NOTE TO USERS

This reproduction is the best copy available.

Kenneth Ho Yan Ip

- - - - - - - - -

A thesis submittedin~confoÏTmïty w i t l the requTrëmënEs- -

for the degree of Master of Applied Science Graduate Department of Electrical and Computer Engineering

University of Toronto

Copyright @ 2001 by Kenneth Ho Yan Ip

National Library 1+1 ofCamda Bibliothèque nationale du Canada

Acquisitions and Acquisitions et Bibliogrâphic Services services bibliographiques

395 WeUingbn Street 395, rue W&ngtwi OnewaON KlA0154 OnawaON K 1 A W canada Canada

The author has granted a non- exclusive licence allowing the National Library of Canada to reproduce, loan, distribute or seii copies of this thesis in rnicrofonn, paper or electronic formats.

The author retains ownership of the copyright in this thesis. Neither the thesis nor substantial extracts fiom it may be printed or othexwise reproduced without the author's permission.

L'auteur a accordé une licence non exclusive permettant à la Bibliothèque nationale du Canada de reproduire, prêter, distribuer ou vendre des copies de cette thèse sous la forme de microfiche/film, de reproduction sur papier ou sur format électronique.

L'auteur conserve la propriété du droit d'auteur qui protège cette thèse. Ni la îhèse ni des extraits substantiels de celle-ci ne doivent être imprimés ou autrement reproduits sans son autorisation.

Abstract

A Compact Four-Element Injection-Locked Scanning Antenna Array

Kenneth Ho Yan Ip

Master of Applied Science

Graduate Department of Electrical and Computer Engineering

University of Toronto

2001

A compact, single layer CPW-fed patch scanning array using injection locking a t

9.83 GHz is presented in this thesis. The unit element for the array is a self oscillating

active patch antenna with a GaAs FET centered behind the patch for tight packing.

The feedback for the oscillator is provided t hrough elec tromagnet ic coupling using a

twin-dot arrangement behind the patch. A low power control signal is injected through

parasitic coupling at the CPW side of the circuit. This parasitic coupling is achieved by

electromagnetic coupling of the locking signal to the gate of the FET. Phase shifting of

the elements is achieved by electronically adjusting the gate voltage of the GaAs FET.

A scan range of - 12" to 9.5* is obtained for tbis array.

Acknowledgements

Regardless of how it seems, this work could not have been completed without the

help of many people. First and foremost, 1 would like to thank my supervisor George

Eleftheriades for his guidance and insight, and also for his patience in seeing this work

through. 1 wodd also like to thank Tommy Kan for his contributions to Chapter 3 of

this Thesis.

1 would like to thank the people in my research group for their invaluable assistance

with the technical aspects of my learning during this effort, including Xidong Wu, Ramesh

Abhari, Micah Stickel, Mina Danesh, Meide Qiu, Tony Grbic, Andrew Pavacic, Michael

Simcoe, Micheal Hickey, Anthony Zegers, Roman Pahuta, Edwin Lau, Gerald Dubois,

and Peter Kremer.

I would like to extend my thanks to the graduates outside my research group for their

support and assistance in the early stages of my graduate study, including Shirley Lam,

Micheal Sun, Gary Choy, Farhad Meshkati, Pravesh Mahtani, and Dickson Cheung.

Next, 1 tvould like to t hank my friends for providing moral support, jokes and much fun

outside of school, including Shirley Lam, Anson Lam, Christy Lam, Gary Yip, Catherine

Lai, Jerry Wang and Andrew Lam.

Lastly, and rnost importalitly, 1 would like to thank my parents and my sister Holly

for their years of love and devotion, and I dedicate this work to them.

iii

Contents

Introduction 1

1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

. . . . . . . . . . . . . . . . . . . . . . . . 1.2 Mechanical Scanning Systems 3

. . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 SwitchedBeamSystems 3

. . . . . . . . . . . . . . . . . 1.4 Conventional Phased Array-Based Systems 4

. . . . . . . . 1.5 Proposed Design: Phased-Shifterless Beam Steering Array 5

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.6 Overview 6

Background 7

. . . . . . . . . . . . . . . . . . . . . . . . . 2.1 Theory of Injection Locking 7

. . . . . . . . . . . . . . . . . 2.1.1 Analysis of Free-Running Oscillator 8

. . . . . . . . . . . . . . . 2.1.2 Analysis of Injection-Locked Oscillators 9

. . . . . . . . . . . . . . . . . . . . . . . . . 2.1.3 PhaseShiftingRange 11

2.2 Review of the Existing Injection-locked Scan Array Architectures . . . . 13

. . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.1 Corporate Feed 13

. . . . . . . . . . . . . . 2.2.2 Unilateral Injection-locked Phased Array 15

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Oscillator Design 18

2.3.1 Prediction of the Large Signal S-Parameters of GaAs FETs . . . . 19

. . . . . . . . . . . . . . . 2.3.2 Embedding Elements of the Oscillator 20

3 A Single Layer CPW-Fed Active Patch Antenna 24

3.1 Active Antenna Configuration . . . . . . . . . . . . . . . . . . . . . . . . 25

3.2 Active Antenna Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

3.2.1 Choiceof Frequency . . . . . . . . . . . . . . . . . . . . . . . . . 26

3.2.2 Choice of Substrate . . . . . . . . . . . . . . . . . . . . . . . . . . 26

3.2.3 Characterization of the MMIC . . . . . . . . . . . . . . . . . . . . 26

3.2.4 Conditions of Oscillation . . . . . . . . . . . . . . . . . . . . . . . 27

. . . . . . . . . . . . . . . . . . . 3.2.5 Overview of the Overall Design 27

. . . . . . . . . 3.2.6 Design of the Two-port CPW-Fed Patch Antenna 28

3.2.7 Oscillation Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

. . . . . . . . . . . . . . . . . . . . . . . 3.3 E'rperimental Setup and Results 34

3.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

4 A Compact Single Layer Injection-Locked Active Antenna 39

4.1 Active Antenna Configuration And Design . . . . . . . . . . . . . . . . . 40

4.1.1 Choice of Frequency . . . . . . . . . . . . . . . . . . . . . . . . . 40

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.2 Configuration 40

4.1.3 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

4.2 ExperimentalResults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

. . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Summary and ConcIusion 50

5 A Compact Four-Element Injection-Locked Scanning Array 52

5.1 Array Configuration And Design . . . . . . . . . . . . . . . . . . . . . . . 52

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1.1 Configuration 52

5.1.2 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

5.2 Experimental Setup and Results . . . . . . . . . . . . . . . . . . . . . . . 56

5.2.1 Radiation Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . 56

5.2.2 EIRP Measurement Setup and Results . . . . . . . . . . . . . . . 59

5.3 Summary and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 61

6 Conclusions and Future Directions 62

6.1 L"onc1usions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

6.2 FutureDirections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

6.2.1 High DC-to-RF Efficiency Osciilator . . . . . . . . . . . . . . . . 63

6.2.2 Enhanced Bandwidt h . . . . . . . . . . . . . . . . . . . . . . . . . 64

6.2.3 Increased Steering Angle using a Unilateral Injection Locking Ar-

chitecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

Contributions 66

A Network Analyzer Calibrat ion: Thru-Reflect-Line 67

A.1 Short-Open-Load-Thru (SOLT) Calibration . . . . . . . . . . . . . . . . 68

A.l . l The HP8722C VNA . . . . . . . . . . . . . . . . . . . . . . . . . 69

A . 2 Thru-Reflect-Line (TRL) Calibrat ion . . . . . . . . . . . . . . . . . . . . 70

A 2 1 The HP8722C VNA . . . . . . . . . . . . . . . . . . . . . . . . . 71

. . . . . . . . . . . . . . . A.3 Implementation of TRL in CPW environment 72

Bibliography 75

Chapter 1

Introduction

1.1 Motivation

The growing demand of high bandwidth data access in future wireless communications

systems has forced engineers to look for inexpensive hardware designs that are compact

and efficient. It is predicted that, in addition to higher operating frequency, future wire-

less systems will require the front-end antenna system to perform scanning or tracking

in order to support higli bandw-idth. An application of this concept requires the use of

directive? beam steering antenna array systems [l] . By exploiting the properties of direc-

tional antema arrays, these systems provide spatial filtering and maximize the response

in the direction of the desired signal, which helps in lowering the filtering requirements

in the subsequent RF stages and improving the overall capacity of cellular systenis.

For the past decades, beam steering antenna systems have mostly been used in the

military, as radars for missile detection and tracking. Recently, these systems have been

adopted in satellite systems, in the form of multi-bearn antenna systems, to provide

worldwide continuous coverage [l, 2). Although applying beam steering systems to cel-

lular communications is envisioned to be a logical approach in enhancing capacity, two

main constraints have prohibited them from entering to the commercial sector: size and

cost. For example, the use of beam steering antenna array systems in laptop computers

for wireless LAN applications is desirable since the use of highly directive antenna ar-

rays offers higher bandwidth and better signal-tenoise ratio (SNR). However, such array

needs to be of reasonably small size and low profile to fit in a mobile unit. In a u t e

motive and aircraft industries, beam steering antennas can be utilized as radar sensors

for collision avoidance in vehicles and as landing/surveillance indicators in commercial

aircraft [3]. Again, cost and low profile requirements have forced engineers to choose

other designs, such as lens antenna implementations for automobiles, over phased array

designs [4].

Having mentioned several potential applications, the architectures for beam steering

antenna systems will be discussed. The most popular architectures for beam steering are

phase-shifter based antenna arrays, switched beam systems, and mechanical scanning

systems [2]. Although these architectures provide viable solutions, switched beam and

phased array antennas are preferable choices thanks to their accurate and fast electronic

steering features. However, both architectures require the use of a feeding network, which

can result in significant feedline losses. As the number of antenna elements increases, the

feedline losses increase proport ionally. To furt her complicate mat ters, transmit-receive

(T-R) modules have to be placed behind each antenna element for full transmit and

receive functionality, which introduces cost and efficiency issues into the pic t u e . Both

architectures will be briefly introduced in later sections.

Less conventionally, novel bearn-steering techniques using injection-locked oscillators

have been demonstrated to perform scanning functions sirnilar to phase-shifter based

systems [5 ] . Compared with conventional phased arrays, these novel systems pose less

stringent specifications on the feeding network, and do not require phase shifters and

power amplifiers. However, the drawback for these systems is that they offer a limited

scanning range. This limited scanning range aspect will be discussed in the next chapter.

CHAPTER 1. INTRODUCTION

1.2 Mechanical Scanning Systems

Beam steering can be reaüzed in a simple fashion - using a highly directive reflector

antenna mounted on top of a mechanical rotator. While this design is relat

sive, easy to build and the antenna performance (radiation pattern) is not

throughout the scan, they tend to be large, bulky and slow [6].

1.3 Switched Beam Systems

Fetd Network 1

ively inexpen-

compromised

Figure 1.1: (a) Switched beam architecture using the Butler matrix. (b) Main beam

radiation patterns for the Butler matrix.

Figure l . l a shows an example of a four element switched bearn system using the Butler

matrix [2]. The feed network consists of four quadrature hybrids and two 45" delay lines

and is used to distribute the RF input signal from the four input ports to the antenna

array with equal magnitudes but different phases. By switching the RF signal between

the input ports, beams pointing at different directions can be formed (Figure l . lb) .

The main advantage of this architecture is that it is relatively low cost and simple to

implement . In addition, this architecture can be extended to 2-D implementations and

also support multiple beams, which is suitable for satellite communications [2]. However,

the scanning is discrete and the feed network increases with the number of antenna

elements, making it less attractive for mobile LAN applications where flexibility, size and

CHAPTER 1. INTRODUCTION

cost are very important.

1.4 Conventional Phased Array-Based Systems

Feed Nmwork

lnput Si,&

Figure 1.2: Conventional architecture for an electronically scanned transmit ter.

Figure 1.2 shows a simplified conventional phased array for transmit ting applications

[2]. It consists of an array of four antenna elements, phase shifters and amplifiers. Feed

networks are used to distribute the RF signal from the source to each antenna element.

