A SELF-STRUCTURING TWO-PORT NETWORK
By
Andrew K. G. Temme

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
MASTER OF SCIENCE
ELECTRICAL ENGINEERING
2011

ABSTRACT
SELF-STRUCTURING TWO-PORT NETWORK
By
Andrew K. G. Temme

This work explores the capabilities of a self-structuring two-port network. Based on the idea
of a self-structuring antenna, a self-structuring two-port network utilizes switches to dynamically reconfigure itself using intelligent search algorithms such as genetic algorithms (GA).
This allows for the S-parameters of the self-structuring two-port network to be adjusted to
match, ideally, any desired set of S-parameters. A combination of simulations and experiments were carried out to characterize such a network. It was found that the constructed
prototype exhibited significant loss of power yet no one mechanism could be identified either
through simulations or experiments as the most significant mechanism. Steps were taken in
the experimental setup to mitigate the power loss with relatively little improvement. Given
the loss of power, the capabilities of the manufactured self-structuring two-port network were
investigated using both random and GA based searches. Results from the random search
suggested that the lowest losses for the network are over the frequency range of approximately 750 MHz to 1.5 GHz. Random search results suggested that switch combinations
exist for matching various types of loads and to create an adjustable attenuator among
other applications. Through the use of a GA certain applications were investigated further.
Accepting the system power loss, the network was fairly successful in matching arbitrary
loads. The network was also used successfully to create an arbitrary attenuator with losses
larger than normally observed in the network. Due to the high power loss observed in both
simulations and in experiments, further research should be directed to creating a two-port
network with lower losses. Doing so would allow for further exploration of this topic.

Soli Deo gloria.

iii

ACKNOWLEDGMENTS

When I made my decision to come to Michigan State University for my undergraduate degree,
I knew very little about the EM group. When I made my decision to stay at MSU for graduate
school, I knew a great deal about the EM group. Without having attended MSU for my
bachelor’s degree, there is little chance that I would be attending MSU now or even studying
EM at another school. The experiences that the undergraduate program offered me here have
shaped my career goals and opportunities. One of the reasons why I attended Michigan State
University for my undergraduate degree was the STARR Charitable Foundation Scholarship.
Due to the significant impact that the STARR Charitable Foundation Scholarship has had
on me and the opportunities afforded to me, I would like express my continued thanks and
apprecaition to the anonymous donors. Their financial support has helped bring me to where
I am today.
Completion of this thesis has required a large amount of work. I received support from
many sources in order to accomplish this.
I would like to first thank my advisor Dr. Edward Rothwell. His has offered guidance and
advice since I first came to MSU. As a research advisor, Dr. Rothwell has helped me develop
my research abilities and skills as well as improving my technical writing. His guidance has
been invaluable throughout this research.
Secondly, I would like to thank Dr. Raoul Ouedraogo for mentoring me since I first joined
the Electromagnetics Research Group (EM Group) in the summer of 2008 and helped him
on his master’s work. He has been a source of information on all things self-structuring and
genetic algorithm.
I would like to thank Dr. Rothwell and Dr. Shanker Balasubramaniam for bringing me
into the EM Group and giving me an opportunity to work on both computational and
experimental EM during my first summer. The EM Group has also been supportive and
helpful during my time as a group member and I would like to thank them for this.
iv

Further, I would like to thank Dr. Rothwell, Dr. Shanker, and Dr. Prem Chahal for
serving on my thesis committee.
A large thanks goes out to Don VanderLaan for all of his help with LabView. Collin
Meierbachtol has been kind enough to offer his L EX 2ε thesis template for use by other
AT
students, upon which this thesis typesetting is based. Andrew Prey and Dr. Raoul Ouedraogo
were kind enough to review this thesis and provide their comments.
Dr. Jack Ross has been kind enough to provide and help the EM Group with his software
genetic algorithm software for the FEKO Suite, GA-FEKO. Your continued support of MSU
is much appreciated.
I would also like to thank my family, Hanna Miller, and Kathryn Bonnen for their support
(and food) during my research and especially during the writing process.
This work has been supported by a scholarship from the IEEE Antennas and Propagation
Society, through a Graduate Office Fellowship, and through a National Science Foundation
Graduate Research Fellowship.
This thesis is copyright c 2011 by Andrew K. G. Temme.

v

Table of Contents
List of Tables

viii

List of Figures

xii

List of Listings

xiii

1 Introduction

1

2 Background and Literature Review
2.1 Self-Structuring Antennas and Systems . . . . .
2.1.1 Self-Structuring Patch Antenna . . . . .
2.1.2 Automatically Tunable Cavity Resonator
2.2 Two-Port Networks . . . . . . . . . . . . . . . .
2.3 Self-Structuring Two-Port Network . . . . . . .
2.4 Power Balance . . . . . . . . . . . . . . . . . . .

. . . . .
. . . . .
System
. . . . .
. . . . .
. . . . .

3 Simulations of a Self-Structuring Two-Port Network
3.1 Initial Investigation and Simulations . . . . . . . . . .
3.2 Power Balance Simulations . . . . . . . . . . . . . . . .
3.2.1 FEKO Suite 6.1 Upgrade . . . . . . . . . . . . .
3.2.2 List of Simulations . . . . . . . . . . . . . . . .
3.2.3 Cavity Simulations . . . . . . . . . . . . . . . .
3.2.4 Coplanar Simulations . . . . . . . . . . . . . . .
3.2.5 Perimeter Post Simulations . . . . . . . . . . .
3.2.6 Switched Shorting Post Simulations . . . . . . .
3.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

4 Experimental Investigation
4.1 Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.1 Two-Port Network . . . . . . . . . . . . . . . . . . .
4.1.2 HP 8753D Network Analyzer . . . . . . . . . . . . .
4.2 Selection of States . . . . . . . . . . . . . . . . . . . . . . .
4.3 LabView Software . . . . . . . . . . . . . . . . . . . . . . . .
4.3.1 Random Search . . . . . . . . . . . . . . . . . . . . .
4.3.2 Standing Wave Ratio . . . . . . . . . . . . . . . . . .
4.3.3 Genetic Algorithm . . . . . . . . . . . . . . . . . . .
4.4 Random Search . . . . . . . . . . . . . . . . . . . . . . . . .
4.4.1 Initial Investigation . . . . . . . . . . . . . . . . . . .
4.4.2 Matching a Wideband Horn Antenna . . . . . . . . .
4.4.3 Power Balance of Self-Structuring Two-Port Network
4.4.4 Mitigation techniques . . . . . . . . . . . . . . . . . .
vi

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

3
3
5
7
7
10
11

.
.
.
.
.
.
.
.
.

14
14
15
18
18
19
22
27
41
41

.
.
.
.
.
.
.
.
.
.
.
.
.

46
46
46
53
54
55
56
58
58
62
62
71
74
76

4.5

4.6

Genetic Algorithm Based Search . . . . .
4.5.1 Optimized transmission . . . . .
4.5.2 Spiral . . . . . . . . . . . . . . .
4.5.3 Dipole Antenna Transmission . .
4.5.4 Attenuator . . . . . . . . . . . . .
Switch Investigation . . . . . . . . . . .
4.6.1 Individual Switch Properties . . .
4.6.2 Switch Removal from Experiment

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

78
82
82
86
90
91
91
91

5 Conclusion
104
5.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
5.2 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
5.3 Topics and Suggestions for Future Study . . . . . . . . . . . . . . . . . . . . 105
Appendix

108

HP 85052D Economy Mechanical Calibration Kit, DC to 26.5 GHz, 3.5 mm 109
Bibliography

112

vii

List of Tables
4.1

Pin and port placement

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

50

4.2

Genetic algorithm search parameters and results. . . . . . . . . . . . . . . .

81

viii

List of Figures
2.1

SSA schematic

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4

2.2

Shorting post, pad, and switch. For interpretation of the references to color
in this and all other figures, the reader is referred to the electronic version of
this thesis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

6

2.3

Automatically tunable cavity resonator. . . . . . . . . . . . . . . . . . . . . .

8

2.4

Prototype self-structuring two-port network. . . . . . . . . . . . . . . . . . .

11

3.1

GA optimized band pass filter centered at f0 =5.005 GHz. . . . . . . . . . . .

16

3.2

Bandpass filter tunability over 4.90–5.05 GHz . . . . . . . . . . . . . . . . .

17

3.3

Free cavity report. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

20

3.4

Free cavity power balance report. . . . . . . . . . . . . . . . . . . . . . . . .

21

3.5

Lossless cavity report. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

23

3.6

Lossless cavity power balance report. . . . . . . . . . . . . . . . . . . . . . .

24

3.7

Lossy cavity report. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

25

3.8

Lossy cavity power balance report. . . . . . . . . . . . . . . . . . . . . . . .

26

3.9

no-sides-lossless-cavity report. . . . . . . . . . . . . . . . . . . . . . . . . . .

28

3.10 no-sides-lossless-cavity power balance report. . . . . . . . . . . . . . . . . . .

29

3.11 no-sides-lossy-cavity report. . . . . . . . . . . . . . . . . . . . . . . . . . . .

30

3.12 no-sides-lossy-cavity power balance report. . . . . . . . . . . . . . . . . . . .

31

3.13 post-cavity-lossless report. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

32

3.14 post-cavity-lossless power balance report. . . . . . . . . . . . . . . . . . . . .

33

3.15 post-cavity-lossless-walls report. . . . . . . . . . . . . . . . . . . . . . . . . .

34

3.16 post-cavity-lossless-walls power balance report. . . . . . . . . . . . . . . . . .

35

ix

3.17 post-cavity-lossy report. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

37

3.18 post-cavity-lossy power balance report. . . . . . . . . . . . . . . . . . . . . .

38

3.19 post-cavity-lossy-15 report. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

39

3.20 post-cavity-lossy-15 power balance report. . . . . . . . . . . . . . . . . . . .

40

3.21 post-cavity-lossy-walls report. . . . . . . . . . . . . . . . . . . . . . . . . . .

42

3.22 post-cavity-lossy-walls power balance report. . . . . . . . . . . . . . . . . . .

43

3.23 rings-reduced report. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

44

3.24 rings-reduced power balance report. . . . . . . . . . . . . . . . . . . . . . . .

45

4.1

Prototype self-structuring two-port network. Reprint of Figure 2.4. . . . . .

47

4.2

Material parameters of the circuit board used to create the self-structuring
two-port network prototype provided by Taconic. . . . . . . . . . . . . . . .

48

4.3

Schematic view of self-structuring two-port. . . . . . . . . . . . . . . . . . .

49

4.4

Genetic algorithm flow chart. . . . . . . . . . . . . . . . . . . . . . . . . . .

60

4.5

Experimental setup for initial random search experiment. . . . . . . . . . . .

62

4.6

Random search results, S11 magnitude, 5-20-21-2001-combine. . . . . . . .

63

4.7

Random search results, S11 phase, 5-20-21-2001-combine. . . . . . . . . .

64

4.8

Random search results, S12 magnitude, 5-20-21-2001-combine. . . . . . . .

65

4.9

Random search results, S12 phase, 5-20-21-2001-combine. . . . . . . . . .

66

4.10 Random search results, S21 magnitude, 5-20-21-2001-combine. . . . . . . .

67

4.11 Random search results, S21 phase, 5-20-21-2001-combine. . . . . . . . . .

68

4.12 Random search results, S22 magnitude, 5-20-21-2001-combine. . . . . . . .

69

4.13 Random search results, S22 phase, 5-20-21-2001-combine. . . . . . . . . .

70

4.14 Wideband horn antenna matching experimental setup. . . . . . . . . . . . .

71

4.15 Random search of possible states and original unmatched antenna performance. 72
4.16 Random search zoomed in. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
x

73

4.17 Maximum and minimum observed power balances. . . . . . . . . . . . . . . .

74

4.18 Average observed power balances. . . . . . . . . . . . . . . . . . . . . . . . .

75

4.19 Copper covered self-structuring two-port network. . . . . . . . . . . . . . . .

77

4.20 Best S21 and S11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

79

4.21 Difference between maximum transmission and minimum reflection. . . . . .

80

4.22 Optimized transmission with 50 Ω loads, 750 MHz to 1.0 GHz . . . . . . . .

83

4.23 Optimized transmission with 50 Ω loads, 1.4GHz to 1.5 GHz . . . . . . . . .

84

4.24 Spiral antenna used for input impedance matching experiments. . . . . . . .

85

4.25 Experimental setup schematic for spiral antenna measurements and optimization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

86

4.26 Spiral antenna transmission optimization results at 1.4 GHz to 1.5 GHz. . .

87

4.27 Spiral antenna system transmission, 1.9 GHz to 2.2 GHz . . . . . . . . . . .

88

4.28 Experimental setup schematic for dipole antenna measurements and optimization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

88

4.29 Dipole antenna throughput results, dataset 7-8-2010-155PM-v. . . . . . . .

89

4.30 20 dB attenuator optimization results. . . . . . . . . . . . . . . . . . . . . .

90

4.31 Switch testing microstrip setup. . . . . . . . . . . . . . . . . . . . . . . . . .

92

4.32 S-parameters of an opened relay switch. . . . . . . . . . . . . . . . . . . . . .

93

4.33 S-parameters of a closed relay switch. . . . . . . . . . . . . . . . . . . . . . .

94

4.34 Power balance of a relay switch. . . . . . . . . . . . . . . . . . . . . . . . . .

95

4.35 S-parameters for state 0x00000000. . . . . . . . . . . . . . . . . . . . . . . .

96

4.36 S-parameters with switches unsoldered from shorting posts. . . . . . . . . . .

97

4.37 Power balance for switches unsoldered from shorting posts. . . . . . . . . . .

98

4.38 Difference in power balance for between open state and unsoldered switches.

99

4.39 Limited frequency range for the difference in power balance for between open
state and unsoldered switches. . . . . . . . . . . . . . . . . . . . . . . . . . . 100

xi

4.40 S-parameters with no switches connected to the two-port network. . . . . . . 102
4.41 Power balance with no switches connected to the two-port network. . . . . . 103

xii

Listings
4.1

Sample configuration file, 5-26-2009-1152AM-v-config.txt.

xiii

. . . . . . . .

