Geaney Ciaran S 201911 MAS Thesis

Reflectarray Antennas for Multiple-Beam Satellite Applications
by
Ciaran Sean Geaney
A thesis submitted in conformity with the requirements

for the degree of Master of Applied Science
Graduate Department of Electrical and Computer Engineering
University of Toronto
c Copyright 2019 by Ciaran Sean Geaney

Abstract
Reflectarray Antennas for Multiple-Beam Satellite Applications
Ciaran Sean Geaney

Master of Applied Science
Graduate Department of Electrical and Computer Engineering
University of Toronto
2019
With the demand for high data rate communications systems continually increasing, there is a need for
high-gain antennas which can meet increasingly difficult requirements while occupying less volume. This
is especially true in satellite communications applications.
Modern communications satellites use a coverage scheme which services a geographical region with a
set of narrow spot beams which differ in frequency and/or polarization, allowing for increased capacity
through frequency reuse. One solution to generate this coverage is to place a set of feeds in front of an
offset reflector whereby each feed generates a separate beam. However, this generally requires multiple
reflectors, which occupy mass and volume on the satellite that could be used for other services. In this
thesis two methods are investigated that could potentially reduce the number of reflectors required. The
results could also be useful for non-satellite applications which employ multiple-beam antennas that use
feed displacement.
ii
To my parents, Columba and Elaine,
and my brother Michael.
iii
Acknowledgements
This work would not have been possible without the contributions and support of others, and in this
section I would like to take the opportunity to thank them. First and foremost, I would like to thank
my supervisor Sean Hum for his support and guidance throughout my graduate studies, with regard to
both research and non-research matters, and for encouraging me to pursue new ideas and opportunities.
His enthusiasm, positive outlook, and patience with me are greatly appreciated, and were much needed
at times. Working with him for the last several years has been a period of professional and personal
growth.
I would like to thank those who I collaborated with. For the reflectarray synthesis work, the dis-
cussions with Eduardo Martinez-de-Rioja and Prof. José Encinar at UPM were valuable to reach our
conclusions. Eduardo’s work on the reflectarray layout was also very helpful. I would like to thank our
former undergraduate thesis student Jianwei Sun for his excellent work on developing the genetic algo-
rithm optimizer. The dual-circular-polarization reflectarray work would not have been possible without
Mehdi Hosseini’s polarizer and initial studies, and so I would like to thank him for this and for the discus-
sions that we had. I would also like to thank our industry partners at MDA, Yves Demers and Mathieu
Riel, for providing practical input regarding applications, and IDS for providing the ADF licenses.
I would like to thank Alex, Bozhou, Lisa, Parinaz, Paul, Talia, Trevor, and Utkarsh for the interesting
and useful discussions which we had, and/or for working on tools which I used in my research work. I
would also like to thank Albert, Ayman, Michael, and Tony for advice regarding simulations, fabrication,
and measurements. To my fellow masters students Kevin, Jeremy, Philip, and Steven, as well as anyone
else who I didn’t mention, thanks for the shared experiences and I wish you all the best for the future.
Finally, I would like to thank my parents, who have been supportive of my journey regardless of
where it has taken me.
iv
Contents
1 Introduction 1
1.1 Antennas for Satellite Communication Applications . . . . . . . . . . . . . . . . . . . . . . 2
1.1.1 Phased Array Antennas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.2 Reflector Antennas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.3 Reflectarray Antennas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2 Multiple Spot Beam Coverages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2.1 Single-Feed-per-Beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2.2 Multiple-Feeds-per-Beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.2.3 Direct Radiating Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3 Motivation and Thesis Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3.1 Bifocal Reflectarray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.3.2 Dual-polarization Reflectarrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.3.3 Thesis Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.4 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2 Background 16
2.1 Antenna Array Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 Reflectarray Background and Design Principles . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2.1 Phase Shift Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.2.2 Phase Derivative Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.2.3 Dual Reflectarrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3 Analysis of Reflectarray Elements and Periodic Structures . . . . . . . . . . . . . . . . . . 25
2.3.1 Element Modelling and Phase Shift Mechanism . . . . . . . . . . . . . . . . . . . . 25
2.3.2 Multi-layer Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.3.3 Rotation of One-port Scattering Parameters . . . . . . . . . . . . . . . . . . . . . . 32
2.3.4 Circularly Polarized S-parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.4 Analysis of Reflectarrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.4.1 Single Reflectarrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.4.2 Dual Reflectarrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.4.3 One-dimensional Dual Reflectarrays . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.5 Antenna Performance Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.5.1 Radiated Power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.5.2 Aperture Efficiency, Directivity, and Gain . . . . . . . . . . . . . . . . . . . . . . . 51
v
2.5.3 Radiated Power, Directivity, and Gain in Two-dimensional Space . . . . . . . . . . 53
2.5.4 Co-polar and Cross-polar Directivity and Gain . . . . . . . . . . . . . . . . . . . . 54
3 Bifocal Reflectarray 55
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.2 Bifocal Reflector Synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.3 Bifocal Reflectarray Synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.4 Application to Multi-beam Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.5 Prime-focus Geometry with BCF = 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.5.1 Reference Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.5.2 Bifocal Reflectarray Synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.6 Exploring the Synthesis Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.6.1 Selection of Different Foci . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.6.2 Initial Phase Derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.6.3 Sweeping the BCF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.7 Bifocal Reflectarray Synthesis with BCF = 1 . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.7.1 Comparison of BCF = 2 Case to a Cassegrain Reflectarray . . . . . . . . . . . . . 80
3.8 Extension of the Prime-focus Geometry to 2D Arrays . . . . . . . . . . . . . . . . . . . . 85
3.8.1 Full-wave Simulation of the Sub-reflectarray . . . . . . . . . . . . . . . . . . . . . . 88
3.9 Other Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
3.9.1 Offset Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
3.9.2 Optimization-Based Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.10 Comments and Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4 Dual-polarization Reflectarrays 98
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.2 Design Considerations for Reflectarray Elements . . . . . . . . . . . . . . . . . . . . . . . 98
4.2.1 Bandwidth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.2.2 Polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.2.3 Periodicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
4.2.4 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.2.5 Reconfigurability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.3 Reflectarray Element Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.3.1 Stacked Jerusalem Cross Element . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.3.2 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
4.3.3 Comparison to Other Dual-LP Elements . . . . . . . . . . . . . . . . . . . . . . . . 110
4.3.4 Comparison to Other Reported Sub-wavelength Elements . . . . . . . . . . . . . . 115
4.4 Extension to Circular Polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
4.4.1 Comparison to Other Dual-CP Control Methods . . . . . . . . . . . . . . . . . . . 119
4.4.2 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
4.4.3 Cross-polar Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
4.5 Reflectarray Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
4.5.1 Antenna System Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
4.5.2 Reflectarray Synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
vi
4.6 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
4.6.1 Feed Horn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
4.6.2 Dual-LP Reflectarray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
4.6.3 Dual-CP Reflectarray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
4.7 Comments and summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
5 Conclusions 157
5.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
5.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
5.2.1 Bifocal Reflectarray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
5.2.2 Dual-Polarization Reflectarray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
5.3 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
Bibliography 162
vii
List of Tables
3.1 Numerical values for the prime-focus reference geometry shown in Fig. 3.6. . . . . . . . . 63
3.2 The feed positions and naming scheme used, and the colours used to plot the resulting
beams and amplitude distributions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.3 Initial conditions used in Section 3.5.2 for the bifocal synthesis procedure. . . . . . . . . 67
3.4 Initial conditions used for the bifocal synthesis procedure in Section 3.6.1. . . . . . . . . 72
3.5 Initial conditions for Section 3.6.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.6 Initial conditions used for the bifocal synthesis procedure in Section 3.6.3. . . . . . . . . 77
3.7 Initial conditions for the bifocal synthesis procedure used in Section 3.7. . . . . . . . . . 80
3.8 Geometrical parameters of the triple co-planar dipole reflectarray unit cell. . . . . . . . . 89
3.9 Comparison of the bifocal reflectarray performance between the ideal phase shifter AF
and MoM/AF simulation methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.1 Geometrical parameters for the Jerusalem cross unit cell pictured in Fig. 4.3. . . . . . . . 105
4.2 Comparison of several two-layer dual-polarized reflectarray elements from the results pre-
sented in Fig. 4.9 and Fig. 4.7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.3 Comparison of different dual-LP reflectarray unit cells based on simulation results. ‘+’,
’0’, and ‘−’ refer to good, acceptable, and poor, respectively. . . . . . . . . . . . . . . . . 117
4.4 Comparison to reduced size reflectarray unit cells reported in the literature. . . . . . . . 117
4.5 Polarizer unit cell dimensions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
4.6 Comparison of the simulated dual-CP reflectarray unit cell performance at 20.5 GHz when
using the existing polarizer or one designed for higher return loss. . . . . . . . . . . . . . 124
4.7 Simulated dual-LP reflectarray performance before and after choosing its phase offset
according to the parametric study of Fig. 4.25. The values for co-polar gain and cross-
polar discrimination are provided at 20.5 GHz. . . . . . . . . . . . . . . . . . . . . . . . . 132
4.8 Comparison to other wideband reflectarray works. . . . . . . . . . . . . . . . . . . . . . . 133
4.9 Simulated dual-CP reflectarray performance before and after choosing phase offset accord-
ing to parametric study of Fig. 4.27 . The values for co-polar gain and cross-polarization
ratio are given at 20.5 GHz. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
4.10 Measured co-polar beam peak directions at 20.5 GHz and their corresponding design di-
rections for the dual-LP reflectarray. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
4.11 Measured co-polar beam peak directions at 20.5 GHz and their corresponding design di-
rections for the dual-CP reflectarray. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
viii
List of Figures
1.1 Usage of radio and microwave communication bands. . . . . . . . . . . . . . . . . . . . . . 2

1.2 Computed atmospheric transmission at the summit of Mauna Kea with a water vapour
level of 0.5 mm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Schematic drawing of a transmit phased array. . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Cross-sectional schematic view of an offset paraboloidal reflector. . . . . . . . . . . . . . 5
1.5 Cross-sectional schematic view of a reflectarray antenna. . . . . . . . . . . . . . . . . . . 6
1.6 A hypothetical example of a four-colour coverage scheme. The map of western Canada is
taken from an online source. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.7 Generation of multiple beams using SFB with a single reflector. . . . . . . . . . . . . . . 9
1.8 Generation of multiple beams using SFB with multiple reflectors. . . . . . . . . . . . . . 10
1.9 Examples of SFB antenna systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.10 Generation of multiple beams using MFB with a single reflector. . . . . . . . . . . . . . . 11
1.11 Examples of prototype MFB feed systems. . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.1 A planar array of antenna elements arranged in a rectangular grid. . . . . . . . . . . . . 17

2.2 The geometry of an offset pencil beam reflectarray, illustrating the conversion from a
spherical wave to a plane wave. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3 Cross-section of a reflectarray directing rays from an incident plane wave to a non-specular
direction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.4 A dual-reflectarray antenna transforming a spherical wave to a plane wave. . . . . . . . . 24
2.5 Analysis of a unit cell using periodic boundary conditions. . . . . . . . . . . . . . . . . . . 26
2.6 The conventions used in this thesis for the plane and angles of incidence. . . . . . . . . . . 27
2.7 The simulated reflection characteristics of the square patch illustrated in Fig. 2.5 at normal
incidence. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.8 Analysis of a multi-layer periodic structure. . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.9 The supercell method of analyzing multi-layer periodic structures. In (a) the loop surface
has a periodicity which is a quarter of that of the patch surface. In (b) the cell period
of the loop surface is not an integer multiple of the patch surface, and their lattices are
rotated with respect to one another. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.10 Example unit cell with the slant polarization convention shown. . . . . . . . . . . . . . . 32
2.11 Transmitted and reflected circularly-polarized waves normally incident on an interface,
illustrating the circular polarization conventions used in this thesis. . . . . . . . . . . . . . 33
2.12 Reflectarray analysis geometry. The element which the feed ray is incident on is being
treated as an xR -polarized uniformly illuminated aperture. . . . . . . . . . . . . . . . . . 37
ix
2.13 Comparison between the feed model of (2.49a) with qE = 28 and qH = 17 to a pyramidal
horn simulated at 20 GHz using HFSS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.14 Analysis of a rectangular aperture considering two orthogonal linear polarizations. . . . . 39
2.15 Dual-reflectarray analysis geometry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.16 Scattering of a TM-polarized incident field from a strip, which is used to determine the
element factor for the analysis of one-dimensional reflectarrays. . . . . . . . . . . . . . . 45
3.1 Examples of early bifocal spatially-fed antennas. . . . . . . . . . . . . . . . . . . . . . . . 56