Such feed networks include corporate-, parallel-, series-, and space-feed networks; Fig-

ure 1.2 shows an exmpIe of a corporate-feed network using circuit power-dividers. By

electronically controlling the phase-shifters a t each antenna element, a constant phase

progression is established dong the array, leading to a scanning radiation pattern.

These systems are conceptually simple and can provide fast electronic beam steering.

Nonetheless, as the number of antenna element increases, the number of amplifiers and

phase shifters as well as the size of the feed network i~crease proportionally [2]. This

drawback has led researchers to explore other techniques in combining the functions of

phase shifters and arnplifiers, as well as minimizing the size of the feeding network [ 5 ] .

CHAPTER 1. INTRODUCTION 5

1.5 Proposed Design: P hased- S hifterless Beam Steer-

ing Array

Square Patch (Front)

hiMIC (bsdr)

70 Q CPW Lille

Figure 1.3: Proposed architecture for an elec tronically scanned transmit ting array.

Scanning antenna arrays utilizing injection-locked oscillators emerge as an attractive

alternative to conventional phase-shifter based systems [5, 71. In these types of novel

beam scanning arrays, the nurnber of circuit power-dividers is reduced by using active

integrated antennas for spatial power combining which leads to reduced feed-line losses.

In addition, phase shifters and amplifiers are replaced by injection-locked phase agile

oscillators for a low cost design [5]. Another added advantage is that the phase noise of

the array can be lowered by using a low-power, stable reference oscillator, making these

types of novel systems an attractive alternative for phased-array applications [5 ] .

The purpose of this work is to design and implement a low cost, compact beam

steering array using injection locked oscillators. The proposed scan array is shown in

Figure 1.3. This scan array is composed of four identical active patch antennas. These

active elements are, by themselves, free-running oscillators with a GaAs FET integrated

with each patch antenna using CPW technology. With the GaAs FETs situated behind

the patch, a compact active antenna element is realized, along with the added advantage

of low spurious radiation. In addition, CPW technology is adopted for an easy integration

wit h electronics, and to avoid via holes and multi-layer substrates when compared wit h

microstrip technology [8].

A low power injection signal is used for synchronization and phase control of the

oscillators. A single 70 R CPW transmission line is utilized to distribute the low power

injection signal in series, in order to maintain a compact configuration (as opposed to a

corporate feed of bulky power dividers). Parasitic coupling between the locking signa)

and the active antenna is achieved by connecting the injection signal line to the open-

circuited stub at the gate of the FET using a pseud+T-junction (see Figures 1.3). Finally,

phase shifts for each active element are achieved by detuning the gate biasing voltages

of the active elements within their locking ranges [7].

1.6 Overview

This Thesis is divided into six Chapters. Chapter 2 reviews the concepts, theory and

previous work related to injection-locked steerable arrays. Chapter 3 describes the pro-

posed CPW-fed slot-coupled patch antenna structure and evaluates the possibility of

integrating it with active devices. In addition, the design and the experimental results

of a 2.6 GHz prototype active antenna are presented. Chapter 4 describes the design

and the experirnental results from a compact version of the active antenna presented in

Chapter 3 at 9.8 GHz. Chapter 5 presents the design and the experimental results of a

linear beam scanning array based on the active antenna discussed in Chapter 4. Finally,

Chapter 6 sumrnarizes the research findings and suggests future research directions.

Chapter 2

Background

This chapter provides the t heory needed to understand the operation of phased-shifterless

scanning arrays. In the first section of this chapter, the theory of injection Iocking is

presented. In the second section, an overview of the existing phased-shifterless array

architectures is reviewed. Finally, in the third section, the oscillator design using an

S-paramet er approach is discussed.

2.1 Theory of Injection Locking

Output Signal

-0 v m Injection Signal

+ Active Device

Figure 2.1: (a) General representation of an injection locked oscillator (b) Mode1 of the

injection locked oscillator.

Figure 2. la shows a diagram of an injection-locked voltage control oscillator (VCO). A

typical VCO consists of a tunable resonant tank, oscillating at a free running oscillation

frequency wo. When a low power injection signal is present in the system, either by direct

connection or through coupling, the output frequency of the VCO will follow the injected

signal over a certain frequency range, Aw,(=locking range). The oscillator is said to be

'locked' to the injection signal.

The oscillator's model is shown in Figure 2.lb [9, 101. In this simplified model.

the active device has a negative load admittance of -Gd(V), which is a function of the

oscillator's voltage. The injection signal is modelled by an injection current with strengt h

of pGL, phase of Q, and an injection frequency, ui,. The resonant tank is modelled by

a shunt LC element . The oscillator load is represented by an admittance, GL .

2.1.1 Analysis of Free-Running Oscillator

In the absence of an injection signal, the oscillator in Figure 2.lb is operating about its

free-running frequency, wo. The equation of the free-running oscillation is given by [9]

where,

Bc =WC

Separating the real and imaginary parts of equation 2.1 yeilds

and a free-running frequency of

1

For oscillator analysis, it is convenient to introduce an important parameter called

the quality factor, which is defined as [I l ]

(average energy stored) Q = w

energy loss /secad

The quality factor for a parallel resonant circuit is

2.1.2 Analysis of Injection-Locked Oscillators

Going back to Figures 2. la and 2.1 b, the output voltage of the oscillator can be expressed

where A(t ) and d ( t ) are the time dependent amplitude and phase which are assumed to

Vary slowly with respect to the injection signal frequency, winj. This latter assumption

allows to neglect higher order terms in the the subsequent analysis.

Referring back to

Current Law (KCL):

Figure 2.lb where a small injection signal is present, by Kirchhoff's

the circuit c m be represented by

d v C- + G L V + Vdt + Gd(V) = Ii,

dt L

where,

Iinj = pGr. COS(U~, t + +)

pGL in equation 2.10 represents the injection signal strength, and @ represents the phase

of the injection current.

Substituting equation 2.8 into 2.9 and performing integration by parts yields,

1 dA + C-- [ A d t + (GL - G d ) + ---- A C O S ( W ~ , ~ + @ ) = I ~ , (2.11)

Multiplying equation 2.11 by cos(winjt + 4) and sin(winjt + Q) in t'trn and integrating

the results over one period yields the following two equations:

1 dA A C-- [ A d t + (GL - G d ) + = pGL COS($J - 4) L w&j A At

Finally, substituting equations 2.7 and 2.5 to 2.12 and 2.13 reveals the folIowing

amplitude and phase dynamics for the injection locked system of Figure 2.1:

where Q is given by equation 2.7.

Note that the amplitude dynamics (Equation 2.15) is not as important as the phase

dynamics of the system and, therefore, not discussed here 1121. On the other hand,

equation 2.14 is an important result known as Adler's equation [9, 131. i t governs the

relationship among the free-running oscillation frequency (w,) , injected frequency (wi, ) ,

and the phase shift of the output signal (Q - 4). The quantity, Aw,, in equation 2.14

is the locking range. The systeni is locked as long as the injected signal is within this

locking range. That is

Iwinj - W O ~ 5 Awm (2.16)

If the injected signal is out of the locking range, the system will lose lock and the

oscillator will oscillate at its own free-running frequency, w,. Xote that the locking range

in equation 2.14 is inversely proportional to the quality factor (Q), while the locking

range is proportional to the injection signal strength, p. For a large locking bandwidth,

the injected power should not be too low, and the VCO should have a reasonably low

qudity factor. Lastly, when the oscillator is locked t.o the injected signal, the phase noise

of the output signal follows the phase noise of the injected signal [9, 141.

Going back to the Adler's equation (Eq. 2-14), at steady-state. 2 = 0, and can be

rearranged as:

Equation 2.17 reveals the relationship between the injected frequency, wi,, and the

free-running frequency, wo, to the phase difference of the output and injected signals,

$J - d- Either by adjusting the injected frequency, wi+, or the free-running frequency,

wo, the phase difference between the injected and output signals can be changed. This

highlights the operating principle of the injection-locked phased arrays. In practice,

the injected signal is fixed, and the free-running frequency is changeci to achieve the

appropriate phase shift .

2.1.3 Phase Shifting Range

In equation 2.17, the inverse sine function gives two possible solutions for the phase

difference in an injection-locked system, but only one of them is correct. The true result

is found from a stability analysis on Adler's equation (Equation 2.14).

Let us assume that the output phase is perturbed by a small value, 64. Substituting

this new output phase #' = 4 + 6 9 into Adler's equation yields

d9 d - + z(69) = - Winj sin(@ - 4) COS(^@) - cos($ - 4 ) sin(&$) (2.19) dt

Since 64 is small, equation 2.19 becomes

Equation 2.21 tells us that the perturbation, 64, will decay (stable) as long as cos(+ -

4) remains positive. In other words, when q!~ - q5 lies in the range:

Hence, perturbation analysis shows that the range of the phase difference, S, - 9, is

limited to &90°, as s h o w in Figure 2.2.

Figure 2.2: Relative phase shift between the oscillator output and injected signal versus

the normalized injected frequency.

From this stability analysis, a clear disadvantage when using injection locking for

beam steering is the resulting Iimited phase sliifting range of f 90" (Figure 2.2). This

lirnited phase shifting range would in turn limit the beam steering angle of the phased

array. To illustrate this point, assume a maximum progressive phase shift of f 90" and

a fixed inter-element spacing of 0.54,. The maximum beam steering angle can be deter-

mined using the following equation (151:

4 sin 8, = - - kod

where q5 is the progressive phase shift between adjacent antemas (in radian), k, is the

free-space wavenumber, d is the inter-element spacing, and 0, is the main beam angle

from broadside. Substituting t hese values into equation 2.23 reveals t hat the maximum

beam steering angle is limited to f 30°. From this example, we infer that it is important

to design a compact active element and a compact feed network in order to achieve the

best scanning range.

Again, it is also worthwhile to note that in equation 2.14, the locking bandwidth is

inversely proportional to the quality factor Q of the resonant tank. The locking band-

width is usually about 1% of the output frequency (narrow range) for a high-Q oscillator,

making it difficult to achieve high tuning accuracy. However, equation 2.14 implies that

a low Q factor for the VCO is desirable because it would increases the locking range. For

unlocked oscillators, the syst em phase noise deteriorat es wit h a low-Q tank. Fortunat ely,

in a locked system, the phase noise of the output frequency depends on the phase noise

of the injected signal [9]. Therefore, the Q factor of the tank does not contribute to the

phase noise performance, rendering a low Q oscillator desirable in these systems.

2.2 Review of the Existing Injection-locked Scan Ar-

ray Architectures

2.2.1 Corporate Feed

Figure 2.3 shows a beam scanning array without phase-shifters. An injected signal is

used to lock the oscillators through a corporate feed network. Once al1 the oscillators

are locked, the phase shifting action is achieved by adjusting the VCO's control voltages

(wl, wz, ~ 3 , ~ 4 ) Assuming there is no mutual coupling involved, equation 2.17 suffi-

ciently describes the phase shifting property of this system. The main advantage of this

architecture is that a high radiated power can be achieved by means of a distributed

Corporate Feed Network -T-

Figure 2.3: Corporate feed network with injection locked VCOs.

array of low power ones. The phase-shifters and power amplifiers are eliminated in this

architecture, thus reducing the cost of the beam-steering array.

RI R 1 Patch Anteana (front) ~1 RI

FET oscillator (back)

Wilkinson power i A ' divider (front)

T Injection Signal rn œ - Z,=SOQ ~ = r o ~ Z?,,=iûûQ R l = l O O Q R2=200R

Figure 2.4: Layout of a corporate feed neta-ork in microstrip realized by Wilkinson power

dividers (161.

Figure 2.4 shows a practical microstrip implementation using a corporate feed net-

work, realized by G . Forma and J .M. Laheurte at 4 GHz [16]. Wilkinson power dividers

have been used for the corporate feed network, and the antenna element was based on

a CPW-fed active patch. An FET has been used to provide the oscillation. Proximity

coupling between the patch antennas and the microstrip feed lines provided a path for

locking the FET oscillators. The corresponding inter-element distance was about 0.75

at 4 GHz. With two eiements, a phase shift of -30" to 30° was observed experimentally,

while a phase shift of -10" to 10" was observed for a four-element array [16].

