the system. The SNR is defined as the power of the signal
divided by the power of the noise and is often used as a measure of the sensitivity of a
system. It was shown that the signal does indeed propagate through the system properly
and the next step is to show that the noise propagates correctly as well. This needs to be
verified because the noise will be affected by certain components differently than the
signal. For example, the excess noise of 35 dB is added to represent a more realistic
source. Table 4.4 shows the flow of the noise through the interferometer setup. Since a 1
KHz BW is used for simulations, the thermal noise and thus the minimum noise level is 144 dBm. The node labels are used similarly to the ones in Figure 4.13.
.
•
Stage B is the lowest noise level of the system which is as expected to be
approximately -144 dBm.
•
Stage C: excess noise of 35 dB is added at the output of the source to represent a
real source.
•
Stage D: due to the subtraction of two signals at the null stage, the noise of the
input to the amplifier is back to its minimum level of -144 dBm.
•
Stage E: the noise is amplified by ~32 dB and also increased by the NF of the
amplifier ~12 dB
72
•
Stage F: the noise signal is amplified by ~32 dB since the null at noise level is
now amplified sufficiently that it is no longer affected by the NF of the amplifier.
The differential connection also allows for removal of common mode noise.
•
Stage G: the noise signal is amplified by ~28 dB since the output connection is
single – ended.
Table 4.4: Noise flow through the interferometer stages.
Stage
B
C
D
E
F
G
Noise Power
Expected
(dBm)
Noise Power
Simulated
(dBm)
-144
-109
-144
-100
-68
-40
-144
-109
-144
-103
-71
-42
Table 4.4 shows good agreement between expected and simulated results. As
previously discussed, discrepancies are small and can be ignored. With these two
parameters of signal and noise the SNR at each stage of the interferometer is calculated
for a tuned null state with a load of 30 Ohm as well as the SNR when the load changes
slightly to 30.1 Ohm and the results are summarized in Table 4.5. The SNR at the null
and perturbed state is positive throughout the system, which means that a signal does
indeed propagate when the load varies and it is not due to simply noise. There is some
degradation in SNR at the null amplification stage, due to the low power input signal at
the cancellation stage which is affected by the NF of the amplifier and not only by the
gain.
73
Table 4.5: SNR calculated for the various stages from noise and signal data.
Stage
B
C
D
E
F
G
SNR Null State
(dB)
SNR Null
Perturbed State
(dB)
84
49
2
1
1
1
84
49
36
27
27
27
Subsequent amplifiers will not contribute further degradation due to the sufficiently large
gain of the nulling amplifier, as can be shown through the Friis equation [24].
4.4 Summary
The above figures and tables have shown that the current topology (section 4.3)
developed in ADS for the interferometer is a valuable model for understanding the
operation of the interferometer, since the values obtained are showing a good agreement
to expected values. The signal and noise both properly propagate through the system. A
null of -144 dBm has been recorded which shows a great improvement over a dual source
operation. This was previously expected and now it has been verified. This happens
because, in a single source, the phase noise is correlated which is not the case with dual
source. Another interesting item to note is the cancellation of the excess noise because
the noise is correlated in the system at the null stage. This means that, at the null stage,
the system not only cancels the reflected signal with a matched signal but also reduces
the noise floor of the system back to its lowest possible value (governed by K*T*B) even
though excess noise was introduced to the system due to imperfections in real
components. The significance of this simulation is to provide a reference for the hardware
assembly of the interferometer which will be discussed in chapter 5.
74
Chapter Five: Hardware Implementation
One goal of this research is to develop a working prototype of a MI that is able to create a
null utilizing a single source. Previous chapters established the basics that are needed to
design and construct the interferometer. Discussions included the theory of operation as
well as a simulation in order to gain insight into the difficulties and sensitive areas that
would be encountered during hardware implementation. This chapter will focus on the
procedures taken throughout the breadboard stages and will provide a validation of
operation by comparing the measured results to previously obtained simulations.
First, a PIN diode attenuator was built and assessed to ensure it behaves according
to the design found through ADS simulations. This is an important step since the
attenuator must be able to achieve a required attenuation range as discussed in the
previous chapter. The coarse and fine controls must operate properly and provide
sufficient resolution to tune a null. Therefore, the attenuator is regarded as a major
contributor to the sensitivity of the interferometer and this is reflected by the amount of
extensive work conducted in Chapter 4 to simulate and understand the behaviour, as well
as in the following sections of this chapter to attain better sensitivity. Several build
iterations had to take place before the attenuator was fully debugged and the whole
system could be assembled. While creating the simulation model of the interferometer,
several assumptions and design decisions had to be made. For example, the output stage
was implemented by using the differential behaviour of the amplifiers. This led to several
debugging stages of the system and in the end resulted in a slightly modified topology
than the one used in simulations.
75
5.1 PIN diode attenuator
Two attenuators were constructed following the simulation topology. The first was built
using MPN-7300 and the second using MADP-000165-01340W chip PIN diodes. The
reason that these diodes were chosen over the originally simulated HPND-4005 diode is
due to their availability in chip form, as it was later found that the original diode would
not be available in this form. These diodes exhibited the required low capacitance at the
off state to provide low insertion loss.
The first iteration of the PIN diode attenuator was built using sample diodes
obtained from AEROFLEX (Plainview, New York) which were the MPN-7300. These
diodes come in chip form as required by design and have a capacitance rating that
matches the required 30fF. Ideally, the diodes will have 30fF capacitance at off state, and
the diodes exhibited behaviour very close to this target. A test sheet was provided with
the diodes which specified the capacitance of the diodes at 0 V bias. The values measured
ranged from 30fF to 100fF with most diodes having a value in the middle of this range.
An attempt was made to try to distinguish which diode had a particular response, but
since there were no clear labels on the package it was not possible to do so. Therefore 8
diodes were chosen out of the 20 which were provided in the sample. Six diodes were
used for the coarse control and two were used for the fine control, as shown in Figure 5.1.
Test measurements were conducted around the future operating frequency of 5
GHz. Agilent’s PNA-X was used to perform a two port calibration in order to measure
the S21 response of the attenuator.
One of the design expectations was that the attenuation at 0 V bias on both control lines
will not exceed 10 dB from 0 to 20 GHz. Lower insertion loss would provide greater
76
range of operation, but 10 dB was decided to be sufficient as a realistic upper limit for the
interferometer performance as the major concern is with small reflections, where depth of
the null will depend on the maximum attenuation that the attenuator can provide. Figure
5.2 shows that this was achieved for 500 MHz to approximately 18 GHz. While the input
and output ports are connected to the NA the bias voltages were varied. Initially V1 (fine
bias) and V2 (coarse bias) are set to 0 V and the insertion loss is found to be 3.1 dB. Then
V2 is set to 1.7 V and the attenuation is at 18.7 dB. Keeping V 2 at 1.7 V V1 is set to 1.7 V
and the attenuation is now at 22.7 dB. This shows that, as expected, the coarse control has
a greater effect on the attenuation. It was also found that, as the voltage increases on both
bias lines, the maximum stable attenuation achieved is approximately 30 dB. Insertion
loss (~ -1.5 dB) and maximum attenuation (~ -23dB) were simulated in chapter 4. It is
seen that around 5 GHz the hardware implementation of the PIN diode attenuator is
comparable or better, where the IL (~ -1 dB) and maximum attenuation (~ -30 dB). Some
of the above results are summarized in Table 5.1.
