electrical and computing engineering filereconfigurable antennas using mems . carla sofia dos reis...

99
Reconfigurable Antennas Using MEMS Carla Sofia dos Reis Medeiros Master’s Degree Dissertation in Electrical and Computing Engineering Jury President: Prof. António Castelo Branco Rodrigues, IST Supervisor: Prof. Carlos António Cardoso Fernandes, IST Co-supervisor: Prof. Jorge Rodrigues da Costa, ISCTE Member: Prof. Custódio José de Oliveira Peixeiro, IST October 2007

Upload: duongthien

Post on 10-Feb-2019

213 views

Category:

Documents


0 download

TRANSCRIPT

Reconfigurable Antennas Using MEMS

Carla Sofia dos Reis Medeiros

Master’s Degree Dissertation in

Electrical and Computing Engineering

Jury President: Prof. António Castelo Branco Rodrigues, IST

Supervisor: Prof. Carlos António Cardoso Fernandes, IST

Co-supervisor: Prof. Jorge Rodrigues da Costa, ISCTE

Member: Prof. Custódio José de Oliveira Peixeiro, IST

October 2007

ACKNOWLEDGEMENTS

I would like to express my most sincere gratitude to Prof. Carlos Fernandes, my supervisor, and

to Prof. Jorge Costa, my co-supervisor, for the support, guidance and dedication through the

elaboration of this work.

To Mr. António Almeida for performing the measurements of the antennas and for the

suggestions and help during this work.

To Mr. Vasco Fred and Mr. Carlos Brito for the fabrication of the antennas and for the useful

recommendations.

To my colleagues at Instituto Superior Técnico and at Instituto de Telecomunicações for the

friendship, support and suggestions.

The work presented on this thesis was developed in the framework of R-META project

““Reconfigurable Low-profile Antennas Using Metamaterials” – POSC/EEA-CPS/61887/2004 – funded

by Fundação de Ciência e Tecnologia, FCT, under the POS_Conhecimeto program co-funded by

FEDER.

i

iii

ABSTRACT

This work investigates the feasibility of using a commercial numerical electromagnetic solver -

WIPL-D Microwave - to completely model reconfigurable patch antennas using packaged RF MEMS

switches. The proposed model takes into account not only the switch RF characteristic through its

scattering matrix, but also the influence of the MEMS encapsulation and DC actuation circuit on

antenna performance in terms of impedance and radiation pattern. To achieve this, a procedure is

proposed using WIPL-D Microwave to combine a 3D EM analysis of the antenna with a microwave

circuit analysis. The MEMS scattering matrix is de-embedded from measurements performed on a

dedicated test circuit and, for comparison purposes, the equivalent lumped element circuits are

calculated.

The selected antenna’s configurations for testing the procedure are based on a square patch

with one or two slots, the MEMS being used to either short or leave the slots open, enabling to switch

the operating frequencies while maintaining good input impedance match and stable radiation

characteristics, in the 2 to 3 GHz frequency band. This simple antenna’s structures enables to focus

on the modelling of commercial packaged MEMS and on the resulting accuracy of the antenna

simulation prediction, rather than focus on the antenna optimization.

The test antennas were designed and manufactured and the experimental results agree well

with simulations, thus validating the proposed modelling procedure: The results were published and

presented at two conferences and submitted to a Journal.

Keywords Reconfigurable patch antennas, Frequency agility, Packaged RF MEMS switches, Full wave

3D modelling.

v

RESUMO

Este trabalho tem como objectivo avaliar a viabilidade de se usar um simulador

electromagnético comercial (WIPL-D Microwave) para a modelação de antenas impressas

reconfiguráveis que incluam interruptores RF MEMS encapsulados. O modelo proposto permite a

análise do desempenho da antena em termos de impedância de entrada e características de

radiação. A simulação tem em consideração não só as características RF dos interruptores, através

da matriz de dispersão, como também a influência do encapsulamento metálico do MEMS e do

respectivo circuito de actuação DC. Com este intuito, propõe-se um procedimento que combina a

análise electromagnética 3D das antenas com a análise de circuitos de microondas utilizando o

WIPL-D Microwave. A matriz de dispersão dos MEMS é extraída dos resultados medidos num circuito

de teste, devidamente elaborado para esse efeito, e é calculado o modelo equivalente do interruptor.

As configurações de antenas seleccionadas para validação do modelo proposto baseiam-se

em antenas impressas quadradas incluindo uma ou duas fendas rectangulares, as quais podem ou

não ser curto-circuitadas pelo interruptor RF MEMS. Deste modo é possível comutar a frequência de

ressonância das antenas mantendo simultaneamente uma boa adaptação da impedância de entrada

assim como as características de radiação da antena, na banda de frequência dos 2 aos 3 GHz. A

estrutura simples das antenas permite manter a ênfase na modelação dos MEMS comercias e na

precisão da simulação, em vez de se focalizar na optimização da antena.

As antenas foram projectadas e fabricadas e os resultados experimentais estão concordantes

com as simulações, corroborando o modelo e o procedimento proposto. Os resultados foram

publicados e apresentados em duas conferências e submetidos para publicação em revista científica.

Palavras-Chave Antenas impressas reconfiguráveis, Comutação na frequência, Interruptores RF MEMS

encapsulados, Modelação Electromagnética 3D.

CONTENTS

Acknowledgements.......................................................................................................................i Abstract....................................................................................................................................... iii Resumo .......................................................................................................................................v Contents ....................................................................................................................................vii List of Figures .............................................................................................................................ix List of Tables ............................................................................................................................ xiii List of Abbreviations ..................................................................................................................xv Chapter 1 - Introduction.............................................................................................................. 1

1.1. Overview .......................................................................................................................... 1 1.2. State of the art.................................................................................................................. 4 1.3. RF MEMS Switches ......................................................................................................... 5 1.4. Thesis Organization ......................................................................................................... 7

Chapter 2 - RF MEMS switches Characterization...................................................................... 9 2.1. Objectives ........................................................................................................................ 9 2.2. WIPL-D EM and WIPL-D MW Overview........................................................................ 10 2.3. RF MEMS switch basic description................................................................................ 10 2.4. S-matrix De-embedding Procedure ............................................................................... 12 2.5. Experimental Issues....................................................................................................... 14 2.6. Influence of the Encapsulation....................................................................................... 21 2.7. Conclusions.................................................................................................................... 26

Chapter 3 – Reconfigurable Antennas Simulation models....................................................... 27 3.1. Objectives ...................................................................................................................... 27 3.2. MEMS Reconfigurable Patch antenna with one slot...................................................... 28 3.3. MEMS Reconfigurable Patch antenna with two slots .................................................... 32 3.4. Patch antenna with two slots using Ideal Switches ....................................................... 36 3.5. Conclusions.................................................................................................................... 38

Chapter 4 – Experimental Results............................................................................................ 39 4.1. Objectives ...................................................................................................................... 39 4.2. MEMS Reconfigurable Patch antenna with one slot...................................................... 40 4.3. MEMS Reconfigurable Patch antenna with two slots .................................................... 43 4.4. Patch antenna with two slots using Ideal Switches ....................................................... 47 4.5. Evaluation of tick substrate de-embedded S-Matrix ...................................................... 49

vii

viii

4.6. Gain and Radiation Efficiency........................................................................................ 51 4.7. Conclusions.................................................................................................................... 52

Chapter 5 – Conclusions and Future Work .............................................................................. 53 Annexes.................................................................................................................................... 57 ANNEX A Manufacturing process ....................................................................................... 59

A.1 Antenna mask ......................................................................................................... 59 A.2 Photolithographic process....................................................................................... 59 A.3 Fabrication process................................................................................................. 60

ANNEX B Antenna analysis and simulation ........................................................................ 61 B.1 Method of Moments (MoM)..................................................................................... 61 B.2 WIPL-D software ..................................................................................................... 62

ANNEX C Study of Patch Antenna Parameters .................................................................. 65 ANNEX D Radiation Efficiency measurements ................................................................... 70 References ............................................................................................................................... 79

LIST OF FIGURES

Figure 1.1 –Operating principle of RF-MEMS switches devices: (a) Resistive series switch; (b)

Capacitive shunt switch. .......................................................................................................................... 6 Figure 1.2 – Teravicta TT712-68CSP RF MEMS switch operating principle. ............................ 7 Figure 2.1 – Teravicta MEMS switch: (a) Front and back photo; (b) Pin description. .............. 11 Figure 2.2 – Teravicta TT712-68CSP RF MEMS switch characteristic curves provided by

manufacturer [2] for an input impedance of 50 Ω.................................................................................. 12 Figure 2.3 – Equivalent model of the measured test circuit which includes the device under

test (DUT). ............................................................................................................................................. 13 Figure 2.4 - Photo of manufactured test circuits: (a) 50 Ω microstrip reference line; (b) Test

circuit, before mounting the MEMS switch. ........................................................................................... 14 Figure 2.5 – Photo of manufactured prototype: (a) Test circuit with MEMS and 100 kΩ

resistors at the DC path; (b) Zoomed view of the MEMS switch and resistors; (c) Zoomed view of the

DC and RF lines that connect to the MEMS.......................................................................................... 14 Figure 2.6 – 50 Ω microstrip reference line: (a) Electromagnetic model; (b) Microwave circuit

model with connectors........................................................................................................................... 15 Figure 2.7 – Measured and simulated return loss of the 50 Ω reference line: (a) Magnitude; (b)

Phase..................................................................................................................................................... 16 Figure 2.8 -– Measured and simulated insertion loss of the 50 Ω reference line: (a) Magnitude;

(b) Phase. .............................................................................................................................................. 16 Figure 2.9 – Test circuit without MEMS: (a) Electromagnetic model; (b) Microwave model with

connectors. ............................................................................................................................................ 17 Figure 2.10 – WIPL-D simulated return loss and insertion loss curves of Line 1..................... 17 Figure 2.11 – Measured and simulated return loss and insertion loss of the test circuit without

the MEMS: (a) s11 and s41 magnitude; (b) s11 and s41 phase. ............................................................... 17 Figure 2.12 – Measured from test circuit, de-embedded MEMS and equivalent circuit curves

for the MEMS in the OFF-state: (a) s21 magnitude; (b) s21 phase. ........................................................ 18 Figure 2.13 – Measured, de-embedded and equivalent circuit curves for the MEMS in the ON-

state: (a) s11 magnitude; (b) s21 magnitude; (c) s21 phase..................................................................... 19 Figure 2.14 – Photo of fabricated MEMS test circuit with 62 mils thickness substrate. ........... 20 Figure 2.15 – Comparison between results of the MEMS S-matrix using the thick or thin test

circuit: (a) MEMS OFF; (b) MEMS ON. ................................................................................................. 20 Figure 2.16 – Layout of the square patch antenna with one slot. ............................................ 21

ix

Figure 2.17 – Square patch antenna with one slot and without the MEMS: (a) Photo; (b)

Measured and simulated results. .......................................................................................................... 22 Figure 2.18 – Photo of square patch antenna with: (a) MEMS switch; (b) Metal piece. .......... 23 Figure 2.19 – Measured input impedance for the antenna with one slot in all three situations.

............................................................................................................................................................... 23 Figure 2.20 – Measured and simulated input impedance for the antenna with the MEMS case

placed at the centre of the slot. ............................................................................................................. 23 Figure 2.21 – Simulation model: (a) antenna with metal case at the centre of the slot; (b) metal

piece. ..................................................................................................................................................... 24 Figure 2.22 – Measured radiation patterns of the antenna with the MEMS case and with the

metallic case at the centre of the slot: (a) E-plane; (b) H-plane. ........................................................... 25 Figure 2.23 - Measured and simulated radiation patterns of the antenna with and without the

metallic case at the centre of the slot: (a) E-plane; (b) H-plane. ........................................................... 25 Figure 3.1 – Patch antenna with one slot: (a) Open configuration; (b) Closed configuration; (c)

Side view. .............................................................................................................................................. 28 Figure 3.2 – Reconfigurable patch antenna with one slot: (a) Electromagnetic simulation

model; (b) Detailed view of MEMS ports. .............................................................................................. 29 Figure 3.3 – Microwave model of the reconfigurable patch antenna with one slot. ................. 30 Figure 3.4 – Simulated input return loss for the reconfigurable patch antenna with one slot. . 31 Figure 3.5 – Surface currents behaviour on patch antenna with one slot: (a) MEMS OFF; (b)

MEMS ON. ............................................................................................................................................ 31 Figure 3.6 – Simulated E and H-planes for the reconfigurable patch antenna with one slot at

both MEMS state compared with Ideal switches: (a) MEMS OFF; (b) MEMS ON. .............................. 32 Figure 3.7 – Reconfigurable patch antenna with two slots layout: (a) Top view; (b) Side view.

............................................................................................................................................................... 33 Figure 3.8 – Frequency reconfigurable patch antenna with two slots simulation model.......... 34 Figure 3.9 – Frequency reconfigurable patch antenna with two slots microwave circuit. ........ 34 Figure 3.10 – Simulated return loss curves for all four possible states of the MEMS switches.

............................................................................................................................................................... 35 Figure 3.11 – Simulated E and H-plane radiation patterns: (a) MEMS #1 OFF; (b) MEMS #1

ON. ........................................................................................................................................................ 35 Figure 3.12 – Simulation model for the patch antenna with ideal switches: (a) EM model; (b)

Open configuration; (b) Closed configuration........................................................................................ 36 Figure 3.13 – Simulated input return losses for the patch antennas with two slots and ideal

switches: (a) Both open or closed; (b) Intermediate states. .................................................................. 37 Figure 3.14 - Simulated radiation patterns for the patch antennas with two slots and ideal

switches: (a) E-plane; (b) H-plane......................................................................................................... 37 Figure 3.15 - Simulated radiation patterns for the patch antennas with two slots and ideal

switches: (a) E-plane; (b) H-plane......................................................................................................... 38 Figure 4.1 – Photo of manufactured antenna with one slot: (a) Top view; (b) Bottom view..... 40

x

Figure 4.2 – Measured and simulated input return loss of the reconfigurable patch antenna

with one slot and without the MEMS. .................................................................................................... 41 Figure 4.3 – Measured and simulated input return loss curves of patch antenna with one slot

and with MEMS switch: (a) MEMS ON; (b) MEMS OFF. ...................................................................... 41 Figure 4.4 – Measured and simulated radiation patterns of the reconfigurable patch antenna

with one slot: (a) MEMS OFF; (b) MEMS ON. ...................................................................................... 42 Figure 4.5 – Photo of fabricated patch antenna with two slots: (a) Top view; (b) Zoomed view

of the MEMS and the resistors. ............................................................................................................. 43 Figure 4.6 – Measured and simulated return loss curves for the patch antenna with two slots

without MEMS switches......................................................................................................................... 43 Figure 4.7 - Measured and simulated return loss curves for the patch antenna with two slots

and with MEMS #1 inserted: (a) OFF; (b) ON. ...................................................................................... 44 Figure 4.8 - Measured and simulated return loss curves of the patch antenna with two slots:

(a) MEMS #1 in the OFF-state; (b) MEMS #1 in the ON-state.............................................................. 45 Figure 4.9 – Measured and simulated radiation patterns of reconfigurable patch antenna with

one slot: (a) MEMS #1(OFF) - #2(OFF); (b) #1(OFF) - #2(ON). ........................................................... 46 Figure 4.10 - Measured and simulated radiation patterns of reconfigurable patch antenna with

one slot: (a) #1(OFF) - #2(OFF); (b) #1(ON) - #2(ON).......................................................................... 47 Figure 4.11 – Photo of manufactures patch antenna with two slots and ideal switches: (a)

(OFF)-(OFF); (b) (ON)-(ON). ................................................................................................................. 48 Figure 4.12 – Measured and simulated input return loss of the patch antennas with ideal

switches: (a) Open configuration; (b) Closed configuration. ................................................................. 48 Figure 4.13 – Measured and simulated radiation patterns at E and H-plane for the patch

antennas with ideal switches: (a) Open configuration; (b) Closed configuration. ................................. 49 Figure 4.14 – Measured and simulated with new S-matrix input return losses for the patch

antenna with one slot............................................................................................................................. 50 Figure 4.15 - Measured and simulated with new S-matrix input return losses for the patch

antenna with two slots and with MEMS #1 only. ................................................................................... 50 Figure 4.16 - Measured and simulated with new S-matrix input return losses for the

reconfigurable patch antenna with two slots. ........................................................................................ 50

Figure A. 1 – (a) Photographic machine; (b) revealing machine.............................................. 59 Figure A. 2 – Stove. .................................................................................................................. 60 Figure A. 3 – Ultra-violet light oven. ......................................................................................... 60