58

Chapter 1
Introduction
This thesis presents a self-structuring two-port network that is capable of being adjusted to
perform as various two-port devices including matching networks, filters, and attenuators.
Such a network is based upon the principles of a self-structuring antenna.
The primary goal of this thesis is to explore the capabilities of a self-structuring twoport network. This thesis chronicles the work carried out by the author while exploring
and experimenting with a manufactured self-structuring two-port network. The research
was carried out with radio frequency (RF) applications in mind but applications certainly
go beyond just transmitters and antennas. Perhaps the application furthest from RF is for
power factor correction in power distribution systems. Within the realm of RF applications,
this work has implications in numerous areas. Currently there is a very strong interest in
reconfigurable filters, especially using microelectromechanical systems (MEMS) technology.
A MEMS implementation of a self-structuring two-port network is desirable due to the
demand for small, multi-frequency radio systems in cellphones and other devices. Section 2.2
further discusses the current state of research in the MEMS field further. Besides a MEMS
implementation, a self-structuring two-port network is desirable in situations when an RF
system is working under changing conditions. The ability to adjust to these changes without
a technician present is very advantageous. The author’s work in this thesis serves as a

1

investigation into the operation and capabilities of a self-structuring two-port network while
lying the foundation for others to implement application specific designs.
Chapter 2 of this thesis presents a more thorough examination of self-structuring antennas
and their influence on this work followed by a discussion of two-port devices and a brief look
at reconfigurable two-port devices. Finally, the self-structuring two-port network is presented
for the reader.
Chapter 3 contains results from simulations of the network. Prior work performed by
a visiting scholar who first investigated the network is discussed, followed by results from
further simulations. Much of the work done in the new simulations focuses on system losses.
Experimental results are presented and discussed in Chapter 4. First the prototype
unit is fully presented, followed by results from a random search of possible states. Next,
results from genetic algorithm based searches are presented. The chapter concludes with
experiments investigating system losses.
A conclusion and topics for future work are presented in Chapter 5.
An appendix is provided that describes how calibration kits may be downloaded to the
HP 8753D network analyzer and how new kits may be defined.
Throughout this thesis, certain types of figures are used repeatedly. One such figure is a
plot of S parameters. A two-port network has four S parameters: S11 , S21 , S12 , S22 . Because
of their physical meaning, these are often paired as S11 and S22 , and S12 and S21 . Sxx
parameters describe how the complex voltage reflected back at a port while Sxy parameters
describe how the complex voltage transmitted from port y to port x. S parameters are
most often plotted following these pairings in dB and degrees. Finally, simulations results
typically include a plot of efficiency versus frequency. Efficiency in this case is defined based
on power radiated over power delivered to the antenna. In this case, the antenna is the
two-port network.

2

Chapter 2
Background and Literature Review
2.1

Self-Structuring Antennas and Systems

A self-structuring antenna (SSA) is a type of antenna that is able to adapt to changes in the
environment as well as to change performance characteristics such as radiation pattern or
input impedance. An SSA is made up of wire segments or patches (collectively known as an
antenna template) with numerous switches placed in between the segments, a sensor, and a
controller with a search algorithm [1, 2]. Figure 2.1 shows a typical self-structuring antenna.
Switches are used to electrically connect or disconnect pieces of the antenna template,
thereby changing properties of the antenna. In the context of an SSA, a state is defined as
a particular permutation of opened and closed switches.
Self-structuring antennas have been investigated in numerous conference papers, journal
articles, and theses [1–16]. A thorough review of SSAs as an antenna class and an exploration
of possible search techniques are presented in [4, 6] (a review of the somewhat related topic of
frequency-reconfigurable antennas is presented in [17]). Topics of particular interest to this
thesis are self-structuring patch antennas [11] and automatically tunable cavity resonator
systems [15].

3

Figure 2.1: Schematic diagram of a self-structuring antenna.

SSA Template

Control
Unit

Transceiver

4

2.1.1

Self-Structuring Patch Antenna

It has been shown in previous research that shorting posts may be used to adjust the operating frequency of a patch antenna [18–21], allow multi-frequency operation [20, 22, 23], adjust
the input impedance [23] and/or reduce the size of the patch [20, 24–28]. Authors utilize
various approaches to the analysis of these effects. Cavity model theory was one of the first
techniques used [22], followed by an extension of nonuniform transmission line theory [19].
Finally analytic methods were dropped due to complexity while computational methods increased in popularity [24, 29]. Effectively the shorting posts are used to adjust the fields in
the substrate between the patch and the ground plane. Multiple posts allow more degrees
of freedom in how the fields are effected and thereby more degrees of freedom to control the
behavior of the antenna [25].
A self-structuring patch antenna is presented in [10, 11]. This patch uses an array of
switch-controlled shorting posts to control the performance and characteristics of the patch
antenna. All posts are permanently connected to the surface of the patch. When a switch
is opened, the shorting post is not connected to the ground plane. When a switch is closed,
the shorting post is connected to the ground plane through the switch, effectively behaving
like a shorting post used in other research [18–28] to allow, for example, adjustment of the
operating frequency or multi-frequency operation.
Figure 4.4 in [11] shows the shorting pin soldered to a disk pad isolated from the ground
plane by an annular slot. Figure 2.2 shows a similar setup used in the self-structuring
two-port network with the addition of a perimeter shorting post.
As in other patches that utilize shorting posts, the cavity fields are perturbed by the posts,
especially when the posts are connected to the ground plane. Unlike most other patches [18–
31], the SSA patch has a large number of shorting posts, 32 posts in [14]. Typically other
patches have shorting posts that are placed after careful analysis and design in order to effect
the fields in a certain manner; see [21]. The SSA patch in [14] has a large number of posts
which results in an extremely large number of switch states, i.e. 232 states. This makes
5

Figure 2.2: Photograph of shorting pin soldered to a disk pad isolated from the ground plane
by an annular slot. The switch, control wires and a perimeter shorting post are also seen.
For interpretation of the references to color in this and all other figures, the reader is referred
to the electronic version of this thesis.

6

analysis impractical, so posts are placed based on patterns, symmetry (or antisymmetry),
experience, and a limited analysis of certain states.

2.1.2

Automatically Tunable Cavity Resonator System

The extent to which a cavity may be tuned using SSA principles was investigated in [15]. In
this paper twenty-six “L” shaped wires with segments of various lengths and orientations were
placed inside of a cavity extending away from one wall. These were then connected to the
cavity wall through switches. Figure 1 of [15] shows a schematic drawing of the automatically
tunable cavity resonator system and Figure 2.3 shows the prototype resonator placed inside
of an anechoic chamber for testing. It was shown that it is possible to change the resonant
frequency of the cavity by selecting different states without changing the physical dimensions
of the cavity or by using other existing tuning methods. It was also shown that the cavity
could be retuned to the original resonant frequency if the cavity fields were perturbed by an
object that was placed into the cavity.
The combination of the SSA patch and tunable cavity resonator concepts points towards
the idea of creating a general self-structuring two-port network using on a copper clad substrate.

2.2

Two-Port Networks

Typical two-port devices include matching networks, filters, attenuators, and phase shifters.
There has been extensive research done on making these devices reconfigurable. One of the
main driving forces behind this research has been the need for mobile devices to operate
at multiple frequencies. Early two-port devices used varactors and potentiometers [32],
whereas recently microelectromechanical systems (MEMS) have been heavily used [33–40].
Two commonly researched devices are the matching network and the filter.
Matching networks are used to transfer the maximum power possible when an impedance

7

Figure 2.3: Automatically tunable cavity resonator system placed inside of an anechoic
chamber for testing.

8

mismatch exists. Maximum power is transferred when a port presents an impedance that is
the complex conjugate of the load attached to said port.
A matching network may take various forms. When impedances are known, resistors,
capacitors, and inductors may be used provided that the frequency is within the operating
range of these devices. Another option is to use transmission lines to transform an impedance
or to use tuning stubs attached to another transmission line. Both discrete components and
transmission lines have been implemented in adjustable value configurations, e.g. varactors
and sliding loads. This allows for tunable matching networks to be created [34–38, 41–43].
These are commonly found in applications where loads are often changed or effected by
outside factors, or when the input impedance of an antenna varies significantly across an
operating band.
Currently most MEMS based matching networks use MEMS switches to place tuning
stubs or capacitors in shunt with a transmission structure, usually a coplanar waveguide
or slow wave structure [33–39, 41–47]. Some devices use other MEMS elements such as
varactors.
Filters are typically adjusted in a similar manner as above when used at the microstrip or
smaller scale. Various valued components are switched into and out of the circuit to achieve
the desired results [32, 40, 48–56]. Often the type of filter (low-pass, high-pass, or band-pass)
is not able to be changed; instead, the center frequency, bandwidth, and shape of the filter
are the adjustable parameters. Recent work is beginning to make switching between types
of filters possible without discretely building each type of filter [54–56].
Before MEMS came into widespread use, a common approach was to use a dual-mode
microstrip filter [48]. A band-pass filter is demonstrated in [48] that utilizes the degenerate
modes found in a dual-mode microstrip to tune the filter. Figure 1 of [48] shows the layout
of the filter. This filter uses a meandered loop with equivalent sides to form a resonator.
The key to the filter is the small square of metal added to the top-right, inside corner of the
resonator as shown in the figure. By adjusting the size of the added metal, i.e. adjusting d,

9

the amount of coupling between degenerate modes may be varied. This leads to variations
in mode splitting which creates the band-pass filter. The filter presented in [48] makes use of
modal interaction which may play a role in the operation of self-structuring patch antennas
and self-structuring two-port networks. A downside to this filter, however, is that it is not
reconfigurable.
The work presented in [56] allows for one structure to be reconfigured and tuned as
either a bandstop or band-pass filter. The circuit in [56] makes use of mutual coupling
between two cavities and the impedance transformation properties of transmission lines. By
asymmetrically tuning the cavities, the coupling can effectively be changed to switch the
filter from a bandstop to band-pass filter and vice-versa. Tuning of a cavity is achieved
by mechanically deforming the cavity using a piezoelectric actuator. An exploded view of
the filter is shown in Figure 7 of [56]. This paper along with [53] are two of the few
papers to utilize cavity perturbation to affect the response of a filter at this scale or smaller.
Cavity perturbation is used in a self-structuring two-port network to effect transmission
characteristics. A self-structuring two-port network is advantageous over the filters presented
in [53, 56] as the network could be switched to other devices as needed whereas the filters
in [53, 56] will always be filters.

2.3

Self-Structuring Two-Port Network

A self-structuring two-port network is based on an extension of SSA principles. The goal is
to use the reconfigurable nature of this two-port to create various types of two-port devices.
Using a design similar to an SSA patch antenna, the two-port adds a second port to the
patch structure and permanently connected shorting posts around the perimeter . Using
thirty-two switched controlled posts, there are 232 or 4,294,967,296 possible states.
2.4 shows a prototype self-structuring two-port network. On the right is port one, on the
left is port two, and at the top is the ribbon cable used to control the switches. Switch control

10

Figure 2.4: Prototype self-structuring two-port network.

and shorting post operation is the same as in the self-structuring patch antenna prototype.
The switches can be seen laid out on the top of the network with control cables running to
the middle two pins of each switch. One side of the switch is connected to the top surface
of the network while the other side of the switch connects to a shorting post. The shorting
post is soldered to a piece of the top surface which has a ring of copper removed from around
it. Finally, perimeter shorting posts can be seen along the perimeter of the network. These
are used to keep power from radiating out of the sides, thereby acting like an antenna.

2.4

Power Balance

When a matching network presents a matched load at all ports, the maximum power will
be transferred to or from the matching network. What is not taken into account in this
description is how much power is lost in the network, known as the insertion loss of the
network. It is possible that given a matching network that is too lossy, less power would be

11

delivered to the load than when no matching network is used. In order to determine how
much power is being lost in a matching network, the power balance of the network should
be examined. Power balance is a way to express the percentage of input power lost by the
network.
Networks may be active, lossless or lossy with regard to power balance. Insertion loss
quantifies the difference between power received at the load and the source power. It is
expressed as a positive number when power is absorbed.
Take an and bn as the amplitudes of the waves entering and exiting port n of a microwave
network, respectively. The average power entering a port is given by Pi = 1 an a∗ and the
n
2
average power leaving a port is given by Po = 1 bn b∗ . If power is applied to port 1, then the
n
2
average source power is the sum of power entering and leaving, i.e. reflected from, port 1.
Regarded input power as positive, then the power leaving port 1 is a negative quantity. The
average source power is then given by
1
Psrc = (a1 a∗ − b1 b∗ ).
1
1
2

(2.1)

According to the conservation of energy, power entering a system must equal the sum of
power lost in and power leaving the system,

Psrc = Pout + Plost .

(2.2)

1
For a two-port network with power applied to port 1, Pout = 2 b2 b∗ . Substituting this
2
expression into Equation 2.2 and dividing by Pin = 1 a1 a∗ gives
1
2
1 (a a∗ − b b∗ )
1 b b∗
P
1 1
2
2 1 1
= 2 2 + lost .
1 a a∗
1 a a∗
Pin
2 1 1
2 1 1

12

(2.3)

S-parameters are defined as

Smn =

bm
an a =0, k=n
k

(2.4)

where m and n may be equal. Equation 2.3 can therefore be written as
P
∗
∗
1 − S11 S11 = S21 S21 + lost .
Pin

(2.5)

Equation 2.5 is called the power balance equation and is generally given as an inequality [57]
∗
∗
1 − S11 S11 ≥ S21 S21 .

(2.6)

S21 S∗ + S11 S∗ = 1.
21
11

(2.7)

If a system is lossless, then

An active network, such as an amplifier, may appear to be lossless or have gain, i.e. the
inequality in (2.6) may be switched to a less than or equal to sign. This is due to an external
power source that is not typically included in the power balance equation.
In this work the term power balance is used to describe the term −Plost /Pin in the
∗
∗
equation −Plost /Pin = S11 S11 + S21 S21 − 1. In some cases, it may be expressed as
a percentage given by −(−Plost /Pin ) × 100%. Throughout this thesis, power balance is
calculated as |S11 |2 + |S21 |2 − 1 or equivalently S11 S∗ + S21 S∗ − 1. The exact definition
11
21
used is typically included in the label of the ordinate axis. Using this definition, a net power
loss results in a value between minus one and zero while a net power gain results in a value
greater than zero.

13

Chapter 3
Simulations of a Self-Structuring
Two-Port Network
In 2007, Chien-Hsien (Jay) Lee was a visiting scholar from National Taiwan University in
the Electromagnetics Research Group at Michigan State University (MSU). As his research
project, Lee began studying the self-structuring two-port network. He designed a circuit
based on Lynn Greetis’ self-structuring patch antenna [10, 11] and conducted simulations
in order to find suitable post placements and evaluate the performance of the network as
a filter. Acceptable results from simulations led to the construction of a prototype unit.
Experimental results showed high system losses, hence further simulations were undertaken
by the author to evaluate the performance of the prototype.