77
Coarse control
Thin film resistor
Fine control
Figure 5.1: PIN diode attenuator using AEROFLEX diodes.
Figure 5.2: MPN-7300 attenuator response at 0V bias with 560 ohm external on bias
lines.
78
Table 5.1: PIN diode attenuator performance test measurements with MPN-7300.
Fine Control (V1)
Fine Control (V2)
Freq (GHz)
Attenuation (dB)
0
0
5
3.1
0
1.7
5
18.7
1.7
1.7
5
22.7
The second iteration of the PIN diode attenuator involved a second set of diodes
provided by a different supplier. The reason for this was to evaluate the effect of different
diodes on performance. The data sheet provided with the diodes was reviewed and the
capacitance was found to be within acceptable values mentioned above. Since most of the
problems were identified through the design of the first attenuator, the second attenuator
was designed with no faults from the first attempt. The only change made to the design
was the value of the thin film resistors from 100 ohm to 500 ohm. This allows us to avoid
using additional external resistors to obtain the 500 ohm resistance that was used in
simulations on the DC bias lines. Similar to the first attenuator, the S21 response shown in
Figure 5.3 was characterized using a NA to confirm the proper behaviour of the
attenuator. Similar tests to the ones made with the first attenuator were conducted and it
was found that the operation of this attenuator is fairly similar to the first attenuator.
However, the insertion loss from 1 GHz to 18 GHz is better suggesting a better match,
and is well within the 10 dB requirement.
79
Figure 5.3: MADP-000165-01340W attenuator response at 0V bias.
5.1.1 DC supply sensitivity
When a null is obtained, it is used to take a measurement, thus the null must remain very
stable and not drift from its set value. In simulations it was found that the major
contribution to variation in the system comes from the attenuator. This is due to the fact
that the attenuator is used to modify the power level of a signal to match the power level
of the reflected signal from the object. The better the power level match, the greater the
cancellation and a deeper null can be obtained. The attenuation is a function of the
resistance value of the PIN diodes which is set by the amount of current flowing through
them. The diodes are biased using a DC power supply by varying the voltage until a low
null can be seen on the NA. During the PIN diode attenuator tests, which involved
varying the bias voltage on both bias lines, it was noticed that the S 21 measurement at a
single frequency is not constant and varies slightly. The variation is quite small but in
80
order to obtain a deep null this minimal variation is enough to shift the null significantly,
since the null is at a very low power level. The reason the S21 measurement varies as it
does is due to the imperfections of the DC power supply. The value that the supply is set
to is not perfectly constant and has some noise. The DC power supply used for initial
testing (Agilent E3631A) has a resolution of 1 mV and can be seen to vary by up to 1
mV. Previous work done trying to approximate this sensitivity in simulations showed that
a 1 mV change in DC bias could shift the null by as much as ~2 dB. Therefore the DC
supply has to be as clean as possible and provide a better resolution than 1 mV step.
Several ideas were discussed and implemented to minimize the noise in the DC source
and desensitize the voltage step to provide improved resolution.
5.1.1.1 Low pass filter
The first attempt to improve the sensitivity of the DC power supply was to try to
eliminate some of the noise which is shifting the null. This was investigated by directly
connecting the DC power supply to an oscilloscope and looking at the signal that the
power supply is outputting. It was noticed that the signal contained higher frequency
components that are unwanted and are a source of variation. In order to eliminate these
effects, a Low Pass Filter (LPF) was implemented at both bias connections of the
attenuator. The idea is that the LPF will only let the desired DC signal into the attenuator
and will eliminate the higher frequency components. The LPF was made from discrete
capacitors and inductors which were soldered together as close as possible to the
attenuator bias and ground. The topology of the LPF used can be seen in Figure 5.4. On
the oscilloscope it was noticed that this LPF indeed cleaned the signal but did not
81
completely eliminate all the unwanted variations. A trial was conducted to increase the
order of the filter by cascading two LPF section. The effect of this was not as drastic and
again did not completely eliminate the unwanted higher frequency components.
Therefore the topology seen in Figure 5.4 was used, since it provides an improvement
over using the output of the DC power supply directly.
Figure 5.4: Low pass filter implemented at attenuator to improve sensitivity.
5.1.1.2 Voltage desensitizing (resolution enhancement)
The second attempt to improve the bias signal that is fed to the attenuator was in the form
of desensitizing the voltage source. The resolution of the power supply was not
sufficiently small which meant that it was very difficult to obtain a null because it was
not possible to choose many attenuation levels between ~5 to ~30 dB.
In order to gain some extra sensitivity, the voltage source was first connected to a
series of resistors in order to reduce the voltage drop across the bias line. This meant that
a 1 mV change on the power supply would correspond to a smaller than 1 mV change on
the bias line. This provided higher resolution for adjustment in order to find a deeper null,
but it did not improve the fact that the signal constantly varied slightly.
82
5.1.1.3 Current source implementation
The resistance of a PIN diode is dependent on the diode’s forward bias characteristic of
series resistance vs. bias current. By varying the bias current across the diode it is
possible to choose the appropriate resistance value in order to obtain the required
attenuation level in the attenuator circuit. This can be done by either using a voltage
source or a current source. A voltage source allows the ability to choose a particular
voltage level which will ideally remain constant while the current is varied depending on
the loading, while a current source allows setting the current to ideally a constant value
and the voltage will vary with the loading to maintain the proper current setting. As
previously mentioned, the first tests of the PIN diode attenuator were conducted with a
voltage source and it was noticed that the response has some drift. Some of this variation
can be attributed to the fact that, for a voltage source, the current is allowed to vary
slightly as function of loading. This means that it would be beneficial to operate the
attenuator using a current source, which should keep the current value from varying thus
keeping the resistance values of the PIN diodes more stable. This also directly translates
to a more stable S21 response of the attenuator. The most basic current source was
constructed using a voltage source and a very high resistance in series with the bias line.
This is a similar procedure to desensitizing the voltage which was mentioned above, with
the exception that the resistance for this method is significantly higher than initially used
for only desensitizing. Therefore these two techniques were combined together into a
single implementation. The improvement provided by this method is noticeable;
originally the null would drift after several minutes, while after applying this method it
83
was possible to keep the null relatively stable for about 30 minutes. This is important for
accurate repeatable measurements.
5.2 Microwave interferometer hardware build
The nulling interferometer complete design schematic developed in chapter 4 was
extensively studied and simulated in order to understand the operation of the MI.
Therefore this diagram was used as a reference throughout initial breadboarding. The first
iteration was built to be identical to the schematic with minor changes, some of which
improve operation and others eliminate unnecessary components.
5.2.1 Hardware deviations from simulations
BPFs used in simulations were omitted as they were believed to have no consequences in
hardware on the harmonic saturation of the TC271 amplifiers because of filtering present
inside the component. During the build a decision was made to connect the directional
coupler in a different manner than in simulations. Since the reflected signal is expected
to be minute, it is unwanted for it to lose more power by being coupled. Therefore the
directional coupler is connected to have the coupling loss present in the incident path and
the reflected signal will propagate along the through path and will encounter less loss by
~12.5 dB. The added loss to the incident path is compensated by using a differential
amplifier with greater gain at the input. For this, the TC931 was chosen as it is a
differential amplifier like the TC271, but with 40 dB of gain. The extra ~12 dB gain at
the input will cover the ~12 dB added loss through the coupler.