Figure B. 1 – WIPL-D PRO loading. ......................................................................................... 62 Figure B. 2 – Example of WIPL-D PRO EM Model. ................................................................. 63 Figure B. 3 - Example of an antenna circuit in WIPL-D Microwave. ........................................ 64

Figure C. 1 – Influence of ground plane size in the antenna’s input impedance. .................... 65 Figure C. 2 - Influence of patch size in the antenna’s input impedance................................... 66

xi

xii

Figure C. 3 - Influence of feed position in the antenna’s input impedance. ............................. 66 Figure C. 4 - Influence of slot’s length in the antenna’s input impedance................................ 67 Figure C. 5 - Influence of slot width in the antenna’s input impedance.................................... 67 Figure C. 6 - Influence of slot’s position in the antenna’s input impedance. ............................ 68 Figure C. 7 - Influence of substrate permittivity in the antenna’s input impedance.................. 68 Figure C. 8 - Influence of dielectric and metallic losses in the antenna’s input impedance. .... 69

Figure D. 1 – Photo of the two cavities..................................................................................... 70 Figure D. 2 – Photo of the inclusion of the antenna prototype within the cavity: (a) front view;

(b) side view. ......................................................................................................................................... 71 Figure D. 3 – Input impedance for the antenna with one slot and ideal switches OFF............ 71 Figure D. 4 - Smith Charts plots for the patch antenna with two slots and ideal switches OFF:

(a) Cavity #1; (b) Cavity #2.................................................................................................................... 72 Figure D. 5 – Input impedance for the antenna with one slot and ideal switches ON.............. 73 Figure D. 6 – Smith Charts plots for the patch antenna with two slots and ideal switches ON:

(a) Cavity #1; (b) Cavity #2.................................................................................................................... 74 Figure D. 7 – Input impedance for the antenna with one MEMS switch ON. ........................... 75 Figure D. 8 - Smith Charts plots for the patch antenna with one MEMS switch OFF: (a) Cavity

#1; (b) Cavity #2 .................................................................................................................................... 76 Figure D. 9 - Input impedance for the antenna with two MEMS switches at the OFF state..... 77 Figure D. 10 - Smith Charts plots for the patch antenna with two MEMS switches OFF: (a)

Cavity #1; (b) Cavity #2 ......................................................................................................................... 78

LIST OF TABLES

Table 2.1 – Dimensions of the square patch antenna with one slot......................................... 21 Table 2.2 – Measured and simulated performance of the square patch antenna with one slot.

............................................................................................................................................................... 22 Table 2.3 – Measured and simulated performance for the antenna with the MEMS case placed

at the centre of the slot. ......................................................................................................................... 24 Table 3.1 - Dimensions of the square patch antenna with one MEMS. ................................... 30 Table 3.2 – Dimensions of the frequency reconfigurable patch antenna with two slots. ......... 34 Table 4.1 - Measured and simulated input return loss values of patch antenna with one slot

and with MEMS switch. ......................................................................................................................... 41 Table 4.2 - Measured and simulated return loss values of the patch antenna with two slots and

with MEMS #1 inserted.......................................................................................................................... 44 Table 4.3 – Measured and simulated return performance of the MEMS reconfigurable patch

antenna with two slots. .......................................................................................................................... 46 Table 4.4 – Directivity, Gain and efficiency of the double-slot antennas.................................. 52

xiii

xv

LIST OF ABBREVIATIONS

BW - Bandwidth

DC – Direct Current;

DUT – Device under Test;

EM – Electromagnetic;

FCT – “Fundação de Ciência e Tecnologia”;

FET – Field Effect Transistor;

GPS – Global Positioning System

IT – “Instituto de Telecomunicações”;

MEMS – Micro-Electromechanical Systems;

MoM – Method of Moment;

MW – Microwave;

PIN – p-type, intrinsic, n-type

RF – Radiofrequency

RLC – Resistor (R), Inductor (L) and Capacitance (C);

R-Meta – Reconfigurable Low-profile Antennas Using Metamaterials;

SMT – Surface Mount Technology;

SPDT – Single Pole Double Throw;

UWB – Ultra-Wide Bandwidth;

WLAN – Wireless Local Area Network;

Chapter 1 - Introduction

1.1. OVERVIEW

Microstrip patch antennas have been extensively investigated in the literature and are very

attractive for satellite and wireless mobile communication applications. Their advantages include light

weight, low profile, low cost, relatively small dimensions and compatibility with integrated circuits. On

the other hand, one of the most common drawbacks of these antennas is their inherent narrow

bandwidth. Nevertheless, several modified configurations have been proposed for multi-band,

broadband or ultra-wide band (UWB) applications.

Current developments in wireless communication technologies call for the integration of

several applications into a single terminal, like WLAN, land mobile and satellite communications, GPS,

etc. Since each application operates in a specific frequency bands, in some cases with different

polarizations or radiation characteristics, different antenna structures would be needed to integrate a

single user terminal. However, due to limitations in terms of physical size of some terminals, this

approach is not practical. With reconfigurable antennas it is possible to accommodate several

applications into a single antenna structure, since they enable to electronically change its operating

frequency and/or radiation patterns by adjusting/modifying in some way the shape of the structure. In

many designs this involves the use of RF switches, which can be either GaAs FET or PIN diodes,

1

varactors or, more recently, Micro-Electromechanical Systems (MEMS) switches. Microstrip antennas

are widely used for reconfigurability due to the flexibility of their physical structure.

Reconfigurable antennas have many advantages when compared to other fixed shape

typologies. In general, reconfigurable antennas can result in a significant reduction for the antenna

size and cost, as well as less complexity in system design and development for the communication

systems. For example, single-port multi-band antennas require the use of narrow band filters for

selecting the application bandwidth to reduce noise in the receiver. In reconfigurable antennas, this

filtering is provided directly by the antenna. Multi-band antennas tend to maintain similar polarization

and radiation characteristics at the different operating bands. This may be a considerable set-back

when different types of radiation pattern and polarisation are required for each application. Pattern

reconfiguration may be used also to reduce interferences and fading in multipath environments.

Electrostatic actuated RF MEMS switches are very attractive for antenna reconfigurability

since they provide high performance at RF. When compared with PIN diodes, MEMS provide better

performance at RF due to their low insertion loss, good linearity, high isolation and very low power

consumption. RF MEMS also provide easy integration with CMOS circuits and printed antennas.

However, the main drawback of MEMS switches is the high actuation voltage that can reach values

ten to fifty times higher than the required for PIN diodes although MEMS require less DC power. In

research, two approaches can be found when integrating RF MEMS with antennas: either the MEMS

is directly constructed and integrated with the antenna wafer during fabrication or a packaged MEMS

is attached to the antenna after fabrication. The latter approach is the addressed in this work. Due to

their characteristics, the main application areas of RF MEMS switches in antennas have been:

wireless communication systems, radar systems for defence applications, automotive radars and

satellite communication systems.

A simulation tool is required for antenna design and analysis, since the reliable and accurate

prediction of the results allows drawing conclusions about the design option and to reduce

development costs. When choosing the appropriate software, aspects regarding the antenna

application and configurations have to be taken into consideration. If the antenna integrates active or

passive components the simulator must allow the insertion of lumped elements or S-matrixes into the

electromagnetic model. In some cases, a 3D modelling may be required, instead of a 2D or 2.5D,

when designing complex configurations or to predict the influences induced by the physical structure

of the embedded components.

The goal of this thesis is to accurately model frequency reconfigurable patch antennas with

commercial packaged MEMS switches and acquire expertise in the design and integration of active

components in radiating structures. For a full characterization of the MEMS switch, the antenna model

must take into account not only the switch RF characteristic, but also the influence of the packaged

MEMS encapsulation and of the DC actuation circuit on antenna performance in terms of impedance

and radiation pattern. Thus, in order to enclose these aspects in the antenna analysis, a commercial

numerical 3D electromagnetic solver combined with a microwave circuit analysis is used – WIPL-D

Microwave [1]. The 3D Electromagnetic solver core from WIPL-D EM is based on the method of

moments (MoM), the antenna being modelled by composite wires, metallic plates and dielectric plates

structures. Thus, the objective of this work is to evaluate how a commercially available

2

electromagnetic solver like WIPL-D Microwave [1] can accomplish this task and predict correctly

antenna radiation patterns and input impedance. For this purpose, a commercially available single-

pole double throw (SPDT) RF MEMS switch (Teravicta TT712-68CSP) [2] is selected to switch the

operating frequency of two reconfigurable test antennas.

In order to validate the modelling and to avoid excessive antenna complexity, simple antenna

configurations are chosen. They consist of rectangular patch antennas with one or two slots, each one

cross-connected by a centred MEMS switch. The first antenna configuration includes only one slot and

one MEMS switch; its ON-state causes an up shift on the antenna resonance frequency, with no

significant change in the radiation pattern. A second more demanding test antenna is also considered,

involving two slots and two MEMS switches, independently controlled, resulting in four operating

frequencies. The MEMS switches are mounted with the contacts facing down for direct connection

with the RF signal lines, in order to minimize the RF path onto the MEMS switch, so that the common

flip-chip mounting with bound-wiring is avoided.

In the antenna simulation, the MEMS is modelled by a 3D representation of the package and

by the measured scattering matrix over the antenna operating bandwidth. This enables more reliable

prediction of the antenna frequency behaviour. The focus of this work is mainly on the development of

appropriate modelling procedures that lead to good agreement between measured and numerical

results in terms of input impedance and radiation patterns rather than the optimization of the antennas

for specific wireless applications specifications.

The new procedure that is proposed in this thesis for the analysis of MEMS reconfigurable

antennas is divided into three steps. The first step consists in obtaining the experimental scattering

matrix of the MEMS for its accurate RF switching characterization within the desired operating

bandwidth. This is achieved, in this work, using a dedicated designed test circuit and an accurate de-

embedding procedure. Afterwards, each antenna model is developed and analysed in WIPL-D 3D

Electromagnetic solver taking into account not only the patch but also its RF feeding structure, the

MEMS encapsulation geometry including its RF lines, DC control lines and vias. The final step

consists in the inclusion of the measured MEMS scattering matrix into the antenna model using WIPL-

D Microwave. The 3D Electromagnetic (EM) module is combined with a microwave circuit analysis

code in WIPL-D Microwave. The proposed analysis procedure is compared with the simple equivalent

model approach for the MEMS switches commonly used in the literature.

This work is developed at Instituto de Telecomunicações, IT, in the framework of R-Meta

project, “Reconfigurable Low-profile Antennas Using Metamaterials” [3], funded by Fundação de

Ciência e Tecnologia, FCT. Although this thesis addresses MEMS reconfigurable antennas, the aim of

the R-Meta project is broader: to design and theoretically characterize reconfigurable metamaterial

surfaces that can be used as ground planes for low-profile antennas.

3

1.2. STATE OF THE ART

Several publications addressing reconfigurable antennas can be found in the literature. Many

of these configurations include ideal switches, PIN diodes, varactors or MEMS that may encompass

several distinct functions on the antennas: for instance they are used to modify the antenna feed

location, to control the electrical length of slots placed within the patch, to connect or disconnect

several elements in antenna arrays or in stacked configurations, or, similarly, to connect parasitic

elements to the radiating patch. In this way it is possible to electronically reconfigure the operating

frequency, the polarization, or the main direction of the radiation beam.

In [4]-[12] patch antennas bare PIN diode switches across slots to modify the antenna

configuration and thus control its input impedance or polarization. The switches determine the

effective electrical length of the current paths on the antenna and their orientation controls which

antenna mode is affected. Varactors diodes in [13]-[14] are also used in a slot configuration to tune the

resonant frequency of the antenna according to the applied voltage. Another method for antenna input

matching consists in adjusting the feed point position. This is achieved electronically by controlling the

length of the inset-feed of the antenna using ideal or MEMS switches [15]-[16]. By controlling the

number and the position of the switches, it is possible to obtain multiple operating frequencies. Ideal

switches in [17]-[19] and PIN diodes [11]-[12] are used to connect or disconnect a parasitic patch to

the radiating element. In these configurations the working principle is very simple: when the switches

are ON, the effective length of the patch increases, when OFF, the operating frequency is determined

mainly by the radiating patch.

Stacked configurations can also be reconfigured for a two-mode operation. For example in

[20]-[21], a planar inverted-F antenna (PIFA) or a stacked patch antenna operation is obtained with

PIN diodes switching ON or OFF the feedings or the shorting pins. In [22] a PIFA configuration uses a

PIN diode between the patch and the ground plane to operate as a loop antenna when the diode is

activated: the switch is used to change the resonant mode of the antenna.

Concerning the radiation pattern reconfigurability, in [23] a simple and compact switched-

beam antenna is proposed, consisting of a centre-fed square patch antenna with PIN diodes

connecting to the ground plane, arranged in such a way that the beam direction is switched by

applying forward or reverse bias. In [24] three parallel metal strips, where only the centre strip is feed,

shift the maximum radiation direction by lengthening or shortening the parasitic strip with respect to

the radiating strip. This is accomplished by closing or opening gaps at each of the parasitic elements

using metal strips, while maintaining the impedance characteristics. A five element array antenna in

[25] is proposed for beam control, composed of a probe fed patch and four parasitic patches with one

slot cross-connected by switches arranged around the main patch. Depending on the diodes state, the

frequency of the parasitic patch is changed and the mutual coupling tilts the beam in the direction of

the patch incorporating the polarized diode. Hilbert Curve Patch Antennas in [26]-[27] change the

radiation patterns while maintain the operating frequency. Opening or closing slots, etched to the

patch, suppresses or strengthens one of the two side lobes of the original configuration.

4

The use of MEMS for antenna applications still suffers some limitations, especially for spatial

applications. This is due the immaturity of some RF MEMS aspects, such as hermetic packaging

issues, reliability and power handling capabilities. However, the RF MEMS characteristics, such as

very low power consumption, fast switching time and broad frequency range are very attractive for

designing phased-array antennas. For example, in [28] the nine-elements of the antenna array are

connected or disconnected, using ideal switches to model the MEMS, so that a dual-band antenna is

developed for satellite or radar applications.

Because of its high performance at RF, work concerning the integration of RF-MEMS with the

antennas can be found in many publications, but still it has not been fully demonstrated. In [29] MEMS

switches are monolithically integrated and fabricated along with a rectangular spiral antenna. When

activating or deactivating the switches, the spiral overall arm length is changed and consequently its

radiation beam is changed. As for frequency reconfigurablility, in [30] a planar 1-iteration Sierpinski

gasket antenna uses RF-MEMS switches to shift the operating frequency, while maintaining the

radiation characteristics. When all switches are OFF, the antenna operation follows a bowtie mode;

conversely, when all switches are ON, the operation mode is the same as a fractal antenna. In a

simple antenna configuration [31], capacitive shunt MEMS switches connect a parasitic L-shaped

patch to the radiating square patch. When the MEMS are turned ON, the effective electrical length

increases and the operating frequency is lower than in the OFF state.

Only few works can be found in the literature regarding reconfigurable antennas with

encapsulated MEMS switches. For example, in [32]-[33] encapsulated MEMS switches are used in a

parasitic-slot antenna array to obtain a reconfigurable reactance and consequently to steer the

antenna radiation pattern. In [34], encapsulated MEMS switches are also used in a square spiral

microstrip antenna to reconfigure the radiation patterns by changing the standing electric field

distribution on the radiator. A frequency reconfigurable PIFA antenna in [35]-[36], uses encapsulated

MEMS switches to control the electrical length of the L-shaped slot and hence the resonance

frequency. In all of these works the RF modelling of the switch is very basic, not taking into account

the effect of the encapsulation in the antenna performance. Full modelling of the RF MEMS switch is

one the main goals of the present work.

1.3. RF MEMS SWITCHES

MEMS switches are devices whose operation is based on the use of mechanical movement to

achieve a short circuit or an open circuit in the RF transmission line. RF MEMS switches can be

mainly categorized by:

o circuit configuration – series or shunt;

o type of switching contacts – resistive or capacitive;

o actuation mechanism – electrostatic, electrothermal.