As implied above, as the number of antenna elements increases, the beam steering

angle decreases. This is because the phase difference between the injected arid output

s i s a l s is limited to a ma.uimum of 3~90". As the number of elements increases, the

progressive phase shift angle decreases. For example, for a tw*element array, the first

element can have a -90" phase shift and the second a 90" phase shift. This results in

an overall 180" progressive phase shift. However, for a three-element array, the maxi-

muin phase difference between the elements is reduced to 90°, and so on. The relation

between the maximum progressive phase shift and the number of element antennas for

this architecture is:

where N is the number of element antennas in the array.

To overcome this problem, a modified feeding method has been proposed by G. Forma

and J.M. Laheurte [16]. Two coupling microstrip lines have been used to feed the injected

signal to the oscillator: The first one has a length of L and the second a length of L + h/2

(see Figure 2.5). A switch is used to select the appropriate path for the injection signal.

With an additional 180" phase-shift provided by the second feed line, this technique gives

a complete 360" phase shifting range.

2.2.2 Unilateral Injection-locked Phased Array

Another architecture proposed by J. Lin, S. Chew, and T. Itoh [17] is shown in Figure

2.6. Instead of feeding the network in parallel, an injected signal locks only to the first

oscillator. The first oscillator in turn locks the second oscillator through an amplifier

and so on. Because the amplifiers isolate the direction of coupling, no backward coupling

occurs arnong the oscillators. Similar to the corporate architecture, once the oscillators

are locked in, the phase shift of each element c m be changed by adjusting the free

Patch Antenna (front) Switch /

FET oscillator (back) rT CPW Iine (back)

Figure 2.5: An extra 180" line with a switch is used to provide a coniplete 360° phase

shifting range [16].

Injected

Active ~ n t e n n a A+ 1- = 9û0

Amplifier

Figure 2.6: A Schematic for the unilateral feeding of the VCOs.

running frequency of each active antenna. Compared wit h the corporate architecture,

the highlight of this design is that the progressive phase shift has a range of f.90°,

independently of the number of elements in the array.

A microstrip implementation of this architecture is shown in Figure 2.7 [l7]. The

oscillation power is provided by a field effect transistor and the free-running frequency

is tuned by adjusting the DC bias voltage of the FETs. Unilateral coupling between the

elements is achieved by using FET buffer amplifiers. For this design, the inter-element

spacing arnong the antennas is 0.79A0 at an operating frequency of 6 GHz. Experimental

results demonstrated a beam scanning range of -27" to 6" by using a two-element array,

while a beam scanning range of -21° to 6 O was achieved by using a three-element array,

Figure 2.7: Layout of the unilateral feeding network by means of microstrip coupling

between the FET oscillators.

- i.e. a degradation of 6" in scanning range was observed in this latter case. The reasons

for this 6" loss in the three-element array are: (1) the difficulty in achieving unilateral

injection locking, (2) the fact that there are clifferences in phase shift ranges among the

antenna elements, and (3) because of additional phase shifts, contributed by the coupling

circuit and amplifiers [17].

As was mentioned before, compared wit h the corporate architecture, the advant age of

this design is that the progressive phase shift has a range of 3190"~ independently of the

number of elements in the array. However, the drawback of the unilateral architecture

is that a 360" phase shift is still not possible, and the cost associated is higher because

of the required extra amplifiers. Nevertheless, a bulky corporate feed network is avoided

which makes t his architecture more suitable for compact 2D array implementations

As mentioned in the previous sections, a f 90" progressive phase shift only provides

a maximum of f 30" scanning range with no grating lobes. To achieve a full phase

shifting range for this architecture, a dummy VCO can be placed in between each antenna

element (Figure 2.8). This dummy VCO can provide an additional 180" phase shift,

thus extending the phase shifting range to a full 360". This assumption was proved in

experiments carried out by K. W. Wong and A.K.Y. Lai [18]. In their paper, two locked

180° phase change 180' phase change

Injected Signal

360° phase change

Figure 2.8: A dummy VCO is used to achieve full 360" phase shiWing range [18].

oscillators are cascaded together. The results showed a phase tuning range of 285". The

80" loss is due to the phase delay of linking the two VCOs together and also because

the two oscillators had slightly different locking bandwidths [BI. The drawback of this

approach is the increased cost for the dummy VCOs. As well, certain arnount of power

is wasted just to provide the extra 180" phase shift. Furthermore, such a system is hard

to achieve because of geometric considerations. Indeed, it would be difficult to place two

VCOs within an inter-element spacing of, Say, 0.75X0at lOG Hz.

2.3 Oscillator Design

This section presents a quasi-linear oscillator design procedure based on the gain sat-

uration approximation [19]. Proposed by Johnson [20]; this design method allows easy

prediction of the large signal S-parameters of GaAs FETs based on their small signal

characteristics. Although less accurate, this approach saves time and cost associat,ed

with the equipment for large signal S-parameter measurements, while giving acceptable

results [19, 201. Finally, based on the predicted large signal S-parameters, an appropriate

embedding network can be calculated to provide maximum oscillator power [2 1, 221.

2.3.1 Prediction of the Large Signal S-Parameters of GaAs FETs

F'rom measured large signal S-parameter datasets, Johnson observed that only the mag-

nitude of S21 reduces significantly compared to the rest S-parameters [20]. He suggested

that , under compression, the energy is converted to harmonic frequencies and dissipated

in device and load resistances. The large signal effects produce primarily resistive changes

which decrease the device transconductance g, and thus, the transmission coefficient Szi.

Therefore, it is reasonable to assume that al1 S-parameters, except the magnitude of S21i

are constant under large signal.

The prediction of the large signal S-parameters is based on the power-gain saturation

characteristics of GaAs FETs [20], in which the output power of a FET is approximated

as follows:

where Go is the small signal gain, Lt is the output power, Pin is the input power, and

PSat is the saturated output power when the device is operated as an amplifier. For an

oscillator, the maximum output power occurs at the point of maximum (Pmt - P,,) , or

where dPouc/dPin = 1. Differentiating equation 2.25 with respect to Pi, and rewriting

the equation results in:

and,

For maximum oscillator power, the gain should be tuned for muxirnurn eficient gain,

G M E . At maximum efficient gain, the two-port added power of the amplifier is maxi-

mized. In terms of S-parameters, this maximum efficient gain is defined as:

where K is the Rollett stability factor.

Finally, dividing equation 2.27 with 2.26 gives:

P,t(max) Go - 1 GhIE(rnax oscillator v e r ) = - -

Pin (max) ln Go

Equation 2.31 relates the large signal gain wit h the small signal gain. The small signal

gain can be found from using the measured small signal S-parameters of the device and

equation 2.29. Once the small signal gain is found, the large signal gain can be obtained

from equation 2.31. Finally, the new value of ISzl 1 can be found by substituting GhlE

back into equation 2.29.

2.3.2 Embedding Elements of the Oscillator

Once the estimated large signal S-parameters are found, the embedding elements for the

oscillator can be calculated from the equations developed by Kotzebue et al. and Vehovec

et al. [21] [22]. The schematic for a general oscillator is shown in Figure 2.9. It comprises

of an active device, represented by two port large signal Y-parameters ( y i j = gij + j b i j ) ,

and a ?r embedding network (YK = GK + jBK). The associated voltage gain of the

device is defined as:

Note that in order to maintain consistency wit h the work developed in [21] and [22], large

signal Y-parameters for the device are adopted in this section. A simple transformation

Figure 2.9: Schematic for the general oscillator consisting of a nonlinear active device

and a reciprocal ir embedding network.

can be performed to convert the device's large signal S-paramaters to Y-parameters and

vice versa.

Refering to Figure 2 -9, the condition for oscillation is satisfied when:

These four currents can be expressed in terms of the Y-parameters of the device and the

ernbedding network. Substituting AVi for arrives at the following design equations in

matrix form:

Figure 2.10: Schematic for the shunt oscillator.

In general, the oscillator embedding network has only one resistive load. So, two of the

three conductances (Gi, GZ, G3) can be assigned to zero, and the other four embedding

elements may be determined. For our purposes, Gl and G2, are assigned zero, and the

schematic in Figure 2.9 is reduced to the shunt oscillator shown in Figure 2.10. The

equations for the embedding elements in Figure 2.10 are [21] :

where:

The voltage gain, A, is optimized for maximum oscillator power . The optimum voltage

gain is dervied in [22] and is given as:

Since equations 2.36-2.39 are given in terms of Y-parameters, a simple transformation

from the large signal S-parameters to Y-parameters is required to find out the embedding

element values of the oscillator.

Chapter 3

A Single Layer 'CPW-Fed Active

Patch Antenna

As discussed in Chapter 2, the novel phase-shifterless array design requires VCOs, anten-

nas, and a proper network for feeding the injection-locking control signal to the VCOs.

From a system design perspective, an overall compact design with a minimum number

of components is desirable for a low cost, practical design. Similarly, from an antenna

design perspective, physically isolating the radiator from ot her circuit ry (feed network

and active devices) is desired for low spurious radiation and low cross-talk effects. Shese

factors make active integrated antennas [23, 241 good candidates for irnplement ing the

unit radiating element of this phase-shifterless array. The active antenna described in

this Chapter is a self-running oscillator with the antenna utilized as both a radiator and

a resonator.

The concept of realizing such active antennas can be made possible by means of a

feedback loop, formed by twin dots electromagnetically (EM) coupled to a patch. Such

designs have been demonstrated in [8, 25, 261. This EM-coupling approach results to a

more compact layout with a reduced number of components. In addition, it offers the

advantage of a DC decoupled feedback loop, thus leading to a simplified bias network.

However, previously reported designs of this kind relied solely on microstrip technology

which requires a twdayer structure and the presence of via holes for providing the

necessary ground for DC bias [8, 25, 261.

In this Chapter, the design of a single layer, tweport CPW-fed patch active antenna

is presented. In the proposed approach, only a single layer substrate is required and no

via holes are necessary. In addition, the overall complexity and therefore the cost of the

structure is reduced. Furthermore, no degradation of performance has been observed,

despite the absence of an isolating ground-plane which is inherent in aperture-coupled

microstrip antenna implementations [8, 25; 261. The experimental setup used and the

corresponding results obtained are presented in this Chapter as well.

3.1 Act ive Ant enna Configuration

Patch (Front) /

Figure 3.1: Layout of the active antenna, E, = 2.33, h = 1.57 mm, WI = 0.8 mm, WC =

0.8mm, LI = lOmm, L, = 16mm, Ds = 12.5 mm, L, =31 mm, Li =72mm, La = 72

mm Dimensions for the 50 0 CPW: CPWsignal = 3 mm, CPW,., = 0.2 mm.

The proposed active antenna is shown in Figure 3.1. As shown, the front side of the

substrate hosts the patch whereas the active circuitry is accommodated at the back side

in 50 0, CPW technology. The patch antenna acts both as a radiator and a reedback

resonator for the oscillator. Two capacitively coupled coplanar slots are used at the back

side of the substrate to provide the necessary positive feedback. With reference to Figure

3.1, the longer centralized slot acts as a feeding slot to the patch [27] whereas the shorter

slot is situated at the edge of the patch to electromagnetically close the feedback loop.

3.2 Active Antenna Design

3.2.1 Choice of Fkequency

The frequency chosen here is 2.78 GHz. This frequency was chosen based on considering

the minimum realizable dimensions (about 0.2 mm at that time) for the passive network

and the frequency performance of the MMIC used.

3.2.2 Choice of Substrate

In order to suppress surface wave losses, an electrically thin substrate with low permit-

tivity was used. With a design frequency of 2.78 GHz, the active antenna was built on a

Rogers RT5870 substrate of E, = 2.33 wit h a thickness h = 1.57 mm, as shown in Figure

3.1.

3.2.3 Characterization of the MMIC

An HP-MGA-64135 MMIC has been used for this design. Although the S-parameters

of the MMIC are available by the manufacturer, these parameters are obtained in a

microstrip environment and will not be accurate when the MMIC is inounted in a CPW

environment. Therefore, a CPW-based Thru-Reflect-Line (TRL) calibration kit was built

on the actual substrate (RT5780), and the MMIC's S-parameters were measured with

the use of a HP8722C vector network analyzer (See Appendix A for a discussion of the

TRL techniques). The results are shown in Table 3.1.