84
No specific requirements were considered for the connections between the various
components of the interferometer and connections to the input and output of the NA for
this first pass. Flexible cables were used due to their availability. If these cables are
found to be an issue, an investigation would be conducted to decide on needed cable
performance.
5.2.2 Design corrections
When all of the components were brought together and following the topology develop in
simulation, the system was not producing an output and only noise was being recorded by
the NA. To debug this matter, the system was broken down to simplified stages and
slowly assembled again to locate problems. It was found that DC blocks and filtering
were required in order to avoid the saturation of the output amplifiers. It was later found
that the filter was needed due to the use of the high gain TC931 amplifier which was
producing higher harmonics than expected. Therefore, the TC931 was replaced by a
TC271 in the next build, although the filter remains unless found unnecessary in the
future. The input signal loss incurred by the coupler was monitored to verify that it has no
significant effect on performance. DC blocks were needed to avoid a DC offset occurring
between the inputs to the amplifier and causing it to saturate. These components had to be
located at each input line of the amplifier. Because two DC blocks or two BPF do not
have exactly the same response, a phase offset will result and the signal to the next stage
contains periodic unwanted cancellations. To avoid this problem, the amplifiers are
connected in single ended instead of differential mode which results in ~ 10 dB of gain
85
loss through the amplification stage. Figure 5.5 shows the resulting block diagram of the
updated interferometer topology with the new modifications.
Figure 5.5: Modified interferometer block diagram.
5.2.3 Final output
Once the necessary changes were implemented, the MI produced an output as shown in
Figure 5.6. This output is generated by modifying several connections on Agilent’s 26.5
GHz PNA-X model N5242A to integrate with the interferometer. The front loops at port
1 are disconnected to allow the connections in Figure 5.7. After the interferometer is
connected to the NA, the sensor of choice is connected to TEST PORT 1 with a cable
acting as a transmission line. For the initial build, a coaxial cable was used as both a
transmission line and sensor, by utilizing the mismatch at the end of the cable to air in
order to obtain a reflection.
86
Null Resulting From
DC Biasing
Potential Nulls
Figure 5.6: Output of MI with dc biasing working properly.
Figure 5.7: Interferometer connections to a network analyzer.
87
For this example the IF frequency was set to 1.0 KHz, and the number of data
points to 401. Lowering the IF frequency would be preferential as it will reduce the noise
and produce a deeper null, although for illustrating the operation of the interferometer
this was not crucial. The output of the interferometer is seen to be band limited, which
depends on the BW of the BPF. For this implementation, a filter is chosen from 4 GHz to
5.3 GHz, due to its availability. Within this frequency range, multiple nulls occur at
periodic intervals. This periodicity is directly related to the length of the physical and
electrical attributes of the transmission line. These are referred to as potential nulls and
any of these can be chosen for the tuning process. The null to tune is chosen by its
frequency location and the particular needs of an application. For this proof-of-concept, a
null at 5.22 GHz is arbitrarily chosen and tuned. First, the coarse voltage is increased
while the null is monitored. When the null ceases to reduce in magnitude, this process
ends and the voltage is no longer varied. The next step is to begin increasing the voltage
on the fine bias control to further attempt to reduce the null, similar to the process
followed for the coarse voltage. Some trial and error adjustment is required to finalize the
voltage values resulting in the best null depth. By conducting this technique, a deeper null
begins to form at 5.22 GHz in Figure 5.6. This shows that indeed the system is
performing as expected. Figure 5.8 shows the final topology of the hardware
implementation that provided the above results.
88
5.3 Comparison to simulation build
Throughout the breadboarding stages of the interferometer, the ADS simulation
schematic was used as a reference design guide. Several modifications such as adding
components and re-routing the source signal had to be made, thus the simulation topology
does not reflect the hardware realization. Since simulations are a powerful tool when
needing to make modifications to circuitry or performing analysis, it is beneficial to have
the ADS simulation agree with the response that is obtained in the hardware
implementation. Figure 5.9 shows the updated ADS schematic that is built to match the
final outcome of the modifications made in hardware. The real DC block and BPF used in
the hardware build are characterized using a NA and their response (S- parameter matrix)
is brought into the simulation to keep the simulation as close to reality as possible.
Amplifier
s
PIN Diode
Attenuator
LPF
BPF
DC Block
Figure 5.8: Microwave interferometer modified topology connection.
89
Figure 5.9: ADS schematic matching hardware implementation.
Even though hardware utilized the TC931 as the input amplifier, the updated ADS
simulation includes a TC271 as future experiments of the interferometer will use the
TC271. It is still possible to do proper comparison between the two designs as the
simulation allows manipulating the conditions to match hardware with both TC271 and
TC931. As long as in both cases the null for the output stages is of the same level, the
results are the same.
Table 5.2 shows the signal propagation through the interferometer in the updated
ADS simulation. The operating point of interest is at – 60 dBm input power, as this will
be the point of comparison to hardware performance. This input power level is chosen,
since at this level it was reasonably simple to locate a null in hardware, as expected. It is
possible to use other power levels, but throughout the initial experiments with the
interferometer this level was often used and therefore was preferred for this comparison.
Columns (a), (d) and (g) show that for a -60 dBm input, it is possible to tune the
interferometer to provide a null of -144 dBm (using 1 KHz BW signal), the output of the
system will be approximately -60 dBm. This means that in order for the hardware to
90
match the performance of the simulation, the output of the system must be -60 dBm when
the input is -60 dBm and the system is at a null state. Figure 5.10 shows that this is
indeed the case, and that simulations predict measured performance. Since the null is
obtained over a narrow band and it is more difficult to tune it in hardware, a null
frequency was first found in hardware at ~ 5.0062 GHz and then fed to the simulation as
a parameter.
5.4 Understanding the null
As previously mentioned, a simulation model is a very powerful tool that can aid in the
understanding of a system. In this case, the simulation model allows us to monitor the
propagating signals in locations that are otherwise inaccessible and very difficult to
observe. For this design, this occurs at the stage where the null is obtained, which is the
differential input to the first output stage amplifier. The difficulty in observing this signal
is that, ideally at this location, the null is at noise floor of the system and its power equals
to thermal noise. Since it is difficult to truly obtain a perfect cancelation, the signal might
have slightly higher power. It is difficult to accurately record power levels that are this
low since this would require a very precise detector with high resolution. For example,
the detector in the NA cannot resolve the null and any small changes of power around it.
Since the interest is to detect small changes, this limitation will not suffice and thus the
null is amplified through the output stage to help the detector resolve the signal. Thus the
signal that the interferometer provides at the output is the amplified null and not the null
itself.
91
Table 5.2: Interferometer signal propagation using 1 KHz BW signal simulation.
Figure 5.10: Interferometer hardware implementation output null response.
92
This means that when the attenuator is being tuned to obtain a null, it is quite
difficult to know how low the output power has to become to reflect a null of thermal
noise value at the first stage of the output of the interferometer. Therefore it is important
to validate that the power seen at the output of the interferometer can be translated back
through the amplifier to the first stage and indeed be of noise levels.
