PASSIVE AND ACTIVE RECONFIGURABLE MICROSTRIP REFLECTARRAY ANTENNAS

Transcription

1 PASSIVE AND ACTIVE RECONFIGURABLE MICROSTRIP REFLECTARRAY ANTENNAS

2

3 PASSIVE AND ACTIVE RECONFIGURABLE MICROSTRIP REFLECTARRAY ANTENNAS PROEFSCHRIFT ter verkrijging van de graad van doctor aan de Technische Universiteit Delft, op gezag van de Rector Magnificus prof. dr. ir. J.T. Fokkema, voorzitter van het College voor Promoties, in het openbaar te verdedigen op dinsdag 30 september 2008 om uur door Mostafa HAJIAN elektrotechnisch ingenieur geboren te Arak, Iran

4 Dit proefschrift is goedgekeurd door de promotor: Prof.dr.ir. L.P. Ligthart Samenstelling promotiecommissie: Rector Magnificus, Prof.dr.ir. L.P. Ligthart, Prof.ir. K. Robers, Dr.Sci. A.G. Yarovoy, Prof.dr.ir. G. Vandenbosch, Prof.dr.ir. E.R. Fledderus, Prof.dr.ir. I.G.M.M. Niemegeers, Prof.dr.ir. W.C. van Etten, voorzitter Technische Universiteit Delft, promotor Technische Universiteit Delft Technische Universiteit Delft Katholieke Universiteit Leuven, België Technische Universiteit Eindhoven, Technische Universiteit Delft Technische Universiteit Twente Thesis Delft University of Technology. With references and with summary in Dutch. ISBN Subject headings: hollow patch antenna, reflectarray antenna, passive- and active antennas, shared aperture antenna. Printed in The Netherlands Copyright c 2008 by M. Hajian All rights reserved. No part of the material protected by this copyright notice may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, recording or by any information storage and retrieval system, without permission from the copyright owner. The work presented in this thesis was financially supported by IRCTR in The Netherlands.

5 To Carla, Lavinia and Ramses

6

7 Contents 1 Introduction Research objective Research lines Novelties and main results Outline of the thesis Microstrip reflectarray antennas Basic microstrip reflectarray Active microstrip reflectarray Design parameters of microstrip reflectarray Feeding Spillover and taper efficiencies Losses Bandwidth Single element considerations Numerical aspects Design parameters Patch dimensions Substrate thickness Dielectric constant Ground plane dimensions Performance parameters Far field phase

8 II CONTENTS Far field magnitude Bandwidth Cross-polarization Resonance frequency Near field Surface currents RF Reflection and transmission of semiconductor material under illumination of light The concept Modelling the reflection and transmission coefficients Propagation constant in semiconductor material Electron-hole pair generation rate Reflection and transmission coefficients Numerical results Experimental results X-band Ka-band Discussions Conclusions Formulation of the integral equation for microstrip reflectarray antennas Numerical validation Singularity Numerical results Conclusions Variable-sized phasing technique Geometry of a variable-sized patch Design procedure Infinite ground plane Truncated ground plane

9 CONTENTS III 6.3 Phase diagram Computational aspects Substrate thickness Surface currents Near-field Array design Measurement results Bandwidth Conclusions Hollow patch: part Design procedure Phase diagram Computational aspects Substrate thickness Patch length Surface currents Near-field Waveguide simulator measurements Conclusions Hollow patch: part Geometry of the rectangular hollow patch Design procedure Phase Diagram Computational aspects Substrate thickness Patch length Slot length Surface currents Near-field Hollow patch with island

10 IV CONTENTS 8.6 Phase Diagram Substrate thickness Patch length Surface currents Conclusion Reconfigurable active MRA with capacitive loading Geometry of a loaded hollow patch Design procedure Phase diagram Computational aspects Substrate thickness Patch length Slot length Slot width Surface currents Near-field Varactors Varactor-based scanning capabilities Technological aspects and experimental results Narrow-band multi-frequency shared aperture antenna Conclusions General conclusions and discussion 119 APPENDICES 125 A Derivation of integral equation 127 A.1 Integral equations A.1.1 Derivation of dyadic Green s function A.1.2 The Scattered field A.1.3 Series expansion for the current A.1.4 Bases function

11 CONTENTS V A.2 Derivation of Green s function A.2.1 Basic apporach A.2.2 g xx component of Green s function A.2.3 g zx component of Green s function B Phase comparison 143 C Phase comparison: rectangular and square hollow patch 145 D Control voltages 147 E Publications by the Author 149 Symbols 155 Summary 165 Samenvatting 167 Acknowledgement 169 About the author 171

12 VI CONTENTS

13 Chapter 1 Introduction The history of antennas dates back to James Clerk Maxwell who unified the theories of electricity and magnetism, and eloquently represented their relations through a set of profound equations best known as Maxwell s Equations. The field of antenna technology is dynamic and over the last 50 years has become an indispensable partner in telecommunications technology, systems and applications. The advantages of radio communications and radar in our daily life has put a significant amount of research and work on designing high performance antennas. While in the past antenna design may have been considered a secondary issue in overall system design, today it plays a critical role. In fact, many radio and radar system successes rely on the design and performance of the antenna. The array antennas are most versatile in antenna systems. They find wide applications not only in space-borne systems, but in many other earthbound missions. The state-of-the-art of antenna array field is primarily related to beam scanning array antennas. With arrays, it is to-day s practice not only to synthesize almost any desired amplitude radiation pattern, but also to scan the main lobe by controlling the relative phase excitation between the elements. However, from theoretical and practical point of view most advanced arrays are complex, and expensive and from technological point of view they are difficult to be realized. The principle as used in reflector antennas was already long time before 1888 in use for optical telescopes. The demands of reflectors for use in radio astronomy, microwave communications, deep space communications, and many other applications have forced a major progress in the development and optimising of sophisticated reflector antennas. In general reflector antennas are not flexible, bulky and the proto-typing requires much time and so manpower. In the past decade the research on analytical and experimental techniques for reflectarray antennas has received considerable attention for re-

14 2 Introduction placement of reflector antennas. In addition to this the attempt has been made to use such reflectarray technology for beam scanning by integrating the antenna with active phase shifters, but this is a rather complicated process. In 1999 the International Research Centre for Telecommunications and Radar (IRCTR) initiated a research program in the field of reflectarray antennas. The result of this research has led to writing this Ph.D thesis. 1.1 Research objective The major and central problem that is investigated in this thesis refers to the design of microstrip reflectarray (MRA) antennas with active devices demonstrating low cost beam scanning capabilities. The primary focus is directed towards the solution for the phasing technique that can be achieved with the microstrip reflectarray structure. The secondary focus is to develop a platform for designing microstrip reflectarray antennas using different type of phasing techniques. 1.2 Research lines Reflectarray antennas are complex antenna systems having a number of features that needs to be investigated for a successful realization. A graphical illustration of the major topics and features of reflectarray antennas to be investigated is provided in Figure 1.1. Attention is paid to the design and analysis of the single cell elementary radiator and the array for a successful realization of this new concept. Reflectarray antenna Primary radiator Array architecture Radiating part Phasing method Phase diagram Beam scan Passive Array Active array Figure 1.1: Research tree. In the recent years microstrip antennas has got numerous attention due to their versatilities as primary radiator in the design of reflectarray

15 Novelties and main results 3 antennas. Due to the proposed complex patch geometry, numerical analysis tools for electromagnetic problems are used to predict their performance as primary radiator. Taking into consideration the three distinctive important aspects of each radiator for MRA, i.e. the radiating part, the phasing technique and the phase diagram, form a fundamental approach for a successful design of an antenna element in the array environments. The main goal of this thesis is to design reflectarray antennas with scanning capabilities of the main beam. Passive and active beam scanning are taken into consideration and numerical electromagnetic tools are used to predict the array performance. Small array antennas are developed on a thin dielectric slab and measured to validate the proposed concept. As will be motivated in the next chapter the mutual coupling issue is not considered in detail. Experimental validation in an array realized in printed circuit board technology has justified the simplification to neglect mutual coupling effect. 1.3 Novelties and main results The approach to design a reflectarray antenna has led to a number of research incentives that will be addressed here. The most relevant novelties presented in this thesis are as follows: proposal of a new concept for a reflectarray antenna on semiconductor material using optical imaging (Chapter 4); modelling and measuring the RF reflection- and transmission coefficients of semiconductor material under illumination of an optical source (Chapter 4); fast calculation of the impedance matrix using multi-node processors Distributed ASCI Supercomputers (DAS) based on a self-developed Method-Of-Moment (MOM) algorithm (Chapter 5). In this way a full wave element analysis is presented, where a number of difficulties such as numerical integration, treatment of singularities could be solved in a classical way; addition to existing designs for variable-sized reflectarray antennas based on MOM (Chapter 6); design procedure and measurement of a new phasing technique; the so-called hollow patch antenna was initiated at IRCTR and has led to results given in Chapter 7; design procedure of active-loaded reflectarray antennas using hollow patch antennas (Chapters 8 and 9); realization of a low cost reconfigurable MRA consisting of hollow patches with active varactors for beam scanning (Chapter 9).

16 4 Introduction The main achievements of this thesis are: modelling and experimental validation of the reflection from and transmission through a semiconductor material under illumination of light are presented in Chapter 4; a fast technique for evaluation of currents excited on the patches is developed in Chapter 5; a design procedure for an elementary variable-sized patch antenna using a numerical software package has been realized and its potentials are demonstrated in Chapter 6; a design procedure for an elementary equal-sized hollow patch antenna is given in Chapters 7 and 8; a design procedure for an elementary active hollow patch antenna and the realization of a low cost reconfigurable active MRA with such antenna elements are described in Chapter Outline of the thesis This thesis is organized in three parts which include the main steps for designing MRA antennas. In the first part the electromagnetic formulation of the elementary radiator is presented. The full wave formulation based on MOM is developed and studied. The second part is dedicated to the design and analysis of the different geometries for elementary radiators which allow different phasing techniques. The third part covers the aspects related to the design and analysis of passive- and active reflectarray antennas for beam scanning purposes. A brief description of each chapter is provided hereafter: Chapter 2 - An overview of the reflectarray antenna is given. The advantages and disadvantages of such array configuration are discussed in more detail. Important aspects of the antenna are presented and explained with some examples. Chapter 3 - This chapter gives an overview of the parameters of a single antenna element which determine the performance of the total antenna system. Some of the parameters are determined for each phasing technique given in subsequent chapters. Chapter 4 - This chapter introduces a new concept for the realization of reflectarray antennas. The antenna uses semiconductor material illuminated optically. The reflection- and transmission coefficients of the semiconductor material under illumination of light is modelled using the S-T matrix. The approach was developed and presented for the first time at IRCTR. Measurements have been done to validate simulation results.

17 Outline of the thesis 5 Chapter 5 - This chapter deals with the modelling and analysis of a single patch antenna on an infinite ground plane. The full wave integral equations are presented using the dyadic Green s function and solved using MOM to predict accurately the current distribution excited on the surface of the patch antenna. Once the current is known, all other parameters including the phase of the scattered field in the far zone can be determined. Computer implementation of the MOM method is discussed subsequently. The efficiency, in terms of computing time requirement, was achieved by adopting a new parallel programming using DAS systems. Chapter 6 - In this chapter the analysis and design of MRA at millimeter wave (MMW) frequencies using a variable-sized phasing technique is explained. Critical parameters and design curves are presented. A number of arrays has been built and measurements are performed to demonstrate the passive beam scanning. Chapter 7 - A new phasing technique, the so-called hollow phasing using slotted patch is introduced in this chapter. Such a geometry form the basis for the realization of active reflectarray antennas using varactors. The dimensions of the slot are altered in both x- andy directions simultaneously to achieve the phasing while the outer patch dimensions remain the same. Design and performance parameters of such phasing technique are analysed. Measurements are done using a waveguide simulator to measure the phase and to validate the simulation results. Chapter 8 - This chapter is the second part of hollow phasing using the slotted patch. Since the active antenna is realized at 6 GHz, this chapter determines specific aspects of the phasing mechanism at this frequency. The focus in this case is that the slot dimensions can be changed in one direction only. Chapter 9 - In this chapter the concept of active reflectarray antennas at 6.0GHz is presented. The hollow patch antenna is integrated with active varactors. The capacitance of such a varactor can dynamically be adopted by applying different voltages. It is shown that such change in capacitance has a considerable effect on the reflected phase of the signal impinging on the hollow patch antenna loaded with a varactor device. The dimensions of the slot are kept constant for all antenna elements in the array. A new antenna geometry is designed for integration with the active device. The design aspects and parameters are studied in detail. Measurements are performed to demonstrate the proposed concept. Chapter 10 - Main conclusions, recommendations and open issues for future work are addressed in this chapter.

18 6 Introduction

19 Chapter 2 Microstrip reflectarray antennas As the name implies, a reflectarray antenna combines some of the best features of reflector- and array antennas. Due to their high efficiency and gain, reflector antennas are most often used in applications where one single main lobe pattern as fixed beam is required. The geometry of the reflector antenna consists of a feed illuminating the reflector surface. Usually the surface has a parabolic shape. In general, reflector antennas and antenna mounts are bulky in size and large in mass. For high-performance applications, they have huge manufacturing costs and their packaging and transportation is difficult. By mechanically moving the feed, which can have often complications in many applications, some limited scanning of the main beam becomes feasible. As an alternative a planar reflectarray antenna has been proposed as a replacement for conventional parabolic reflector antennas. It was first investigated by Berry [1]. The reflectarray antenna employs a number of isolated antenna elements without any power divider network and is in many cases printed on a flat surface [2]-[16]. Figure 2.1 shows the basic geometry of a microstrip reflectarray using patch antennas. The flat reflector, as shown in Figure 2.1, is composed of a thin slab of dielectric material having one side completely covered with a thin layer of metal (serving as a ground plane) and the other side etched with many metallic microstrip patches. A feed antenna illuminates the array antenna, the individual elements are designed to scatter the incident field with proper phase to form a planar phase surface in front of the array aperture, as suggested in Figure 2.1. This approach is especially desirable at higher frequencies where the loss of a microstrip feed network is unacceptable. The rapid development in microstrip antenna technology has led to the use of microstrip antennas in a variety of reflectarray configurations. The appreciation for microstrip antennas is mainly due to their inherent compactness, lightweight and low manufacturing cost and its beam scanning capabilities once it is integrated with a controlling phase

20 8 Microstrip reflectarray antennas device. In general, the feed may be positioned at an arbitrary angle from reflectarray, but is assumed to be far enough from the reflectarray so that the incident field can be approximated by a plane wave. x Co-phase wave front h Microstrip antenna element Feed antenna f FP Substrate Ground plane f FP = focal distance Path delay z O from feed antenna to MRA Figure 2.1: Geometry of the microstrip reflectarray antenna. Phase shifting elements correct for path delay, and create a co-phase wavefront at a given angle. As indicated in Figure 2.1 the basic design principle requires that the phase ψ i of the field reflected from an element in the reflectarray be chosen so that the total phase delay from the feed to a fixed aperture plane in front of the reflectarray is constant for all elements. An antenna with such a configuration is called Microstrip ReflectArray (MRA) antenna. The advantages of MRA antennas are as follows [16]: Surface mounting: because of its thin and flat reflecting surface, the antenna can be mounted more easy onto a vehicle, spacecraft, building walls or roof tops with less support structure (less mass and volume) than needed for parabolic reflectors in Direct Broadcast Satellite (DBS) receiver stations. The antenna can also be mounted on a conformal (i.e. curved) structure. The phase deviation caused by the curved structure can be compensated by a number of techniques which will be explained in the subsequent chapters; Scanning beam potentials: the main beam of the antenna can be

21 Basic microstrip reflectarray 9 scanned in space, passive as well as active. theme in this thesis; This will be a central High reliability: since the element devices of the array can be highly decoupled from each other, the failure of a few elements will have a minor impact on the performance of such an antenna system; Low manufacturing cost: the reflectarray uses a printed microstrip antenna and can be fabricated using a simple and low cost etching process. This is specifically true for the high-performance aircraft, spacecraft, and satellites and missile applications where size, weight, cost, performance, ease of installation, and aerodynamics are constraints; Weight and compactness; Large aperture dimension capability: due to the fact that no power combiner/divider is needed, the resistive insertion loss of thousands of microstrip patches in the reflectarray is the same as that of a single patch element; Mutual coupling: as extracted from literature on printed and flat structures of arrays employing patch antennas, the mutual coupling effect between the MRA antenna elements is low and is not considered in this thesis [16]-[21]; Side lobe levels: The sidelobe levels of MRA with electrically large dimensions (gain higher than 30 db) are generally low, since the illuminated field coming from the feed horn is tapered. The elements on the edge of the array receive less energy compared to the center element. These effects result into a strong tapering across the array aperture, which leads to low sidelobe level. Disadvantages of MRA are: Narrow bandwidth: because of the resonance behaviour of the patch element the maximum bandwidth is in the order of few percent of the resonance frequency. It is possible to increase the bandwidth by using a two or three layered stacked array [22]; Efficiency: a high reflection coefficient and low insertion loss allow for achieving relatively good efficiency for electrically large apertures. It has been shown that the efficiency of the antenna can be in order of 50% to 70% [3, 8]. 2.1 Basic microstrip reflectarray The basic geometry of a MRA configuration is shown in Figure 2.1. There is a great deal of flexibility in choosing the feeding method such as prime-focus,

22 10 Microstrip reflectarray antennas offset and cassegrain feeding. Linear, dual and circular polarization can be obtained. Since it is desired to form a planar phase surface in front of the aperture, one of the key feature of MRA is how the individual elements can be made to scatter with a desired phase. There are a number of different techniques to control the phase. One method is to use identical elements with variable length stubs to control the reflection phase [4, 5]. Usually this is not the optimal approach since the delay lines require space and as its length increases it becomes part of the radiating process. This unwanted radiation increases the cross polarization component [5]. A better approach was chosen by Pozar in which the compensation was done by using patches with variable sizes [8, 9]. This technique will also get attention in this thesis. Another method is by rotating the antenna element [7]. These techniques can be viewed as methods to shift slightly the resonance frequency of the patch which effects the phase of the reflected field. The new phasing technique proposed in this thesis is the active hollow patch phasing technique. It uses identical patches with a slot of fixed dimensions loaded with active device. 2.2 Active microstrip reflectarray In active phased array antennas, each radiating elements is equipped with a phase shifter. Beams are formed by shifting the phase of the signal emitted from each radiating element, to steer the beams in the desired direction. This property of active antennas can be used in many applications such as: space probe communication, weather research, radio-frequency identification, broadcasting, low-profile and low cost in-motion satellite TV reception system, earth science missions, earth based sensors for remote sensing, and application to satellite TV. Hence due to the enormous applications there is a great interest for developing advanced high gain phased arrays in many companies and universities. There is the experience that such systems are extremely expensive, electronically complex, and beyond the reach of many applications where cost constraints prevail. In some applications, only limited beam steering is required; in such cases, a simplified beam scanning facility, can be selected. For this thesis work it was decided to focus on the design and analysis of a low cost array antenna with limited-scan angle, ±18 based on the active MRA concept. In the past different single element phasing techniques have been considered for use in active MRA. Hollow patches are assumed to be easily modified for active controlling the phase shift [23]. Research has been done loading a rectangular patch with varactors at each radiating edge [24]. In [25] an aperture-coupled patch with perpendicular feeds is used. Another possibility is presented in [26] using a slotted patch loaded with Micro-Electro- Mechanical-System (MEMS). This thesis offers a new approach using a capacitive element in combination with a hollow patch. This method shows