3.2 Rappaport’s offset dual-reflector synthesis. . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.3 Iterative ray-tracing procedure. (a) A ray is traced from FA to a known point on the
sub-reflectarray, and the phase derivative at this point and the intersection of the main
reflectarray with the z-axis are used to find the next main reflectarray point. (b) The
desired ray direction is used to determined the required phase derivative at the new main
reflectarray point. (c) The phase derivative on the main reflectarray is used to reflect a
ray incident at −α which determines the next sub-reflectarray point. (d) A ray is traced
from the new sub-reflectarray point to focus FB to complete the iteration. The new sub-
reflectarray point and phase derivative are used with step (a) to start the next iteration.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.4 Flow chart of the bifocal reflectarray synthesis procedure. RA is an abbreviation for
reflectarray. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.5 An example prime-focus paraboloidal reflector. (a) The feed horns are placed as close
as physically possible such that each feed is displaced by its diameter. (b) The feed
displacements are adjusted to give the required beam spacing, resulting in feeds that are
physically overlapping. The +θ and −θ arcs in the left-hand side of (a) and (b) indicate
the conventions used for the positive negative angular directions shown in the pattern
plots. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.6 Symmetric centre-fed dual reflectarray geometry. The locations of the real focus (FR )
and virtual focus (FV ) are shown for the reference geometry and are equidistant from the
sub-reflectarray. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.7 The phase and amplitude distributions on the reflectarrays for the prime-focus reference
dual-reflectarray shown in Fig 3.6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.8 The beams produced by the reference dual-reflectarray using a 55 mm feed spacing. . . . 65
3.9 Ray-tracing procedure for the upper half of the geometry shown in Fig. 3.6 with BCF = 2
and the foci located two feeds from the centre (F2 and F6 ). The blue lines are rays
traced from FA (transmit mode), the red lines are rays traced to FB (receive mode). The
circles indicate the two foci, and the points on the two reflectarrays obtained from the
synthesis. The blue points were obtained from the transmit rays, apart from the initial
sub-reflectarray point, and the red from the receive rays. In (e) the thick lines represent
the main reflectarray and sub-reflectarray, and the feed positions are shown. . . . . . . . 68
3.10 (a)–(b) The phase derivatives obtained from the ray-tracing procedure using the initial
conditions given in Table 3.3. The circles are the points which were obtained from the syn-
thesis procedure, and the curves are the result of interpolating over the reflectarray points.
(b) The phase distributions obtained by numerically integrating the phase derivatives. . 69
x
3.11 The (a)–(b) amplitude distributions on the reflectarrays and (c) the beams produced when
the bifocal synthesis with the parameters given in Table 3.3 is applied to the prime-focus
dual-reflectarray with a cluster of 7 feeds. . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.12 Comparison of the bifocal synthesis results when different separations between the focal
points are used. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.13 Comparison of the patterns produced using different foci spacings. . . . . . . . . . . . . . 74
3.14 Performance of the bifocal synthesis procedure as the initial phase derivative is swept
using the initial conditions given in Table 3.5. . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.15 Phase distributions on the sub- and main reflectarrays as the initial phase derivative is
swept. The values in the legend are given in rad/m. . . . . . . . . . . . . . . . . . . . . . 75
3.16 Some selected patterns resulting from the bifocal synthesis as the initial phase derivative
is swept. The value of the initial sub-reflectarray phase derivative corresponding to each
set of patterns is given in the caption. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.17 Some performance metrics of the bifocal synthesis procedure as applied to the configura-
tion of Section 3.5.2 as the BCF is swept from 0.45 to 2.5. . . . . . . . . . . . . . . . . . . 77
3.18 Phase distributions on the reflectarrays resulting form the bifocal synthesis procedure as
the BCF is swept. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.19 Some selected patterns resulting from the bifocal synthesis procedure as the BCF is swept. 79
3.20 Phase derivatives, and resulting phase and amplitude distributions on the reflectarrays
with BCF = 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.21 Beams produced by the bifocal reflectarray (solid) with BCF = 1 compared with the single
focus reference (dashed). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
3.22 Tracing the rays from a plane wave incident at broadside to estimate the focal point of
the main reflectarray. The real focus is to the left of the sub-reflectarray, and located at
F4 , and the estimated virtual focus is to the right. . . . . . . . . . . . . . . . . . . . . . . 82
3.23 Comparison of the bifocal and similar Cassegrain reflectarray phase distributions. . . . . 83
3.24 Comparison of the beams produced by the bifocal and similar Cassegrain reflectarrays.
The solid lines are from the bifocal reflectarray, and the dashed from the Cassegrain. . . 83
3.25 Amplitude distribution on the main reflectarray of the similar Cassegrain configuration. . 83
3.26 Two typical dual-reflector antenna geometries. . . . . . . . . . . . . . . . . . . . . . . . . 84
3.27 Points obtained by rotating the phase distributions of Fig. 3.10c and Fig. 3.10d . . . . . . 86
3.28 Interpolated phase distributions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
3.29 Wrapped phase distributions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.30 Comparison of the beams produced by the bifocal synthesis from a 1D reflectarray, and
when the phase distributions have been rotated to form a 2D reflectarray. For the 2D
reflectarray, the elevation cuts of the co-polar gain patterns are being plotted. . . . . . . 87
3.31 Gain contours in direction cosine space for the 2D bifocal reflectarray and its folded
reference geometry, where u = sin θ cos φ and v = sin θ sin φ. . . . . . . . . . . . . . . . . . 88
3.32 (a) Top view and (b) side view of the triple co-planar dipole reflectarray unit cell with
lA1 = 5 mm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
xi
3.33 (Left) The full meshed sub-reflectarray layout consisting of triple coplanar dipole unit
cells, and (right) a zoomed-in portion of the layout. The layout was done by Eduardo
Martinez de Rioja from Prof. Encinar’s group at UPM, and meshing was performed using
Gmsh. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
3.34 Normalized magnitude of the x-component of the incident field on the main reflectarray
at 20 GHz for four different feed positions using the (a)–(d) array factor with ideal phase
shifters on the sub-reflectarray and a q = 1 feed model, (e)–(h) MoM with a λ0 /2 dipole
feed and the real reflectarray layout. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
3.35 Phase of the x-component of the main reflectarray incident field at 20 GHz for four different
feed positions using (a)–(d) array factor with ideal phase shifters on sub-reflectarray and
q = 1 feed model, and (e)–(f) MoM with a λ0 /2 dipole feed. . . . . . . . . . . . . . . . . . 91
3.36 Principal plane cuts for half of the feeds at 20 GHz computed using AF for a main re-
flectarray incident field that was simulated using MoM with a real sub-reflectarray layout
(labelled MoM/AF), and AF assuming ideal phase shifters on the sub-reflectarray (la-
belled AF). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
3.37 Example of defining a set of masks based on a desired radiation pattern. . . . . . . . . . 93
3.38 Single point masks with BCF = 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.39 Comparison of the phase distributions obtained from the bifocal synthesis and GA optimizer. 94
3.40 Comparison of the beams produced by the bifocal synthesis procedure and the GA op-
timizer. The solid lines are from the bifocal reflectarray, and the dashed from the GA.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3.41 Amplitude distributions on the main reflectarray resulting from the GA optimized and
bifocal reflectarrays for the lower half of the feed cluster (F1 to F4 ). The bifocal results
are shown as dashed lines, and the GA results as solid lines. . . . . . . . . . . . . . . . . . 96
4.1 Example rectangular patch element showing the polarizations which its dimensions are
aligned with. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.2 Dual circular polarization control concepts from the literature. α is the angular separation
between the RHCP and LHCP beams, and f1 and f2 are the operating frequencies. . . . 102
4.3 Stacked Jerusalem cross reflectarray unit cell geometry. . . . . . . . . . . . . . . . . . . . 106
4.4 Element variation as lx is increased with ly fixed at 2.7 mm. . . . . . . . . . . . . . . . . 107
4.5 The simulated co-polar reflection response of the two-layer Jerusalem cross element to a
normally incident x0 -polarized plane wave at (a)–(b) 20.5 GHz for several values of ly as
lx is swept, and (c)–(d) for several values of lx with ly = 2.9 mm as frequency is swept. . . 108
4.6 Simulated reflection response of the two-layer Jerusalem cross element to a TM-polarized
incident plane wave as the angle of incidence is swept in its E-plane and H-plane at
20.5 GHz with ly = 2.9 mm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.7 Simulated reflection response of the stacked Jerusalem cross element as both lx and ly are
swept at 20.5 GHz and normal incidence. . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
4.8 Other dual-polarization reflectarray elements that were considered. In (a) px = py =
3.75 mm, lx,l − lx,u = ly,l − ly,u = 0.5 mm (b) lx,l /lx,u = ly,l /ly,u = 0.7, px = py = 3.75 mm
for the quarter-wavelength cell and px = py = 7.50 mm for the half-wavelength cell. . . . 111
xii
4.9 Simulated co-polar reflection responses of some dual-LP reflectarray unit cells to a nor-
mally incident VP wave at 20.5 GHz. (a)–(d) λ0 /4 × λ0 /4 two-layer stacked rectangular
patches, (e)–(h) λ0 /2 × λ0 /2 two-layer stacked rectangular patches, (i)–(l) λ0 /4 × λ0 /4
two-layer stacked crossed dipoles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
4.10 Simulated co-polar reflection response of the stacked λ0 /4 × λ0 /4 crossed-dipole element
illustrated in Fig. 4.8a to a normally incident TM-polarized plane wave (a)–(b) at 20.5 GHz
for several values of ly as lx is swept, and (c)–(d) as a function of frequency at several
values of lx with ly = 2.9 mm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
4.11 Simulated co-polar reflection response of a two-layer λ0 /2×λ0 /2 stacked rectangular patch
unit cell, pictured in Fig. 4.8b, to a normally incident TM-polarized plane wave (a)–(b)
at 20.5 GHz for several values of ly as lx is swept, and (c)–(d) for several values of lx with
ly = 4.1 mm as frequency is swept. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
4.12 Simulated TM polarization reflection response of the stacked λ0 /4 × λ0 /4 crossed-dipole
element illustrated in Fig. 4.8a as lx is swept with ly = 2.9 mm at several angles of incidence.115
4.13 Simulated TM polarization reflection response of the two-layer λ0 /2 × λ0 /2 stacked rect-
angular patch unit cell, pictured in Fig. 4.8b, as lx is swept with ly = 4.1 mm at several
angles of incidence. Note that the y-axis limits are different from the other angle of
incidence plots on account of the increased variation. . . . . . . . . . . . . . . . . . . . . 116
4.14 Polarizer unit cell. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
4.15 Reflectarray–polarizer stack-up, with an illustration of how the two are combined to im-
plement independent dual-CP control. The height of the air gap is 8 mm. . . . . . . . . . 119
4.16 Transmission line model of full structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
4.17 Simulated reflection characteristics of the cascaded polarizer–reflectarray at 20.5 GHz and
normal incidence. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.18 Generation of cross-polarization from polarizer mismatch. . . . . . . . . . . . . . . . . . . 123
4.19 Axial ratio of the reflected fields when a RHCP wave is incident on the dual-CP reflectarray
unit cell as the air gap size is varied with ly = 2.7 mm. . . . . . . . . . . . . . . . . . . . . 123
4.20 Cascaded polarizer–reflectarray cross-polar reflection coefficient magnitude at 20.5 GHz
as the size of the air gap between the reflectarray and polarizer is varied. . . . . . . . . . 124
4.21 Simulated reflection characteristics of the cascaded polarizer–reflectarray at 20.5 GHz and
normal incidence when a polarizer unit cell with higher return loss is used. . . . . . . . . 125
4.22 The antenna system geometry with the reflectarray’s equivalent paraboloidal reflector.
(x, y, z) is the global coordinate system and (xR , yR , zR ) is the reflectarray coordinate
system, where y and yR are the same. The transverse dimension of the reflectarray, in
the y-direction, is 19.5 cm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
4.23 Required phase distributions for both beams. The beam directions in the global coordinate
system are (a) θ = 0◦ and (b) θ = 5◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
4.24 Phase wrapping routine applied to the required phase distribution to produce the HP
beam for the dual-LP reflectarray with ψ0 = 500◦ . . . . . . . . . . . . . . . . . . . . . . . 129
4.25 Results of the phase offset parametric study for the dual-LP reflectarray. . . . . . . . . . 131
4.26 (a)–(b) Computed performance of the dual-LP reflectarray, and (c)–(d) distributions of
assigned element lengths before and after adding the phase offsets. . . . . . . . . . . . . . 132
4.27 Results of sweeping the phase offsets of both polarizations for the dual-CP reflectarray. . 135
xiii
4.28 (a)–(b) Computed dual-CP reflectarray performance, and (c)–(d) distributions of assigned
elements before and after adding the phase offset. . . . . . . . . . . . . . . . . . . . . . . 136
4.29 Fabricated reflectarray prototypes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
4.30 Dual-LP reflectarray with mounting structure and feed horn, set up in near-field scanner. 138
4.31 Standard gain horn used to determine absolute gain, set up in near-field scanner. . . . . 139
4.32 Measured normalized magnitude patterns of the dual-polarization feed horn with a com-
parison to the cosq θ model used. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
4.33 Measured co-polarized phase patterns of the dual-polarization feed when the origin of the
measurement coordinate system is located at the mouth of the horn. . . . . . . . . . . . 141
4.34 Measured co-polarized phase patterns of the dual-polarization feed when the origin of the
measurement coordinate system is located about 2 cm inside the horn. . . . . . . . . . . 142
4.35 Measured patterns cuts of the dual-LP reflectarray at some selected frequencies when the
feed is vertically polarized, with a comparison to AF predictions. The black dash-dot line
shows the design direction of the beams in the elevation plane. . . . . . . . . . . . . . . . 145
4.36 Measured patterns cuts of the dual-LP reflectarray at some selected frequencies when the
feed is horizontally polarized, with a comparison to AF predictions. The black dash-dot
line shows the design direction of the beams in the elevation plane. . . . . . . . . . . . . . 146
4.37 Dual-LP reflectarray measured patterns at some selected frequencies across two-dimensional
space. In (a)–(h) the feed is vertically polarized, and in (i)–(p) it is horizontally polarized. 147
4.38 Measured (a) co-polar gain, (b) cross-polar discrimination, and (c) total directivity vari-
ation with frequency of the dual-LP reflectarray with a comparison to array factor pre-
dictions. The values are taken at the intersections of the principal plane cuts. . . . . . . 148
4.39 Measured pattern cuts of the dual-CP reflectarray resulting from the synthesized RHCP
feed at some selected frequencies, with a comparison to AF predictions. The black dash-
dot line shows the design direction of the beams in the elevation plane. . . . . . . . . . . . 152
4.40 Measured pattern cuts of the dual-CP reflectarray resulting from the synthesized LHCP
feed at some selected frequencies, with a comparison to AF predictions. The black dash-
dot line shows the design direction of the beams in the elevation plane. . . . . . . . . . . . 153
4.41 The measured patterns of the dual-CP reflectarray in direction cosine space at some
selected frequencies. By post-processing the LP measurements, in (a)–(h) the feed is
operating in RHCP and in (i)–(p) it is operating in LHCP. . . . . . . . . . . . . . . . . . 154
4.42 Measured co-polar gain and cross-polar discrimination variation with frequency of the
dual-CP reflectarray in the principal planes of each beam with a comparison to array
factor predictions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
4.43 (a) Simulated transmitted cross-polarization of the polarizer for incident RHCP as the
angle of incidence is varied in the φ = 0 plane. (b) The angle of incidence across the
truncated reflectarray surface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
xiv
List of Acronyms
AF Array factor
AR Axial ratio
BCF Beam compression factor
BDF Beam deviation factor
BFN Beamforming network
BPE Beam pointing error
CAD Computer aided design
CP Circular polarization
CPSS Circular polarization selective surface
DFT Discrete Fourier transform
DGR Dual-gridded reflector
DRA Direct radiating array
FBW Fractional bandwidth
FEBI Finite Element Boundary Integral
GA Genetic algorithm
GEO Geostationary Earth orbit
GSM Generalized scattering matrix
HP Horizontal polarization
IDFT Inverse discrete Fourier transform
LEO Low Earth orbit
LH-CPSS Left-handed circular polarization selective surface
LHCP Left hand circular polarization
LP Linear polarization
LUT Look-up table
MFB Multiple-feeds-per-beam
MoM Method of Moments
OMT Ortho-mode transducer
PEC Perfect electric conductor
PMC Perfect magnetic conductor
RADAR Radio detection and ranging
RF Radio frequency
RH-CPSS Right-handed circular polarization selective surface
RHCP Right hand circular polarization
xv
SFB Single-feed-per-beam
SGH Standard gain horn
SLL Side lobe level
TE Transverse electric
TM Transverse magnetic
UPM Technical University of Madrid
VP Vertical polarization
VRT Variable rotation technique
XPD Cross-polar discrimination
xvi
Chapter 1
Introduction
One of the most significant innovations of the 19th century is wireless communication, whereby infor-
mation can be modulated on to a high frequency carrier signal and transmitted through free space via
electromagnetic waves. This allows people from all over the world to communicate with one another
simply by having an appropriate receiver to pick up and demodulate the transmitted signals. These
technologies have had a significant impact on our society by providing us with increased access to in-
formation, and by transforming our economies and industries, and over the last few decades they have
become ubiquitous. Wireless technologies are also by no means limited to communications, and play
important roles in radio detection and ranging (RADAR), navigation, and sensing applications, to name
a few examples.
A crucial part of any wireless system is the antenna, which acts as an interface between propagating
electromagnetic waves, and fields which are confined to cables, printed circuit boards, or waveguides.
Depending on the application they are used for, antennas can take on many forms. Cellular phone
antennas are small enough to be contained in a device that fits inside one’s pocket, whereas antennas
used for radio astronomy and deep space communications can be as large as a building. In any case, the
growth and development of wireless technology, and its foray into new applications, has increased the
demands on antennas: they must provide more functionality while fitting into a smaller form factor.
While the networks of today often rely on antennas with fairly broad patterns, such as sector antennas
found on base stations, or omnidirectional antennas used in cell phones or routers, modern high-capacity
communications systems are moving towards high-gain directional antennas. High-gain antennas focus
power over a small angular region, allowing for transmit power levels to be reduced while still meeting the
margins required to maintain a high quality link. The narrow beams of high-gain antennas also provide
spatial isolation between communications channels. This can enable techniques to be applied which
make more efficient use of the already crowded frequency spectrum. Having multiple isolated channels
also opens up opportunities for signal processing algorithms to be applied to improve the capacity of a
communications system.
Part of the increased interest in high-gain antennas comes from the use of higher frequencies in modern
and future communications systems. As the dimensions of an antenna scale with wavelength, they can
be made smaller and more lightweight at higher frequencies. This allows for the use of electrically large
apertures, which are required to achieve high gain, that do not occupy much volume. One motivation for
exploring higher frequencies is that many potential interferers are present in the lower frequency bands
1
Chapter 1. Introduction 2
due to the prevalence of devices communicating using standards such as LTE, WiFi, and Bluetooth.
This is visible from Fig. 1.1 which illustrates the usage of different frequency bands by application.
Another factor is that higher frequency bands offer more bandwidth. As an example consider a 20 MHz
communications channel: at 20 GHz this occupies a significantly smaller fractional bandwidth than at
2 GHz. This allows for higher channel capacities to be achieved.
Figure 1.1: Usage of radio and microwave communication bands [1].
High-gain antennas are often used in satellite communications due to the large distance over which
communications satellites must transmit, and the narrow region on the ground which they cover. Mass
and volume are highly constrained on satellites, and so innovative high-gain antenna solutions are re-
quired which are lightweight and occupy less space while meeting challenging electrical requirements.
These factors are becoming even more important as the demand for high data rate communications
services continues to increase. Satellites are able to fulfil this demand in areas where it may be diffi-
cult or not economically feasible to construct extensive terrestrial networks, such as rural communities,
developing nations, and regions with difficult terrain. Additionally, smaller lightweight satellites for
which mass and volume are even more constrained are becoming popular for both scientific and commu-
nications applications. In the following sections some antenna solutions and requirements for satellite
communications will be outlined, and some potential methods to reduce the required antenna volume
for a specific coverage scheme will be proposed.
1.1 Antennas for Satellite Communication Applications

There are a wide variety of satellite communications applications, and as such there are many different
requirements on antenna characteristics such as gain, bandwidth, operating frequency, polarization, and
power handling. The requirements will also depend on the orbit which the satellite is located in. For
example, a satellite in low Earth orbit (LEO) will have a finite window of time for which it is in view
of the ground station. In this situation an antenna with a beam that is steerable over a wide angular
range may be desirable in order to maintain a good link as the satellite moves across the sky. However,
in geostationary Earth orbit (GEO) the satellite and ground station do not move relative to each other.
In this scenario a steerable antenna may not be as important as one which has much higher gain in order
to overcome the higher free space path losses due to being located farther away. In this thesis the main
focus will be on high-gain antennas with fixed patterns for potential use in GEO satellite applications.
There are a number of factors that differentiate satellite communications applications from terrestrial
ones. These factors play a role in what sort of antennas are chosen. First of all, communications satellites
tend to operate at high frequencies. While these frequency bands offer higher bandwidths, they are also
more prone to environmental factors such as rain which can severely attenuate signals. Another factor
that satellites must consider is the transmission characteristics of the Earth’s atmosphere. An example
atmospheric transmission curve is plotted in Fig. 1.2, where several absorption bands can be identified
where electromagnetic waves do not propagate. These bands must be avoided for links to a ground
station. Additionally, when electromagnetic waves travel through the ionosphere their polarization is
rotated due to the Faraday effect. For linearly-polarized waves this can cause polarization mismatch
between the satellite and ground station antennas, resulting in losses. Polarization mismatch for linearly-
polarized antennas can also occur due to motion of the satellite. For these reasons the use of circularly-
polarized antennas is common in satellite applications.
100
90
80
70
Transmission [%]
60
50
40
30
20
10
0
0 50 100 150 200 250 300
Frequency [GHz]
Figure 1.2: Computed atmospheric transmission at the summit of Mauna Kea with a water vapour level
of 0.5 mm [2].
High-gain satellite antennas tend to fall into two categories: phased array antennas and spatially-
fed antennas. For the latter case reflector antennas are the most common solution, although recently
reflectarrays have been garnering more attention. A characteristic which all of these antennas have is
that their gain is proportional to the size of their aperture. This offers a simple way to tailor the antenna
to fit the gain requirements for a given application, which may be simpler for some technologies than
others. These types of antennas will be described briefly in the following sections, and a more detailed
analysis will follow in Chapter 2.
1.1.1 Phased Array Antennas

Phased array antennas consist of a collection of antennas, referred to as elements, which are excited
with a particular amplitude and phase in order to produce a desired radiation pattern. An illustration
of a phased array operating in transmit mode is provided in Fig. 1.3. In the example shown the element
excitations are delayed such that rays from each of them will add up constructively in the far-field
to produce a planar wavefront, resulting in a directive radiation pattern. This shows two desirable
characteristics of a phased array: by appropriately exciting a number of antenna elements the gain can
be increased far beyond what is achievable from a single element, and the beam produced by the array
can be scanned to various positions in space without having to mechanically steer it. Different radiation
patterns can be formed by appropriately adjusting the amplitude and phase of the signals driving the
elements, which can be used to produce nulls in the direction of interfering signals, to reduce radiation
in undesired directions, or to produce a beam with a specific shape. These adjustments can be made
dynamically through the use of electronically tunable components.
Antenna elements
Power amplifiers
PA
PA
PA
PA
PA
PA
PA
PA
ϕ ϕ ϕ ϕ ϕ ϕ ϕ ϕ Phase shifters
Power splitting network
Transmitter output
Figure 1.3: Schematic drawing of a transmit phased array.
Phased arrays are highly flexible, as the radiation characteristics of the array can be adjusted simply
by changing the excitation of the elements. However, practically implementing the required excitation
requires the design of a feed network. This can become quite complex and challenging to lay out for arrays
containing large numbers of elements, which are required to achieve high gain, and the feed network will
additionally have significant losses. Large phased arrays also require large quantities of electronics, such
as power amplifiers, low noise amplifiers, and phase shifters. While these requirements may be somewhat
alleviated for a digital beamforming array, where delays can be applied to the baseband signal rather
than the radio frequency (RF) signal, a large amount of processing power is still required. All of these
components are expensive, and as such phased arrays are not always practical solutions for fields outside
of aerospace and defence where low quantity custom solutions are the norm, and performance is valued
over cost. However, this may change as the frequencies used for communications increase, resulting in
smaller antennas, and as higher power and integration levels become achievable in RF integrated circuits.
1.1.2 Reflector Antennas

Reflector antennas are a fairly popular type of directional antenna and are often seen on residential
buildings and cellular communications towers, and at airports. An example is shown in Fig. 1.4, where
it is visible that the reflector antenna consists of two components: a reflective surface, and a feed.
In some applications multiple reflective surfaces may be used. The goal of the reflective surface is to
intercept radiation from the feed and use the shape of its surface to produce a desired radiation pattern.
Commonly the shape of the reflector is paraboloidal, due to its focusing properties. If a spherical wave
source is placed at the focus of a paraboloidal reflector, it will then reflect the rays from the source such
that the reflected rays are all parallel. If a plane that is normal to the reflector axis is drawn, this has
the effect of equalizing the distance travelled by each ray from the source to the plane. This produces a
pencil beam, the gain of which is generally far beyond that of the feed, although the gain will depend
on the reflector size.
Reflector
Parent reflector
C
Feed
F
Figure 1.4: Cross-sectional schematic view of an offset paraboloidal reflector.
The main attributes of a reflector antenna which determine its radiation characteristics are its focal
length F and diameter D. These terms are commonly expressed as a ratio, called the F/D ratio of the
reflector antenna. For an offset reflector antenna, only a portion of the full ‘parent’ paraboloid is used
to form the reflector. This is generally done to reduce blockage of the reflector’s main beam by the feed
and its associated electronics and mounting structure. The portion of the paraboloid which is used is
determined by another parameter, the clearance C.
While reflector antennas are commonly used to form narrow high-gain pencil beams, the reflector
curvature can be modified to shape its radiation pattern. In satellite applications this is commonly done
to produce a contoured beam which focuses the radiation pattern onto a geographical region. Reflectors
can also be flat or made up of flat sections, and can have only one curved dimension such as in the case
of cylindrical reflectors.
Some advantages of paraboloidal reflectors are that they have very large bandwidths, as their mode
of operation does not depend on frequency, and they can achieve very high gain. However, a downside
is their size. A high-gain reflector must be electrically large, which at lower frequencies can result in a
very large physical size. This is a drawback for space applications where space and weight are highly
constrained. A large reflector will also likely have to be stowed away during launch, and deployment
in space presents another challenge. For weight reduction the reflectors can be made up of a meshed
material rather than being solid, and to reduce their profile, arrangements with multiple reflectors can
be used. A common multiple reflector arrangement uses two reflectors, a larger main reflector and a
smaller sub-reflector, which is called a dual-reflector antenna.
1.1.3 Reflectarray Antennas

A more recent, but more uncommon, high-gain antenna solution is the reflectarray. An example reflec-
tarray is illustrated in Fig. 1.5, where it is set up to be equivalent to the reflector antenna shown in
Fig. 1.4, and as such it is described by the same geometric parameters.
Ground plane
Elements
Parent reflector
C
Feed
F
Figure 1.5: Cross-sectional schematic view of a reflectarray antenna.
A reflectarray consists of a collection of antenna elements which are fed spatially by a feed antenna.
The antenna elements are generally microstrip-based and are printed on a flat grounded dielectric sub-
strate, although this does not have to be the case. A reflectarray is essentially a hybrid between a phased
array and a reflector antenna. In one sense it is a phased array where the elements are fed spatially from
a feed antenna rather than being driven directly through a transmission line or waveguide based feed
network. In another sense it is a discretized version of a reflector antenna. In any case the functionality
is the same. The reflectarray elements are designed to produce a phase shift which equalizes the electri-
cal length of the ray paths from a spherical source to an aperture plane, producing a narrow high-gain
beam.
As reflectarrays compensate electrical length rather than physical path length, this introduces fre-
quency dependence, resulting in limited bandwidth. This has always been one of the major drawbacks
of reflectarrays, and has been the focus of much research work. Another drawback is their large size,
although for applications where a reflector is already being considered a reflectarray-based solution does
not present a disadvantage as it will have a comparable profile. To reduce their profile, reflectarrays can
be used in dual-reflectarray configurations in a similar manner to their dual-reflector counterparts.
Reflectarrays have several characteristics which lend themselves to space applications. A major ad-
vantage of reflectarrays is that they can produce multiple types of patterns using the same mechanical
structure, as the distribution of elements can simply be changed. This is desirable for satellite applica-
tions, as it allows for pattern synthesis and optimization techniques to be applied without impacting the
mechanical design of the satellite. For curved reflectors their surface must instead be shaped, resulting in
new mechanical characteristics which must be considered. Printed reflectarrays are simple to manufac-
ture as they can take advantage of mature and widely used fabrication processes used to produce printed
circuit boards, such as photolithography. They are also lightweight, and can be made deployable in the
event that they must be stowed during launch. Reflectarrays offer a lot of flexibility with regard to the
manipulation of electromagnetic waves, as both the elements and their distribution in the reflectarray
can be engineered. This makes them promising for applications where high levels of isolation are needed
between different frequency bands and polarizations.
1.2 Multiple Spot Beam Coverages