In order to translate the power level of the null at the output of the interferometer
to the pre-amplified null, the topology in Figure 5.11 was used in both simulation and
hardware. In this case, the output stage of the interferometer consisting of three cascaded
amplifiers is connected together to depict its operation in the interferometer. The input to
the system is terminated by 50 Ohm matched loads in order to supply the network with an
input of thermal noise level. This mimics the condition that the cancellation is well made
and the null is at the thermal noise level. The interconnection of the amplifiers is kept the
same as in the system and the output of the system is fed to a receiver on the NA. The test
is performed at 1 KHz signal BW as well as 10 Hz.
Figure 5.12 shows that if the input to the output stage of the interferometer is
thermal noise (which is generated by the 50 Ohm matched loads), the output seen by the
system is approximately -50 dBm at 1 KHz BW and -70 dBm at 10 Hz. These are slightly
rough numbers since the output was viewed using a NA. When the output is viewed on a
spectrum analyzer with proper averaging, more realistic values are measured, indicating
outputs closer to approximately -55 dBm and -75 dBm, respectively. The difference
between these numbers makes sense since a reduction of a factor of 100 in BW results in
an improvement of 20 dB in thermal noise. The thermal noise for 1 kHz signal is -144
dBm and for 10 Hz is -164 dBm which is 20 dB lower as expected. This is supported by
93
the input and output power levels shown in Table 5.3. The discrepancies between the
power levels of the hardware implementation and simulation can be attributed the slight
differences in component response and some loss in components not explicitly modeled,
such as the cable from the output to the receiver.
These results clearly show that when the interferometer output was tuned to be
approximately -60 dBm at ~5.0062 GHz (as shown in Figure 5.10), the input to the
amplification stage was fed with a signal approximately at -144 dBm. For the input to
have such low power, the matching signal had to be well matched to the reflected signal
in order to provide optimal cancellation. Therefore the null at the output cannot be tuned
to be lower than -60 dBm. At this point the system is most sensitive to changes as the
interferometer caused the system to have the maximum dynamic range it is able to
achieve dictated by the thermal noise.
Figure 5.11: Output stage of MI.
94
Table 5.3: Signal propagation through the output stage of the MI.
IF (Hz)
PIN (dBm)
Pout (dBm)
Pout1 (dBm)
Pout2 (dBm)
1000
-144
-103
-79
-58
10
-164
-123
-99
-78
Figure 5.12: Output of cascaded amplifier output stage of interferometer with 50
Ohm matched loads on differential lines.
5.5 Summary
The analysis and discussion provided in this chapter have shown that by following the
ADS simulations developed in chapter 4 and addition of several modifications, it is
possible to construct a hardware implementation of the MI. Once the simulations are
upgraded to match the new topology developed throughout breadboarding, the
performance of both implementations is compared. The findings show that simulation
results and test measurements agree with each other, implying that both implementations
95
are well matched. This agreement means that for future developments the simulations can
be used to understand any effects on the device, and even integration of this device with
others. This chapter also covers points targeted at understanding the null, because
knowing its behaviour is important to be able to produce meaningful measurements.
Lastly, prior to attempting a nulling process it is important to know the depth of the null
that can be achieved under set conditions. The main parameters that govern the depth of
the null at the output of the system are noise floor and the gain at the output stage of the
interferometer.
96
Chapter Six: Experimental validations and applications
So far the focus has been on understanding the theory of operation of the interferometer
as well as the steps taken for its realization. With this knowledge, it is now possible to
begin comprehensive experiments to validate the interferometer. For this study, our main
concern is to achieve improved sensitivity which is shown through a comparison with the
current measurement methodology used for TSAR. Aside from the comparison
validation, there is also an interest to explore possible applications that can benefit from
this improvement. For example, the interferometer can be implemented to perform
Micro-Electro-Mechanical Systems (MEMS) Capacitor measurements. These capacitors
are appropriate candidates because they have very small capacitance resulting in high
impedance, which makes them relatively difficult to measure and characterize accurately.
Three experimental setups were investigated where each setup is built to
specifically target a key expectation of the interferometer’s performance. The first
involved a mini-anechoic chamber to compare the performance of the interferometer to
TSAR’s DMM under ideal conditions. The second setup is also concerned with a
performance comparison, although in this case the test bed is less than ideal and
incorporates the oil tank previously used in chapters 2 and 3. Lastly, a MEMS
measurement test station is used to investigate the ability of the interferometer to
integrate into already available systems, to show its ability to enhance sensitivity and to
demonstrate versatility. These experiments led to several interesting observations and
conclusions, which will be discussed throughout this chapter.
97
6.1 Mini anechoic chamber
In order to perform a fair comparison between the two measurement methodologies of
Interferometery and DMM, a test setup was developed as shown in Figure 6.1. This
involved the construction of a miniature anechoic chamber to provide an environment
with minimal reflections. To do this the inner walls of the chamber are padded with
Eccosorb AN absorbers (Emerson & Cuming, Randolph, MA). The sensor in this case is
a horn antenna that operates well around 4-5 GHz as it is well matched to free space. The
purpose of the antenna (#3160 Standard Gain Horn, ETS-LINDGREN, Cedar Park,
Texas) is to emit microwaves towards a tube inserted through the top of the chamber and
detect changes in the solution flowing through the tube. To do so, water is pumped
through the tube at a constant rate of 185 ml/min for a period of time until the system
stabilizes thermally (receivers in the NA reach temp stability after ~90 min) and air
bubbles are eliminated. This is verified by measuring the reference for each case and
waiting until it stops varying. Once stability is reached, glycerin (
the water solution (
) is added to
) in increments of 1 ml at 15 second interval and the
dielectric constant is monitored/measured by an external probe system [39].
The probe system consists of an open ended coaxial probe, which measures the
reflection coefficient at the calibration plane of the probe [39]. This plane is located at the
connection between the probe and the port of the NA. In this case, the same NA (PNA-X)
is used for both the interferometry and dielectric probe setups. The reflection coefficient
is then de-embedded to obtain the aperture reflection coefficient, which is then translated
to a dielectric constant measure using a rational function model technique [40].
98
Coaxial probe
Figure 6.1: Mini anechoic chamber setup using the MI with a PNA-X (N5242A
26.5GHz) as a platform. Top view made available by Jeremie Bourqui (University of
Calgary) shows dimensions in mm for Horn antenna (outside chamber), tube
insertion location and chamber measurements.
99
6.1.1 Interferometer measurements of low contrast glycerin
With the interferometer connected to the test setup (as shown in Figure 6.1), it is now
possible to take a measurement. The first step is to turn on the pump to allow water flow
in the tube system in order to let the system stabilize.
To determine stability, a null is found at a particular frequency in this case ~4.82
GHz. The null is then constantly recorded for a period of time. During this period if the
null begins to drift, the bias voltages are adjusted to regain the null. This process is
continued until the null no longer drifts. In general it would take the system a couple of
hours to stabilize, approximately 2- 3 hours would be sufficient. This is very important as
even a minor change in temperature will cause drift in the null during the test and
invalidate the measurements.
Figure 6.2 shows the stability of the null for the experimental measurement
chosen at 4.8187 GHz, with an IF of 10 Hz at an input power of -60 dBm. In order to
visually see the null on the NA, a BW of 155.858 kHz is chosen with 201 data points. For
this test the system has water flowing through the tubes and no glycerin is added. It can
be seen that the null behaviour is quite stable over a time interval of 160 seconds. This
provides assurance that an accurate and repeatable measurement can be made with
confidence.