23 Design parameters of microstrip reflectarray 11 advantages compared to other techniques. For evaluating the active MRA it is important to consider the following aspects: Beam steering: Not only the capability of beam steering is important but also the scanning range and the quality of the pencil beam needs to be evaluated. The sensitivity of the signal controlling the phase should receive separate attention; Control circuit: The complexity of the control circuit is an important issue especially for large array antennas. The number of control signals and the method of control might become a complex issue in the design. For example if each array element requires two switches (four choices for the phase) a 5x5 array already needs 50 signals to be connected and controlled. The control unit requires additional electronics which can introduce parasites etc; Durability and Robustness: Durability and robustness are considered to be advantageous for passive MRA. Integrating the MRA with active components may put a limitation on system lifetime, since these components are less robust and may lead to signal distortions. 2.3 Design parameters of microstrip reflectarray In this section the most important design parameters of MRA are listed and compared to those of reflector antennas, if appropriate Feeding The MRA can use feeding techniques similar to that of the reflector. They have both the same subtended angle [8]. In this thesis the feeding is a horizontally-polarized plane wave along the main axis of the array, normal to the antenna. For MRA measurements a horn- or waveguide feed is used. The feed is placed at the focal point of the antenna which is usually is at the far-field of MRA. Hence it can be assumed that the MRA is illuminated with the plane waves Spillover and taper efficiencies The efficiency of MRA is dominated by the spillover and taper efficiencies. To illustrate the spillover and taper efficiencies the following feed pattern is assumed G f (θ) =cos n (θ) (2.1) where n = 2, 4, 6, 8,... For this feed pattern the spillover and amplitude taper efficiencies can be found in closed-form expressions. Based on the

24 12 Microstrip reflectarray antennas definitions and analysis given in [8] and [27] the spillover and amplitude taper efficiencies for a circular array become η s =1 cos (n+1) θ 0 (2.2) 2n (1 cos ( n 2 1) θ 0 ) 2 η t = tan 2 (θ 0 ) ( n 2 1)2 (1 cos n (2.3) (θ 0 ) where η s and η t are the spillover and taper efficiency, respectively and θ 0 is the subtended half-angle, defined in Figure 2.2. The aperture efficiency as function of subtended angle is depicted in Figure 2.3 for different values of n. D a = Aperture size D a f FP θ 0 θ 0 feed z Figure 2.2: Subtended angle of reflectarray n=8 n=6 n=4 n=2 Aperture efficiency Subtended half angle [degree] Figure 2.3: Aperture efficiency for a circular MRA versus subtended halfangle.

25 Design parameters of microstrip reflectarray 13 As in the case of a reflector antenna the maximum efficiency is around 80% and is maximized for a selected optimum value of θ 0. Although this is the efficiency for a circular MRA, it still gives a good indication of the efficiency behaviour as function of subtended angle for non-circular MRA as will be investigated in this thesis Losses MRA suffers from dielectric loss, copper loss, and surface wave excitation. The losses due to the dielectric and copper are usually higher than losses due to surface wave excitation [8]. Figure 2.4 presents the dielectric losses as function of normalized patch length at 35.0GHz. The meanings of the parameters are defined in the Figure 3.1. It indicates that the losses can be several db for thin substrates with a high-loss tangent tanδ of the dielectric, where δ is the loss angle h = mm h = mm Loss [db] tanδ = tanδ = Normalized Patch Length ΔL/L 0 Figure 2.4: Loss for microstrip patches versus patch size for two different substrate thickness, h. f = 35.0GHz, resonance length L 0 = 2.5mm, patch width W = 3.3mm, ε r =2.33[8] Bandwidth The bandwidth of the reflectarray is considered to be the frequency range in which the designed array operates within acceptable levels [27]. The bandwidth performance of a microstrip reflectarray can be limited by four factors [3]: microstrip patch element, element spacing, the feed antenna bandwidth, and differential spatial phase delay. According to [8] the bandwidth of MRA is mainly dominated and determined by the bandwidth of the single element. Each element of the array, independent of the phasing technique, uses the

26 14 Microstrip reflectarray antennas patch antenna within ±5% of the nominal resonance frequency. Because of the nonlinear behaviour of the phase curve, deviations from a linear phase curve greater than a few percent, will occur due to changes in frequency. This effect limits the bandwidth of the MRA. Random phase errors can occur by the processing tolerances in the flatness of the array, by the etching process and by uncertainties in the phase center of the feed. In the classical approach the bandwidth is defined at the frequencies for which the return loss is below a certain threshold, usually taken to be -10dB. However, for the MRA this is not the case. Hence in this thesis the bandwidth is connected to the drop in gain by (±2dB) in comparison with the radiation pattern at the center frequency. For the evaluation of MRA most of the parameters given in the previous subsections will hardly be further considered in this thesis. Also no attempt is made to obtain favourable antenna specifications concerning the elements and array structure. The main focus is to demonstrate the concept of employing reflectarrays for passive- and active beam scanning. The different parameters discussed in this chapter provide background information for the evaluation of such antennas. Some of the parameters will be determined and the here-given terminology is used throughout the text of coming chapters.

27 Chapter 3 Single element considerations In this thesis the array design is based on the phase characteristics of the elementary radiator and requires therefore detailed single element analysis. In later chapters it will be justified to design the array using single element data. Mutual coupling between array elements is presumed negligible because of sufficiently large inter-element spacing and thin substrate. The elementary radiator is microstrip antenna. Figure 3.1 illustrates the configuration of antenna element. z Fringing effect Patch in z = h y Substrate, ɛ r W x h L 0 GP size λ 0 2 Ground plane GP size λ 0 2 Figure 3.1: Microstrip antenna. Microstrip antennas, as shown in Figure 3.1, consist of a very thin metallic strip (patch) placed a small fraction of a wavelength (h <<λ 0, usually 0.003λ 0 h 0.05λ 0,whereλ 0 is the free-space wavelength) above a ground plane. The space between the patch and ground plane is filled with substrate with dielecric constant of ɛ r. Because the dimensions of the

28 16 Single element considerations patch are finite along the length and width, based on the transmission line theory, the fields at the edges of the patch undergo fringing [27]. This is illustrated along the length in Figure 3.1. Fringing makes the patch look wider electrically compared to its physical dimensions and need to be taken into consideration for determining the resonance frequency. In this chapter an overview on the elementary radiator and the so-called single cell parameters is given which will play a major role in realizing phase-shifting per antenna element in an array. 3.1 Numerical aspects The single elements and array results are obtained using specific-made software and a commercial package offering a full-wave solution based on the MOM. The different phasing techniques require different modelling. For all optional techniques, the study of the sampling the patch geometry is an important issue. The dimensions and form of the geometry determine how dense the configuration needs to be sampled in space, in order to create accurate numerical results. If some parts of the geometry are small compared to the operational wavelength, these parts need to be sampled with high sampling rate. To reduce the run time applying non-uniform sampling becomes an important issue. Hence all models in this thesis are sampled using non-uniform meshing. The sampling issues are addressed separately for each phasing technique. The results for the rectangular patch show that errors resulting from sampling are predictable in most cases. An error in the sampling causes a shift in the resonance frequency. It means that even though the antenna would not operate at the desired frequency, it would work at some near-by frequency and can still give the correct scan angle. 3.2 Design parameters The basic design parameters are briefly discussed in this section. These parameters play an important role in the design process of the single antenna element and array configuration Patch dimensions The rectangular patch antenna is used as antenna element in the different phasing techniques considered in this thesis. The dimensions partly control the resonance frequency for each phasing technique and determine the reflected phase characteristics.

29 Design parameters Substrate thickness The substrate thickness is related to the losses, phase sensitivity and phase range of the single array element. In each phasing technique there is a tradeoff between the phase range and phase sensitivity. A decrease in substrate thickness yields an increase in phase range and sensitivity. The influence of substrate thickness is examined for the various phasing techniques addressed in this thesis. Due to the large steps in discrete thickness values provided by the laminates manufacturers [28], the substrate thickness is not a major design parameter. Small adjustments of the resonance frequency or the phase sensitivity can not be obtained by selecting an adjusted substrate thickness Dielectric constant Not only the free space wavelength λ 0 is of importance to the propagation properties but also the propagation inside the dielectric has to be taken into account. The wavelength inside the dielectric is given by [27] λ r = c f ɛ r (3.1) where c is the speed of light in the free space, f is the frequency and ɛ r is the relative dielectric permittivity of the substrate. Decreasing the dielectric constant increases the resonance frequency. The dielectric constant for microstrip arrays is chosen in the lower range [27]. The dielectric constant for the substrate is fixed throughout the thesis, unless specified differently, at ɛ r = At this point it can be concluded that in the design of MRA (using commercial microstrip laminates) the dielectric permittivities and the height of the substrates are less suitable to be used as tuning parameters Ground plane dimensions The reflection from the ground plane is an important parameter that needs to be taken into consideration while designing the reflectarrays [8]. The ground plane has a considerable effect on the reflected phase and needs to be studied carefully. The ground plane dimensions are based on an element spacing of λ 0 /2 which is most often selected in array designs and is depicted in Figure 3.1.

30 18 Single element considerations 3.3 Performance parameters The single element performance is evaluated using parameters specified in the following subsections. Most parameters are directly obtained from simulation or measurement results Far field phase The most important parameter needed for the design of MRA is the reflected phase at the broadside as function of a variable U used in the phasing technique. This parameter gives the designer the ability to design an array and to predict its performance. The variable U can represent: the dimensions of the patch, the height of the substrate, the sampling or any other parameters which could effect the scatter phase. The phase center of the elementary antenna at the origin is the reference phase for the phase curves presented in this thesis. For the phase curves the following aspects should be considered: Phase range The phase range is the complete change in phase values that can be reached bythesingleelement. Aphaserangeof360 ensures the largest possible scan range. Phase sensitivity The phase sensitivity can be described as the steepness of the phase versus a variable. This varibale can be different for the different phasing technique. In equation form it can be described as the differential of the curve ψ sens = δψ (3.2) δu where ψ is the reflected phase, and U is the parameter for a specific phasing technique. The phase sensitivity is maximum at the center of the phase curve which corresponds to the resonance point. The phase sensitivity and phase error determine the bandwidth of the single element. Phase error The phase error is closely related to the bandwidth of a single element. It can be considered as the difference in phase caused by a shift in the frequency. The phase curve at one frequency differs from the phase curve at another

31 Performance parameters 19 frequency and the difference between them give the phase errors. phase errors limit the MRA bandwidth. These Far field magnitude The magnitude of the electric or magnetic field at the far zone gives an indication of the radiated power by an antenna element. The maximum radiation depends on the effective area of the antenna element in its environment Bandwidth As just suggested the bandwidth of the single element in a MRA is determined by the steepness of the phase curve. A more formal definition can be the frequency range in which the phase errors are kept within a certain limit. The bandwidth for a phasing technique can be derived from the reflected phase as function of frequency. Moreover, this describes also the range in which the spacing between the phase curves for different parameters is within certain boundaries. This can be then seen also as bandwidth definition of the single element. At last, the estimation of the bandwidth can be determined by studying the radiation pattern of the array for each phasing technique Cross-polarization In MRA a linear polarized wave is assumed to excite surface currents on the microstrip antenna element. Depending on the geometry, the surface currents have a current pattern with minor components perpendicular to the polarization of the incoming wave. These components are responsible for some cross-polarization. In this thesis cross-polarization is usually measured Resonance frequency The resonance frequency of a single microstrip antenna element is the frequency at which the wave matches the electrical length of the current excited on the patch. At the resonance frequency the imaginary part of the surface current is zero [29]. For the microwave circuits integrated with the element the maximum power is radiated and no power is reflected back to the source at this resonance frequency. As was mentioned the phase curve is most sensitive to small changes in the resonance frequency and resonance length. Hence the accurate determination of these parameters plays a major role in accurate design of the antenna system.

32 20 Single element considerations Near field The radiating edges [27] based on the transmission line and cavity model determine the radiation characteristics of the microstrip antenna element. Study of the near field confirms such phenomena. The surface current distribution can provide an indication of the radiating edges; near field analysis quantifies the effect of the radiating edge as will be shown in subsequent chapters Surface currents The induced surface currents determine the antenna properties [29]. In order to have a thorough understanding of a phase-shifting technique, study of the induced surface currents is essential. An accurate study of the surface currents can offer extensive knowledge of the patch behaviour. The important parameters of the antenna can be determined by the distribution of the surface current.

33 Chapter 4 RF Reflection and transmission of semiconductor material under illumination of light potential approach to realize a reflectarray is based on the use of semiconductor material, which material properties alter when it is illuminated by a light source. This concept is studied in this chapter. It is well known that at microwave frequencies, antenna beam steering is commonly realized using phased array antennas. Such approach is expensive and require time consuming design procedures [30]. As the frequency increases, the complexity and costs make the realization of phased array antennas even more, demanding and difficult. A preliminary study of implementing a non-mechanical and low-weight antenna that can form and scan the beam is presented in this chapter. 4.1 The concept The antenna uses a semiconductor wafer, a MMW source and an array of lasers for the optical masking (Figure 4.1). Each optical laser in the array illuminates a number of pixels in the semiconductor wafer. In the semiconductor wafer the spatially varying density of charge carriers changes the properties of the semiconductor. The portions of the semiconductor that is illuminated with light becomes a conductor [31]. The duration of the semiconductor at this state depends on the carrier lifetime of electrons and holes, τ n. Hence those portions can reflect- or transmit the incident field coming

34 22 Semiconductor material under illumination of light from an electromagnetic wave source. y x Metalization (Dipole) Semiconductor material z MMW Feed source Light illumination: holographic imaging Array of low power laser Addressing the array Look up table Figure 4.1: The antenna system based on the optical technology, in this case dipoles elements are projected. Since the optical illumination can be manipulated locally, the spatial distribution of carriers can be chosen in such a way that a reflected or transmitted electromagnetic wave can form a beam of desired shape and direction. By changing the optical masking, beam scanning in space becomes feasible [32, 33]. The antenna elements in the array configuration can be dipoleor patch-like. The concept of such an antenna for beam-scanning based on optical imaging is depicted in Figure 4.2. Semiconductor: not illuminated Semiconductor: illuminated with light (a) (b) (c) Figure 4.2: Projection of microstrip reflectarray on semiconductor with beam scanning capabilities over a time interval τ n with different element geometries: (a) equal size; (b) variable size; (c) rotation. Specific projections of the light source on the semiconductor correspond to each desired scan angle. Prior knowledge of the coordinates for each element in the array geometry is a must and needs to be stored in advance

35 Modelling the reflection and transmission coefficients 23 in a look-up table for all possible elevation- and azimuth scanning angles, θ scan = θ 0, φ scan = φ 0, respectively. Based on the data from the look-up table, the low power lasers in the array will be addressed individually to be on or off. The lasers that are in on state determine the optical masking. In this way the wafer can be arbitrary masked as function of the time. Since the duration τ n of the masking depends heavily on the carrier lifetime of the semiconductor material, the masking can be reconfigured rapidly. In this way the radiation from a MMW source can be scanned through space. In order to understand the behaviour of the semiconductor material illuminated by the light source, a silicon wafer is encapsulated between two air-filled microwave (MW) or MMW waveguides and is illuminated with high power laser pulses. The wave reflected from and transmitted through semiconductor material is modelled using the S-T matrix. The propagation constant in the semiconductor material depends on the density of the electron-hole pairs generated by the optical source. The generated electron-hole pair changes the conductivity of the material so that it becomes more a good conductor instead of insulator. It is demonstrated that, once a semiconductor material is illuminated with an optical source, a nearly complete reflection can occur. In order to validate the theoretical results measurements have been done. The measurements are carried out on a silicon wafer in order to get an indication of the magnitude of reflection and transmission of signals in the MMW and MW region. This property of semiconductor material can be used to realize a reconfigurable reflectarray antenna that can scan the beam of an incoming wave generated by the fixed feed. 4.2 Modelling the reflection and transmission coefficients Figure 4.3 presents the overall T-matrix that relates the input-port to the output-port of a microwave network. Based on the transmission line theory, Γ in κ 1 T A linear two-port network ι 2 ξ a Waveguide b ι 1 (a) κ 2 (b) Figure 4.3: (a) Two-port network for determining the reflection and transmission of a microwave device; (b) aperture waveguide with width of a and height of b. the input reflection coefficient, Γ in, and output transmission coefficient, ξ, are related to the wave components and the S-matrix and T-matrix compo-

36 24 Semiconductor material under illumination of light nents via [34] Γ in =S 11 ι 1 κ2 κ = T 12 1 =0 T 22 ξ =S 21 ι 2 κ2 κ = 1 (4.1) 1 =0 T 22 where κ and ι are the incident and scattered waves, respectively. The conversion relations between the scattering matrix, S, and the transmission matrix, T, is given by [34] [ T = and [ S = T 11 T 12 T 21 T 22 S 11 S 12 S 21 S 22 ] = ] = [ S12 S 21 S 11 S 22 S 21 S 11 S 21 [ T12 S 22 S 21 1 S 21 T 11 T 22 T 12 T 21 T 22 T 22 1 T 22 T 21 T 22 ] ] (4.2) (4.3) For further analysis the geometry consisting of the silicon wafer of finite length l s encapsulated by two air-filled hollow rectangular waveguides is modelled as shown in Figure 4.4. Moreover, Figure 4.4 illustrates the S-T matrices for each waveguide section and for each transition, the corresponding lengths and their respective propagation constants. The T-matrices of l 2 l s l 1 γ 0 γ g γ 0 S l2 S 2 S s S 1 S l1 T l2 T 2 T s T 1 T l1 T Figure 4.4: Configuration of S-T matrices with the silicon wafer. the network becomes [35] [ ] e γ 0 l 1 0 T l1 = 0 e γ 0l 1 [ 1 T 1 = 2 γg + γ 0 γ g γ 0 γ g γ 0 γ g γ 0 γ g + γ 0 ] (4.4) (4.5)

37 Modelling the reflection and transmission coefficients 25 [ ] e γ gl s 0 T s = (4.6) 0 e γgls [ ] 1 T 2 = 2 γg + γ 0 γ 0 γ g (4.7) γ g γ 0 γ 0 γ g γ g + γ 0 [ ] e γ 0 l 2 0 T l2 = 0 e γ 0l 2 (4.8) where T 1 and T 2 represents the interface between waveguides with different propagation constants. γ g is the propagation constant in the silicon and will be determined in the next section. γ 0 is the propagation constant in the air-filled waveguide and is given by { j ω γ 0 = 2 μɛ 0 ( π a )2 if ω 2 μɛ 0 > ( π a )2 ( π a )2 ω 2 (4.9) μɛ 0 otherwise where a is the width of the waveguide cross section. j is imaginary unit, ω is the radial frequency of the signal, ɛ 0 is the free space permittivity, μ = μ 0 is the free space permeability Propagation constant in semiconductor material The propagation constant γ g of electromagnetic waves in semiconductor material is derived from a similar procedure as used in Equation (4.9). It is noted that the propagation constant γ g in the guide differs from the intrinsic propagation constant γ s of a material. γ g is given by [36] γ 2 g = γ 2 s + k 2 c (4.10) where k c =( π a ) is called the cut-off wave number. The electrical properties of a semiconductor material under illumination from an optical source will change. In general the propagation constant in a medium is given as [36] γ s = jωμ(σ + jωɛ) (4.11) where σ is the conductivity of the material. The conductivity for semiconductor is given by [37] σ = N n q n μ n + N p q p μ p (4.12) where μ n and μ p are the mobility of electrons and holes in cm2 Vsec. N n and N p are the electron and hole concentration in cm 3, which will be altered when extra electron-hole pairs are generated. In general in thermal equilibrium we have N n =N p = N 0,whereN 0 is the intrinsic carrier concentration. Substituting Equation (4.12) in (4.11) and combining with Equation (4.10) leads to the propagation constant in the semiconductor material. The only unknown parameter is the concentration of optically generated electron-hole pairs which will be addressed in the next subsection.