3.1

Initial Investigation and Simulations

Simulations of the two-port network were carried out using FEKO Suite 5.2, a commercial
method of moments EM solver provided by EM Software and Systems [58]. The genetic
algorithm (GA) optimizer GA-FEKO [59] was used to optimization switch states.
A design similar to a patch SSA was first simulated. In order to keep the network from
radiating, shorting posts were placed around the perimeter of the network. These posts
14

essentially create the side walls of a cavity. Lee then investigated appropriate board size
and thickness, various methods of feeding the network, and the placement of the switchcontrolled shorting posts. An infinite ground plane was used so that the half-space Green’s
function could be used thereby reducing computational complexity. This made the use of a
genetic algorithm for optimization practical.
A major goal of Lee’s research was to learn how the frequency response of the network
could be adjusted or tuned. This mainly took the form of searching for various types of
filter responses. Searches were performed using either random searches or genetic algorithm
based searches. Figure 3.1 shows the response of the circuit configured as a band pass filter,
found using a GA, centered at f0 =5.005 GHz. This center frequency may be moved or tuned
across a range of frequencies by changing the switch state, as demonstrated by Figure 3.2,
over the band 4.90 GHz to 5.05 GHz.

3.2

Power Balance Simulations

As discussed in Section 2.4, the experimental setup exhibited an unexpectantly high power
loss, even after various mitigation techniques were applied. A new series of simulations was
carried out in an attempt to better understand the loss mechanisms involved in this project.
These simulations were run in a newer version of the FEKO Suite than previously empoyed
and used a finite ground plane. As the objective was to investigate power losses in the experimental system, the physical and material properties from the constructed self-structuring
two-port were used, see Section 4.1.1 for specific values. The increased computational complexity of this model was acceptable because optimization was unnecessary. While short
solution times were not required, extremely long solution times were undesired as well. As
such, the medium density mesh included in CAD FEKO was used.

15

Figure 3.1: GA optimized band pass filter centered at f0 =5.005 GHz.

16

Figure 3.2: Demonstration of the ability to tune the center frequency of a bandpass filter
over the frequency range of 4.90 GHz to 5.05 GHz.

17

3.2.1

FEKO Suite 6.1 Upgrade

Previous simulations were carried out in FEKO Suite 5.2. While learning FEKO, the author
noticed that when simulating cavities and patch antennas, antenna efficiency was larger
than one hundred percent. This was observed in multiple simulations including the patch
antenna example included with FEKO Suite 5.2. This example, when run in FEKO Suite
6.1, provided acceptable results. It is suspected that something in the configuration of the
solver used in version 5.2 was changed since the time that Lee ran the GA optimization
simulations.
The author visited the EM Software and Systems booth at the 2011 IEEE APS/URSI
conference in Spokane, WA. After discussing this problem with them, FEKO Suite 6.1 and
a trial license were obtained. Besides changes in the computational engine, other changes
included a redesign of CAD FEKO, the GUI based model input program. This allowed for
the creation of models and simulations without the direct editing of a card file. It should be
noted that card files may still be used in the new version.

3.2.2

List of Simulations

Below is a brief description of the simulations run to investigate power losses in the experimental setup. All simulations were calculated at twenty-six frequencies between 700 MHz
and 1.5 GHz (32 MHz step) except in post-cavity-lossy-15 as noted. TLY5 is the name
of the Taconic circuit board used to construct the experimental network, see Section 4.1.1.
• free-cavity - PEC cavity with a free space interior
• lossless-cavity - PEC cavity with a lossless TLY5 interior
• lossy-cavity - Copper cavity with a lossy TLY5 interior
• no-sides-lossless-cavity - PEC top and bottom with a lossless TLY5 dielectric
• no-sides-lossy-cavity - Copper top and bottom with a lossy TLY5 dielectric
18

• post-cavity-lossless - PEC top and bottom with a lossless TLY5 dielectric and
perimeter shorting posts
• post-cavity-lossless-walls - PEC cavity with a lossless TLY5 interior and perimeter shorting posts
• post-cavity-lossy - Copper top and bottom with a lossy TLY5 dielectric and perimeter shorting posts
• post-cavity-lossy-15 - Copper top and bottom with a lossy TLY5 dielectric and
perimeter shorting posts at thirteen frequency points between 1.5 GHz and 1.884 GHz
(32 MHz steps)
• post-cavity-lossy-walls - Copper cavity with a lossy TLY5 dielectric and perimeter
shorting posts
• rings-reduced - Copper cavity with a lossy TLY5 dielectric, perimeter shorting posts,
and sixteen switched shorting posts

3.2.3

Cavity Simulations

Three cavities were simulated to investigate the properties of a cavity having the same
dimensions (42 x 27 x 0.5 cm) as the self-structuring two port network. All cavities had
two ports placed in the same locations as on the constructed two port network. The first
cavity simulated, free-cavity, consists of PEC walls and a free space interior. Results from
this simulation are shown in Figures 3.3 and 3.4. One can see that the cavity demonstrates
reciprocity as do all simulations. Seen in the plots for efficiency and power balance are nonrealistic values (negative efficiencies and net power gain). These values, or errors, are on the
order of hundredths of a percent. These errors are most likely numerical artifacts because
of the relatively small active power. Even with these errors, it is easy to see that power is
maintained in the system as expected.
19

Figure 3.3: Free cavity report.

S11

S12

S21

S22

free−cavity.dat

150
S Parameter Phase (deg)

S Parameter Magnitude (dB)

0
−5
−10
−15
−20
−25

100
50
0
−50
−100
−150

−30

0.8
1
1.2
1.4
Frequency (GHz)

Active Power

0.8
1
1.2
1.4
Frequency (GHz)

Loss Power

0.02

0.2
0.15
0.1
Efficiency (%)

Power (W)

0.015

0.01

0.005

0.05
0
−0.05
−0.1
−0.15

0

−0.2

0.8
1
1.2
1.4
Frequency (GHz)

20

0.8
1
1.2
1.4
Frequency (GHz)

Figure 3.4: Free cavity power balance report.
−4

8

free−cavity.dat

x 10

Power Balance = |S11|2+|S21|2−1

6
4
2
0
−2
−4
−6
−8
−10
0.7

0.8

0.9

1
1.1
1.2
Frequency (GHz)

21

1.3

1.4

1.5

The second cavity, lossless-cavity, had PEC walls and a lossless TLY5 interior dielectric with a relative permittivity of r = 2.2. Results from this simulation are shown in
Figures 3.5 and 3.6. The graph of antenna efficiency shows that no power is lost due to
radiation. Figure 3.6 shows that no power is lost. Again though, there are errors in the
calculations as the power balance is larger than 0, representing a net power gain.
The third cavity, lossy-cavity, was designed to have system losses. It consisted of
17.78 µm thick, i.e. 1/2 oz, Copper walls on all six sides and a lossy TLY5 interior dielectric
with a relative permittivity of r = 2.2 and a loss tangent of δ = 0.0010. Results from this
simulation are shown in Figures 3.7 and 3.8. Here one is able to see that the simulations
exhibit no unrealistic values. This improvement can most likely be attributed to the added
loss. A technique used early on in computational electromagnetics was to add small amounts
of loss in order to account for slight divergences in the solution. It is possible that the two
previous simulations were causing problems for the FEKO solver, for which the loss helped in
this case. This is plausible because the FEKO solver in FEKO Suite 5.2 had significant error
for the same and simpler cavity simulations. Very little power was lost in this simulation
due to radiation; however, at some frequencies up to twenty-five percent of the power was
lost. While it is expected that some power would be lost, this amount is more than expected.
Examination of the output files revealed that most of the power was lost in the dielectric
and metal, but a primary loss mechanism was not pinpointed.

3.2.4

Coplanar Simulations

In order to show how much power is lost through the sidewalls, the sidewalls were removed
from the cavities. This is important to evaluate in order to determine the effectiveness of
the perimeter shorting posts.
Only two simulations with TLY5 dielectric were used for this new set of simulations and
for the remaining simulations. The two TLY5 dielectric models from the previous section
were modified for use in these simulations. As mentioned above, the four side walls of each
22

Figure 3.5: Lossless cavity report.

S11

S12

S21

S22

lossless−cavity.dat

150
S Parameter Phase (deg)

S Parameter Magnitude (dB)

0
−5
−10
−15
−20
−25

100
50
0
−50
−100
−150

−30

0.8
1
1.2
1.4
Frequency (GHz)

Active Power

0.8
1
1.2
1.4
Frequency (GHz)

Loss Power

0.5
Efficiency (%)

1

0.015
Power (W)

0.02

0.01

0.005

0

0

−0.5

−1

0.8
1
1.2
1.4
Frequency (GHz)

23

0.8
1
1.2
1.4
Frequency (GHz)

Figure 3.6: Lossless cavity power balance report.

lossless−cavity.dat

−3

14

x 10

Power Balance = |S11|2+|S21|2−1

12
10
8
6
4
2
0
−2
0.7

0.8

0.9

1
1.1
1.2
Frequency (GHz)

24

1.3

1.4

1.5

Figure 3.7: Lossy cavity report.

S11

S12

S21

S22

lossy−cavity.dat

150
S Parameter Phase (deg)

S Parameter Magnitude (dB)

0
−5
−10
−15
−20
−25

100
50
0
−50
−100
−150

−30

0.8
1
1.2
1.4
Frequency (GHz)

Active Power

0.8
1
1.2
1.4
Frequency (GHz)

Loss Power

0.02

−11

3

x 10

2.5
Efficiency (%)

Power (W)

0.015

0.01

2
1.5
1

0.005
0.5
0

0

0.8
1
1.2
1.4
Frequency (GHz)

25

0.8
1
1.2
1.4
Frequency (GHz)

Figure 3.8: Lossy cavity power balance report.

lossy−cavity.dat
0

Power Balance = |S11|2+|S21|2−1

−0.05
−0.1
−0.15
−0.2
−0.25
−0.3
−0.35
0.7

0.8

0.9

1
1.1
1.2
Frequency (GHz)

26

1.3

1.4

cavity were removed to create a coplanar structure. All other parameters and geometries
remained unchanged.
Figures 3.9 and 3.10 show the results for the lossless dielectric simulation, no-sideslossless-cavity. Unlike the results presented in the last section, the active and loss power
are noticeably different. Looking at the plot of efficiency shows that at four frequencies,
over twenty percent of the delivered power is radiated, and at one of those four frequencies,
over sixty-five percent is radiated. Figure 3.10 shows that at almost all frequencies, over ten
percent of the power is lost.
The results from the lossy simulation, no-sides-lossy-cavity, are shown in Figure 3.11
and 3.12. While the results vary from the lossless case, the behavior is similar. As the power
loss is similar in both simulations, one can see that without walls, the primary loss mechanism
is radiation. This same observation was made by Lee and resulted in the two port network
having perimeter shorting posts.

3.2.5

Perimeter Post Simulations

The thirty-eight perimeter shorting posts were added to the lossless and lossy TLY5 cavity
and coplanar simulations. In addition, the lossy TLY coplanar simulation was run at frequencies above 1.5 GHz. Together, these simulations demonstrated the effectiveness of the
shorting posts and point towards loss mechanisms other than radiation.
Figures 3.13 and 3.14 show the results from a lossless TLY5 coplanar simulation with
perimeter shorting posts, post-cavity-lossless. It is easily seen that power is lost due to
radiation and other loss mechanisms. Figures 3.15 and 3.16 show the results from a lossless
TLY5 cavity simulation with perimeter shorting posts. In this simulation, one sees behavior
similar to that of the lossless-cavity simulation. There is essentially no radiated power
while the overall power balance is almost zero. Here, errors in the numerical solver cause
power generation.
Figures 3.17 through 3.20 show the results from a lossy TLY5 coplanar simulation with
27

Figure 3.9: no-sides-lossless-cavity report.

S11

S12

S21

S22

no−sides−lossless−cavity.dat

150

−5

S Parameter Phase (deg)

S Parameter Magnitude (dB)

0

−10
−15
−20
−25
−30
−35
−40

100
50
0
−50
−100
−150

0.8
1
1.2
1.4
Frequency (GHz)

Active Power

0.8
1
1.2
1.4
Frequency (GHz)

Loss Power

0.02

80
70
60
Efficiency (%)

Power (W)

0.015

0.01

0.005

50
40
30
20
10

0

0

0.8
1
1.2
1.4
Frequency (GHz)

28

0.8
1
1.2
1.4
Frequency (GHz)

Figure 3.10: no-sides-lossless-cavity power balance report.

no−sides−lossless−cavity.dat
0

Power Balance = |S11|2+|S21|2−1

−0.1
−0.2
−0.3
−0.4
−0.5
−0.6
−0.7
0.7

0.8

0.9

1
1.1
1.2
Frequency (GHz)

29

1.3

1.4

1.5

Figure 3.11: no-sides-lossy-cavity report.

S11

S12

S21

S22

no−sides−lossy−cavity.dat
150

S Parameter Phase (deg)

S Parameter Magnitude (dB)

0
−10
−20
−30
−40
−50

100
50
0
−50
−100
−150

−60

0.8
1
1.2
1.4
Frequency (GHz)
Active Power

0.8
1
1.2
1.4
Frequency (GHz)

Loss Power

0.02

60
50
Efficiency (%)

Power (W)

0.015

0.01

40
30
20

0.005
10
0

0

0.8
1
1.2
1.4
Frequency (GHz)

30

0.8
1
1.2
1.4
Frequency (GHz)

Figure 3.12: no-sides-lossy-cavity power balance report.

no−sides−lossy−cavity.dat
0

Power Balance = |S11|2+|S21|2−1

−0.1
−0.2
−0.3
−0.4
−0.5
−0.6
−0.7
0.7

0.8

0.9

1
1.1
1.2
Frequency (GHz)

31

1.3

1.4

1.5

Figure 3.13: post-cavity-lossless report.

S11

S12

S21

S22

post−cavity−lossless.dat

150
S Parameter Phase (deg)

S Parameter Magnitude (dB)

0
−5
−10
−15
−20
−25

100
50
0
−50
−100
−150

−30

0.8
1
1.2
1.4
Frequency (GHz)
Active Power

0.8
1
1.2
1.4
Frequency (GHz)

Loss Power

0.02

80
70
60
Efficiency (%)

Power (W)

0.015

0.01

0.005

50
40
30
20
10

0

0

0.8
1
1.2
1.4
Frequency (GHz)

32

0.8
1
1.2
1.4
Frequency (GHz)

Figure 3.14: post-cavity-lossless power balance report.

post−cavity−lossless.dat
0

Power Balance = |S11|2+|S21|2−1

−0.1
−0.2
−0.3
−0.4
−0.5
−0.6
−0.7
0.7

0.8

0.9

1
1.1
1.2
Frequency (GHz)

33

1.3

1.4

1.5

Figure 3.15: post-cavity-lossless-walls report.