The conditions at which Figure 6.2 is obtained are repeated, with the exception
that this time a perturbation to the system will be introduced in the form of glycerin
addition to the water flow. Figure 6.3 shows the output of the interferometer over a
period of time with the addition of glycerin at 15 second intervals and Figure 6.4 shows
the change in the dielectric constant of the mixture. It can be clearly seen at what time the
100
glycerin is added by the spikes in the response. These spikes do not actually occur due to
the glycerin itself, but rather due to the air bubbles occurring when the glycerin is added
using the syringe. The addition of the glycerin is actually noted by the shift of the whole
system response at ~95 seconds. At this point consecutive additions of glycerin increase
the output level. The interferometer detected the dilution of glycerin after the sixth
addition. This corresponds to a 0.35% change in the dielectric constant from its initial
System Output (dBm)
value.
-50
-55
-60
-65
-70
-75
-80
-85
-90
-95
-100
0
20
40
60
80
100
120
140
Time (s)
Figure 6.2: Interferometer output when the system is dormant.
160
Relative Change (unitless)
101
30
25
20
15
10
5
0
-5
-10
-15
-20
-25
-30
Noise floor shift
indicating initial
detection after
6th addition
0
10
20
30
40
50
60
70
80
90 100 110 120 130 140 150 160
Time (s)
Figure 6.3: System response to the addition of glycerin using an interferometry.
71.7
Dielctric Constant
71.6
71.5
71.4
71.3
71.2
71.1
71
0
10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160
Time (s)
Figure 6.4: Dielectric constant change as water is diluted with glycerin in room
temperature at 4.8 GHz.
102
6.1.2 Comparison to current technique
The end goal of this experiment has been to determine which system is more sensitive to
the addition of glycerin. Figure 6.3 confirmed that the interferometer is indeed able to
detect the perturbation after a certain amount is added. The next step was to perform the
same experiment with the DMM and to compare the two responses. The test setup
discussed above was kept similar for DMM, including the setting on the NA with the
exception of port power. Test port power had to change to -36.9 dBm to keep the input
power to the horn antenna the same as for the interferometer (MI output) case. The
response with no perturbation was recorded to verify that there is no drift, making the
measurement valid. Then the same amount of glycerin at 15 second intervals is added,
and the response of the system is recorded.
Figure 6.5 shows the response of both systems to the addition of glycerin. To
obtain these plots that data points leading to the first glycerin insertion are averaged and
the value is used as a reference point. The reference is subtracted from all of the data
points in the trace providing a relative change from an initial point. This is done on both
plots to be able to overlap them and determine which is more sensitive. It is clearly seen
that both systems detect the air bubbles occurring from the process of adding the glycerin
at approximately the same time interval. It is also noted that both plots overlap well on
each other and no major differences are seen. This suggests that both systems detect the
glycerin at the same time and detect consecutive additions with the same sensitivity. Even
though this experiment shows that for this case the interferometer is not able to provide
greater sensitivity than DMM, it also suggests that, under ideal test conditions, the
performance of both systems is identical.
Relative Change (unitless)
103
30
25
20
15
10
5
0
-5
-10
-15
-20
-25
-30
0
10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160
Time (s)
Glycerin Gain Interferometer
Glycerin Gain DataMemory
Figure 6.5: Interferometer and DMM performance comparison by looking at
relative changes for anechoic chamber results.
6.2 Oil tank test bed – Low relative permittivity contrast comparison
The anechoic chamber experiment, although providing positive detection results, did not
show improvement due to the ideality of the setup (minimal unwanted reflections). The
next step was to revisit the oil tank experiment previously conducted in chapter 3 for the
analysis of the two source interferometer sensitivity. The reason for this is to move away
from an ideal case where the reflections are very minimal and do not obstruct detection of
reflections from the target. The absorbers used in the setup of this experiment help reduce
reflections but do not eliminate all reflections. The antenna is fairly well matched but
does radiate in the backwards direction, causing unwanted reflections. Only the Acrylic
rod is used which provides a very low contrast change. Both measurement systems are
104
used to determine which approach can detect the rod with greater sensitivity. Figure 6.6
shows the components used for this test.
Canola oil
εr=2.5
σ=0.04 S/m
Acrylic rod
εr=2.6
σ=6.25x10-15 S/m3
Figure 6.6: Oil tank test bed for revisited low contrast comparison.
6.2.1 Interferometer sensitivity response to low contrast objects
In chapter 3 it was discussed and shown that due to the low contrast between Canola oil
and the Acrylic rod, it is quite difficult to detect the insertion of this rod into a tank filled
with Canola oil. The dual source interferometry method was not able to detect this
change due to its relatively poor sensitivity; although DMM did detect a very small
change, it was deemed insignificant for proper validation of detection.
For the interferometer, the procedure discussed in chapter 3 is followed to
perform this measurement at ~4.466 GHz with an IF of 10 Hz at -60 dBm and 201 data
105
points. Figure 6.7 shows the output of the interferometer system as a rod is inserted into
System Output (dBm)
the test bed.
-40
-45
-50
-55
-60
-65
-70
-75
-80
-85
-90
-95
-100
0
25
50
75
100
125
150
175
200
225
250
275
300
325
350
Time (s)
Figure 6.7: Interferometer response to Acrylic rod insertion in an oil test bed.
Unlike before, the single source interferometer is able to detect the insertion of the
Acrylic rod. It is also interesting to note that the insertion and removal points can be
observed by the “overshoot” response. This is due to the extra perturbations caused when
the rod is moving. The null did drift slightly while taking the measurement, which is
noticed by the increase in the null power from ~ -70 dBm to ~ -63 dBm, although this
drift did not obscure the detection of the rod. For the best case, the difference between the
null (system dormant ~ -70 dBm) and Acrylic rod (system perturbed ~ -53 dBm) is 17
dB. This clearly shows that the rod is detected even though its contrast is small, and there
is also plenty of room to detect objects with even smaller contrast ratio to the Canola oil.
106
6.2.2 Comparison of techniques
To compare both measurement systems, the oil tank experiment is repeated with the
DMM technique and its result is compared against the interferometer result from Figure
6.7. The comparison of the data is again done by looking at the relative change between
the response when the system is dormant and where the system is at perturbed state.
Figure 6.8 clearly shows that the interferometer technique is able to detect the Acrylic rod
with a greater margin between the noise floor of the system and the response of the
object. The relative change with the DMM is 10 as opposed to interferometry having a
delta of 17. This test was repeated multiple times and the same response was observed,
thus providing confidence that detection is a real response.
This result shows that under realistic test conditions, the interferometry technique
is able to provide greater sensitivity compared to DMM. In this case study, there is a delta
of 7 between the two techniques, which means that it would be possible to find a contrast
case where DMM will not detect a change while interferometry will. There is room for
improvement for the interferometer, as it does show a degraded null over time as opposed
to DMM, which maintains a noise floor at a constant level quite well. This could be
addressed in the future with a temperature compensation feedback loop. Another key
point is that the performance of the interferometer can easily be greatly improved by
using a low noise amplifier (LNA) at the null stage since the current amplifier has 12 dB
of NF, which is fairly large and causes an unwanted rise in noise floor. Therefore, the
reflection has to be large enough to cause a change, which will overcome the NF for the
interferometer to detect it.