Two main types of electrostatic actuated MEMS switch configurations are commonly found in the

literature: resistive series switches and capacitive shunt switches [37]. Resistive series switch, which

working principle is shown in Figure 1.1a, consists in a cantilever beam which is electrostatically

5

attracted to the substrate to close an open transmission line. This type of switches is attractive for use

at low frequencies, from DC up to a few GHz, where the contact resistance is small thus minimizing

losses. These losses increase with frequency. The capacitive shunt MEMS switch consists of a

suspended bridge that is electrically connected to the RF ground, shown in Figure 1.1b. This switch is

not adequate for DC signals. In fact, the OFF-state for RF signals can be obtained by almost short

circuiting the RF line to the ground plane. A DC controlled electrostatic attractive force is enough to

bend the bridge close to the ground however without touching it to promote the RF signal shunt that

interrupts the transmission line. This switch configuration retains a low insertion loss, but only provides

good isolation above 10 GHz.

In the literature, MEMS switches are usually modelled using lumped element circuits for the

antenna analysis, according to the MEMS typology and switch state [37]. A simple model can be

derived for the resistive series switch using a capacitor for the OFF-sate and a resistor for the ON

state [37]-[39]. However, in capacitive shunt switches the equivalent model corresponds to a shunted

RLC circuit, where the switch state is determined by the capacitance value [37], [39]. In [32] and [36]

the modelling of the packaged MEMS switches consists on the resistive series typology equivalent

circuit, combined with transmission line sections. In [35] a single resistance or capacitance is used for

modelling, depending of the MEMS switch’s state.

(a) (b)

Figure 1.1 –Operating principle of RF-MEMS switches devices: (a) Resistive series switch; (b) Capacitive shunt switch.

The Teravicta TT712-68CSP switch consists of a series cantilever-beam with electrostatic

actuation [2]. The metal beam is attached to an input signal electrode (source) and is suspended

above a control electrode (gate) and an output signal electrode (drain). When a sufficiently large

voltage (+68 V) is applied to the gate relative to the source, the resulting electrostatic force pulls the

beam toward the drain until contact. At that point, the switch is closed and a signal path is formed from

the source to the drain, through the metallic beam. To maintain closure, no quiescent current is

required, leading to an ultra low power consumption device.

6

Figure 1.2 – Teravicta TT712-68CSP RF MEMS switch operating principle.

At the moment, RF MEMS switches are an emerging technology that still needs improvement.

For example, lower actuation voltages, improved reliability, hot-switching durability (RF power level

above 0 dBm), packaging, cost issues and limitations due to the substrate materials used for

construction, are topics under research.

1.4. THESIS ORGANIZATION

The details of the developed work are described in three main chapters. In Chapter Two, the

de-embedding procedure is demonstrated and explained, including the printed circuits required to

measure the scattering matrix of the MEMS, as well as the effect of the MEMS’ metallic encapsulation

both in terms of input return loss or radiation pattern characteristics. The main results in this chapter

were published and presented at a conference [40].

In Chapter Three, the simulations models for the two selected antenna configurations are

presented. Simulation models with RF MEMS switches are explained and numerical results of input

return loss and radiation patterns are shown.

In Chapter Four, the experimental results obtained with the manufactured prototypes are

discussed and compared with the simulation model and with the lumped element approach. A study

concerning losses due to the MEMS switches is also presented. The work included in this chapter

concerning the numerical and experimental results of the antennas was published in two conferences

[40]-[41] and, after a weighty analysis of the simulation model and of the de-embedding procedure,

submitted to a journal [42].

The main conclusions and future work, regarding this project, are addressed in Chapter Five.

7

Chapter 2 - RF MEMS switches Characterization

2.1. OBJECTIVES

The criteria for selecting a convenient MEMS switch for antenna integration were limited by the

small number of current commercially available package MEMS and by the difficulties in acquiring

small quantities. The most import parameters to take into account when evaluating the MEMS RF

performance are the isolation, which corresponds to scattering matrix element s21 when the MEMS is

OFF, and the insertion loss, element s21 when it is ON. In addition, the size of the package is also an

important factor due to the limitations that it may induce in the antenna design and performance. A

large encapsulation will considerable limit the possible configuration choices for the antennas under

development and the number of possible localization for the MEMS switches in the antenna. The

package physical integration with the printed circuits is also an important factor: for example, if the RF

contacts can be mounted directly over the metallic surfaces or if wire bonding is required. The MEMS

switch used in this work is the Teravicta TT712-68CSP [2], as previously referred.

9

In this chapter, the required steps to completely model packaged MEMS switches in

reconfigurable antennas are presented. As first step, the simulation tool is briefly introduced. Then,

several steps required to experimentally extract the RF MEMS scattering matrix using a test circuit are

described and, for comparison with a simpler common approach, the equivalent circuit lumped

elements are calculated for each operating state of the switch. The manufacturing techniques used for

integrating the MEMS switch into the test circuit are also presented. To finalise, the results of a study

concerning the RF MEMS metallic package RF influence are shown.

2.2. WIPL-D EM AND WIPL-D MW OVERVIEW

WIPL-D software package is used in this work because of the long positive experience at IT

with this tool and because it enables a full electromagnetic (EM) characterization of 3D structures,

together with microwave (MW) circuit analysis. The package comprises two complementary tools for

the analysis and optimization of electromagnetic structures: the WIPL-D EM (Pro) solver and the

WIPL-D Microwave tool.

WIPL-D EM is a 3D electromagnetic solver which is based on the Method of Moments (MoM).

It enables to model complex 3D structures formed by wires, metallic plates and dielectrics and to

calculate its EM behaviour both in terms of radiation characteristic and of impedance. The S-matrix of

the model is calculated with reference to arbitrary number of generators (ports) included in the model,

which can be used either for RF feeding purposes or for the inclusion of external microwave

components (through its scattering matrix). One point that is very relevant for the present work is that

WIPL-D requires that the generators are always attached to wires, which may become a nuisance

when modelling external components ports as will be discussed and ahead as well as the way to

circumvent its implications.

The WIPL-D MW tool enables to perform a microwave circuit analysis using and combining

predefined library closed-form models of microwave circuit elements in four implementation

technologies: microstrip, coplanar waveguide, rectangular waveguide, coaxial, lumped elements and

idealized device models. Importantly, it also enables the inclusion of previously analysed 3D EM

structures into the microwave circuit. This can be done either by importing the whole structure or just

the calculated S-matrix data.

The WIPL-D tool includes a sophisticated optimization engine which is used throughout this

work.

2.3. RF MEMS SWITCH BASIC DESCRIPTION

The resistive series Teravicta RF MEMS switch (TT712-68CSP) [2] has a compact hermetic

chip-scale package with dimensions 3.25 mm x 4.5 mm x 1.25 mm. Its top face is metallic and the RF

and DC contacts are solder spheres at the bottom side of the case, Figure 2.1a. The pin description is

shown in Figure 2.1b; each DC control voltage input pin independently actuates its respective RF

output path. At 0 V the switch is at the OFF-state; when an actuation voltage of the order of +68V is

10

supplied to a DC control pin, the RF path between the RFIN contact and the selected RFOUT is closed

and the MEMS is switched ON.

According to the manufacturer, the main characteristics of the referred MEMS switch are wide

frequency range, from DC to 7 GHz: low power consumption, linearity, small size and 30 W peak RF

power handling capability. Manufacturer curves are shown in Figure 2.2 for insertion loss, return loss

(which corresponds to the s11 element of the scattering matrix) and for isolation. An ideal switch

presents an insertion loss of 0 dB and an isolation of −∞ dB within its frequency range. From Figure

2.2, it can be observed that the MEMS switch has a good performance up to 7 GHz, where the

insertion loss is less than 0.5 dB and the isolation values are below -15 dB, however somewhat far

from the ideal values. In the present work, the band of interest is between 2 and 3 GHz, where the the

MEMS present nearly the best performance.

The manufacturer results were obtained using a 12 mil thick Rogers R04003 substrate, with a

permittivity value of 3.38. However, the substrate which will be used to manufacture the antennas in

the present work is the Rogers Duroid 5880, with permittivity value of 2.2 and thickness of 10 mils and

preferably 62 mils in order to favour some bandwidth. Because a different substrate is used, it is

necessary to re-measure the experimental scattering matrix of the MEMS when mounted on this

substrate. In this way the MEMS switch S-matrix is measured under the same conditions and possible

unexpected behaviours are assessed in the de-embedding procedure.

(a) (b)

Figure 2.1 – Teravicta MEMS switch: (a) Front and back photo; (b) Pin description.

For optimal performance in the MEMS operating range (DC to 7 GHz), the manufacturer

recommends the use of external circuit components: 100 kΩ resistors, inserted in series at the DC

lines and shorting the RF output line, to be mounted as close as possible to the MEMS device.

However, to avoid clogging the printed circuits (and the antenna), only the resistors in the DC paths

were included to prevent coupling between RF and DC lines. The +68V actuation voltage can be

optionally supplied by a charge pump (also provided by the manufacturer) which operates from a 3V

supply voltage, however requiring additional circuit components. As before, to avoid clogging the

antenna, the 68V actuation voltage was chosen to be fed directly into the MEMS. In the proposed

antenna configurations to be shown ahead, only one of the two output ports of the SPDT switch is

used.

11

Figure 2.2 – Teravicta TT712-68CSP RF MEMS switch characteristic curves provided by

manufacturer [2] for an input impedance of 50 Ω.

2.4. S-MATRIX DE-EMBEDDING PROCEDURE

The next step concerns measuring the RF MEMS switch scattering matrix at both operating

states (OFF and ON). For this purpose, the MEMS must be in practice integrated into a transmission

line, which will be referred as the “test circuit”. This means that the measured S-matrix will refer to the

combined effect of the MEMS and of the auxiliary transmission line. A de-embedding process must be

used to numerically extract the device S-matrix from the measured S-matrix of the test circuit with the

MEMS.

The test circuit is composed by two 50Ω microstrip lines, each one connected to an RF port of

the MEMS switch. A couple of methods can be found in the literature to perform the de-embedding.

Although some procedures may seem very similar to each other, they generally differ in accuracy:

these procedures may depend on the desired application for the device, the frequency range of

measurement and the typology and characteristics of the device.

The de-embedding method adopted in this work [44] combines measured and simulated results

and can be applied to surface mount device (SMT) in general. For this procedure, two test circuits are

required: a 50 Ω microstrip line for reference purposes and a test circuit for measuring the MEMS

switch RF characteristics.

Some methods consider that the test circuit characteristics are negligible when compared to the

performance of the device under test (DUT). However, as very high performance characteristics are at

stake in the case of MEMS switches, it is no longer suitable to include the test circuit characteristics in

the device S-parameters. It is necessary to de-embed the measured results for an accurate

characterization of the DUT. The de-embedding process used in this work requires both the measured

results and the simulated matrix of each microstrip line in the test circuit as a 2-port network, with the

ground plane as reference. Then, simple matrix manipulation is used to extract the MEMS scattering

matrix from the measurements.

During measurements, the network analyser equipment can only measure the S-parameters of

the complete test circuit structure. The de-embedding procedure used in this work [44] considers that

12

the measured S-matrix can be expressed in terms of the corresponding transmission matrix (TMeasured)

re-arranged as the multiplication of three separate transmission T-matrices as shown in Figure 2.3 and

expressed by equation (1).

Figure 2.3 – Equivalent model of the measured test circuit which includes the device under test (DUT).

[ ] [ ]1Measured Line DUT LineT T T T

2⎡ ⎤ ⎡= ⎤⎣ ⎦ ⎣ ⎦

2

(1)

TDUT corresponds to the transmission matrix of the device under test. Rearranging the matrix

multiplication, the DUT de-embedded T-matrix is obtained from (2).

[ ] [ ]1

1 1

DUT Line Measured LineT T T T− −

⎡ ⎤ ⎡ ⎤= ⎣ ⎦ ⎣ ⎦ (2)

Equations (3) are used to convert S-matrix into T-matrix and vice-versa.

⎥⎥⎥⎥

⎢⎢⎢⎢

−−

=⎥⎦

⎤⎢⎣

22

21

22

22

21122211

22

12

2221

1211

1TT

T

TTTTT

TT

ssss

;

⎥⎥⎥⎥

⎢⎢⎢⎢

−−

−=⎥

⎤⎢⎣

2121

22

21

11

21

21122211

2221

1211

1ss

sss

sssss

TTTT

(3)

Two circuits were designed and manufactured on 30 mm x 10 mm Rogers Duroid 5880

substrate with a 10 mils (0.254 mm) thickness and 2.2 of permittivity: a 50 Ω microstrip reference line

(Figure 2.4a) and the test circuit (Figure 2.4b) prepared for the MEMS insertion. The latter circuit is

formed by two 50 Ω microstrip lines, with half the length and same width of the reference line, and the

required DC control paths for actuating the MEMS switch. Since the substrate thickness determines

the width of the 50 Ω microstrip line, a thin substrate was chosen so that the width of the transmission

line can be similar to the radius of the MEMS’ contact spheres, Figure 2.1a. In this way, no mismatch

and reflections are expected to occur at the RF input and output pins of the MEMS switch. The 0.73

mm width of the transmission line was calculated and adjusted through simulations using the EM

model in WIPL-D and the calculator in WIPL-D Microwave [1]..

13

Line 2

Line 1

(a) (b) Figure 2.4 - Photo of manufactured test circuits: (a) 50 Ω microstrip reference line; (b) Test

circuit, before mounting the MEMS switch.

2.5. EXPERIMENTAL ISSUES

The RF and DC contacts of the selected MEMS switch are at the bottom face of the case. In

order to minimize the RF path onto the MEMS switch, this was mounted directly on the top of the

circuit, with the contacts facing down for direct connection with the RF lines. Therefore, the common

flip-chip mounting with wire-bonding was avoided. The MEMS was soldered to the test circuit (Figure

2.5a) using hot air flux method following the manufacturer recommendations [2]. The major difficulty of

this mounting is to ensure that the MEMS contacts at the bottom side are correctly aligned with the

corresponding metal paths and that the bonding is firm and neat. Only with these conditions the

measurements of the MEMS switches are repeatable and reliable. Because the IT antenna Lab is not

equipped for precision Integrated Circuit mounting, this process being manual, slight perforations were

inserted in the test circuit top face to guide the MEMS alignment, see Figure 2.5b.

Since the charge pump for voltage control is not included in the circuit as previously explained,

an actuation voltage of +68 V DC is required to perform the switching operation, Figure 2.5a.

(a) (b) (c)

Figure 2.5 – Photo of manufactured prototype: (a) Test circuit with MEMS and 100 kΩ resistors at the DC path; (b) Zoomed view of the MEMS switch and resistors; (c) Zoomed view of the DC and RF lines that

connect to the MEMS.

In order to minimize RF coupling to the DC paths, 100 KΩ resistors were introduced in series

with each DC line (Figure 2.5a and Figure 2.5b), as recommended by the manufacturer.

Three steps were taken to de-embed the MEMS S-matrix – SDUT – out from the measured S-

matrix – SMeasured – of the test circuit. As first step, the S-matrix from the 50 Ω reference line

(transmission line) without the switch was measured and compared to the model calculated in WIPLD.

0 V

resistors68 V

MEMS

14

However, the electromagnetic model in Figure 2.6a does not include the SMA connectors which are

attached to the transmission line circuit in Figure 2.4a and are included in the measurements. The

connectors introduce additional length and mismatches to the experimental results. To account for

these disturbances, the connectors were inserted into the transmission line simulation using coaxial

lines in WIPL-D Microwave, shown in Figure 2.6b. For each frequency, WIPL-D Microwave calculates

de S-matrix of the reference line by combining the previously calculated S-matrix of the transmission

line electromagnetic model with the coaxial lines used to simulate the connectors. The coaxial lines

are defined by the dielectric outer radius (Rout), dielectric permittivity (εr), the radius of the inner

conductor (Rin) and the total length of the line (Lc)., In addition to the coaxial line, the connector’s

model can include, if required, an inductor or a shunt capacitor to model mismatches or degradation of

the connectors.

(a) (b)

Figure 2.6 – 50 Ω microstrip reference line: (a) Electromagnetic model; (b) Microwave circuit model with connectors.

The model of the SMA connectors is fine tuned by comparing the measured 50 Ω reference line

results with simulations. The SMA connectors parameters that produced the best match were: Rout =

1.06 mm, εr = 2.2, Rin = 0.31 mm and Lc = 9.6 mm. Actually, in these models only the length of the

connectors (Lc) was adjusted to match the phase of the measured insertion loss, the remaining

parameters correspond to the physical properties of the connectors.

The comparison between measured and simulated s11 elements (return loss) with and without

the connectors is presented in Figure 2.7 and the s21 elements (insertion loss) in Figure 2.8. Since the

transmission line presents very low insertion losses in the measured frequency range – less than 0.3

dB – and a return loss value below -17 dB, it was not required to account for any mismatches in the

connectors’ simulation model with capacitors or inductors. Therefore, when the connectors are

included in the reference line simulations, only the phase of the resulting scattering matrix is altered.