38 26 Semiconductor material under illumination of light Electron-hole pair generation rate The electron-hole pair concentration generated by the optical source is derived from the following approach. It is shown that photons with energy greater than the band-gap energy E g of the material can be absorbed in the semiconductor, thereby creating electron-hole pairs. The intensity I v (x) is given in the unit of energy cm 2 sec ; αi ν(x) is the rate in which energy is absorbed per unit volume. α is the absorption coefficient in cm 1 and depends on the semiconductor material and wavelength of the optical source. If we assume that one absorbed photon at energy hν creates one electron-hole pair, then the generation rate of electron-hole pairs equals [38] dg dt = αi ν hν (4.13) where hν is the photon energy in ev and is related to the wavelength of the optical source via E = hν = 1.24 (4.14) λ(in μm) The generated electron-hole during a time interval Δτ is given as δn(t =Δτ) = dg Number of electrons Δτ,in unit of dt cm 3 (4.15) where Δτ is duration of the pulse. Assuming that there is no spatial variation in the excess carrier concentration and that at time t = 0 the electron-hole pairs have been generated uniformly based on Equation (4.15), the time dependence of the generated electron-hole pairs can then be given as [38] δn(t) =δn(0)e t τn (4.16) where τ n is the carrier life time of electrons and holes, δn(0) is the concentration of excess carriers which exist at t = 0. Note that for pure (undoped) Si in a thermal equilibrium N 0 = per cm 3 at room temperature while the total carrier concentration that can be generated is: per cm 3. Electron-hole pair concentration as indicated in Equation (4.12) is adjusted by the generated electron-hole pair given by Equation (4.16) Reflection and transmission coefficients The overall T-matrix of the microwave network depicted in the Figure 4.4 is determined by the matrix multiplication [ ] T T = 11 T 12 = T T 21 T l2 T 2 T s T 1 T l1 (4.17) 22

39 Numerical results 27 Inserting Equations (4.4)-(4.8) in (4.17), and after some manipulations and rearranging, the elements of the total T-matrix yields T 11 = 1 [(γ g + γ 0 ) 2 e γgls (γ g γ 0 ) 2 e γgls ]e γ 0(l 1 +l 2 ) (4.18) 4γ g γ 0 T 12 = 1 4γ g γ 0 [(γ g 2 γ 0 2 )(e γgls e γgls )]e γ 0(l 1 l 2 ) T 21 = 1 4γ g γ 0 [(γ g 2 γ 0 2 )(e γgls e γgls )]e γ 0(l 1 l 2 ) (4.19) (4.20) T 22 = 1 [(γ g + γ 0 ) 2 e γgls (γ g γ 0 ) 2 e γgls ]e γ 0(l 1 +l 2 ) (4.21) 4γ g γ 0 Substituting Equation (4.19) and (4.21) into equation (4.1) leads to the expression for the input reflection Γ in = (γ g 2 γ 2 0 )(e γgls e γgls )e 2γ 0l 2 (γ g + γ 0 ) 2 e γgls (γ g γ 0 ) 2 (4.22) e γgls The transmission coefficient is given as [35] ξ = 4γ g γ 0 [(γ g + γ 0 ) 2 e γgls (γ g γ 0 ) 2 e γgls ]e γ 0(l 1 +l 2 ) (4.23) 4.3 Numerical results Using the Equations (4.9) to (4.16), (4.22) and (4.23), the reflection and transmission coefficient of a silicon wafer under illumination of an optical source is determined. The waveguide is operating in the dominant TE 10 mode. The operational frequency is 35.0GHz. The wafer is encapsulated by two air-filled waveguides. The dimension of the cross-section of the waveguide is 3.5mm 7mm. The thickness of the silicon wafer is 0.3mm. The optical pulse has a Gaussian shape with a wavelength of 1.06μm and a pulse duration of 10ns. At this wavelength the absorption coefficient of silicon is α = 10cm 1 [38], with a typical value for the mobility of silicon μ n = 1350 cm 2 Vsec [37]. The reflection and transmission as function of time are shown in Figure 4.5. The parameter is the energy density of the optical source in mj cm illuminating the semiconductor at time t = 0. The conductivity of the silicon 2 wafer as a function of time and for different values of the energy density of the optical source is plotted at linear and logarithmic scales in Figures 4.5(c) and 4.5(d), respectively. It is evident that this considerable change in conductivity leads to a nearly total reflection of the electromagnetic waves. A carrier lifetime of τ n =10μs has been chosen. It is shown that when the generation rate decreases, the input reflection also decreases. The results are in good agreement with results published [39].

40 28 Semiconductor material under illumination of light 1 1 Reflection coefficient Energy Intensity mj/cm Transmission coefficient Energy Intensity mj/cm Counductivity (S/m) Time [μsec] 2.5 x (a) Energy Intensity mj/cm Logarithmic Counductivity (S/m) Time [μsec] (b) Energy Intensity mj/cm Time [μsec] (c) Time [μsec] (d) Figure 4.5: (a) Reflection, (b) transmission and (c) the conductivity σ (d) the conductivity in logarithmic scales of silicon after being illuminated by an optical source for different values of optical energy density in mj cm Experimental results The measurements presented in this chapter were performed at the Applied Physics department of the Faculty-Applied Sciences of Delft University of Technology. Measurements have been done to validate the theoretical results and to determine experimentally the reflection- and transmission of electromagnetic waves incident on a piece of semiconductor material which is illuminated by an optical source. A silicon wafer with a thickness of 0.3mm is encapsulated between two waveguides. One of the waveguides had a slot on the rear side for illuminating the wafer with the laser pulse. Two detectors (DC to 50GHz) are used to measure the reflection and transmission of the MW or MMW source. A mirror and a lens are used to focus the laser light into the waveguide. The laser was an Nd +3 Yttrium Aluminium Garnet (Y 3 Al5O 1 2 or YAG) operating at an infrared wavelength of 1.06μm, with a maximum energy of 300mJ and a Guassian pulse with duration of

41 Experimental results ns leading to a peak power of 15MW [40]. A three-channel digital oscilloscope, TDS784A, is used for simultaneously measuring the reflection and transmission and synchronizing X-band The measurement set-up in the X-band is depicted in Figure 4.6. S Synchronization Lens Waveguide Silicon Input microwave signal Light Mirror Coupler Slot Reflection Detector for measuring Transmission Attenuator f=9.47ghz Laser P max=16dbm 20dB LNA Detector for measuring Reflection Ch1 Ch2 TDS 784A Ch3 S Figure 4.6: Reflection- and transmission measurement set up at X-band. The silicon wafer is encapsulated by two open waveguides operating in the dominant TE 10 mode and having aperture dimensions of b = 10.0mm and a = 22.0mm. For generating the incident electromagnetic waves on the silicon wafer, a MW source at 9.47GHz with a maximum output power of 16dBm has been used. The source is connected to the hollow waveguides via a -20dB wide-band coupler. The coupler is simultaneously used for measuring the wave reflected by the silicon; use is made of a detector, which subsequently is connected to one of the channels of the digital oscilloscope. The transmission wave through the silicon wafer is measured by the second channel of the digital oscilloscope using another detector. The third channel of the oscilloscope is used for synchronization. The mirror and lens were used for aligning and focusing the laser light. Before switching-on the laser, the reflection caused by the silicon (high dielectric constant) and the open-ended waveguide is tuned away with a matching stub. The relative reflection- and transmission voltage before and after switching-on the laser is depicted in Figure 4.7. The straight line represents the reference voltage. The portion of the reflected pulse between 42μs <t<70μs becomeslower than the reference signal. This phenomena could not be explained. It can be concluded that a complete reflection occurred after switching-on the laser.

42 30 Semiconductor material under illumination of light Reference pulse Reflected voltage (mv) Reflected pulse Reference pulse Transmitted voltage (mv) Transmit pulse Time (μs) (a) Time (μs) (b) Figure 4.7: Measured X-Band (a) reflection and (b) transmission pulse after the silicon is illuminated with light, the laser pulse of 20ns occurs at t=0 sec Ka-band Figure 4.8 presents the block diagram of the measurement set-up at Ka-band and is almost similar to the measurement set-up in the X-band. Waveguide MMW source Silicon Waveguide Coupler 3dB coupler Adaptor 2.4 mm Reflection set-up mm Transition HP8474C Detector.01-33GHz Slot Mirror Light 10dB coupler Oscilloscope Laser Lens Synchronization Transmission set-up Adaptor 2.4 mm mm Transition HP8474C Detector.01-33GHz Figure 4.8: Reflection- and transmission measurement set up at Ka-band. A picture of the measurement set-up with the IRCTR AB MMW Net- Work Analyzer (NWA) system is presented in the Figure 4.9. For generating the incident electromagnetic waves on the silicon wafer, a MMW source at 32.0GHz with a output power of 3dBm is used. The source is connected to the hollow waveguides where the silicon is located via a -3dB and -10dB

43 Discussions 31 Laser (a) (b) Figure 4.9: (a) IRCTR AB mm NWA (10MHz up to 110GHz) employed as the MMW source for the measurement, (b) measurement set-up at the department of Applied Physics, Faculty of Applied-Sciences of Delft University of Technology. wide-band waveguide coupler. Moreover, both couplers are simultaneously used to measure the reflection wave through an adaptor and a coaxial transition from 2.4mm to 3.5mm. This signal is then measured by a detector, which subsequently is connected to one of the channels of the digital oscilloscope. The transmission wave through the silicon wafer is measured by the second channel of the digital oscilloscope using another detector through an adaptor and a coaxial transition from 2.4mm to 3.5mm. Similar to the measurement set-up in the X-band, the third channel of the oscilloscope is used for synchronization. The laser pulse is guided into the waveguide containing the silicon via a mirror and a lens. In this case the silicon wafer is encapsulated by two open waveguides operating in the dominant TE 10 mode having an aperture with b =3.5mmanda = 7.0mm. Before turning-on the laser, the reflection caused by the silicon (high dielectric constant) is tuned away with a MMW tuner. Figure 4.10 depicts the relative transmission and reflection of the silicon wafer before and after the laser is switched on. The figures present measured results for a single shot measurement, and a measurement averaged over 16 laser pulses. For clarity the average measurement is inverted. An on/off ratio of almost 20dB is measured. A complete reflection occurs after the laser is turned on. Figure 4.11 shows the comparison between simulated and measured results. A good agreement is observed. 4.5 Discussions After examining the results given in the previous sections it can be concluded that the transmission and reflection is independent of the operational fre-

44 32 Semiconductor material under illumination of light Transmission coeficient (mv) Single shot Average Transmitted voltage [mv] Single shot Average Time (μs) (a) Time [μs] Figure 4.10: Measured (a) reflection and (b) transmission pulse after the silicon is illuminated with light at Ka-band, the laser is turned on at t=0 sec. 1 1 (b) Reflection coefficient Measurement Simulation Transmission coefficient Measurement Simulation Time [μsec] Time [μsec] (a) (b) Figure 4.11: Comparison between the measured and simulated results, (a) reflection (b) transmission. quency. In this section the effect of some of the features of the semiconductor material and optical source on the overall performance of the antenna system is briefly addressed. If the antenna would be used for short range communications or radar applications, the system would operate at low microwave power. The following aspects should then be taken into account. Dimensions: Consider a MMW collision avoidance radar sensor for short range detection using a beam switching antenna with 3.0dB beamwidth of 8.8 and 4 in the azimuth- and elevation plane respectively [41]. This beamwidth requirement leads to a directive antenna with directivity

45 Discussions 33 D = 30dB. An array antenna with square elements could satisfy such a requirement. Assuming that each antenna occupies a space of 0.5λ 0,withλ 0 representing the free-space wavelength, at a center frequency of f = 35.0GHz (λ 0 = 8.6mm) the total size of the antenna would be cm 2. Similair results are indicated in [8]. Carrier lifetime: This parameter has a direct effect on the radar range and the duration of the beam steered towards a certain elevation and azimuth angle. The silicon material used for the measurements had a carrier lifetime of 10μs. This leads to a radar range of maximum 1.5km. Diffusion length: Diffusion length is a measure of how far the density of generated electron-hole pair concentration has spatially propagated in the x- and y-direction by diffusion in time t. This may leads to the distortion of the projected antenna geometry on the semiconductor material. This parameter has effect on the dimensions of the antenna element which effect the scan angle and side lobe level. In the design process the error caused by the diffusion length can be calibrated in advance. The diffusion length, L n, is related to the carrier life time via L n = D n τ n (4.24) where D n is diffusion coefficient. For the used silicon D n =15 cm2 sec and τ n =10μs. This leads to L n = 0.122mm. The effect of diffusion length is shown in Figure 4.12 for an 1 5 beam scanning array antenna using variable-sized elements in order to obtain scan angels at 0, no diffusion with diffusion 40 no diffusion with diffusion E plane radiation pattern [db] E plane radiation pattern [db] Elevation angle [Degree] (a) Elevation angle [Degree] (b) Figure 4.12: The effect of the diffusion length on the pattern of a variablesized 1x5 linear array antenna at f = 35.0GHz for (a) 0,(b)10.

46 34 Semiconductor material under illumination of light Examining Figure 4.12, it can be concluded that the diffusion length affects the antenna pattern due to the small wavelength at such high frequencies. This effect becomes less at the lower frequency bands. Generated heat in the silicon wafer caused by the MMW source: MMW applications operate usually at low transmit power. To have an indication of the heat generated in the silicon wafer the following consideration is appropriate. Due to the high reflection less than 1% of the power is absorbed in the semiconductor material. Assuming a transmit power at the feed of 1W and neglecting the free space losses this power will impinge on the silicon wafer. Due to the high reflection the absorbed power in the wafer is in the order of 10mW. The power density for a array antenna is then 0.15 mw at Ka-band. In cm this case no cooling is needed. If the antenna would be 2 part of a highpower radar air cooling can be used. Part of the silicon wafer area is not illuminated by the laser. The energy gap of the silicon is much higher than the energy of the MMW source meaning that the nonilluminated parts of the wafer are transparent for the electromagneticwaves. Laser power intensity: The intensity of the photon flux decreases exponentially as function of distance in the semiconductor material according I v (x) =I v0 e αx (4.25) where x is the distance and α is the absorption coefficient. I v0 is the incident power density of the optical source. At the operating wavelength the absorption coefficient of the silicon is 10cm 1. The thickness of the silicon wafer used for the measurement was 0.3mm. The power density of the optical source after this distance is I v (x) I v0 = e =0.74 (4.26) This means that around 30% of the optical power is absorbed and 70% passes through the silicon. It is therefore advantageous to use the laser power in a more efficient way. It is observed during the measurement that by a fraction of the maximum pulse energy (in order of 1 mj cm ) the silicon was saturated and a complete reflection were occurred. For 2 a wafer with an area of 100cm 2 a laser with 100mJ is needed. The maximum power for a 20ns laser pulse would be 5MW. However, due to the carrier life time of 10μs, a laser with power of 10kW would satisfy the requirements. In case only 30% of this power is needed for total reflection, then 3kW source would satisfy the requirements. Figure 4.13 shows the simulated reflection- and transmission coefficient using lasers with lower energy pulse.

47 Conclusions Reflection coefficient Energy Intensity mj/cm Transmission coefficient Energy Intensity mj/cm Time [μsec] (a) Time [μsec] (b) Figure 4.13: (a) Reflection, (b) transmission of silicon after being illuminated by an optical source with lower optical energy density in mj cm 2. Generated heat in the silicon wafer caused by the laser source: An equilibrium electron-hole concentration that has led to some of the results reported in this chapter would correspond to a very high temperature near the melting point of silicon. Using a laser with a lower power would decrease this effect. However, extensive research is needed for controlling the generated heat and amount of cooling that is needed for the final design. Based on the theoretical- and measurement results presented in this chapter, it can be concluded that by manipulation of the semiconductor material, a reflectarray antenna can be realized to form and scan a beam. Due to the high cost of laser, the optics components and mechanical support structures, it was decided to shift the research into the direction of beam scanning based on MRA. This approach will be addressed in subsequent chapters. 4.6 Conclusions A piece of silicon wafer with thickness of 0.3mm is encapsulated between two media in a closed X-band or Ka-band waveguide structure. The silicon wafer is illuminated with a high power laser pulse. The numerical and measurement results show that there is considerable change in the conductivity of the silicon wafer before and after switching-on the laser. The incident laser pulse on the wafer is responsible for a considerable change in reflection and transmission of the electromagnetic waves. This feature of semiconductor materials can be used to realize beam scanning antennas. The duration in which the beam can be directed towards a certain point in space depends heavily on the carrier life time of electron-hole pairs of the semiconductor

48 36 Semiconductor material under illumination of light material. The measurement results are modelled using the S-T model. There is a good agreement between simulation and measurement results. Elements of novelty A new concept for the realization of a reflectarray antenna using optical imaging has been proposed. In particular, the reflection and transmission properties of a silicon wafer was formulated and an analytical solution was proposed. Moreover, it was demonstrated that there is nearly complete reflection of electromagnetic waves in a closed waveguide structure by enlighting the silicon wafer. Experimental investigations of a silicon wafer illuminated by a laser pulse were performed by implementing a unique and complex measurement set-up. It was demonstrated experimentally that a complete reflection occurred after the silicon is illuminated by a laser pulse. A complete overview of this investigation is presented in the references [31], [42].