Relative Change (dB)
107
30
25
20
15
10
5
0
-5
-10
-15
-20
-25
-30
0
25
50
75 100 125 150 175 200 225 250 275 300 325 350 375 400
Time (s)
Interferometer
Data Memory
10 per. Mov. Avg. (Interferometer)
10 per. Mov. Avg. (Data Memory)
Figure 6.8: Interferometer and DMM performance comparison by looking at
relative changes for oil tank experiment.
6.3 MEMS
The purpose of the MI is to enhance the sensitivity of a measurement system and provide
the ability to measure very small differences between objects. One example of a practical
application for this device is manufacturing processes that produce elements with great
precision. The interferometer can be used to determine if the values are within a set
tolerance from a known reference (null). In order to show this practicality, the MI is
integrated into a MEMS measurement setup shown in Figure 6.9. This will validate the
ability of the interferometer to distinguish small value changes in MEMS capacitors using
information provided by the null shift.
108
Figure 6.9: MEMS capacitor measurement setup.
6.3.1 Capacitance sensitivity
The first set of measurements using the MEMS capacitor setup involved looking at the
ability to obtain a null and its properties as the capacitor value at the sensor varies. To do
this, two capacitors [41] with relatively close values were chosen in order to determine if
it is possible to null one of them and detect the change caused by the other. The values of
the capacitors were first obtained using a standard single port, Short-Open-Load
calibration with a Cascade Microtech Impedance Standard Substrate and a Cascade
Microtech probe (Cascade Microtech Inc., Beaverton, Oregon). Reflection data (S 11) for
each element were obtained and the capacitance was calculated to be ~40 fF and ~57 fF.
The next step was to connect the probe to the interferometer test setup and to bring it in
contact with the 40 fF capacitor. An example of a MEMS capacitor (top view) is shown
in the RHS of Figure 6.9. In order to connect to these elements three traces are used. The
two outer traces connect to a movable bridge and the central trace connects to the signal
109
line. By applying DC it is possible to move the bridge down and vary the capacitance.
The tip of the probe is made of three connectors (ground-signal-ground) that line up with
the measurement traces. The tip is positioned on the trace and then pushed slightly
forward to scratch the surface and make good contact. For this experiment the DC was
not varied and the interest was to observe the frequency shift of the null, as the sensor is
brought up against different elements. An experiment in which DC is varied has been
considered as part of future work.
The settings on the NA were as follows; test port power -60 dBm at port 1, IF of
10 Hz and 401 data points over 4.7887 – 4.889 GHz. While the sensor remains in contact
with the capacitor, the biasing voltage on the attenuator is varied until a null power level
of approximately -65 dBm is reached. To do so, the coarse bias is increased as long as it
improves the depth of the null; the moment an increase of voltage results in decrease of
null performance, the voltage is taken to its previous state. To tune the null further, the
fine bias voltage is then varied similarly to the coarse bias. With the null in place the
system is ready to detect a change in capacitance. For this, the sensor is moved towards
the 57 fF capacitor on the wafer and the result is shown in Figure 6.10. It is clearly
observed that the system is able to detect a change between the two elements, which is
expressed through the frequency shift of the null.
110
Figure 6.10: Interferometer output at nulled state (red) and shifted null (green) in
response to a change.
6.3.2 Capacitance quantification
The null can provide multiple pieces of information through amplitude and frequency
shifts. A frequency shift translates to a capacitive or inductive change, while an
amplitude shift translates to a resistive change. In the previous section, it was shown that
the interferometer is able to detect the capacitance change between elements by
monitoring the frequency shift of the null. To further investigate this, several capacitors
were chosen with values ranging from ~40 fF to ~368 fF. A capacitor of value ~107 fF
was placed in contact with the sensor and a null was generated. From this reference point,
the rest of the chosen capacitors were brought in contact with the sensor and the
frequencies to which the reference null had shifted were recorded. Figure 6.11 shows that
there is a linear relationship between the null’s shifted frequency and the capacitance
value of the element.
111
300
Capacitance (fF)
250
200
150
100
50
0
4.427
4.428
4.429
4.43
4.431
Frequency (GHz)
4.432
4.433
Figure 6.11: Linear relationship between null frequency shifts and capacitance
values.
This means that with prior information (known reference or look-up table) the
interferometer is able to provide means to determine quantities about objects, such as
capacitance.
112
6.4 Conclusion
The focus of this chapter has been to provide experimental evidence to the claims made
regarding the improvement potential of the interferometer for sensitivity. This was shown
through three experiments each focusing on a specific goal. The anechoic chamber
experiment has shown that under ideal conditions the interferometer can match a
technique that is currently used for data collection in breast cancer detection. The oil tank
experiment clearly shows that in a more realistic case where conditions are less than
ideal, the interferometer is more sensitive to minute changes. This is a significant result
as it confirms the ability of the interferometer to enhance the sensitivity of a measurement
system. The interferometer was also integrated into a second measurement system, which
is used to characterize MEMS capacitors. First it is shown that the interferometer indeed
operates well with the system and remains sensitive. Then, a potential application of the
data provided by the interferometer’s null is explored.
An important benefit of the interferometer is that there is no design constraint to a
particular sensor and the device can operate with a multitude of sensors. Since the
interferometer is able to improve the sensitivity of a measurement system, it does not
need to have a perfectly matched sensor to obtain high sensitivity. This allows for sensors
used with this device to have looser constraints and system sensitivity to be maintained.
By being able to loosen design constraints on sensors it is possible to save money and
time needed for their development. Lastly, this study demonstrated both the theory and
design of the MI, as well as providing a successful implementation. From the summary of
results it is clearly noticed that a tool has been developed that is able to tune a
113
measurement system to provide increased sensitivity by repositioning a system from low
to high resolution region through the creation of a virtual match.
114
Chapter Seven: Conclusion
Relative measure microwave interferometry has been introduced as a potential method to
enhance sensitivity of microwave measurement systems. Interferometry provides the
means to cancel unwanted reflections that degrade sensitivity by creating a virtual match.
This approach focuses on the relative information that can be obtained from the null’s
magnitude and frequency shifts. Magnitude shift relates to a resistive change, while a
frequency shift can be translated to capacitive or inductive variation. With a null created
at a known or unknown reference, minute changes around the characteristics of the
reference can now be detected with great sensitivity. The first MI prototype showed
promise of these claims and sparked interest for further exploration, although it suffered
from repeatability due to instability of output.
This thesis, focused on sensitivity enhancement of microwave measurement
systems, launched an investigation into the necessary steps of realizing a stable and
repeatable implementation of the MI. To do so, the first prototype was analyzed and
broken down to its main components. Three internal and three external components were
identified as splitter, attenuator, amplifier, source, coupler and receiver. This allowed the
construction of a simplified two source interferometric setup in order to perform proofof-concept testing with a stable implementation. Experiments were set up to study
repeatability, sensitivity and ability to eliminate reflections. Even though, in comparison
to DMM, the sensitivity of this implementation was fairly poor, repeatability was
improved. Promising results were also seen when the interferometer was able to bring out
the response of an object, which in DMM was masked by the response of the antenna.
115
This also provided evidence that the interferometer is not limited to a single frequency
and can collect wideband data for processing.