Due to the large number of wireless communications applications and wireless infrastructure operators,
spectrum is a very valuable resource in any wireless communications system, and satellite systems are
no exception. Qualitatively, the spectral efficiency of a wireless link refers to how efficiently it is able to
utilize the available bandwidth. Some factors which spectral efficiency depends on are the modulation and
multiplexing schemes employed, as well as the characteristics of the channel. As the allocated spectrum
for a given application is divided into multiple sub-bands, one way to improve spectral efficiency is to
use the same sub-bands to transmit multiple data streams. This is called frequency reuse, and a simple
method of implementing it is through spatially isolating multiple links. Spatial isolation can be achieved
by using antennas which produce multiple beams at different positions.
Modern high-speed communications satellites use multiple beam coverage schemes in which a geo-
graphical area is divided in to a set of small, slightly overlapping spots [3, 4]. A hypothetical coverage
pattern is illustrated in Fig. 1.6, where each of the circles is the coverage area on the Earth resulting from
a high-gain antenna transmitting from a satellite located in GEO. Two frequency sub-bands, centred at
the frequencies f1 and f2 , and two polarizations are used. Each frequency/polarization combination is
represented by one colour, resulting in this type of coverage being termed a four-colour coverage scheme.
It is visible from Fig. 1.6 that each beam is never adjacent to a beam having the same colour. As such,
the adjacent beams are ideally completely isolated and do not interfere with each other, allowing them
to operate as independent communications channels. This allows for a high degree of frequency reuse
to be achieved, increasing the spectral efficiency and capacity of the satellite. The overlap between
the beams ensures that there are no gaps in coverage across the region. Other coverage schemes with
different numbers of colours are also possible, for example three- or seven-colour schemes, although the
four-colour case generally gives the best compromise between performance and capacity [3, 5]. Note that
in Fig. 1.6, f1 and f2 are in closely adjacent frequency bands. Taking these to be the transmit bands of
the satellite, another four-colour set of spots will need to be generated for the receive bands.
Although increased capacity is generally the main motivation behind multiple-beam schemes, they
can also be used to tailor coverage according to language or population density [3]. There are multiple
ways in which a multiple spot beam coverage pattern can be generated, which will be covered in the
following sections.
A : f1 , Pol. 1
C D C D C D
B : f2 , Pol. 1
A B A B A B C : f1 , Pol. 2
D : f2 , Pol. 2
C D C D C D
A B A B A B
C D C D C D
A B A B A
Figure 1.6: A hypothetical example of a four-colour coverage scheme. The map of western Canada is
taken from an online source [6].
1.2.1 Single-Feed-per-Beam
The simplest solution to generate a multiple spot beam coverage pattern is to place a set of feeds,
generally referred to as a feed cluster, in front of a reflector. The shape of the reflector is generally
paraboloidal. Due to being displaced from the focus of the reflector, each feed will produce a beam
which is scanned at a different angle based on how much it has been displaced. This is called single-
feed-per-beam (SFB), and this scenario is illustrated in Fig. 1.7, considering a portion of the feeds used
to generate one row of the coverage pattern shown in Fig. 1.6. The reflector does not have to be purely
paraboloidal and may be shaped to correct for beam aberrations incurred from the feeds not being
located at the reflector’s focal point.
When reflector antennas are used, the feeds must be appropriately sized to prevent spillover losses,
which are losses resulting from power radiated from the feeds which is not captured by the reflector.
This sets a constraint on how wide the beamwidth of the feed antennas can be. A commonly-used feed
for reflectors is a horn antenna. The beamwidth of a horn is inversely related to the size of its aperture,
and as such it will generally have a minimum physical size in order to achieve an acceptable amount
Figure 1.7: Generation of multiple beams using SFB with a single reflector.
of spillover loss. As the closest that the feeds in the feed cluster can be placed is such that they are
touching, this results in a minimum spacing of the feeds. Together with the F/D ratio of the reflector
this sets a minimum achievable angular separation of the beams generated by the feed cluster. This is
shown in Fig. 1.7, and in general the achievable beam spacing does not produce beams which overlap
at the level set by the coverage requirements. While the sizes of the feeds could be reduced to obtain
the correct spacing of the beams, this would result in a fairly significant amount of loss due to spillover
[7, 8], which is generally unacceptable due to the stringent requirements of a GEO satellite link budget.
One way to deal with this problem is to increase the size of the reflector to produce an oversized
reflector. In this case the reflector would be shaped instead of paraboloidal in order to provide the
required coverage pattern. This method is not feasible for very narrow spot beams as there are practical
limits on how large the reflector can be made [8].
A more popular solution to the spillover problem is illustrated in Fig. 1.8. Instead of using a single
reflector fed by a single feed cluster, multiple feed clusters and reflectors can be introduced. The beams
produced by each reflector/feed cluster pair can then be interleaved, achieving the desired multiple
beam pattern. In the example shown four reflectors and feed clusters are used. This is fairly typical
[3, 8, 11, 10, 4], particularly for the higher frequency bands where the required reflector and feed sizes
are smaller, although the number of reflectors would depend on the requirements and coverage pattern of
the application. Some examples of reflectors and feeds for multiple-reflector SFB antennas are shown in
Fig. 1.9. The four-reflector solution is applicable if each reflector and feed cluster is shared between the
transmit and receive functions, otherwise double the amount of feeds and reflectors would be required.
In the example illustrated in Fig. 1.8 each reflector only produces one type of beam. An advantage
of this configuration is that the feeds can be properly sized in order to obtain good spillover efficiency.
However, a clear downside of this method is the number of reflectors which are required. Using multiple
reflectors takes up valuable space on the satellite which could be used for other services or instruments,
and adds mass and volume which contribute to launch costs and may increase the complexity of the
satellite’s mechanical design.
B
C
A
D
B
C
A
D
Figure 1.8: Generation of multiple beams using SFB with multiple reflectors.
1.2.2 Multiple-Feeds-per-Beam
While SFB uses a single independent feed to generate each beam, another solution is to use multiple
feeds to generate each beam [3, 8, 12, 13, 14, 5, 15]. A beamforming network (BFN) is used to generate
the correct excitation at each feed element to produce the multiple beam pattern, and by using multiple
feeds to generate each beam the feeds can be shared between beams. This is illustrated in Fig. 1.10,
where one of the feeds is shared between the sets used to generate each beam. By sharing feeds between
beams they can be effectively overlapped, reducing the feed spacing beyond what is possible in SFB,
and thus reducing the angular separation of the beams. This method is called multiple-feeds-per-beam
(MFB). Some examples of feed arrays used for MFB applications are shown in Fig. 1.11.
Compared to SFB the major advantage of MFB is the reduction in the number of reflectors, as for
one frequency band the entire four-colour pattern can be generated using a single reflector. MFB can
also be advantageous in applications where a very large number of beams at large scan angles is required
[15], or at lower frequencies where the feeds and reflectors required for SFB would be bulky and heavy
[3, 5]. However, a downside is the design of the BFN, which is a complex structure consisting of many
couplers and power dividers. MFB also has some drawbacks with regard to gain performance, and it
is not always possible to use the same reflector and feed cluster for both transmit and receive [8, 10].
Using different reflectors for transmit and receive is not always a disadvantage, as they can be optimally
sized for each band [3]. Considering a receive band which is higher in frequency than the transmit band,
using a common reflector for transmit and receive will result in the reflector being oversized for receive
operation.
(a) Antenna system for the Anik F2 satellite [9] (b) CAD model of a SFB feed cluster [5]
(c) Eurostar 3000 Ka-band multi-beam mission reflectors [10]
Figure 1.9: Examples of SFB antenna systems.
BF
N
Figure 1.10: Generation of multiple beams using MFB with a single reflector.
1.2.3 Direct Radiating Arrays
Finally, multiple beam coverages can be generated using a phased array antenna without a reflector. In
satellite applications this is generally referred to as a direct direct radiating array (DRA). This gives
(a) Medusa MFB feed array [3] (b) MFB focal plane array mock-up [14]
Figure 1.11: Examples of prototype MFB feed systems.
a lot of design flexibility, as the array excitation directly controls the coverage on the ground without
having to consider effects such as spillover which are present when the array is used to feed a reflector.
However, due to the large aperture size required to generate narrow spot beams, DRA solutions require
a very large number of elements and a complex BFN to generate the required excitation. Additionally,
as multiple feeds are shared between beams the power amplifiers driving the elements need to be backed
off from their saturated power level in order to maintain adequate electrical performance, leading to a
degradation in efficiency [8].
1.3 Motivation and Thesis Goals
Of the solutions discussed to generate multiple-beam antenna patterns, SFB is the simplest as it does
not require the design of a complex BFN. However, the fact that it needs multiple reflectors to generate
the correct beam spacing is a significant drawback.
The main objective of this thesis is to investigate design techniques which could be applied to reduce
the number of reflectors required for SFB systems. Reducing the number of reflectors would eliminate
the drawbacks of SFB while retaining its simplicity, leading to a cost-effective solution for satellite
multiple-beam antennas. This will involve two studies, which will be described in the following sections.
While reflector antennas are often the de facto solution for high-gain satellite antennas, and some of
the techniques that will be investigated are derived from previous work on reflector antennas, in this
work reflectarray-based solutions will be explored as they offer a high degree of design flexibility and
are simple to manufacture. Although reflectarrays are a fairly mature technology, it is evident from the
review provided in this chapter that they are rarely used in current satellite missions. The design methods
explored in this work aim to add additional capability to reflectarrays, which would give them advantages
over traditionally-used reflector antennas, aiding their adoption into real-world space applications.
While the focus of this work is satellite communications systems implemented using SFB, the findings
could be applied in a more general sense to multiple-beam reflectarray antennas.
1.3.1 Bifocal Reflectarray

The main reason why SFB systems need multiple reflectors to generate overlapping spot beams is that
the minimum spacing between the feeds is dictated by their size, which is difficult to reduce without
increasing spillover losses, as reducing the feed aperture diameter will increase its beamwidth. However,
in this scenario it is assumed that the reflector which generates the beams is paraboloidal, and as such
the angular spacing of the beams for a given feed spacing is set by the reflector F/D ratio. There
have been numerous works in the past on design methods for reflectors and lenses which are bifocal
and have two focal points [16, 17, 18, 19]. These synthesis techniques allow for the beam scan angles of
two separate focal points to be specified, giving some additional degrees of freedom over a single focus
reflector. Taking the two foci to be feeds in a feed cluster, by controlling their scan angles it follows
that the scan angles of the entire cluster will also be affected. This characteristic could potentially be
used to reduce the beam spacing of the cluster beyond what is achievable with a single focus reflector,
without changing the reflector or feed sizes, which would allow for the four-colour beam pattern to be
generated with fewer reflectors. While bifocal reflector designs generally require a smaller sub-reflector
to be introduced, this is typically much smaller than the large main reflector, and would still result in a
reduced overall antenna volume.
In this work a ray tracing-based technique for designing bifocal shaped dual-reflector antennas [19]
will be adapted to planar dual-reflectarrays by modifying the law of reflection used in the initial synthesis
procedure to account for surfaces which have a phase gradient. We then attempt to use this technique
to reduce the angular separation between the beams generated by a feed cluster without modifying the
feeds, or their positions, in any way. The original method was used for designing shaped solid reflectors,
which are challenging to fabricate, and some solutions may not be manufacturable. Reflectarrays are a
promising alternative in this case, as to implement the synthesis results only changes to the phase/element
distributions are needed.
1.3.2 Dual-polarization Reflectarrays

As was noted previously, each of the reflectors and feed cluster pairs in Fig. 1.8 generates the beams
for one frequency/polarization combination of the four-colour scheme, which results in a four-reflector
solution. However, if one reflector could be used to generate the sets of beams for both frequencies or
both polarizations then the number of reflectors could be reduced to two, which is a fairly significant
improvement. This can be done for two orthogonal polarizations if the reflector has an independent
response depending on the polarization of the incident fields. A well-known solution for applications
using linear polarization is the dual-gridded reflector (DGR) [20], which uses two reflectors placed in
front of each other. The front reflector is polarization-selective, which allows it to reflect fields in
one polarization while transmitting orthogonally polarized fields which are reflected by the back solid
reflector. By adjusting the angles of the two reflectors independent beams in both polarizations can be
formed. Designing a reflector with an independent response for the two carrier frequencies would be
more challenging, as they are located in closely adjacent frequency bands.
While DGRs have existed for some time, they operate in linear polarization, which is not suitable
for applications requiring circular polarization. An additional drawback is that the two surfaces making
up the DGR must be positioned and angled precisely with respect to one another, which can complicate
manufacturing. A simple alternative that could be used to a DGR is a dual-polarization reflectarray.
This performs the same function, but can use a single surface, as the reflectarray elements can be
designed to have an independent response to orthogonally linearly-polarized waves. In this second study
a dual-polarization reflectarray element will be designed which has a strongly decoupled response for two
orthogonal linear polarizations. As a tangential objective, the reflectarray element design will focus on
bandwidth, as this is a well-known drawback of reflectarrays which needs to be addressed for them to be
further developed and be used in practical applications. A lot of recent work has focused on improving
the electrical characteristics of reflectarrays from the element design level, but generally the focus has
not been on dual-polarization applications.
Although a dual-polarization reflectarray-based solution could offer some advantages with regards
to ease of manufacturing, this still does not solve the problem of producing independent beams in
the orthogonal senses of circular polarization. To address this, a linear-to-circular polarizer will be
used in combination with a dual-linear polarization reflectarray. By placing a polarizer in front of the
reflectarray, assuming the use of a circularly-polarized feed the incident circularly-polarized fields will
be converted to linear polarization. The reflectarray will ideally only see linearly-polarized fields, for
which it can independently control the reflection phases of orthogonal polarizations, while the polarizer
handles the necessary conversions between linear and circular polarization. This gives a structure which
can independently control orthogonally circularly-polarized fields.
1.3.3 Thesis Goals

From the previous discussion there are a number of goals which this thesis seeks to accomplish, and in
this section they will be stated.
For the bifocal reflectarray we set out to:
1. Apply a ray-tracing procedure for synthesizing dual-reflector antennas [19] to dual-reflectarrays,

and use it to reduce the beam spacing produced by a feed cluster from that of a reference case.
2. Explore the parameters of the synthesis procedure to observe their impact on performance.
For the dual-polarization reflectarrays we aim to:
1. Design a wideband dual-polarization reflectarray unit cell which can independently control the
reflection phase of orthogonally linearly-polarized waves.
2. Extend the operation of the linearly-polarized reflectarray unit cell to dual-circular polarization
by using a linear-to-circular polarizer.
3. Design two proof-of-concept reflectarray prototypes which are each capable of generating two in-
dependent beams in orthogonal polarizations. One prototype will operate in linear polarization
and the other in circular polarization.
Wherever possible, results will be validated experimentally through measurements. If this is not
feasible, then full-wave simulation results will be incorporated for validation.
1.4 Thesis Outline

This thesis is organized as follows: In Chapter 2 some theoretical background for the other chapters will
be provided, which includes the basic operation and design of reflectarrays, and analysis methods for
antenna arrays, periodic structures, and reflectarrays. Some definitions will also be covered. Chapter
3 will cover the use of reflectarray synthesis techniques to compress the beam spacing of a reflectarray
antenna system fed by a feed cluster. The main technique is based on a ray-tracing procedure used
to design shaped curved reflectors, but it will be applied to reflectarrays. The results of a tangential
investigation into using genetic algorithm based optimization will also be presented. In Chapter 4, the
design of dual-polarization reflectarrays will be presented. This will start with a literature overview of
dual-polarization reflectarray elements, and the design of the element used in this work. This reflectarray
element will be combined with a linear-to-circular polarizer to form a structure which can independently
control orthogonally circularly-polarized waves. The design of two prototypes will be presented, one
operating in linear polarization and the other in circular polarization, and methods of improving per-
formance will be considered. Both computational and measurement results will be provided. Finally,
Chapter 5 concludes this thesis and suggests avenues through which this work could be improved and
extended in the future.
Chapter 2
Background
In this chapter an overview of the theory behind the operation of reflectarrays and their analysis will
be given, which will be necessary for understanding the later chapters. This includes an overview of
antenna arrays, of which reflectarrays can be considered a subset, and periodic structures. Conventions
and performance metrics used in this thesis will be defined. Additionally as both one-dimensional and
two-dimensional reflectarrays will be explored in the later chapters, the methods of analyzing both will
be described.
2.1 Antenna Array Background
An antenna array is simply an arrangement of multiple antennas. First let us consider one element located
at position r0 , as shown in Fig. 2.1. Its electric field at an observation location r can be expressed as
0
e−jk|r−r |
E (r) = E0 (r) , (2.1)
|r − r0 |
where E0 (r) is a factor that determines the polarization and magnitude of the element’s fields as a
function of position, and k is the wavenumber of the medium that it is surrounded by. The distance
term can be re-written as
p
|r − r0 | = (r − r0 ) · (r − r0 )
p
= r2 + r02 − 2r · r0
s (2.2)
0 2 0
r r
=r 1+ − 2r̂ · r̂0 .
r r
Note that the term r̂ · r̂0 depends only on the spherical coordinate angles and not the magnitude of the
radial components r or r0 . Expanding the square root term in a Taylor series about r0 /r = 0 then gives
s 0 2
r
0
r
0
r 1h i r 0 2
0 0 0 2 (2.3)
1+ − 2r̂ · r̂ = 1 − r̂ · r̂ + 1 + (r̂ · r̂) + ··· ,
r r r 2 r
16
Chapter 2. Background 17
where only the first two terms of the series expansion have been shown. Substituting (2.3) into (2.2)
yields
r02 h 2
i
|r − r0 | = r − r̂ · r0 + 1 + (r̂0 · r̂) + · · · . (2.4)
2r
For antenna problems the region of interest is typically far away from the source, such that r r0 . Then
the second order derivative term in (2.4) can be neglected. Returning to (2.1), using the first term of
(2.4) for the magnitude variation and the first two terms for the phase variation gives
0 0
e−jk|r−r | e−jk(r−r̂·r )
≈ , (2.5)
|r − r0 | r
which is called the far field approximation. Substituting (2.5) into (2.1) and splitting up the complex
exponential term then gives
e−jkr jkr̂·r0
E (r) ≈ E0 (r) e . (2.6)
r
where it is assumed that E0 (r) does not vary with the position of the element.
θ r − r0
φ
r0
y
∆x
x
∆y
Figure 2.1: A planar array of antenna elements arranged in a rectangular grid.
Now consider a configuration of N elements, each weighted by its own amplitude and phase coefficient
Ii = ai ejψi . Neglecting coupling between the elements and applying (2.6), the fields from all the elements
can be obtained through superposition as
N
e−jkr X jkr̂·r0i
E (r) = E0 (r) Ii e , (2.7)
r i=1
where the first term has been factored out as it does not change with the summation index. The subscript
on r0i indicates that the element positions are discrete. (2.7) shows that the total fields from the array
can be obtained as a product of two terms, a vector element factor EF and a scalar array factor (AF) as
E (θ, φ) = EF (θ, φ) · AF (θ, φ) , (2.8)
where
e−jkr
EF (θ, φ) = E0 (r) (2.9a)
r
N
X 0
AF (θ, φ) = Ii ejkr̂·ri . (2.9b)
i=1
This is called the principle of pattern multiplication, where the element factor is a function of the
type of element used in the array, and the array factor depends on the geometry and excitation of the
array [21]. The element factor determines the polarization characteristics of the array, whereas the
array factor can increase the directivity of the array far beyond what a single element can achieve,
and determines the shape of the overall radiation pattern. Knowing that in the most general case that
r̂ = x̂ sin θ cos φ + ŷ sin θ sin φ + ẑ cos θ and r0i = x̂x0i + ŷyi0 + ẑzi0 , the array factor can be expressed as
N
X 0
AF (θ, φ) = ai ejψi ejkr̂·ri
i=1
(2.10)
N
x0i cos φ+yi0 sin φ+zi0
ai ej (ψi +k[ cos θ ])
X
sin θ sin θ
= .
i=1
A common application of an antenna array is to produce a pencil beam in a specified direction. This can
be done if the waves travelling from each array element to a plane normal to the desired scan direction
add constructively in phase, forming a plane wave. From (2.10) this is achieved when the argument of
the complex exponential goes to zero, giving the scan condition for a beam in the (θ0 , φ0 ) direction as
ψi = −k (x0i sin θ0 cos φ0 + yi0 sin θ0 sin φ0 + zi0 cos θ0 ) . (2.11)
Now consider the case of a planar array in a coordinate system that has its x and y axes aligned with
the surface of the array, such that zi0 = 0. From (2.10) the AF of this configuration is given by
N
0 0
ai ej (ψi +kx xi +ky yi ) ,
X
AF (θ, φ) = (2.12)
i=1
where
kx = k sin θ cos φ (2.13a)

ky = k sin θ sin φ. (2.13b)
Let us further assume that the elements are placed in a rectangular grid with spacing ∆x in the x-
direction and ∆y in the y-direction. This geometry is shown in Fig. 2.1. Then introducing two new
indices m and n, the positions of each element are given by x0i = m∆x and yi0 = n∆y. If necessary,
an appropriate shift of coordinates can be made to account for the origin of the array. Splitting the
summation of (2.12) such that there are M elements in the x-direction and N elements in the y-direction,
and absorbing the element phase and amplitude into an excitation coefficient Imn = amn ejψmn then gives
M X
X N
AF (θ, φ) = Imn ej(kx m∆x+ky n∆y) . (2.14)
m=1 n=1
The double summation term has the form of a two-dimensional inverse discrete Fourier transform (IDFT).
For a discrete function f (p, q) which has a Fourier representation F (m, n), the discrete Fourier transform
pairs are given by
M X
X N
F (m, n) = f (p, q)e−j2π(mp/M +nq/N ) (2.15a)
p=1 q=1
M N
1 XX
f (p, q) = F (m, n)ej2π(mp/M +nq/N ) . (2.15b)
M N m=1 n=1
Then, by comparing (2.15b) and (2.14), the array factor can be computed using the IDFT as
AF (kx , ky ) = M N · IDFT2 (Imn ) , (2.16)
where IDFT2 is the 2D IDFT. Taking (p, q) to be the indices of the 2D IDFT result, these indices can
be related to the spatial frequencies kx and ky by comparing (2.15b) and (2.14):
2πp
kx = (2.17a)
M ∆x
2πq
ky = . (2.17b)
N ∆y
kx and ky can then be related back to the spherical coordinate angles using (2.13), where the conventional
pattern cuts can then be taken. This formulation is advantageous, as there are algorithms available to
quickly compute the IDFT. To obtain better resolution in the computed patterns, M and N do not have
to be the number of elements in the array, and can be increased by zero-padding the array excitation, as
is often done when taking the discrete Fourier transform (DFT) of one-dimensional time-domain signals.
2.2 Reflectarray Background and Design Principles
There are numerous ways to look at how a reflectarray works, and even more variations on how to
design them. In this section the two methods that will be explored are looking at the reflectarray as a
discrete array of phase shifters, similar to the antenna arrays previously discussed, and as a surface with
non-constant phase variation that can redirect incident plane waves to a non-specular reflection angle.
Dual-reflectarrays will also be introduced, in which the reflectarray is fed by another reflectarray rather
than being fed directly by the feed, as they will be explored in Chapter 3.
2.2.1 Phase Shift Approach
A reflectarray is an array of scatterers that has been engineered to produce a desired radiation pattern
when fed spatially by a feed antenna. Knowing the incident fields produced on the reflectarray from
the feed, and the desired radiation pattern, the required aperture phase distribution of the reflectarray
can be determined. This phase distribution can then be implemented by designing each element in
the reflectarray to scatter the incident field with an appropriate phase shift. In most applications the
scattering elements in the reflectarray are printed microstrip based structures which are backed by a
planar grounded dielectric, as this is simple to manufacture. However, this does not have to be the
case and alternatives are possible. For example the first published reflectarray was made up of short
circuited rectangular waveguides [22], reflectarray elements have been fabricated using three-dimensional
structures such as a perforated dielectric [23] or machined metal [24], and non-planar reflectarrays made
up of panels [25] or a doubly curved surface [26] have been demonstrated.
A cross-section of a planar reflectarray which is designed to produce a pencil beam is shown in
Fig. 2.2. In one sense reflectarrays can be viewed as antenna arrays as described in the previous section,
but without the elements being driven directly by a transmission line based feed network. In another
sense, a reflectarray can be viewed as having the same functionality as a reflector antenna: when fed from
a source of spherical waves they both equalize the differing path lengths from the source, collimating
the rays to form a plane wave. However, paraboloidal reflectors use their curvature to compensate
the physical distance, or time delay, that each ray must travel. Reflectarrays equalize the electrical
path lengths of the rays, which introduces frequency dependence, and reflectarrays are thus narrowband
by nature. Techniques have been investigated to increase their bandwidth, which will be described in
Chapter 4.
In Fig.2.2 a horn antenna is shown as the feed for the reflectarray. Horn antennas are popular as feeds
for spatially fed apertures as they produce a tapered illumination, and can be designed to support large
bandwidths. The former allows for losses resulting from power radiated by the feed, but not intercepted
by the aperture, to be reduced. This loss is known as spillover and will be discussed later in this chapter.
In order to produce a pencil beam in the direction (θ0 , φ0 ) with unit vector r̂0 , the reflectarray must
impose a constant phase delay on each ray traced from the focal point to a plane normal to r̂0 . Tracing
the rays for one element in Fig. 2.2, at the planar wavefront the total phase delay amounts to the path
travelled from the focal point to the reflectarray, the reflection phase incurred from the reflectarray
element, and the path from the reflectarray to the wavefront. This yields the condition
−k (|r0 − F|) + ψi − k |r0 | = C + 2πN, (2.18)
where k is the wavenumber, F is the position vector of the focal point, r0 is the position vector of the
reflectarray element, and r0 is the vector from the reflectarray element to one of the planar wavefronts.
C is an arbitrary constant, and N is an integer. An integer multiple of 2π can be added, as the phase
will be mathematically the same. It is visible from Fig. 2.2 that the magnitude of the output ray can
be expressed as
|r0 | = a + b, (2.19)
where a is the distance from each reflectarray element to a reference wavefront intersecting the top of
b
a
x Planar
wavefronts
r00
r0
r0 − F
r0
Reflectarray
Feed horn
F
z
y
Figure 2.2: The geometry of an offset pencil beam reflectarray, illustrating the conversion from a spherical
wave to a plane wave.
the reflectarray, located at r00 , and b is the distance from this wavefront to one of the others. After the
rays from the feed have been collimated they will be parallel, and perpendicular to the wavefronts, and
so b is a constant. However, a changes with position, and is the extra distance travelled by each reflected
ray to the outgoing wavefront in the r̂0 direction. To produce a plane wave the reflected rays will be
parallel, and so a can be expressed as the component of the displacement vector along the reflectarray
surface (r00 − r0 ) in the outgoing ray direction as
a = (r00 − r0 ) · r̂0 = r00 · r̂0 − r0 · r̂0 , (2.20)
where the term r00 · r̂0 is a constant. Substituting (2.19) and (2.20) into (2.18) then gives
−k (|r0 − F|) + ψi − k (r00 · r̂0 − r0 · r̂0 + b) = C + 2πN. (2.21)
After solving for the required reflection phase of each element we arrive at
ψi = k (|r0 − F| − r0 · r̂0 ) + ψ0 + 2πN, (2.22)
where ψ0 is an arbitrary constant and the result of lumping all of the constant terms in (2.21) together.
(2.22) gives the required phase distribution of the reflectarray to produce a pencil beam when fed by a
point source. Note that this is the same condition as required for an array with driven elements, with
the extra stipulation that the reflectarray elements must also compensate the phase incurred from the
different path lengths from the feed to the reflectarray.