49 Chapter 5 Formulation of the integral equation for microstrip reflectarray antennas ne of the most important steps towards the realization of MRA is the determination of the amplitude and phase of the excited current on the patch antenna element. The microstrip antenna analysis is complex due to the presence of the dielectric inhomogeneity, narrow band electrical characteristics and a wide variety of patch and substrate configurations. The analytical models which account for the dielectric substrate in a rigorous manner are referred to as the full wave solutions [43, 29]. These mathematical models use exact Green s function in a MOM solution for the dielectric substrate and allow for including the space wave radiations, surface wave modes, dielectric loss and coupling to external elements. MoM analysis is a well known technique for accurate calculations of antenna parameters [43]-[45]. It can be used to determine all major antenna parameters under consideration. The usage of the Green s function in a mathematical model ensures that predictions are accurate and versatile [46]-[49]. Features of a full wave solution include: Accuracy: full wave analysis results into accurate prediction of the input impedance, radiation pattern, mutual coupling, radar cross section, etc; Completeness: the model includes the effects of surface waves, and space wave radiation; Versatility: the model can be used for arbitrarily shaped microstrip elements and arrays, various types of feeding networks and for radiation as well as for scattering problems;

50 38 Integral equation for microstrip reflectarray antennas Computational efficiency: codes that are required to solve a computational complex mathematical model require careful programming to be efficient as well as accurate. In this chapter results of a full wave solution are presented based on the calculated complex (amplitude and phase) current coefficient on the patch antenna element. This is important for determining the reflected field coming from a MRA. The currents along the patch element are described in so-called Entire Basis (EB) and Piece Wise Sinusiodal (PWS) expansion modes. Extensive use has been made of the integral equations MOM-based software algorithms as derived in Appendix A. The complete derivation is addressed in [50]. To validate the numerical solution the analysis was first carried out at S-band [44]. The frequency band was then shifted to Ka-band. The motivations to develop our own software were: to investigate approximations and error propagation issues. The numerical solution of Maxwell s equations necessitate often simplifications and neglections. These impact and the presence of systematical errors can then be studied; commercial software packages are expensive tools without direct access to the source code. It has been demonstrated that improved insight can be achieved with the in-house implementation of MoM-based algorithms. 5.1 Numerical validation This section gives the results of the EB and PWS modes at S- and Ka-band. Before presenting the results it is worthwhile to address the singularity in the Green s function. This aspect is important because the integral needs to be solved independently from the inherent singularity in the Green s function in order to avoid numerical difficulties Singularity Examining Equations (A.49) and (A.50) reveals that there is a singularity and integration in Equation (A.51) should therefore be done carefully. The procedure is such that the location of the singularity is determined by searching for the zeroes of Equation (A.50) in advance. For this purpose use is made of the Least Minimum Square (LMS) method and in the integral Equation (A.51) integration over the singularity for β = β 0 is excluded

51 Numerical validation 39 according 2π = 0 ( )dk x dk y [ β 0 ϑ ( )dβ + 0 β0 +ϑ β 0 ϑ ( )dβ + β 0 +ϑ ] ( )dβ dα (5.1) Usually ϑ is chosen to be a very small value. The integral over the singularity then becomes [44] I singular = β0 +ϑ β 0 ϑ Applying the residue theory leads to ( )dβ = β0 +ϑ β 0 ϑ f(β) dβ (5.2) T m (β) I singular = πjf(β 0) T m (β 0 ) (5.3) where the prime represent the derivative of function T m with respect to β. The value of ϑ is chosen such that it guarantees the convergence and stability of the solution to the Equation (5.2). In this thesis ϑ is chosen to be ϑ = Numerical results First analysis for both EB and PWS modes was carried out at S-band ( GHz) to validate the correctness of the integral equations and the codes in comparison with literature [44]. The patch has square dimensions with L = W = 4.02cm, ɛ r = 2.55, h = 1.59mm, and tanδ = The impedance matrix as given by Equation (A.51) was implemented using Equations (A.54), (A.56), and (A.60) for the current distribution, and Equation (5.3) for determining the singularity. 3-divisions i.e. M = N = 3 along the x- andy-axis, were selected. The impedance matrix of dimension 9 9 for each frequency was determined and it is referred to as 9 PWS. Figure 5.1 represents the comparison of 1 EB with 1, 3 and 9 PWS expansion modes respectively. There is a good agreement with the results reported in [29] and [44]. Analysis was then carried out at Ka-band ( GHz) for various PWS expansion modes. The initial dimensions of the patch started from the transmission line theory which led to L = 2.6mm, W =3.4mmforɛ r = 2.33, and h = 0.381mm. These parameters were then optimised using the analytical model presented in this chapter and compared to the results from the MOM-based EM simulator software package FEKO [51]. Figure 5.2(a) shows the comparison for 1, 3 and 9 PWS modes. In both cases of S- and Ka-band the 9 PWS performance is less accurate because in the case of 9 PWS modes we assume a current distribution over the width of the patch which is non-uniform.

52 40 Integral equation for microstrip reflectarray antennas EB and 1PWS expansion Mode Comparison EB with 3 PWS expansion Mode Complex Current Coefficient Magnitude [A/m] EB 1 PWS Real Imaginary Complex Current Coefficient[A/m] EB 3 PWS Real Imaginary Frequency [GHz] Frequency [GHz] (a) EB with 9 PWS expansion modes (b) Complex Current Coefficient[A/m] EB 9 PWS Real Imaginary Frequency [GHz] (c) Figure 5.1: Comparison of the excited current on the patch as function of frequency for: (a) 1 EB vs 1 PWS mode, (b) 1 EB vs 3 PWS mode, (c) 1 EB vs 9 PWS modes. There is a small shift in the resonance frequency. The analysis has led to the dimensions of L = 2.45mm, W =3.3mmforf r = 34.5GHz. For 3 and 9 PWS expansion mode, the solution of equation (A.46) gives 3 and 9 current coefficents. The average of the real and imaginary parts of the current from the segments along the resonance length is taken for the comparison. Figure 5.2(b) presents the comparison between our own results with 1 and 3 PWS modes and the results from FEKO. There is a good agreement between both results. Distributed ASCI supercomputer 2 (DAS-2) The acquisition time of the MATLAB codes was considerably improved by using DAS (Distributed ASCI Supercomputer) namely by a factor of almost 30. The reasons for such a long processing time are:

53 Numerical validation 41 Current on the patch [A/m] PWS 3 PWS 9 PWS Freq (GHz) (a) Complex Current Coefficient Magnitude [A/m] Real Imaginary 1 PWS 3 PWS FEKO Freq (GHz) (b) Figure 5.2: Comparison of the excited current on the patch as function of frequency for: (a) 1, 3 and 9 PWS expansion modes, (b) FEKO with 1 and 3 PWS expansion mode. the type of processor and amount of memory, Pentium IV processors, 1.2 GHz with 512MB of RAM; the codes were written in MATLAB which requires a lot of overheads during the execution; the frequency spacing was small, about 600 frequency points for generating one figure; MATLAB becomes extremely slow when more for loops are used. DAS is a wide-area distributed computer network consisting of 200 Dual Pentium-III nodes. The machine is built out of clusters of workstations, which are interconnected by SurfNet, the Dutch university Internet backbone for wide-area communication, whereas Myrinet, a popular multi- Gigabit LAN, is used for local communications. The DAS machine is used for research on parallel and distributed computing. DAS-2 is a five-cluster wide-area distributed system designed by the Advanced School for Computing and Imaging (ASCI). DAS-2 located at the five universities. The cluster at the Vrije Universiteit contains 72 nodes, the other four clusters have 32 nodes (200 nodes with 400 CPUs in total). The system was built by IBM. The operating system on DAS-2 is RedHat Linux. Figure 5.3 shows DAS-2 systems at Delft University of Technology, and VU University Amsterdam. The DAS-2 system is funded by NWO (the Netherlands organization for scientific research) and the participating universities. The following institutes and organisations are directly involved in the realization and running of DAS-2:

54 42 Integral equation for microstrip reflectarray antennas Vrije Universiteit, Amsterdam (VU); Leiden University (LU); University of Amsterdam (UvA); Delft University of Technology (TUD); The MultimediaN Consortium (UvA-MN). (b) (a) Figure 5.3: DAS-2 systems at: (a) Delft University of Technology, (b) VU University Amsterdam. Each node contains the following hardware: Two 1-GHz Pentium-IIIs; at least 1 GB RAM (1.5GB for the nodes in Leiden and UvA, and 2GB for two large nodes at the VU); A 20GB local IDE disk (80GB for Leiden and UvA); A Myrinet interface card; A Fast Ethernet interface (on-board).

55 Conclusions 43 It is worth noting that building the next generation of DAS system, DAS-3, is in progress with more-advanced technology using revolutionary Optical interconnects. As one of its distinguishing features, DAS-3 employs a very novel internal wide area interconnect based on optical paths. DAS-3 consists of 272 computer nodes. The system is currently being built [52]. 5.2 Conclusions The detail of the dyadic Green s function for a current source has been derived in Appendix A. This function is used to set up electric-field integral equations for analysing microstrip patch antennas. The equations form the fundamental tools for accurate prediction of patch antenna parameters needed for designing MRA. The integral equations were solved using Galerkin s method and allow for accurately predicting the design parameters for the patch antennas. This is essential for reliable designing MRA antennas. There is a good agreement observed between the results presented here with the results reported in the literature. Elements of novelty The integral equations for determining the complex current excited on the patch were derived in Appendix A. The numerical results from own made software based on MOM agree with the results obtained in the literatures and FEKO. It is worth observing the efficiency, in terms of computing times requirement of the MOM computer implementations. This computational performance was achieved by adopting a new parallel programming using DAS systems. The computing times requirement is down-scaled with a factor of 30 compared to the computing times requirement on Pentium IV processors, 1.2GHz with 512MB of RAM.

56 44 Integral equation for microstrip reflectarray antennas

57 Chapter 6 Variable-sized phasing technique The variable-sized rectangular patch demonstrates the basics of the microstrip single element design for analysis of MRA. In this apporoach the patches with variable size is used to control the phase. Documentation on the performance and design of variable-sized patches is available, making it an excellent starting point to test the simulation environment. It is also a valuable approach for comparing with other phasing techniques. A clear overview of the theoretical and practical design of reflectarrays using microstrip patch elements of variable size is provided in [8] and [9]. In a MRA using a variable-sized technique the dimension of the patch in the direction of the incoming wave polarization is altered in order to shift the resonance frequency. This shift in resonance leads to the change of the reflected phase. The phase diagram allow for designing MRA antennas based on such variable-sized phasing technique. The results presented in this Chapter verify the correctness of the simulation model and offers a comparison method for other phasing techniques described. For simplicity and to demonstrate the concept of passive scanning it is decided to scan the beam only in one plane. 6.1 Geometry of a variable-sized patch The geometry of the Ka-band single variable-sized element is presented in Figure 6.1. The structure consists of a metal patch of variable length L on a grounded substrate. We assume a plane wave with linear horizontal polarization along the positive y-direction and with normal incidence on the patch. We change this coordinate system compared to the last chapter to be compatible with FEKO. The system parameters are presented in table 6.1.

58 46 Variable-sized phasing technique z Incident plane wave E =â y E 0 Substrate Metal W y h= 0.381mm x L GP size < λ0 2 =4.3mm GP size < λ0 2 =4.3mm Figure 6.1: Geometry of a variable-sized rectangular patch. Table 6.1: Parameters of the variable-sized Patch. Parameter Value Operating Frequency f = 34.5GHz Substrate Thickness h = 0.381mm Substrate Permittivity ɛ r =2.33 Loss Tangent tan δ = Patch Width, Length 3.3mm, Lmm Ground plane dimensions <λ 0 /2, λ 0 /2 Meshing λ 0 / Design procedure Infinite ground plane The general design procedure in MRA antennas including variable-sized patch is to determine the phase diagram. This phase diagram provides the data needed for designing the individual elements in the array. The procedure is started with selecting the dimensions of a single element to resonate at the desired frequency. The first step is dimensioning the single element using the transmission line model as reported in Chapter 5. The results are then optimised using the full wave MOM solution. The dielectric constant and thickness of the substrate are ɛ r = 2.33, h = 0.381mm respectively with

59 Design procedure 47 a loss tangent tanδ = The length and width of the microstrip patch based on the transmission line theory is given by [27] W = c 2 (6.1) 2f r ɛ r +1 [ ] 1 ɛ e = ɛ r +1 + ɛ r h 2 (6.2) 2 2 W ( )( ) W ɛ e +0.3 ΔL h =0.412 h ( )( ) (6.3) W ɛ e h +0.8 c f r = 2(L +2ΔL) (6.4) ɛ e where f r, ɛ e are the resonant frequency and the effective dielectric permittivity respectively. L and W are the length and the width of the patch along the y- andx-direction. Due to the fringing effect discussed in Chapter 3, Figure 3.1, dimensions of the patch along its length have been extended by ΔL. Equations (6.1) to (6.4) are used to calculate the initial dimensions of the patch; we found L = 2.6mm, W =3.4mmforɛ r = 2.33, and h = 0.381mm (see Chapter 5). The results are then optimised employing the infinite ground plane model explained in Chapter 5 using 1 and 3 PWS expansion modes and validated by the FEKO-result; we found L = 2.45mm and W = 3.3mm, respectively. The unknown current coefficients calculated as function of frequency have been shown in Figure 5.2(b) and compared with FEKO. The resonance frequency is in the range 34.5GHz 35.0GHz. The length of the rectangular microstrip patch was varied to observe the phase shift around the resonance frequency. The variation of phase with respect to patch length is shown in Figure 6.2. By examining Figure 6.2(a), a variation of 160 is observed. For design purposes it is required to stay within the slope of the curve. Note that in this case the effect of a limited ground plane has not been taken into consideration Truncated ground plane An important aspect to analyse the phase diagram is the effect of the ground plane size. In fact the total scattered field consists of two components: the specular reflection from the grounded dielectric substrate and the field reradiated by the patch. The total reflection coefficient is reported in detail in [8] and will not be discussed here. The reflection from the ground plane has a very relevant effect on the phase of the reradiated field. It dominates the range of the phase variation one can accomplish. In order to keep the analysis simple and because of the most-often-selected element spacing in the array

60 48 Variable-sized phasing technique FEKO MOM 3 PWS Truncated GP Infinite GP GP size = λ 0 /2 λ 0 / Phase [Degree] Phase [Degree] Length of Patch [mm] (a) Length of Patch [mm] (b) Figure 6.2: Phase length variation of a rectangular microstrip patch at 34.5GHz on: (a) infinite ground plane, comparison FEKO and our MOM results as derived in Chapter 4, (b) truncated ground plane, comparison with infinite ground plane. configuration, the antenna element is designed for a square ground plane, thus W gp = L gp = λ 0 /2. Figure 6.2(b) shows the phase comparison between the model using an infinite- and a truncated ground plane as function of the resonance length respectively. Examining Figure 6.2(b) the resonance frequency is shifted slightly and the phase range is affected considerably. In Appendix B it is demonstrated that the difference between the phase diagram of a single element with truncated ground plane and the phase diagram in the array is negligible. 6.3 Phase diagram The phase diagrams for a patch of variable-sized are shown in Figure 6.3 and demonstrate a characteristic behaviour. Figure 6.3(a) shows the reflected phase as function of patch length L for different frequencies. This figure represents the data needed for the array design. The total phase range at the operating frequency f is 327 and the maximum phase sensitivity is 29 /0.1mm. An increase in frequency causes a decrease in phase range and sensitivity. Increasing the frequency gives similar results as an increase in substrate thickness. The relative thickness of the substrate compared to the wavelength becomes for higher frequency thicker and a thicker substrate relates to a less sensitive system and a decrease in phase range. Figure 6.3(b) shows that the resonance frequency shifts to lower values of L for higher frequencies f. Increasing the frequency results in a reduced wavelength which is directly related to the patch resonance length.

61 Phase diagram GHz 34.5 GHz 36. GHz mm 2.35 mm 2.45 mm 2.55 mm 2.65 mm Phase [Degree] Phase [Degree] Length of Patch [mm] (a) Freq [GHz] (b) Figure 6.3: Phase Diagrams of a variable-sized patch as function of: (a) patch length, (b) frequency. A fundamental property of the variable-sized antenna element phasing technique is therefor: f L L res f res An increase in operational frequency f causes a decrease in resonance length L res. Vice-versa a decrease in the patch length L causes an increase in the resonance frequency f res. This property is a fundamental criterion for designing MRA. Because of the rapid change around resonance, most MRA elements have lengths within ± 5% of the nominal resonant length L res. Hence for a design at 34.5GHz a phase range of 290 is achieved if patch lengths between the mm are used. The neutral patch length L res is selected at 2.45mm, i.e. in the middle of the phase drop region. The variation of patch lengths of the array elements is almost ±20.0% of L res. The estimation of the bandwidth of the single element and the array can be determined from this figure. The progressive phase envelope for beam steering not far from broad side, which is obtainable from a progressive length in the range 2.1mm L 2.6 mm, is almost constant between GHz. This frequency range is ±4.35% around the operating frequency f res. The total phase range over this band equals 220. This bandwidth is verified in Figure 6.3(b) and is validated in subsection Figure 6.3(b) shows the reflected phase as function of frequency. The resonant length of the patch, L, is considered as parameter. This set of curves is very helpful in understanding the frequency behaviour of the antenna element operating in an array system. The shift in the resonance frequency for different rectangular patches is in accordance to equation 6.4. For example, the difference in resonance frequency of a patch with resonant length of L=2.25mm and a

62 50 Variable-sized phasing technique patch of L=2.65mm becomes 5.178GHz. Figure 6.3(b) supports this shift in frequency, where a difference of about 4.5GHz between the phase curves of L= 2.25mm and L= 2.65mm is presented. Figure 6.3(b) suggests that a phase range of variable-sized patch single element is around 360. Thisisnot observed in Figure 6.3(a). The reason for this higher phase range is caused by the relationship that increasing the frequency is electrically equivalent to decreasing the wavelength which corresponds to a decrease in patch length L. Increasing the frequency results into a larger complete system (antenna element plus truncated ground plane) relative to the operating wavelength, and not only the patch length increases. It means that increasing the frequency is electrically the same as increasing substrate thickness, patch length and substrate dimension at the desired frequency. Figure 6.4 gives the magnitude of the electric far-field, verifying the resonance shifts. Note that in the electric far-field, the factor e jk 0 r r is suppressed. r is the distance that tends to infinity and k 0 is propagation constant in air. The electric far-field is therefore in volts [51]. Figure 6.4(a) and 6.4(b) shows the magnitude of the far field as function of patch length L and frequency, with either the frequency or the patch length as parameters. The figures demonstrate the single-stage bandstop filter behaviour of the patch antenna. In general the patch behaves as a series-connected parallel resonant circuit. In this case the loss is maximum at resonance frequency and hence the magnitude of the electric field decreases at the resonance frequency GHz 34.5 GHz 36 GHz mm 2.5 mm 2.6 mm Electric Far field [mv] Electric Far field [mv] Length of Patch [mm] Freq [GHz] (a) Figure 6.4: Magnitude of electric field in the far zone for a variable-sized patch as function of (a) the resonance length for different frequency, (b) the frequency for different length. (b)