To move away from the simplicity of a two source interferometer, a
comprehensive ADS simulation model of the proposed single source implementation was
developed. First the model was built with realistic components, to quickly verify
operation and determine key components. It was quickly found that the null is highly
sensitive to the response of the attenuator and the phase shifting properties of the
transmission line cable. As a result, when the ideal components were replaced with
realistic models, the attenuator was carefully designed from ground up, while the
transmission cable was allowed to vary to act as a phase shifter. The complete model was
verified through an analysis of signal and noise propagation at the various stages of the
circuit, while monitoring the SNR. Propagation of both signal and noise were as expected
and the output can be tuned to achieve a null, which was governed by the thermal noise.
The interferometer’s hardware components were brought together and assembled
to match the designed schematic in ADS. Several debugging stages had to take place
prior to achieving an output. Extra care was taken building the attenuator and it was
tested separately to verify its behaviour and performance with applied coarse and fine
biasing voltages. Debugging work led to modification of circuitry, such as the need for a
BPF and a DC block, as well as having to move away from a complete differential
topology of output amplifiers. The ADS schematic was then modified to match this new
topology. This did degrade the output gain, but it is still sufficiently high. Experiments
done to validate the agreements between hardware implementation and simulations
showed excellent agreement.
116
Lastly, the prototype MI was integrated into three different measurement systems.
The mini-anechoic chamber test showed that under ideal conditions the interferometer
can match DMM, when attempting to sense low contrast changes in a solution. The oil
tank procedure confirmed that for a realistic setup the interferometer has improved
sensitivity of ~7 dB over DMM, when sensing a perturbation with a low contrast object.
The MEMS capacitors were used as an example of a possible implementation of the
interferometer, which showed great promise as the interferometer was able to integrate
into the system with ease and behave as expected. The first two tests involved looking at
the magnitude shift of the null at a single frequency, while the last test involved recording
the frequency shift of the null. This clearly shows that both magnitude and frequency
shifts in null can provide meaningful information.
This is the first experimental work developing a reflection based MI tool that is
independent of application, provides wideband time domain application, contains a
comprehensive simulation model in agreement with hardware, covers in-depth analysis of
the null and utilizes relative measurements.
7.1 Future Work
The phase shifting properties of the circuit are set by the length of cable to the sensor.
This is disadvantageous as it requires a physical change of the cable to shift the frequency
location of the null. By adding a phase shifter in the attenuator path, it would be possible
to eliminate the need for this. This is not simple work as it would pose a great difficulty
in tuning, since the phase shifter will also provide some unwanted attenuation which
would throw off the attenuation achieved by the attenuator. Therefore there will be a need
117
to tune more parameters at a time to obtain a null, although this can be tackled by
developing an optimization procedure.
The sensitivity of the current implementation of the MI can be further enhanced
by using a lower NF amplifier as the first amplifier at the output stage where the null is
formed. This change can greatly increase the SNR of the interferometer and further
improve sensitivity. The reason for this is the 12 dB NF of the amplifier. This means that
the change in the null must overcome the 12 dB NF before the change can be pulled out
of the noise floor and recorded.
Currently the interferometer operates by recording changes in dB that occur at a
single frequency where the null is tuned. The nulls do occur periodically and it should be
possible by means of more accurate tuning to obtain a deep null at those multiple
frequencies rather than at a single point. This would allow the interferometer to take
multiple measurements and provide results across multiple frequencies at the same time.
Simple tests have shown that, by modifying the transmission line length, the frequency
step between nulls can be altered in order to collect more or fewer points within a
frequency range.
Lastly, the interferometer in its current implementation is built from separate
components that connect together. Prototyping the interferometer as a single unit where
for example traces can replace cables can greatly improve the stability of the null.
118
Bibliography
[1]
E. Zastrow, S. C. Hagness, B. D. Van Veen, and J. E. Medow, "Time-Multiplexed
Beamforming for Noninvasive Microwave Hyperthermia Treatment," Biomedical
Engineering, IEEE Transactions on, vol. 58, pp. 1574-1584, 2011.
[2]
Y. Deshan, J. M. Bertram, M. C. Converse, A. P. O'Rourke, J. G. Webster, S. C.
Hagness, J. A. Will, and D. M. Mahvi, "A floating sleeve antenna yields localized
hepatic microwave ablation," Biomedical Engineering, IEEE Transactions on,
vol. 53, pp. 533-537, 2006.
[3]
N. K. Nikolova, "Microwave Imaging for Breast Cancer," Microwave Magazine,
IEEE, vol. 12, pp. 78-94, 2011.
[4]
S. Y. Semenov, A. E. Bulyshev, A. Abubakar, V. G. Posukh, Y. E. Sizov, A. E.
Souvorov, P. M. van den Berg, and T. C. Williams, "Microwave-tomographic
imaging of the high dielectric-contrast objects using different imagereconstruction approaches," Microwave Theory and Techniques, IEEE
Transactions on, vol. 53, pp. 2284-2294, 2005.
[5]
N. H. Farhat, “Microwave holography and coherent tomography,” in L. E. Larsen
and J. H. Jacobi , Eds., Medical Applications of Microwave Imaging. New York:
IEEE Press, 1986, pp. 66–81.
[6]
M. Klemm, J. A. Leendertz, D. Gibbins, I. J. Craddock, A. Preece, and R.
Benjamin, "Microwave Radar-Based Differential Breast Cancer Imaging:
Imaging in Homogeneous Breast Phantoms and Low Contrast Scenarios,"
Antennas and Propagation, IEEE Transactions on, vol. 58, pp. 2337-2344, 2010.
[7]
M. Klemm, I. J. Craddock, J. A. Leendertz, A. Preece, D. R. Gibbins, M. Shere,
and R. Benjamin, "Clinical trials of a UWB imaging radar for breast cancer," in
Antennas and Propagation (EuCAP), 2010 Proceedings of the Fourth European
Conference on, 2010, pp. 1-4.
[8]
R. K. Amineh, M. Ravan, A. Khalatpour, and N. K. Nikolova, "ThreeDimensional Near-Field Microwave Holography Using Reflected and
Transmitted Signals," Antennas and Propagation, IEEE Transactions on, vol. 59,
pp. 4777-4789, 2011.
119
[9]
D. J. Kurrant and E. C. Fear, "Regional estimation of the dielectric properties of
the breast: Skin, adipose, and fibroglandular tissues," in Antennas and
Propagation (EUCAP), Proceedings of the 5th European Conference on, 2011,
pp. 2920-2924.
[10]
T. Williams, E. C. Fear, and D. W. Westwick, "Tissue sensing adaptive radar for
breast cancer detection: investigations of reflections from the skin," in Antennas
and Propagation Society International Symposium, 2004. IEEE, 2004, pp. 24362439 Vol.3.
[11]
M. Lazebnik, D. Popovic, L. McCartney, C. B. Watkins, M. J. Lindstrom, J.
Harter, S. Sewall, T. Ogilvie, A. Magliocco, T. M. Breslin, W. Temple, D. Mew,
J. H. Booske, M. Okoniewski, and S. C. Hagness, "A large-scale study of the
ultrawideband microwave dielectric properties of normal, benign and malignant
breast tissues obtained from cancer surgeries," Physics in Medicine and Biology,
vol. 52, p. 6093, 2007.
[12]
G. Bellizzi, O. M. Bucci, and I. Catapano, "Microwave Cancer Imaging
Exploiting Magnetic Nanoparticles as Contrast Agent," Biomedical Engineering,
IEEE Transactions on, vol. 58, pp. 2528-2536, 2011.