S11

S12

S21

S22

post−cavity−lossless−walls.dat

150

10

S Parameter Phase (deg)

S Parameter Magnitude (dB)

20

0
−10
−20
−30
−40
−50
−60

100
50
0
−50
−100
−150

0.8
1
1.2
1.4
Frequency (GHz)

Active Power

0.8
1
1.2
1.4
Frequency (GHz)

Loss Power

0.5
Efficiency (%)

1

0.015
Power (W)

0.02

0.01

0.005

0

0

−0.5

−1

0.8
1
1.2
1.4
Frequency (GHz)

34

0.8
1
1.2
1.4
Frequency (GHz)

Figure 3.16: post-cavity-lossless-walls power balance report.

post−cavity−lossless−walls.dat
0.035

Power Balance = |S11|2+|S21|2−1

0.03
0.025
0.02
0.015
0.01
0.005
0
0.7

0.8

0.9

1
1.1
1.2
Frequency (GHz)

35

1.3

1.4

perimeter shorting posts, post-cavity-lossy. The first two figures cover the usual frequency range while the last two figures show an extended range from 1.5 GHz to 1.8 GHz.
As expected, the system exhibits loss from both radiation and other loss mechanisms. The
perimeter shorting posts appear to behave less like a wall as the distance between shorting
posts becomes significantly greater than a wavelength. A higher overall power loss is noted
suggesting that in general power loss increases with frequency. This behavior is no surprise
as this is typical with most materials.
Figures 3.21 and 3.22 show the results from a lossy TLY5 cavity simulation with perimeter shorting posts, post-cavity-lossy-walls. Due to the addition of walls to form a cavity,
the radiation efficiency has decreased to almost zero. Power loss is reduced but still present.
These results suggest that the perimeter shorting posts only effect power loss through radiation.

36

Figure 3.17: post-cavity-lossy report.

S11

S12

S21

S22

post−cavity−lossy.dat

150
S Parameter Phase (deg)

S Parameter Magnitude (dB)

0
−5
−10
−15
−20
−25

100
50
0
−50
−100
−150

−30

0.8
1
1.2
1.4
Frequency (GHz)

Active Power

0.8
1
1.2
1.4
Frequency (GHz)

Loss Power

0.02

60
50
Efficiency (%)

Power (W)

0.015

0.01

40
30
20

0.005
10
0

0

0.8
1
1.2
1.4
Frequency (GHz)

37

0.8
1
1.2
1.4
Frequency (GHz)

Figure 3.18: post-cavity-lossy power balance report.

post−cavity−lossy.dat
0

Power Balance = |S11|2+|S21|2−1

−0.1
−0.2
−0.3
−0.4
−0.5
−0.6
−0.7
0.7

0.8

0.9

1
1.1
1.2
Frequency (GHz)

38

1.3

1.4

1.5

Figure 3.19: post-cavity-lossy-15 report.

S11

S12

S21

S22

post−cavity−lossy−15.dat

150
S Parameter Phase (deg)

S Parameter Magnitude (dB)

0

−5

−10

−15

100
50
0
−50
−100
−150

−20
1.5

1.6
1.7
1.8
Frequency (GHz)
Active Power

1.5

1.6
1.7
1.8
Frequency (GHz)

Loss Power

0.02

60
50
Efficiency (%)

Power (W)

0.015

0.01

40
30
20

0.005
10
0
1.5

0
1.5

1.6
1.7
1.8
Frequency (GHz)

39

1.6
1.7
1.8
Frequency (GHz)

Figure 3.20: post-cavity-lossy-15 power balance report.

post−cavity−lossy−15.dat
−0.1

Power Balance = |S11|2+|S21|2−1

−0.15
−0.2
−0.25
−0.3
−0.35
−0.4
−0.45
−0.5
−0.55
−0.6
1.5

1.55

1.6

1.65
1.7
1.75
Frequency (GHz)

40

1.8

1.85

3.2.6

Switched Shorting Post Simulations

Finally a simulation, rings-reduced, was run that involved sixteen of the thirty-two switched
shorting posts. The posts and annular rings were simulated with no switch, effectively simulating an all open state. Material properties and geometry were the same as in post-cavitylossy. Only sixteen posts could be simulated due to computational costs. Figures 3.23
and 3.24 show the results from this simulation. These results suggest a decrease in power
loss as compared to prior simulations. Further, it demonstrates that, especially in comparison to radiated loss from the sides of a coplanar two port network, the annular rings
contribute very little to the power loss.

3.3

Conclusion

Various simulations were carried out to study the self-structuring two-port network. Initial
simulations consisted of lossless materials and an infinite ground plane. The goal was to
explore the capabilities of the network in a timely manner. These simulations also provided
guidance when it came time to manufacture a self-structuring two-port network. Results
from these initial simulations demonstrated the ability to create a filter with a tunable
center frequency.
New simulations were carried out after unsatisfactory performance of the manufactured
self-structuring two-port network during experiments. The goal of these simulations was
to determine sources of significant power loss. Actual physical dimensions and material
properties were used in an attempt to model the constructed network. Overall, no one
loss mechanism was identified as significant. Perimeter shorting posts and copper tape
around the edges was shown to be effective at reducing radiated power. Radiation from
the rings around the shorting posts was not significant. These simulations suggested that
while dielectric losses were present, they were not sufficient to cause the losses observed in
experiments.

41

Figure 3.21: post-cavity-lossy-walls report.

S11

S12

S21

S22

post−cavity−lossy−walls.dat

150
S Parameter Phase (deg)

S Parameter Magnitude (dB)

0

−5

−10

−15

100
50
0
−50
−100
−150

−20

0.8
1
1.2
1.4
Frequency (GHz)
Active Power

0.8
1
1.2
1.4
Frequency (GHz)

Loss Power
−11

0.02

4

x 10

3
Efficiency (%)

Power (W)

0.015

0.01

2
1
0

0.005
−1
0

−2

0.8
1
1.2
1.4
Frequency (GHz)

42

0.8
1
1.2
1.4
Frequency (GHz)

Figure 3.22: post-cavity-lossy-walls power balance report.

post−cavity−lossy−walls.dat
0

Power Balance = |S11|2+|S21|2−1

−0.02
−0.04
−0.06
−0.08
−0.1
−0.12
−0.14
−0.16
−0.18
0.7

0.8

0.9

1
1.1
1.2
Frequency (GHz)

43

1.3

1.4

Figure 3.23: rings-reduced report.

S11

S21

rings−reduced.dat
150

−5

S Parameter Phase (deg)

S Parameter Magnitude (dB)

0

−10
−15
−20
−25

100
50
0
−50
−100
−150

−30

0.8

1
1.2
1.4
Frequency (GHz)

Active Power

0.8

1
1.2
1.4
Frequency (GHz)

0.8

1
1.2
1.4
Frequency (GHz)

Loss Power

0.02

1.5

Efficiency (%)

Power (W)

0.015

0.01

1

0.5

0.005

0

0.8

0

1
1.2
1.4
Frequency (GHz)

44

Figure 3.24: rings-reduced power balance report.

rings−reduced.dat
0

Power Balance = |S11|2+|S21|2−1

−0.05
−0.1
−0.15
−0.2
−0.25
−0.3
−0.35
−0.4
0.7

0.8

0.9

1
1.1
1.2
Frequency (GHz)

45

1.3

1.4

Chapter 4
Experimental Investigation
A prototype self-structuring two-port network was constructed and used to experimentally
explore the network as a two-port device. The initial investigation was a random search of
a small percentage of the possible switch configurations. Genetic algorithm based searches
were then used to explore the characteristics of the network. Results from this stage of experimentation suggested a high-loss network. Mitigation was performed to reduce radiation
losses. Random and genetic algorithm searches were carried out post-mitigation. Continued
power loss points to an inefficient two-port network.

4.1

Hardware

Before leaving MSU, Chien-Hsien Lee manufactured the two-port network used in the experiments for this thesis. Its construction is described below. A photograph of the completed
prototype is shown in Figure 2.4, reprinted here as Figure 4.1 for the reader’s convienence.

4.1.1

Two-Port Network

The self-structuring two-port network was constructed on a 42 x 27 x 0.5 cm Taconic TLY-5
double-sided circuit board (relative permittivity, r , of 2.20, dissipation factor, δt , of 0.0010).

46

Figure 4.1: Prototype self-structuring two-port network. Reprint of Figure 2.4.

A copy of the letter stating the parameters of the board is shown in Figure 4.2.
Figure 4.3 is a schematic of the network. A 36 x 21 cm rectangle centered on the board
was subdivided into a 3 x 3 cm grid. Thirty-eight perimeter shorting posts were placed
around the outside of this rectangle at grid intersections. Throughout the rectangle, thirtytwo switch controlled shorting posts were placed on the grid along with two SMA female
panel mount connectors. Using the perimeter post in the bottom left corner of Figure 4.1 as
the origin, Table 4.1 gives the coordinates of each post and port.

47

Figure 4.2: Material parameters of the circuit board used to create the self-structuring twoport network prototype provided by Taconic.

48

Figure 4.3: Schematic view of self-structuring two-port.

49

Table 4.1: Pin and port placement taking the bottom left perimeter pin as the origin.
Element
Bottom, Left Corner
of Board
Perimeter Post 1
Perimeter Post 2
Perimeter Post 3
Perimeter Post 4
Perimeter Post 5
Perimeter Post 6
Perimeter Post 7
Perimeter Post 8
Perimeter Post 9
Perimeter Post 10
Perimeter Post 11
Perimeter Post 12
Perimeter Post 13
Perimeter Post 14
Perimeter Post 15
Perimeter Post 16
Perimeter Post 17
Perimeter Post 18
Perimeter Post 19
Perimeter Post 20
Perimeter Post 21
Perimeter Post 22
Perimeter Post 23
Perimeter Post 24
Perimeter Post 25
Perimeter Post 26
Perimeter Post 27
Perimeter Post 28
Perimeter Post 29
Perimeter Post 30
Perimeter Post 31
Perimeter Post 32
Perimeter Post 33
Perimeter Post 34
Perimeter Post 35
Perimeter Post 36
Perimeter Post 37
Perimeter Post 38

x (cm) y (cm)
-3

0
3
6
9
12
15
18
21
21
21
21
21
21
21
21
21
21
21
21
21
18
15
12
9
6
3
0
0
0
0
0
0
0
0
0
0
0
0

x (cm) y (cm)

-3

0
0
0
0
0
0
0
0
3
6
9
12
15
18
21
24
27
30
33
36
36
36
36
36
36
36
36
33
30
27
24
21
18
15
12
9
6
3

Element

Port 1
Port 2
Switched Post 1
Switched Post 2
Switched Post 3
Switched Post 4
Switched Post 5
Switched Post 6
Switched Post 7
Switched Post 8
Switched Post 9
Switched Post 10
Switched Post 11
Switched Post 12
Switched Post 13
Switched Post 14
Switched Post 15
Switched Post 16
Switched Post 17
Switched Post 18
Switched Post 19
Switched Post 20
Switched Post 21
Switched Post 22
Switched Post 23
Switched Post 24
Switched Post 25
Switched Post 26
Switched Post 27
Switched Post 28
Switched Post 29
Switched Post 30
Switched Post 31
Switched Post 32

50

33
3
33
30
30
27
27
27
27
27
24
24
25
21
21
21
18
18
3
6
6
9
9
9
9
9
12
12
12
15
15
15
18
18

12
9
3
3
9
3
9
12
15
18
6
12
18
6
12
18
15
18
18
18
12
18
12
9
6
3
15
26
3
15
9
3
6
3

Each port had a 0.127 cm diameter hole drilled for it. All posts, perimeter and switchcontrolled, had 0.1 cm diameter holes drilled for them. An annular ring of copper with
an inner diameter of 1.4 cm and width of 0.1 cm was removed from around each switchcontrolled shorting post on the top surface of the device. This was done in order to isolate
the post from the rest of the surface.
Coto 5 V reed relay switches were used to control the switched posts. Switches were
affixed to the network using clear tape. Typical time required to mechanically close a switch
is 350 µs and 100 µs to open [60]. The two inside pins on each relay are the control pins
and the two outside pins are the switched pins. Switches are not polarized. Each pad
and associated switch are covered with electrical tape. Control wires are potted about
one inch from the switch. Dow Corning 748 noncorrosive silicone sealant (McMaster-Carr
#74915A654) was used as the potting compound. The entire top side of the network is
covered with copper tape in an effort to minimize radiation. See Section 4.4.4 for more
details.
Placed on the bottom side of the circuit in each corner is a rubber foot. For most
experiments, the network was kept in a shallow cardboard box lid. This was done for ease of
transportation, setup, and storage. An extra layer of insulation was included for protection
from test equipment and surfaces.
Switch control is accomplished through a National Instruments PCI-DIO-96 digital input/output (DIO) board. The board has ninety-six DIO spread connections across two
50-pin ribbon cables. Pins 49 and 50 on each cable are +5 VDC and ground, respectively.
Each DIO pin can supply 2.5 mA; however, each switch requires 250 mA. To supply this
current, Toshiba TD62783APG source driver chips are used. Each chip is capable of simultaneously driving eight outputs. The cable carrying pins 1 – 50 is connected to a circuit
board which has an external power supply and four source driver chips. The outputs of these
drivers are connected to the network through a 4.5 in long, 64-pin ribbon cable which carries
a signal and ground for each switch. This short cable is split in half with a ferrite on each

51

half and then one large ferrite around the entire cable. The ribbon cable’s connector is glued
to the edge of the network, female side up so that the control wires for each switch can be
plugged into the connector as seen in the top of Figure 4.1.
Mechanical problems often arise in the connection between the ribbon cable and the
switches. The most common problems are the following: one or both control wires for a
switch may come out of the connector, the control pins on the switches may break, or the
wires need re-soldered to the switches. Often, attempting to fix one problem will cause
another one of these problems. These mechanical problems have been observed by the
author with all SSA prototypes at MSU that utilize ribbon cable control and/or discrete
Coto switches. Another common problem exhibited by most prototypes is a relatively short
lifetime for the ribbon cable connectors, both male and female. Often times, the plastic
side walls of the male connectors will break and the pins will become bent after repeated
connections. This is due to the difficulty in connecting and disconnecting ribbon cables.
Female connectors are difficult to assembly and crimp properly leading to the connector not
staying together. Caution must be exercised when working with ribbon cables.
Future investigations should consider using different connectors and control methods to
reduce mechanical failures in experiments.
Mechanical problems on the network board were mitigated in two ways. First, control
wires were mechanically potted near the switches to provide strain relief and to protect
the switch pins. Second, control wires were hot glued to the ribbon cable connector after
verifying that the switch was operational. The switches were checked again after all control
wires were hot glued.
A few words regarding the Dow Corning sealant used as the potting compound are offered.
Setting time is one problem; working time is around 1 hr, with complete curing occurring
in approximately 24 hrs. This relatively lengthy time means that whatever is being potted
must be able to support itself or have a support system during this period. Hot glue, on the
other hand, requires only a few seconds or tens of seconds of support. The downside of hot

52

glue is that it does not adhere to copper tape well. The Dow Corning sealant was selected
for multiple reasons including being non-corrosive, non-conductive, giving off few fumes, its
ability to bond to all materials of interest, and having a relatively low cost-to-volume ratio.
There are multiple methods used to check for proper switch operation. The simplest
check is to listen for the switches to operate. The next check should be performed routinely,
especially after an experiment has been moved or any control cables adjusted. Each switch
is turned on (either individually or all at once) and the output voltage measured. When
operating properly, the voltage should be between 3.50 V and 3.75 V, optimally 3.64 V for
the constructed control board and network. If the output is 5 V, there is most likely an
open circuit somewhere. The first place to check is the connection between the ribbon cable
and the switch control wires. Then the switch pins should be checked. If the output is 0 V,
disconnect DC power and look for a short. Leaving power applied may cause the driver to
heat up considerably. This leads to another testing procedure which is to check the drivers
for high heat. When functioning properly, the drivers should not feel warm.
Over the past three years of working with the Toshiba driver chips, there has only been
one time that a chip quit functioning. In this case the defective output would always be at
5 VDC with no current to operate the switch. It was determined to be malfunctioning after
switching the orientation of the output ribbon cable ( i.e. pin 32 became pin 1) and also
after switching SSA devices.