Table 3.1: Measured S-parameters for the HP-MGA-64135 MMIC at a bias voltage of

1 Mag. S12 1 Pha. S12 (deg.) 1 Mag. S22 1 Pha. S 2 ~ (deg.) 1

1ov. -

3.2.4 Conditions of Oscillation

To achieve oscillation, the Barkhausen oscillation criterion must be met at the design

frequency [28]. In terms of S parameters:

Req. (GHz)

S21 (Passive Network) + SÎl (Ampli f i er ) = O (dB) (3-1)

Mag. S21

~(LoopGain) = 2n?r, where (n E integer) (3.2)

The Sll of the passive network and of the MMIC are designed to match to a 50 R

load.

Pha. S21 (deg.) Evlag. Sll

3.2.5 Overview of the Overall Design

Pha. SI1 (deg.)

The design of the structure is based on HP-ADS and HP-Momentum. First, the size of

the patch and the main feed-slot have been designed to achieve antenna resonance at

2.78 GHz, maximize the front-twback radiation ratio and yield a good return ioss on

the feeding CPW line. Subsequently, the magnitude of the transmission coefficient S21

has been adjusted, by the proper choice of the position (Ds) and geometry ( L I ) of the

offset slot (See Figure 3.1) , to meet the oscillation condition of equation 3.1 at 2.78 GHz.

Finally, the length of the intercomecting 50 R CPW line has been adjusted for achieving

the required 2n loop phase shift, dictated by equation 3.2.

3.2.6 Design of the Two-port CPW-Fed Patch Antenna

Patch (Front)

J, J h-

T - f CPW (Back)

Reference Plane ~eference Pisne Port 1 Port 2

Figure 3.2: Layout of the two-port CPW-fed patch antenna.

As shown in Figure 3.2, the two-port CPW-fed patch antenna has been simulated using

HP-Momentum, and a parametric analysis has been performed. The results of this

analysis are similar t.o the ones found in [29] and [25]. It was observed that the input

impedance of the main slot (SI1 of reference port 1) is heavily dependent on the width

CPWsignar, while the front-teback radiation ratio depends solely on the length of the

patch (L,) for a k e d substrate thickness [29]. In addition, the insertion loss (S2J is

dependent on the separation distance between the slots (D,) [25].

As an illustration, Figures 3.3 and 3.4 show the effect-s of varying the distance, D,,

m m -

oa- 1 lm

Figure 3.3: Effect of Ds on (a) the coupling strength d B (Szl) and (b) the return loss Si 1.

and the width, CPWsignal, on Sll and Szl, respectively. With reference to Figure 3.2,

Figure 3.3a shows that the coupling strength (S21 in dB) of the tweport CPW-fed patch

antenna decreases dramatically as the distance between the coupling slots increases. The

reasoning behind this result is that the magnetic-field under the patch decreases when

moving towards the radiating edges of the patch. Kence, the coupling strength between

the two slots decreases as LI, is increased. On the other hand, Figure 3.3b shows that

the variation of Ds has a rninor effect on the return loss, Sl1, of the structure.

To illustrate the effects of the width, CPWSignai, on the return loss Sll , three similar

aspect ratios (CPWsignal :CPWga, = 2:0.15, 3:0.2, and 4:0.25) representing the same

characteristic impedance of 50 0 are used for this analysis. With reference to Figure 3.2,

as the width of CPWsipal increases, the radius of the input irnpedance locus dramatically

increases (see Figure 3.4b), while the coupling strength S2i increases only moderately with

a longer C P Wsignai.

The result in Figure 3.4 can be explained by considering the distribution of the tan-

gential electric field, E, in the aperture of the main slot [29]. This tangential electric

field E is related to the magnetic currents, M = -2 x E, of the CPW line. This mag-

netic current M, in turn, is mainly responsible for the electromagnetic coupling with the

Figure 3.4: Effect of signal width CPli;i,nar on (a) the coupling strength and

(b) the return loss SI1.

magnetic-field under the patch [30]. For the main coupling slot (See Figure 3.5), the

tangential electric field mainly concentrates in the area of CPWsignal- Therefore, varying

CPWsignal changes the coupling surface of the aperture, leading to significant changes in

the return loss Sil of the structure. The length of the short slot Lf (see Figure 3.2) is

not as important as CPM'siw,l for this capacitivly coupled case.

The maximum front-teback radiation ratio occurs at the natural resonant frequency

of the patch antenna [29] [25]. Therefore, based on the cavity mode1 [15], the length

Figure 3.5: Magnitude of the tangential electric field of the main coupling slot

Figure 3.6: Simulated Eplane radiation patterns of the of the CPW-fed patch antenna

for different D. and CPWSQnai.

L, of the patch antenna was chosen to be 31 mm for a natural resonant frequency of

around 2.8 GHz. As demonstrated in Figure 3.6, the simulated front-t*back radiation

ratio when varying Ds, CPWsigmil and WC remains around 18 dB at a frequency of 2.78

GHz. This result indicates that varying the parameters D,; CPW,,,,,i, and WC has little

effect on the radiation pattern for the patch antenna.

Lastly, the remaining parameters (Wf, CIi,, LI, and L,) have a rninor effect on the

coupling strength and the return loss of the two-port CPW-fed patch antenna in Figure

3.2, and so they are used for the fine tuning of Sll and S21 of the structure.

Using HP-Momentum, the final dimensions of the CPW-fed patch antenna are listed

in Table 3.2, whereas the simulated S-parameters are shown in Figure 3.7. As shown in

Figure 3.7, the simulated Sll of this structure is -21 dB, whereas IS2]l is 9 dB. The IS2il

= 9 dB of the passive structure is deliberately set higher than the measured gain S21 of

the MMIC, which is 13 dB (See Table 3.1). This is a common practice to ensure that

Figure 3.7: Simulated S-parameters of the two-port patch antenna.

the oscillation wiH start.

Table 3.2: Designed dimensions for the two-port CPW-fed patch antenna.

3.2.7 Oscillation Test

The overall oscillation test has been performed using the schematic part of the HP-ADS.

The schematic of the overall structure is shown in Figure 3.8. The simulated S-parameters

of the two-port antenna are trsnsferred to HP-ADS, as well as the measured S-parameters

of the MMIC to be used. With reference to Figure 3.1 and equation 3.2, the simulated

CPW lengths required to start the oscillation are Li + LÎ = 54.9 mm, and Li + L2

= 144 mm. In other words, both lengths (54.9 mm and 144 mm) satisfy the condition

- CPW CPW2

2 - 1 t O h m S u a i t - ' + C F w S u a l - S c a r r - 2 Cr t r W r J nm

S u b i 1-"CPWSubl. ' 5 2 P S l o o - 3 . 5 G H z t-O. 2 mn P o i n t a i 2 0 1 L - 7 2 n m

SHP' ~ ~ i ~ ~ - / g i a a ~ / i a i ~ n n e / a e s ~ q n ~ ~ ~ r o - ~ r ~ / a o c / U ~ ~ C J ~ P '

Figure 3.8: Schematic

M o g n e t u o c o f t h e L o o p G a i n

of the oscillation test in HP-ADS

P h a s e o f t h e L O O P G a i n

m2 f r e 02. 7 8 0 C f i z

r n a g q o s c t c s t w i i

Figure 3.9:

m 1 f r e q = 2 . 7 8 0 C H z

b r a . . S ( t . ! > ) - 1 . 1 1 2 p h a s e ( o s c t e s t w l i b r a . . S ( 1 . 1 ) ) = - 0 . 2 1 2

Simulated results of the oscillation analysis.

(Back View) (Front View)

Figure 3.10: Photographs of the active patch antenna.

governed by equation 3.2. Due to layout considerations, the length of 54.9 mm is not

enough to connect the two-port antenna with the MMIC. Therefore, the length of 144

mm is used. Figures 3.8 and 3.9 show the simulation setup and the corresponding results

for the oscillation test. As shown, for a length of 144 mm, the simulated loop gain at 2.78

GHz has a magnitude of 1.112 and a phase of -0.21Z0, indicating that the oscillation

condition is satisfied at 2.78 GHz.

3.3 Experimental Setup and Results

The active antenna has been fabricated in the Etching Laboratory of the University of

Toronto, and corresponding photographs are shown in Figure 3.10. The circuit was biased

at 10 V DC and 50 mA by means of a DC bias low-pass filter, consisting of a couple of

chip capacitors and a lumped inductor, situated in a coplanar fashion at the output of

the MMIC. The RF-spectrurn was measured at a distance of 5m away from the active

antenna using a HP8563E spectrum analyzer connected to a receiving horn-anteniia and

Center freq = 2.7556GHz Res. B W = lOWz Video B W = 10Wz

Frequency (GHz)

Figure 3.11: Measured spectrum of the radiated RF power.

is shown in Figure 3.11. The oscillation frequency is found to be 2.76 GHz which is close

to the design frequency of 2.78 GHz. In addition, the phase noise of the oscillation is

rneasured manually using the same spectrum analyzer. With the resolution bandwidth

set to RBW = l k h , the sideband power %?as found to be Psideband = -57.5 dB at a 100

kHz offset with respect to the Parrie, = O dB carrier power. These results enable to find

the single-sideband phase noise of the oscillator, which is defined as the power ratio of the

single-sideband carrier to the carrier power spectral density at a certain frequency offset

away froni the carrier in a 1-Hz bandwidth [3?]. In quantitative terms, this becomes

The single-sideband phase noise is estimated to be -87.5dBc/Hz at a 100 kHz offset

from the carrier. It should be mentioned that this is inconsistent with a -100dBc/Hz

phase noise, measured using the phase noise software utility for the HP8563E spectrum

analyzer. So, the measured -100dBc/Hz is judged to be overoptimistic. Similar inaccu-

4 cross-pl J

Angle (Degree)

Figure 3.12: Measured (a) Eplane and (b) H-plane active radiation pattern.

racies, arising when using the same phase noise utility with free-running oscillators have

been observed and reported in [32].

The active antenna patterns of the device was tested in the anechoic chamber of

the University of Toronto. Figures 3.12a and 3.12b show the measured E and H-plane

radiation patterns. As shown, the worse cross-polarization appears in the Eplane but

does not exceed the level of 15 dB. On the other hand, the measured front-to-back ratio

has been found to be 12 dB. Both the rneasured cross-polarization levels and front-to-

back ratio compare favorably with those of the twdayer microstrip structures of [25, 81,

indicating no degradation of the performance of the proposed single-layer CPW approach.

It is noted that Figures 3.12a and 3.12b have considerable ripples in their radiation

patterns, which is a result from the drifting of the free-running oscillation. This problem

can be solved by locking the free-running signal with a stable reference signal, which will

be discussed in later chapters.

Finally, the effective isotropic radiated power (EIKP) has been measured, based on

the method described in [8]. The effective isotropic radiated power (EIRP) is given by

the following expression (331:

Meter m

Standard Gain Hom Standard Gain Hom

Figure 3.13: EIRP measurement setup (R = 5.015 m).

Active Antenna

Prec Xo EIRP = PtranSGtrans = -(P)-~ Grec 47&

m R

where Pt,,, and Pr,, axe the power transmit ted and received by the standard gain horn,

respectively. Gtra,, and Gr, are the gain of the standard gain horns, and R = 5.015 m

is the separation distance between the standard gain horn and the test antenna.

The rneasurement setup is shown in Figure 3.13. First, using the standard gain horn,

the signal generator transmits RF power at the frequency of interest, i.e. 2.76 GHz. The

transmitted power level presented to the standard gain horn, Pt,,,,, can be measured by

the power meter whereas the gain of the standard horn Gr,, can be obtained from the

manufacturer's datasheet. The transmitted EIRP can then be calculated frorn Eq. 3.4.