Depending on the application, a radiation pattern other than a pencil beam may be required. Then
the reflectarray phase distribution will differ from that imposed by (2.22), and will involve array synthesis
or optimization techniques.
2.2.2 Phase Derivative Approach
An alternative approach to describing the principle of operation of a reflectarray and synthesizing its
phase distribution involves modifying the law of reflection to account for the phase shift as a function
of position on the reflectarray’s surface. This has been applied to synthesize a shaped beam reflectarray
in which both the phase and amplitude distribution can be controlled [27], and to produce anomalous
reflection and refraction of infrared radiation as well as generate different kinds of beams [28]. It will
also be used to synthesize the reflectarray phase distributions in Chapter 3.
dr
θi θr
di
n̂ n̂ θr n̂ n̂
θi
ψ1 ψ2 ψ1 ψ2
∆x ∆x
(a) Incident ray (b) Reflected ray
Figure 2.3: Cross-section of a reflectarray directing rays from an incident plane wave to a non-specular
direction.
The cross-section of a reflectarray illuminated by a plane wave is shown in Fig. 2.3. In this one-
dimensional analysis the elements are spaced by ∆x, and considering two adjacent elements they impart
phase shifts on the incident rays of ψ1 and ψ2 . The incident plane wave arrives at an angle of θi , and
the reflectarray is designed to reflect it at an angle θr . Tracing the ray paths for two adjacent elements,
it is visible that the incident rays must travel an extra distance di to the element on the right, and the
reflected rays must travel an additional distance dr from the element on the left. From Fig. 2.3 these
distances are
di = ∆x sin θi (2.23a)
dr = ∆x sin θr . (2.23b)
For the reflected wave to have a planar wavefront, the electrical distance travelled by the two rays must
be the same, and so these differences in path length must be compensated by the reflectarray. This gives
the condition
ψ1 − kdr = ψ2 − kdi
(2.24)
ψ1 − k∆x sin θr = ψ2 − k∆x sin θi ,
which when re-arranged yields
ψ2 − ψ1
k (sin θi − sin θr ) = . (2.25)
∆x
Taking the limit as ∆x → 0 the term on the right hand side of (2.25) becomes the derivative of phase with
respect to position. After some re-arranging this gives an alternative design equation for the reflectarray
phase distribution:
1 dψ (x)
sin θi − sin θr = . (2.26)
k dx
If the incident ray directions are known and the output ray directions have been specified, the required
phase derivatives across the reflectarray surface can be determined from (2.26) and then integrated
to determine the required phase distribution. This formulation can be extended to two-dimensional
reflectarrays using the relations [27]
1 ∂ψ (x, y)
sin θr cos φr + sin θi cos φi = − (2.27a)
k ∂x
1 ∂ψ (x, y)
sin θr sin φr + sin θi sin φi = − , (2.27b)
k ∂y
where now the y direction must also be considered, and so the total derivatives have become partial
derivatives. (θi , φi ) and (θr , φr ) are the angles of incidence and reflection, respectively.
2.2.3 Dual Reflectarrays

Based on the application, a reflector antenna system may be composed of multiple reflectors. In particular
the addition of a second reflector can allow the feed and its associated active components to be placed in
a more convenient location, make the antenna system more compact, or add more degrees of freedom to
improve performance. The classical microwave dual-reflector antenna geometries, such as Cassegrain and
Gregorian reflectors, are based on optical techniques used for telescopes [29]. In these geometries all of
the reflectors are derived from conic sections, yielding reflector shapes such as paraboloids, hyperboloids,
and ellipsoids. As it is possible to design a reflectarray which emulates a paraboloidal reflector over a
limited frequency range by equalizing the electrical path lengths of the rays from the feed rather than
their physical length, it is also possible to develop planar analogues to the other components of curved
dual-reflector antennas.
The cross-section of a dual-reflectarray geometry is illustrated in Fig. 2.4. The geometry is similar
to that of Fig. 2.2 except that a sub-reflectarray is added, and now the feed of the antenna system faces
the +z- rather than the −z-direction. The system has both a real focal point, located at FR , and a
virtual focal point located at FV . The virtual focal point is a consequence of the sub-reflectarray, which
reflects the rays from the real focus in a manner such that they appear to be coming from another point
source. The electrical characteristics of the sub-reflectarray are determined by the distances to the real
and virtual foci. Given the locations of the two foci and the position vectors of the sub-reflectarray
elements, the required phase of each subreflectarray element ψs,i can be determined as [30]
ψs,i = k (|r0s − FR | − |r0s − FV |) . (2.28)
Main
reflectarray
Planar
wavefronts
RV
r0s RR FV
FR
z Sub-
y Real feed reflectarray
Virtual feed
Figure 2.4: A dual-reflectarray antenna transforming a spherical wave to a plane wave.
Depending on the placements of the real and virtual foci relative to the sub-reflectarray, it can have
various shapes. If the virtual focus is closer than the real focus, the sub-reflectarray phase distribution
will be convex as in a Cassegrain reflector system, and the rays from the virtual focus will be more
spread than those from the real focus. If the virtual focus is further away from the sub-reflectarray than
the real focus the sub-reflectarray phase will be concave, and the rays from the virtual focus will be
bundled more tightly than those from the real focus. In another case the virual focus can be placed on
the same side as the real focus, forming a Gregorian reflector/reflectarray antenna [29, 30] where the
sub-reflectarray phase distribution is ellipsoidal.
In all of these cases, the sub-reflectarray is performing a transformation from one spherical wave
source to another. Thus, the main reflectarray can be designed as a typical pencil beam reflectarray
using (2.22) where its focal point is the virtual focus of the dual-reflectarray antenna. This is perhaps the
simplest way to perform dual-reflectarray synthesis, where the two reflectarrays are treated separately
and the sub-reflectarray simply shifts the focal point. This is commonly done to reduce the length of
the reflector system from that of the equivalent single reflector configuration. Other synthesis methods
will be explored in Chapter 3.
2.3 Analysis of Reflectarray Elements and Periodic Structures

In the previous section, the behaviour of reflectarrays and methods to synthesize their phase distributions
were introduced. However, reflection based phase shifters must be designed to actually construct the
reflectarray, which requires characterization of the scattering elements used. This is rooted in the analysis
of periodic structures, which will be given an overview in this section, along with methods to analyze
multi-layer structures, and conversions necessary to design the structures presented in Chapter 4.
2.3.1 Element Modelling and Phase Shift Mechanism

There are multiple steps involved in the design of reflectarrays, which vary in complexity. The traditional
approach is to compute the phase required as a function of position on the reflectarray surface, either
using (2.22) to produce a pencil beam, or optimization based or array synthesis methods to produce
other kinds of patterns. The phase distribution must then be realized, which requires knowledge of how
the reflectarray elements can be varied in order produce a change in reflection phase.
Reflectarray elements are typically characterized by simulating each unit cell in a periodic environ-
ment illuminated by an incident plane wave Ei . The unit cell is one element of a structure that is
repeated infinitely in a periodic manner across a plane. Periodic simulations are generally not very
computationally intense and can be performed using standard desktop computers, unless the element
structure is complex or has a high quality factor resonance. They also to some degree take into account
mutual coupling between elements, and can be verified experimentally. For example, unit cells can be
placed in a rectangular waveguide which approximates the infinite array assumption through image the-
ory, or quasi-optical measurements can be performed which measure the scattering response to incident
plane waves of a sample containing many repeated unit cells.
Considering one unit cell of the overall structure illuminated on both sides by plane waves, as pictured
in Fig. 2.5, the scattered fields can be obtained in terms of the incident fields as
    
TE TE,TE TE,TM TE,TE TE,TM TE
Es,1 S11 S11 S12 S12 Ei,1
 TM   TM,TE TM,TM TM,TE TM,TM   TM 
Es,1  S11 S11 S12 S12  Ei,1 
 E TE  =  S TE,TE TE,TM   TE  , (2.29)
    
TE,TM TE,TE
 s,2   21 S21 S22 S22   Ei,2 
TM TM,TE TM,TM TM,TE TM,TM TM
Es,2 S21 S21 S22 S22 Ei,2
where the the subscripts on the field terms i and s indicate the incident and scattered fields, respectively,
and the numerical subscript refers to the side of the unit cell, or the port. The matrix relating the incident
and scattered fields is the generalized scattering matrix (GSM) of the unit cell, and the superscripts
indicate the transverse electric (TE) and transverse magnetic (TM) polarizations, as indicated in Fig. 2.5.
While the TE and TM polarizations are commonly used, any other orthogonal polarization basis could
be used to characterize the response of the unit cell, such as slant polarization or circular polarization
(CP). The GSM of the element is analogous to the scattering parameters (S-parameters) used to relate
the incident and reflected travelling or power waves between the ports of a microwave circuit.
While many kinds of truly periodic structures exist, such as frequency selective surfaces or polar-
ization converters, it is clear that in order to produce a characteristic that changes with position that
all of the elements in a reflectarray cannot be the same. When the reflectarray elements are analyzed
as being periodic, when in reality they will not be, this is called a quasi-periodic or local periodicity
assumption. This assumption is commonly used when designing structures of this type, such as meta-
Ei,2
TM
Port 2
Es,2
TM
TE
Ei,1
TE Unit cell
Es,1
x
Port 1
y
Figure 2.5: Analysis of a unit cell using periodic boundary conditions.
surfaces, transmitarrays, and lenses. For reflectarrays the local periodicity assumption has been found
to not be a significant source of error, as was demonstrated in a recent work where the computed and
measured radiation patterns of a reflectarray which was designed to have an exaggerated lack of period-
icity were compared [31]. In addition to neglecting the variations in the elements making up reflectarray,
the quasi-periodic assumption also neglects the fact that the reflectarray is finite. Effects such as edge
diffraction are ignored. However, in a typical reflectarray application the illumination level at the edge
of the reflectarray is intentionally low, which reduces the edge effects. Considerations can be made
to design reflectarrays in a way such that in the actual application they more closely approximate the
conditions under which the individual elements where characterized.
There are several ways to implement the periodic boundary conditions required for simulation. One
method is to surround the unit cell with alternating perfect electric conductor (PEC) and perfect mag-
netic conductor (PMC) boundaries, where the cell is repeated infinitely by image theory. Modern
computer aided design (CAD) tools also allow for periodic boundaries to be implemented. The second
method is used in this work using Ansys HFSS, where the boundaries are set up as master/slave and
Floquet ports are used as the excitation.
While Fig. 2.5 illustrates the unit cell being excited by normally incident plane waves, the angle of
incidence can be varied. Fig. 2.6 shows how the angles of incidence and plane of incidence are defined in
this thesis. As the reflectarray is fed by a spherical wave source, the angle of incidence of the rays from
the feed will vary across the reflectarray surface and this will produce deviations in the characteristics
from the normally incident response. In order to simulate these effects, the angle of incidence can be
varied when characterizing the unit cell. To reduce the number of variations that need to be simulated,
the reflectarray can be divided into discrete zones by angle of incidence rather than considering the
angles of incidence of each individual element. Another approach is to design the unit cells to be highly
stable with angle of incidence to avoid these effects, although accounting for them will increase accuracy.
In this work the normally incident characteristics are used unless otherwise specified. For reflectarrays
TE
θi
TM
φi
y
Figure 2.6: The conventions used in this thesis for the plane and angles of incidence.
using an F/D ratio of one or greater the variation in angle of incidence will not be very high, and so in
these cases it may be acceptable to design the reflectarray assuming normal incidence.
It should be noted that the matrix entries in (2.29) are in general matrices themselves. A plane
wave incident on an infinite periodic structure acts as a periodic excitation, the solution to which is
an infinite series of plane waves [32, 33]. Depending on the angle of arrival of the incident plane wave
and the periodicity of the elements, the scattered plane waves will either propagate or be evanescent.
These scattered plane waves are referred to as Floquet modes TEmn and TMmn , where m and n are the
mode indices. The fundamental modes TE00 and TM00 propagate in the specular direction. Typically
in reflectarray design the element periodicity is chosen such that only the fundamental mode propagates,
as higher order propagating modes correspond to grating lobes and will produce radiation in undesired
directions. In this thesis only the fundamental modes are considered, and so the terms in the GSM
of (2.29) are all complex-valued scalars. However, higher-order modes are relevant for larger element
spacings, or if the fields of interest are in a region close to the surface of the infinite array where the
evanescent modes have not decayed enough for their effects to be negligible, such as when multiple
periodic structures are placed in close proximity.
While (2.29) deals with the case where the periodic structure is excited from both sides and both
reflection and transmission are considered, reflectarrays are typically backed by a ground plane, which
will not transmit any fields assuming the ground plane is infinite. Thus, for reflectarrays it is enough to
consider only a one-port reflection GSM which takes the form
" # " #" #
TE,TE TE,TM
EsTE S11 S11 EiTE
= TM,TE TM,TM . (2.30)
EsTM S11 S11 EiTM
As an example, a rectangular patch, as shown in Fig. 2.5, with half-wavelength periodicity at 20 GHz
backed by a ground plane was simulated in a one-port periodic environment in HFSS. The conductors
were modelled as PEC sheets, and the dielectric has a thickness of 0.787 mm and the electrical properties
r = 3 and tan δ = 0.0013. The TM co-polar reflection phase as the patch width is varied symmetrically
is shown in Fig. 2.7. It is visible from Fig. 2.7a that changing the patch width shifts its resonant
frequency, which is the frequency where its reflection phase is 0◦ . Varying a dimension of the element,
as illustrated in this example, is a common phase shift mechanism used in reflectarray design. When
the reflection phase is plotted against the characteristic which produces its variation, as in Fig. 2.7b, the
resulting curve allows for reflectarray element variations to be assigned based on the required phase at
each position in the reflectarray.
200
200
3.0 mm
3.2 mm
100
Phase [deg.]
3.4 mm 100
Phase [deg.]
3.6 mm
3.8 mm
0 4.0 mm 0
4.2 mm
4.4 mm
-100 4.6 mm -100
4.8 mm
-200
16 18 20 22 24 26 -200
1 2 3 4 5 6 7
Frequency [GHz]
Width [mm]
(a) Variation with frequency for some selected patch
(b) Variation with length at 20 GHz
dimensions
Figure 2.7: The simulated reflection characteristics of the square patch illustrated in Fig. 2.5 at normal
incidence.
Although the phase shift mechanism shown in the example of Fig. 2.7 involved varying the patch
geometry, other mechanisms include loading the element with continuously or discretely tunable lumped
components, or rotating the element. A more detailed review of different kinds of reflectarray unit cells,
their mode of operation, and design will be given in Chapter 4.
2.3.2 Multi-layer Structures

The situation often arises where a periodic structure has multiple layers. Multi-layer structures can pro-
vide more degrees of freedom in the design procedure, or can be used to decouple different functionalities
to allow for a more modular design approach to be used. A case of a three-layer structure is illustrated
in Fig. 2.8, where each structure is separated by an air gap. This does not have to be the case, and the
air gaps could have zero length in the case that all the surfaces have been bonded or fastened together.
To analyze a multi-layer structure a cascade method can be used. This approach comes from the
analysis of microwave circuits where it is common to characterize each block or component in a circuit
individually using scattering parameters. The characteristics of the full circuit can then be determined
by cascading the S-parameters of each component. If two structures have the GSMs SA and SB , then
they can be cascaded to determine their combined effect using the relations [33]
B A −1 B A
SAB A A

11 = S11 + S12 I − S11 S22 S11 S21 (2.31a)
−1
SAB A B A
SB

12 = S12 I − S11 S22 12 (2.31b)
−1
SAB B A B
SA

21 = S21 I − S22 S11 21 (2.31c)
AB B B A B −1 A B

S22 = S22 + S21 I − S22 S11 S22 S12 . (2.31d)
S3 Ei,2
SAG
S2 Port 2
SAG
S1
Es,2
Ei,1
d2
d1
Es,1
x
Port 1
y
Figure 2.8: Analysis of a multi-layer periodic structure.
In the example shown in Fig. 2.8 the three surfaces have the known GSMs S1 , S2 , and S3 , and are
separated by air gaps which are characterized by SAG . By applying (2.31) multiple times to account for
each of the surfaces and the air gaps, the response of the full multi-layer structure can be determined.
For the case when only the fundamental modes are considered, the size of each term in (2.31) is 4 × 4.
As noted previously, in general when higher-order Floquet modes are taken into account these terms
will be matrices whose size depends on the number of modes considered. However, if the surfaces are
spaced far enough apart the effects of higher-order mode coupling will be small and can be neglected.
In order to determine whether the distance between the structures is large enough for this assumption
to be valid, the attenuation factor of each mode can be computed from its complex wavenumber.
The GSMs of some common structures can be used as building blocks to construct a multi-layer
structure. The air gaps in between the surfaces in Fig. 2.8 can be treated as a dielectric slab which has
the GSM [33]
 
0 0 e−jk0 d 0
e−jk0 d 
 
 0 0 0
SAG =
e−jk0 d
, (2.32)
 0 0 0  
0 e−jk0 d 0 0
where k0 is the free-space wavenumber and d is the size of the air gap between the surfaces. Although
an air gap is considered here, the wavenumber can be modified to account for the permittivity of the
material between the structures if a dielectric is used. By treating the space between the layers of the
structure in this manner, this region does not have to be discretized and the fields in it do not have to
be solved for. This can offer a speed advantage for simulations if the spacing between layers is large, or
if the overall structure has many layers.
Another case of interest are the S-parameters of a ground plane. This situation arises if a ground
plane is to be added to a structure, or if the one-port GSM of a unit cell with a ground plane has been
determined and the two-port GSM is needed for it to be cascaded with another structure. The two-port
GSM of a unit cell backed by a ground plane on its second port will have the form
" # " #" #
Es,1 S11 T12 Ei,1
= , (2.33)
Es,2 T21 ΓGP Ei,2
where S11 is the one-port reflection matrix of the unit cell, which is backed by a ground plane, T12
and T21 the transmission matrices between Port 1 and Port 2, and ΓGP is the refection matrix of a
perfectly conducting ground plane. Taking the ground plane to be infinite and homogeneous, applying
the boundary conditions for the tangential electric field yields
" #
0 0
T12 = T21 = (2.34a)
0 0
" #
−1 0
ΓGP = . (2.34b)
0 −1
Now the full 4 × 4 GSM of the structure is known and it can be used in the cascade analysis. If only a
ground plane by itself is considered, S11 in (2.33) can be replaced by ΓGP .
An alternative analysis method to cascading the GSMs of each layer in the structure is to form a
unit cell which contains all of the layers, which will be termed a ‘supercell’. For the case illustrated in
Fig. 2.8 where all of the layers have the same periodicity the supercell can simply be formed by placing
ports on each end of the full structure and surrounding it with periodic boundaries. However, for cases
where the cell periods are not the same constructing the supercell becomes more complicated.
Two other cases are illustrated in Fig 2.9 where two periodic structures are considered which do
not have the same periodicity. One surface is made up of rectangular loops and the other of larger
rectangular patches. In Fig 2.9a the cell period of the loop surface is a quarter of that of the patch
surface. The cell periods of the two layers can be made equal by choosing the supercell period to be the
least common multiple of the cell periods of each layer. In this case this can be done by including four
of the loop unit cells rather than one. As the surface is periodic and infinite, this is essentially the same
as including one element in the unit cell, and the supercell can be constructed as before.
In Fig. 2.9b the periodicities of the two surfaces are not integer multiples of one another, and addi-
tionally the lattice of the loop surface is rotated by 45◦ with respect to that of the patch surface. In
this case it is challenging and perhaps not possible to find a supercell period which contains an integer
multiple of elements for both surfaces. Instead the supercell period can be set such that some of the
loop elements are truncated, as shown in Fig. 2.9b. If this unit cell is repeated it will form the same loop
surface as before truncation. However, care must be taken to ensure that the unit cell with truncated
elements actually represents the periodic surface.
An advantage of the supercell method is that it takes into account all of the interactions between
the surfaces. However, a drawback is that multiple elements need to be included from some or all of
the layers if their periodicities are different, and the air gaps between the structures become part of
the simulation domain. This will result in increased simulation times and computational requirements
compared to the cascading method. Additionally, the supercell is less modular, as if any changes are
Ei,2
Port 2
Es,2
Ei,1
x
Es,1 y
Port 1
z
(a)
Ei,2
Port 2
Es,2
Ei,1
x
Es,1 y
Port 1
z
(b)
Figure 2.9: The supercell method of analyzing multi-layer periodic structures. In (a) the loop surface
has a periodicity which is a quarter of that of the patch surface. In (b) the cell period of the loop surface
is not an integer multiple of the patch surface, and their lattices are rotated with respect to one another.
made to any of the layers the entire supercell must be re-simulated. Thus, in this work the cascade
method will be used to analyze multi-layer structures.
2.3.3 Rotation of One-port Scattering Parameters