4.1.2

HP 8753D Network Analyzer

Measurements were taken using an HP 8753D network analyzer controlled by a National
Instruments USB-to-GPIB converter (product name: NI GPIB-USB-HS). The analyzer was
calibrated using an HP 85052D Economy Mechanical Calibration Kit (DC to 26.5 GHz,
3.5 mm) for most experiments. The calibration table for this calibration kit is not readily
available in a format usable by the HP 8753D. Using VNA Cal Kit Manager [61], the calibration table for the HP 8510C was edited and then downloaded to the network analyzer.
53

Appendix 5.3 contains the procedures for editing and downloading the cal kit table to the
HP 8753D.
When adapters of various types were used in the same experiment, such as when connecting to a large horn antenna having an N-type connector, the analyzer was calibrated using
the 3.5 mm cal kit and then an SMA-to-N type connector attached to the port requiring it.
The effect of this adapter is assumed to be negligible. Using existing cal kit definitions and
VNA Cal Kit Manager, it is possible to create a cal kit for an SMA connection on one cable
and an N type connection on the other cable.

4.2

Selection of States

States for random testing are determined from a state file, not a dynamic random number
generator. The state file is parsed sequentially for as many states as needed. The file used
throughout the research for this thesis consisted of one hundred thousand unique states that
were generated by a National Institute of Standards and Technology (NIST) random number
generator. The file begins with states for all switches open, all closed, and each switch closed
individually and then has the one hundred thousand random states. Overall, there are one
hundred thousand thirty four states in the file, stored as a text file with each state on a
line. Each state is encoded as a thirty-two character long string of 0’s and 1’s. States are
read like a binary number with switch 32 corresponding to the left most bit and switch 1
corresponding to the right most bit.
A file of random states is used instead of a random number generator for multiple reasons.
First, it lends repeatability to experiments. While each experiment is not random when
compared to others, the states used are random. Second, using a NIST generated list ensures
a statistically random sample. Third, a set file reduces programming complexity as states
do not have to be generated. Finally one is able to use the random state file as a means to
associate results with states since the random state file is static. The random state file has

54

only three repeated states: 0x5D124738, 0xCFBF5D29, and 0xF013B91A. Repeated states can
be found using the terminal command sort file-name | uniq -d.

4.3

LabView Software

Data accquisition and experiment control were carried out using software written in National
Instruments’ (NI) LabView environment [62]. LabView allows for easy interface between NI
hardware and many other devices using various protocols. The ECE shop at MSU has
developed numerous LabView applications for use in the teaching labs. Because of this and
the availability of an NI USB-to-GPIB adapter, LabView was selected for this project.
LabView is programmed by creating a block diagram of the program. Blocks essentially
represent functions with data being passed between blocks by wires. This design philosophy
encourages programs that use a parallel flow instead of a sequential flow.
LabView provides many ready-to-use blocks for both common and uncommon tasks,
e.g. plotting. Program development in LabView requires a change in mindset and forces
the programmer to change his normal programming style. In addition, proper layout of a
program in LabView can be difficult to achieve as the “automatic cleanup” feature often
makes the program more difficult to follow. It is often difficult to get an overview of a large
LabView program because LabView programs are represented as having three dimensions
(height, width, and depth) where in a traditional program such as FORTRAN, C++, or
Matlab the program really only has a length. The depth of a LabView program is often
displayed in structures such as conditional loops. Only one condition can be displayed at
a time. As is common in almost all languages, data types can be confusing at first. While
attempting to remove this confusion, LabView has created its own problems with data types.
LabView’s strongest attribute is the large, local library of functions available to the user
that in other languages would have to be written by the programmer or sought out in a remote
code library. Interfacing with various hardware devices is somewhat simplified in LabView.

55

Also fairly straightforward is data import and export. Mathematical processing of data
can be cumbersome, however, as anything beyond simple arithmetic requires a complicated
implementation.
Another strength of LabView is the debugging process. LabView provides many ways to
track data through a program, provides breakpoints, and very thorough “syntax” checking.
The LabView integrated development environment (IDE) gives the user immediate feedback
on type mismatches and improperly connected wires.
While the IDE provides helpful debugging features, these features are probably the only
helpful thing about the IDE. There is no way to zoom in or out on a block diagram in
order to get an overview of the program. Rearranging the block diagram via cut-and-paste
or drag-and-drop is implemented, but functionally does not work. Insertion of blank space
for visual reasons or for development reasons is extremely difficult. Formatting or layout
of the block diagram for ease of reading, programming, and understanding is difficult if not
properly done from the beginning.
LabView has great potential; however, programming anything beyond a very basic application involves many hurdles, most notably the IDE itself. Future research should consider
using Matlab or another programming language.
The newly acquired Agilent ENA E5071C has Microsoft Visual Basic installed natively.
Future data acquisition and automation will most likely be carried out in Visual Basic.
The various programs created using LabView to run experiments are described below.

4.3.1

Random Search

The first program created performs a random search experiment. Using the NIST generated
random number pool, the program sets a switch state and then records the measured Sparameters before setting the next state.
To begin, the user configures the network analyzer through the software. Parameters
that may be adjusted include start and stop frequency, number of points, averaging, IF
56

bandwidth, power level, communications time-out, and the GPIB address of the network
analyzer.
Next the user selects the NIST random pool file to be used and the folder to which data
should be saved. The user then enters how many random states to read from the beginning
of the file, the amount of time allowed for the switches to settle before a reading is taken,
and the type of experiment. Experiments may be one of three types: setup, testing, or
valid. Setup and testing are used for debugging purposes while a valid type experiment is
used when data is to be recorded and stored. Finally, the user may enter any annotations
regarding the experiment.
Besides running experiments, the user may configure the network analyzer and then enter
a state and take a reading for that state. This allows for data from individual states to be
selectively gathered.
Once the program is started, the current iteration number and state are displayed. The
state is shown as a binary number and also using a series of thirty-two LED type indictors.
A progress bar shows the percentage of the experiment which has been completed. At the
bottom of the window, the magnitude and phase for measured S-parameters are displayed.
When an experiment is run, the software creates a folder for that experiment. This
folder will be a subdirectory of either a folder titled “Valid” or “Other” depending on if the
experiment is setup as a valid experiment or either a setup or testing experiment, respectively.
These directories are placed in the folder selected when setting up the experiment. The
folder for each experiment is titled mm-dd-yyyy-{v,s,t} where mm-dd-yyyy is the date on
which the experiment was started and the suffix tells the type of experiment: valid, setup, or
testing. Within the folder, data files have the same prefix. Appended to this prefix is either
-config, -freq, -sxx, or -swr. Each of these files stores the experiment configuration, the
frequency list, S-parameter data (xx tells which parameter), or standing wave ratio (SWR)
data, respectively. Each line of an S-parameter file stores the data for one state. Each data
point has a real and imaginary value separated by a comma. Data points are then separated

57

Listing 4.1: Sample configuration file, 5-26-2009-1152AM-v-config.txt.
Date : 5/26/2009
Time : 1 1 : 5 2 AM
Number o f S t a t e s : 1
S t a r t ( Hz ) :
7 . 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0E+08
Stop ( Hz ) :
1 . 5 0 0 0 0 0 0 0 0 0 0 0 0 0 0E+09
Num P o i n t s : 26
Avg F a c t o r : 1
IF BW ( Hz ) :
3 . 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0E+03
Power (dBm) :
1 . 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0E+01
S e t t l e Time (ms) : 30
VISA Timeout (ms) : 30000
A n n o ta t io n s :
Antenna t o t−l i n e t o c a l e d p o r t . No 2 p o r t network .
F i n i s h Time : 5/26/2009 1 1 : 5 2 AM

1
2
3
4
5
6
7
8
9
10
11
12
13
14

by commas. If S-parameter data is imported into Excel as a .csv file, there will be twice the
number of columns as frequency points and the same number of rows as states. Frequency
files are a comma separated list of frequency points stored on one line.
Listing 4.1 lists an example configuration file. This file contains entries that describe the
configuration of the network analyzer, start and end dates and times, and any annotations
that were made by the user.

4.3.2

Standing Wave Ratio

A variation of the above program was created to evaluate the SWR observed when using the
two-port network. The program measures S11 , calculates the SWR and stores both the S11
and SWR data.

4.3.3

Genetic Algorithm

To find an optimized state for the self-structuring two-port network, a genetic algorithm (GA)
was implemented in LabView. This program was based on the random search program. As
such, a great deal of the interface, configuration, and output files are the same.
58

A genetic algorithm uses the evolutionary principle of survival of the fittest to find an
optimized solution. The GA used in this thesis were guided primarily by [63] and previous
work done at MSU by Raoul Ouedraogo in [13, 16]. Figure 4.4 is a flowchart of the GA used
in this thesis. This GA minimizes the cost function rather than maximizing the fitness.

59

Figure 4.4: Genetic algorithm flow chart.

60

Initially the user configures the network analyzer. Once configured, the program allows
the user to set parameters for the experiment and for the genetic algorithm. Currently
genes are only paired using random pairing. Future versions may include pairing methods
such as top-to-bottom; roulette wheel-weighted random; roulette wheel-cost weighted; and
tournament selection.
The first version of this program only optimized at one frequency. The second version
is able to optimize at multiple frequencies. When multiple frequencies are desired, the user
selects which frequencies to optimize, enters a goal value for each frequency in dB and
indicates whether the final value is to be lower than, higher than, or equal to the goal value.
The cost function must be edited in the block diagram because of the limitations of
LabView. The cost function, fc (·), is given by

fc (Sxy,1 , Sxy,2 , ..., Sxy,n ) =










n S
i=1 xy,i − gi ,

if diri = −1

n
i=1 Sxy,i − gi , if diri = 0









n g −S
xy,i ,
i=1 i

(4.1)

if diri = +1

where x, y ∈ {1, 2}, Sxy,i is the S-parameter at the ith frequency, n is the number of
frequencies to be optimized, gi is the goal at the ith frequency, and diri is the direction for
optimization at the ith frequency. If diri = −1, 0, +1, then the value is optimized to be less
than, equal to, or greater than the goal, respectively.
GA experiments output an additional file with the suffix -states.txt. This is a list of
evaluated states in an order that corresponds to the order of results in other files for that
experiment.

61

Figure 4.5: Experimental setup for initial random search experiment.

4.4

Random Search

Initial experiments used a random search to explore the capabilities of the two-port network.
This random search consisted of setting a switch state and recording the S-parameters before
moving onto another state.

4.4.1

Initial Investigation

Characterization of the self-structuring two-port network began by measuring all four Sparameters for thirty thousand states at twenty-six frequencies between 700 MHZ and
1.5 GHz, i.e. in 32 MHz steps. Switches were allowed to settle for 30 ms before S-parameter
measurements were taken. The experimental setup is schematically shown in Figure 4.5.
This experiment is stored as dataset 5-20-21-2011-combine.
Figure 4.6 through Figure 4.13 show plots of magnitude and phase for all four S-parameters.
Each frequency point from each state is plotted as a ‘+’. One can see that most states have
an insertion loss greater than 10 dB. In the middle of the measured frequency band, |S11 |
and |S22 | of -20 dB were easily reached while at either end of the band it was difficult to
reach -10 dB. Phase values ranged from −180◦ to +180◦ for all S parameters except for S11
where the phase primarily ranged from −180◦ to 0◦ .

62

Figure 4.6: Random search results, S11 magnitude, 5-20-21-2001-combine.

0
−5

|S11| (dB)

−10

−15

−20

−25

−30

−35

0.7

0.8

0.9

1
1.1
1.2
Frequency (GHz)

63

1.3

1.4

1.5

1.6

Figure 4.7: Random search results, S11 phase, 5-20-21-2001-combine.

180
150
120
90

∠ S11 (deg)

60
30
0
−30
−60
−90
−120
−150
−180

0.7

0.8

0.9

1
1.1
1.2
Frequency (GHz)

64

1.3

1.4

1.5

1.6

Figure 4.8: Random search results, S12 magnitude, 5-20-21-2001-combine.

0
−10
−20

|S12| (dB)

−30
−40
−50
−60
−70
−80
−90

0.7

0.8

0.9

1
1.1
1.2
Frequency (GHz)

65

1.3

1.4

1.5

1.6

Figure 4.9: Random search results, S12 phase, 5-20-21-2001-combine.

180
150
120
90

∠ S12 (deg)

60
30
0
−30
−60
−90
−120
−150
−180

0.7

0.8

0.9

1
1.1
1.2
Frequency (GHz)

66

1.3

1.4

1.5

1.6

Figure 4.10: Random search results, S21 magnitude, 5-20-21-2001-combine.

0
−10
−20

|S21| (dB)

−30
−40
−50
−60
−70
−80
−90

0.7

0.8

0.9

1
1.1
1.2
Frequency (GHz)

67

1.3

1.4

1.5

1.6

Figure 4.11: Random search results, S21 phase, 5-20-21-2001-combine.