This value corresponds to the value read from the spectrum analyzer, Pspdrumo by the

foIlowing equation:

Specuum Anai yzcr

k = EIRP

PIpedrurn (3-5)

where k is a constant which represents the cumulative losses of the cable and free space.

The spectrum andyzer power, Pspedrumr is measured using the finest resolution band-

width. Finally, the transmitting standard gain horn is replaced by the active antenna,

and the EIRP of the active antenna is calculated from equation 3.5.

Using the above setup and procedure, an EIRP of 20 dBm was calculated from mea-

surements. The MMIC amplifier specifications indicate a typical output power of 12.5

dBm at the oscillating frequency. This implies an antenna gain of 20 - 12.5 = 7.5 dB

which is consistent with typical expected values for patch antennas.

3.4 Conclusion

A single layer active antema oscillator based on 50 R, CPW technology has been suc-

cessfully designed and tested at 2.76 GHz. The active antenna utilizes electrornagnet ic

coupling for closing the feedback loop. The structure leads to a layout with no via-

holes, a reduced component count and a simplified DC bias network. Nonetheless? a long

delay line is required for oscillation, which makes this design less attractive for array

applications.

The active antenna achieves an EIRP of 20 dBm, a front-teback ratio of 12 dB and

the cross-polarization level is better than -15 dB on the principal planes. On the other

hand, the measured phase noise is -87.5 dBc/Hz at a 100 kHz offset away from the carrier.

Chapter 4

A Compact Single Layer

Injection-Locked Active Antenna

In the previous Chapter, a single layer CPW-based active patch antenna unit-ce11 was

presented. The main advantages of the structure in Figure 3.1 are that no via holes are

necessarÿ for grounding, only a single layer is required, and that electromagnetic coupling

of the patch is used for closing the feedback loop, which further reduces the complexity of

the antenna structure. These benefits do not compromise the qudity of the performance

when compared to aperture coupled designs [8, 251.

However, the design of the previous Chapter requires a long delay line for adjusting

the phase of the feedback loop, thus forcing the device to be placed outside of the patch.

In addition, that structure does not feature injection locking capabilities. Bot h of t hese

deficiencies make the design less suitable for array implementations.

In this Chapter, a significantly improved design and an associated design methodol-

ogy are presented. In this improved version, the active element (GaAs FET) is centered

behind the patch antenna, thus making it suitable for array applications. In particular,

the GaAs FET is embedded in a CPW fed twin-slot arrangement, which is electromagnet-

ically coupled to a patch antenna resonator. Two open-circuited CPW stubs are utilized

for gate and drain matching. This proposed approach inherits the advantages from the

previous' design, with the added benefit of a compact size. In addition, a low power

injection signal serves to st abilize the oscillation t hrough parasitic coupling to the gate

of the FET at the CPW side of the circuit, thereby avoiding the need of using microstrip

couplers at the patch side 1161; this leads to lower parasitic and cross-polarized radiation.

In this Chapter, the configuration and the design methodology of the improved active

antenna element are presented, followed by the design of the injection locking feed net-

work. Next, the experimental procedure and the results obtained for the active antenna

are described, including the radiation patterns when the active element is locked a t 9.81

GHz, the phase-noise performance, the effective isotropic radiated power (EIRP), the

dc-terf efficiency and the locking range of the active antenna.

4.1 Active Antenna Configuration And Design

4.1.1 Choice of Fkequency

The frequency chosen here is 9.81 GHz, as opposed to 2.78 GHz previously used. This

frequency was chosen because the overall dimension for a corresponding array implemen-

tation will be smaller, which makes the etching process easier. Moreover, a new etching

technique was developed at that time which allows the fabricated microwave circuits to

work up to a frequency of 30 GHz [31]. Most importantly, it is envisioned that the oper-

ating frequency of future wireless communications systems wili keep increasing in order

to support broad bandwidth applications.

4.1.2 Configuration

The active antenna proposed here is depicted in Figure 4.1. As shown, the front side

of the substrate hosts the patch whereas the active circuitry is accommodated a t the

back side in 50 0 CPW technology. The substrate thickness is chosen to be 1.57 mm for

Square Patch (Front)

f - f MMIC (back)

Figure 4.1: Layout of the compact active antenna unit cell, E, = 2.33, h = 1.57 mm, L,

= 9 mm, L, = 9.5 mm, Ld = 13.6 mm, WC = 0.2 mm, W , = 0.4 mm, Dg = 2.3 mm,

CPWgigmr = 2.5 mm, CPW,,, = 0.1 mm, CPWlsignai = 3 mm, CPWl,,, = 0.2 mm,

L = 5 mH, bat, = -0.785 V, b r a i n = 3.5 V.

a good compromise among surface-wave excitation, bandwidt h, and front-to-back ratio.

The injection locked active antenna is designed to operate at 9.81 GHz using an ATF-

26884 GaAs FET from Hewlett Paclirtrd as the active element. Compared with the design

presented in the previous Chapter where the device was located outside of the patch. the

new design centers the device in between the coupling dots and behind the patch for

compactness. For matching, two open-circuited CPW stubs are used at the gate and

drain of the FET as shown in Figure 4.1. For DC biasing, two discrete inductors with L

= 5 mH soldered on the board with silver epoxy are utilized as RF chokes, as s h o w in

Figure 4.1. Furthermore, the injection locking signal is fed on a 50 52 terminated CPW

transmission line as also shown in Figure 4.1. Parasitic coupling between the locking

signal and the active antenna is achieved by comecting the injection signal line to the

open-circuited stub at the gate of the FET using a pseudeT-junction (see Figure 4.1).

Coupling at the gate of the active antenna has the advantage of lower power leakage

of the injection signal and thus lower parasitic radiation than when coupling to the

drain. In addition, this approach avoids the use of microstrip coupling to the patch-side

for implementing the injection locking feed network in contrast to [16]: which helps to

maint ain low cross-polarization and low parasi t ic radiation at broadside.

4.1.3 Design

Port 1 A

r FET

- XI

1 (Lc CPWsw, W 3

Figure 4.2: Optimum 9.81 GHz oscillator circuit, RL = 32.3 R, X l / w = 0.665 nH, X&

= 2.08 nH, X 3 / u = 0.628 nH.

Figure 4.2 shows the shunt feedback equivalent circuit of the overall structure. The

shunt embedding network for the FET is represented by a load resistance RL, and three

reactances XI, X2 and X3. The current source, Iinj , represents the injection signal. The

resistance, RL, and the reactance, X3, mode1 the radiation resistance and reactance of the

patch antenna respectively ; whereas Xi and Xz represent the cumulative reac tances of

the capacitive coupling slots and the open circuited CPW matching stubs. The physical

dimensions of Fig. 4.1, aifecting the elements of the equivalent circuit are marked in Fig.

4.2 for convenience.

The oscillator design follows the procedure outlined in Chapter 2 and [20, 21, 351

for calculating the shunt embedding network for the FET. Using the sarne procedure

described in Chapter 3, the small signal S-parameters of the FET were measured over a

frequency range from 9-11 GHz using a HP8722C vector network analyzer and a custom

made TRL calibration kit for the specific CPW environment. As mentioned in Chapter

2, reasonable values for the large signal S-parameters c m be estimated by appropriately

modifying the magnitude of the small signal SÎ1 while keeping the remaining small signal

parameters the same [21, 361. The estimated large signal S21 was obtained as follows:

First, using the small signal parameters of the device, the corresponding small signal

gain, Go, was calculated from the following equations [20] :

where K is the Rollett stability factor. Next, the maximum efficient gain at the point of

maximum oscillator power, GhfE, was calculated using the following expression:

Equation 4.3 has been derived in Chapter 2 of this Thesis using the approximated power-

gain saturation characterist ics of a FET power amplifier [20].

Once GkfE has been calculated, the large signal IS2iI was obtained by substituting Go

with GhIE into 4.1. In this design, the device has a measured small signal 1 Sm 1 = 7.8 dB

and a Go = 11.1 dB corresponding to a G M ~ = 6.7 dB and yielding a large signai 1 = 3.6 dB. With these large signal S-parameter values, the shunt feedback embedding

Figure 4.3: Layout of the slot-coupled patch, L, = 2.7 mm, W, = 0.4 mm, D, = 2.3 mm,

CPWsi,( = 2.5 mm, CPW,,, = 0.1 mm.

circuit, shown in Figure 4.2, was calculated fiom a set of equations given in Chapter 2

of this Thesis for achieving an optimum oscillation condition at the design frequency of

9.81 GHz [21]. With reference to Figure 4.2, the embedding elements were catculated

and are

Table

listed in Tabie 4.1:

4.1: Calculated values for the shunt feedback embedding circuit in Fiove 4.3.

After cdculating the embedding circuit from the measured S-parameters of the FET,

a distributed version of the embedding circuit was redized by the slot-coupled patch

and the two matching stubs at the gate and drain of the active device. The design of

the slot-coupled patch was based on HP-Momentum, a method of moments full-wave

electromagnetic planar solver. A symmetric structure for the slot-coupled patch was

adopted for a simple design and is shown in Figure 4.3.

First, using HP-Momentum, the dimension of the patch, L, (see Fig. 4.1), was

designed to achieve a condition close to antenna resonance at 9.81 GHz, and to yield

Figure 4.4: Schematic for the final passive structure, RL = 43.6 R, X3/w = 0.314 pF7

a front-to-back ratio of 24 dB [25, 291. In addition, a fùced distance of D. = 2.3 min

between the coupling slots was maintained for accommodating the FET symmetrically

in the center of the active antenna (see Figure 4.3). The S-parameters of the resulting

structure were then translated into the equivalent circuit shown in Figure 4.4, which

consists of the complex shunt radiation impedance, RL + jX3, and two shunt reactances,

Xslot, for representing the capacitance of the coupling slots. Subsequently, a paramet ric

analysis relating the dimensions L,, LI,, and W, (see Figure 4.3) with the equivalent

circuit in Figure 4.4 was undertaken, based on HP-Momentum. It was found that the

equivalent lumped elements RL, X3 and XscOt are less sensitive to variations of dimensions

D, and W, . On the other hand, there is greater sensitivity on the length of the coupling

dots L, (see Figure 4.3). For this latter case, the variation of RL, X3 and Xsi,t as a

function of L, is shown in Figure 4.5. Based on Figure 4.5, the length of the slots L, was

chosen such that the required embedding RL value of equation 4a was closely achieved.

In this example, a value of L, = 2.7 mm was chosen, yielding an RL = 43.6 R which

Figure 4.5: RL, X3 and X f i t for different slot lengths, L,

is close to the value of RL = 32.3 R required in Table 4.1. It was not judged prudent

to choose an even longer slot L, to further reduce the patch resistance RL, in order

to contain parasitic radiation from the dots and the CPw lines, and to maintain the

integrity of the CPW ground plane. F'urthermore, it should be noted that increasing the

patch dimension, L, above resonance, would swing the patch reactance X3 in Fig. 4.5

from capacitive to inductive as required in Table 4.1 but a t the expense of a longer patch

and a lower front-teback ratio. A size L, of 9 mm was chosen for this compact design.

Subsequently, the lengths of the two open-ended CPW stubs, represented by Xcat,

and in Figure 4.4, were optirnized in order to meet the oscillation condition.

Specifically) the length of the stubs were computed such that the reactances of the stubs,

XGate and Xorain, plus the reactances of the coupling slots, XSiot, become equal to the

calculated values, XI and X2, of the embedding network, i.e.:

Based on this procedure, the minimum possible lengths for the gate and drain stubs

were determined to be 9.5 mm and 1.1 mm, respectively. However, a one-half guided

wavelength line was added to the drain stub resulting to a value of 13.6 mm. Adding

this extra section to the drain stub avoids the unwanted coupling between the open

circuited stub and the patch antenna, at the expense of a larger active antenna size. The

overall structure was verified and fine tuned using the oscillation test in the schematic

part of HP-ADS. Finally, the layout of the pseudeT-junction (see Figure 4.1) for the

injection locking network was simulated using HP-Momentum, and a coupling level of

-12 dB has been obtained. Injection locking was applied at the gate stub of the FET via

the pseudo-T-j unct ion to rninimize the power leakage and parasitic radiation from the

injected signal through the active antenna. The layout of the injection locking network

is shown in Figure 4.1. As shown, the injection line is terminated to a 50 R resistor to

avoid standing waves dong the line.