It can sometimes be convenient to obtain the scattering matrix of a unit cell in one basis, and then use
these results to calculate the response in a rotated basis. For example, in a cascade of several surfaces
which are rotated with respect to one another, the surfaces may be designed in their own coordinate
systems. However, in the full structure the rotated S-parameters are required. In this manner the
surfaces may be used by themselves, or combined in a modular manner with other surfaces, without
requiring multiple sets of simulations. This will be applied in Chapter 4 when a reflectarray is integrated
with a linear-to-circular polarizer.
TM0
TM
TE0
γ
TE
Figure 2.10: Example unit cell with the slant polarization convention shown.
Consider the case that the one-port GSM has been determined with respect to the TE0 and TM0
polarizations as
" 0
# " 0 0 0 0 #" 0
#
TE ,TE TE ,TM
EsTE S11 S11 EiTE
0 = 0
TM ,TE 0 0
TM ,TM 0 0 . (2.35)
EsTM S11 S11 EiTM
However, the response to the TE and TM polarizations is desired, which are rotated by an angle γ from
their primed counterparts. The polarizations are illustrated in Fig. 2.10, where the primed polarizations
are aligned with an example dipole-type element. The unit vectors of the two sets of polarizations can
be expressed in terms of each other as
âTE = âTE0 cos γ − âTM0 sin γ

(2.36)
âTM = âTE0 sin γ + âTM0 cos γ
âTE0 = âTE cos γ + âTM sin γ
(2.37)
âTM0 = −âTE sin γ + âTM cos γ.
(2.37) can be used to transform the incident fields in (2.35) to the un-primed polarization basis, and the
resulting scattered fields can be transformed to the un-primed system using (2.36) to give
" # " #" 0 0 0 0 #" #" #
TE ,TE TE ,TM
EsTE cos γ − sin γ S11 S11 cos γ sin γ EiTE
= TM ,TE 0 0 0
TM ,TM 0 , (2.38)
EsTM sin γ cos γ S11 S11 − sin γ cos γ EiTM
which is essentially the result of applying rotation matrices. When simplified (2.38) gives expressions
for the S-parameters in the un-primed polarization basis as
TE,TE TE0 ,TE0 TM0 ,TM0 1

TE0 ,TM0 TM0 ,TE0

S11 = S11 cos2 γ + S11 sin2 γ − sin 2α S11 + S11
2
TE,TM 1
TE0 ,TE0 TM0 ,TM0

TE0 ,TM0 TM0 ,TE0
S11 = sin 2α S11 − S11 + S11 cos2 γ − S11 sin2 γ
2 (2.39)
TM,TE 1
TE0 ,TE0 TM0 ,TM0

TE0 ,TM0 TM0 ,TE0
S11 = sin 2α S11 − S11 − S11 sin2 γ + S11 cos2 γ
2
TM,TM TE0 ,TE0 TM0 ,TM0 1
TE0 ,TM0 TM0 ,TE0

S11 = S11 sin2 γ + S11 cos2 γ + sin 2α S11 + S11 .
2
For the special case when γ = 45◦ , which will be of interest in Chapter 4,
TE,TE 1 TE0 ,TE0 TM ,TM 0 0

TE ,TM TM ,TE 0 0 0 0

S11 = S + S11 − S11 − S11
2 11
TE,TM 1 0 0 0 0 0 0 0 0

S11 = S TE ,TE TM ,TM
− S11 TE ,TM
+ S11 TM ,TE
− S11
2 11 (2.40)
TM,TE 1 TE0 ,TE0 TM0 ,TM0 TE0 ,TM0 TM0 ,TE0

S11 = S − S11 − S11 + S11
2 11
TM,TM 1 0 0 0 0 0 0 0 0

S11 = S TE ,TE TM ,TM
+ S11 TE ,TM
+ S11 TM ,TE
+ S11 .
2 11
2.3.4 Circularly Polarized S-parameters
Er RHCP
Er LHCP
Ei RHCP
Ei LHCP
TM
x
z TE
y
Et RHCP
Et LHCP
Figure 2.11: Transmitted and reflected circularly-polarized waves normally incident on an interface,
illustrating the circular polarization conventions used in this thesis.
When working with structures that convert between linear and circular polarization it is necessary to
know the relations between the two polarization bases. The conventions used in this thesis for circular
polarization are illustrated in Fig 2.11, where the right hand circular polarization (RHCP) and left
hand circular polarization (LHCP) basis vectors can be defined in terms of the TE and TM polarization
vectors as:
1
âR,i = âR,t = √ (âTM − jâTE ) (2.41a)
2
1
âL,i = âL,t = √ (âTM + jâTE ) (2.41b)
2
1
âR,r = √ (âTM + jâTE ) (2.41c)
2
1
âL,r = √ (âTM − jâTE ) (2.41d)
2
The subscripts R and L indicate RHCP and LHCP, respectively, and the subscripts i, t, r the incident,
transmission, and reflection directions. The different definition for the reflected waves from the incident
and transmitted ones comes from the different propagation directions, as RHCP and LHCP are based
on the right-hand convention applied to the direction of travel. These expressions allow for the incident
and reflected CP fields to be determined by projecting the TE and TM fields onto a CP basis as
EiR = Ei · âR,i (2.42a)

EiL = Ei · âL,i (2.42b)
ErR = Ei · âR,r (2.42c)
ErL = Ei · âL,r , (2.42d)
which yields
" "# #" #
EiR1 j 1 EiTE
L
=√ (2.43a)
Ei 2 −j 1 EiTM
" # " #" #
ErR 1 −j 1 ErTE
=√ . (2.43b)
ErL 2 j 1 ErTM
The CP transmission response from incident linearly-polarized waves can be determined by applying the
same techniques. This is useful in the case of linear-to-circular polarizers. The difference in sign on the
imaginary terms between (2.41) and (2.43) arises from conjugation when taking the dot product between
complex numbers. The TE- and TM-polarized fields can be expressed in terms of the circularly-polarized
fields by solving for them in (2.43) by taking the inverse of the polarization conversion matrix:
" "# #" #
EiTE 1 −j j EiR
=√ (2.44a)
EiTM 2 1 1 EiL
" # " #" #
ErTE 1 j −j ErR
=√ . (2.44b)
ErTM 2 1 1 ErL
Now as the relations between the linear polarization (LP) and CP bases are known, the LP and CP
GSMs can be written in terms of each other using an analysis based on a method in the literature [34].
Substituting (2.44a) into (2.30) and the result into (2.43b), where the scattered fields are taken to be
the reflected fields, then gives:

" # " #" #" #" #
TE,TE TE,TM
ErR 1 −j 1 S11 S11 −j j EiR
= TM,TE TM,TM . (2.45)
ErL 2 j 1 S11 S11 1 1 EiL
This can be simplified to obtain the CP S-parameters in terms of the TE- and TM- polarized S-parameters
as
RR 1 h TE,TE TM,TM

TE,TM TM,TE
i
S11 = −S11 + S11 − j S11 + S11 (2.46a)
2
RL 1 h TE,TE TM,TM

TE,TM TM,TE
i
S11 = S11 + S11 − j S11 − S11 (2.46b)
2
LR 1 h TE,TE TM,TM

TE,TM TM,TE
i
S11 = S11 + S11 + j S11 − S11 (2.46c)
2
LL 1 h TE,TE TM,TM

TE,TM TM,TE
i
S11 = −S11 + S11 + j S11 + S11 , (2.46d)
2
where the one-port CP reflection matrix takes the form:

" # " #" #
ErR RR
S11 RL
S11 EiR
= . (2.47)
ErL LR
S11 LL
S11 EiL
Applying a similar analysis, substituting (2.43a) into (2.47) and the result into (2.44b) yields the TE/TM
reflection terms in terms of the CP ones as
TE,TE 1 RR LL RL LR

S11 = −S11 − S11 + S11 + S11 (2.48a)
2
TE,TM j RR LL RL LR

S11 = S11 − S11 + S11 − S11 (2.48b)
2
TM,TE j RR LL RL LR

S11 = S11 − S11 − S11 + S11 (2.48c)
2
TM,TM 1 RR LL RL LR

S11 = S11 + S11 + S11 + S11 , (2.48d)
2
which will be used in the analysis of CP reflectarrays.
2.4 Analysis of Reflectarrays

2.4.1 Single Reflectarrays
Reflectarrays are challenging to analyze for a number of reasons. First of all, due to their complex
and arbitrary geometry they do not lend themselves to closed-form solutions of the type that can be
obtained for simplified analyses of paraboloidal reflectors. Second, they pose a multi-scale problem: to
achieve the necessary gain for the applications they are usually proposed for, a reflectarray can be tens
of wavelengths in diameter, with a feed that is located a similar distance away. However, the reflectarray
is made up of scatterers with a typical maximum spacing of λ0 /2, with element features that can be
less than λ0 /10, where λ0 is the free space wavelength. Additionally, more complex reflectarrays can
be made of two or three layers, and in some cases are integrated with other electromagnetic structures,
such as phased arrays, frequency selective surfaces, or polarizers in close vicinity. The multi-scale nature
of the problem, which may contain little or no symmetry that can be exploited, can result in high
computational complexity that makes analysis using commercially available electromagnetic simulation
tools difficult and time consuming, or even prohibitive. Furthermore, for designs based on iterative
methods or optimization routines it is desirable to have a fast analysis technique for initial designs.
In this section, a technique to quickly compute reflectarray patterns will be described. It is based
on the array factor analysis of the Section 2.1, and is an implementation of a method in the literature
[35]. To more accurately and rigorously compute the patterns more sophisticated techniques such as the
Method of Moments (MoM) or Finite Element Boundary Integral (FEBI) can be used, but array factor
analyses have shown reasonable agreement with measurements for moderate- to large-sized reflectarrays.
The procedure that will be followed is to first compute the fields incident on the reflectarray surface
resulting from the feed. The GSMs obtained from periodic structure simulations will then be used
to compute the reflected field at the surface of the reflectarray. Finally, this will be treated as the
excitation of an antenna array, and the IDFT will be used to compute the far-field pattern. This
procedure is analogous to the aperture distribution method used for reflector antennas [21, 36], in which
the near-field is computed across a plane normal to the rays leaving the reflector, and is used to compute
the far-field pattern by treating it as the excitation of an aperture.
Feed Model
For the first part of this analysis, we need a way to compute the patterns resulting from the feed. A
common feed model is the cosq θ model which has fields of the form [36]
e−jkrf
Ef,X = E0 θ̂f cosqE θf cos φf − φ̂f cosqH θf sin φf (2.49a)
rf
−jkrf
e
Ef,Y = E0 θ̂f cosqE θf sin φf − φ̂f cosqH θf cos φf , (2.49b)
rf
where (2.49a) represents an x-polarized feed, and (2.49b) represents a y-polarized feed. It assumes that
the feed only radiates towards the reflectarray (0◦ ≤ θf ≤ 90◦ and 0◦ ≤ φf ≤ 360◦ ), and does not radiate
in the backward direction. The subscript f indicates that the spherical coordinates and unit vectors
are referenced to a feed-centered coordinate system, as illustrated in Fig. 2.12, as the feed will likely
be displaced and rotated relative to the reflector or reflectarray. The superscripts qE and qH are used
to model the beamwidth of the main lobe of the feed in its E-plane and H-plane, respectively. Higher
values of qE and qH can be used to model a more directive feed such as a horn antenna, whereas lower
values can model less directive feeds such as dipoles or patches.
A comparison between the feed model of (2.49a) and a pyramidal horn at 20 GHz is given in Fig. 2.13.
The pyramidal horn was simulated using HFSS, and is fed by a WR-51 (12.95 mm×6.48 mm) rectangular
waveguide. The horn has a 70 mm linear taper, and its aperture size is 43 mm × 43 mm. It is visible that
the cosq θ feed model can closely approximate the main lobe of the horn, which is unequal in its E-plane
and H-plane. It does not model the sidelobes, but in a typical reflectarray design the reflectarray is only
illuminated by the main lobe of the horn, and as the sidelobes are low they can be neglected in the inital
design procedure.
Other methods to model feed antennas include Gaussian beams, spherical wave expansions, comput-
ing the feed patterns via a technique such as mode matching or the finite element method, and measured
feed patterns. While both cosq θ and Gaussian beam models only model the main lobe of the feed an-
tenna, the others can be used to model the entire feed amplitude and phase pattern at the expense of
xR
Ea
yR
zR
xf
θf
φf
zf
yf
Figure 2.12: Reflectarray analysis geometry. The element which the feed ray is incident on is being
treated as an xR -polarized uniformly illuminated aperture.
0 0
Gain, normalized [dB]
Gain, normalized [dB]
-10 -10
-20 -20
cos q 3, q=28.0 cos q 3, q=17.0
Pyramidal horn Pyramidal horn
-30 -30
-60 -40 -20 0 20 40 60 -60 -40 -20 0 20 40 60
Theta [deg.] Theta [deg.]
(a) E-plane pattern cut (b) H-plane pattern cut
Figure 2.13: Comparison between the feed model of (2.49a) with qE = 28 and qH = 17 to a pyramidal
horn simulated at 20 GHz using HFSS.
higher complexity, and the requirement of having a specific feed antenna to model. Due to its simplicity
and ability to model a large number of feeds at an abstract level, and widespread use in the reflector
and reflectarray antenna community, the cosq θ model will be used in this thesis.
Element Factor
To compute the reflectarray patterns using the array factor method, an element factor is needed to
model the elements. Although for some more traditional elements, such as microstrip patches, closed
form equations exist, there are a wide variety of reflectarray elements available to choose from. It would
not be possible to derive expressions for the fields scattered by each one, and in some cases the geometry
is complex and makes deriving an accurate closed-form model difficult. To this end, each reflectarray
unit cell will be modelled as a uniformly illuminated aperture on a perfectly conducting ground plane,
as shown for one element in Fig. 2.12. This methodology is common in the reflectarray community.
Let us consider a rectangular aperture in the xy-plane, pictured in Fig. 2.14, extending over the
region −∆x/2 ≤ x0 ≤ ∆x/2 and −∆y/2 ≤ y 0 ≤ ∆y/2. We will look at two cases: the x-polarized case
where the aperture field is EX Y
a = x̂E0x and the y-polarized case where Ea = ŷE0y . Assuming that the
aperture is mounted on an infinite PEC ground plane, Schelkunoff’s equivalence principle and image
theory can be applied to replace the aperture with an equivalent magnetic current over the same region,
as illustrated in Fig. 2.14b, having the value
M = 2Ea × n̂, (2.50)
where n̂ is the normal vector of the aperture, which in this case is ẑ. Substituting in the aperture fields
then gives
MX = 2x̂E0x × ẑ = −ŷ2E0x (2.51a)

MY = 2ŷE0y × ẑ = x̂2E0y , (2.51b)
where the subscripts X and Y indicated the polarization of the aperture field. These magnetic currents
then give rise to the electric vector potential
¨
e−jkr 0
F= Mejkr ·r̂ dS 0 . (2.52)
4π r
S
Evaluating the integrals over the aperture region, and neglecting the r̂ component of the Cartesian unit
vectors when transforming from rectangular to spherical coordinates, gives
e−jkr
∆x∆yE

0x kx ∆x ky ∆y
FX = − θ̂ cos θ sin φ + φ̂ cos φ sinc sinc (2.53a)
2π r 2 2
−jkr
∆x∆yE

0y e kx ∆x ky ∆y
FY = θ̂ sin φ + φ̂ cos θ cos φ sinc sinc , (2.53b)
2π r 2 2
where in this case the sinc function is defined as sinc(x) = sin(x)/x. Applying the relations in the
far-field that Eθ ' −jωηFφ and Eφ ' jωηFθ [21] gives the electric field for each polarization as
e−jkr
jk∆x∆yE

0x kx ∆x ky ∆y
EX = θ̂ cos φ − φ̂ cos θ sin φ sinc sinc (2.54a)
2π r 2 2
jk∆x∆yE e−jkr
kx ∆x

ky ∆y

0y
EY = θ̂ sin φ + φ̂ cos θ cos φ sinc sinc , (2.54b)
2π r 2 2
where the approximation has been replaced by an equality with the understanding that these relations
are only valid in the far-field.
x
x
PEC
, µ
EX
a
MY
y EYa
y MX
∆x
∆x
z z
∆y ∆y
(a) Aperture mounted on an infinite perfectly conduct- (b) Equivalent magnetic current distributions
ing ground plane
Figure 2.14: Analysis of a rectangular aperture considering two orthogonal linear polarizations.
Radiated Fields
Now with a feed model and element factor, the array factor method outlined in Section 2.1 can be used
to compute the radiation pattern of the reflectarray. First, considering one feed polarization (x or y),
either (2.49a) or (2.49b) is used to calculate the incident field across the reflectarray surface in the feed
coordinate system (xf , yf , zf ). From this, the fields can be projected onto the reflectarray surface to
obtain the tangential fields across the reflectarray in its own coordinate system (xr , yR , zR ). The different
coordinate systems are illustrated in Fig. 2.12, and in general they are offset and rotated with respect
to each other.
The reflected near-field of the reflectarray can be determined using the GSM from (2.30) as
Er (xR , yR ) = x̂R Er,x + ŷR Er,y = S · Ei (xR , yR ) , (2.55)
where Ei refers to the tangential components of the incident fields from the feed with respect to the
reflectarray surface. The fields in (2.55) are expressed in the (xR , yR , zR ) coordinate system in which
the xR and yR axes are tangent to the reflectarray, and the zR axis is normal to its surface. The
reflectarray lies in the plane zR = 0. Although in (2.30) the TE and TM polarizations were used, in
HFSS under normal incidence the convention is that âTM = x̂R and âTM = −ŷR , and so the appropriate
transformation of polarization can be performed.
From (2.49) and (2.55) it is clear that no matter what the feed polarization, that both xR and yR
components will be present in the incident field on the reflectarray. Thus, to calculate both the co-
and cross-polarized components, the xR component will be treated as the excitation of the x-polarized
aperture of (2.54a) and the yR component as the excitation of the y-polarized aperture of (2.54b). The
results will then be summed to obtain the total field. Although cross-polarization will result simply from
the geometry of the problem, it will mostly be determined by the magnitude of the cross-polar terms in
the element GSMs, and the cross-polarization level of the feed.
As in most of the reflectarray elements will be located away from the origin, the r terms in the
−jkr
e /r factor of (2.54) should be replaced by |r − r0 | to account for the different origins of the elements.
However, this will be neglected for the 1/r factor. Applying the far-field approximation of (2.5) and
summing the contributions of each element, where the elements are modelled by taking E0x = Er,x and
E0y = Er,y in (2.54) and summing both terms, gives:
h i
E (θ, φ) =K θ̂ (cos φ + sin φ) + φ̂ cos θ (cos φ − sin φ)
M X
X N . (2.56)
[Er,x (xR , yR ) + Er,y (xR , yR )] ej(kx xR +ky yR )
m=1 n=1
K is a common term given by
e−jkr

jk∆x∆y kx ∆x ky ∆y
K= sinc sinc . (2.57)
2π 2 2 r
Recognizing the double summation as the IDFTs of the two components of the reflected near-field, (2.16)
can be applied to obtain
h i
E (θ, φ) = θ̂ (cos φ + sin φ) + φ̂ cos θ (cos φ − sin φ) KM N [IDFT2 {Er,x } + IDFT2 {Er,y }] . (2.58)
This now gives us the vector electric field radiated by the reflectarray, in the reflectarray coordinate
system. Depending on the application, additional coordinate transformations may need to be applied to
observe the fields in a global coordinate system, or one with its z-axis aligned with the pattern peak.
It should be noted in the above analysis that when applying the equivalence principle only the
magnetic currents were used, as for an aperture on an infinite perfectly conducting ground plane the
electric currents are shorted out [21]. However, for an arbitrary aperture in space both the electric and
magnetic currents should be accounted for. This has been done in the literature, and has shown good
correlation with measurement results [31].
From (2.16) it is visible that using the IDFT to calculate the AF will give it as a function of kx
and ky . Due to the IDFT, this will result in points that are spaced in a rectangular grid of kx and
ky points. However, due to the non-linear relationship between (kx , ky ) and the spherical coordinate
variables (θ, φ), the spacing in (θ, φ) will not be uniform. Thus, it will be convenient to introduce the
additional coordinate variables
kx
u= = sin θ cos φ (2.59a)
k0
ky
v= = sin θ sin φ (2.59b)
k0
kz p
w= = cos θ = 1 − u2 − v 2 . (2.59c)
k0
The (u, v, w) coordinate system will be termed direction cosine space, where u, v, and w are the direction
cosines. (2.59) shows how the direction cosines can be converted back to spherical coordinates to obtain
‘cuts’ which are commonly used to characterize antenna radiation patterns. (2.59) can also be rearranged
to express some common trigonometric functions in terms of the direction cosines as
p
sin θ = u2 + v 2 (2.60a)
cos θ = w (2.60b)
v
sin φ = √ (2.60c)
u + v2
2
u
cos φ = √ (2.60d)
u + v2
2
v
tan φ = . (2.60e)
u
(2.60) then allows (2.58) to expressed in direction cosine space as
θ̂ (u + v) + φ̂w (u − v)
E (u, v) = √ KM N [IDFT2 {Er,x } + IDFT2 {Er,y }] . (2.61)
u2 + v 2
Circular Polarization
The analysis so far has assumed that the reflectarray and feed are both linearly polarized. Now the
case when both are circularly polarized will be considered, which requires modifications to the feed
model. The circularly polarized feed models can be formed from an appropriately phase shifted linear
combination of the x-polarized and y-polarized feed models in (2.49) as:
1
Ef,R = √ (jEX,f + EY,f ) (2.62a)
2
1
Ef,L = √ (EX,f + jEY,f ) . (2.62b)
2
In the CP case the element GSMs are typically expressed in a CP basis. To use the analysis developed in
this section, they must be converted to an LP basis using (2.48) before applying (2.55) to determine the
reflected fields. After computing the radiated fields using (2.61), the CP fields can then be determined
by projecting them onto the CP basis vectors. In terms of the spherical coordinate unit vectors, the CP
basis vectors are given by
1
âR = √ θ̂ − j φ̂ (2.63a)
2
1
âL = √ θ̂ + j φ̂ , (2.63b)
2
which gives the CP radiated fields as:
1
ER (u, v) = E · âR = √ (Eθ + jEφ ) (2.64a)
2
1
EL (u, v) = E · âL = √ (Eθ − jEφ ) . (2.64b)
2
Depending on which feed polarization is used, either (2.64a) or (2.64b) will correspond to the co-polar
or cross-polar component.
2.4.2 Dual Reflectarrays