An accurate characterization of the connectors and microstrip line lengths is required in order to

extract correctly the magnitude and especially the phase of all the elements of the MEMS scattering

matrix in each operating state. Simulations have shown that slight changes in the phase of the SDUT

matrix elements may result in considerable shifts of the antenna’s operating frequency at both

operating states of the MEMS.

15

(a) (b)

Figure 2.7 – Measured and simulated return loss of the 50 Ω reference line: (a) Magnitude; (b) Phase.

(a) (b)

Figure 2.8 -– Measured and simulated insertion loss of the 50 Ω reference line: (a) Magnitude; (b) Phase.

In step two, the S-matrixes – SLine_1, SLine_2 – of the left and the right open sections of the RF

lines at each side of the MEMS (labelled as Line 1 and Line 2 in Figure 2.4b) were calculated in WIPL-

D Microwave, excluding the MEMS and using the previously obtained SMA connectors’ model. The

electromagnetic model is in Figure 2.9a and the microwave circuit is in Figure 2.9b. The S-matrix of

the left side line is obtained with ports 1 and 2 of Figure 2.8b and the right side line with ports 3 and 4.

The magnitude of the element S11 and S21 of the simulated scattering matrix for Line 1 are shown in

Figure 2.10. For Line 2 the results are identical.

To confirm the connector and test circuit models, this was first measured without the MEMS,

before mounting the resistors. The comparison between measured and simulated results of the

elements S11 and S41 are shown in Figure 2.11. Results agree quite well, both in terms of magnitude

and phase, however a resonance occurs around 5.5 GHz. This occurs due to coupling effects in the

DC wires, caused by the proximity of the RF and DC paths: the resistors were included after the

MEMS soldering, to minimize the referred coupling.

16

4

2 3 Port 1

(a) (b) Figure 2.9 – Test circuit without MEMS: (a) Electromagnetic model; (b) Microwave model with connectors.

Figure 2.10 – WIPL-D simulated return loss and insertion loss curves of Line 1.

(a) (b)

Figure 2.11 – Measured and simulated return loss and insertion loss of the test circuit without the MEMS: (a) s11 and s41 magnitude; (b) s11 and s41 phase.

To finalise, in step three the simple T-matrix manipulation involving SMeasured, SLine_1 and SLine_2

is used to extract the desired MEMS S-matrix SDUT. The measured and de-embedded results are in

Figure 2.12 for the OFF-state of the MEMS and in Figure 2.13 for the ON-state. Because the insertion

losses of both line sections in the test circuit are very low, the differences between the measured and

de-embedded results are mainly in the phase of the S-matrix elements.

Measured isolation for the OFF-state, ranges between -24 and -12 dB and for the ON-state,

insertion loss ranges from -0.13 to -1.5 dB. Manufacturer reference values for insertion loss are quite

17

similar. However, measured isolation presents 5 dB degradation when compared to the manufacturer

nominal curve. The discrepancy between manufacturer and measured isolation values may be

explained by the absence of the recommended resistors at the RF output line. Even so, the results are

considered more than adequate to demonstrate the objective of this work.

For comparison purposes, the equivalent lumped component models for each state of the

MEMS switch [37] were also calculated from the previously de-embedded scattering matrix at fc = 2.75

GHz. It is recalled that this is an approximate model that is often used in the literature. Considering the

switch configuration, the equivalent circuit consists of a resistor for the ON-state (R = 2.2 Ω) or a

capacitor (C = 0.102 pF) for the OFF-state. These lumped elements are calculated considering the

four elements of the de-embedded MEMS S-matrix. Adjustments in phase values can be performed

adding ideal transmission lines [36]. These lumped elements are calculated performing a fitting with

the four elements of the de-embedded MEMS S-matrix at fc, using equation (4) reproduced from [45].

11 22 12 210

21 MEMS ON

11 22 12 210

21 MEMS OFF

(1 )(1 )Re2

(1 )(1 )1 2 Im2c

s s s sR Zs

s s s sf ZC s

⎡ ⎤+ + −= ⎢ ⎥

⎣ ⎦

⎡ ⎤+ + −= π ⎢ ⎥

⎣ ⎦

(4)

Figure 2.12 and Figure 2.13 also show the results from this simple equivalent model

superimposed on the MEMS de-embedded S-matrix curves for the 1-7 GHz frequency range. The

resulting isolation and insertion loss magnitude curves are quite similar to the de-embedded values in

the 2.5 to 3 GHz frequency band. However, the discrepancy between phase values of the calculated

lumped model and measured S-Matrix in the measured bandwidth increases away from fc. It is

emphasized however that, even at the fc frequency, the agreement is not perfect because the

equivalent circuit is calculated not from one element of the S-matrix, but from all the four elements of

the measured S-Matrix and uses a single lumped element for each state of the MEMS. Such

description is thus insufficient when broad range of frequencies is involved in the antenna design.

(a) (b)

Figure 2.12 – Measured from test circuit, de-embedded MEMS and equivalent circuit curves for the MEMS in the OFF-state: (a) s21 magnitude; (b) s21 phase.

18

(a)

(b) (c)

Figure 2.13 – Measured, de-embedded and equivalent circuit curves for the MEMS in the ON-state: (a) s11 magnitude; (b) s21 magnitude; (c) s21 phase.

As previously mentioned, it is advisable that the MEMS test circuit used for S-matrix de-

embedding resembles as close as possible the MEMS mounting conditions at the antennas. In this

any the influence from the substrate characteristics and from other unexpected effects may also be

accounted in the de-embedding procedure.

However, this approach may not always be feasible namely due to limitations on the width of

the transmission lines or due to the frequency band of operation. Limited by the size of the RF contact

pads below the MEMS case, compatible width of 50 Ω microstrip transmission lines impose a

substrate thickness of the order of 10 mils as previously used. However, at the 2 to 3 GHz band of

interest for this work, an antenna configuration on a 10 mils thickness Duroid 5880 substrate presents

a very narrow impedance bandwidth. A 62 mils thickness substrate is somewhat more favourable, but

in this case the width of the transmission line for the test circuit becomes approximately the length of

the MEMS’ case and undesirable stray radiation may occur.

The previously described tests and procedures were repeated for a new test circuit using a 62

mils (1.5748 mm) thickness RT DUROID 5880 substrate, Figure 2.14. The simulated width for the

transmission line for this substrate thickness is now 4.5 mm. As stressed before, this large dimension

for the transmission line width is unfavourabe due to required abrupt transition for the contacts with the

connecting MEMS pins. This may be responsible for mismatches at the MEMS input and output pins.

In addition, the inner conductor radius of the connectors is relatively small compared with the

19

transmission line with: this transition adds discontinuities to the test circuit model resulting in a larger

discrepancy for the measured and reference values of the MEMS S-matrix.

Figure 2.14 – Photo of fabricated MEMS test circuit with 62 mils thickness substrate.

The same three-step de-embedding procedure was performed to extract the MEMS S-matrix.

The de-embedded S-Matrixes of the MEMS using the two test circuits, with 10 mils and 62 mils

thickness are shown in Figure 2.15. Measured results with this test circuit show a significant difference

when comparing with those obtained using the thin line, particularly the phase curves. It can be

observed that for the OFF-state of the MEMS, isolation deteriorates by about 5 dB. When the switch is

at the ON-state the insertion loss is decreased. However, the obtained de-embedded insertion loss

values are not completely accurate since there are even some frequencies with losses slightly above

0dB. An important aspect to note is that, the de-embedded phase of the MEMS switch becomes close

to zero at both states. This demonstrates that the MEMS S-matrix depends on the substrate in which

the MEMS is inserted and poses an addition challenge on finding the correct way to use the MEMS

de-embedded S-matrix on the antenna simulator.

(a)

(b)

Figure 2.15 – Comparison between results of the MEMS S-matrix using the thick or thin test circuit: (a) MEMS OFF; (b) MEMS ON.

20

2.6. INFLUENCE OF THE ENCAPSULATION

This study intends to evaluate the influence of the MEMS’ metallic encapsulation on the

antennas behaviour. This is done by measuring the input return loss and radiation patterns of an

antenna with and without the MEMS switch. For this purpose, a simple antenna configuration was

chosen: a probe-fed square patch antenna with a single slot, shown in Figure 2.16. First, the antenna

is measured isolated, without the switch or the needed RF and DC lines. Then, for comparison

purposes, the RF MEMS switch is placed at the centre of the slot, again without any polarization

circuits or RF paths. Therefore the MEMS is in no way electrically connected with the antenna. The

difference between measured data in the two cases allows to qualitatively estimating the influence of

the encapsulation in the antenna performance.

Figure 2.16 – Layout of the square patch antenna with one slot.

The antenna was simulated and optimized using WIPL-D (described in ANNEX B) so that a

resonance frequency occurs around 3 GHz. The patch was fabricated on a RT Rogers Duroid 5880 62

mils (1.5748 mm) thickness, with 2.2 of permittivity and tangent loss 0.0009. The substrate total size

was 40 mm x 37 mm and the dimensions of the patch are presented in Table 2.1. All the simulations

were performed using WIPL-D Electromagnetic [1], which is based in the Method of Moments. The

simulation models include metallic and dielectric losses.

Dimensions (mm)

L = W 28.20

Ls 18.58

Ws 4.00

Ps 21.10

Xf 12.94

Yf = 0.5L 14.10

Table 2.1 – Dimensions of the square patch antenna with one slot.

Ls L

W

Ws

(Xf, Yf)

Ps

Feeding point

21

Measurements of the antenna prototype in Figure 2.17a without the RF MEMS switches in

place, allow to fine tune the simulation model and thus create a good reference for comparison with

the case involving the MEMS. As shown in Figure 2.17b and in Table 2.2, a very good agreement is

obtained between measured and simulated results, only a decrease of 0.15 % in the impedance

bandwidth (BW), frequency band below -10 dB, is verified. To obtain this kind of agreement it was

enough to increase the substrate permittivity by 1% (to 2.22), within the manufacturer tolerance

specification [43].

(a) (b)

Figure 2.17 – Square patch antenna with one slot and without the MEMS: (a) Photo; (b) Measured and simulated results.

Measured Simulated Shift

Fr 2.982 GHz 2.983 GHz 0.03%

BW 0.94 % 1.09% 0.15%

Table 2.2 – Measured and simulated performance of the square patch antenna with one slot.

The antenna was next measured with the MEMS switch positioned at the centre of the slot, as

shown in Figure 2.18a. This experience revealed that the presence of the MEMS case induces a slight

decrease in the resonance frequency of the antenna. The subsequent step consisted in finding a good

model for the MEMS encapsulation. For this purpose, the same measurement of the antenna was

performed by replacing the MEMS with a metallic box of equal dimensions. The obtained experimental

results matched exactly the previous ones with the MEMS. The comparison between the measured

results is shown in Figure 2.19. Analysing these results, it can be concluded that the metal box is a

very good experimental model for the encapsulation of this particular MEMS package. Analysing the

input return losses, the antenna without MEMS has an experimental resonant frequency at 2.98 GHz

and with MEMS at 2.97 GHz, which corresponds to a frequency shift of 0.35%.

22

(a) (b)

Figure 2.18 – Photo of square patch antenna with: (a) MEMS switch; (b) Metal piece.

Figure 2.19 – Measured input impedance for the antenna with one slot in all three situations.

The previous metal box was then added to the WIPL-D antenna simulation model (Figure

2.21a). The metallic package was positioned 0.01 mm above the patch layer. The detail of the case is

shown in Figure 2.21b; it even reproduces the small upper indentation presented in the MEMS’

encapsulation seen in Figure 2.1a. The corresponding simulated results are shown in Figure 2.20 and

in Table 2.3. As can be verified, the numerical results agree very well with measurements.

Figure 2.20 – Measured and simulated input impedance for the antenna with the MEMS case placed at the

centre of the slot.

23

Measured Simulated Shift

Fr 2.97 GHz 2.969 GHz 0.03%

BW 0.98 % 1.1% 0.12%

Table 2.3 – Measured and simulated performance for the antenna with the MEMS case placed at the centre of the slot.

So far, the validation of the metal box to model the MEMS’ encapsulation has been only

performed in terms of input impedance. Therefore to extend our tests the radiation pattern

measurements of the antenna with the MEMS and with the metal piece were also performed.

Measurements are shown in Figure 2.22. When comparing the results, the radiation pattern

characteristics are mostly identical and even the peak cross-polarization, in both planes, are very

similar between measurements. It appears that for this antenna configuration, the presence of the

MEMS encapsulation does not degrades the polarization. In fact, the peak cross-polarization is below

-30 dB at the E-plane and below -23 dB at the H-plane and the -3dB beamwidth is about 90º for the E-

plane and 100º for the H-plane.

(a) (b)

Figure 2.21 – Simulation model: (a) antenna with metal case at the centre of the slot; (b) metal piece.

24

(a) (b)

Figure 2.22 – Measured radiation patterns of the antenna with the MEMS case and with the metallic case at the centre of the slot: (a) E-plane; (b) H-plane.

Measured and simulated co-polar radiation patterns show very good agreement up to 120º in

both planes, Figure 2.23. Beyond this limit, the antenna supporting tower and positioner shadows the

measured radiation. Since the cross-polarization peak levels are low in both planes, it is difficult to

predict correctly its behaviour with the simulation. Nevertheless, an adequate agreement is obtained in

the H-plane, where the numerical peak cross-polarization level is 10 dB below the experimental result.

Simulations of the antenna without the metal box are also shown in Figure 2.23. The results

indicate that the radiation characteristics are not affected by the MEMS’ encapsulation.

(a) (b)

Figure 2.23 - Measured and simulated radiation patterns of the antenna with and without the metallic case at the centre of the slot: (a) E-plane; (b) H-plane.

The results in this Section confirm that the metal box is an excellent model to be used in the

simulators to take into account this MEMS packaging effects on the input return loss and on the

radiation pattern of the antenna.

25

26

2.7. CONCLUSIONS

In this chapter, the steps required to model the RF behaviour of the Teravicta’s TT712-68CSP

MEMS switch were presented and explained. The several steps required to experimentally extract the

MEMS switch RF characteristic were described and the importance of an accurate de-embedding

procedure was also pointed out.

It was demonstrated that the physical presence of the MEMS metallic case induces a shift

down in the antenna’s resonance frequency. Also, according to simulations the radiation

characteristics are not affected by the metallic encapsulation. On the other hand, measurements

showed that when replacing the MEMS with a metal piece of equal dimensions, the same impedance

and pattern behaviours are obtained. These results were reproduced by simulations including a metal

case to model the MEMS’ encapsulation.

Chapter 3 – Reconfigurable Antennas Simulation models

3.1. OBJECTIVES

In this chapter, the proposed simulation model for reconfigurable antennas with RF MEMS

switches is described. Two frequency reconfigurable patch antennas are presented for testing the

model. The chosen antenna configurations are maintained very simple so that the focus of this work is

mainly in obtaining a accurate prediction of the antenna performance with the MEMS without the

possible additional uncertainties from a complex antenna geometry. The electromagnetic simulations

include the MEMS encapsulation and the required DC feeding circuit to actuate the switch. The

measured MEMS S-Matrix is included in the simulations using microwave circuit analysis.

The first antenna configuration includes one RF MEMS switch to toggle between two operating

frequencies. The second configuration is a more demanding test since it has two MEMS switches

allowing four operating states. For comparison purposes, a passive prototype of this second antenna

with ideal switches is also analysed. The aim of the two antennas is to validate the simulation model.

No real application specifications are satisfied by these antennas, such as frequency bandwidth or

27

radiation characteristics. However, as explained, the main goal of this study is to establish a reliable

model for the MEMS and not to optimize an antenna for a specific application. The frequency band of

interest for the antennas presented in this study is from 2 to 3 GHz.

3.2. MEMS RECONFIGURABLE PATCH ANTENNA WITH ONE SLOT

The first antenna configuration that is addressed in this work is very common in the literature

[4]-[8] and has been extensively characterized. It consists on a probe fed square patch on a finite

ground plane with a rectangular slot cross-connected by a centred switch, which can either short or

leave the slot open. With the switch at the OFF state, currents flow around the slot and the average

length of the current path is the longest. Hence the antenna resonates at the minimum operating

frequency. Conversely, when the switch is turned ON, most of the electric currents flow through the

switch, decreasing the path length of the current. Therefore, the corresponding resonance frequency

is the highest. The ratio between the two operating frequencies increases mainly with either the

distance between of the slots to the feeding point or with their respective length and width.