4.2 Experimental Results

The active antenna was built on a Duroid 5870 substrate of E, = 2.33 with a thichess

of h = 1.57 mm, as shown in Figure 4.1. Air bridges were built on top of the CPW lines,

especially around the pseud*T-junction and the injection line, to suppress the parasi tic

slot-line mode. A free-running oscillation frequency of 9.817 GHz f 28 MHz has been

measured using a HP8563E spectrum analyzer. On the other hand, a HP83620B series

swept signal generator was used to provide a low noise injection signal to the active

antenna. When the active antenna was locked to the injection signal, a locking range of

10 MHz and 31 MHz was obtained for an injection power level of -10 dBm and O dBm,

respectively. The measured characteristics of the active antenna are summarized in Table

4.2. The RF-spectrum of the free-running and locked signals with a O dBm of injected

power as measured with the HP8563E spectrum analyzer are shown in Figure 4.6. Using

the same spectrum analyzer, the phase noise of the locked and unlocked signals has been

measured manually based on the following method: With the resolution bandwidth of

the spectrum analyzer set to RBW = 1 kHz, sideband power levels were measured and

averaged nt 100 kHz away hom the carrier. The following relation then applies [31],

Pnoise = Psideband - Parrie, - 10log(RBW) d B

Table 4.2: Summary of the measured characteristics of the active antenna.

I V* = 3 . 5 ~ I I 1 min 1 max 1

l Free Running Il I l I 1 Fkequency II f,,(GHz) 1 9.789 1 9.845 1

Locking Range

( ffree-runnins =9-81 GHz)

Using this procedure, the phase noise of the locked and unlocked signal has been

calculated to be -107.5 dBc/Hz and -63.28 dBc/Hz, respectively, both at a 100 kHz

offset away from the carrier. Note that the phase noise software utility for the HP8563E

is not reliable when measuring the phase noise of an unlocked active antenna because it

assumes that the frequency of the signal being examined does not drift in time, which

is not the case for an unlocked active antenna. It can be concluded that a cleaner

spectrum is observed when the active antenna is locked to a stable injection signal,

which is consistent with phase noise analysis of injection locking [14].

The active antenna patterns of the unit-ce11 when locked have been tested in the

anechoic charnber of the University of Toronto. Figures 4.7a and 4.7b show the measured

Figure 4.6: Measured spectrum of the fiee running and locked signals.

E and H-plane radiation patterns a t 9.81GHz for injection signal levels of O and 5 dBm,

together with the cepolarized patterns simulated using HP-moment m. As shown, the

worse cross-polarization appears in the H-plane but does not exceed the level of 15 dB.

On the other hand, the measured front-to-back ratio has been found to be better than

15 dB. Both the measured cross-polarization levels and the front-ti~back ratio compare

favorably with those of the two-layer microstrip structures of [25] and [8], indicating no

performance degradation of the proposed single-layer CPW approach. In addition, a close

comparison of Figures 4.7a and 4.7b indicates that the power of the injection signal did

not affect the radiation patterns, suggesting a low parasitic radiation from the injection

locking network.

Finally, based on the method described from the previous Chapter and in [8], the

effective isotropic radiated power (EIRP) of the locked patterns a t 9.81 GHz has been

measured to be 19.6 dBm, at a DC bias condition of Vdrain and Idrai* of 3.5 V and 28

mA, respect ively. From the measured radiation patterns, the directivity of the antenna is

estimated to be D = 8.3 dB resulting to an effective transmitter power, Peff = EIRP - D

of 11.3 dBm, and a dc-to-rf efficiency of 14%. These results are summarized in Table 4.3.

Figure 4.7: Measured (a) Eplane and (b) H-plane active radiation pattern at 9.81 GHz.

4.3 Summary and Conclusion

A compact, injection locked, single layer active antenna oscillator based on 50 R CPW

technology has been successfully designed and tested a t 9.81 GHz. This CPW-based

design eliminates the use of via-holes, and tightly integrates the FET device with the

patch antenna thus making it suitable for injection-locked phased-array applications. By

utilizing slot coupling from the device to the patch for closing the feedback loop, the

number of lumped element components is minimized, and the size of the unit-ce11 as well

as parasitic radiation are reduced. In addition, an injection locking path is established at

the CPW side through parasitic coupling to the gate of the FET, leading to low spurious

CHAPTER 4. A COMPACT SINGLE LAYER INJECTION-LOCKED ACTIVE ANTENNA

Table 4.3: Surnmary of the measured performance of the active antenna.

1 EIRP 1 19.6 dBm 1

leakage of the injected power.

Despite the tight packing of the device to the antenna, a modular design method-

ology has been presented which allows the implementation of the proposed layout in a

systematic way. The final measured results for the designed active antenna at 9.81 GHz

demonstrate clean E- and H-plane patterns, exhibiting a cross-polarization level better

than -15 dB. The active antenna achieves an EIRP of 19.6 dBm, a front-to-back ratio of

15 dB, and a locking range of 31 MHz. On the other hand, the measured phase noise

of the locked signal is -107.5 dBc/Hz at a 100 kHz offset away from the carrier. This

compact unit-ce11 design can be utilized as the building block in phase-shifterless beam

steering arrays.

Chapter 5

Compact Four-Element

Injection-Locked Scanning Array

In the previous chapter, a compact CPW-based single layer active antenna with injection

locking capabilities was presented. This design was extended to a compact linear scanning

array and is presented in this chapter. Next, the experimental procedure and the results

obtained for the array are described, including the scanned radiation patterns and the

effective isotropic radiated power (EIRP) .

5.1 Array Configuration And Design

5.1.1 Configuration

The layout and photographes of the four-element injection-locked array are shown in

Figures 5.1 and 5.2. The structure of the active antenna unit ce11 is based on the design

presented in Chapter 4 (See Figure 4.1). This unit ce11 features a simple and compact

design, while maintaining good radiation characteristics. An expanded view of the active

antenna element is shown in Figure 5.3. Note that the dimensions in Figure 5.3 are

different from those of Figure 4.1 in Chapter 4. This is due to a slightly different S-

square Patcb (Front)

Squnrc Paîcbes (Front)

4 vcO :r / / \

Figure 5.1: Layout of the injection-locked array, E, = 2.33, h = 1.57 mm, ACpW = 23.65

mm, A, = 30.52 mm, 70 R CPWsimal = 1 mm, 70 R CPW,,, = 0.2 mm, 59.5 R CPWsiwai

= 1.2 mm, 59.5 fl CPW,., = 0.1 mm, 50 R CPWsiwal = 1.26 mm, 50 R CPW,,, = 0.07

mm.

parameters measured in the new batch of ATF-26884 GaAs FETs used when compared

to what as originally reported in the previous chapter.

With reference to Figure 5.3, the front side of the substrate hosts the patch whereas

the active circuitry is accommodated at the back side in CPW technology. A Duroid 5870

substrate of e, = 2.33 with a thickness of 1.57 mm and an ATF-26884 GaAs FET froni

Agilent Technologies have been used in this design. For matching, two open-circuited

CPW stubs have been used a t the gate and the drain of the FET as shoan in Figure 5.3.

For DC biasing, two discrete inductors and capacitors with L = 5 mH and C = 47 pF

were soldered on the board with silver epoxy and were utilized as RF chokes, as sliown

in Figure 5.3. A four-element array has been built using the active unit ce11 described

previously (See Figure 5.1). For a series design, a single 70 0 CPW transmission line was

used to feed the injection signal as also shown in Figure 5.1. Series feed was chosen to

distribute the injection signal in order to maintain a compact configuration (as opposed

to a corporate feed using bulky power dividers). Parasitic coupling between the locking

signal and the active antenna was achieved by connecting the injection signal line to the

(Front View) (Back mew)

Figure 5.2: Photographs of the injection-locked array.

open-circuited stub at the gate of the FET using a pseudo-T-junction (see Figures 5.1

and 5.3). Air bridges were built on top of the CPW lines, especially around the pseudo-

T-junction and the injection line, to suppress the parasitic dot-line mode. To maintain a

zero phase difference of the injected signal arnong the active antennas, the length of the

interconnecting CPW line between two adjacent patches was kept to about one CPW

wavelength X C p W . This corresponds to an inter-element spacing of roughly 0.78 free-

space wavelengths A,. A 59.50 quater wavelength transformer was used to match the

70Cl line with the 50Q CPW line and the external injection locking cable. Due to the

soldering tolerances and the S-parameters variation of the discrete GaAs FETs used,

the active elements were tested individually for determining ttieir free-running frequency

range. A given FET was replaced with a new one if its free-running frequency range

was not suitable. This procedure was repeated until a satisfactory frequency range was

obtained for each FET.

5.1.2 Design

The oscillator design for each active element follows the procedure outlined in the pre-

vious chapter. A new batch of ATF-26884 GaAs FETs was used for this design, which

Figure 5.3: Detailed layout of the active element, L, = 9.21 mm, Lg = 8 mm, Ld = 10.7

mm, WC = 0.2 mm, W, = 0.4 mm, D. = 2.3 mm, CPWsipal = 2.3 mm, CPWg., = 0.2

mm, L = 5 mH, C = 47pF.

exhibits slightly different S-parameters than what was originally reported from the pre-

vious Chapter. Consequently, the dimensions of the active elements were modified to

compensate for this effect. The lengths and widths of the CPW stubs were adjusted to

yield a free running oscillation frequency of 9.83 GHz (See Figure 5.3).

The design of this novel phased array is based on Adler's equation of injection locking

described in Chapter 2 [13]. When the injected signal locks to the free-running signal of

the oscillator, a phase difference A4 can be created between the oscillating signai and

the injected signal and is related by the following equation (Refer to chapter 2):

where Ui, is the injected signal frequency, w, is the free-running frequency of the oscil-

lator, and 2Awm is the locking bandwidth. According to theory presented in Chapter 2,

the maximum possible phase difference is f !JO0.

Rom Figure 1, the array is a conventional series feed injection Iocking array similar

to the corporate feed network described in Chapter 2 of this thesis. As mentioned in

Chapter 2, the maximum progressive phase shift (A@) between two adjacent elements

that c m be achieved by this type of array is

where A&lma, = 90' is the maximum phase difference between the oscillating signal and

the injected signal, and m > 2 is the number of elements in the array. When the active

elements are locked and radiate at the maximum progressive phase shift Ac$(max), the

corresponding maximum scanning angle for the array is

where Xo is the free space wavelength and d is the antenna inter-element spacing. For

the design of Figure 5.1, d = 0.78A0 and m = 4, yielding a maximum progressive

phase shift and a corresponding scanning range of A@(max) = f 60° and Orna, = f l2.4",

respect ively.

5.2 Experimental Setup and Results

5.2.1 Radiation Patterns

The active antenna patterns for each element have been tested in the anechoic charnber

of the University of Toronto. A Wiltron 360SS69 sweep generator was used as the source

for the control injection signal (See Figure 5.1). The injection power presented at the

input 50 Tt CPW line is -4.5 dB. The E and H-plane radiation patterns and the phase

noise of the elements have been individually measured and found similar to the ones

reported in Figures 4.7a and 4.7b of Chapter 4.

180

Figure 5.4: Measured H-plane pattern of the array at broadside.

Figure 5.5: Measured H-plane scan pattern at the "left" edge of the tuning range at 9.83

GHz.

Figure 5.6: Measured H-plane scan pattern at the "right" edge of the tuning range at

9.83 GHz.

Subsequently, the active radiation patterns of the array have been measured and are

shonm in Figures 5.4, 5.5 and 5.6. The phase shifts for the active elements are achieved by

varying the gate DC b i s voltages of the GaAs FETs. A measured scan angle from -12"

to 9.5" is observed as shown in Figures 5.5-5.6. It is believed that the interconnecting

feedline in Figure 5.1 is not exactly equal to 1 X C p i ~ and thus adds a constant phase

shift to the antenna elements, resulting in an asymmetric scanning range. This 1.02 A,

feedline provides an extra - ïO phase shift to the antenna elements. From equation 5.3,

the maximum progressive phase shift between adjacent antenna elements is calculated

to be AO(max) = 53" and -67". This leads to an asymmetric scan angle of O,,, = 11"

and -13.g0, respectively, which is in good agreement with the measured scan angle of

9.5" and - 12". On the other hand, the highest cross-polarization level of the array was

found to be below 23 dB at broadside as shown in Figures 5.4, 5.5 and 5.6.