180
150
120
90

∠ S21 (deg)

60
30
0
−30
−60
−90
−120
−150
−180

0.7

0.8

0.9

1
1.1
1.2
Frequency (GHz)

68

1.3

1.4

1.5

1.6

Figure 4.12: Random search results, S22 magnitude, 5-20-21-2001-combine.

10
0

|S22| (dB)

−10

−20

−30

−40

−50

−60

0.7

0.8

0.9

1
1.1
1.2
Frequency (GHz)

69

1.3

1.4

1.5

1.6

Figure 4.13: Random search results, S22 phase, 5-20-21-2001-combine.

180
150
120
90

∠ S21 (deg)

60
30
0
−30
−60
−90
−120
−150
−180

0.7

0.8

0.9

1
1.1
1.2
Frequency (GHz)

70

1.3

1.4

1.5

1.6

Figure 4.14: Wideband horn antenna matching experimental setup.

4.4.2

Matching a Wideband Horn Antenna

One possible use for a self-structuring two-port network is as a matching network. In order
to explore using the network as a matching network, a American Electronic Laboratories’
wideband, 500 MHz to 6 GHz, horn antenna was connected to port 2 of the self-structuring
two-port network while the network analyzer was connected to port 1. The LabView random
search program was used to explore all one hundred thousand states in the NIST generated
pool in addition to each switch operated by itself and all switches open and closed together
for a total of 100,034 states or 0.0023% of all possible states. Data was taken at twenty-six
frequencies between 700 MHz and 1.5 GHz in 32 MHz steps. Switches were allowed to settle
for 30 ms. The experimental setup is shown in Figure 4.14.
Figure 4.15 shows the SWR of the wideband horn antenna without the self-structuring
two-port network as a solid red line, data set 5-26-1152AM-v, and the SWR for all explored
states as blue ‘+’s, data set 5-25-2009-1123AM-v. Figure 4.16 only shows SWR values
between one and eight. The majority of states produce an SWR that is higher than the
original SWR of the antenna. It is also clearly seen that there are many states that produce
an SWR below that of the unmatched antenna expect at 1.43 GHz. These results seem
very promising at first glance; however, the data only shows how much power is reflected
back to the analyzer from the input of the matching network. What is not known from this
data is how much power is making it to the antenna versus being lost by the self-structuring
two-port network.

71

Figure 4.15: Random search of possible states and original unmatched antenna performance.

72

Figure 4.16: Random search zoomed in.

73

Figure 4.17: Maximum and minimum observed power balances.

Maximum and Minimum Power Balances Observed
1
No Tape

S11S* +S21S* (dB)
11
21

Sides

0.8

2nd Layer

0.6
0.4
0.2
0

4.4.3

1st Layer

0.8

1
1.2
Frequency (GHz)

1.4

Power Balance of Self-Structuring Two-Port Network

The power balance for the self-structuring two-port network was calculated for each state
in the initial random search dataset, dataset 5-20-21-2011-combine. It should be noted
∗
∗
that in this case, the power balance is calculated as S11 S11 + S21 S21 giving values of zero
for complete loss and one for no loss. Figure 4.17 shows the maximum and minimum values
for this data set and Figure 4.18 shows the average values for this dataset as solid blue
lines. The minimum and maximum values are the extremes for all measured points at each
frequency, independent of other values from the same state. The other data series plotted
in these figures will be discussed in Section 4.4.4.
74

Figure 4.18: Average observed power balances.
Average Power Loss of Two Port Network
1
No Tape
Sides
1st Layer
2nd Layer

0.9

S11S* +S21S* (dB)
11
21

0.8

0.7

0.6

0.5

0.4

0.7

0.8

0.9

1

1.1

1.2

Frequency (GHz)

75

1.3

1.4

1.5

One can see that the self-structuring two-port network is on average very lossy. The
best performance seen in the 5-20-21-2011-combine dataset is a power balance of 94.6%
at 1.244 GHz while the worst performance was 2.5% at 1.084 GHz.
It was unclear at this point in the research as to what loss mechanisms were playing the
most significant roles.

4.4.4

Mitigation techniques

In order to increase the efficiency, and thereby, the usefulness of the self-structuring two-port
network, various attempts were made to reduce the loss of the network. Because the design
and construction of the network are similar to those of a patch SSA, it was thought that
radiation was the primary loss mechanism. In order to reduce the radiated emissions from
the network, copper tape was first applied to the edges of the board with the thought that the
perimeter shorting posts were allowing the network to radiate. The maximum and minimum
power balances for this case are plotted in Figure 4.17 and the average value plotted in
Figure 4.18 as dotted green lines.
As the power loss was still very high, the shorting post pads and switches were insulated
with electrical tape before the entire network was covered with a layer of copper tape. After
measuring the losses, a second layer of tape was applied in an effort to close any gaps that
might have existed in the first layer. Figure 4.19 shows the network after copper tape was
applied. The maximum and minimum power balance for the first and second layer of tape
are plotted as red, dash-dash lines and black, dash-dot lines, respectively, in Figure 4.17.
The average values are plotted in Figure 4.18 using the same line styles.
Figures 4.17 and 4.18 suggest that the application of the copper tape does have some
effect on the network; however, the effect is not a noticeable improvement with regards to
the power balance of the network. This suggests that power is lost through a mechanism
other than radiation.
Another possible source of power loss is common-mode current on the control cables.
76

Figure 4.19: Copper covered self-structuring two-port network.

77

Three toroids were placed onto the control ribbon cable in order to reduce the commonmode current. The cable was split in half with a toroid placed around each half. The third
toroid was placed around the entire ribbon cable. This setup can be seen at the top of
Figure 4.19. Analysis of the power balance after application of the toroids did not show a
significant reduction in power loss.
Given that the network is typically lossy, it is necessary to determine at what frequencies are associated with the highest loss and which are associated with the lowest loss.
This was indirectly evaluated by finding the lowest reflection value (min(S11 )) and the
highest transmission values (max(S21 )) at each frequency across all measured states. Figure 4.20 shows the best transmission and reflection values found in a random search, dataset
6-23-2010-090PM-v. The difference between theses two values is shown in Figure 4.21.
Values greater than 0 suggest that the network transfers more power than it reflects. The
largest values tend to be below 3 GHz with two relatively wideband areas from 750 MHz
to 1 GHz and 1.9 GHz to 2.2 GHz. These bands are most useful when transmission is a
desirable characteristic of the two-port network.

4.5

Genetic Algorithm Based Search

Keeping in mind that the network is fairly lossy, several experiments were undertaken to
optimize the self-structuring two-port network as a matching network with various loads
and as an attenuator. The LabView based GA described in Section 4.3.3 was used. The
parameters and results from the GA optimizations are shown in Table 4.2.

78

Figure 4.20: Best S21 and S11 .

Maximum and Minimum Observed S Parameters
S Parameter Magnitude (dB)

0

−20

−40
Best S21 (Max)

−60

−80
0

Best S11 (Min)
1

2
3
4
Frequency (GHz)

79

5

6

Figure 4.21: Difference between maximum transmission and minimum reflection.

60

max(S21)−min(S11) in dB

40

20

0

−20

−40

−60

−80
0

1

2

3
4
Frequency (GHz)

80

5

6

81

Keep Top %

Mutation Prob.

Num. Evaluated

Dataset
Best State
Cost
6-30-2010-950PM-v-config.txt 0xFED66FBE 96.67 -10.79 1.72 -13.16 -6.93 750-1000
7-1-2010-147AM-v-config.txt
0xCE299D61 100.59 -11.14 3.07 -16.14 -7.84 1400-1500
7-2-2010-231PM-v-config.txt
0x07BA1662 487.50 -46.32 2.71 -51.87 -43.50 1400-1500
7-2-2010-851PM-v-config.txt
0x87373662
484.45 -46.04 2.48 -51.05 -43.67 1400-1500
7-6-2010-1032AM-v-config.txt 0x5001C3DE 619.82 -58.35 5.47 -65.81 -48.63 1900-2200
7-8-2010-155PM-v-config.txt
0x86B8D4F2 636.14 -59.83 3.08 -64.45 -55.16 1400-1500
0x3F6D640A
8.35 -19.85 1.08 -21.57 -17.70 750-1000
7-8-2010-613PM-v-config.txt
- All GA searches were optimized at eleven, linearly spaced frequency points in the specified range.
- All GA searches used a population size of 250 individuals and a maximum of 200 generations.
- New individuals were created using random pairing, mutated one bit.
- The top 10% of the population was not mutated.
- All searches reached the maximum number of generations before reaching the goal cost.

Goal (dB)

Freq (MHz)

Maximum (dB)

Minimum (dB)

St. Dev., σ (dB)

Mean, µ (dB)

Table 4.2: Genetic algorithm search parameters and results.

-2
-2
-2
-2
-2
-2
-20

0.10
0.10
0.10
0.15
0.10
0.10
0.10

0.75
0.75
0.75
0.80
0.70
0.70
0.75

16174
16251
14773
14317
17401
16079
16293

4.5.1

Optimized transmission

To begin optimization, each port of the self-structuring two-port was connected to the network analyzer as shown in Figure 4.5. Two optimizations were performed, the first from
750 MHz to 1 GHz (dataset 6-30-2010-950PM-v) and the second from 1.4 GHz to 1.5 GHz
(dataset 7-1-2010-147AM-v). A goal of -2 dB was set for all frequency points.
Figure 4.22 shows one of the lowest cost states, 0xFED66FBE. This figure shows that the
goal of -2 dB across the band was not met. As this state was found through optimization,
this results suggests that the network will typically have about 10 dB of insertion loss. This
particular state had an average insertion loss of µ = −10.79 dB and a standard deviation of
σ = 1.72 dB. A similar result is seen in Figure 4.23. This response shows a larger range of
transmission values than in the lower frequency band; however, the general value of 10 dB
remains (µ = −11.14 dB, σ = 3.07 dB).
These results point to the manufactured self-structuring two-port network having an
insertion loss of at least 10 dB for frequencies below 1.6 GHz. The author believes that
when optimized as in these results, losses caused by changes in the states are reduced leaving
only losses that result from the construction and physical properties of the network. This
conclusion is supported by the random search results as shown in Figure 4.10. The lowest
insertion loss found across the band during the random search was roughly 10 dB.

4.5.2

Spiral Antenna

In the previous section, the self-structuring two-port network was used to match 50 Ω loads
on each port. In order to provide a more challenging and realistic experiment, a spiral
antenna was connected to Port 2 of the network as shown in Figure 4.25. A picture of
the spiral antenna used in this experiment is shown in Figure 4.24. Port 1 of the network
analyzer was connected to Port 1 of the network while Port 2 of the analyzer was connected
to the wideband horn antenna used in Section 4.4.2.
The solid blue lines in Figure 4.26 and Figure 4.27, show the transmission through the
82

Figure 4.22: Optimized transmission with 50 Ω loads, 750 MHz to 1.0 GHz, dataset
6-30-2010-950PM-v.

Optimized Transmission
−6

|S21| (dB)

−8

−10

−12

−14
0.75

0.8

0.85
0.9
Frequency (GHz)

83

0.95

1

Figure 4.23:
Optimized transmission with 50 Ω loads,
7-1-2010-147AM-v.

1.4GHz to 1.5 GHz,

Optimized Transmission
−8

|S21| (dB)

−10
−12
−14
−16
1.4

1.42

1.44
1.46
Frequency (GHz)

84

1.48

1.5

Figure 4.24: Spiral antenna used for input impedance matching experiments. Large numbers
on the ruler are inches.

85

Figure 4.25: Experimental setup schematic for spiral antenna measurements and optimization.

system when the spiral antenna was connected directly to the analyzer. The dashed lines in
these figures show the transmission when the self-structuring two-port network was optimized
for transmission. Figure 4.26 is from two different optimizations, dataset 7-2-2010-231PM-v
is Matching State 1 using state 0x07BA1662 and dataset 7-2-2010-851PM is Matching State
2 using state 0x87373662. Figure 4.27 uses 7-6-2010-1032AM-v and shows the transmission
for state 0x5001C3DE. One can see that even after optimization, the insertion loss of the
network is still relatively high.

4.5.3

Dipole Antenna Transmission

Further transmission testing was conducting using a dipole antenna. The experimental setup
is similar to that used for the spiral antenna in Section 4.5.2 and is shown in Figure 4.28.
Results of the optimization at 1.4 to 1.5 GHz, dataset 7-8-2010-155PM-v, are shown in
Figure 4.29. As was seen before, the network exhibits an insertion loss of roughly 10 dB.
The original transmission shape was closely followed. This is promising because it suggests
that the network is able to be used as a matching network albeit with a fairly consistent
insertion loss.

86

Figure 4.26: Spiral antenna transmission optimization results at 1.4 GHz to 1.5 GHz.

Transmit/Receive System Characteristics
−30
−32
−34
−36

|S21| (dB)

−38
−40

No Matching
Matching State 1
Matching State 2

−42
−44
−46
−48
−50
−52
1.4

1.42

1.44

1.46

Frequency (GHz)

87

1.48

1.5

Figure 4.27:
Spiral antenna
dataset 7-6-2010-1032AM-v.

system

transmission,

1.9

GHz

to

2.2

GHz,

Transmit/Receive System Characteristics
−35
−40

|S21| (dB)

−45
−50

No Matching
Matching State 1

−55
−60
−65
−70
1.9

2
2.1
Frequency (GHz)

2.2

Figure 4.28: Experimental setup schematic for dipole antenna measurements and optimization.

88

Figure 4.29: Dipole antenna throughput results, dataset 7-8-2010-155PM-v.

Dipole Antenna Matching

|S21| (dB)

−50

Unmatched
Matching Network

−55

−60

−65
1.4

1.42

1.44
1.46
Frequency (GHz)

89

1.48

1.5

Figure 4.30: 20 dB attenuator optimization results.

−17.5
−18
−18.5

|S21|

−19
−19.5
−20
−20.5
−21
−21.5
−22
0.75

4.5.4

0.8

0.85
0.9
0.95
Frequency (GHz)

1

1.05

Attenuator

A self-structuring two-port network may be used to create other types of devices besides
matching networks. The use as an adjustable attenuator was evaluated using the setup
shown in Figure 4.5. An attenuation value of 20 dB was set as the goal. This value was
selected because it was larger than the typical insertion loss of the self-structuring two-port
network.
Figure 4.30 shows the final optimized state, 0x3F6D640A. One can see that after optimization, an attenuator with 20±2 dB was created. In this frequency range, the average
value is µ = 19.85 dB with a standard deviation of σ = 1.08 dB.
Optimizations to create a filter were not carried out.