63 Phase diagram Computational aspects An important issue when using numerical solutions is the meshing. In this section the meshing for the variable-sized phasing is taken into consideration. Homogeneous meshing is applied since there is no discontinuities in the patch geometry. Figure 6.5 shows the results for the variable-sized patch. Phase [Degree] λ 0 /20 λ 0 /25 λ 0 /30 Phase [Degree] L=4.5 mm L=3.5 mm λ 0 /30 λ 0 /25 λ 0/ Length of Patch [mm] (a) Freq [GHz] (b) difference: λ 0 /20 and λ 0 /30 difference: λ 0 /25 and λ 0 /30 Phase difference [Degree] Length of Patch [mm] Figure 6.5: Meshing for a variable-sized patches: (a) length of the patch at f = 34.5GHz, (b) fixed length, (c) phase error between different meshing at f = 34.5GHz. Figure 6.5(a) shows the reflected phase as function of patch length L for different meshing. Examining Figure 6.5(a), it seems that different meshing has a minor effect on the phase diagram. However this is not the case. Figure 6.5(c) shows the absolute phase difference for different meshing. The figure illustrates how sensitive the system is for changes in meshing at resonance. The phase difference between the λ 0 /20 and λ 0 /30 curves for L = 2.5mm is 20. A phase error of such magnitude is not acceptable for the array design. Figure 6.5(b) shows the phase of the variable-sized patch as (c)

64 52 Variable-sized phasing technique function of frequency for two resonant lengths of the patch. By examining Figure 6.5(b) it can be concluded that the phase curves for reduced size sampling as function of frequency will shift slightly to a lower resonance. This shift is comparable for different rectangular patch sizes and especially the difference between λ 0 /20 and λ 0 /30 is considerable. This result indicates that the phase error might be large when studying the patch behaviour. The sampling for variable-sized patches that is applied in the remaining part of this chapter is λ 0 /30. It is shown that such a sampling offers the optimal combination of computational speed and accuracy. For variablesized systems the effect of meshing is negligible above λ 0 /30. Figure 6.6 depicts the sampled geometry of the patch antenna generated in FEKO. For more complex systems meshing is studied individually in order to ensure the correctness, accuracy and convergence of the phase diagrams. Figure 6.6: Meshing VZ geometry generated in FEKO Substrate thickness The microstrip elements are printed on a substrate which can be manufactured with different standard thicknesses. Figures 6.7(a) and 6.7(b) show the reflected phase as function of patch length L at a resonance frequency of f res = 34.5GHz and as function of frequency at a resonance length L = 2.45mm for different substrate thicknesses, respectively. The trade-off between phase range ψ range, phase sensitivity ψ sens, and substrate thickness is clearly visible. Because manufacturers can guarantee an accuracy of the metal patches within a few μm a very sensitive system to the patch length doesn t give a problem. A fundamental property that can be drawn from the results related to substrate thickness is: h = Ψ range & Ψ sens 6.4 Surface currents Excited surface currents determine the performance and behaviour of the microstrip elements. Study of the surface currents provides insight in oper-

65 Surface currents 53 Phase [Degree] h = mm h = mm h = mm f =34.5 GHz Phase [Degree] mm mm mm L = 2.45 mm Length of Patch [mm] (a) Freq [GHz] (b) Figure 6.7: The reflected phase for different substrate thicknesses as function of (a) length at f res = 34.5GHz, (b) frequency at resonance length L = 2.45mm. ation of any elementary radiator and in this case the variable-sized patches. They demonstrate the basics in surface current behaviour of a microstrip system and give an indication of the bandwidth of the antenna system. L = 2.25 mm L = 2.45 mm L = 2.65 mm Figure 6.8: The electric current excited on the patch for different resonant lengths at the resonance frequency of f = 34.5GHz. If the electrical length of the patch current matches the operational wavelength, the maximum currents are excited. Figure 6.8 shows the surface current distribution at resonance- and outside resonance length at the desired resonance frequency of f = 34.5GHz. The instantaneous surface currents are directed along the y-direction. The distribution shows a considerable increase in the magnitude of the surface currents at resonance. Figure 6.9 shows the surface current distribution at resonance- and outside resonance frequencies at a resonance length L = 2.45mm.

66 54 Variable-sized phasing technique f=33.5ghz f=34.5ghz f=35.5ghz Figure 6.9: The electric current excited on the patch for different frequencies at a resonance length L = 2.45mm Near-field The excited surface currents in the rectangular metal plate result into a single re-radiating antenna. The impinging energy is reradiated at the edges of the plate. L=2.25 mm L=2.45 mm L=2.65 mm Figure 6.10: Magnitude of the electric near-field for variable-sized patch at the resonance frequency of f = 34.5GHz. Figure 6.10 shows the magnitude of the electric field at a distance of 1.0mm from the antenna directly above the patch. The size of the near-field scan plane is λ 0 2 λ 0 2. Two radiating edges are visible. The radiating surface area is not confined to the patch area. The radiating area is much larger due to the fringing effect. This demonstrates the concept of effective lengths [27].

67 Array design Array design In order to demonstrate the concept of passive beam switching using such a phasing technique, it was decided to design a 1x5 element array to scan the main beam over elevation angles of θ scan =0,10 and 15 respectively. For the design the required progressive phase is calculated using the expression Θ= k 0 d y sinθ scan (6.5) where Θ, d y represents the required progressive phase shift and the centerto-center distance between the successive elements along the y- direction respectively. k 0 = 2π λ 0 is the free space wave number and θ scan represents the elevation angle where the main beam of the array is required to be oriented. The array element spacing is chosen to be d y =0.5λ 0 along the y-direction. From Equation (6.5) the progressive phase shift for θ scan =0,10 and 15 isθ=0, and respectively. The lengths corresponding to the required progressive phase shifts are then interpolated from the design curve shown in Figure 6.3(a). It was decided to switch the beam only in the elevation plane (φ =90,-90 θ 90 ); it means that the dimension of the patch in the vertical plane remains the same. The performance of an 1x5 array is then evaluated by using FEKO. Table 6.2 shows the value of the parameters used for the simulations. Table 6.2: Parameters for the array design using a variable-sized patch. Parameter Value Operating Frequency f = 34.5GHz Substrate Thickness h = 0.381mm Substrate Permittivity ɛ r =2.33 Loss Tangent tan δ = Patch Width, Length 3.3mm, Lmm d y λ 0 /2mm Meshing λ 0 /25 The antenna elements with the desired length are simulated in FEKO to verify and validate the phase variation curve. Due to beam switching in the E-plane, the lengths of the patch antennas are changed along the y-direction. Table 6.3 shows the corresponding length of the individual antenna element in the array. The array is illuminated with an incident horizontally polarized plane wave with unit amplitude. Figure 6.11 presents the E-plane pattern simulated by FEKO. The beams are oriented towards the correct elevation angle, hence validating the correctness of the phase variation design curve. It was observed that the reflection from the ground plane had much impact on the simulation results.

68 56 Variable-sized phasing technique Table 6.3: Resonant length for each antenna in the array. θ scan =10 Column number Length of the patch (in mm) θ scan =15 Column number Length of the patch (in mm) Electric far field [db] θ scan = 0 θ scan = 10 θ scan = Elevation angle [Degree] Figure 6.11: Theoretical E-plane radiation patterns of 1 5 using variablesized patch for θ scan =0,10 and Measurement results Based on the numerical results given in the previous section three 5 5 antenna arrays were built and their radiation patterns were measured. The measurements presented in this thesis were performed in the anechoic chamber of IRCTR (also known as DUCAT - Delft University Chamber for Antenna Test) using unique IRCTR MMW measurement facilities. DUCAT is a moderate sized (3m 3m 6m) anechoic chamber that allows the measurement of most of the antenna parameters (scattering parameters, radiation patterns, gain, etc.) in frequency domain and in time domain, as well. The shielding of the chamber is optimum for frequencies above 2.0GHz up to 18.0GHz and it is at least 120dB. All walls are covered with pyramidal shaped absorbers. It is found that the walls have reflection coefficients of less than -36.0dB. The maximum distance between the antenna under test and the standard gain antenna is around 3.5m. The measurements in frequency domain (the ones presented in this thesis) are performed using the

69 57 Array design (a) (b) (c) Figure 6.12: (a) Realized variable-sized MRA, from left to right array for θscan = 0, 10, 15, dx = dy = λ0 /2; (b) IRCTR 8510 XF MMW NWA used for pattern measurements; (c) manufactured array antenna mounted on the foam column of DUCAT. NWA HP-8510XF (up to 110.0GHz). Moreover, a PC (Personal Computer) is used to control the equipment and to execute automated measurements controlling the antenna positioners. A photo of the prototype antennas, the measurement equipment and measurement set-up is depicted in Figure The radiation properties of the manufactured antennas were investigated by means of measurement software and hardware in the anechoic chamber available at IRCTR. The array was located on a tiny cylinder with a height of 30.0cm mounted on top of the DUCAT low reflecting column. The miniature cylinder was made from foam material with relative dielectric permittivity near to r = 1.0, so that the measurement results are not affected by the set-up. To decrease the effect of blocking, a Standard Gain Horn (SGH), mounted at a distance of 45.0cm from the arrays, is used as an offset feed. The aligning was such that the array was illuminated within the 3.0dB beamwidth of the feed. The measurement probe was also a SGH

70 58 Variable-sized phasing technique (same as feed horn) and located in the far field of the array. Figure 6.13 shows the comparison between simulations and measurements for the three array antennas. E plane Radiation Pattern [db] Measured Simualted Cross polar E plane Radiation Pattern [db] Measured Simualted Cross polar Elevation angle [degree] Elevation angle [degree] (a) 0 (b) E plane Radiation Pattern [db] Measured Simualted Cross polar Elevation angle [degree] (c) Figure 6.13: Comparison between the theoretical and measurement results (co-and cross-polar) (a) θ scan =0,(b)θ scan =10,(c)θ scan =15. The results demonstrate that the concept is working and that MRA with variable-sized patches can be used to switch the beam. The size of the ground plane had much effect on the radiation pattern. The ground plane dominated the radiation pattern since only a few elements were used in the array. It was then necessary to tune the ground plane dimension. There was a difference in optimal ground plane dimensions between simulations and measurements. FEKO uses an ideal condition for the analysis of the antenna; to achieve such conditions at Ka-band frequencies is a difficult task. Although the offset angle of the feed is low, it still could affect the measurement results. The amplitude tapering of the feed was not considered in the simulation results. This has some effect on the measurement. In general it can be concluded that there is a good agreement between simulation and measurement results, proving the concept.

71 Conclusions Bandwidth It has been shown that the dominant factor controlling the reflectarray bandwidth is the bandwidth of the radiating element. The effect of non-constant path delays over the surface of the MRA has little significance unless the aperture is electrically very large. In fact the dominant factor that limits the reflectarray bandwidth is generally the element bandwidth. Figure 6.14 shows the measured radiation pattern for the array scanning at θ scan =15 at three frequencies. The shift in the maximum of the pattern can be neglected leading to a bandwidth more than 1.0GHz. The results for θ scan = 10 are not shown here, but a measured bandwidth of over 1.5GHz can be reported here. E plane Radiation Pattern [db] f = 34.0 GHz f = 34.5 GHz f = 35.0 GHz Elevation angle [degree] Figure 6.14: Radiation pattern measurement for θ scan =15 at three different frequencies. 6.6 Conclusions In this chapter the MRA concept based on variable-sized patch antennas is presented at MMW band frequencies. This technique forms the basic technology and a useful reference for other phasing techniques presented in the subsequent chapters. No conclusions regarding the loss budget and efficiency of the array antenna are given here because of limitations in the manufactured array geometry and in the measurement set-up. The results given in this chapter clearly demonstrate the concept. Results shown in this chapter illustrate potentials and capabilities of MRA at MMW band frequencies. A complete overview of this investigation is presented in the reference [9].

72 60 Variable-sized phasing technique

73 Chapter 7 Hollow patch: part 1 In a MRA using a hollow patch all elements have the same sizes. The electrical length of the elements is varied by the insertion of a slot in the patch having different dimensions. The surface currents are forced around the slot leading to a larger electrical current length compared to the patch without slot. In this apporach the scatter phase is controlled by the dimension of the slot. This chapter demonstrates the use of a hollow patch and explains its electrical behaviour. Main advantages of the hollow patch are firstly that the dimensions of the antenna elements stay the same and secondly that the sensitivity to the slot length is controllable by dimensioning the patch and slot. The geometry of a single rectangular element patch with square slot is depicted in Figure 7.1. The geometry consists of a metal patch with a slot of variable width W h and length L h on a grounded substrate. The excitation is a plane wave with linear polarization along the positive y-direction. The system parameters are shown in table 7.1. It is noted that the slot dimensions are variables of the system. 7.1 Design procedure The presence of the slot offers additional design parameters compared to the variable-sized case. The slot enables the designer to use the geometry of the patch to set the sensitivity and phase range. The sensitivity to the slot width is high so that a small variation in the dimension of the slot leads to a large shift in the electrical length of the current. The smaller the slot compared to the patch width the lower the sensitivity. The dimensioning of the patch and slot offer a powerful design tool. Increasing the slot size however also increases the cross-polar current components. Extensive study has shown that more holes or holes at different positions on the patch turned into worse results. A rectangular- and circular hollow patch yielded lower a phase range. We expected at least an improvement in the average level of cross-

74 62 Hollow patch: part 1 Incident plane wave Substrate Metal z E =â y E 0 L h h= 0.381mm W h L W GP size < y λ 0 2 =4.48mm x GP size < λ0 2 =4.48mm Figure 7.1: Geometry of a hollow rectangular patch. Parameter Table 7.1: Parameters of the hollow Patch. Value Operating Frequency f = 32.0GHz Substrate Thickness h = 0.381mm Substrate Permittivity ɛ r =2.33 Loss Tangent tan δ = Patch Width, Length 3.5, 2.5mm Slot Width, Length W h, L h Ground plane dimensions <λ 0 /2 λ 0 /2 Smart sampling λ 0 /25, λ 0 /60 for patch and slot respectively. polarization with a circular hole, but we didn t notice any improvements in our simulations. By introducing a slot on the patch, the resonant length can only be increased, since the path of the currents is lengthened. The slot makes path length of the current electrically longer and hence the resonant frequency will only be lowered. In the variable-sized scenario, the resonant length could be either increased or decreased.

75 Phase diagram Phase diagram To work around the resonance frequency, the hollow patch must consequently be designed so that it resonates at the operational frequency with an intermediate size of the slot. Therefore, a patch is designed with dimensions of 3.3mm 2.5mm that resonates at 35.0GHz without any slot and resonates at 32.0GHz 32.5GHz with a square slot of intermediate size within the range of 0.9mm W h = L h 1.1mm. Figure 7.2(a) depicts the reflected phase as function of slot dimensions for different frequencies GHz 32.0 GHz 32.5 GHz 33.0 GHz 33.5 GHz mm 0.9mm 1.0mm 1.1mm 1.2mm Phase [Degree] 50 Phase [Degree] Dimensions of the slot [mm] (a) Freq [GHz] (b) Figure 7.2: Phase diagrams of a hollow patch as function of (a) slot dimensions, (b) frequency. The phase curves corresponding to a small slot width compared to the patch length, have the same form as the phase curve in the variable-sized scenario. Figure 7.2(b) presents the reflected phase as function of frequency for different slot dimensions. The fundamental property in slotted systems is W h,l h f res i.e. an increase in slot dimensions W h and L h, results into a decrease of the resonance frequency f res. W h has a more dominant effect on the phase diagram than L h due to the polarization of the incoming wave and the geometry of the patch. The excited surface currents are forced around the slot. As stated before, an increase in slot size leads to an increase in the electrical path length of the current (the current distance is lengthened). The resonance frequency corresponds to the electrical length of the current of the hollow patch. If the resonance frequency is lower the electrical length is larger and vice-versa. In Appendix C, it is shown that the slot length L h parallel to the polarization of the incoming wave has a minor effect on the resonance frequency.

76 64 Hollow patch: part 1 From Figure 7.2 we learn that the maximum phase range at the operating frequency f is 300 and the maximum phase sensitivity is 25.5 /0.1mm. The phase range for a system using elements with slot dimensions varying between mm is 265. As demonstrated in Figure 7.2(b) the shift between the phase curve of a patch with a slot of dimensions of 0.8mm and 1.0mm is about 0.2GHz whereas the shift between a patch with a slot of 1.0mm and 1.2mm is about 0.3GHz. This is due to the non-linear relationship between the electrical length of the current of the patch system and its resonance frequency Computational aspects The patch element in the variable-sized phasing technique does not contain a non-homogeneous geometry which is small compared to the operating wavelength. For the variable-sized phasing technique non-homogeneous sampling does not offer any advantage. A slot introduces a discontinuity in the geometry of the patch which includes the substrate area in the slot and the metal surrounding the slot edges. This paragraph demonstrates that these areas must be sampled more dense in order to achieve accurate results Phase [Degree] λ 0 /20 λ /30 0 λ 0 /40 SS1 SS2 Phase [Degree] λ /20 0 λ /30 0 λ 0 /40 SS1 SS Dimensions of the slot [mm] (a) Freq [GHz] (b) Figure 7.3: Meshing effects for hollow patch depending on (a) dimensions characteristics at f =32.0GHz, (b) frequencycharacteristics; slotdimensions W h = L h = 1.0mm. Hence inhomogeneous meshing (so-called smart sampling) is applied for the hollow patch. Examining Figure 7.1, the slot edges are sampled with a sampling size of λ 0 /60 and the rest of geometry is sampled with λ 0 /25. This sampling scheme is denoted as SS1. The sampling scheme with λ 0 /60 and λ 0 /28 is denoted as SS2. Analysis of the meshing required for a hollow patch is shown in Figure 7.3. Figure 7.3(a) presents the phase as function of

77 Phase diagram 65 the slot dimensions for different meshing constants. The convergence in the phase curves for a denser meshing is apparent. For other phasing techniques this convergence is also an indication of minimal meshing needed for accurate results. Sampling of λ 0 /20, λ 0 /30, and λ 0 /40 curves give the results for homogeneous meshing. Figure 7.3(b) shows the frequency response of the phase diagram for different sampling. In this case the results are close to each other, since the whole geometry is electrically changed by changing the frequency. At resonance the antenna is very sensitive and requires a fine meshing for accurate results. Taking SS2, i.e. λ 0 /28:λ 0 /60 as the reference scheme for the sampling, the absolute phase differences between homogeneous sampling of λ 0 /20, λ 0 /30, λ 0 /40 and inhomogeneous sampling of λ 0 /25:λ 0 /60 have been calculated. Figure 7.4(a) and 7.4(b) shows the phase error between different meshing for a hollow patch Phase [Degree] SS2 : λ 0 /20 SS2 : λ 0 /30 SS2 : λ 0 /40 SS2 : SS1 Phase [Degree] SS2 : λ /20 0 SS2 : λ /30 0 SS2 : λ 0 /40 SS2 : SS Dimensions of the slot [mm] (a) Freq [GHz] (b) Figure 7.4: Error caused by different meshing for hollow patch depending on (a) dimensions characteristics at f = 32.0GHz, (b) frequency characteristics; slot dimensions W h = L h = 1.0mm. The phase error is reduced from 40 to 0 by the inhomogeneous meshing. The frequency behaviour for the homogeneous and inhomogeneous case shows that the error at resonance is highest. The phase error is highest in the phase drop region for different meshing schemes. The dimensions of the array elements are taken from this region. The results presented in this section demonstrate that accurate meshing for each phasing technique is crucial. The convergence in phase error has been studied resulting into a meshing configuration with an error of 0 for the complete range. Inhomogeneous meshing of λ 0 /25:λ 0 /60 provides an adequate phase accuracy with acceptable computational requirements. Figure 7.5 depicts the sampled geometry of patch antenna generated in FEKO.