Active A r n y S u d a r d Cain Horn I

Test signal (Reccive Cable) I 1

finj

1

I ' ',,- Tut signal (Reccive Cable)

I Wiltmn 3630Al - 1 R e a i v t r Rcfericnce signal

Ta Ft,

F R

1 I I

Test Positioaer I 1

Pm

L 1

Cwpler Wiltron3630~[ r Rccviver 1 Rcfcrcnccsignîl

Figure 5.7: EIRP measurement setup (R = 5.015 m).

Wilrrcm 36ûSS69 Fmqucncy Swccpcr

5.2.2 EIRP Measurement Setup and Results

T a % (5 )

Since the direction of the main bearn cannot be known precisely - a priori, the EIRP

measurement setup described in Chapter 3 is not suitable for measuring the EIRP for

the array. Therefore, a modified measurement procedure (see Figures 5.7a and 5.7b)

was employed in order to determine the EIRP of the array as well as the EIRP of the

individual elernent S.

For convinence, the expression of EIRP (equation 3.4) is again stated here.

With reference to Figure 5.7a and equation 5.4, the received power, Pr,, and the gain of

the standard horn antenna, Grec, must first be determined in order to find the EIRP of

the array. Using the tweantenna method [15], the gain of the standard horn antenna has

been measured to be Grec = 12.55 dBi f 0.15 dBi within a frequency range of 9.7-9.9 GHz.

This measured result compares favourably with the data provided by the manufacturer

(12 dBi Q 10 GHz).

Figures 5.7a and 5.7b show the EIRP measurement setup for the active array. First,

with the setup shown in Figure 5.7a, the radiation pattern of the active array has been

measured a t 9.83 GHz by means of the ratio Ta: Ra = Parmg. Next, using the calibration

setup s h m in Figure 5.7b, the through measurement of the received cable (Pthtu =

T,:Ra) has been measured at a frequency of 9.83 GHz. Next, the power received by the

standard gain horn was calculated by the expression:

Finally, the EIRP of the active array has been calculated by substituting equation 5.5

and the worse case antenna gain (Gr, = 12.7 dBi) to equation 5.4.

Based on the method described above, the rneasured EIRP for the single elernent is,

on average, 17.0 dBm. On the other hand, the measured EIRP of the patterns in Figures

5.4 to 5.6 are 29.8 dBm f 0.5 dB, at an average DC bias condition of Vdrain = 3.5 V and

Idrai* = 115 mA. From the measured radiation patterns, the directivity of the array is

estimated to be D = 15 dB, resulting in an effective transmitter power, P, = EIRP - D

of 14.8 dBm. These results for the single element and the array are summarized in Table

Table 5.1: Sunimary of the measured characteristics of the single el ement and the array.

II Single Element (ave.) Array (ave.)

9.83 GHz

29.8 dBm

-- -

9.83 GHz

17 dBm

finj

EIRP

T-

5.3 Summary and Conclusion

A phase-shifterless beam scanning array based on a compact uniplanar unit-ce11 has been

demonstrated. A maximum beam scanning range from -12O and 9.5* has been achieved

which is in good agreement with the theoretical range of -12.4" to 12.4O. The tight

packing nature of this design makes it well suited for 2D array implementations.

Chapter 6

Conclusions and Future Directions

6.1 Conclusions

In this Thesis, two active antenna oscillator designs have been examined, built and ex-

perimentally evaluated. Based on a uniplanar unit-ce11 presented in Chapter 4, a phase-

shifterless beam scanning array has been demonstrated. The active element of tliis array

is a compact, injection locked, single layer antenna oscillator based on 50 0 CPW tech-

nology. By utilizing slot coupling from the device to the patch for closing the feedback

loop, the nuniber of lumped element components is minimized, and the size of the unit-

cell, as well as parasitic radiation are reduced. In addition, an injection locking path is

established at the CPW side through parasitic coupling to the gate of the FET, leading

to low spurious leakage of the injected power. This design eliminates the rise of via-holes,

and tightly integrates the FET device with the patch antenna. A maximum beam scan-

ning range from a four-element linear array of -12" to 9.5O has been achieved which is

in good agreement wit h the theoretical range of - 12A0 to 12.4".

CHAPTER 6. CONCLUSIONS AND FUTURE DIRECTIONS

Reference plane 1 Reference plane 2

Figure 6.1: Layout of the CPW-fed rectangular patch.

6.2 Future Direct ions

Future research directions of this work include improving the DC-teRF efficiency of

the oscillator, increasing the bandwidth of the patch antenna, increasing the maximum

steering angle of the array and eventually implement 2-D scanning arrays.

6.2.1 High DC-to-RF Efficiency Oscillator

The presented unit cells in Chapters 3 and 4 have not been optimized for high DC-to-RF

efficiency. Higher efficiency can be achieved by operating the underlying amplifier in a

class E mode [37] [38]. It should be pointed out that due to the CPW slot coupling used

at the center of the patch (see Fig. 6.1), the input ernbedding impedance presented to

the device at second harmonic closely corresponds to an open-circuit. This assertion has

been verified both using HP-Momentum as well as experimentally and it is the primary

requirement for class E operation (see Fig. 6.2) [39].

CHAPTER 6. CONCLUSIONS AND FUTURE DIRECTIONS 64

( Design frequency [x] = 2.585 GHz 1 znd harmonic [O] = 5.165 GHz

O

S. I~GHZ.-O.UW~ dB 6.395GHr-1333 dB

4 -

-10 - g 85GHr -1 1.83 dB - - .1s - cn

.20 -

.2s - 6.98CHr-24.14 dB . - 2 3

Figure 6.2: bleasured harmonic response of the patch of Figure 6.1 (a) at reference plane

2 and (b) a t reference plane 1.

6.2.2 Enhanced Bandwidth

In addition to high efficiency, wider bandwidth is desirable because this leads to: 1.

Higher capacity for wireless communications; 2. Easier synchronization and locking of

the active elements for array irnplementat ions.

Due to the fact that patch antennas are inherently narrowband, the active elements

presented in Chapter 3 and 4 also acquire a narrow bandwidth. Therefore, to increase

the bandwidth of the active elements, the existing patcli antenna structure needs to be

modified. One possibility would be to add parasitic elements next to the main radiating

patch, with slightly different lengths and widths [40, 411. This will introduce multiple

resonant frequencies close to the original resonant frequency, enhancing the bandwidth of

the patch antenna while maintaining a clean radiation pattern. However, t his approach

potentially increases the size of the unit ce11 (t hus reducing the maximum scan-angle) .

Another possibility for wider bandwidth would be to stack another microstrip an-

tenna, with slightly different dimension, on top of the original one (see Figure 6.3) [42].

CHAPTER 6. CONCLUSIONS AND FUTURE DIRECTIONS

CPW/Ground plane

Figure 6.3: Stacked patches for increasing the operating bandwidt h.

Again, the additional microstrip antenna introduces another resonant frequency close to

the original one, which increases the frequency bandwidth in excess of 20% for a return

loss of below 10 dB [42].

6.2.3 Increased Steering Angle using a Unilateral Injection Lock-

ing Architecture

The linear phase-shifterless array presented in Chapter 6 exhibits a maximum bearn

scanning range from -12" and 9.5". The scan angle can be increased by modifying

it based on the unilateral locking architecture presented in Chapter 2. According to

equation 2.23, the theoretical scan angle can be increased to & 18.ï0, independently of

the number of elements in the array.

To further increase the scan angle, the inter-element spacing can be decreased to 0.5Xo

for obtaining scan angle of f 30". This is feasible due to the cornpactness of the unit cell,

although caution must be exercised to account for the increased mutual coupling among

the patches.

Contributions

K.H.Y. Ip, T.M.Y Kan, and G.V. Eleftheriades, "A single-layer CPW-fed active

patch antenna," IEEE Microwave and Guided Waves Letters, vol. 10, pp. 6466, Feb.

2000.

K.H.Y. Ip and G.V. Eleftheriades, "A compact CPW-based single layer injection-

locked active antenna for array applications," To appear in the Feb. 2002 issue of the

IEEE Transactions on Microwave Theory and Techniques, URL:

http://www. waves.utoronto.ca/prof/gelefth/jpub/index. html.

K.H .Y. Ip and G.V. Eleftheriades, "A compact single layer injection-locked linear

scanning array," Accepted to the IEEE Mircowave and Wireless Components Let ters.

lSt prize winner in Canadian Student Cornpetition of XXVIth General Assembly of

the International Union of Radio Science.

Appendix A

Network Analyzer Calibration:

Thru-Reflect-Line

The function of a vector network analyzer (VNA) is to measure the S-parameters of

microwave circuits and devices. To make these nieasurements, the device must be con-

nected to the vector network analyzer using connectors, cables and possibly transitions.

However, these connecting components introduce measuring errors (in terms of phase

delay, losses, and mismatches) which must be accounted for prior to measuring the de-

vice of interest. Figure A.1 illustrates a comrnon representation of a network analyzer

measurement for tweport devices. The error introduced by the connection components

and by the network analyzer are lumped together in the two-port S-parameters error

boxes A and B, while the device under test (DUT) is represented by its own two-port

S-parameters as shown in Figure Al. To accurately measure the device (at the reference

planes of the DUT), a calibration procedure is used to characterize the error boxes be-

fore measureing the DUT. This allows that the effects of these errors are removed from

subsequent measurements.

I 1 I I

Rcfcrmcc Refmnce Refercncc Referrncc pl- for plane for plane for plme for V~ 'A pon 1 & v i a pon 1 & v i a pon 2 VNA pon 2

Figure A.l: (a) Block diagram and (b) signal Aow graph for a tweport measurement.

A. 1 Short-Open-Load-Thru (SOLT) Calibration

One way to calibrate a network analyzer is to connect three or more known loads (such as

shorts, opens and matched loads) to the network analyzer for calculating the S-parameters

of the error boxes. For a Short-Open-Load-Thru (SOLT) calibration, four different con-

nections are utilized to determine the S-parameters of the error boxes. These four con-

nections are shown in Figure A.2 in terms of signal flow graphs. The Short connection is

made by connecting a short to ports 1 and 2. The Open connection is made by connect-

h g an open to ports 1 and 2. The Load connection is made by connecting a matched

ioad to ports 1 and 2. Finally, the Thru connection is made by connecting port I to port

2. After determining the S-parameters of the error boxes, a simple translation can be

performed from the network andyzer's reference plane to the DUT's reference plane.

The main disadvantage of the SOLT calibration scheme is that it is generally difficult

to realize perfect opens and shorts at microwave frequencies. For example, a stray capac-

itance of the open circuit connector always exists. The value of this parasitic capacitance

must be known and incorporated during the de-embedding calculations. Therefore, most

of the commerical SOLT calibration kits corne with cal kit definition files which contain

(.I

1 '(m-b bl

II I I Ji* II a:

Figure A.2: Signal flow graphes for (a) short (b) open (c) matched load and (d) thru

connection.

t hese parasitic values for accurate rneasurements.

Another disadvantage of SOLT calibration is that while the commericd SOLT calibra-

tion kits are readily available for coaxial environment, they are rarely found for microstrip

or CPW environments. The reason for this is that microstrip and CPW environments

allows different substrate permittivities and thicknesses which make the a-priori charac-

terization of the parasitics difficult . Therefore, different calibration schemes are needed

for microstrip and CPW environments.

A . l . l The HP8722C VNA

Similar to the error box model in Figure A.l , the ifP8722C VNA uses a 12-term error

model for the two-port error correction. For reference purposes, the 12-term error model

is shown in Figure A.3 and the associated error corrected S-parameters for the DUT are

shown in equations A. 1-A.4 [43].