90

4.6

Switch Investigation

The high loss of power observed in experiments, even after mitigation attempts, dictated that
further loss mechanisms be investigated. Simulations, as described in Section 3.2 were used
for computational investigation. Experimentally, the power behavior of the Coto switches
was investigated.

4.6.1

Individual Switch Properties

A 50 Ω microstrip line was manufactured to measure the S-parameters of the switches. The
circuit board with the switch used for testing is shown in Figure 4.31. A +5 VDC power
supply was used to open and close the switch. The full two port S-parameters for the switch
were measured in both open, Figure 4.32, and closed, Figure 4.32, states. The power balance
for an open and closed switch is plotted in Figure 4.34.
These results suggest that a significant amount of power is being lost in the switch. It is
believed that these losses come mostly from coupling into the control wires and into the relay
coil inside the switch. As power loss is additive, the sum of these losses added across thirtytwo switches has the potential to drastically effect the performance of the self-structuring,
two-port network.

4.6.2

Switch Removal from Experiment

Due to the relatively high losses observed in the relay switches, the effect of these switches
while connected to the network was investigated. First the copper covering was removed and
the two port was measured in an all open state. These results are plotted in Figure 4.35.
Next, the switches were disconnected from the shorting posts, but left connected to the
control wires and the top side of the board. This configuration was then measured and
the results are shown in Figure 4.36. The power balance from this experiment is shown in
Figure 4.37. The difference between the unsoldered switches and the open state is shown in

91

Figure 4.31: Switch testing microstrip setup.

92

Figure 4.32: S-parameters of an opened relay switch.

Opened Switch S−Parameters
0

S parameter (dB)

−5
−10
−15
−20
S11

−25

S22
−30

0.5

1

1.5

2

2.5
3
3.5
4
Frequency (GHz)

4.5

5

5.5

6

0

S parameter (dB)

−10
−20
−30
−40
−50
−60
S12

−70
−80

S21
0.5

1

1.5

2

2.5
3
3.5
4
Frequency (GHz)

93

4.5

5

5.5

6

Figure 4.33: S-parameters of a closed relay switch.

Closed Switch S−Parameters
0

S parameter (dB)

−10
−20
−30
−40
S11

−50

S22
−60

0.5

1

1.5

2

2.5
3
3.5
4
Frequency (GHz)

4.5

5

5.5

6

1.5

2

2.5
3
3.5
4
Frequency (GHz)

4.5

5

5.5

6

S parameter (dB)

5

0

−5

−10
S12
S21
−15

0.5

1

94

Figure 4.34: Power balance of a relay switch.

Switch Power Balance
0.1

Power Balance = S11S* +S21S* −1
11
S21

0
−0.1
−0.2
−0.3
−0.4
−0.5
−0.6
−0.7
−0.8

Opened
Closed
1

2
3
4
Frequency (GHz)

95

5

6

Figure 4.35: S-parameters for state 0x00000000.

Connected Switches S−Parameters, all open
10

S parameter (dB)

5
0
−5
−10
−15
−20
S11

−25
−30
0

S22
1

2

3
4
Frequency (GHz)

5

6

5

6

0

S parameter (dB)

−20
−40
−60
−80

S12
S21

−100
0

1

2

3
4
Frequency (GHz)

96

Figure 4.36: S-parameters with switches unsoldered from shorting posts.

Unsoldered Switches S−Parameters

S parameter (dB)

0

−5

−10

−15
0

1
S11

2
S22

3
4
Frequency (GHz)

5

6

0

S parameter (dB)

−20
−40
−60
−80

S12
S21

−100
0

1

2

3
4
Frequency (GHz)

97

5

6

Figure 4.37: Power balance for switches unsoldered from shorting posts.

Switch Power Balance
0
Soldered
Unsoldered

Power Balance = S11S* +S21S* −1
11
S21

−0.1
−0.2
−0.3
−0.4
−0.5
−0.6
−0.7
−0.8
−0.9
−1
0

1

2

3
4
Frequency (GHz)

5

6

Figure 4.38 and with a limited frequency range of 700 MHz to 1.5 GHz in Figure 4.39. As
can be seen in these figures, there seems to be no significant difference between having and
not having the switches soldered to the shorting posts. This either means that power was
not being lost through the switches or that power was coupling into the switches through
the control wires and the other leg of the switch which had remained soldered to the top of
the two-port.
To investigate further, the switches were completely removed (electrically) from the twoport. Control wires and most switches were disconnected from one another but remained on

98

Power balance difference (Connected − Unsoldered)

S21 difference (Connected − Unsoldered) (dB)

Figure 4.38: Difference in power balance for between open state and unsoldered switches.

Unsoldered Switches S−Parameters
40
30
20
10
0
−10
−20
−30
−40
0

1

2

3
4
Frequency (GHz)

5

6

2
3
4
Frequency (GHz)
Mean Difference: −0.0179

5

6

1

0.5

0

−0.5

−1
0

1

99

Power balance difference (Connected − Unsoldered)

S21 difference (Connected − Unsoldered) (dB)

Figure 4.39: Limited frequency range for the difference in power balance for between open
state and unsoldered switches.

Unsoldered Switches S−Parameters
20
10
0
−10
−20
−30
−40
0.7

0.8

0.9

1
1.1
1.2
Frequency (GHz)

1.3

1.4

1.5

0.8

0.9

1
1.1
1.2
Frequency (GHz)

1.3

1.4

1.5

1

0.5

0

−0.5

−1
0.7

100

the board due to the potting used to secure them earlier in experiments. The S-parameters
for this setup are shown in Figure 4.40 and the power balance for the network is shown in
Figure 4.41.

As is seen in the figure, there is still significant power loss even after the

switches have been disconnected. This suggests that, while the switches may have been a
sink for some power, the switches did not contribute significantly to the overall performance
of the two port network.At this time, no one power loss mechanism has been identified as
most significant. Instead, it is thought that a combination of various mechanisms leads to
the high losses observed.

101

Figure 4.40: S-parameters with no switches connected to the two-port network.

No switches connected
10
S11

S−parameter (dB)

5

S22

0
−5
−10
−15
−20
−25
−30
0

1

2

3
4
Frequency (GHz)

5

6

0

S−parameter (dB)

−20
−40
−60
−80

S21
S12

−100
0

1

2

3
4
Frequency (GHz)

102

5

6

Figure 4.41: Power balance with no switches connected to the two-port network.

Power Balance for Self Structuring 2 Port Netowrk
0

Power Loss = S11S* +S21S* −1
11
21

−0.1
−0.2
−0.3
−0.4
−0.5
−0.6
−0.7
−0.8
−0.9
−1

1

2
3
4
Frequency (GHz)

103

5

6

Chapter 5
Conclusion
5.1

Conclusion

This work has explored the capabilities of a self-structuring two-port network. A combination
of simulations and experiments were carried out to characterize the network.
It was found that the constructed prototype exhibited significant loss of power. Research
was undertaken to explore this topic further. While various loss mechanisms were identified,
no one mechanism could be pinpointed as the most significant. Steps were taken in the
experimental setup to mitigate the power lost with relatively little improvement.
Bearing in mind the lossy nature of the network, the capabilities of it were investigated
using both random and genetic algorithm based searches. Results from the random search
suggested that the lowest losses for the network are over the frequency range of approximately
750 MHz to 1.5 GHz. It was in this frequency range that most experiments were performed.
Random search results suggested that switch combinations existed for matching various types
of loads, given the insertion loss of the network. Results also pointed to the ability to vary
the insertion loss to levels below that caused by power loss.
Using a genetic algorithm, select applications of the self-structuring two-port network
were investigated. One such application was for the optimization of transmission through

104

the network. Accepting the system power loss, the network was fairly successful in matching
arbitrary loads. The network was also used successfully to create an arbitrary attenuator
with losses larger than normally observed in the network.
While the results obtained are promising, the actual performance of the manufactured
two-port network is, overall, disappointing. Due to the high power loss observed in both
simulations and in experiments, further research should be directed to creating a two-port
network with lower losses. Doing so would allow for further exploration of this topic.

5.2

Contributions

This work has made various contributions to the scientific community. First, the work has
been presented at the APS/URSI conference in 2009 and 2010 [64, 65]. It was also presented
at the 2010 Graduate Academic Conference at Michigan State University [66]. and through
numerous posters.

5.3

Topics and Suggestions for Future Study

After completing this work, it is the opinion of the author that the following items should
be considered for, or when conducting, future work:
• Future investigations should consider using different connectors and control methods
to reduce mechanical failures in experiments. If ribbon cables continued to be used,
proper crimping tools should be purchased. Extreme care should always be exercised
when connecting and disconnecting ribbon cables in order to extend their usable life.
• Any future prototype should consider redesigning how switches are connected to the
control wires. Aspects of the design to be considered include switch orientation, methods of affixing the switch to the surface, methods of shorting the circuit under test,
methods of connecting to the circuit under test, methods of connecting the control
105

wires to the switch pins, stress relief of control wires, and routing of the control wires.
Designers should consider the harsh environment experienced in a laboratory setting
including repeated stresses.
• The driver IC circuit should be updated to include a circuit that continually monitors
and gives feedback on the condition of each output. An example of this is to have an
LED that is one color when the output is within a given range to show that the switch
is functioning; is another color in order to show that the output is staying high, i.e.
the switch is not connected; and is a third color to show that there is a short in the
control wires.
• When using SMA and N type connectors, a cal kit should be created from the Type
N cal kit, HP 85032B, and the SMA cal kit, HP 85052D.
• Future investigations should consider using a substrate with different properties in an
attempt to reduce the insertion loss.
• Future investigations should be considered that study how phase is effected by the
network.
• Consideration should be given to constructing a self-structuring two-port network using
MEMS techniques. This would considerably decrease the size of the network while
hopefully decreasing losses.
• A very interesting potential application for a self-structuring two-port network is for
power factor correction on the power grid. The power factor of a device is the ratio of
the real power to the apparent power delivered to the device. The best performance
is reached when the power factor is unity. For most consumers, the power factor of a
device, or even their house, is not a major concern. For industrial customers the power
factor is important. Often times such customers attempt to correct their power factor
by installing what are essentially large capacitors or inductors.
106

Future research should consider creating a self-structuring two-port network capable
of operating on the power grid.
• In the future a microstrip based self-structuring two-port network could be constructed.
This may take the form of placing switches in a network of microstrip lines. These lines
may be laid out out in a pattern similar to the templates used in early SSA’s [1, 2].

107

Appendix

108

HP 85052D Economy Mechanical
Calibration Kit, DC to 26.5 GHz,
3.5 mm
This appendix outlines procedures for editing and downloading the definitions for the HP
85052D Calibration Kit to the HP 8753D Network Analyer. This procedure is done using
VNA Cal Kit Manager [61] and is written for version 2.10. The procedure can be adopted
to define any calibration kit for the network analyzer such as a combined SMA and N type
kit. Refer to the user manual of the cal kit [67] for more information and calibration tables.
1. Open VNA Cal Kit Manager
2. Open the 85053D calibration kit definition from the Agilent 85052D.ckm file found in
the same directory as the VNA Cal Kit Manager program
(default location: C:/Program Files/VNA Cal Kit Manager 2/
Agilent 85052D.ckm)
3. Save the file under a new name: File->Save As...
4. Go to the Standard Definitions tab
5. In the Standard Type column of the table, select none for standards #5 through #8
6. Copy the entries for standards #9, #11, and #13 to standards #3 through #5
109

7. Select none for standards #9, #11, and #13
8. Select the Class Assignments tab
9. Select 8753D/E from the VNA menu at the top of the program window
10. In Column A of the Class Assignments table, enter 2, 1, 3, 2, 1, 3, 5,5,5,5, 3, 3 for rows
S11A through Response & Isolation
11. Under the Cal Kit Properties tab, enter a label for the calibration kit definition and
update the description
12. After make sure the 8753D is turned on, select Check GPIB from the GPIB menu. A
window saying GPIB Check Passed and HEWLETT PACKARD,8573D,0,6.14 should be
displayed. Click OK. If it is not, ensure that the GPIB interface between the computer
and the analyzer is properly setup.
13. Select Send Kit to VNA from the GPIB menu. This can also be done using the button
on the toolbar.
14. A success window should show up after transfering the cal kit to the VNA. Click OK
15. On the 8753D, select the SAVE USER KIT soft key
16. Press the CAL button on the front of the 8753D
17. Select CAL KIT
18. Select SELECT CAL KIT
19. Select USER KIT
20. Press the CAL button on the front of the 8753D
21. The Cal Kit Label entered in the definition should now be displayed under the CAL
KIT soft key item
110

22. Calibrate using the standard procedures

111

Bibliography

112

Bibliography
[1] Coleman, C.M. and Rothwell, E.J. and Ross, J.E., “Self-structuring Antennas,” in Antennas and Propagation Society International Symposium, 2000. IEEE, vol. 3, pp. 1256
–1259 vol.3, 2000.
[2] Coleman, C.M. and Rothwell, E.J. and Ross, J.E. and Nagy, L.L., “Self-structuring
Antennas,” Antennas and Propagation Magazine, IEEE, vol. 44, pp. 11 – 23, June
2002.
[3] Perry, B.T. and Coleman, C.M. and Basch, B.F. and Rothwell, E.J. and Ross, J.E.,
“Self-structuring Antenna for Television Reception,” in Antennas and Propagation Society International Symposium, 2001. IEEE, vol. 1, pp. 162 –165 vol.1, 2001.
[4] Coleman, Christopher, Self-Structuring Antennas. Ph.d. dissertation, Michigan State
University, May 2002.
[5] Perry, Bradley, A Self-Structuring Antenna Prototype. M.s. thesis, Michigan State
University, August 2002.
[6] Coleman, C.M. and Rothwell, E.J. and Ross, J.E., “Investigation of Simulated Annealing, Ant-Colony Optimization, and Genetic Algorithms for Self-Structuring Antennas,”
Antennas and Propagation, IEEE Transactions on, vol. 52, pp. 1007 – 1014, April 2004.
[7] Yumei Guo and Yingzeng Yin and Huili Zheng and Jingli Guo, “Self-Structuring Antenna for Multi-Band Operation,” in Microwave, Antenna, Propagation and EMC Technologies for Wireless Communications, 2005. MAPE 2005. IEEE International Symposium on, vol. 1, pp. 350 – 353 Vol. 1, August 2005.
[8] Jessberger, Joachim, Investigation of the Near-Field Properties of the Self-Structuring
Antenna. M.s. thesis, Michigan State University and Studienarbeit, Kaiserslautern
University, October 2005.
[9] Perry, B.T. and Rothwell, E.J. and Nagy, L.L., “Analysis of Switch Failures in a
Self-Structuring Antenna System,” Antennas and Wireless Propagation Letters, IEEE,
vol. 4, pp. 68 – 70, 2005.
[10] Greetis, L.M. and Rothwell, E.J., “A Self-Structuring Patch Antenna,” in Antennas and
Propagation Society International Symposium, 2008. AP-S 2008. IEEE, pp. 1 –4, July
2008.
[11] Greetis, Lynn, A Self-Structuring Patch Antenna. M.s. thesis, Michigan State University, May 2008.
[12] Ouedraogo, R.O. and Rothwell, E.J., “Self-Tuning Cavity Resonator,” in Antennas and
Propagation Society International Symposium, 2008. AP-S 2008. IEEE, pp. 1 –4, July
2008.
113