In the previous section a method used to calculate the fields radiated by a reflectarray using a feed model
was described. However, there are situations which use two reflectarrays, a sub-reflectarray and main
reflectarray, and for these cases modifications must be made to compute the incident field on the main
reflectarray. Again, the analysis follows a previously published work [35].
The far-field region of an antenna, in which the phase and amplitude of its pattern do not depend
on the radial coordinate r apart from the e−jkr /r factor, is given by [21]
2D2
r> , (2.65)
λ
where the coordinate system is centred on the antenna, and D is its maximum dimension. For a
traditional reflectarray, a feed that is physically much smaller than the reflectarray is used. Assuming
that the reflector does not have a very small F/D ratio it is likely that that reflectarray is in the far-field
zone of the feed, and so models such as (2.49) can be used. However, a sub-reflectarray is typically
much larger than a feed horn, and in some cases can even be the same size as the main reflectarray.
Thus, in general the main reflectarray will not be in the far-field region of the sub reflectarray unless
the separation between the two is very large, and (2.61) cannot be used to determine the incident field
on the main-reflectarray. The approach taken here will be to take into account the different distances
and angles between each element on the main reflectarray and each element on the sub-reflectarray.
The analysis geometry is illustrated in Fig. 2.15. Let us assume that the main reflectarray has M ×N
elements, each of which is indexed as (m, n), and the sub-reflectarray has P × Q elements with indices
(p, q). The main reflectarray also has element spacings ∆xm in the x direction, and ∆ym in the y
direction, and the sub-reflectarray ∆xs and ∆ys . As the feed, sub-reflectarray, and main reflectarray
can be arbitrarily positioned and angled, there are four coordinate systems to consider. These are
the global coordinate system (x, y, z), feed coordinate system (xf , yf , zf ), main reflectarray coordinate
system (xm , ym , zm ), and sub-reflectarray coordinate system (xs , ys , zs ). The global coordinate system
provides a reference for the other three. The feed coordinate system is centred on the phase center of
the feed and is rotated such that its z-axis points normal to its aperture. The main and sub-reflectarray
coordinate systems have their origins located on their surfaces with their z-axes pointing normal to them.
Additionally, due to the main reflectarray not necessarily being in the far-field of the sub-reflectarray
the different origins of each sub-reflectarray element must be considered. Each sub-reflectarray element
has its own coordinate system (xpq , ypq , zpq ) which is a translated version of (xs , ys , zs ).
First, using the feed model of (2.49) and transforming the resulting fields to the sub-reflectarray
coordinate system, its incident field Eis (xs , ys ) can be calculated at each element. (2.55) can then be
xs φpq
xpq
ypq
ys
zpq
zs
rpq
xf θpq
θf
φf
zf
yf
xm
zm
ym
Figure 2.15: Dual-reflectarray analysis geometry.
applied to determine the reflected fields. Next, modelling the sub-reflectarray as an array of uniformly
illuminated rectangular apertures, as was done in the previous section, the incident field on the main
reflectarray can be determined after performing a coordinate transformation to obtain the fields in the
main reflectarray coordinate system. However, for the dual-reflectarray case the contributions of each
sub-reflectarray element are calculated separately and then summed to take into account the different
distances and angles between each sub-reflectarray and main reflectarray element. As before, to deter-
mine cross-polarization effects an x- and y-polarized aperture are used to model each element and the
results are summed vectorially.
Substituting the standard definitions for the spherical coordinate unit vectors into (2.54) gives
jk∆x∆yE0x e−jkr

kx ∆x ky ∆y
EX = (x̂ cos θ − ẑ sin θ cos φ) sinc sinc (2.66a)
2π r 2 2
−jkr

jk∆x∆yE0y e kx ∆x ky ∆y
EY = (ŷ cos θ − ẑ sin θ sin φ) sinc sinc . (2.66b)
2π r 2 2
This will allow for the contribution of each aperture to be summed, as the spherical unit vectors change
with position. Then, adding the contribution of each element on the sub-reflectarray at an element on
the main reflectarray yields
Q
P X
X
Eim (m, n) = Kpq {(x̂s cos θpq − ẑs sin θpq sin φpq ) Er,xs (p, q)
p q (2.67)
+ (ŷs cos θpq − ẑs sin θpq sin φpq ) Er,ys (p, q)} ,
where
jk∆xs ∆ys e−jkrpq

k sin θpq cos φpq k sin θpq sin φpq
Kpq = sinc sinc . (2.68)
2π rpq 2 2
Here x̂s , ŷs , and ẑs are the Cartesian unit vectors of the sub-reflectarray coordinate system, and
(rpq , θpq , φpq ) is the position in spherical coordinates of element (m, n) on the main reflectarray with
respect to element (p, q) on the sub-reflectarray. Er,s (p, q) = x̂s Er,xs (p, q) + ŷs Er,ys (p, q) is the reflected
field at each sub-reflectarray element. Now that the incident field on the main reflectarray is known,
computing its radiated fields is the same as for the single reflectarray case and (2.61) can be used. This
can then be broken down into co-polar and cross-polar components as before.
2.4.3 One-dimensional Dual Reflectarrays
In the previous sections, the reflectarrays that were analyzed were two-dimensional. Although in reality
when a reflectarray is fabricated it is typically a planar two-dimensional structure, it can be useful to
analyze and synthesize one-dimensional reflectarrays. This will be used in the reflectarray synthesis work
in Chapter 3.
For a one-dimensional reflectarray, the same array factor method will be again applied. However,
as it will be assumed that the problem is infinite in the horizontal direction and variations only take
place in the vertical direction, the space over which the fields are calculated is reduced from three
dimensions to two. Only the case in which the electric field vector is in the plane of incidence will be
considered. This requires modifications to be made to the element factor. Two element factors will
be considered: one where the pattern of the array elements is treated as scattering of TM-polarized
waves from a rectangular strip at normal incidence using the physical optics method, and one where
the reflectarray is considered as an array of magnetic line currents. The first method was used for the
one-dimensional reflectarray work in this thesis, but the second is more analogous to what was used for
the two-dimensional reflectarray analysis.
The feed model used is similar to (2.49). Restricting our analysis to the xz-plane (φ = 0, −π/2 ≤
θ ≤ π/2) and assuming no variation in the y-direction gives:
e−jkrf
Ef = θ̂f E0 √ cosq θ. (2.69)
rf
Array of strips / physical optics
In this analysis, the geometry is in the xz-plane and the strip extends to infinity in the +y and −y
directions, as illustrated in Fig. 2.16. Although typically two-dimensional problems are analyzed in
polar coordinates in the xy-plane, the geometry used here corresponds to a vertical cross-section of
the three-dimensional reflectarray geometries used in this thesis. Thus the spherical coordinates r and
θ are still used, but with the understanding that the region of interest is restricted to φ = 0 and
−π/2 ≤ θ ≤ π/2.
+∞ r
y
θ ∆x
−∞
z
Ei
ki
Hi
Figure 2.16: Scattering of a TM-polarized incident field from a strip, which is used to determine the
element factor for the analysis of one-dimensional reflectarrays.
Consider a plane wave normally incident on the strip of the form:
Ei = x̂E0 e−jkz (2.70a)

1 1 E0
Hi = k̂ × Ei = − ẑ × x̂E0 e−jkz = −ŷ e−jkz . (2.70b)
η η η
At z = 0 this will give the physical optics current density [37]
E0 2E0
Js = 2n̂ × Hi = −2ẑ × ŷ = x̂ , (2.71)
η η
This give rise to the magnetic vector potential

¨
µ e−jkR 0
A= Js dS
4π S R
ˆ
∆x/2 ˆ
+∞ √ (2.72)
0 2 0 2
µE0 e−jk |r−r | +(y−y )
= x̂ q dx0 dy 0 ,
2πη 0 2 0 2
−∆x/2 −∞ |r − r | + (y − y )
where r = x̂x + ẑz and r0 = x̂x0 . The observation points lie in the y = 0 plane. This can be transformed
into [37]
ˆ
∆x/2
jµE0 (2)
A = −x̂ H0 (k |r − r0 |) dx0 , (2.73)
2η
−∆x/2
(2)
where H0 is the Hankel function of the second kind and order zero. Taking the large argument
approximation of the Hankel function [38] yields
s
2j 0
e−jk|r−r | .
(2)
H0 (k |r − r0 |) ≈ (2.74)
πk |r − r0 |
Taking θ to be negative for x < 0 in the φ = 0 plane where the analysis takes place, the far-field
approximation from Section 2.1 can be applied to the distance term to give
|r − r0 | ≈ r − r̂ · r0 = r − r0 cos (θ − θ0 ) = r − x0 sin θ. (2.75)
Retaining both terms for phase and only the first term for amplitude as before, substituting (2.75) into
(2.74) and the result into (2.73) gives
r ˆ
∆x/2 r
j e−jkρ e−jkρ

jµE0 0 jµE0 ∆x j k∆x sin θ
A ≈ −x̂ √ ejkx sin θ
dx0 = −x̂ sinc √ . (2.76)
η 2πk ρ η 2πk 2 ρ
−∆x/2
Transforming (2.76) to spherical coordinates and applying the relation in the far-field that E ' −θ̂jωAθ +
φ̂Aφ [36] gives
r
k∆x sin θ e−jkρ

jk
E = −θ̂E0 ∆x cos θ sinc √ (2.77a)
2π 2 ρ
r
k∆x sin θ e−jkρ

jk
= − (x̂ cos θ − ẑ sin θ) E0 ∆x cos θ sinc √ , (2.77b)
2π 2 ρ
which will be used as the element factor for the one-dimensional reflectarray analysis. Taking u = sin θ,
(2.77) can also be expressed in direction cosine space as

r
k∆xu e−jkρ

jk p
E = −θ̂E0 ∆x 1 − u2 sinc √ (2.78a)
2π 2 ρ
r
k∆xu e−jkρ

p
2
jk p
2
= − x̂ 1 − u − ẑu E0 ∆x 1 − u sinc √ , (2.78b)
2π 2 ρ
which will be convenient when the DFT is used to compute the radiation patterns.
The dual-reflectarray analysis follows the same procedure as in Section 2.4.2, but applied to one-
dimensional reflectarrays. The incident field on the sub-reflectarray can be computed using (2.69),
and then projected onto the reflectarray. The tangential component of the reflected field can then be
computed at each reflectarray element from the reflection coefficient as
Er,s = x̂s Γ · Ei,s , (2.79)
where Γ is the xs -polarized reflection coefficient of the element. Then (2.77b) can be used to compute
the incident field at reach main reflectarray element resulting from all sub-reflectarray elements giving
r P
k∆x sin θp e−jkrp

jk X
Ei,m = −∆x Er,x (p) (x̂s cos θp − ẑ sin θp ) cos θp sinc √ , (2.80)
2π p 2 rp
where the sub-reflectarray is taken to have P elements and p is the element index, and Er,x (p) is the
result from (2.79) at each element. The coordinate system (rp , θp ) is aligned with the sub-reflectarray
coordinate system, but with its origin shifted to the center of element p. Computing the tangential
reflected field across the main reflectarray by projecting the incident field onto the main reflectarray
coordinate system, and then multiplying by the reflection coefficient of each element as in (2.79), the
radiated fields of the dual-reflectarray can be computed using the array factor method in a similar
manner to the two-dimensional reflectarray analysis. This gives:
M
r
jk (1 − u2 ) e−jkr X

kx ∆x 0
Em = −θ̂ sinc √ Er,m (x0 ) ejkx x (2.81a)
2π 2 r m
r
jk (1 − u2 ) e−jkr

kx ∆x
= −θ̂ sinc √ M · IDFT {Er,m (x0 )} , (2.81b)
2π 2 r
where there are M elements in the main reflectarray, and kx = k sin θ. The relationship u = sin θ can
be used to convert the patterns to a function of θ.
Array of Slots / Magnetic Line Currents
Consider a slot in the xy-plane that is infinite in the ±z-directions and finite in the y-direction with
width ∆y. Although the geometry is different from the previous section, a coordinate transformation
could be performed to transfer between either coordinate system, or the reflectarray geometry could be
adjusted to account for differences.
Assuming a vertically-polarized uniform aperture field Ea = ŷE0 , application of the equivalence
principle as before gives a magnetic current of M = −ẑ2E0 . In the far-field in two dimensions, the
electric vector potential resulting from a magnetic current distribution is given by [39]
r ¨
j e−jkρ 0
F = −j √ M (ρ0 ) ejkρ ·ρ̂ dρ, (2.82)
8πk ρ S
where the large argument approximation of the Hankel function has been applied, and ρ = x̂x + ŷy is
the radial vector in cylindrical coordinates. Integrating over the slot, the source coordinate is ρ0 = ŷy 0
and dρ = dy 0 . Substituting these expressions in and taking the dot product in the argument of the
exponential gives
r ˆ
∆y/2
j e−jkρ 0
F = −j √ −ŷ2E0 ejk sin φy dy 0 , (2.83)
8πk ρ
−∆y/2
which when evaluated results in:

r
j e−jkρ

k∆y sin φ
F = j2E0 ∆y √ sinc . (2.84)
8πk ρ 2
In the two-dimensional far field, in the absence of J, the electric field is related to the electric vector
potential by [39]

E = −jωη φ̂Fz − ẑFφ , (2.85)
which gives the electric field radiated by the slot as:

r
jk e−jkρ

k∆y sin φ
E = φ̂E0 ∆y √ sinc . (2.86)
2π ρ 2
This element factor is similar to (2.77a), which was used for the one-dimensional reflectarray analysis in
this thesis, but expressed in a different coordinate system.
2.5 Antenna Performance Metrics
In order to characterize the performance of antennas and make quantitative comparisons between dif-
ferent antennas or design methods, there are several metrics that are used. Some common ones are
directivity, gain, aperture efficiency, side lobe level (SLL), and cross-polar discrimination (XPD). For
circularly polarized antennas, axial ratio (AR) may be specified rather than XPD.
The definitions of SLL and XPD are generally straightforward and visible from context. SLL is
the amplitude of the minor lobe closest to the main lobe, relative to the main lobe amplitude. A low
SLL is desirable, as for a directional antenna side lobes indicate power being radiated in unwanted
directions. XPD is the point-by-point ratio of the co-polarized pattern to the cross-polarized pattern,
and for a directional antenna it is generally quoted in the direction of the co-polarized pattern’s main
lobe. It corresponds to how much isolation the antenna can provide between the wanted and unwanted
polarizations. The other performance metrics are a bit more nuanced, and so they will be described in
more detail in this section.
2.5.1 Radiated Power
To calculate many antenna performance parameters, the power radiated by the antenna must be known.
From standard electromagnetic theory, the average power flux in a region of space is given by the
Poynting vector S:
1
S= E × H∗ . (2.87)
2
Assuming that E and H are the fields of a plane wave that propagates in the r̂ direction, they are related
by
1
H= r̂ × E, (2.88)
η
where η is the intrinsic impedance of the medium through which E and H propagate. In this thesis it
p
will be assumed to be free space, for which η = η0 = µ0 /0 ≈ 377 Ω. In the far-field, where the r̂
component of E is negligible, the electric and magnetic fields can be expressed as
E = θ̂Eθ + φ̂Eφ (2.89a)

1
H= φ̂Eθ − θ̂Eφ . (2.89b)
η
(2.87) and (2.89) then give an expression for the Poynting vector in the far-field in terms of the electric
field only:
1 2 2

S = r̂ |Eθ | + |Eφ | . (2.90)
2η
(2.87) can be integrated over a closed surface to find the total power passing through it. For a surface
which encloses an antenna this gives the power which it radiates, Prad , which can be expressed as
‹
Prad = S · ds. (2.91)
S
In this case the integration will be performed over the far-field sphere, and so ds = r̂r2 dΩ. Substi-
tuting (2.90) into (2.91) then yields
‹
r2
2 2

Prad = |Eθ | + |Eφ | dΩ (2.92a)
S 2η
‹ 2
r
2 2

= |Eθ | + |Eφ | sin θdθdφ, (2.92b)
S 2η
where the integrand in (2.92a) is called the radiation intensity U (θ, φ):
r2 2 2

U (θ, φ) = |Eθ | + |Eφ | . (2.93)
2η
Radiation intensity expresses the power per unit solid angle in a region, with units of Watts per steradian
(W/sr). Note that by multiplying the square of the electric field magnitude by r2 that this removes the
inverse square distance term. This means that U is a quantity that is independent of distance.
While the expressions in (2.92) use spherical coordinates, direction cosine space can be convenient
to calculate the radiated power, as when computing the radiated fields using the IDFT the results are
obtained over a uniform grid in (u, v). The solid angle in (2.92a) can be written in terms of the direction
cosines using the expression
∂ (θ, φ)
dΩ = sin θdθdφ = sin θ dudv, (2.94)
∂ (u, v)
∂(θ,φ)
where ∂(u,v) is the Jacobian determinant:
∂θ ∂θ
∂ (θ, φ) ∂u ∂v
= . (2.95)
∂ (u, v) ∂φ ∂φ
∂u ∂v
From (2.60a) and (2.60e) the spherical coordinate angles can be expressed in terms of the direction
cosines as:
p
θ = sin−1 u2 + v 2 (2.96a)
v
φ = tan−1 . (2.96b)
u
Taking the derivatives of (2.96) with respect to u and v gives the entries in the Jacobian:
∂θ u
= √ (2.97a)
∂u w u2 + v 2
∂θ v
= √ (2.97b)
∂v w u2 + v 2
∂φ −v
= 2 (2.97c)
∂u u + v2
∂φ u
= 2 . (2.97d)
∂v u + v2
Substituting (2.97) into (2.95) and evaluating gives the Jacobian determinant as:
∂ (θ, φ) 1
= √ . (2.98)
∂ (u, v) w u2 + v 2
Then substituting (2.98) and (2.60a) into (2.94) gives an expression for the solid angle in direction cosine
space:
dudv
dΩ = . (2.99)
w
Substituting (2.99) into (2.92) then gives an expression for computing the radiated power in direction
cosine space:
‹
r2 2 2 dudv

Prad = |Eθ | + |Eφ | . (2.100)
S 2η w
While the integration in (2.92) takes place over a closed surface, the analysis used here only predicts
radiation in front of the reflectarray and so back-radiation should be neglected. Thus, the integration
√
limits are 0 ≤ θ ≤ π/2, 0 ≤ φ ≤ 2π, or equivalently u2 + v 2 ≤ 1. This gives the following expressions
for Prad :
ˆπ/2ˆ2π
r2 2 2

Prad = |Eθ | + |Eφ | sin θdθdφ (2.101a)
2η
0 0
¨
r2 2 2 dudv

= √ |Eθ | + |Eφ | . (2.101b)
u2 +v 2 ≤1 2η w
Both expressions in (2.101) correspond to a hemisphere in front of the antenna, which could be either
the reflectarray or the feed. Note that like the reflectarray, from (2.49) the feed model also assumes that
no power is radiated in the backward direction.
2.5.2 Aperture Efficiency, Directivity, and Gain

The power radiated by the antenna, as expressed by (2.101), can be used to define some important
performance metrics. As the fields from the feed can be predicted by (2.49), and those of the reflectarray
can be computed from (2.61), (2.101) can be used to compute the power radiated from both. Then the
aperture efficiency of the reflectarray ε can be defined as ratio of the power radiated by the reflectarray
Prad,RA to that radiated by the feed Prad,f :
Prad,RA
ε= (2.102)
Prad,f
The efficiency of the feed has been neglected, as this work is focused on the design of the reflectarray.
Aperture efficiency can be broken down into several terms, in a similar manner to what is done for
reflector antennas [21, 36]:
ε = εs εt εp εx εb εd . (2.103)
Here εs is spillover efficiency, which represents the power radiated by the feed antenna that is not in-
tercepted by the reflectarray, and ‘spills over’ its edges. The next term εt is taper efficiency, which is
a measure of the reduction in directivity resulting from having a non-uniform amplitude distribution
across the reflectarray’s surface. As with reflector antenna design, with reflectarrays there is a compro-
mise between the spillover and taper efficiencies. As a more sharply tapered amplitude distribution is
introduced less losses will be present due to spillover, resulting from the lower edge illumination, but the
directivity will be reduced due to the taper. Conversely, as the edge illumination is raised directivity
will increase at the expense of increased spillover losses.
εp and εx are the phase and polarization efficiencies, respectively. Phase efficiency represents the
reduction in directivity resulting from phase errors across the aperture that will be present due to
fabrication errors, quantization of the element phases, and assumptions or compromises made during
the design phase. For reflectarrays additional phase errors beyond those contributed by the elements
can be introduced if the feed does not have a flat phase pattern over the region which illuminates the
reflectarray, or through errors in positioning of the reflectarray and feed. Polarization efficiency is a
measure of the power lost to radiation produced in the undesired polarization, which is generated by
both the feed and the reflectarray elements.
The next term εb is blockage efficiency, which represents losses due to the fields radiated by the
reflectarray being ‘blocked’ by the feed horn and its mounting structure, and for dual-reflectarrays
additionally the sub-reflectarray. This can be reduced by using an offset configuration, in which the
components that would cause blockage are moved away from the main beam direction. Finally, εd is
radiation efficiency, which represents losses due to Ohmic dissipation in the conductors and dielectric
of the reflectarray. These can be reduced by using high quality microwave substrates to construct the
reflectarray, which have a low loss tangent.
Now let us consider an isotropic radiator, which radiates power equally in all directions. Thus, its
radiation intensity will be a constant which will be termed U0 . Re-visiting (2.92), this gives its radiated
power as
‹
Prad = U0 dΩ = 4πU0 , (2.104)
S
where the final term results as there as 4π steradians in the unit sphere. This gives the radiation intensity
of an isotropic radiator as:
Prad
U0 = . (2.105)
4π
The directivity pattern of an antenna is its ability to radiate power more or less in certain directions
compared to a reference antenna. Alternatively viewing the antenna in receive mode, it describes its
sensitivity as a function of space. As (2.105) describes how the power radiated by an antenna would
be distributed if it were an isotropic radiator, using this as the reference the directivity pattern can be
defined as
U (θ, φ) U (θ, φ)
D (θ, φ) = = . (2.106)
U0 Prad /4π
In this case the units of D are decibels relative to isotropic (dBi). The units of D will differ if a reference
other than an isotropic antenna is used. For example, if a loss-less half-wave dipole is used as the
directivity reference the units of D are dBd.
To take into account the efficiency of an antenna, its gain pattern can be computed by referencing
its input power rather than its radiated power:
U (θ, φ) U (θ, φ)
G (θ, φ) = = . (2.107)
U0 Pin /4π
For a reflectarray, neglecting the efficiency of the feed the input power is the power radiated by the feed,
and thus Pin = Prad,f .
From (2.106) and (2.107) it follows that directivity and gain are related by efficiency:
G (θ, φ) = εD (θ, φ) . (2.108)