Figure A.3: (a) Forward model (b) reverse model signal flow graphes for HP SOLT

calibration.

A.2 Thru-Reflect-Line (TRL) Calibration

Similar to the SOLT calibration, the Thru-Reflect-Line calibration requires the use of

t hree connections to completely characterize the error boxes (See Figure A.4) [44]. For

Figure A.4: (a) Thru (b) Reflect and (c) Line signal flow graphes for the TFU calibration.

a Thru connection, port 1 and port 2 are connected directly. For a Reflect connection, a

high impedance load, ïL, is connected to ports 1 and to port 2. The value of rL need not

to be an exact open or short. Finally, for a Line connection, a matched line, e-T', with

a phase of about ,414 is connected between port 1 and port 2. Again, the phase of tliis

matched line need not to be exactly X/4. The main advantage of the TRL calibration is

that this scheme does not rely on known standard loads. This makes the TRL calibra-

tion scheme favourable for microstrip and CPW environment. The disadvantage of this

scheme is that the measurement bandwidth is narrow. If a large measurement bandwidth

is needed, anot her similar calibration technique called m-LRL (multi-Line-Reflec t-Line)

can be used [45].

A.2.1 The HPS722C VNA

As oppose to the complete four-channel receiver architecture, the HP8722C VNA is a

three-channel receiver architecture and supports a modified TRL* calibration scheme

(Not true TRL, see Figure A.5). This TRL* scheme assumes that the source and load

Reference planc for VNA pon 1

Reference plane for VNA port 2

- a- - Reference olme for

- A- - Rcference plane for . -

VNA port 1 VNA port 2

(b)

Figure A.5: (a) Three-channel receiver and (b) four-channel receiver architecture in HP

series network analyzers.

matching coefficients of the test ports are equal (i.e. t here is port-impedance symmetry

between forward and reverse measurements) . While this assumption is true for the four-

channel receiver VN A, t his is only a fair assumption for t hree-channel receiver network

analyzer. As a result, the HP8722C VNA gives less accurate measurements for the TRL*

scheme [43].

The requirements for the thru, reflect , and line calibrating components can be found

in [46]. The main requirements for the HP8722C TRL calibration kit are listed below-.

Thru: a simple connection of the two ports directly (zero length line).

Reflect: a short, or near short circuit reflection load.

Line: a matched line with a length of X/4 at midband.

A.3 Implementation of TRL in CPW environment

Following the requirements listed from the previous section, custom-made TRL standards

for a CPW environment were made. The dimensions are shown in Figure A.6, and the

measured SI1 for the ATF26884 GaAs MESFET is shown in Figure A.7. As shown, a

magnitude ripple of about 0.5 dB exists in Figure A.7a as well as a phase ripple of about

CPW

Figure A.6: (a) Thru (b) Reflect and (c) Line realized in CPW environment, CPWsignal

= 1.9 mm,CPW,., = 0.8 mm, LmI = 8 mm, LLine = 3.13 mm, e, = 2.33, h = 1.57 mm.

3". These ripples are mainly caused by the imperfection for the TRL* scheme for the

HP8722C VNA.

0.5 dB ripple -5

f r c q , GHz f r e q . GHz

Figure A.7: Measured Sii of the ATF-26884 GaAs FET using the custom-made TRL

cdibrat ion kit.

Bibliography

[1] B. Pattan, Robust Modulation Methods and Smart Antennas, Prentice Hall, New

Jersey, 2000.

[2] R.C. Hansen, Phased Array Antennas, Wiley, New York, 1997.

[3] Atul Jain, "Applications of millimeter-wave radars to airport surface surveillance,"

in IEEE Digital Avionics Systems Conference, Oct. 1994, pp. 528 - 533.

[4] P.L. Lowbridge, P.M. Brigginshaw, and B. Kumar, "Millimet re wave technology for

collision avoidance and cruise control," in IEEE Automotive Electronics Conference,

June 1991, pp. 150 - 154.

[5] R.A. York and T. Itoh, "Injection- and phase-locking techniques for beam control,"

in IEEE Bansactions on Microwave Theory and Techniques, Nov. 1998, vol. 46, pp.

1920-1929.

[6] M.I. Skolnik, The Radar Handbook, McGraw-Hill, New York, 2nd edition, 1990.

[7] P.L. Liao, and R.A. York, "A new phase-shifterless beam-scanning technique using

arrays of coupled oscillators," in IEEE Transactions on Mic.r.owave Theory and

Techniques, Oct. 1993, vol. 41, pp. 1810-1815.

[8] J.F. Zürcher and Y. Brand, "An active stripslot-foam inverted patch antenna," in

Microwave and Optical Technology Letters, Oct. 1996: vol. 13, pp. 114-119.

[9] K. Kurokawa, "Injection-locking of solid-state microwave oscillators," in Proc. IEEE,

Oct. 1973, vol. 61, pp. 1386-1409.

[IO] K.D. Stephan, LLInter-injection-locked oscillators for power combining and phased

arrays," in IEEE Transactions on Microwaue Theory and Techniques, Oct. 1986,

vol. 34, pp. 1017-1025.

[1 l] D.M. Pozar, Microwaue Engineering, Wiley, New York, 2nd edition, 1998.

[12] R.A. York and Z.B. PopoviC, Active and Quasi-Optical Arrays for Solid-state Power

Combining, Wiley, New York, 1997.

1131 R. Adler, "A study of locking phenornena in oscillators," in Proc. IEEE, Oct. 1973,

vol. 61, pp. 1380-1385.

[14] H.C. Chang, X. Cao, U.K. Mishra, and R.A. York, "Phase noise in externally

injection-locked oscillator arrays," in IEEE Transactions on Mzcrowave Theoq and

Techniques, Nov. 1997, vol. 45, pp. 2035-2042.

[15] C. Balanis, Antenna Theory, John Wiley Sons Inc., New York, 2nd edition, 1997.

[16] G. Forma and J.-M. Laheurte, "Design for injection-locked oscillator arrays," in

Electronics Letters, April 1998, vol. 34, pp. 683-684.

[lï] J. Lin, S.S. Cliew, and T. Itoh, "A unilateral injection-locking type active phased

array for beam scanning," in IEEE MTT-S Int. Microwave Sgmp. Dzg., May 1994,

vol. 2, pp. 1231-1234.

[18] K. W. Wong and A.K.Y. Lai, "Phase tuning beyond 180 degrees by injection-locked

oscillators," in IEEE Tencon'93/Beijing, 1993, vol. 2, pp. 1-4.

[19] A M . El-Tager, E.A. Abdallah, A. Abdel-Nazir, "Design of microwave oscillators

using linear, quasi-linear and nonlinear methods," in 1999 Asia Pacific Microwaue

Conference, Nov. 1999, vol. 2, pp. 464-467.

[20] K.M. Johnson, "Large signal GaAs MESFET oscillator design," in IEEE Transac-

tions on Microwave Theory and Techniques, March 1979, vol. 27, pp. 217-227.

[21] K. L. Kotzebue and W. J. Parrish, "The use of large signal sparameters in mi-

crowave oscillator design," in Proceedings 1975 Int. Microwave Symp. on Circuits

and Systems, 1975.

[22] M. Vehovec, L. Houselander, and R. Spence, "On oscillator design for maximum

power," in IEEE 7Fansactions on Circuit Theory, Sept. 1968, vol. 15, pp. 281-283.

[23] K.C. Gupta and P.S. Hall, Analysis and Design of Integrated Circuit Antenna Mod-

ules, Wiley, New York, 2000.

[24] J. Lin and T. Itoh, 'LActive integrated antennas," in IEEE Transactions on Mi-

crozuaue Theory and Techniques, Dec. 1994, vol. 42, pp. 2186-2194.

1251 W. J. Tseng, and S.J. Chung, "Analysis and application of a two-port aperture-

coupled microstrip antenna," in IEEE Transactions on Microwave Theory and

Techniques, May 1998, vol. 46, pp. 530-535.

[26] H. Shiomi, K. Hayata, Y. Kido, and S. Kawasaki, "Quasi-optical patch resonator

antenna array," in Asia Pacifie Microwave Conference Proceedings, 1997, vol. 3, pp.

945-948.

[27] L. Giauffret, J.M Laheurte, and A. Papienik, "Study of various shapes of the cou-

pling dot in cpw-fed microstrip antennas," in IEEE 7kansactions on Antennas and

Propagation, Apr. 1997, vol. 45, pp. 642-647.

[28] A.S. Sedra and K.C. Smith, Microelectronic Circuits, Oxford University Press, New

York, 3rd edition, 1991.

[29] L. Giaufhet and J.M. Laheurte, "Parametric study of the coupling aperture in

cpw-fed microstrip antennas," in IEE Proceedings - Microwaves, Antennas and

Propagation, June 1999, vol. 146, pp. 169-174.

[30] L. Giaufhet and J. M. Laheurte, "Theoretical and experimental characterisation of

cpw-fed microstrip antennas," in IEEE Proceedings - Microwaves, Antennas and

Propagation, Feb. 1996, vol. 143, pp. 13-17.

[31] R.W. Rhea, Oscillator Design and Cornputer Simulation, McGraw-Hill, New York,

2nd edition, 1997.

[32] A.R. Brown, High-Q integrated nicrornachined components for a 28 GHz fiont-end

transceiver, Ph. D. thesis, University of Michigan, Ann Arbor, 1999.

[33] M. Gouker , "Toward standard figures-of-merit for spatial and quasi-opt ical power-

combined arrays," in IEEE Transactions on Mz'crowave Theory and Techniques, Jul.

1995, vol. 43, pp. 1614-1617.

[34] A. Grbic, "A novel leaky millimeter-wave linear slot antenna array," h1I.A .Sc. Thesis,

University of Toronto, 2001.

[351 B.K. Kormanyos and G.M. Rebeiz, "Oscillator design for maximum added power,"

in IEEE Microwave and Guided Waves Letters, June 1994, vol. 4? pp. 205-207.

[36] P. Liao, and R.A. York, "A six-element beam-scanning array," in IEEE Microwave

and Guided Wave Letters, Jan. 1994, vol. 4, pp. 20-22.

[37] E. W. Bryerton, W.A. Shiroma, and Z.B. Popovié, "A 5-ghz high-efficiency class-e

oscillator," in IEEE Microwave and Guided Waves Letters, Dec. 1996, vol. 6, pp.

441-443.

[38] V. Radisic, Y. Qian, and T. Itoh, "Novel architectures for high-efficiency ampli-

fiers for wireless applications," in IEEE Pansactions on Mzcrowave Theory and

Techniques, Nov. 1998, vol. 46, pp. 1901-1909.

[39] M.D. Weiss, and Z. PopoviC, "A 10 ghz high-efficiency active antenna," in 1999

IEEE MTT-S Int. Microwave Symp. Dig., June 1999, vol. 2, pp. 663466.

[40] C.K. Aanandan, P. Mohanan, and K.G. Nair, "Broad-band gap coupled microstrip

antenna," in IEEE ïkansactions on Antennas and Propagation, Oct. 1990, vol. 38,

pp. 1581-1586.

1411 J.F. Zürcher and F.E. Gardiol, Broadband Patch Antennas, Artech House, Boston,

1995.

[42] F. Croq and D. M. Pozar, "Millimeter-wave design of wide-band aperture-coupled

stacked microstrip antennas," in IEEE Transactions on Antennas Propagation, Dec.

1991, vol. 39, pp. 1770-1776.

[43] D. Ballo, "Network analyzer basics," 1998 Back to Basics Seminar, 1998.

[44] G.F. Engen and C.A. Hoer, "Thru-Reflect-Line: an improved technique for cali-

brating the dual six-port automatic network analyzer," in IEEE Tkansactions on

Mzcrowave Theory and Techniques, Dec. 1979, vol. 27, pp. 987-998.

[45] R.B. Marks, "A multiline method of network analyzer calibration," in IEEE Trans-

actions on Microwave Theory and Techniques, Jul. 1991, vol. 39, pp. 1205-1215.

[46] "In-fixture microstrip device measurement using TRL* calibration, ," Hewlett

Packard Product Note 8720-2, 1991.