As previously referred, the antenna configuration is intentionally simple to focus on the MEMS

RF modelling within patch antennas. The open and closed ideal configurations for the patch antenna

with one slot are presented in Figure 3.1a and in Figure 3.1b, respectively. Although the MEMS switch

presents a very good RF performance, its behaviour is not ideal, as shown by measured results in the

previous chapter. For this reason, it is expected that some of the RF currents flow through the MEMS

switch opening the OFF-state and another fraction is reflected when the switch is at the ON-state.

substrate

(a) (b) (c) Figure 3.1 – Patch antenna with one slot: (a) Open configuration; (b) Closed configuration; (c) Side view.

Because the RF and DC contacts of the selected MEMS of the selected MEMS switch are at

the bottom face of the case, a few changes had to be performed in the antenna configuration. New DC

and RF paths were inserted in the antenna slot to use and actuate the MEMS switch. The final

simulation model is shown in Figure 3.2. The MEMS is placed within the middle of slot, on the top face

of the antenna. By mounting the switch with the contacts facing down there is a direct connection with

28

the slot edges through the RF lines. This way, the RF path is minimized. The DC control was chosen

to be fed from the back of the antenna and then passed through vias and metallic lines into the

MEMS. In this way, the influence of the DC circuit on the antenna RF performance is reduced.

The antenna simulation model was developed and analysed using WIPL-D 3D EM Solver [1].

As shown in Figure 3.2a, it includes the antenna and its RF feeding, the MEMS’ DC actuation circuit

and the MEMS metallic package positioned 0.01 mm above the patch layer. A detailed view of the

metal paths underneath the MEMS is shown in Figure 3.2b. In addition, two ports (generators), dark

blue dots in Figure 3.2b, were attached to thin wires above the antenna slot, yellow lines in the

simulation model. The simulated S-matrix of the electromagnetic model is calculated with reference to

these two MEMS ports and the feeding point. In fact, the 3D EM solver calculates a 3×3 scattering

matrix of the presented configuration. Latter, using a microwave circuit simulator, the measured RF

MEMS S-matrix previously obtained for the 10 mils test circuit is inserted in the model.

MEMS RF Ports

(a) (b)

Figure 3.2 – Reconfigurable patch antenna with one slot: (a) Electromagnetic simulation model; (b) Detailed view of MEMS ports.

The WIPL-D EM solver takes into account the electrical length of wires where the MEMS’

ports are attached. For this reason, the position of the ports was exhaustively analysed and several

possibilities were tested to include the MEMS using a thicker substrate than the one where the MEMS

was measured. The inclusion of the MEMS into the antenna simulation model was therefore

performed using wires in air over the slot with an effective electrical length close to the wires

configuration at the test circuit (embedded in the thin substrate).

The 3D antenna simulation models include the 62 mils thickness (1.5748 mm) Duroid 5880

substrate with a 2.22 of permittivity, dielectric (tangent loss = 0.0009), metallic losses, and a 3 GHz

compatible meshing, including expanded meshing at the metallic edges (edging). The accuracy and

current expansion level used for the simulations is enhanced 2. The simulation microwave model is

shown in Figure 3.3. The cyan rectangle labelled “Antenna” contains the S-matrix of the antenna

Ls L

Ws

(Xf, Yf)

Ps

Feeding

point

W

29

electromagnetic model, “PORT_1” is the antenna feeding and “MEMS” is the measured S-matrix of

the switch in the ON or OFF state.

Antenna

Figure 3.3 – Microwave model of the reconfigurable patch antenna with one slot.

The initial dimensions of the antenna were estimated based upon the general dimensions of a

square patch antenna without the slot and with a resonance frequency at 2.5 GHz. Subsequently, the

antenna optimization was preformed using WIPL-D Optimizer within the microwave circuit analysis in

Figure 3.3 for each operating state of the MEMS switch. The optimization model criteria included

|s11|<-30 dB as the cost function for an unrestricted frequency band, simultaneously for both operating

states of the MEMS switch, without any attempt to match the antenna to any especial application.

However, it was required a compromise for the return loss levels between the two operating states.

The parameters used to optimize the antenna correspond to the patch dimensions (L), the slot

length (Ls) and position (Ps) and also the feeding position along the x-axis (Xf). The feeding position in

the vertical direction (Yf) is at the centre of the slot for ensuring linear polarization. As for the slot’s

width (Ws) is maintained equal to 4 mm due to the large dimensions of the MEMS. The final

dimensions for the antenna are shown in Table 3.1. The optimizations were performed considering a

square 50 x 50 mm2 dielectric wafer. A study regarding the influence of the dimensions, the

characteristics of effects of the dielectric and metallic losses on the patch performance is presented in

ANNEX C.

Dimensions (mm)

L = W 35.10

Ls 22.00

Ws 4.00

Ps 27.35

Xf 14.20

Yf = 0.5L 17.55

Table 3.1 - Dimensions of the square patch antenna with one MEMS. When the MEMS switch is at the OFF-state the simulated operating frequency is f1 = 2.4 GHz

and when turned ON the frequency shifts up to f2 = 2.62 GHz. The return loss simulated curves are

shown in Figure 3.4 and compared with the results obtained for an antenna with ideal switches and

with equal dimensions. The ideal switches were obtained connecting or disconnecting the floating

wires above the slot. Figure 3.4 shows that the operating frequencies with MEMS switches are below

30

the frequencies obtained with the ideal switches, especially for the OFF-state. From these results it

can be concluded that the MEMS switch adds additional length to this antenna configuration.

Figure 3.4 – Simulated input return loss for the reconfigurable patch antenna with one slot.

The near-H field intensity at the patch layer for both operating states of the MEMS is shown in

Figure 3.5. As previously mentioned the current distribution depends on the switch state and therefore

determines the operating frequencies. When the MEMS is OFF the currents are concentrated around

the slot, while in the ON-state the main intensity is verified in the area surrounding the MEMS ports.

However, in the latter state, some areas of high intensity still occur around the slot.

(a) (b)

Figure 3.5 – Surface currents behaviour on patch antenna with one slot: (a) MEMS OFF; (b) MEMS ON.

To include the MEMS’ influence on the radiation patterns, these are calculated in WIPL-D

Microwave. The simulated radiation patterns at both operating frequencies in the E and H-plane are

shown in Figure 3.6. The -3 dB beamwidth is about 86º in the E-plane and about 92º in the H-plane at

both operating frequencies and the cross-polarization is very low for the OFF-state. When the switch is

on the cross-polarization level increases about 15 dB in the E-plane and 10 dB in the H-plane.

31

Nevertheless, the general numerical cross-polarization levels in any operating state is always below -

20dB.

As demonstrated in the previous chapter, the physical presence of the RF MEMS switch in this

example has almost neglectable influence upon the antenna radiation patterns simulated

characteristics. Comparing the E and H-plane radiation patterns of the antenna with MEMS and with

ideal switches, the results are almost identical both in terms of -3dB beamwidth and cross-polarization

levels. However, for higher operating frequencies configurations a foremost influence is expected for

the 3D encapsulation on the performance of the antennas.

(a)

(b) Figure 3.6 – Simulated E and H-planes for the reconfigurable patch antenna with one slot at

both MEMS state compared with Ideal switches: (a) MEMS OFF; (b) MEMS ON.

3.3. MEMS RECONFIGURABLE PATCH ANTENNA WITH TWO SLOTS

In this section, a MEMS reconfigurable patch antenna with two slots is proposed. Similar

configurations using PIN diodes can be found in [4] and in [11]-[12]. The working principle of this

antenna is very similar to the previous configuration, except that the possible number of switch states

is increased. In fact, since each RF MEMS switch cross-connecting the slot is independently actuated,

the two switches can produce four different path lengths for the currents, Figure 3.7. With the switches

at the OFF-state, the two slots are open and currents flowing around the slots over a longer path

produce a lower resonance. Conversely, the currents shortest path and higher resonance frequency

32

occur for both the MEMS switches are turned ON. Intermediate frequencies correspond to the states

where only one of the switches is ON. The ratio between the maximum and minimum frequencies

increases mainly with either the distance between the slots or with their respective length. The

proximity of one of the MEMS to the feeding probe poses an additional strain to test the model

reliability. In fact, some of the reflected current from this MEMS with greatly influence the antenna RF

input.

MEMS #1 MEMS #2

(a) (b) Figure 3.7 – Reconfigurable patch antenna with two slots layout: (a) Top view; (b) Side view.

The electromagnetic simulation model is shown in Figure 3.8: it includes the MEMS’ metallic

encapsulation and identical DC and RF paths below the switches as in the previous Section. However,

due to the additional slot and the DC feeding paths the simulation complexity is considerably

increased: the number of unknowns for the EM model and the simulation time for each frequency

analysis is significantly enlarged, when comparing to the previous configuration with one slot. For

these reasons and due to the software licence limitations, the accuracy and current expansion level

used in simulations was reduced to the enhanced 1 mode. The simulation model considers 2.22 for

the permittivity of the substrate, dielectric losses (loss tangent 0.0009), metallic losses, edging and

meshing performed at 3 GHz.

Each MEMS switch is also linked to the antenna electromagnetic model using two ports

attached to thin wires, as discussed in the previous Section. For this purpose, the microwave circuit in

Figure 3.9 includes two MEMS S-matrix. Each of them can be either the measured ON or OFF MEMS

scattering matrix depending upon the desired antenna configuration. The numerical S-matrix of the

antenna obtained with the electromagnetic model now includes five ports: four for inserting the two

MEMS switches into the simulation (two for each switch) and another for the RF feed of the antenna.

The goal, when optimizing this configuration, was to obtain at least three operating

frequencies, using the same cost function as before for each of the MEMS possible states. The

parameters used for matching the antenna were mainly the slots position and length, the position of

the feeding point and the size of the rectangular patch. To simplify the optimization, it was considered

that each slot had the same dimensions and that they were placed at the same distance from the

correspondent adjacent edge of the patch. The final dimensions for the patch are shown in Table 3.2

and the antenna substrate total size is 50 x 50 mm2.

33

Figure 3.8 – Frequency reconfigurable patch antenna with two slots simulation model.

Figure 3.9 – Frequency reconfigurable patch antenna with two slots microwave circuit.

Dimensions (mm)

L = W 34.00

Ls 17.40

Ws 4.00

Ps 11.50

Xf 11.00

Yf = 0.5L 17.00

Table 3.2 – Dimensions of the frequency reconfigurable patch antenna with two slots.

The simulated input return loss is shown in Figure 3.10 for the different combinations of the

switch states. The simulated resonance frequencies occur at f1 = 2.65 GHz when both switches are at

the OFF-state, at f2 = 2.72 GHz, MEMS #1(OFF)-#2(OFF), at f3 = 2.73 GHz, MEMS #1(ON)-#2(OFF),

and at f4 = 2.85 GHz, when both switches are ON. The second state, MEMS #1(OFF)-#2(ON), has

approximately the same operating frequency as the third state, because the slots are at the same

distance from the patch. However, this latter state produces a faint resonance, input return loss level

above -10 dB. Nevertheless, this fact is irrelevant since the objective of this work is only to prove that

simulations agree with measurements even if the antenna does not present the best performance.

Ls L

Ws

(Xf, Yf)

W

Ps Ps

Feeding

point

34

Figure 3.10 – Simulated return loss curves for all four possible states of the MEMS switches.

The simulated radiation patterns for the four possible states for the MEMS #(1) and #(2) and at

the E and H-plane are shown in Figure 3.11. Measured and simulated co-polar radiation patterns are

very stable at the four operating frequencies. The numerical -3 dB beamwidth is about 85º at the E-

plane and about 90º at the H-plane in all the states. For the intermediate states MEMS #1(OFF)-

#2(ON) and MEMS #1(ON)-#2(OFF) the peak cross-polarizations increases due to the asymmetrical

behaviour of the currents on the patch. At this latter state the cross-polarization is the highest due to

the proximity of the closed switch to the feeding point.

(a)

(b) Figure 3.11 – Simulated E and H-plane radiation patterns: (a) MEMS #1 OFF; (b) MEMS #1 ON.

35

3.4. PATCH ANTENNA WITH TWO SLOTS USING IDEAL SWITCHES

For comparing the results and drawing conclusions about the MEMS influence on the antenna

performance, the double-slot antenna configuration but using ideal switches was also simulated for

further fabrication. It was observed that the ideal switches could either be modelled using wires or

metal strips (either both open or both closed). Both approaches produce the same input return loss

result in the simulator. However, for manufacturing purposes, the metal strip approach is more

adequate and therefore the simulations were performed with the same choice.

The electromagnetic model for these antennas is shown in Figure 3.12a and open and closed

configurations are shown in detail in Figure 3.12b and Figure 3.12c, respectively. The slots include the

previously used DC feeding circuit with slight modifications to include the RF metal strips for the ON

sate.

Simulated results using the previous dimensions show a good impedance match occurred in

all switch states. The EM simulation model considers 2.22 permittivity for the substrate, dielectric

losses (tangent loss 0.0009), metallic losses, edging, meshing performed at 3 GHz and the accuracy

and current expansion level used is enhanced 2.

(a) (b) (c)

Figure 3.12 – Simulation model for the patch antenna with ideal switches: (a) EM model; (b) Open configuration; (b) Closed configuration.

The simulated return losses for both antennas are shown in Figure 3.13a and it can be

observed that the simulated resonant frequencies occur at 2.75 GHz for the open configuration and at

2.86 GHz for the closed configuration of the ideal switches. Comparing with the simulated operating

frequencies for the antenna with both MEMS OFF (2.65 GHz) or ON (2.85 GHz), again it is verified

that the MEMS switches introduce additional length to the antenna, in particular at the OFF-state. The

intermediated states were also simulated with ideal switches, although not manufactured. The

numerical input return loss is shown in Figure 3.13b: the resonance frequency occurs at 2.8 GHz for

both states with ideal switches, which corresponds to a considerably shorter current path than for the

same state configuration with MEMS (2.72 GHz).

36

(a) (b)

Figure 3.13 – Simulated input return losses for the patch antennas with two slots and ideal switches: (a) Both open or closed; (b) Intermediate states.

Regarding radiation patterns simulations in Figure 3.14 and in Figure 3.15, the co-polarization

is very stable in all the operating states. The -3dB beamwidth is similar at the four operating

frequencies, 85º for the E-plane and 88º for the H-plane and the cross-polarization is very low in both

planes and at all frequencies, below -25 dB. Due to the similarity of the obtained curves, it can be

concluded that the radiation patterns are nearly unaffected by the operating state of the ideal switches.

However, in the previous configuration including the two MEMS switches, radiation pattern simulations

including the non-ideal S-matrix of the MEMS switches show a 10 dB higher peak cross-polarization

for the intermediate states, #1(OFF)-#2(ON) and #1(ON)-#2(OFF).

(a) (b)

Figure 3.14 - Simulated radiation patterns for the patch antennas with two slots and ideal switches: (a) E-plane; (b) H-plane.

37

(a) (b)

Figure 3.15 - Simulated radiation patterns for the patch antennas with two slots and ideal switches: (a) E-plane; (b) H-plane.

3.5. CONCLUSIONS

Simulations have shown that the MEMS RF switches present a different behaviour when

compared with ideal switches. The use of physical switches introduces an additional length to the

antenna configuration when compared with ideal switches. This allows reducing the antenna size and

increases the frequency ratio of the different resonances. Comparing the measured and simulated

radiation patterns of the reconfigurable antenna with two MEMS or with ideal switches, the MEMS

presence influences the cross-polarization levels, but the the co-polarization characteristics are

practically unchanged.

The configurations analysed in this chapter have a very simple working principle and allow to

focus on the MEMS RF modelling. The complexity of the simulations is however quite significant due

to the DC and RF thin lines that connect to the MEMS and because of the electromagnetic model of

the MEMS encapsulation. The effectiveness of the proposed modelling procedure and quality of the

presented simulation results is assessed in the following Chapter, by comparing it against

measurements.

38

Chapter 4 – Experimental Results

4.1. OBJECTIVES

In this chapter, the experimental results of the manufactured prototypes described in the

previous chapter are presented. The input return loss and radiation patterns at the E and H-plane are

obtained experimentally and compared with the numerical results. The fabrication details are also

explained, such as the DC supply circuit for the MEMS and the difficulties observed during

measurements are described.

To fine tune the simulation models for the antennas and to ensure that the discrepancies

between measured and simulated curves do not result from modelling errors, first the antennas were

measured without the MEMS switches. Then, each switch at the time is inserted in the antennas and

measured. In this way it was easier to account for the numerical and experimental results

discrepancies, to detected inaccuracies in the design of the antennas and to draw conclusions.