78 66 Hollow patch: part 1 Figure 7.5: Meshed hollowed patch geometry generated in FEKO Substrate thickness Figure 7.6 shows the phase diagram of the hollow patch for different substrate thickness. The trade-off between sensitivity and phase range still holds. Figure 7.6(a) shows the reflected phase as function of slot dimensions for h = mm h = mm h = mm h=0.381 mm h = mm h = mm Phase [Degree] 50 0 Phase [Degree] Dimensions of the slot [mm] (a) Freq [GHz] (b) Figure 7.6: The reflected phase of a square hollow patch for analysis of substrate thickness as function of (a) slot dimensions at f = 32.0GHz, (b) frequency; W h = L h = 1.0mm. different substrate thicknesses at f = 32.0GHz. The phase range decreases for substrate thickness higher than 0.381mm. This is not so clearly visible in Figure 7.6(b). Figure 7.6(b) shows the reflected phase as function of frequency for a fixed slot size of 1.0mm and different substrate thicknesses. The phase curve is shifted to the lower frequency as the substrate thickness increases. The same phenomena occurred for a variable-sized system with fixed patch dimensions.

79 Phase diagram Patch length Figure 7.7 shows the reflected phase of a hollow patch for different patch lengths. The length of the patch offer us another powerful design aspect. Sensitivity can be adjusted by changing this parameter. 210 L = 2.3 mm L = 2.5 mm L = 2.7 mm L = 2.3 mm L = 2.5 mm L = 2.7 mm Phase [Degree] 50 Phase [Degree] Dimensions of slot [mm] (a) Freq [GHz] (b) Figure 7.7: The reflected phase of a hollow patch as function of (a) dimensions at f = 32.0GHz, (b) frequency; L h = W h = 1.0mm. In the variable-sized case the sensitivity follows directly from the substrate thickness and permittivity. For a variable-sized element system the sensitivity remains the same by varying the patch length. The resonance frequency is lower for the higher patch lengths. For a hollow patch the phase curves for different length L are shown in Figure 7.7(a). As can clearly be seen the slot width to patch length ratio determines the sensitivity. An increase in patch length is electrically similar to a decrease in the slot width to patch length ratio and hence leads to a decrease in sensitivity. Figure 7.7(b) shows the reflected phase as function of frequency for patches of different length L with a fixed slot size of L h = W h = 1.0mm. The shift in frequency is similar to that of the variable-sized case. The change in the sensitivity is negligible. The patch length is directly related to the electrical length of the current and hence the resonance frequency. Varying the length of the patch leads to a fundamental property for the hollow patch phasing technique. L = ψ sens f res meaning that an increase in patch length, L, results in a decrease of phase sensitivity ψ sens and resonance frequency f res. The resonance frequency is determined by the length of the patch and slot dimensions. The resonance can be shifted slightly by changing the slot dimensions. For the higher frequency shifts the patch length should be adjusted.

80 68 Hollow patch: part Surface currents Analysis of the surface currents clarify the fundamental properties of the hollow patch phasing technique. Figure 7.8 shows the current distribution at the resonance frequency for different dimensions of the slot. L h =0.4mm L h =0.8mm L h =1.0mm L h =1.3mm L h =1.6mm Figure 7.8: The electric current excited on the patch for different resonant lengths at f = 32.0GHz; L h = W h. The figures demonstrate the difference in electrical length of the surface currents which determines the maximum excited surface currents at the resonance frequency. The smaller the slot width, the lower the average current distance which results in a higher resonance frequency. Surface current distributions are also studied at the resonance slot dimension of L h = W h = 1.0mm for different frequencies as depicted in Figure 7.9. The current is maximum at the resonance frequency of f = 32.0GHz and slot dimensions at resonance. The amplitude and distribution of the current is nearly similar for f = 31.0GHz and 33.0GHz around the center frequency respectively, which give an indication of bandwidth. The amplitude and distribution of the current givs an indication concerning the bandwidth of the antenna. Similar results concerning the bandwidth have been reported in [55]. f = 30.0GHz f = 31.0GHz f = 32.0GHz f = 33.0GHz f = 34.0GHz Figure 7.9: The electric current excited on the patch for different frequencies and L h = W h = 1.0mm Near-field Figure 7.10 shows for different slot dimensions at f = 32.0GHz the magnitude of the electric near-field at a distance of 1.0mm.

81 Waveguide simulator measurements 69 L h = W h =0.9mm L h = W h =1.0mm L h = W h =1.1mm Figure 7.10: Electric near-field for a hollow patch at f = 32.0 GHz and different slots dimensions. f = 31.0GHz f = 32.0GHz f = 33.0GHz Figure 7.11: Electric near field for a hollow patch with L h = W h =1.0mm at different frequencies. The size of the near-field scan plane is λ 0 2 λ 0 2. The near-field is highest for L h = W h = 0.9mm and slightly lower for L h = W h = 1.0mm. Figure 7.11 shows for different frequencies the magnitude of the electric near-field at a distance of 1.0mm from the hollow patch with resonance dimension of the slot, L h = W h = 1.0mm. It is obvious and expected that the near-field is dominant at the resonance frequency of f = 32.0GHz. 7.4 Waveguide simulator measurements To measure the phase of an element, the element needs to be integrated in a large array antenna. This method is costly and time-consuming. This approach is therefor not attractive. Hence the waveguide simulator (WGS) measurement technique is proposed [53]-[54]. The antenna element is encapsulated between two air-filled waveguides. It is important that the polarization of the fundamental TE 10 mode propagating in the waveguide is in the same direction as the resonant length of the patch. The measurement setup is presented in Figure 7.12.

82 70 Hollow patch: part 1 Location of patch Extension waveguide (a) (b) Closed WG See Figure (b) Up-converter heads (c) Figure 7.12: Waveguide simulator setup, (a) patch and waveguide, (b) closed structure, (c) measurement set-up. The extension waveguide is added to create a closed waveguide structure. The HPXF (DC-110GHz) NWA (discussed in chapter 6, Subsection 6.5.1), was used for the frequency sweep measurement. The NWA provided frequency steps (801 points) from 25.0 to 40.0GHz. The hollow patch with slot dimensions Wh = Lh = 1.0mm is used for the measurements. The NWA measured the required S11 parameter. The waveguide dimensions were 7.0mm 3.3mm 218.0mm. It is worthwhile to report that the measurement set up was costly. To give an indication, the HF coaxial cable at MMW band to connect 2.4mm to 1.5mm was about 3000 Euros. The measurements were calibrated using the standard waveguide calibration set. The result of the frequency response measurement is presented in Figure 7.13 (solid line). The outcome of the numerical simulations are also provided in this figure. It is evident from the figure that there is a good agreement between simulations and measurement.

83 Conclusions WGS measurement WG FEKO simulation far field FEKO simulation Phase (deg) Freq (GHz) Figure 7.13: Comparison between the measurement and theoretical results. 7.5 Conclusions In this chapter a new phasing technique developed at IRCTR for a reflectarray is proposed. A theoretical study was first carried out, then measurements were performed on some prototypes which showed good agreement with simulations. The hollow patch phasing has the advantage of avoiding non-uniform spacing between the elements. Another advantage is an easier design, since the choice of the distance between adjacent elements is no longer influenced by the geometry of the single element in the array configuration. Elements of novelty A new concept of phasing technique for MRA is introduced. The analysis, design and performance of such technique is studied in detail. The concept was demonstrated theoretically and validated experimentally using a waveguide simulator. A complete overview of this investigation is presented in the reference [55].

84 72 Hollow patch: part 1

85 Chapter 8 Hollow patch: part 2 A main challenge in this PhD thesis is to analyse and design an active MRA at the MMW frequency band. For the passive phasing techniques discussed in the previous chapters (Chapters 6 and 7), in order to achieve the phasing, it is necessary to alter the patch geometry. However, this is not the case when active devices are integrated into the patch. In this case the active devices would give an extra degree of freedom which leads to the versatility of keeping the elementary radiator geometry fixed. Hence it is necessary to re-design the geometry of the antenna element. The original idea was to load the hollow patch with MEMS. In this approach the length of the slot can be altered adaptively leading to a different length of the electric current. The analysis and design of such a hollow patch was first considered and will be presented in this chapter. In our research on switches the use of a tunable capacitance was suggested. The properties of the diode-based varactors will be covered in the next chapter. The capacitance of such an active component can be changed adaptively by applying different D.C. voltages across the varactor. Moreover for the integration of the varactors a new design concept of the hollow patch was needed. This will be discussed in section 8.5. Due to the complexity of actual MMW antenna designs and in order to prevent complications in manufacturing and to have a certain degree of confidence in the design process, it was suggested by DIMES scientists to integrate varactors devices with MRA at a lower frequency, i.e., 6.0GHz. It has been demonstrated that as the frequency increases the Q value of varactors decreases with 1 f. In this case the varactor behaves more or less as a resistor instead of a capacitor. Technically it was possible to integrate a varactor chip at Ka-band; however we then had to face the risk that the design may not be successful. Hence to demonstrate the concept with a certain degree of confidence it was decided for this lower frequency. In the next two chapters the design, analysis and measurement of active MRA at

86 74 Hollow patch: part 2 GP size < λ 0 2 = 23.0mm L = 12.0mm W h = 10.3mm W = 15.0mm GP size < λ 0 2 = 23.0mm Side view Patch Substrate h = 1.575mm L h = 0.55mm Ground plane Figure 8.1: Geometry of the 6.0GHz hollowed patch. 6.0GHz will be presented. We therefore made the design of the hollow patch at 6.0GHz, employing the technique addressed in the last chapter. Hence, it is important first to analyse the hollow patch characteristics at such a frequency using the rectangular slot. As indicated in the last chapter all elements have the same dimensions. The electrical length of the current is varied by insertion of slots of different dimensions. The slots in the 6.0GHz case are rectangular and the approach slightly differs from the one in the previous chapter. However, the same phenomena occurred and surface currents are forced around the slot leading to a longer electrical length of the current compared to the patch without slot. 8.1 Geometry of the rectangular hollow patch The geometry of the single element hollow patch with the rectangular slot is depicted in Figure 8.1. The structure consists of a metal patch with a slot of width W h and length L h on a grounded substrate. The difference with the geometry given in the previous chapter is that in 6.0GHz case the length of the slot is kept constant and the phasing is achieved by changing the width of the slot. This makes the geometry simpler and easier for integrating the MEMS or varactor. The system parameters are presented in table 8.1. The slot width is the variable of the system and is oriented perpendicular to the polarization of the incoming wave.

87 Design procedure 75 Table 8.1: Parameters of the hollowed Patch at 6.0GHz. Parameter Value Operating Frequency f = 6.0GHz Substrate Thickness h = 1.575mm Substrate Permittivity ɛ r =2.33 Loss Tangent tan δ = Patch Width, Length 15.0, 12.0mm Slot Width, Length W h, 0.55mm Ground plane dimensions <λ 0 /2, λ 0 /2 Smart sampling SS λ 0 /35, λ 0 /100, λ 0 /120 patch, patch current, slot 8.2 Design procedure As was stated earlier the slot enables the designer to use the geometry of the patch to set the sensitivity and phase range. The sensitivity to the slot width is high so that a small variation in the dimension leads to a large shift in the electrical length of the current. The smaller the slot compared to the patch width the lower the sensitivity. It is shown in [56] for the variable-sized case that the phase drop region at 6.0GHz occurs when the electrical length of the current is varied between mm. A hollowed patch must therefore vary its electrical length close to this region. The same design concept as in the previous chapter has been used where the selected slot width and patch length was determined so that: the antenna resonates at 6.0GHz, and the resonant length corresponds to the middle of the phase drop region. The dimensions of the hollow patch in this chapter are in the same size order as the hollow patch considered in [57] and the varactor- loaded patch as discussed in the next chapter. 8.3 Phase Diagram The hollow patch element presented in this chapter is not optimised. Figure 8.2(a) presents the reflected phase as function of the slot width for different frequencies. The resonance width is 10.3mm. Figure 8.2(b) represent the phase diagram as function of the frequency for different values of the slot width.

88 76 Hollow patch: part W h = 9.30 mm W h = 10.3 mm W h = 11.3 mm Phase [Degree] GHz 6.0 GHz 6.2 GHz Phase [Degree] Slot width [mm] (a) Freq [GHz] (b) Figure 8.2: Phase diagrams of a rectangular hollow patch as function of (a) slot width at different frequency, (b) frequency for different width of the slot. As was demonstrated in section 7.2, the same conclusion can also be drawn for the rectangular hollow patch, meaning W h f res i.e. an increase in the slot width W h results in a decrease of the resonance frequency f res. The excited surface currents are forced to flow around the slot. The phase curve becomes more sensitive as the width of the slot approaches the width of the patch. The maximum phase range at the operating frequency f is 320 and the maximum phase sensitivity is around /mm. The phase range for a system using elements with slot widths varying between mm is around 290. Figure 8.2(b) shows also the non-linearity of the phase curve in frequency shifts for different values of the slot width. The shift between the phase curve of a patch with a slot of 11.3mm and 10.3mm is about 0.275GHz whereas the shift between a patch with a slot of 10.3mm and 9.3mm is about 0.25GHz. This is due to the non-linear relationship between the length of the electrical current and the resonance frequency of such a patch system Computational aspects As indicated in the last chapter the hollow patch introduces a discontinuity in the geometry. Smart sampling is applied for the simulation results. As it is demonstrated in Figures 8.8 and 8.9, the current excited on the patch has a complex distribution and therefore that part of the patch needs to be sampled accurately to achieve reliable results. Hence, number of simulations have been performed to study the convergence of the phase curve. The results have led to the following: the slot area is meshed with a sampling of λ 0 /120; part of the patch where current has a major contribution is meshed

89 Phase Diagram 77 with λ 0 /100; all other areas are meshed with λ 0 /35. This sampling scheme is designated with SS2. Figure 8.3(a) shows the phase as function of slot width for different meshing constants. The convergence of the phase curves for dense meshing is apparent. The λ 0 /20 and λ 0 /30 curves are results of homogeneous meshing and λ 0 /25:λ 0 /40 is the result of inhomogeneous sampling. The error is maximum at resonance length and frequency. Figure 8.3(b) shows the phase error between different meshing for a hollow patch λ 0 /20 λ 0 /25+λ 0 /40 λ 0 /35: λ 0 /120: λ 0 /70 λ 0 /35: λ 0 /120: λ 0 / λ /20 : SS2 0 λ /25; λ /40 : SS2 0 0 SS1 : SS2 Phase [Degree] Phase [Degree] Slot width [mm] (a) Slot width [mm] (b) Figure 8.3: Meshing (a) effect of meshing and slot width on phase (b) phase error; SS1 = λ 0 /35:λ 0 /70:λ 0 /120; SS2 = λ 0 / 35:λ 0 /100:λ 0 /120. The absolute phase difference is shown between the phase results for homogeneous sampling of λ 0 /25, λ 0 /25:λ 0 / 40, and SS1 and phase results for the SS2 meshing. In all cases the phase error is highest in the phase drop region. The convergence in phase error has been studied in detail. We learned that with the proposed inhomogeneous meshing convergence is reached at expenses of computational requirements. It means that inhomogeneous meshing of SS2 provides an adequate phase accuracy. Figure 8.4 depicts the sampled geometry of patch antenna generated in FEKO. Figure 8.4: Meshing hollow patch geometry generated in FEKO.

90 78 Hollow patch: part Substrate thickness Figure 8.5 shows the phase behaviour of the hollow patch for different substrate thicknesses. Figure 8.5(a) shows the reflected phase as function of slot width for different substrate thicknesses. The phase for a substrate with a low thickness becomes sensitive and it shifts the resonance frequency considerably. Figure 8.5(b) shows the reflected phase as function of frequency for a fixed slot width of 10.3mm and different substrate thicknesses. As in the square hollow patch, the phase curve is shifted to the lower frequency as the substrate thickness decreases. Phase [Degree] h =0.508 h =0.787 h =1.575 Phase [Degree] h =0.508 h =0.787 h = Slot width [mm] (a) Freq [GHz] (b) Figure 8.5: Phase diagram for the analysis of substrate thickness for a rectangular hollow patch as function of (a) slot dimensions at f =6.0GHz,(b) frequency; W h = 10.3mm, L h = 0.55mm Patch length Figure 8.6(a) shows the phase curve for different patch length L. As discussed earlier the slot width to patch ratio determines the sensitivity. An increase in the length of the patch is electrically similar to a decrease in this ratio and hence leads to a decrease in sensitivity. Figure 8.6(b) shows the reflected phase as function of frequency for different L, withw h = 10.3mm. As in the square hollow patch, the same conclusion can be drawn, i.e. L = Ψ sens & f res

91 Phase Diagram L = 11 mm L = 12 mm L = 13 mm Phase [Degree] Phase [Degree] L = 11 mm L = 12 mm L = 13 mm Slot width [mm] (a) Freq [GHz] (b) Figure 8.6: Path length analysis for the 6.0GHz hollow patch as function of (a) slot width (b) frequency Slot length Since the induced surface currents must flow around the slot, it is obvious that if the slot length increases the electric length that the current travel is increased leading to a decrease in resonance frequency Phase [Degree] L h = 0.30mm L h = 0.55mm L h = 0.80mm Phase [Degree] L h = 0.30 mm L h = 0.55 mm L h = 0.80 mm Slot width [mm] (a) Freq [GHz] (b) Figure 8.7: The phase diagram for the rectangular 6.0GHz hollow patch as function of (a) slot width (b) frequency. The slot length is not of great importance for the design. The thinner the slot the smoother is the phase curve, especially for high values of the slot width. Figure 8.7 shows the effect of slot length variation on the phase curve. The main purpose of the slot is to interrupt the currents in the direction of the desired polarization. The slot length is chosen to be 0.55mm.

92 80 Hollow patch: part Surface currents Figure 8.8 shows the current distribution for different slot widths at resonance frequency. W h =9.3mm W h = 10.3mm W h = 11.3mm Figure 8.8: Surface currents for a rectangular hollow patch at 6.0GHz for different slot widths. f =5.9GHz f =6.0GHz f =6.1GHz Figure 8.9: Surface currents on the patch for different frequencies at W h = 10.3mm. Since W h = 10.3mm is closer to the resonance frequency of 6.0GHz the magnitude of the current on the patch is higher than in the other cases. At the resonance frequency the excited surface currents are maximum. The smaller the slot width, the lower the average electric length of the current which leads to a higher resonance frequency Near-field The current distribution provides an accurate indication for the behaviour of the antenna system. The changes in the current distribution directly affect the radiation characteristics. Figure 8.10 shows the magnitude of the electric

93 Surface currents 81 field at 1.0mm distance of the antenna for different values of the slot width, W h. The main part of the radiation is originated from the edges. W h =9.3mm W h = 10.3mm W h = 11.3mm Figure 8.10: The magnitude of the electric field at 6.0GHz in the near-field zone for a hollow patch. f =5.8GHz f =6.0GHz f =6.2GHz Figure 8.11: The magnitude of the electric field in the near-field zone at different frequencies; W h = 10.3mm. Based on the transmission-line model, the microstrip antenna can be represented by two radiating slots along the length of the patch (each of width W and height of h). Instead of two radiating edges for the variablesized system four radiating edges are identified. The surface currents are interrupted by the slot and this leads to additional radiating edges (equivalence principle). The size of the radiating edges at the slot depend on the slot width. In the case of W h = 9.3mm this radiating edge is therefore smaller than for the W h = 11.3mm case. The patch with a slot width of W h = 10.3mm is closest to its resonance frequency at 6.0GHz and therefore radiates more than in the other configurations. Figure 8.11 shows the magnitude of the electric field at 1.0mm distance of the antenna for different frequencies. Comparing Figure 8.10 with Figure 8.11 learns that the frequency has a less relevant effect on the near-field. Changing the frequency affects the complete antenna geometry which is not the case when the width of the slot is altered.