[13]
A. Mashal, B. Sitharaman, X. Li, P. K. Avti, A. V. Sahakian, J. H. Booske, and S.
C. Hagness, "Toward carbon-nanotube-based theranostic agents for microwave
detection and treatment of breast cancer: enhanced dielectric and heating response
of tissue-mimicking materials," IEEE Trans Biomed Eng, vol. 57, pp. 1831-4,
Aug 2010.
[14]
A. Mashal, J. H. Booske, and S. C. Hagness, "Toward contrast-enhanced
microwave-induced thermoacoustic imaging of breast cancer: An experimental
study of the effects of microbubbles on simple thermoacoustic targets," Physics in
Medicine and Biology, vol. 54, pp. 641-650, 2009.
[15]
P. Hariharan, Basics of Interferometry. Burlington, MA, USA: Academic Press,
2006.
[16]
G. W. Swenson, Jr. and N. C. Mathur, "The interferometer in radio astronomy,"
Proceedings of the IEEE, vol. 56, pp. 2114-2130, 1968.
[17]
J. K. Wallace, "Common-path interferometric wavefront sensing for space
telescopes," in Aerospace Conference, 2011 IEEE, 2011, pp. 1-7.
120
[18]
M. E. Tobar, E. N. Ivanov, J. G. Hartnett, J. M. le Floch, and P. L. Stanwix, "High
precision microwave interferometers and oscillators for applied and fundamental
physics applications," in Electromagnetics in Advanced Applications (ICEAA),
2010 International Conference on, 2010, pp. 803-806.
[19]
D. W. Griffin, "MW interferometers for biological studies," Microwave Journal,
vol. 21, pp. 69-72, 1978.
[20]
M. Pieraccini, M. Fratini, F. Parrini, and C. Atzeni, "Dynamic Monitoring of
Bridges Using a High-Speed Coherent Radar," Geoscience and Remote Sensing,
IEEE Transactions on, vol. 44, pp. 3284-3288, 2006.
[21]
G. De Pasquale, G. Bernardini, P. P. Ricci, and C. Gentile, "Ambient vibration
testing of bridges by non-contact microwave interferometer," IEEE Aerospace
and Electronic Systems Magazine, vol. 25, pp. 19-26, 2010.
[22]
C. Atzeni, A. Bicci, D. Dei, M. Fratini, and M. Pieraccini, "Remote Survey of the
Leaning Tower of Pisa by Interferometric Sensing," Geoscience and Remote
Sensing Letters, IEEE, vol. 7, pp. 185-189, 2010.
[23]
Agilent Technologies. (2004, August 31). Network Analyzer Basics [Online].
Available: http://cp.literature.agilent.com/litweb/pdf/5965-7917E.pdf
[24]
D. M. Pozar, Microwave Engineering: John Wiley & Sons, 2011.
[25]
H. Tanbakuchi, L. C. Betts, and D. Blackham, "Impedance Analyzer," US Patent
7161358, January 9, 2007.
[26]
R. Debroucke, J. F. Larchanche, D. Theron, D. Ducatteau, H. Tanbakuchi, and C.
Gaquiere, "Sub-femtoFarad MOS varactor characterization tools," in Microwave
Conference (EuMC), 2010 European, 2010, pp. 783-786.
[27]
Agilent Technologies. (2000, September). Understanding and Improving Network
Analyzer
Dynamic
Range
[Online].
Available:
http://cp.literature.agilent.com/litweb/pdf/5980-2778EN.pdf
[28]
J. Bourqui, M. A. Campbell, T. Williams, and E. C. Fear, "Antenna Evaluation for
Ultra-Wideband Microwave Imaging," International Journal of Antennas and
Propagation, vol. 2010, 2010.
[29]
C. J. Shannon, M. Okoniewski, and E. C. Fear, "A dielectric filled ultra-wideband
antenna for breast cancer detection," in Antennas and Propagation Society
International Symposium, 2003. IEEE, 2003, pp. 218-221 vol.1.
121
[30]
J. Bourqui, M. Okoniewski, and E. C. Fear, "Balanced Antipodal Vivaldi Antenna
for Breast Cancer Detection," in Antennas and Propagation, 2007. EuCAP 2007.
The Second European Conference on, 2007, pp. 1-5.
[31]
Z. Hanqiao, L. Chaojiang, R. Divan, A. Hoffmann, and W. Pingshan, "Broadband
Mag-Noise of Patterned Permalloy Thin Films," Magnetics, IEEE Transactions
on, vol. 46, pp. 2442-2445, 2010.
[32]
Z. Hanqiao, R. Divan, and W. Pingshan, "Ferromagnetic resonance of a single
magnetic nanowire measured with an on-chip microwave interferometer," Review
of Scientific Instruments, vol. 82, p. 054704 (4 pp.), 2011.
[33]
R. Debroucke, D. Gloria, D. Ducatteau, D. Theron, H. Tanbakuchi, and C.
Gaquiere, "attoF MOS varactor RF measurement VNA coupled with
interferometer," in Microwave Measurement Conference (ARFTG), 2011 77th
ARFTG, 2011, pp. 1-2.
[34]
G. A. Ferrier, S. F. Romanuik, D. J. Thomson, G. E. Bridges, and M. R. Freeman,
"A microwave interferometric system for simultaneous actuation and detection of
single biological cells," Lab on a Chip, vol. 9, pp. 3406-12, 2009.
[35]
E. N. Ivanov and M. E. Tobar, "Noise measurements beyond the standard thermal
noise limit," in Frequency Control Symposium, 2008 IEEE International, 2008,
pp. 79-80.
[36]
J. M. Sill and E. C. Fear, "Tissue sensing adaptive radar for breast cancer
detection: preliminary experimental results," in Microwave Symposium Digest,
2005 IEEE MTT-S International, 2005, p. 4 pp.
[37]
Avago Technologies. (2006, June 21). HMPP-3865 MiniPAK PIN Diode High
Isolation SPDT Switch Design for 1.9 GHz and 2.45 GHz Applications [Online].
Available: http://www.avagotech.com/docs/5988-8470EN
[38]
Agilent Technologies. (2006, September). Circuit Envelope Simulation [Online].
Available: http://cp.literature.agilent.com/litweb/pdf/ads2006/pdf/cktsimenv.pdf
[39]
D. Popovic, L. McCartney, C. Beasley, M. Lazebnik, M. Okoniewski, S. C.
Hagness, and J. H. Booske, "Precision open-ended coaxial probes for in vivo and
ex vivo dielectric spectroscopy of biological tissues at microwave frequencies,"
Microwave Theory and Techniques, IEEE Transactions on, vol. 53, pp. 17131722, 2005.
122
[40]
J. M. Anderson, C. L. Sibbald, and S. S. Stuchly, "Dielectric measurements using
a rational function model," Microwave Theory and Techniques, IEEE
Transactions on, vol. 42, pp. 199-204, 1994.
[41]
B. Wu, M. Okoniewski, and M. E. Potter, "Design and Fabrication of a Ternary
Switch for MEMS-Controlled Reflectarray Elements," Antennas and Wireless
Propagation Letters, IEEE, vol. 8, pp. 998-1001, 2009.
Документ
Категория
Без категории
Просмотров
0
Размер файла
6 692 Кб
Теги
sdewsdweddes
1/--страниц
Пожаловаться на содержимое документа