The simulated results of the patch using the MEMS equivalent model and with the MEMS

measured S-matrix using the 62 mils test circuit, which was calculated in Chapter 2, are also

presented and discussed.

39

4.2. MEMS RECONFIGURABLE PATCH ANTENNA WITH ONE SLOT

The first antenna configuration was manufactured using the method described in ANNEX A:

the top and bottom view of the manufactured patch antenna with one slot is in Figure 4.1. The detail of

the bottom side shows the DC wires passing through the antenna ground plane onto the DC pads on

the antenna front face. In the subsequent antennas, the same precautions mentioned when

manufacturing the test circuits, concerning the MEMS’ soldering and mounting, were followed.

(a) (b)

Figure 4.1 – Photo of manufactured antenna with one slot: (a) Top view; (b) Bottom view.

The mounting of the switches and DC wires were performed at successive steps to allow for

intermediate measurements. First, the antenna was measured isolated, without the RF MEMS or the

DC wires. Then an intermediate measurement of the input return loss of the antenna without the

MEMS but with the DC wires was also performed. The results showed that the wires have no effect on

the resonance frequency of the passive prototype, because the frequency shift was insignificant. Next,

the MEMS switch was integrated at the antenna and the whole structure was re-measured.

The return loss of the single-slot antenna without the MEMS is shown in Figure 4.2.The

measured resonance frequency occurs at 2.475 GHz and matches very well with the simulated value

of 2.48 GHz: this corresponds to a frequency shift of just 0.2% between the two results. Bandwidth

values are also very similar: the simulated results show a slightly larger operating band. The

agreement of the resonance depth is also good.

The input return loss for the antenna with the MEMS for both operating states is presented in

Figure 4.3. Comparing with the simulated results, a slight shift down of 0.8 % is observed in the

measured resonance frequency, but the return loss levels are very similar, only a 3 dB difference is

verified for the ON state of the MEMS. From the results in Figure 4.3 and Table 4.1, it can be

observed that the discrepancy between measurements and simulated curves is independent of the

switch state for this antenna configuration.

40

Figure 4.2 – Measured and simulated input return loss of the reconfigurable patch antenna with

one slot and without the MEMS.

To achieve these measured results, a few difficulties had to be overcome. For example, the

precise and reliable MEMS’ soldering to the antenna metal paths was challenging because the MEMS

contacts are below its casing, between the MEMS bottom face and the top face of the antenna. The

correct hot air flow temperature profile versus time had to be mastered to ensure a reliable soldering

without damaging the MEMS and to ensure repeatable measurements. Another important aspect to

consider is to verify for the presence of coupling effects between the DC and RF paths, this can be

confirmed if the RF performance of the antenna is affected by the DC wires.

(a) (b)

Figure 4.3 – Measured and simulated input return loss curves of patch antenna with one slot and with MEMS switch: (a) MEMS ON; (b) MEMS OFF.

Measured Simulated Shift

MEMS OFF 2.38 GHz 2.40 GHz 0.80%

MEMS ON 2.60 GHz 2.62 GHz 0.76%

Table 4.1 - Measured and simulated input return loss values of patch antenna with one slot and with MEMS switch.

41

The measured results were also compared with the simulated input return losses for the

simplified equivalent model of the MEMS switch discussed in Chapter 2. When the switch is in the

OFF-state, the MEMS is represented by a 0.102 pF capacitor, conversely, at the ON-state the

insertion losses are quantified with a 2.2 Ω resistor. WIPL-D Microwave simulations were repeated for

the previous antenna’s electromagnetic model but replacing the MEMS’ measured S-matrix

representation by the lumped elements approach for each operating state of the switch. These new

simulated results show a larger discrepancy with respect to measurements than the simulated curves

obtained using the measured MEMS S-matrix. Now, as shown in Figure 4.3 a shift of around 2 % is

observed between measurements and simulations.

The radiation patterns of the antenna with MEMS were measured at the two operating

frequencies and in the E and H-plane. Results in Figure 4.4 show that the co-polar components curves

agree very well up to 120º, beyond this limit the measurements are shadowed: the -3 dB beamwidth is

larger at the H-plane at both operating frequencies, which agrees well with simulations. Cross-

polarization level is low, less than -17 dB for both switch states in the E-plane and less than -10 dB in

the H-plane.

(a)

(b) Figure 4.4 – Measured and simulated radiation patterns of the reconfigurable patch antenna with one slot:

(a) MEMS OFF; (b) MEMS ON.

42

4.3. MEMS RECONFIGURABLE PATCH ANTENNA WITH TWO SLOTS

As before, the second antenna configuration was manufactured using the method described in

ANNEX A. The top view of the manufactured patch antenna with two slots is shown in Figure 4.5. In

order to exclude any fabrication and design modelling errors and to fine tune the model, the same

sequential partial mounting and measurement step were also applied to this antenna. First, the

antenna was measured isolated without the MEMS switches and excluding the DC feeding wires. The

measured and simulated s11 resonance match exactly at 2.747 GHz, as shown in Figure 4.6 and the

only difference between these two curves is in the return loss level: measurements are 5dB above the

simulated results.

Resistors

MEMS #1 MEMS #2

(a) (b) Figure 4.5 – Photo of fabricated patch antenna with two slots: (a) Top view; (b) Zoomed view of the MEMS

and the resistors.

Figure 4.6 – Measured and simulated return loss curves for the patch antenna with two slots without

MEMS switches.

43

Then, only the MEMS #(1) was inserted in the antenna and input return loss was measured. It

was experimentally verified that there were large coupling effects between the RF and DC paths. For

example, simple changes in the DC wires bending or position changed the antenna return loss curve.

This experience revealed that due to the proximity of the RF feed point, the currents that flow through

the MEMS #(1) are higher than the current across the second MEMS of this antenna or across the

MEMS in the previous single-slot configuration. For this reason, it was necessary to include the 100

kΩ resistors in series with the DC lines for both MEMS in order to reduce the coupling effects. For

symmetry reasons, this procedure was performed in both MEMS and not only in the first one. With the

resistors, the measurements appeared to be extremely reliable, repeatable and no coupling effects

were detected. The comparison between measured and simulated input return losses is shown in

Figure 4.7 and in Table 4.2, for each operating state of the switch #1. It can be observe that the input

return losses agree very well, the discrepancy between resonant frequencies is quite good, less than

0.8% for the ON-sate and less than 0.5% for the OFF-state. As before, Figure 4.7 also shows

superimposed on the measurements the simulation results corresponding to the MEMS equivalent

lumped circuits and once again the agreement of this model with measurements are not so good.

(a) (b)

Figure 4.7 - Measured and simulated return loss curves for the patch antenna with two slots and with MEMS #1 inserted: (a) OFF; (b) ON.

Measured Simulated Shift

MEMS OFF 2.675 GHz 2.687 GHz 0.45%

MEMS ON 2.774 GHz 2.795 GHz 0.76%

Table 4.2 - Measured and simulated return loss values of the patch antenna with two slots and with MEMS #1 inserted.

The subsequent and final step was the insertion of the second MEMS into the antenna

prototype with the correspondent resistors in the DC lines and with the DC wires. The measured and

simulated results for the antenna four possible combinations of the two MEMS switching states are

shown in Figure 4.8. As expected from the simulations, it is noted that the antenna produces at least

44

three well defined resonant frequencies (f1, f3 and f4), corresponding to MEMS #1(OFF)-#2(OFF),

#1(ON)-#2(OFF) and #1(ON)-#2(ON). The #1(OFF)-#2(ON) switch state combination produces a faint

resonance, f2.

Again, the numerical and experimental reflection levels at the resonance are very similar.

Even with the increased number of MEMS switches, the discrepancy between measured and

simulated operating frequencies is reasonably low: about 1.1 % for the first three states, MEMS

#1(OFF)-#2(OFF), #1(OFF)-#2(ON) and #1(ON)-#2(OFF); and a 1.8% frequency shift at the fourth

state, when both switches are ON. As verified in the measurement of the antenna with only one RF

MEMS, when this switch is ON the discrepancy between results tends to be larger than when it is

turned OFF.

It is stressed out that the alternative common lumped elements approach to model the MEMS

produces a higher discrepancy in the input return loss characteristic, as shown in Figure 4.8. Results

show that, even though the equivalent model insertion loss and isolation magnitude curves are similar

to measurements in the 2 to 3 GHz band (Figure 2.12 and Figure 2.13), high discrepancies in the

resonance frequency occur due to the different S-Matrix phase values.

(a)

(b)

Figure 4.8 - Measured and simulated return loss curves of the patch antenna with two slots: (a) MEMS #1 in the OFF-state; (b) MEMS #1 in the ON-state.

45

MEMS #1 MEMS #2 Measured Simulated Shift

OFF OFF 2.62 GHz 2.65 GHz 1.15%

OFF ON 2.69 GHz 2.72 GHz 1.12%

ON OFF 2.70 GHz 2.73 GHz 1.11 %

ON ON 2.8 GHz 2.85 GHz 1.78 %

Table 4.3 – Measured and simulated return performance of the MEMS reconfigurable patch antenna with two slots.

Once again, measured and simulated co-polar radiation patterns at these four frequencies

(Error! Reference source not found. and Figure 4.10) show good agreement up to 120º for all

resonance frequencies and planes. The -3 dB beamwidth is larger in the H-plane at all operating

frequencies and agree very well with simulations.

Cross-polarization level is low, less than -20 dB in the E-plane and less than -15 dB in the H-

plane, when switches are both OFF or both ON (the radiation patterns for these states are very

similar). At the #1(OFF)-#2(ON) state the cross-polarization is below -15 dB, except for a small

angular region and at the #1(ON)-#2(OFF) state, the peak cross-polarization increases up to -10 dB in

both planes. Except for the #1(ON)-#2-(OFF) state, the measured cross-polarization levels are higher

than simulation predictions and this is generally worse in the E-plane. This may be related with

induced RF currents in the long DC wires in the measurement set-up.

(a)

(b) Figure 4.9 – Measured and simulated radiation patterns of reconfigurable patch antenna with one slot: (a)

MEMS #1(OFF) - #2(OFF); (b) #1(OFF) - #2(ON).

46

(a)

(b) Figure 4.10 - Measured and simulated radiation patterns of reconfigurable patch

antenna with one slot: (a) #1(OFF) - #2(OFF); (b) #1(ON) - #2(ON).

4.4. PATCH ANTENNA WITH TWO SLOTS USING IDEAL SWITCHES

The two antenna configurations using ideal switches were also manufactured using the

method described in ANNEX A. Once again the objective is to serve as reference to evaluate the

influence of the real MEMS on the antenna performance. The top view of the patch antennas with two

slots are shown in Figure 4.11a and in Figure 4.11b, for the open and closed configurations

respectively. The measured input impedance agrees very well for both configurations and is shown in

Figure 4.12. When the ideal switches are OFF the experimental resonance frequency occurs exactly

at the simulated value of 2.75 GHz. For the ON-state the measured resonant frequency is 2.85 GHz,

corresponding to a shift down of 0.35% comparing with the 2.86 GHz simulated resonance.

47

(a) (b)

Figure 4.11 – Photo of manufactures patch antenna with two slots and ideal switches: (a) (OFF)-(OFF); (b) (ON)-(ON).

(a) (b)

Figure 4.12 – Measured and simulated input return loss of the patch antennas with ideal switches: (a) Open configuration; (b) Closed configuration.

The measured and simulated radiation patterns are shown in Figure 4.13. The co-polar

components are very similar up to ± 120º and the cross-polarization is very low, in both planes and

operating frequencies. Since the simulated results are 50 dB below the co-polarization it is expected

that the measured curves cannot match the very low simulated level but nevertheless the measures

cross-polarization level is comfortably below -25 dB. As predicted by the simulation, the cross-

polarization is higher in the H-plane and results show that at the #1-OFF #2-OFF state the peak is

about 5 dB higher than in the opposite state. This is due to the currents behaviour on the patch; since

the slot is open the currents flow around the slots.

48

(a)

(b) Figure 4.13 – Measured and simulated radiation patterns at E and H-plane for the patch antennas with

ideal switches: (a) Open configuration; (b) Closed configuration.

4.5. EVALUATION OF TICK SUBSTRATE DE-EMBEDDED S-MATRIX

It is recalled that all the previous simulations are based on the MEMS S-matrix that was

obtained from the 10 mils thickness test circuit. The present Section alternatively compares

measurements with simulated results using the measured MEMS S-matrix obtained with the 62 mils

test circuit, equal to the antenna substrate thickness. In this case, the MEMS ports are defined in the

WIPL-D model through wires from the patch to the ground, instead of wires defined over the slot in air

(see Figure 3.2b). In this way, the antenna simulation procedure is in accordance to what is performed

in the de-embedding procedure. The results for all the antenna configurations are shown in Figure

4.14, Figure 4.15 and Figure 4.16. It is observable that, for all the configurations, the frequency shift

obtained for the MEMS in the OFF-state is smaller, but the same does not occur for the ON-state of

the MEMS.

49

Figure 4.14 – Measured and simulated with new S-matrix input return losses for the patch antenna with

one slot.

Figure 4.15 - Measured and simulated with new S-matrix input return losses for the patch antenna with

two slots and with MEMS #1 only.

Figure 4.16 - Measured and simulated with new S-matrix input return losses for the reconfigurable patch

antenna with two slots.

As might be expected, there is a more noticeable discrepancy between measured and

simulated results then what was found with the first de-embedding. This is because the microstrip line

50

is not fully adapted at the MEMS input/output ports as previously noted and because the connector-

line transition is not fully characterized by simulation. However, the main discrepancy occurs at the

ON-state of the MEMS due to the inaccuracy of measurements for this MEMS switch state.

This shows that the simulation model is not as accurate as before because the impedance of

the transmission line influences the MEMS’ S-matrix. To perform a correct de-embedding, the test

circuit lines should be adapted in the measured frequency band.

4.6. GAIN AND RADIATION EFFICIENCY

To try to quantify the radiation losses of the fabricated antenna prototypes due to the MEMS,

two alternative approaches were considered: in one approach the efficiency was calculated through

the ratio of the measured gain to the directivity; in the second approach a direct efficiency

measurement method (to be described next) is adopted.

To infer the part that the MEMS plays in the efficiency value, two antenna prototypes were

compared: the double-slot antenna configuration with MEMS and the same antenna configuration with

narrow metal strips playing as ideal switches. Measured gain, simulated directivity and calculated

radiation efficiency values of the antenna with MEMS at the resonances are shown in Table 4.4: the

slight increase of the antenna gain with the operating frequency is mainly due to the increase of the

antenna directivity. As it can be observed in Table 4.4, the experimental gain values obtained for the

ideal switches configurations are very similar for both switch states but the calculated efficiency is

about 10% higher than for the configuration with MEMS.

The direct approach for the experimental determination of the antenna efficiency follows the

method described in [46] and in [47]. It is based on the Weller cap model [48] and in the closed

waveguide measurements method [49]. Only two impedance measurements for each antenna

configurations are required for this method, one in free space and the other inside a cavity (see

ANNEX D). Appropriate methods for frequency reconfigurable antennas integrating active elements

were not found in literature. The difficulty stems from the need to introduce the MEMS actuation

voltage inside the measurement cavity, but the DC wires inevitably link the supposedly closed cavity

with the outer volume, thus precluding the correct evaluation of the active antenna efficiency. The

results of the active antennas were not repeatable due to non-measurable radiation of the metallic

pieces and DC wires: these measurements presented an uncertainty of around 3% for the radiation

efficiency, which is significant when trying to determine antenna efficiencies in the range

80%85%.The method and detailed results are presented in ANNEX D.

For the patch antennas with MEMS switches #1(OFF)-#2(OFF) and for the antennas with ideal

switches the measured radiation efficiency value obtained with the direct method are in Table 4.4: a

decrease of almost 10% is verified when the two MEMS are included.

For both antenna configurations (double-slot with MEMS or double-slot with ideal switches)

the direct method provides higher efficiency values than those obtained through gain/directivity. This

may be related with the fact that the simulated directivity was used in the first method instead of the

51

calculated directivity due to practical limitations to measure the antenna radiation pattern for all solid

angle

However, both methods agree that when comparing the two antenna configurations, the

MEMS degrade efficiency by approximately 10% (0.5 dB).