94 82 Hollow patch: part Hollow patch with island The geometry of the hollow patch presented and analysed in the last section is suitable for integrating MEMS switches in an elementary radiator. As mentioned, our research focuses on employing an active tunable capacitor based on varactor diode technology. Due to the geometry of the varactor device, the hollow patch geometry needed to be re-designed. The analysis and design of such an elementary patch antenna is presented in this section. The geometry of the hollow patch with the rectangular slot has to be adjusted with an additional island as is depicted in Figure GP size < λ 0 2 = 23mm L = 12.4mm W h = 10.0mm 2.1 mm W = 15mm GP size < λ 0 2 = 23mm 2.1 mm Patch Side view Substrate h = mm L h = 0.55mm Ground plane Figure 8.12: Geometry of the 6.0GHz hollow patch with island. The island is introduced in the antenna geometry for including the varactor chip in the slot. To intgerate the active device it is required that the dimensions of the patch antenna to be fixed. Nevertheless, it is worthwhile to analyse the phase diagram of such a geometry. The system parameters are presented in table 8.2. The design is similar to the procedure presented in section Phase Diagram Figure 8.13(a) shows the reflected phase as function of the slot width for different frequencies. The sampling scheme is SS2 defined in subsection Due to the existence of the island the length and width of the slot of the patch had to be adjusted to L = 12.4mm and W h = 10.0mm to have the patch

95 Phase Diagram 83 Table 8.2: Parameters of the hollow patch with island at 6.0GHz. Parameter Value Operating Frequency f = 6.0GHz Substrate Thickness h = 1.575mm Substrate Permittivity ɛ r =2.33 Loss Tangent tan δ = Patch Width, Length 15, 12.4mm Slot Width, Length W h, 0.55mm Island dimensions W island = L island =2.1mm Substrate Width, Length <λ 0 /2, λ 0 /2 Smart sampling SS2 λ 0 /35, λ 0 /100, λ 0 /120 patch, patch current, slot resonant at f = 6.0GHz. It seems that the island makes the electric length of the current slightly longer leading to the shift in resonance frequency. Figure 8.14 depicts the sampled geometry of patch antenna generated in FEKO GHz 6.0 GHz 6.2 GHz W h = 9.0 mm W h = 10.0 mm W h = 11.0 mm Phase [Degree] Phase [Degree] Slot width[mm] (a) Freq [GHz] (b) Figure 8.13: Phase diagrams of a rectangular hollow patch with island as function of (a) slot width at different frequency, (b) frequency for different slot widths. Comparing the phase diagram of the hollow patch with island depicted in Figure 8.13(a) with Figure 8.2(a) reveals the same behaviour of the reflected phase. Figure 8.13(b) presents the phase diagram as function of frequency for different values of the slot width. Comparing Figure 8.13(b) with Figure 8.2(b), the phase diagram for the hollow patch with island is less sensitive. Moreover the phase diagram for W h = 11.0mm is shifted slightly to the higher frequency.

96 84 Hollow patch: part 2 Figure 8.14: Meshing hollow patch geometry with island generated in FEKO. As was demonstrated in Chapter 7 and Appendix C the length of the slot perpendicular to the incoming wave polarization dominates the reflected phase. Figure 8.15(a) depicts the phase diagram as function of frequency for W h = 10.0mm and different island sizes, hence confirming this conclusion. The length of the island in this region has a marginal effect on the phase response as demonstrated in Appendix C. However the island can have an effect on the resonance frequency. Figure 8.15(b) shows the phase response as function of frequency for different lengths of the patch and dimensions of the island. It can be concluded from Figure 8.15(b) that the island shortens the electric length of the current and hence leads to an increase in the resonance frequency. In order to make the patch resonant at the same frequency of f r = 6.0GHz, the length of the patch is then increased from L = 12.0mm to L = 12.4mm L island = W island = 1.7 mm L island = W island = 2.1 mm L island = W island = 2.5 mm L = 12.0 mm L = 12.4 mm L = 12.0 mm L = 12.4 mm Phase [Degree] Phase [Degree] W island =L island = 0.55mm Freq [GHz] (a) 0 W island =L island = 2.1mm Freq [GHz] (b) Figure 8.15: Phase diagrams of a rectangular hollow patch with island as function of frequency for (a) island dimensions, (b) length of the patch and island dimensions.

97 Surface currents Substrate thickness Figure 8.16 shows the phase behaviour of the hollowed patch with island for different substrate thicknesses. A similar behaviour as in the case of the rectangular hollow patch is observed. A substrate with thickness h = 3.175mm shifts the resonance frequency considerably and hence the phase diagram becomes useless h = mm h = mm h = mm f =6.0GHz h = mm h = mm h = mm Phase [Degree] Phase [Degree] Slot width[mm] (a) Freq [GHz] (b) Figure 8.16: Phase diagram for the analysis of substrate thickness for rectangular hollow patch with island as function of (a) slot dimensions at f = 6.0GHz, (b) frequency Patch length Figure 8.17(a) shows the phase curve for different patch length L. There is similarity (almost identical) with the phase curve of the rectangular hollow patch depicted in Figure 8.6(a). Figure 8.17(b) shows the effect of L on the phase diagram as function of frequency. The shift in frequency for increasing the length by 1.0mm is higher than in the rectangular hollow case shown in Figure 8.6(b). 8.7 Surface currents Figure 8.18 visualizes the current distribution for different slot widths at f = 6.0GHz. At the resonance frequency the excited surface currents are maximum. When comparing with the rectangular hollow patch, it seems that in this case the current components deflect earlier and is leading to a decrease in electric length. This deflection is caused by the presence of the island. Hence the resonance frequency is shifted slightly. Figure 8.19 shows

98 86 Hollow patch: part L = 11.4 mm L = 12.4 mm L = 13.4 mm f = 6.0 GHz L = 11.4 mm L = 12.4 mm L = 13.4 mm f = 6.0 GHz W h = 10.0mm Phase [Degree] Phase [Degree] Slot width[mm] (a) Freq [GHz] (b) Figure 8.17: Path length analysis for the 6.0GHz hollow patch with island as function of (a) slot width (b) frequency. the current distribution at different frequencies, where W h = 10.0mm. The slot width has more effect on the amplitude of the current than the frequency. W h =9.0mm W h = 10.0mm W h = 11.0mm Figure 8.18: Surface currents at 6.0GHz for a rectangular hollow patch with island and different slot widths. 8.8 Conclusion For integrating active devices into the patch antenna using the available technology we decided to shift the frequency from Ka-band to C-band. A new design concept of the patch at this frequency band has been analysed. The here given design of a hollow patch with rectangular slot can be used for integrating the MEMS switches. Since the switch changes the electrical length of the slot, the analysis has given insight into the behaviour of the reflected phase as function of different parameters by changing the slot width. In this way the electric length of the current is altered causing a

99 Conclusion 87 f =5.9GHz f =6.0GHz f =6.1GHz Figure 8.19: Electric currents on the patch for different frequencies and W h = 10.0mm. change in the reflected phase. A second design has been presented for integrating varactor devices. A square island is added at the center of the slot. The island decreases the electric length of the current leading to a higher resonance frequency. In both approaches a complete phase shift of almost 360 is achieved for 8.0mm W h 12.0mm. Novelty Two new design concepts for elementary patch antenna based on the hollow patch have been presented. The designs are suitable for integrating active devices (such as MEMS and varactors) into the elementary antennas in an array.

100 88 Hollow patch: part 2

101 Chapter 9 Reconfigurable active MRA with capacitive loading The electronically tunable microstrip reflectarray technology appears to be very interesting for beam scanning applications. It provides advantages over the active phased array realized with transmit/receive modules such as elimination of a beamforming network, which can be complex and lossy for a high-directivity antenna. The technology offers therefore the potential to be simpler and a lower cost alternative for an active phased array. In this chapter the relevant aspects related to the design of a reconfigurable active MRA using capacitive loading hollow patches are presented. Based on the phasing techniques addressed in the previous two chapters employing hollow patches, the concept of beam steering employing active loaded antenna is presented here. The initial approach was to imitate the electrical behaviour of the hollow patch with variable slot width by integrating active switches over the slot. The width of the slot is then adaptively reconfigured using the switches. The active switch is a set of MEMS switches, which can monolithically be integrated and packaged onto the same substrate. Previous works also has shown that it is possible to reconfigure the beam of a reflectarray antenna mechanically but also electrically using MEMS [58]-[60]. However, employing such an approach leads to a more complex system since for reconfiguring continuously the beam, the required number of MEMS would increase substantially. Moreover, the beam steering would be discrete and not continuous. In [58] MEMS are used to tune the return losses of the antenna for different frequencies. A MEMS-based structure using patch resonators of adjustable height has been proposed in [59] leading to a very complex configuration. Moreover, the radiation patterns show a rather poor performance. Furthermore, the availability and configurability of the switches were of primary concern for such a system and were leading to the basic question: How complex becomes the control part? During our research on switches, tunable capacitances as an alternative solution came in our mind. In this approach,

102 90 Reconfigurable active MRA with capacitive loading the behaviour of a system with tunable capacitance is not fully represented by a hollow patch system since the electrical length of the slot is not directly switchable as would be the case with the use of switches. Switches affect directly the electrical size of a slot whereas a capacitor introduces a second contribution to the resonance frequency. In [61], the authors suggest a technology which consists of a resonant microstrip patch, aperture-coupled to a transmission line and loaded with two diodes. The total impedance of the transmission line can be varied by changing the reverse bias voltage applied simultaneously on the two diodes, and hence creating a phase variation in the reflection coefficient. The suggested technology is different and more complex than the concept introduced here. The phasing technique introduced in this chapter make use of hollow patch elements which can vary the reflected phase by varying the capacitance of the varactor diode through different biasing. In a joint research program between IRCTR and DIMES 1 the design of the active MRA using diode-based varactors is addressed. The varactor chips have been designed and developed at DIMES and act as tunable capacitive devices. The fundamental theory and design of the varactor is addressed in [62, 63]. The analysis of the hollow patch has given the fundament for the development of our active MRA and was discussed in the two previous chapters. 9.1 Geometry of a loaded hollow patch The ensemble of an elementary radiator consisting of a hollow patch loaded with the varactor is depicted in Figures 9.1 and 9.2. The system parameters are presented in Table 9.1. The phasing technique achieved by using a loaded varactor diode is more complex than the phasing techniques addressed in previous chapters. The capacitance of the varactor is controlled by applying DC biasing signals. The central part of the patch consists of the varactor chip with an additional metalized plate. Two vias are used to transfer the Bias and GND control signals between the two sides of the PCB (see Figure 9.2). The connection between points A and B through the chip represents the RF path. The chip can be mounted on top of the metal plate in the center of the slot. The chip has five connection points on top which are wired to the patch laminate using bondwires. The capacitance value of the varactor is varied using a bias voltage (indicated in Figure 9.2) between -12 and 0 Volts. The chip requires a GND signal which is also presented in the figure. On the bottom, the patch consists of a ground plane with islands directly under the center of the chip. The bias signal is transferred using a metallized via from top to bottom of the patch as can also be seen in Figure 9.2(b). To ensure decoupling of the control signal from the RF signal, a Surface Mounted Device (SMD) capacitor is integrated on the bottom of the patch. 1 Delft Institute for Micro-Electronics and Submicrontechnology

103 Geometry of a loaded hollow patch 91 GP size < λ 0 2 = 23mm L = 11.8mm ℵ= 0.3mm W h = 13mm chip W = 15mm GP size < λ 0 2 = 23mm Patch Varactor Substrate h = 1.575mm L h =0.4mm Ground plane Side view Figure 9.1: Geometry of the varactor-loaded hollow patch. Table 9.1: Parameters of the varactor-loaded hollow patch. Parameter Value Operating Frequency f = 6.0GHz Substrate Thickness h = 1.575mm Substrate Permittivity ɛ r =2.45 Loss Tangent tanδ = Patch Width, Length 15mm, 11.8mm Slot Width, Length 13, 0.4mm Distance chip to edge of slot ℵ =0.3mm Ground plane dimensions λ 0 /2, λ 0 /2 Chip Thickness 0.5mm Chip Width 0.65mm Chip Length 0.9mm Resonance capacitance C 0 = 0.55pF 0.6pF Smart sampling λ 0 /170, λ 0 /42, λ 0 /170, λ 0 /33 for chip, additional chip structure, slot edges and patch respectively.

104 92 Reconfigurable active MRA with capacitive loading Bond wire Metal Substrate Decoupling SMD capacitor A Varactor Chip B SMD capacitor Metal via Bias via Bias via GND via GND (a) (b) Figure 9.2: Geometry of the varactor-loaded hollow Patch, (a) front side, (b) back side. 9.2 Design procedure The design of the elementary radiator in Figures 9.1 and 9.2 can be seen as a search for the optimum parameters of the geometry that can provide a suitable phase diagram over a specific frequency range. The design procedure is very similar to that of the SMD MRA considering the available range of C values [57]. The operational frequency is given by the dimensions of the patch and the capacitance of the varactor device. The first step in the design procedure is the determination of the necessary capacitance range for different frequencies around C 0. C 0 is the capacitor value of the varactor for which the hollow patch is at resonance. This C 0 can be found using our numerical approach in order to give resonance at the desired frequency. The numerical results are based on the hollow patch model with an ideal capacitance bridging the slot, i.e. by connecting one edge of the slot to the other edge. The C 0 capacitance required for different operational frequencies are presented in Table 9.2. The varactor introduces constraints in the geometry. The dimensions of the chip as indicated in Table 9.1 requires a metal island for mounting; this island is extended by 0.3mm around the chip. Furthermore, in order to isolate the chip from the microstrip laminate a space between the chip and the edges of the slot has been created. The length of this space is ℵ. The effect of ℵ will be discussed in Subsection

105 Phase diagram 93 Table 9.2: Operational frequencies and their corresponding C 0. Parameter Value Operational Frequency Capacitance Value C 0 2GHz 1.92pF 6GHz 0.6pF 12GHz 0.32pF 32GHz 0.10pF 9.3 Phase diagram The phase diagram for the hollow patch loaded with the varactor is different from diagrams obtained with other phasing techniques addressed previously. In order to determine the performance of the phase diagram we select one variable while all other parameters in Table 9.1 remain fixed. The capacitance of the isolated element is controllable and creates so the possibility that each element can obtain a specific value in the phase diagram. The capacitance is directly proportional to the corresponding bias voltage which subsequently control the reflected phase of each elementary cell in the array architecture Phase [Degree] f = 5.8GHz f = 6.0GHz f = 6.2GHz L = 11.8mm Phase [Degree] C = 0.4pF C = 0.6pF C = 0.8pF L = 11.8mm Capacitance [pf] (a) Freq [GHz] (b) Figure 9.3: Phase diagram of a hollow patch loaded with tunable varactor as function of (a) capacitance, (b) frequency. Figure 9.3(a) illustrates the reflected phase as function of the capacitance of the wire segment across the hollow patch for different frequencies. The phase range for a varying capacitance 0.3pF C 0.8pF is about 250 with a maximum sensitivity of 40 per 0.05pF (40 /0.05pF). A bias voltage from -12 to 0V leads to a controllability of 0.05pF/Volt for this range.

106 94 Reconfigurable active MRA with capacitive loading The phase sensitivity matches the controllability resulting in a maximum sensitivity to the control signal of about 40 /Volt. The capacitance of the varactor can be adjusted from 0.2 C 1.1pF, hence the slope of the phase diagram is exactly in the desired range. The performance of the phase curves at frequencies f =5.8GHzandf = 6.2GHz is deteriorated which demonstrate a narrow band behaviour of the antenna. The phase response for f = 6.2GHz is too sensitive for designing the array, and the phase response for f = 5.8GHz is out of the desired capacitance region and hence not usable. Moreover, due to the additional capacitance added by the varactor device to the lumped element network of the patch, the patch will be more sensitive in the lower frequency range. Figure 9.3(b) presents the frequency response of the phase diagram for different values of the capacitance across the slot. The resonance frequency is shifted to higher frequencies as C decreases. This leads to the fundamental property that relates the change in capacitance to the shift in resonance frequency according: C = f res An increase in the capacitor value C causes a decrease in the resonance frequency f res. The phase curves indicate a small frequency band in which the resonance is shifted. The maximum shift in frequency for a 0.4 C 0.8pF varactor capacitor range is 200MHz Computational aspects The numerical analysis of the varactor-loaded elementary radiator requires smart sampling. In view of an automatic design procedure, the details of the hollow patch geometry including the chip are replicated in the numerical model. Figure 9.4 depicts the sampled geometry of varactor loaded hollow patch antenna generated in FEKO. Figure 9.4: Meshed varactor loaded hollow patch geometry generated in FEKO. Four different sampling constants for different parts of the antenna element system are used: chip part is meshed with λ 0 /170, additional structures for the chip are meshed with λ 0 /42, slot edges are meshed with λ 0 /170, remaining edges with λ 0 /33. The structure of the chip is considered to be

107 Phase diagram 95 a dielectric box with a dielectric permittivity of ɛ r = 4, covered by a metal surface. The additional structure consists of one metal island (required for mounting the chip) and another for the bondwire connections. The capacitor value is implemented in FEKO by a wire segment just above the chip. This wire segment extends over the chip located at the center of the slot and connected from one side of the edge slot to the other Substrate thickness Figure 9.5(a) presents the reflected phase as function of the capacitance for different standard substrate thicknesses. The fundamental property of the Phase [Degree] h = 0.787mm h = 1.575mm h = 3.175mm f = 6.0GHz Phase [Degree] h = 0.787mm h = 1.575mm h = 3.175mm C=0.6pF Capacitance [pf] (a) Freq [GHz] (b) Figure 9.5: Phase diagram for the analysis for a varactor-loaded rectangular hollow patch with various substrate thicknesses as function of (a) capacitance at f = 6.0GHz, (b) frequency with C 0 =0.6pF. substrate thickness identified in chapter 3 is applicable in these figures. An increase in substrate thickness decreases the sensitivity and phase range. The phase curve for h = 1.575mm demonstrates an optimal combination of phase range and sensitivity. Figure 9.5(b) depicts the frequency behaviour of the reflected phase for different substrate thickness and a varactor of 0.6pF. It is worth noting that the frequency response for different substrate thicknesses resembles that of the hollow patch phasing technique Patch length As indicated previously the patch length is an important design parameter affecting the antenna resonance frequency. Hence adjustment in the patch length would shift the phase curves. The patch length is used to position the phase curve in the desired region. Figure 9.6(a) depicts the reflected phase as function of the capacitance for different lengths L of the patch

108 96 Reconfigurable active MRA with capacitive loading at f = 6.0GHz. Depending on the specifications of the varactor, i.e. the C-V characteristic, the capacitance range determines a maximum possible phase range. The varactor of 0.3pF C 0.9pF with L = 11.8mm gives the optimal phase curve due to the smoothness of the phase diagram. The phase behaviour for a varactor with capacitance range 0.3pF C 0.8pF and L = 12.0mm has a better phase range, however it becomes more sensitive. Figure 9.6(b) gives the reflected phase as function of frequency for different patch lengths L and C 0 =0.6pF Phase [Degree] L = 11.6mm L = 11.8mm L = 12.0mm f = 6.0GHz Phase [Degree] L = 11.6mm L = 11.8mm L = 12.0mm C = 0.6pF Capacitance [pf] (a) Freq [GHz] (b) Figure 9.6: Effect of the patch length on the phase diagram for a varactorloaded patch as function of (a) capacitance at f = 6.0GHz (b) frequency with C 0 =0.6pF Slot length We should keep in mind that the parameters determining the dimensions of the slot need to be fixed for integrating the varactor and they are therefore less suitable to be used for tuning. However, the gap length and the spacing between the island and the edge of the slot have been adjusted in the design process. In the analysis of the reflected phase, the slot length can be divided into variation of: slot length, L h, in the region outside the chip, distance ℵ between the chip and the edge of the slot. The parameters which define the length of the slot are illustrated in Figure 9.1. The reflected phase as function of capacitance for different values of slot length L h is plotted in Figure 9.7(a) at f =6.0GHz. AsL h increases, the phase curve becomes more sensitive which indicates a narrow band behaviour of the antennas as shown in [57]. Figure 9.7(b) presents the frequency response of the reflected phase for different slot length L h and fixed capacitance value of 0.6pF. An increase in slot length results in a decrease of the resonance frequency and vice-versa. Only a small frequency shift is introduced by a slot length variation between mm.