2.5.3 Radiated Power, Directivity, and Gain in Two-dimensional Space
For the analysis of one-dimensional reflectarrays, it is necessary to modify the gain and directivity
equations for two-dimensional space. Using the coordinate system shown in Fig. 2.16, the electric and
magnetic fields can be written as:
E = θ̂Eθ (2.109a)
1 Eθ
H = r̂ × E = φ̂ . (2.109b)
η η
Applying (2.87) this gives the Poynting vector as
2
|Eθ |
S = r̂ , (2.110)
2η
which when integrated over a circle of radius r gives the radiated power in two dimensions:
ˆπ ˆπ 2
r |Eθ |
Prad = S · r̂rdθ = dθ. (2.111)
2η
−π −π
The integrand is visible as the radiation intensity in two-dimensional space:

2
r |Eθ |
U (θ) = . (2.112)
2η
√
Note that by multiplying by r that U (θ) is independent of distance, as |Eθ | will roll off with 1/ r. This
is consistent with the three-dimensional definition of radiation intensity.
Assuming U (θ) to be uniform in all directions with a value of U0 gives a radiated power from (2.112)
of
Prad = 2πU0 . (2.113)
Similar to (2.106), the directivity in two dimensions can then be defined as:
U (θ)
D (θ) = . (2.114)
Prad /2π
Gain can then be determined in the same manner as (2.108), by multiplying directivity by efficiency.
Re-visiting (2.111), making the change of variable u = sin θ gives:
ˆ1 2
r |E |
Prad = √ θ du. (2.115)
2η 1 − u2
−1
This allows for the radiated power to be computed in direction cosine space, which will again be useful
due to computation of the reflectarray fields using the IDFT.
2.5.4 Co-polar and Cross-polar Directivity and Gain

(2.58) and (2.61) give the reflectarray fields in spherical coordinates. However, for applications where
high-gain antennas such as reflectarrays are used and there are requirements on isolation between po-
larizations, for example satellite communications, it is usually desirable to express the fields in terms of
co-polarized and cross-polarized components. In general the co-polarized and cross-polarized fields are
not Eθ and Eφ , and are instead typically expressed in terms of Ludwig’s third definition of polarization
as [40, 41]
" # " #" #
Eco cos φ − sin φ Eθ
EX (θ, φ) = = (2.116a)
Ecr − sin φ − cos φ Eφ
" # " #" #
Eco sin φ cos φ Eθ
EY (θ, φ) = = , (2.116b)
Ecr cos φ − sin φ Eφ
where Eco and Ecr are the co-polarized and cross-polarized components of the electric field, respectively.
There are two definitions of co-polarization and cross-polarization, as the definitions will change depend-
ing on whether the antenna is x-polarized or y-polarized. This is indicated by the subscript on E in
(2.116a) and (2.116b). For circular polarization, the co-polarized and cross-polarized components are
simply RHCP and LHCP, as given in (2.64)
Rather than the total directivity and gain of an antenna as in the previous section, it is often desired
to break these down into co-polar and cross-polar components. However, directivity and gain are scalar
quantities, and so this decomposition must be done on the fields first.
From the definitions of the co-polar and cross-polar fields given in (2.116), co-polar and cross-polar
radiation intensity can be calculated as in (2.93):
r2 2
Uco (θ, φ) = |Eco (θ, φ)| (2.117a)
2η
r2 2
Ucr (θ, φ) = |Ecr (θ, φ)| . (2.117b)
2η
By (2.106), the co-polar and cross-polar directivity an then be calculated as
Uco (θ, φ)
Dco (θ, φ) = (2.118a)
Prad /4π
Ucr (θ, φ)
Dcr (θ, φ) = , (2.118b)
Prad /4π
where Prad is the total radiated power. If the input and radiated power are known this an then be
used to calculate the radiation efficiency, which allows the co-polar and cross-polar gain patterns to be
calculated as:
Gco (θ, φ) = εDco (θ, φ) (2.119a)

Gcr (θ, φ) = εDcr (θ, φ) . (2.119b)
The above equations can also be expressed in terms of the direction cosines for convenience.
Chapter 3
Bifocal Reflectarray
The first design method that was investigated to design multibeam reflectarrays is a bifocal synthesis
procedure. Before delving into how bifocal techniques have been applied to our application, a discussion
of the motivation and background for multifocal techniques will be provided.
3.1 Introduction
As a result of the increased use of multiple-beam antenna systems in satellite communications and radar
applications, antennas were needed that are able to scan over a wide angular range. There are several
challenges associated with achieving wide angle scanning. For spatially-fed reflector and lens antennas
that scan by mechanically displacing a feed antenna, or that use a cluster of feed antennas to generate
simultaneous or switched multiple beams, effects such as reduced gain due to phase errors and spillover
are introduced as the aperture is fed away from its focal point. Additionally, scan aberrations such as
coma lobes begin to appear that limit the usable scan range [42, 16]. Both mechanically and electrically
steered aperture antennas will also see a reduction in gain due to the reduction in aperture size as they
are scanned away from broadside.
The effects listed above have motivated investigations since the 1950s into design techniques that
mitigate them and increase the scan range of antennas. Lenses with wider scan capability were designed
by using two foci for both constrained [16] and dielectric [17] lenses. These techniques were later
extended to prime-focus reflector antennas [18]. For lenses, as waves travelling to or from the feed must
pass through both surfaces of the lens, modifying the inner and outer lens contours gives enough degrees
of freedom to design for two foci. In the case of reflectors, a second reflector must be introduced to give
two surfaces that can be modified, yielding a configuration that is similar to a Cassegrain or Gregorian
dual-reflector antenna. Some examples of bifocal antennas are shown in Fig. 3.1. As a the sub-reflector in
a prime-focus dual-reflector configuration can cause significant blockage, methods of designing symmetric
bifocal reflectors were extended to dual-offset dual-reflector antennas by Rappaport [19]. Research into
bifocal reflector antennas for wider scan capability continues with work being done with solid offset
shaped reflectors [43], and reflectarrays [44]. In the reflectarray case a symmetric configuration can be
used without aperture blockage if the sub-reflectarray is polarization-selective.
What all of the above methods have in common is that they use of two foci instead of a single focus.
This gives two perfectly focused points which have high quality beams, and the scanned beams between
55
Chapter 3. Bifocal Reflectarray 56
(b) Bifocal symmetric dual-reflector

(a) Un-zoned (left) and zoned (right) bifocal dielectric lens antenna [17].
antenna [18]. The two foci are la-
The two foci are labelled as F 0 and F 00 .
belled as A and B.
Figure 3.1: Examples of early bifocal spatially-fed antennas.
the two foci also generally have good performance [19]. Additionally, for the symmetric case the use of
two foci gives a focal ring instead of just a single focal point, allowing for even more well-focused feeds
to be accommodated.
3.2 Bifocal Reflector Synthesis
Many of the techniques mentioned in the previous section require an antenna geometry with some sort of
symmetry. However, a method developed by Rappaport [19] allows for a completely arbitrary geometry
to be specified. This synthesis procedure will be described in this section, and in the next section it will
be applied to reflectarrays.
The Rappaport synthesis method, pictured in Fig. 3.2, is an iterative ray-tracing procedure in which
a set of parameters and initial conditions must be set, and from these conditions the reflectors grow.
The results of the synthesis are two sets of coordinates and normal vectors, one set for the main reflector
and the other for the sub-reflector, which can be used to form the surfaces of both reflectors. Although
reflectors are three-dimensional structures, the Rappaport synthesis takes place in the xz-plane and is
later extended to three dimensions.
The synthesis procedure proceeds as follows. First some parameters must be specified. These include
the coordinates of the two foci, FA and FB , and the respective scan angles ±α of the beams produced
from them. The path length L from each focal point to its corresponding scanned aperture plane must
also be specified, and this must be equal for both foci. Finally, as a starting point to grow the reflectors,
an initial sub-reflector coordinate S0 and its normal vector n̂S0 must be chosen.
After these parameters have been specified, a ray r1k is traced from FA to Sk , where the subscript k
indicates the iteration number. For the first iteration, Sk and the normal vector at this point are inputs
to the procedure. The reflected ray direction is given by the vector form of the law of reflection from a
Main reflector
r3k
Mk +α
−α
n̂mk
r03k
r02k
Sub-reflector
r2k
x n̂sk+1 Sk+1
r01k
n̂sk Sk
FB
r1k
y z
FA
Figure 3.2: Rappaport’s offset dual-reflector synthesis [19].
planar surface
r̂r = r̂i − 2 (n̂ · r̂i ) n̂, (3.1)
where r̂r is the unit vector of the reflected ray, r̂i is the unit vector of the incident ray, and n̂ is the
normal vector of the surface at the point of incidence.
The reflected ray r2k , can be determined from (3.1) and the total transmit ray path length. This
gives the next main reflector point Mk . The scan angle α of the output ray determines r̂3k , which can
then be used with r̂2k in (3.1) to solve for the required normal vector n̂mk on the main reflector at Mk .
The next step of the synthesis procedure takes placed in the receive mode. Tracing a ray r̂03k from
the direction of the received ray to the new main reflector point Mk and applying (3.1) as before gives
the ray r̂02k , and together with the total receive path length this gives the new sub-reflector point Sk+1 .
The subscript on the sub-reflectarray point has been incremented, as this new point will be used for the
next iteration. The last ray r01k is then traced from Sk+1 to FB , and the required sub-reflectarray normal
at Sk+1 can be computed from (3.1), as r̂02k and r̂01k are known. The next iteration can be started by
tracing a new ray from FA to Sk+1 , and the synthesis procedure continues until the required reflector
sizes have been reached.

x x
dΦmk
θm θm
dxm
(Mxk , Mzk ) n̂m
r 3k θr
θb θi α θb
r 2k r2k
θs θs
θoff θr θoff
(Sxk , Szk )
n̂s θ
i
FB r 1k dΦs (Sk ) FB
dxs
FA FA
z z
(a) (b)
x x
θm dΦm (Mk ) θm
dxm
n̂m
θi θr
θb θb
α
r03k
r02k
r02k θs θi θs dΦs (Sk+1 )

(Sxk+1 , Szk+1 )
n̂s θr dxs
θoff θoff r01k
FB FB
FA FA
z z
(c) (d)
Figure 3.3: Iterative ray-tracing procedure. (a) A ray is traced from FA to a known point on the sub-
reflectarray, and the phase derivative at this point and the intersection of the main reflectarray with
the z-axis are used to find the next main reflectarray point. (b) The desired ray direction is used to
determined the required phase derivative at the new main reflectarray point. (c) The phase derivative on
the main reflectarray is used to reflect a ray incident at −α which determines the next sub-reflectarray
point. (d) A ray is traced from the new sub-reflectarray point to focus FB to complete the iteration.
The new sub-reflectarray point and phase derivative are used with step (a) to start the next iteration.
3.3 Bifocal Reflectarray Synthesis

Now let us consider a case where both of the above curved reflectors are replaced by reflectarrays to
form a dual-reflectarray antenna. Dual-reflectarray antennas, or reflectors with a reflectarray as the
sub-reflector, have been investigated for various applications in recent years, although not as much as
their single surface counterparts. In one work, two reflectarrays in a folded configuration were introduced
to enable control of both the amplitude and phase on a main reflectarray [27] to produce a sector beam,
whereas normally the amplitude distribution is set by the feed and the reflectarray can only impart a
phase shift. A reconfigurable sub-reflectarray was proposed to feed a curved reflector as a concept for
achieving electrical beam steering on an earth observation satellite [45], a similar method to which is
feeding a main paraboloidal reflector with an electrically reconfigurable lens [46]. As dual-reflectarrays
have two surfaces on which their phase distributions can be adjusted they also offer more degrees of
freedom to improve performance, such as by adjusting both phase distributions to improve bandwidth
and cross-polarization [47].
Rappaport’s bifocal synthesis can be applied to reflectarrays by modifying the law of reflection used.
In the original procedure, the rays are reflected using the standard law for plane-wave reflection at a
planar interface, i.e. the rays are reflected in the specular direction. Although the reflectors are curved,
it is assumed that they can be treated as being locally planar. However, the situation changes if we
consider reflection from an array of elements spaced periodically with a phase gradient, as described
in Section 2.2.2. In this case the angles of incidence and reflection are related to the derivative of the
phase gradient along the surface, as well as the wavenumber of the medium through which the waves
propagate, as described by (2.26). In this chapter we will assume that the reflectarrays are surrounded
by free space, and so k = k0 in (2.26).
The bifocal synthesis as applied to reflectarrays is illustrated in Fig. 3.3. To start the synthesis
several initial conditions are required: the locations of the two foci FA and FB , an initial point on
the sub-reflectarray S0 and the phase derivative dΦs /dxs at this point, the desired scan angles from
each focus ±α with respect to a beam axis, the point of intersection of the main-reflectarray with the
z-axis Mx0 , and the tilt angles of the feed cluster, sub-reflectarray, main reflectarray, and beam axis,
which are θoff , θs , θm , and θb , respectively. In the most general version of this synthesis the tilt angles
of both reflectarrays, feed positions, beam axis, and scan angles can all be arbitrarily set without any
requirements on symmetry.
The synthesis procedure begins by launching a ray r1k from the lower focal point FA to the sub-
reflectarray point Sk . On the first iteration Sk and dΦs /dxs at this point must be specified. At Sk the
angle of incidence θi can be determined by taking the dot product of the incident ray direction r̂1k and
the sub-reflectarray normal vector n̂s . As the phase derivative at Sk is known, (2.26) can be used to
determine the angle of reflection θr . Together with the tilt angle of the main reflectarray θm and its
intersection point with the z-axis, this gives the reflected ray r2k . The new main reflectarray point is
then Mk = Sk + r2k . Next, given the desired output ray direction r̂3k and the incident ray direction
r̂2k at Mk , the angles of incidence and reflection can be determined using the main reflectarray normal
vector n̂m . (2.26) can then be re-arranged to solve for the required main reflectarray phase derivative
dΦm /dxm at Mk .
Now analyzing the antenna system in receive mode, a ray r03k is launched from an angle −α from the
beam axis to the point Mk . The phase derivative at this point gives the reflected ray direction r̂20 k , and
the point at which the ray is incident on the sub-reflectarray can be determined from θs and S0 to give
the new sub-reflectarray point as Sk+1 = Mk + r02k . As before, the subscript has been incremented to
indicate that this point will be used in the next iteration.
Finally, a ray r01k is traced from Sk+1 to the second focal point, FB . Using the incident and reflected
ray directions r̂20 k and r̂10 k with n̂s , the required phase derivative at Sk+1 is determined using (2.26).
Specify FA , FB ,
dΦs0 (S0 )
S0 , ,
dxs
θm , θs , θb , α
Trace rays from FA

to +α direction to
dΦm (Mk )
get Mk ,
dxm
Trace rays from −α

direction to FB to
dΦs (Sk+1 )
get Sk+1 ,
dxs
Are RAs
large
No enough?
Yes
Interpolate
dΦs dΦm
and
dxs dxm
Numerically integrate
dΦs dΦm
and
dxs dxm
Truncate RAs
to required size
Figure 3.4: Flow chart of the bifocal reflectarray synthesis procedure. RA is an abbreviation for reflec-
tarray.
Tracing a ray from FA to Sk+1 , the iterative procedure continues until the desired reflectarray sizes have
been reached. The obtained phase derivatives can then be interpolated and numerically integrated to
determine the required phase distributions for the reflectarrays, and the reflectarrays can be truncated
to fit the requirements of the application. A flow chart of the synthesis procedure is given in Fig. 3.4.
3.4 Application to Multi-beam Problem

While the original bifocal synthesis procedure was motivated by increasing the scan range of reflector
antennas, here we will apply it to something akin to the opposite problem. We will first consider a
two-dimensional case where beams are scanned in the xz-plane, as this is where the bifocal synthesis
procedure is performed. To assist with explaining the problem, a geometry and some requirements will
be provided.
Consider a prime-focus parabolic reflector with a diameter of 1.8 m with an F/D ratio of 1.54 oper-
ating at 20 GHz which is fed by a planar feed cluster, as shown in Fig. 3.5. While other arrangements
of feeds are possible, for example feeds which are positioned on a curved surface, a planar feed cluster
is used in this thesis as this is the most feasible case for space applications. For the application being
considered, the feed horn diameter to produce good spillover efficiency on the reflector will be assumed to
be 55 mm, and the antenna system must produce beams that are separated by 0.56◦ . The beam spacing
is realistic for the application under consideration, as is the gain for an aperture of this size [3]. The feed
size is based on feedback from Prof. José Encinar at the Technical University of Madrid (UPM). The
patterns on the right-hand side of Fig. 3.5 were generated using TICRA GRASP, a commercial physical
optics solver used for simulating reflector antenna problems.
In a SFB system a feed cluster is used to generate multiple beams. The feeds are not phased together,
and each element of the cluster simply generates its own independent beam. In this scenario the closest
that the feeds can be placed together is such that are touching. That is to say that the minimum feed
displacement is the diameter of the feed, which in this case is assumed to be 55 mm. For the case shown
here, this results in a set of beams that are separated by about 1.12◦ , as shown in Fig. 3.5a. However,
from our requirements we need to generate beams that have an angular separation that is half of this.
If we reposition each feed in the cluster such that the required 0.56◦ beam spacing is produced, it turns
out that each feed must be displaced by 26.5 mm as can be seen in Fig. 3.5b. However, as this is smaller
than the horn diameter the feeds would have to overlap, which is not physically possible.
In a sense there are two factors that contribute to the minimum achievable beam spacing for this
type of system: the size of the feeds, and the F/D ratio of the reflector. The beam scanning behaviour
of reflector antennas has been investigated in detail, and for cases where minor deviations are made from
the focal point approximate expressions can be developed that relate the angular displacement of the
feed to the angular displacement of the beam through the beam deviation factor (BDF) [36]. As the F/D
ratio of the reflector is increased, for a constant feed displacement the scan angle of the resulting beam
will decrease. This seems to suggest that the angular separation of the beams could be reduced by using
a larger F/D reflector. However, as the reflector diameter is fixed by the required gain and beamwidth
for the application, using a larger F/D will result in a longer focal length antenna. To maintain good
spillover efficiency more directive feeds must be used which will typically have larger apertures, and so
the beam spacing again becomes limited due to the feed sizes.
As outlined in Section 1.2, there are two general solutions to correct the beam spacing: using SFB
the beams from multiple reflectors and feed clusters can be interleaved, or a MFB architecture can be
adopted which uses a phased array as the feed for a single reflector. Of the two methods, SFB is simpler
to implement as it does not require the design of a complex BFN, but using multiple reflectors is a major
drawback as they add mass and volume in an environment where these factors are highly constrained. A
more desirable implementation of SFB would be one in which a single aperture can be used to generate
the correct multiple-beam pattern.
55.0 mm feed displacements

60
+θ z
Copolar gain [dBi]

40
1.80 m
−θ
20
-20
-5 -2.5 0 2.5 5
2.77 m Theta [deg.]
(a)
x
27.5 mm feed displacements

60
+θ z
Copolar gain [dBi]
40
1.80 m
−θ
20
-20
-5 -2.5 0 2.5 5
2.77 m Theta [deg.]
(b)
Figure 3.5: An example prime-focus paraboloidal reflector. (a) The feed horns are placed as close as
physically possible such that each feed is displaced by its diameter. (b) The feed displacements are
adjusted to give the required beam spacing, resulting in feeds that are physically overlapping. The +θ
and −θ arcs in the left-hand side of (a) and (b) indicate the conventions used for the positive negative
angular directions shown in the pattern plots.
Let us now revisit the bifocal synthesis procedure of Fig. 3.3. As inputs to the synthesis procedure,
the locations of two focal points FA and FB are required, as well as the directions that beams will be
scanned from these points, ±α. Considering the five feed case shown in Fig. 3.5, the problem can be
re-stated in a way that uses these variables. If we let the phase centres of the feeds at the edges of the
cluster, located at the top and bottom, be FA and FB , they scan at angles ±α as a result of the reflector
curvature. These scan directions are larger than we would want them to be. However, since the bifocal
synthesis allows us to specify α, we can choose the angles that the two foci scan at, which should allow
us to correct the scan angles from their ‘natural’ locations. Assuming that a uniformly spaced cluster of
feeds will produce a fairly uniformly spaced set of beams, our proposed concept is that by specifying the
scan angles of the beams produced from the two foci, that the angular spacing of the full set of beams
can also be specified. Previous work indicates that for a bifocal antenna the beams resulting from feeds
placed in between the two foci should also have good performance [19, 44], and so the resulting antenna
should be able to provide our required coverage pattern. Although this introduces a sub-reflector, this
would reduce the number of large main apertures required. To facilitate comparisons later, changing
the beam scan angles from their original locations will be referred to as ‘beam compression’, and we can
define a beam compression factor (BCF) to be
∆θsf
BCF = , (3.2)
∆θdes
where ∆θsf is the angular separation of the beams for the single-focus design, and ∆θdes is the desired
separation. For example, in Fig 3.5 for the single-focus case the beams have an angular separation of
1.12◦ , but we require them to to be spaced by 0.56◦ . This gives a BCF of 2.
The application of the bifocal synthesis to offset reflectarrays has been presented in related works
where a reflectarray was synthesized from an equivalent shaped reflector [48], and where a reflectarray
was synthesized directly and applied to the same multi-beam problem [49, 50].
3.5 Prime-focus Geometry with BCF = 2