#1 #2 Measured

Gain Simulated

Directivity Calculated

Efficiency (G/D) Measured

Efficiency

OFF OFF 4.3 dBi 5.7 dBi 72 % 82 %

ON OFF 4.5 dBi 6.1 dBi 69 % --- MEMS

ON ON 5.6 dBi 6.7 dBi 78 % ---

OFF OFF 6.1 dBi 7.02 dBi 81 % 90 % Ideal

Switches ON ON 6.5 dBi 7.11 dBi 87 % 92 %

Table 4.4 – Directivity, Gain and efficiency of the double-slot antennas.

4.7. CONCLUSIONS

Numerical and experimental results of the antennas show very good agreement, especially for

the input return loss curves, thus confirming the adequacy of the proposed MEMS modelling

procedure. A detailed observation of the results show that the agreement between measurement and

simulation slightly degrades when the MEMS is actuated by the DC control voltage (at the ON-state).

This means that some fine tuning of the modelling may be required in the future for this switch state.

The usual simple approach found in the literature of modelling the MEMS with the lumped

elements deviates the simulated results further more from the measurements and the discrepancy is

larger as the frequency ratio increases. This occurs because the modelling accomplished by the

equivalent circuits does not fully characterize the MEMS switch throughout all the operating

bandwidth, but instead it is calculated at a particular frequency. This kind of modelling is therefore not

appropriate for the optimization of large frequency span antennas, unlike what happens with the

adopted scattering matrix representation of the MEMS.

A considerable effort was dedicated to solving prototyping problems; the technology is

reasonably dominated in the Lab and reliable antennas prototypes with MEMS can now be produced

at IT.

52

Chapter 5 – Conclusions and Future Work

This thesis evaluates the feasibility of modelling a MEMS reconfigurable antenna considering

both the MEMS RF response and its 3D package effects on the final antenna performance, using a

commercial EM solver (WIPL-D). This work is not intended to optimize the antenna performance to

fulfil given specifications for a dedicated application, but rather to test the accuracy of the MEMS

model and evaluate its adequacy for simple and complex configurations, involving one or more MEMS

switches.

For inserting the MEMS switch into the simulation models two types of characterization were

performed: using a dedicated test circuit, the MEMS S-matrix was measured and de-embedded. Then

the effects of its metallic encapsulation, when no actuation voltage is applied to the MEMS, were

measured and demonstrated by the simulation model. Therefore, the model used for the RF MEMS

switch includes its measured S-matrix and the physical description of the package, which are used in

WIPL-D Microwave to combine the 3D EM antenna analysis with a microwave circuit analysis.

Two reconfigurable antenna’s prototypes using MEMS switches have been designed and

manufactured. To develop expertise in the MEMS reconfigurable antenna subject, this work has

53

evolved in incremental steps with full simulation and experimental characterization at each step up to

the creation of fully functional prototypes.

The modelling procedure was developed and tested first on a simple configuration antenna – a

single slot patch antenna with the MEMS switch placed across the slot. In order to assess it adequacy

for arbitrary antenna configurations, the modelling procedure was then blindly applied to a slightly

more complex antenna structure, with two slots and MEMS switches providing four resonances.

The measurements were performed and compared with simulations and good results were

achieved for the input return loss and radiation characteristics. The observable frequency shift was

around 1% for each of the structures. The agreement between measured and simulated co-polar

radiation patterns was quite good while some discrepancies occurred for the cross-polar component,

mainly at the ON-state of the MEMS. Using simple configurations for the antennas, any discrepancies

due to the passive elements were purged and the discrepancies that arisen between numerical and

experimental results were mainly due to inaccuracies of the de-embedding procedure or unexpected

effects on the MEMS behaviour.

To assess for the losses due to the MEMS switches, passive prototypes were simulated and

manufactured. Comparison of peak gain values for the antennas with two slots with and without

MEMS indicated that the MEMS losses decrease the antenna efficiency by at least 10%. Antenna

efficiency was also measured and calculated through a direct method (Wheeler cap) but the two

methods demonstrated a 10% divergence for the radiation efficiency values. The Wheeler cap based

method may not suitable for these antenna configurations due to the occurrence of non-measurable

radiation of the metallic pieces and DC wires inside the cavities and to the DC actuation unfeasibility

within the caps.

In the progress of this work a few experimental impairments had to be overcome. First,

because the contacts of the MEMS are facing down, they provide no access for conventional soldering

to the antenna RF pads. Temperature controlled hot air flux technique had to be used. Upon soldering,

it is not immediately possible to ascertain if all the contacts are well soldered. This can only be verified

measuring the circuit and verifying the reliability of the results.

Regarding the reconfigurable antenna with two slots, one of the MEMS is positioned near the

patch feeding point, where antenna currents are more intense. This was responsible for a few

unpredictable results, such as coupling effects between the RF and DC signals. However, this effect

was overcome with the inclusion of resistors in the DC actuation paths of the MEMS.

Analysing the measured and simulated results a few conclusion can be drawn. First, for an

accurate de-embedding process, the lines that connect to the MEMS ports at the test circuit must

behave as transmission lines to avoid any radiation by the MEMS and lines. If not, when extracting the

MEMS S-matrix from measurements, these effects will have repercussions on the antenna

measurements. This is observable when a wider transmission line was used to measure the switch in

Chapter 4 and the results showed a larger discrepancy with the MEMS at the ON-state. Insertion loss

values are more sensitive to the test circuit characteristics rendering in a less accurate de-embedding

process compared to the MEMS OFF-state. Moreover, even though the equivalent model insertion

54

55

loss and isolation magnitude curves are similar to measurements in the 2 to 3 GHz band, high

discrepancies in the resonance frequency results occur due to different S-Matrix phase values.

The MEMS modelling proposed in this work can be directly extended for higher operating

frequencies, for higher bandwidths and for other more complex MEMS 3D antenna configurations,

where it is likely to outperform even further the equivalent lumped element models for the MEMS

which are limited in bandwidth and which neglect the external effects of its casing.

The goals that were set for this thesis were fully accomplished. It is pointed out that MEMS

switches is an emerging technology and these are still in a development phase, the employment of

these components, packaged or not, for antenna reconfigurability has not been fully demonstrated in

the literature. In this sense the present thesis adds new knowledge in the field. The experimental and

simulated results presented and discussed in this thesis were published and presented at two

conferences, [40] and [41] and one paper was submitted to a Journal [42].

The next step for this work will be to proceed to antenna optimization for real application

specifications with a large frequency ratio (around 2.5) and using a more complex 3D structure for a

more demanding test, such as stacked antennas. Another parallel issue that deserves further study is

the modelling of the MEMS casing influence when it is in the ON-state; the proposed metal box model

is enough for the OFF-state, but it is apparently insufficient for the ON-state.

ANNEXES

ANNEX A Manufacturing process

A.1 Antenna mask

For the manufacturing process, the first step is to design the antenna mask. WIPL-D software

is used for designing the layout and subsequently the structure is printed using wipldmask in black and

white, with the correct vertical and horizontal scale adjustments for printing. Metallic surfaces are

printed in black and the antenna scale normally used is 4:1 or 5:1.

A.2 Photolithographic process

In the next step, the previous mask is transcribed into photographic paper, first in the negative

mode and then this mask in transcribed in the positive mode. This way the manufacturing errors are

reduced, when compared with the method without the negative mode. The machines in Figure A. 1

are used for this process.

(a) (b)

Figure A. 1 – (a) Photographic machine; (b) revealing machine.

59

A.3 Fabrication process

It is necessary to prepare the substrate before starting the fabrication process. First it is

cleaned due to the oxidation induced by the copper and due to thin plastic layer that it is applied for

protection. Then, a photo-sensitive polish is applied and the substrate is dried in the stove shown in

Figure A. 2.

Figure A. 2 – Stove.

After dry, it is exposed to ultra-violet light (Figure A. 3), the photographic mask is applied over

the substrate, in a way that the area to be metallic is not illuminated with the UV light.

Figure A. 3 – Ultra-violet light oven.

This action is performed in an oven with vacuum, so that the positive and negative

photographic papers are well positioned. At naked eye some defects may not be detected, therefore it

is necessary to verify for these flaws using the microscope and correct them, such as white dots.

Then, the substrate is dived into a caustic soda solution. The metal that was previously

illuminated with the UV radiation is now more sensitive and is dissolved. This process takes about 2

minutes at a temperature of 30ºC to avoid burning the circuit.

Finally, the copper that is protecting the substrate is removed in a hiperclorate acid solution for

about 15 minutes and the remaining polish is also detached.

60

ANNEX B Antenna analysis and simulation

B.1 Method of Moments (MoM)

The Method of moments (MoM) or boundary element method (BEM) is a numerical

computational method applicable to problems involving currents on metallic and dielectric structures

and radiation in free space. Is a full wave solution of Maxwell’s integral equations in the frequency

domain, where only the structure in question is discretised, not free space.

It requires calculating boundary values, rather than values throughout the space defined by a

partial differential equation and it is significantly more efficient in terms of computational resources for

problems where there is a small surface/volume ratio. Conceptually, it works by constructing a "mesh"

over the modelled surface. Boundary element formulations typically give rise to fully populated

matrices, meaning that the storage data and computational time will tend to grow according to the

square of the problem size. Compression techniques can be used to ameliorate these problems,

though at the cost of added complexity. However heavily depending on the problem being solved and

on the geometry of the structure in analyse.

MoM is applicable to problems for which Green's functions can be calculated, multilayered

dielectric media, e.g. substrates for microstrip. The special Green's function formulation implements

2D infinite planes with finite thickness to handle each layer of the dielectric. Conducting surfaces and

wires inside the dielectric layers have to be discretised, but not the dielectric planes themselves.

Metallic surfaces and wires can be arbitrarily oriented in the media and are allowed to cross multiple

layers.

For the antenna analysis, the Method of Moments consists in the division of the structure in

analysis in triangular or rectangular shapes. This is done in such a way so that the size of each

segment is at lest a tenth of the wavelength. The definition of the parameters for each segment is very

important, since it influences the precision level of the results. The smaller the elements shape and

greater the number of elements involved, more accurate is the solution. However, a compromise must

exist between the number of elements and their size, since the computational time and memory used

increases with the number of unknowns.

The small elements, in which the surface is divided, are called basis functions and can change

depending of the complexity of the structure in analysis. The method is based on solving the equation:

1

M

i ii

J Jα=

=∑ ,

61

Where J and Ji are the total and basis current density. To solve this equation, boundary

conditions are required to be satisfied at discrete points at the surface, due to the computational

capacity that would be necessary to solve this for all points.

B.2 WIPL-D software

This software package serves fast and accurate design and simulation tool for projects

involving microwave circuits, components, and antennas. Integrating WIPL-D Pro 3D EM solver with

WIPL-D Microwave, it enables easy inclusion of 3D models into the circuit as well as their optimization

from within the circuit.

• WIPL-D EM PRO

WIPL-D Pro (Figure B. 1) is a 3D electromagnetic solver that provides fast and accurate

analysis of arbitrary metallic and dielectric/magnetic structures. Enables to create complex 3D models

using wires, plates and predefined 3D objects as building blocks (Figure B. 2). The Preview window

enables to see how the changes you make influence the model. Also, data defining node coordinates,

wires, plates and other entities can be easily accessed. By defining the coordinates with symbols, it is

possible to control the model dimensions.

Quick and efficient calculations, based on the Method of Moments, yield high-precision results

(distribution of currents over surfaces, near field pattern, far field pattern and circuit parameters).

Figure B. 1 – WIPL-D PRO loading.

62

Figure B. 2 – Example of WIPL-D PRO EM Model.

Application areas include:

• arbitrary 3D antennas and antenna arrays

• transmission lines and waveguides

• multilayered microwave circuits over finite or infinite substrate

• metallic and/or dielectric scatterers of arbitrary shapes

• circuit parameters of multiport structures (Y, Z or S)

• electromagnetic compatibility problems

• WIPL-D Microwave

WIPL-D Microwave (Figure B. 3) enables to accurately extract circuit parameters from 3D EM

analyzed structures. It is possible to use the predefined library, or interactively build composite metallic

and dielectric 3D models. The component library includes closed-form models in 4 implementation

technologies: microstrip, coplanar waveguide, rectangular waveguide, coaxial, lumped elements and

many idealized device models are also available.

It capture allows easy circuit modelling and the import of standard-format data files

(Touchstone) is also supported.

This software helps to develop complex structures as:

• RF and microwave filters,

• Matching structures,

63

• Resonators,

• Directional couplers,

• Power dividers, and,

• Connectors.

Moreover, it is possible to simulate and optimize various antennas by combining the power of

circuit and 3D EM solvers, such as:

Microstrip antennas embedded in finite lossy dielectric/magnetic materials,

Horn-type feeds for reflector antennas,

Phased arrays along with their matching circuitry, and

Handset antenna in the vicinity of human head.

WIPL-D Microwave easily creates frequency response plots for s-parameters, impedance and

admittance parameters, voltages and currents. For 3D EM components, plots both 2D and 3D graphs

of radiation pattern, near field distribution and distribution of surface currents.

Figure B. 3 - Example of an antenna circuit in WIPL-D Microwave.

64

ANNEX C Study of Patch Antenna Parameters

In this annex, a brief study cornering the influence of parameters of the patch antenna with

one slot shown in Figure 3.2a with dimensions of Table 3.1 but using ideal switches OFF is presented.

The parameters analysed correspond to the ground plane dimensions, patch size (W), feed position

(Xf), slot length (Ls), width (Ws) and position (Ps), and also the influence of the substrate permittivity (εr)

and of the metallic and dielectric losses.

The first parameter analysed is the antenna ground plane and substrate dimensions. From

Figure C. 1 it can be observed that the impedance is barely affected by the dimensions of the ground

plane: mainly is the return loss level that is affected.

Figure C. 1 – Influence of ground plane size in the antenna’s input impedance.

However, concerning the overall dimensions of the patch, W, the resonance frequency shifts

down when the patch is larger and up when smaller, as shown in Figure C. 2. This is due to the

increase or decrease of the total currents path on the patch.

As for the position of the feed along the x-axis direction (Xf), it allows to tune the resonance

frequency and obtain a good impedance match.

65

Figure C. 2 - Influence of patch size in the antenna’s input impedance.

Figure C. 3 - Influence of feed position in the antenna’s input impedance.

The length of the slots (Ls) determines the operating frequency: the input impedance with a

variation of 4 mm in the slots length is shown in Figure C. 4. When this is decreased, the total currents

path is decreased and the resonance frequency is the highest. Conversely, when increased the total

currents path the resonance frequency decreases.

As expected, the same principle of the slot length is applied to the width of the slot (Ws) and to

the slot’s position (Ps), and therefore the same influence on the antenna is verified. Input impedance

results are shown in Figure C. 6 and in Figure C. 7 for the slot’s width and position, respectively.

66

Figure C. 4 - Influence of slot’s length in the antenna’s input impedance.

Figure C. 5 - Influence of slot width in the antenna’s input impedance.

67

Figure C. 6 - Influence of slot’s position in the antenna’s input impedance.

Small variations of 1% in the substrate permittivity (εr ) affect the resonant frequency as shown

in Figure C. 7. However, the loss level is maintained. The higher the permittivity the lower is the

resonance frequency.

Figure C. 7 - Influence of substrate permittivity in the antenna’s input impedance.

Regarding the metallic and substrate losses, the input impedance results are show in Figure

C. 8. It is observable that only the impedance depth is affected by the inclusion or exclusion of the

68

losses in the simulation: when the losses are assessed by the simulation the return loss level

decreases.

Figure C. 8 - Influence of dielectric and metallic losses in the antenna’s input impedance.

69

ANNEX D Radiation Efficiency measurements

The efficiency measurements were performed using the method described in [46]-[47]. Its

principle is very simple and requires only two return loss measurements for each antenna: in free

space and within a cavity. The cavities used for this measurements were dimensioned in work [47]

and can support the frequency range of the antennas proposed in this work: the smaller cavity has a 9

cm diameter and 15 cm length (cavity #1) and the larger cavity has a 14 cm diameter and 20 cm

length (cavity #2), shown in Figure D. 1.

Figure D. 1 – Photo of the two cavities.

For the data processing, a program based on Matlab was developed. This tool receives, in two

different files, the free space return loss measurements as well as measurements with the antenna

inside the cap. The plot of these data in the Smith Chart results in two arcs of circumference, each one

with its radius and centre. Once the center and radius of the cavity measurements is estimated, the

radiation efficiency is calculated using equation (5). minsΔ is the minimum distance between the

resonance frequency and the estimated circumference at the smith chart, maxsΔ is the maximum

distance and 11 fsS is the return loss magnitude at the resonance frequency of the antenna in free

space. These measurements and numerical analyses were performed for the passive antennas with

ideal switches and for the reconfigurable antenna with one and two slots and with the MEMS switches

at the OFF states (Figure D. 2). Due to the isolation requirements for the cavities, it was not possible

to actuate the MEMS switches.