109 Phase diagram Phase [Degree] L h = 0.25mm L = 0.40mm h L = 0.55mm h f = 6.0GHz Phase [Degree] L = 0.25mm h L h = 0.40mm L = 0.55mm h C = 0.6pF Capacitance [pf] Freq [GHz] (a) (b) Figure 9.7: Effect of the slot length on the phase diagram for a varactorloaded patch as function of (a) capacitance at f = 6.0GHz, (b) frequency with C 0 =0.6pF. The chip structure is separated by a distance ℵ=0.3mm from the slot for a proper isolation and assembly. Figure 9.8(a) depicts the reflected phase as function of capacitance for different values of ℵ at f =6.0GHz. Figure 9.8(b) shows the frequency response of the reflected phase for different values of ℵ with a fixed capacitance value of 0.6pF. This indicates that changing ℵ causes an almost negligible shift in the resonance frequency and vice-versa. It can be concluded that ℵ has a small effect on the phase curve Phase [Degree] ℵ = 0.2mm ℵ = 0.3mm ℵ = 0.4mm f = 6.0GHz Phase [Degree] ℵ = 0.2mm ℵ = 0.3mm ℵ = 0.4mm C = 0.6pF Capacitance [pf] (a) Freq [GHz] (b) Figure 9.8: Effect of ℵ on the phase diagram as function of (a) capacitance at f = 6.0GHz, (b) frequency with C 0 =0.6pF.

110 98 Reconfigurable active MRA with capacitive loading Slot width The main difference in the surface current distribution using an SMD Capacitor [57] and a varactor is the presence of the bondwire. For the SMD case the surface currents travel through the SMD over its complete width while for a varactor the surface currents passes through the bondwire. Furthermore, an SMD can be considered as a bridge with a certain width where currents can travel at any point of the bridge, where the varactor is more of a narrow passage where the currents can only travel through the very narrow wire segment. Hence, for the varactor-loaded hollow patch, this leads to an elementary radiator very sensitive to parameters such as slot width. Figure 9.9(a) presents the reflected phase as function of the capacitance for different values of the slot width at f =6.0GHz Phase [Degree] W = 12.0mm h W h = 13.0mm W h = 14.0mm f = 6.0GHz Phase [Degree] W = 12.0mm h W h = 13.0mm W h = 14.0mm C = 0.6pF Capacitance [pf] (a) Freq [GHz] (b) Figure 9.9: Effect of the slot width on the phase diagram for a varactorloaded patch as function of (a) capacitance at f = 6.0GHz, (b) frequency with C 0 =0.6pF. A dominating effect of the slot width on the phase diagram is visible. Figure 9.9(b) depicts the frequency response of the reflected phase for different values of the slot width and a fixed capacitance of 0.6pF. A decrease in slot width leads to an increase in the resonance frequency due to the smaller path length currents need to travel. 9.4 Surface currents Figure 9.10 presents the excited surface currents for different varactor capacitor values demonstrating that at C = 0.5pF 0.6pF the antenna is closest to resonance. It is worth while to note that for C = 0.4pF the magnitude of the currents around the slot is less than for C = 0.6pF. The resonance point attributed by the varactor is more dominant at lower capacitance values

111 Surface currents 99 C =0.4pF C =0.5pF C =0.6pF C =0.7pF C =0.8pF Figure 9.10: Surface currents for different varactor capacitor values at f = 6.0GHz. f =5.8GHz f =6.0GHz f =6.2GHz Figure 9.11: Surface currents for different frequencies and C 0 =0.6pF. (hence for C = 0.4pF the frequency shift is larger) than for C = 0.8pF. This conclusion is confirmed by the phase diagram (see Figure 9.3(b)). Figure 9.11 shows the surface currents for different frequencies at C = 0.6pF. The reason for the non-symmetry behaviour of the surface currents is discussed in the next subsection Near-field Figure 9.12 shows the magnitude of the electric field in the near zone of the patch at a distance of 1.0mm for different values of the varactor capacitance. At resonance the electric component of the near-field is at maximum. The effect of the asymmetric chip slot is visible in the radiation pattern for C = 0.6pF. As indicated in Subsection 6.4.1, the radiating edge which can be demonstrated using the cavity theory [27] is also clearly visible here. The electric field is most dominant at resonance and around a capacitance of C = 0.6pF 0.7pF. The non-symmetry behaviour of the near-field and current distribution may be caused by

112 100 Reconfigurable active MRA with capacitive loading C =0.4pF C =0.5pF C =0.6pF C =0.7pF C =0.8pF Figure 9.12: Magnitude of the electric field in the near zone for different varactor capacitor values at f = 6.0GHz. the inductive character of the patch; this gives a frequency dependency which may be non-linear the number of capacitance values; this is not large enough to see a symmetrical behaviour the physical connection of the capacitor to the patch; this might present strong parasitic effect. 9.5 Varactors The tunable varactors are designed and developed at DIMES. It uses diodebased circuit topologies, which can act as high-q distortion-free tunable capacitive elements. These diodes are implemented in a novel ultra low-loss Silicon on Glass (SoG) technology and result in measured Qs of over 200 at 2.0GHz. The diode configuration in Figure 9.13 can be applied to realize a voltage-controlled variable capacitor [62]-[64]. Bondwire to antenna Bondwire to antenna R R R R=100kΩ GND BIAS GND Figure 9.13: Topology of varactor diode. Due to its nonlinear behaviour of a varactor using single diode which leads to higher order distortion responses, a diode configuration as in Figure 9.13 has been employed. Theoretically this configuration is free from distortion. It has been demonstrated that for this topology the measured third

113 Varactors 101 order distortion, IM3, shows an improvement compared to the traditional single varactor tuning which is more than 30dB [62]. Another example of a more-traditional approach is the MEMS capacitor, which in its most popular implementation is able to switch between two fixed capacitance values [65, 66]. MEMS capacitors provide a high Q at moderate capacitance values but they require non-standard processing, expensive packaging techniques, high control voltages, and their reliability and switching speed are still poor compared to semiconductor-based solutions [62]. Other tuning techniques are based on voltage-variable dielectrics but exhibit similar drawbacks of manufacturability and performance [67]. Hence due to the drawbacks of MEMS capacitors, this new topology based on a voltage control varactor diode has been worked out. Figure 9.14 represents the detailed layout of the varactor chip processed at DIMES. These Low Distortion Varactor Stack (LDVS) components are, due to their easy implementation and inherently high performance, suitable for use in a variety of high-q tunable circuits, including filters, switches, phase shifters and matching networks [64]. Diode Metal Location for bondwire: chip to antenna Location for bondwire: chip to antenna The chulk seen through the substrate glass. Si R μm GND BIAS GND 891.5μm Figure 9.14: The layout of the varactor device. Figure 9.15 depicts the measured frequency response and C-V characteristicsoftwotypesofvaractordeviceswhicharebasedontwodifferent junction technologies. Device A has a low Q factor with a low leakage current, while device B has a high Q factor with a high leakage current, which means that in order to achieve the same capacitance value for both devices, device B needs a higher tuning voltage. Both varactors are part of a research program performed in DIMES, on different diode junctions, doping profiles, and processing technology.

114 102 Reconfigurable active MRA with capacitive loading Dice A f = 6 GHz 1.2 Dice A Capacitance [PF] Capacitance [PF] Increasing Voltage Voltage [V] (a) Freq [GHz] (b) Capacitance [PF] Dice B f = 6 GHz Capacitance [PF] Increasing Voltage Dice B Voltage [V] (c) Freq [GHz] Figure 9.15: Measured characteristics of two types of varactor devices for integrating into the antennas. Top figures give for device A the Capacitance as function (a) voltage at 6.0GHz, (b) frequency and different voltages. Bottom figures give similar curves but for device B. (d) 9.6 Varactor-based scanning capabilities This section indicates the beam scanning capabilities of the capacitanceloaded hollow patch in an array architecture. The main objective is to use the controlled capacitive element for beam scanning purposes. In order to analyse the concept an 1 6 uniform linear array is implemented and simulatedinfekotoscanthebeaminthee-plane. Tables 9.3 and 9.4 present the array parameters. In Table 9.4, A to F indicate the capacitance values of the varactor chip loaded on the elements in the array architecture. The phase diagram that is employed in the design of the array is depicted in Figure 9.3(a). Due to the complexity of the design, it was decided to scan the beam in one plane only. For a controllable varactor with 0.37pF

115 Technological aspects and experimental results 103 C 0.98pF the beam can be scanned from -15 θ scan 15. The controllability however must be with accuracies in the order of 0.02pF. Figure 9.16 presents the simulation results for the different array configurations and hence demonstrating the potential of our concept. The scanning range can be increased using 1 5 linear array, i.e. for 0.2pF C 1.0pF the beam can be scanned from -22 to 22. Table 9.3: Array parameters. Frequency 6.0GHz Element spacing d y = 23.0mm Linear array 1 6 Elements Patch dimensions 15.0mm 11.8mm Slot 13.0mm 0.4mm Varactor range 0.37pF C 0.98pF Substrate thickness h = 1.575mm Dielectric premittivity ɛ r = 2.45 Tan δ 5e-04 Table 9.4: Varactors capacitance for an 1 6 linear array, scanning -15 θ 15. Array Capacitance values per element Scan angle A: ( )pF +15 B: ( )pF +10 C: ( )pF +5 D: ( )pF -5 E: ( )pF -10 F: ( )pF Technological aspects and experimental results Since the complexity of the varactor-based design is high, replicating the antenna model used in the numerical analysis into the physical structure is not always fully possible and very much depending on the available technology. In the case of the integrated active MRA antenna the most complicated aspects are:

116 104 Reconfigurable active MRA with capacitive loading E plane radiation pattern [db] θ scan =5 θ scan =10 θ scan = Elevation angle [Degree] Figure 9.16: Simulated results for an 1 6 linear array demonstrating the beam scanning capabilities of an active MRA antenna. integration of the varactors into the elementary radiator as part of the array architecture; wiring of the varactor chip for steering the biasing signals; measurement set-up for evaluating the radiation pattern. The technological process associated to the hollow MRA laminates is well established and manufacturing accuracy is very high. Moreover, for isolated elements it was necessary to metalize the via for bond wiring purposes. This is a well developed technique in microstrip technology and has been adopted in this work. Studying in detail the antenna topology under investigation, there are a few aspects to be considered carefully in the manufacturing process: separating the RF signal of the antenna from the DC control signals of the varactor chip. Hence the miniaturized SMD capacitor needs to be integrated carefully on metallic islands on the back side of each elementary patch in the array; bond wiring of the chip for RF signaling, biasing and grounding; methodology and low cost technology for providing the chip with the correct voltage during the pattern measurements; effect of inductive coupling of the bondwire and the metallic ground plane on the resonance frequency of the isolated element and the element in the array. In the manufacturing process it is worth noting that it is not possible to contact the bondwire to Copper. Hence for bondwiring, vias with a diameter

117 Technological aspects and experimental results All dimensions in mm Metal Substrate 1.25 Chip (a) (b) Figure 9.17: Dimensions of the slot for mounting the chip. (a) top view, (b) bottom view. of 0.3mm have been drilled in the PCB laminate and were made of golden material using bondable gold. The bondwires with diameter of 17μm are integrated using the Wedge Wedge process technique [68]. The bondwire has a Gaussian bell form with a height of 300μm. The top and bottom part of the slot where the chip is mounted is depicted in Figure The dimensions are in millimeter. The varactor chip with a thickness of 500μm is assembled (glued) on top of the metalized island as presented in the Figure 9.17(a). The glue had a thickness of 60μm. The glue is not conductive and is carefully injected and spread manually over the metallic island. The chip is subsequently mounted on the glue. The complete antenna is then kept in the oven for over two hours with a maximum temperature of 100 C. Table 9.5 shows the type of varactor devices integrated in the 6 6 antenna array. Originally the complete set of varactor devices which were to be integrated in the array antenna, was of type B (see Figure 9.15). Unfortunately after testing, a number of varactors were damaged and it was required to be replaced them by type A. The capacitance performance of the chips has been measured for f = 6.0GHz at three different tuning voltages V tunning = -8, -3, 0 V. The measured capacitances per column as function of the tuning voltage are presented in Figure Examining Figure 9.18 reveals differences in capacitance of device A and B due to different junc-

118 106 Reconfigurable active MRA with capacitive loading Table 9.5: The varactor type integrated into the antenna. B B B B B B B B B A A A B B B A A B B B B B B B B B B A A A B B B A A B tions employed during the processing. Moreover, due to the accuracy and repeatability of the processing technology there was difference in the capacitance within each type. It is worth noting that the measurement set-up in this case was slightly different from the measurement set-up leading to the results presented in section 9.5. In the first case a miniaturized probe was located exactly on top of the chip allowing for measuring return loss S 11. In this way no additional cable which can introduce inductive effects was included. In order to demonstrate the reconfigurability concept and to validate the simulated results, a 6 6 planar array was manufactured. The hollow patches were loaded with varactor devices where each capacitance is adjusted to vary the reflected phase. A photograph is depicted in Figure The antenna was built in printed technology similar to the one presented in Subsection using high frequency laminated material TLX C1/C1 from Taconic with a thickness of 1.57mm and ɛ r = The primary radiator used in the array configuration is identical with the element discussed in section 9.1. The principal dimensions of this elementary antenna are given in Table 9.3. Figure 9.20 represents the integrated varactor device in the elementary radiator. Though the varactor device is difficult to be recognized, the bondwires are clearly depicted in the figure. In order to block the leakage of the RF signal to the DC part, an SMD ceramic multilayer capacitor chip is integrated (soldered) on the bottom side of each elementary radiator, see Figure 9.2. The SMD capacitor is Philips series NP050 and its value is such that around the operational frequency it behaves as a short circuit and hence blocking the leakage of the RF signal entering the DC circuit. As depicted in Figure 9.19(a), the control signals include 36 wires for the biasing (yellow wires)- and 6 lines for the grounding (blue wires) which have been soldered on the bottom side of each radiator in the array. Each column of the array requires one line for grounding. Experimental results The experimental results presented in this chapter were performed in DUCAT. The measurements in frequency domain are performed using the following equipment: a HP 81 PNWA (Personal Network Analyser) operating from 10MHz to 50GHz), PCI-766 Digital-to-Analog converter (DAC) with an ad-

119 Technological aspects and experimental results Capacitance [PF] P(1,1) P(2,1) P(3,1) P(4,1) P(5,1) P(6,1) Capacitance [PF] P(1,2) P(2,2) P(3,2) P(4,2) P(5,2) P(6,2) Tuning voltage [V] Tuning voltage [V] Capacitance [PF] P(1,3) P(2,3) P(3,3) P(4,3) P(5,3) P(6,3) Capacitance [PF] P(1,4) P(2,4) P(3,4) P(4,4) P(5,4) P(6,4) Tuning voltage [V] Tuning voltage [V] Capacitance [PF] P(1,5) P(2,5) P(3,5) P(4,5) P(5,5) P(6,5) Capacitance [PF] P(1,6) P(2,6) P(3,6) P(4,6) P(5,6) P(6,6) Tuning voltage [V] Tuning voltage [V] Figure 9.18: Measured characteristics of the varactor chips integrated in the array antenna. vanced multi-channel analogue output board with optimised 16 channels each with full 16-bit resolution. The DAC was already available in the laboratory and in order to keep the costs low it was decided to use this one and not purchasing an extra DAC with more channels. The drawback is that in this case, a DAC channel is used to provide the information on the number of each individual antenna element simultaneously.

120 108 Reconfigurable active MRA with capacitive loading (a) Figure 9.19: Manufactured active antennas. (a) rear view; (b) front view. (b) Figure 9.20: Assembled hollowed patch. The feed antenna was WR159 opertioanl in 4.9GHz f 7.05GHz frequency-band. A picture of the DUCAT anechoic chamber, focusing on the antenna under test mounted on the column, is presented in Figures 9.21(a) and 9.21(b). Figure 9.21(c), depicts the measurement equipment PNA NWA: E8364B 10MHz to 50GHz and the control system. Originally 36 analogue signals were required for biasing the complete antenna array. To simplify and reduce costs of the measurement procedure, the architecture of the antenna was slightly changed with respect to the model used in the numerical analysis. Despite of not being able to control each individual element(imposedbythehighcostofpurchasinganextradacwithan32 pins output) it has been decided that demonstrating the concept of the beam scanning should remain unchanged.

121 109 Technological aspects and experimental results (a) (b) (c) Figure 9.21: Manufactured active antennas mounted on the column of DUCAT and PNA NWA. (a) front view; (b) rear view; (c) PNA NWA. The frequency and voltage responses of each varactor device were different from each other due to the technological process used at DIMES to manufacture the chips. Moreover, as was indicated in Table 9.5, 10 elements in the array were integrated with A type varactor chips. Consequently the optimisation of the measured radiation patterns was a difficult task since we were forced to simultaneously address a number of the radiators in the array with the same biasing voltage. However, semi-optimisation has been performed by maximizing the gain, minimizing the side lobe-levels and nulls of the pattern by the constraints of imposing the biasing voltages on the varactors. This is actually performed through a trial and error approach, which actually was a long process. A PC has been used to control and execute automatically the DAC output during the pattern measurements.