[13] Ouedraogo, Raoul Ouatagom, Jr., A Self Tuning Electromagnetic Shutter. M.s. thesis,
Michigan State University, August 2008.
[14] Greetis, L. and Ouedraogo, R. and Greetis, B. and Rothwell, E.J., “A Self-Structuring
Patch Antenna: Simulation and Prototype,” Antennas and Propagation Magazine,
IEEE, vol. 52, pp. 114 –123, February 2010.
[15] Ouedraogo, R.O. and Rothwell, E.J. and Shih-Yuan Chen and Greetis, B.J., “An Automatically Tunable Cavity Resonator System,” Microwave Theory and Techniques, IEEE
Transactions on, vol. 58, pp. 894 –902, April 2010.
[16] Ouedraogo, R.O. and Rothwell, E.J. and Shih-Yuan Chen and Temme, A., “A SelfTuning Electromagnetic Shutter,” Antennas and Propagation, IEEE Transactions on,
vol. 59, pp. 513 –519, February 2011.
[17] Songnan Yang and Chunna Zhang and Helen Pan and Fathy, A. and Nair, V.,
“Frequency-Reconfigurable Antennas for Multiradio Wireless Platforms,” Microwave
Magazine, IEEE, vol. 10, pp. 66 –83, February 2009.
[18] Schaubert, D. and Farrar, F. and Sindoris, A. and Hayes, S., “Microstrip antennas
with frequency agility and polarization diversity,” Antennas and Propagation, IEEE
Transactions on, vol. 29, pp. 118 – 123, January 1981.
[19] Lan, G. and Sengupta, D.L., “Frequency Agile Circular Microstrip Antennas,” in Microwave Symposium Digest, 1985 IEEE MTT-S International, pp. 693 –695, June 1985.
[20] Chakravarty, T. and De, A., “Design of Tunable Modes and Dual-band Circular Patch
Antenna Using Shorting Posts,” Microwaves, Antennas and Propagation, IEE Proceedings, vol. 146, pp. 224 –228, June 1999.
[21] Bonefacic, D. and Bartolic, J. and Germ, M., “Shorted Patch Antenna with PIN Diode
Operating Band Switching,” in Microwave Conference, 2006. 36th European, pp. 862
–865, September 2006.
[22] Zhong, S.S. and Lo, Y.T., “Single-element Rectangular Microstrip Antenna for DualFrequency Operation,” Electronics Letters, vol. 19, pp. 298 –300, April 1983.
[23] Bao Wang and Yuen Lo, “Microstrip Antennas for Dual-Frequency Operation,” Antennas and Propagation, IEEE Transactions on, vol. 32, pp. 938 – 943, September 1984.
[24] Waterhouse, R., “Small Microstrip Patch Antenna,” Electronics Letters, vol. 31, pp. 604
–605, April 1995.
[25] Waterhouse, R.B. and Targonski, S.D. and Kokotoff, D.M., “Design and Performance of
Small Printed Antennas,” Antennas and Propagation, IEEE Transactions on, vol. 46,
pp. 1629 –1633, November 1998.
[26] Kan, H.K. and Waterhouse, R.B., “A Small Printed Antenna for Wireless Communication Handset Terminals,” in Microwave Conference, 2000 Asia-Pacific, pp. 723 –726,
December 2000.
114

[27] Shackelford, A.K. and Sang-Yick Leong and Kai-Fong Lee, “Simulation of a Probe-fed
Notched Patch Antenna with a Shorting Post,” in Antennas and Propagation Society
International Symposium, 2001. IEEE, vol. 2, pp. 708 –711, July 2001.
[28] Kan, H.K. and Waterhouse, R.B. and Lee, A.Y.J. and Pavlickovski, D., “Variations
of the Shorted Patch Antenna,” in Antennas and Propagation Society International
Symposium, 2005 IEEE, vol. 2B, pp. 230 – 233, July 2005.
[29] Zavosh, F. and Aberle, J.T., “Improving the Performance of Microstrip-patch Antennas,” Antennas and Propagation Magazine, IEEE, vol. 38, pp. 7 –12, August 1996.
[30] Jun Zhao and Raman, S., “Design of ”chip-scale” patch antennas for 5-6 GHz wireless
microsystems,” in Antennas and Propagation Society International Symposium, 2001.
IEEE, vol. 2, pp. 6 –9, July 2001.
[31] Shaojun Fang and Jun Zheng and Xiuzhen Luan, “A Novel Multi Arc Slot Antenna
for WLAN Applications,” in Antenna Technology: Small Antennas and Novel Metamaterials, 2005. IWAT 2005. IEEE International Workshop on, pp. 209 – 212, March
2005.
[32] Hunter, I.C. and Rhodes, J.D., “Electronically Tunable Microwave Bandpass Filters,”
Microwave Theory and Techniques, IEEE Transactions on, vol. 30, pp. 1354 –1360,
September 1982.
[33] Casini, F. and Gatti, R.V. and Perrone, V. and Sorrentino, R., “A New Approach to
the Analysis and Synthesis of Lossy Reconfigurable Matching Networks,” in Microwave
Conference, 2009. EuMC 2009. European, pp. 1235 –1238, October 2009.
[34] Qin Shen and Barker, N.S., “Distributed MEMS Tunable Matching Network using
Minimal-Contact RF-MEMS Varactors,” Microwave Theory and Techniques, IEEE
Transactions on, vol. 54, pp. 2646 –2658, June 2006.
[35] Unlu, M. and Topalli, K. and Atasoy, H.I. and Temocin, E.U. and Istanbulluoglu, I. and
Bayraktar, O. and Demir, S. and Civi, O.A. and Koc, S. and Akin, T., “A Reconfigurable
RF MEMS Triple Stub Impedance Matching Network,” in Microwave Conference, 2006.
36th European, pp. 1370 –1373, September 2006.
[36] Vaha-Heikkila, T. and Rebeiz, G.M., “A 20-50 GHz Reconfigurable Matching Network
for Power Amplifier Applications,” in Microwave Symposium Digest, 2004 IEEE MTT-S
International, vol. 2, pp. 717 – 720 Vol.2, June 2004.
[37] V¨h¨-Heikkil¨, Tauno and Rebeiz, Gabriel M., “A 4–18-GHz Reconfigurable RF MEMS
a a
a
Matching Network for Power Amplifier Applications,” International Journal of RF and
Microwave Computer-Aided Engineering, vol. 14, no. 4, pp. 356–372, 2004.
[38] Domingue, F. and Fouladi, S. and Mansour, R.R., “A Reconfigurable impedance Matching Network Using Dual-Beam MEMS Switches for an Extended Operating Frequency
Range,” in Microwave Symposium Digest (MTT), 2010 IEEE MTT-S International,
pp. 1552 –1555, May 2010.
115

[39] Domingue, F. and Fouladi, S. and Kouki, A.B. and Mansour, R.R., “Design Methodology and Optimization of Distributed MEMS Matching Networks for Low-MicrowaveFrequency Applications,” Microwave Theory and Techniques, IEEE Transactions on,
vol. 57, pp. 3030 –3041, December 2009.
[40] Karim, MF. and Ai-Qun Liu and Alphones, A. and Aibin Yu, “A Reconfigurable Micromachined Switching Filter Using Periodic Structures,” Microwave Theory and Techniques, IEEE Transactions on, vol. 55, pp. 1154 –1162, June 2007.
[41] Q. Shen and N. Barker, “Reconfigurable Matching with a 10-30 GHz Distributed RFMEMS Tuner,” in Microwave Conference Proceedings, 2005. APMC 2005. Asia-Pacific
Conference Proceedings, vol. 4, p. 4 pp., December 2005.
[42] Fouladi, S. and Akhavan, A. and Mansour, R.R., “A Novel Reconfigurable Impedance
Matching Network Using DGS and MEMS Switches for Millimeter-Wave Applications,”
in Microwave Symposium Digest, 2008 IEEE MTT-S International, pp. 145 –148, June
2008.
[43] Domingue, F. and Fouladi, S. and Mansour, R. R., “A Reconfigurable impedance Matching Network Using Dual-Beam MEMS Switches for an Extended Operating Frequency
Range,” in Microwave Symposium Digest (MTT), 2010 IEEE MTT-S International,
p. 1, May 2010.
[44] Yumin Lu and Peroulis, D. and Mohammadi, S. and Katehi, L.P.B., “A MEMS reconfigurable matching network for a class AB amplifier,” Microwave and Wireless Components
Letters, IEEE, vol. 13, pp. 437 – 439, October 2003.
[45] Draskovic, D. and Budimir, D., “Optically Reconfigurable Matching Networks,” in Microwave Photonics, 2008. Jointly held with the 2008 Asia-Pacific Microwave Photonics
Conference. MWP/APMP 2008. International Topicsl Meeting on, pp. 135 –137, October 2008.
[46] Draskovic, D. and Christodoulou, C. and Budimir, D., “Optically Reconfigurable RF
circuits,” in Antennas and Propagation (EuCAP), 2010 Proceedings of the Fourth European Conference on, pp. 1 –3, April 2010.
[47] Malmqvist, R. and Samuelsson, C. and Rantakari, P. and Vaha-Heikkila, T. and Smith,
D. and Varis, J. and Baggen, R., “RF MEMS and MMIC Based Reconfigurable Matching Networks for Adaptive Multi-Band RF Front-Ends,” in RF Front-ends for Software
Defined and Cognitive Radio Solutions (IMWS), 2010 IEEE International Microwave
Workshop Series on, pp. 1 –4, February 2010.
[48] Hong, J.S. and Lancaster, M.J., “Microstrip Bandpass Filter Using Degenerate Modes
of a Novel Meander Loop Resonator,” Microwave and Guided Wave Letters, IEEE,
vol. 5, pp. 371 –372, November 1995.
[49] Abbaspour-Tamijani, A. and Dussopt, L. and Rebeiz, G.M., “Miniature and tunable
filters using MEMS capacitors,” Microwave Theory and Techniques, IEEE Transactions
on, vol. 51, pp. 1878 – 1885, July 2003.
116

[50] Cen Ong and Okoniewski, M., “MEMS-Switchable Coupled Resonator Microwave Bandpass Filters,” Microwave Theory and Techniques, IEEE Transactions on, vol. 56,
pp. 1747 –1755, July 2008.
[51] Peng Wen Wong and Hunter, I.C., “A New Class of Low-Loss High-Linearity Electronically Reconfigurable Microwave Filter,” Microwave Theory and Techniques, IEEE
Transactions on, vol. 56, pp. 1945 –1953, August 2008.
[52] Perruisseau-Carrier, Julien and Pardo-Carrera, Pablo, “Recent Developments on Reconfigurable Band-Filtering Antennas,” in Antennas and Propagation (EuCAP), 2010
Proceedings of the Fourth European Conference on, pp. 1 –4, April 2010.
[53] Xiaoguang Liu and Katehi, L.P.B. and Chappell, W.J. and Peroulis, D., “High-Q Tunable Microwave Cavity Resonators and Filters Using SOI-Based RF MEMS Tuners,”
Microelectromechanical Systems, Journal of, vol. 19, pp. 774 –784, August 2010.
[54] Karim, M.F. and Liu, A.Q. and Alphones, A. and Yu, A.B., “A Novel Reconfigurable
Filter Using Periodic Structures,” in Microwave Symposium Digest, 2006. IEEE MTT-S
International, pp. 943 –946, June 2006.
[55] Yi-Ming Chen and Sheng-Fuh Chang and Cheng-Yu Chou and Kun-Hsing Liu, “A
Reconfigurable Bandpass-Bandstop Filter Based on Varactor-Loaded Closed-Ring Resonators [Technical Committee],” Microwave Magazine, IEEE, vol. 10, pp. 138 –140,
February 2009.
[56] Naglich, E.J. and Juseop Lee and Peroulis, D. and Chappell, W.J., “A Tunable
Bandpass-to-Bandstop Reconfigurable Filter With Independent Bandwidths and Tunable Response Shape,” Microwave Theory and Techniques, IEEE Transactions on,
vol. 58, pp. 3770 –3779, December 2010.
[57] Ramo, Simon and Whinnery, John R. and Van Duzer, Theodore, Fields and Waves in
Communication Electronics. John Wiley and Sons, Inc., 3rd ed., 1994.
[58] EM Software & Systems-S.A. (Pty) Ltd., “FEKO - EM Simulation Software.”
http://www.feko.info/.
[59] Jack Ross and Associates, LLC., “GA-Suite.” http://www.johnross.com/gasuite.html.
[60] Coto
Technology,
“9011,
9012
&
9117
http://www.cotorelay.com/9011 9012 9117.pdf.

Miniature

SIP

Relays.”

[61] Barry Brown and Agilent, “VNA Help - Microwave Network Analyzer Measurement
Information.” http://na.tm.agilent.com/vnahelp/.
[62] National Instruments, “NI LabView.” http://www.ni.com/labview/.
[63] Randy L. Haupt and Sue Ellen Haupt, Practical Genetic Algorithms. New York : Wiley,
Second Ed. ed., 2004.

117

[64] Andrew Temme and Chien-Hsien Lee and Raoul Ouedraogo and Shih-Yuan Chen and
Brian J. Greetis and Edward J. Rothwell, “A Self-Structuring 2-Port Network,” 2009
USNC/URSI National Radio Science Meeting, June 2009.
[65] Andrew K. Temme and Edward J. Rothwell, “Optimization of a Self-Structuring TwoPort Network for Use as a Wideband RF Matching Network,” 2010 USNC/URSI National Radio Science Meeting, July 2010.
[66] Andrew K. Temme, “A Self-Structuring Two-Port Network,” 2010 Graduate Academic
Conference Michigan State University, March 2010.
[67] Agilent, “Agilent Technologies 85052D 3.5mm Economy Calibration Kit.” User’s and
Service Guide, September 2007.

118