70

21 1max min 11

2 1( ) ( ) 1

r

fss s S

η − −= ⋅Δ + Δ −

(5)

(a) (b) Figure D. 2 – Photo of the inclusion of the antenna prototype within the cavity: (a) front view; (b) side

view.

The return loss measurements for the antenna with two slots and ideal switches OFF is shown

in Figure D. 3. Both measurements on the two cavities are very similar and, in the frequency

bandwidth that enclosures the resonant frequency in free space, no cavities resonances occur. The

radiation efficiency values were obtained using the method previously described and the estimated

circumferences are shown in Figure D. 4 for both cavities. The calculated radiation efficiency is 91 %

and 88 % for the measurements on cavity #1 and cavity #2, respectively.

Figure D. 3 – Input impedance for the antenna with one slot and ideal switches OFF.

71

(a)

minsΔ

maxsΔ

Measured - Free space

Measured – Cavity

Estimated Circumference

Measured - Free space

Measured – Cavity

Estimated Circumference

(b) Figure D. 4 - Smith Charts plots for the patch antenna with two slots and ideal switches OFF: (a) Cavity

#1; (b) Cavity #2.

The return loss measurements for the antenna with two slots and ideal switches ON is shown

in Figure D. 5 . Again, both measurements on the cavities are very similar and in the frequency

bandwidth that enclosures the resonant frequency in free space no cavities resonances occur.

However, small ripple occurs when measuring the antenna on the large cavity. The radiation efficiency

72

values were obtained using the estimated circumferences in Figure D. 6 and are 91 % and 93 % on

measurements within cavity #1 and cavity #2, respectively.

Figure D. 5 – Input impedance for the antenna with one slot and ideal switches ON.

Then, the active configurations with MEMS switches were measured, without being actuated,

inside both cavities. The return loss measurements for the antenna with one slots using a MEMS

switch at the OFF state is shown in Figure D. 7. In the frequency bandwidth that enclosures the

resonant frequency in free space no cavities resonances occurred for measurements on cavity #1.

However, for measurements within cavity #2 a few cavity resonances were verified. The radiation

efficiency values were obtained using the estimated circumferences in Figure D. 8, an efficiency of

87% was obtained using cavity #1 and #2. However, to obtain these values the resonances on cavity

#2 were discarded by the numerical calculation.

73

(a)

(b) Figure D. 6 – Smith Charts plots for the patch antenna with two slots and ideal switches ON: (a) Cavity #1;

(b) Cavity #2

Measured - Free space

Measured – Cavity

Estimated Circumference

Measured - Free space

Measured – Cavity

Estimated Circumference

74

Figure D. 7 – Input impedance for the antenna with one MEMS switch ON.

The return loss measurements for the antenna with two MEMS switches at the OFF state is

shown in Figure D. 9. The radiation efficiency values were obtained using the estimated

circumferences in Figure D. 10 and correspond to 82% and 83% concerning the measurements on

cavity #1 and on cavity #2, respectively. To obtain this efficiency values the number of points used to

estimate the circumference were reduced, due to the proximity of the cavity resonances to the

operating frequency of the antenna measured in free space, especially at cavity #2.

75

(a)

(b)

Measured - Free space

Measured – Cavity

Estimated Circumference

Measured - Free space

Measured – Cavity

Estimated Circumference

Figure D. 8 - Smith Charts plots for the patch antenna with one MEMS switch OFF: (a) Cavity #1; (b) Cavity #2

76

Figure D. 9 - Input impedance for the antenna with two MEMS switches at the OFF state.

Observing the results, the measurements using cavity #1 are more reliable than cavity #2.

When no MEMS switches are included, the radiation efficiency is around 91 %, when one MEMS

switched is included a decrease of 4 % is observed and with two MEMS the decrease is about 8%

when compared with ideal switches.

However, when measuring the active antennas within the cavities a few resonances occurred

and the circumference estimation was limited by the number of points used, especially for the antenna

with two MEMS inside cavity #2. The uncertainty of these results is around 3% and due to radiating

effects of the MEMS encapsulation and DC wires, the measured results were not always reliable or

repeatable, especially for the double-slot antenna.

77

78

(a)

(b)

Measured - Free space

Measured – Cavity

Estimated Circumference

Measured - Free space

Measured – Cavity

Estimated Circumference

Figure D. 10 - Smith Charts plots for the patch antenna with two MEMS switches OFF: (a) Cavity #1; (b) Cavity #2

REFERENCES

[1] WIPL-D: http://www.wipl-d.com/

[2] Teravicta: http://www.teravicta.com/

[3] Project R-Meta: http://www.it.pt/project_detail_p.asp?ID=528

[4] Yang, F., Rahmat-Samii, Y., “Patch antenna with switchable slots (PASS):

reconfigurable design for wireless communications”, Antennas and Propagation Society International

Symposium, IEEE, Vol. 1, June 2002, pp. 462 – 465.

[5] Yang, F., Rahmat-Samii, Y., “Patch antennas with switchable slots (PASS) in wireless

communications: concepts, designs, and applications”, Antennas and Propagation Magazine, IEEE,

Vol. 47, No. 2, April 2005, pp. 13 – 29.

[6] Nanbo, Jin, Yang, F., Rahmat-Samii, Y., “A Novel Reconfigurable Patch Antenna with

both frequency and polarization diversities for wireless communication”, Antennas and Propagation

Society International Symposium, IEEE, Vol. 2, June 2004, pp.1796 – 1799.

[7] Yang, F., Rahmat-Samii, Y., “A single layer dual band circularly polarized microstrip

antenna for GPS applications”, Antennas and Propagation Society International Symposium, IEEE,

Vol. 4, June 2002, pp. 720 – 723.

[8] Yang, F., Rahmat-Samii, Y., “A compact dual band circularly polarized antenna design

for Mars rover mission”, Antennas and Propagation Society International Symposium, Vol. 3, June

2003, pp. 858 – 861.

[9] Shynu, S. V., Augustin, G., Aanandan, C. K., Mohanan, P., Vasudevan , K., “A

reconfigurable dual-frequency slot-loaded microstrip antenna controlled by PIN diodes”, Microwave

and Optical Technology Letters, Vol. 44, No. 4, February 2005, pp. 374-376.

[10] Sung, Y.J., Kim, B.Y., Jang, T.U., Kim, Y.-S., “Switchable triangular microstrip patch

antenna for dual-frequency operation”, Antennas and Propagation Society International Symposium,

Vol. 1, June 2004, pp. 265 – 268.

[11] Medeiros, C. R., Castela, A., Costa, J.R., Fernandes, C.A., " Evaluation of Modelling

Accuracy of Reconfigurable Patch Antennas ", Proc Conf. on Telecommunications - ConfTele ,

Peniche , Portugal , Vol. 1 , pp. 13 - 16 , May 2007.

[12] Medeiros, C. R., Castela, A., Antenas Reconfiguráveis para multi-serviços, Final Year

Project, Instituto Superior Técnico, Lisboa, Portugal, 2006.

[13] Shynu, S. V., Augustin, G., Aanandan, C. K., Mohanan, P., Vasudevan , K.,

“Development of a varactor-controlled dual-frequency reconfigurable microstrip antenna”, Microwave

and Optical Technology Letters, Vol. 46, No. 4, August 2005, pp.375-377.

79

[14] Shynu, S. V., Augustin, G., Aanandan, C. K., Mohanan, P., Vasudevan, K., “Triple Slot

Arm Reconfigurable Dual Frequency Microstrip Antenna Using Varactors”, Antennas and Propagation

Society International Symposium, Vol. 2B, July 2005, pp. 609 – 612.

[15] Onat, S., Alatan, L., Demir, S., “Design of triple-band reconfigurable microstrip

antenna employing RF-MEMS switches”, Antennas and Propagation Society International

Symposium, Vol. 2, June 2004, pp.1812 – 1815.

[16] Lee, A.W.M., Kagan, S.K., Wong, M., Singh, R.S., Brown, E.R., “Measurement and

FEM modeling of a reconfigurable-patch antenna for use in the wideband gapfiller satellite system”,

Antennas and Propagation Society International Symposium, Vol. 1, June 2003, pp. 379 – 382.

[17] Zhang, C., Yang, S., Pan, H. K., Fathy, A., El-Ghazaly, S., Nair, V., “Development of

Reconfigurable Mini-Nested Patches Antenna for Universal Wireless Receiver Using MEMS”,

Antennas and Propagation Society International Symposium, Vol. 1, July 2006, pp. 205-208.

[18] Liu, S., Lee, M., Jung, C., Li, G.P., Flaviis, F., “A Frequency-Reconfigurable Circularly

Polarized Patch Antenna by Integrating MEMS Switches”, Antennas and Propagation Society

International Symposium, Vol. 2A, July 2005, pp. 413 – 416.

[19] Liu, S. Lee, M., Bachman, M., Li, G.-P., De Flaviis, F., “A Frequency-Selectable Patch

Antenna of Circular Polarization with Integrated MEMS Switches”, IEEE Wireless Communications

and Applied Computational Electromagnetics, April 2005, pp. 195-198.

[20] Kunda, V.K., Ali, M., “Reconfigurable Stacked Patch Antenna for Satellite and

Terrestrial Applications”, Topical Conference on Wireless Communication Technology, IEEE, October

2003, pp. 152-153.

[21] Ali, M., Sayem, A., Kunda, V. K., “A Reconfigurable Stacked Microstrip Patch Antenna

for Satellite and Terrestrial Links”, Transactions on Vehicular Technology, IEEE, Vol. 56, Issue 2,

March 2007, pp. 426-435.

[22] Chen, Q., Kurahashi, M., Sawaya, K., “Dual-mode Patch Antenna with PIN Diode

Switch”, 6th International Symposium on Antennas, Propagation and EM Theory, IEEE, November

2003, pp. 66-69.

[23] Ngamyanyaporn, P., Krairiksh, M., “Switched-beam single patch antenna”, Electronic

Letters, Vol. 38, Issue 1, Jan. 2002, pp. 7-8.

[24] Zhang, S., Huff, G.H., Feng, J., Bernhard, J.T., “ A Pattern Reconfigurable Microstrip

Parasitic Array”, IEEE Transactions on Antennas and Propagation, Vol. 52, No.10, Oct. 2004, pp.

2773-2776.

[25] Nishiyama, E., Hisadomi, R., Aikawa, M, ”Beam controllable microstrip antenna with

switching diode”, Antennas and Propagation International Symposium, IEEE, July 2006, pp. 2337-

2340.

[26] -Song Yang, Bing-Zhong Wang, Yong Zhang Xue, “Reconfigurable Hilbert curve patch

antenna”, Antennas and Propagation Society International Symposium, IEEE, vol. 2B, pp. 613-616,

July 2005.

80

[27] Zhang, Y., Wang, B., Yang, X., Wu, W., “A fractal Hilbert microstrip antenna with

reconfigurable radiation patterns”, Antennas and Propagation Society International Symposium, IEEE,

vol. 3A, July 2005, pp. 254-257.

[28] Weedon, W.H., Payne, W.J., Rebeiz, G.M., “MEMS-switched reconfigurable

antennas”, Antennas and Propagation Society International Symposium, Vol. 3, July 2001, pp. 654 –

657.

[29] Jung C., Lee, M., Li, G., De Flaviis, F., “Reconfigurable Scan-Beam Single-Arm Spiral

Antenna Integrated With RF-MEMS Switches”, IEEE Transactions on Antennas and Propagation, Vol.

54, No.2, Feb. 2006, pp. 455-463.

[30] Anagnostou, D.E., Guizhen Zheng, Chryssomallis, M.T., Lyke, J.C., Ponchak, G.E.,

Papapolymerou, J., Christodoulou, C.G., “ Design, Fabrications, and Measurements of an RF-MEMS-

Based Self-Similar Reconfigurable Antenna”, IEEE Transactions on Antennas and Propagation, Vol.

54, No.2, Feb. 2006, pp. 422-432.

[31] Onat, S., Alatan, L., Demir, S., Unlu, M., Akin, T., “Design of a Re-Configurable Dual

Frequency Microstrip Antenna with Integrated RF MEMS Switches”, Antennas and Propagation

Society International Symposium, Vol. 2A, July 2005, pp. 384 – 387.

[32] Petit, L., Dussopt, L., Lahuerte, J.M, “MEMS-Switches Parasitic-Antenna Array for

Radiation Pattern Diversity”, IEEE Transactions on Antennas and Propagation, Vol. 54, No.9, Sept.

2006, pp. 2624-2631.

[33] Coutts, G., Mansour, R., Chaudhuri, S., “A MEMS-Based Electronically Steerable

Switched Parasitic Antenna Array”, Antennas and Propagation Society International Symposium, Vol.

2A, July 2005, pp. 404-407.

[34] Huff, G., Bernhard J., “Integration of Packaged RF MEMS Switches With Radiation

Pattern Reconfigurable Square Spiral Microstrip Antennas”, IEEE Transactions on Antennas and

Propagation, Vol. 54, No. 2, Feb. 2006, pp. 422-432.

[35] Panaïa, P., Luxey, C., Jacquemod, G., Staraj, R., Kossiavas, G., Dussopt, L.,

Vacherand, F. and Billard, C. “MEMS-Based Reconfigurable Antennas”, IEEE International

Symposium on Industrial Electronics, Vol. 1, 2004, pp. 175-179.

[36] Panaïa, P., Luxey, C., Jacquemod, G., Staraj R.; Petit, L.; Dusspot, L.; “Multistandard

Reconfigurable PIFA Antenna”, Microwave Optical and Technology Letters, vol. 48, No.10, October

2006, pp. 1975-1977.

[37] Rebeiz, G, Muldavin, J., “RF MEMS Switches and Switch Circuits”, IEEE Microwave

Magazine, December 2001, pp. 59-71.

[38] Dimitrakopoulos, N., Hartley, A., Miles, R. and Pollard, R., “Microwave Simulation of

an X-Band RF MEMS Switch”, 1st EMRS DTC Technical Conference, Edinburgh, UK, May 2004.

[39] Milosavljevic, Z., “RF MEMS Switches”, Microwave Review, Vol.10, No.1, 2004

[40] Medeiros, C. R., Costa, J.R., Fernandes, C.A., Kolundzija, B., " Modelling of a MEMS

Reconfigurable Antenna Using WIPL-D ", Proc ESA Antenna Workshop on Multiple Beam and

Reconfigurable Antennas , Noordwijk , Netherlands , Vol. 1 , April 2007, pp. 369 – 372.

81

82

[41] Medeiros, C. R., Costa, J.R., Fernandes, C.A., Kolundzija, B., " MEMS Reconfigurable

Antenna Patch Antenna Analysis Using WIPL-D ", Proc 8th International Symposium on RF MEMS and

RF Microsystems, Barcelona, Spain, Vol. 1, June 2007, pp. 67-70.

[42] Medeiros, C. R., Costa, J.R., Fernandes, C.A., Kolundzija, B., “Modelling of

Frequency Reconfigurable Antennas using RF MEMS Switches”, Microwave Optical and Technology

Letters, SUBMITED August 2007.

[43] Rogers DUROID 5880: http://www.rogerscorporation.com/mwu/pdf/5000data.pdf

[44] Agilent, De-embedding and Embedding S-Parameter Networks Using a Vector

Network Analyzer, Application Note 1364-1, [ONLINE] http://cp.literature.agilent.com/litweb/pdf/5980-

2784EN.pdf.

[45] Pozar, D.M., Microwave Engineering, John Wiley & Sons, 2nd ed., New York, 1998.

[46] Geissler, M., Litschke, O., Heberling, D., Waldow, P., Wolff, I., “An improved method

for measuring the radiation efficiency of mobile devices“, Antennas and Propagation Society

International Symposium, IEEE, Vol. 4, June 2003, pp. 743-746.

[47] Mendes, C., Radiation Efficiency of Small Printed Antennas, Master’s degree

Dissertation, Instituto Superior Técnico, Lisboa, Portugal, 2007.

[48] Wheeler, H.A., “The radiansphere around a small antenna", Proceedings of the IRE,

August 1959, pp. 1325-1331.

[49] Johnston, R. H. and McRory, J. G., “An improved small antenna radiation-efficiency

measurement method", IEEE Antennas and Propagation Magazine, vol. 40, October 1998, pp. 40-48.