3.5.1 Reference Case
Let us again consider the example geometry of the previous section, but using a reference case that
we can start from and use to compare with the bifocal results. As the synthesis procedure requires
two reflectors, this can be achieved using a dual-reflectarray. In this configuration a sub-reflectarray is
introduced to maintain the same F/D as a single reflector prime-focus configuration, but with a reduced
profile. The geometry that will be used as a starting point for the synthesis procedure and as a reference
is shown in Fig 3.6. Numerical values for the geometric parameters are given in Table 3.1.
As the sub-reflectarray size is a significant fraction of the main reflectarray size it will introduce
blockage. This will not be accounted for as it is outside of the scope of this thesis, but for more practical
applications the bifocal synthesis can be extended to dual-offset reflectarray configurations which reduce
the effects of blockage. If only one polarization is required, the sub-reflectarray can additionally be
made polarization-selective [27]. This allows for the co-polarized fields radiated by the reflectarray
to pass through the sub-reflectarray, ideally eliminating the effects of blocking a portion of the main
reflectarray aperture.
In Fig. 3.6 the real and virtual foci FR and FV are equidistant from the sub-reflectarray, which
imposes a flat phase distribution on it. The focal point of the main reflectarray is FV which gives it a
parabolic phase profile. The resulting phase distributions for this configuration are shown in Fig. 3.7a
and Fig. 3.7b. The radiation pattern as computed from an array factor analysis from this set of feeds
Parameter Value
Dm 1.80 m
dms 2.27 m
df s 0.50 m
Ds 0.68 m
F/D 1.54
Table 3.1: Numerical values for the prime-focus reference geometry shown in Fig. 3.6.
Main Feed Sub-

reflectarray cluster reflectarray
Ds
Dm z FR FV
df s df s
dms
Figure 3.6: Symmetric centre-fed dual reflectarray geometry. The locations of the real focus (FR ) and
virtual focus (FV ) are shown for the reference geometry and are equidistant from the sub-reflectarray.
Feed Position (z, x) [mm] Colour

F1 (1770, −165) Black
F2 (1770, −110) Magenta
F3 (1770, −55) Cyan
F4 (1770, 0) Purple
F5 (1770, 55) Red
F6 (1770, 110) Green
F7 (1770, 165) Blue
Table 3.2: The feed positions and naming scheme used, and the colours used to plot the resulting beams
and amplitude distributions.
is shown in Fig. 3.8. A list of the feed positions and their corresponding colours in the pattern and
illumination plots is provided in Table 3.2, along with a numbering scheme which will be used later. The
feeds were modelled using (2.69) with q = 28. This q factor was selected to match the simulated −12 dB
beamwidth of a horn antenna designed by Prof. José Encinar’s group at UPM, which is around 36◦ .
There are several observations to be made from Fig. 3.8. First, the gain is much lower than what
would be expected for an aperture of this size. Looking at Fig. 3.5, the gain for a 1.8 m reflector should be
greater than 50 dB, but the peak gain is 28 dB in Fig. 3.8. This is because the dual-reflectarray is being
analyzed in 2D, in order to facilitate comparisons with the results of the bifocal synthesis procedure,
whereas the prime-focus solid reflector was analyzed in 3D. The two-dimensional analysis method for
dual reflectarrays was described in Section 2.4.3.
Another point of note is that the gain of all the beams is fairly constant. For the results shown in this
chapter the reflectarrays are assumed to be lossless, and so there will be three factors that contribute to
gain reduction: phase errors, the amplitude distribution on the main reflectarray, and spillover losses.
In this configuration the sub-reflectarray has been sized to maintain good spillover efficiency, and with
a flat phase distribution the sub-reflectarray does not provide any magnification of the feed, so spillover
is not an issue. However, while the gain is flat across the feeds, aberrations are introduced to the beams
1 0
Phase [deg.] 0.5 -1000
Phase [deg.]
0
-2000
-0.5
-3000
-1
-0.4 -0.2 0 0.2 0.4 -1 -0.5 0 0.5 1
x [m] x [m]
(a) Sub-reflectarray (b) Main reflectarray
0 0
|Ei|, normalized [dB]

-5 -5
-10 -10
-15 -15
-20 -20
-0.3 -0.2 -0.1 0 0.1 0.2 0.3 -0.5 0 0.5
x [m] x [m]
(c) Sub-reflectarray (d) Main reflectarray
Figure 3.7: The phase and amplitude distributions on the reflectarrays for the prime-focus reference
dual-reflectarray shown in Fig 3.6.
30
20
10
Gain [dB]
-10
-20
-30
-5 -4 -3 -2 -1 0 1 2 3 4 5
Theta [deg.]
Figure 3.8: The beams produced by the reference dual-reflectarray using a 55 mm feed spacing.
produced as the feeds are moved further from the focal point.
The normalized amplitude distributions from the set of feeds are shown in Fig. 3.7c and Fig. 3.7d
for both the sub- and main reflectarrays, respectively. The edge illumination on the sub-reflectarray is
always below −10 dB, and on the main reflectarray it only rises to slightly above −10 dB, showing that
most of the power from the feeds is intercepted by the two reflectarrays.
Finally, the sizes of the reflectarrays are quite large. At 20 GHz the free-space wavelength λ0 is
15 mm, meaning that the diameter of the main reflectarray is 120λ0 , and the sub-reflectarray 40λ0 .
Fabricating the full-size structures would pose both electrical and mechanical challenges that are outside
of the scope of this thesis, as the investigation here is only into whether the bifocal synthesis can be
applied to correct the beam spacing of our reference case. However, investigations have been done into
constructed large reflectarrays by constructing the full structure out of multiple panels [51, 25, 52] or
using a curved surface [53, 54, 26].
3.5.2 Bifocal Reflectarray Synthesis

From the patterns shown in Fig. 3.8 the beams are spaced at 1.12◦ . In this section we will attempt
to correct this by applying the bifocal synthesis with BCF = 2 to the cluster of seven feeds shown in
Fig. 3.6. As shown in Fig. 3.3 and Fig. 3.4 there are a number of initial conditions for the synthesis
procedure, and so variations on these will be considered to see how performance is affected. In this thesis
only the geometry of Fig. 3.6 will be considered, and so θs = θm = θb = 0◦ for all cases.
To produce the required 0.56◦ beam spacing, in the feed cluster the feeds which are adjacent to the
central one should produce beams at ±0.56◦ , the feeds which are displaced two feed-diameters from the
centre must have their beams scanned at ±1.12◦ , and so on. Initially, let us assign FA and FB to be
the feeds displaced two feed-diameters from the centre, which means that α = 1.12◦ should be used for
the synthesis procedure. To generalize this, for a row of equally spaced feeds where the foci are located
symmetrically about the central feed, assuming that the beams produced by the feed cluster are spaced
uniformly
α = Nf α0 , (3.3)
where Nf is the number of feeds away from the centre that each focus is located at, and α0 is the desired
beam spacing. For the BCF = 2 case, α0 = 0.56◦ regardless of how far apart the foci are spaced.
The initial sub-reflectarray point will be chosen to lie at the intersection of the sub-reflectarray
with the z-axis, and have an initial phase derivative of 0 rad/m. This means that the initial ray will be
reflected in the specular direction as per (2.26). The initial conditions are summarized in Table 3.3. Four
iterations of the synthesis procedure are shown in Fig. 3.9, and overlay of where the reflectarrays and feed
horns would be placed is provided in Fig. 3.9e. Note that this many iterations are not actually required,
as is visible from Fig. 3.9e, as we only need to generate enough points to synthesize a reflectarray of the
desired size. The required reflectarray sizes are given in Table 3.1. As the geometry is symmetric, only
the top half of each reflectarray is synthesized. The phase derivative obtained from the top half of the
reflectarray is then mirrored to the bottom half to obtain the phase derivative for the entire reflectarray.
From the rays traced, the phase derivative at each of the points was solved for using (2.26). These
phase derivatives are plotted in Fig. 3.10a and Fig. 3.10b. As these points span a region larger than
the required sub-reflectarray and main reflectarray sizes, they were interpolated across the reflectarray
surfaces with a spacing of λ0 /2, which is 7.5 mm at 20 GHz. For the results shown in Fig. 3.10 cubic
spline interpolation was used, which was implemented using a built-in MATLAB function. These curves
were then integrated numerically using the MATLAB function cumtrapz, which performs cumulative
trapezoidal integration. The resulting phase distributions which were assigned to the sub- and main
reflectarrays are shown in Fig. 3.10c and Fig. 3.10d, respectively. Comparing these results to Fig. 3.7b
Parameter Value
FA F2
FB F6
α0 0.56◦
S0 (z, x) = (0, 0)m
dΦs /dx0s 0 rad/m
BCF 2
Table 3.3: Initial conditions used in Section 3.5.2 for the bifocal synthesis procedure.
and Fig. 3.7b it is visible that while the main reflectarray still has a parabolic phase profile, the sub-
reflectarray phase distribution has gone from being flat to being hyperbolic.
The resulting patterns are shown in Fig. 3.11c. The beam spacing has indeed been reduced from
1.12◦ to roughly 0.56◦ . However, when compared to Fig. 3.8 the peak gain has dropped from 28 dB
to 27.6 dB, and from the centre to edge feeds the gain rolls off by about 2.3 dB. Recall from Fig. 3.8
that for the reference case the gain was almost flat across all of the beams. The edge beams are also
quite distorted, which is a consequence of the highly asymmetric amplitude distributions on the main
reflectarray that the bifocal synthesis procedure has produced.
The illumination from the seven feeds is shown in Fig. 3.11a and Fig. 3.11b. The sub-reflectarray
amplitude distribution is the same as for the reference case (see Fig. 3.7c), but on the main reflectarray
the edge illumination level has gone from below −10 dB to around −4 dB for the centre feed. For the
edge feeds the amplitude taper is very asymmetric and the peak illumination level is actually on the
edge of the reflectarray. The high edge illumination levels indicate a large amount of spillover, which is
reflected in the gain reduction. Spillover efficiency is included in the gain calculations, and was computed
by assuming that the feed horn radiates 1 W and calculating the power radiated by the dual-reflectarray
from its far-field pattern using (2.115). The difference in the power radiated by the reflectarray and the
feed is then due to spillover, as element losses were not considered.
6 6
5 5
4 4
3 3
x [m]
x [m]
2 2
1 1
0 0
-1 -1
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
z [m] z [m]
(a) Iteration 1 (b) Iteration 2
6 6
5 5
4 4
3 3
x [m]
x [m]
2 2
1 1
0 0
-1 -1
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
z [m] z [m]
(c) Iteration 3 (d) Iteration 4
3
x [m]
1
Sub-
0 Main Feed
reflectarray
reflectarray cluster
-1
0 0.5 1 1.5 2 2.5
z [m]
(e) Iteration 4 with feed and reflectarray locations shown
Figure 3.9: Ray-tracing procedure for the upper half of the geometry shown in Fig. 3.6 with BCF = 2
and the foci located two feeds from the centre (F2 and F6 ). The blue lines are rays traced from FA
(transmit mode), the red lines are rays traced to FB (receive mode). The circles indicate the two foci,
and the points on the two reflectarrays obtained from the synthesis. The blue points were obtained from
the transmit rays, apart from the initial sub-reflectarray point, and the red from the receive rays. In (e)
the thick lines represent the main reflectarray and sub-reflectarray, and the feed positions are shown.
150 400
Phase derivative [rad/m]

100
200
50
0 0
-50
-200
-100
-150 -400
-1 -0.5 0 0.5 1 -6 -4 -2 0 2 4 6
x [m] x [m]
2000 0
1500 -1000
Phase [deg.]
Phase [deg.]
1000 -2000
500 -3000
0 -4000
-0.4 -0.2 0 0.2 0.4 -1 -0.5 0 0.5 1
x [m] x [m]
Figure 3.10: (a)–(b) The phase derivatives obtained from the ray-tracing procedure using the initial con-
ditions given in Table 3.3. The circles are the points which were obtained from the synthesis procedure,
and the curves are the result of interpolating over the reflectarray points. (b) The phase distributions
obtained by numerically integrating the phase derivatives.
0 0

-5 -5
-10 -10
-15 -15
-20 -20
-0.3 -0.2 -0.1 0 0.1 0.2 0.3 -0.5 0 0.5
x [m] x [m]
30
20
10
Gain [dB]
-10
-20
-30
-5 -4 -3 -2 -1 0 1 2 3 4 5
Theta [deg.]
(c)
Figure 3.11: The (a)–(b) amplitude distributions on the reflectarrays and (c) the beams produced when
the bifocal synthesis with the parameters given in Table 3.3 is applied to the prime-focus dual-reflectarray
with a cluster of 7 feeds.
3.6 Exploring the Synthesis Parameters

It was observed for the BCF = 2 case of the previous section that while the bifocal synthesis procedure
was able to compress the angular separation of the beams, it was not able to do this while maintaining
good spillover efficiency. However, as the synthesis procedure is dependent on a number of initial
conditions, in this section some parametric studies will be done to see how varying these conditions
affects performance.
In order to assess the performance of the synthesis procedure in a more quantitative manner, there
are some specific parameters that should be observed. The peak gain of each beam, and how fast the gain
rolls off as the feeds are displaced, will be useful as a given coverage pattern typically has requirements
on antenna gain in order to satisfy the link budget. For our application it is desirable to have near
constant gain across all the beams. Related to this, the spillover efficiency and edge illumination on the
main reflectarray are important. For typical reflector and reflectarray designs it is desired for the edge
illumination to have a maximum level of around 10 dB below the peak illumination level in order to have
good spillover efficiency without sacrificing taper efficiency [21]. Finally, the location of each beam is
important, as the beams must have a specific spacing in order to produce the required coverage pattern.
To observe this, a beam pointing error (BPE) will be introduced that is simply the absolute value of the
difference between the desired beam location θdes and the actual peak location θp :
BPEi = |θdes,i − θp,i | . (3.4)
The subscript i indicates the beam/feed number, as the BPE will vary for each beam. The beam/feed
numbering scheme is provided in Table 3.2. Although there are many parameters in the geometry of
the antenna system that could be explored, such as the locations of the feeds, initial sub-reflectarray
point, and initial z-position of the main reflectarray, as well as the offset angles of the feed cluster and
the two reflectarrays, these parameters will not be explored in this section. In a practical application
these geometric parameters are likely set by mechanical constraints, or how space is allotted to other
electrical/mechanical systems on the satellite.
3.6.1 Selection of Different Foci

Considering the BCF = 2 case with seven feeds, there are several options for the selection of the two
foci required as inputs to the bifocal synthesis procedure. Considering only cases where the foci FA and
FB are placed symmetrically about the axis of the centre feed these are:
• One feed from centre, F3 and F5 .
• Two feeds from centre, F2 and F6 , where each focus is located one feed away from the edge feed.
This was the configuration used in Section 3.5.2.
• Three feeds from centre, F1 and F7 , where each focus is located on the edge of the feed cluster.
Combinations of the above could be selected, but these will not be considered here as the beams for this
application are to be scanned symmetrically about the z-axis. The initial conditions used are summarized
in Table 3.4.
The results of performing the synthesis procedure with different foci separations are shown in Fig. 3.12
and Fig. 3.13. From Fig. 3.12a it is visible that there is more variation in the phase derivative obtained
Parameter Value
FA Swept, F1 to F3
FB Swept, F7 to F5
α0 0.56◦
S0 (z, x) = (0, 0)m
dΦs /dx0s 0 rad/m
BCF 2
Table 3.4: Initial conditions used for the bifocal synthesis procedure in Section 3.6.1.
on the sub-reflectarray as the spacing between the two focal points is increased. Compared to the case
when the foci are located two feeds from the centre, in the resulting sub-reflectarray phase distributions
the phase range required increases by around 500◦ when the foci are located three feeds from the centre.
However, there is not much of a difference between the cases where the foci were located one or two
feeds from the central feed. For the main reflectarray (see Fig. 3.12b and Fig. 3.12d) there appears to
be little variation between the three cases. Finally, looking Fig. 3.12e and Fig. 3.12f the illumination on
the main reflectarray is similar for all three cases, although for the case when the foci are located three
feeds from centre the edge illumination level is higher for the edge feeds and lower for the central ones
compared to the case when the foci are located two feeds from the central one. The illumination results
are only shown for one half of the feeds (F1 to F4 ) due to symmetry.
The radiation patterns in Fig. 3.13 show that the sidelobe levels of the beams appear to drop as the
foci separation is increased. The case when the foci are located two feeds from the centre has the most
tightly spaced beams, although the different in beam spacing is not very large. The gain roll-off between
the beams is similar for all of the cases.
Overall, it appears that using different foci separations does not have a large impact on the results of
the bifocal synthesis procedure, especially regarding the high main reflectarray edge illumination levels,
which was the main problem with the results of Section 3.5.2. Thus, the original spacing between the
focal points was used for the rest of the results.
3.6.2 Initial Phase Derivative

One of the inputs to the ray-tracing synthesis procedure is the first point on the sub-reflectarray and its
phase derivative. By (2.26), as in Rappaport’s original synthesis method where the intial sub-reflector
normal vector is specified, specifying the initial phase derivative determines the direction of the first
reflected ray. If it is set to zero the ray will simply be reflected in the specular direction, but if a non-
zero value is enforced this can be adjusted. The phase derivative of the initial sub-reflectarray point was
swept from −30 rad/m to +30 rad/m, and the other initial conditions are given in Table 3.5.
The results are summarized in Fig. 3.14. Due to the symmetry of the geometry, the trend as the
initial phase derivative is swept is only observed for the lower portion of the feed cluster (feeds F1 to F4 ).
It can be observed from Fig. 3.14a that pointing errors were introduced by choosing phase derivatives
other than 0 rad/m. Looking at Fig. 3.14b the peak gain for some of the beams can be increased slightly,
possibly due to increased spillover efficiency as seen in Fig. 3.14d, but at the more extreme values of the
phase derivatives considered the peak gain of each beam is reduced.
The phase distributions on both reflectarrays are shown in Fig. 3.15 for several values of the initial
sub-reflectarray phase derivative. It is visible that as the initial phase derivative is swept from negative
Phase derivative [rad/m] 200 400

Foci 1 from center
Foci 2 from center
100 Foci 3 from center 200
0 0
Foci 1 from center

-100 -200 Foci 2 from center
Foci 3 from center
-200 -400
-1 -0.5 0 0.5 1 -6 -4 -2 0 2 4 6
x [m] x [m]
(a) (b)
2500 0
2000
-1000
Phase [deg.]
Phase [deg.]
Foci 1 from center
1500 Foci 2 from center
-2000 Foci 3 from center
1000
Foci 1 from center -3000
500 Foci 2 from center
Foci 3 from center
0 -4000
-0.4 -0.2 0 0.2 0.4 -1 -0.5 0 0.5 1
x [m] x [m]
(c) (d)
0 0
-5 -5
Foci 2 from center
Foci 2 from center
Foci 1 from center
-10 -10 Foci 3 from center
F1
F1
F2
F2
-15 F3 -15
F3
F4
F4
-20 -20
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1
x [m] x [m]
(e) (f)
Figure 3.12: Comparison of the bifocal synthesis results when different separations between the focal
points are used.
Parameter Value
FA F2
FB F6
α0 0.56◦
S0 (z, x) = (0, 0)m
dΦs /dx0s Swept, −30 rad/m to 30 rad/m
BCF 2
Table 3.5: Initial conditions for Section 3.6.2

30
20
10
Gain [dB]
-10
Foci 2 from center

-30
-5 -4 -3 -2 -1 0 1 2 3 4 5
Theta [deg.]
(a)
30
20
10
Gain [dB]
-10
Foci 2 from center
-30
-5 -4 -3 -2 -1 0 1 2 3 4 5
Theta [deg.]
(b)
Figure 3.13: Comparison of the patterns produced using different foci spacings.
to positive that the main reflectarray focal length decreases, which is visible from the shallower phase
distributions. At values away from 0 rad/m the sub-reflectarray phase distribution becomes asymmetric,
even though the geometry is symmetric.
Some sample patterns from this investigation are shown in Fig. 3.16. It can be seen that away from
the −10 rad/m and +10 rad/m cases the beams become distorted, especially for the positive initial
phase derivative values. For this reason, and as the maximum edge taper level could not be improved
significantly (see Fig. 3.14c), an initial phase derivative of 0 rad/m was kept. However, perhaps this
initial condition could play a bigger role in offset configurations that do not have so much symmetry.
3.6.3 Sweeping the BCF

For the reference case considered in Section 3.5.1 it was found that the BCF needed to satisfy the
requirements of our application is two. However, this produced a solution with poor spillover efficiency,
and varying several of the synthesis parameters while keeping the geometry fixed did not yield any
0.4 28
F1
0.3
Peak gain [dB]

F2
BPE [deg.]
F3 26
F4
0.2 F1
F2
24
0.1 F3
F4
0 22
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30
Inital phase derivative [rad/m] Inital phase derivative [rad/m]
(a) (b)
Maximum edge illumination [dB]
0 90
Spillover efficiency [%]

-1
80
-2
70
-3
60
-4
-5 50
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30
Inital phase derivative [rad/m] Inital phase derivative [rad/m]
(c) (d)
Figure 3.14: Performance of the bifocal synthesis procedure as the initial phase derivative is swept using
the initial conditions given in Table 3.5.
2500 0
-30.0 -5.0 20.0
-500 -25.0 0.0 25.0
-20.0 5.0 30.0
2000 -15.0 10.0
-1000 -10.0 15.0
Phase [deg.]
Phase [deg.]
1500 -1500
-2000
1000 -2500
-3000
500
-3500
0 -4000
-0.4 -0.2 0 0.2 0.4 -1 -0.5 0 0.5 1
x [m] x [m]
Figure 3.15: Phase distributions on the sub- and main reflectarrays as the initial phase derivative is
swept. The values in the legend are given in rad/m.
significant efficiency improvements. Thus, it may be worthwhile to take a look at what happens when
a BCF not equal to two is considered. In this section the BCF will be swept over the range 0.4 to
2.5, while keeping the other synthesis parameters as they were in the original case of Section 3.5.2.
This corresponds to sweeping the beam spacing from 2.40◦ to 0.45◦ . Note that a BCF less than one
30 30
20 20
10 10
Gain [dB]
Gain [dB]
0 0
-10 -10
-20 -20
-30 -30
-20 -16 -12 -8 -4 0 4 8 12 16 20 -20 -16 -12 -8 -4 0 4 8 12 16 20
(a) −30 rad/m (b) −20 rad/m
30 30
20 20
10 10
Gain [dB]
Gain [dB]
0 0
-10 -10
-20 -20
-30 -30
-20 -16 -12 -8 -4 0 4 8 12 16 20 -20 -16 -12 -8 -4 0 4 8 12 16 20
(c) −10 rad/m (d) +10 rad/m
30 30
20 20
10 10
Gain [dB]
Gain [dB]
0 0
-10 -10
-20 -20
-30 -30
-20 -16 -12 -8 -4 0 4 8 12 16 20 -20 -16 -12 -8 -4 0 4 8 12 16 20
(e) +20 rad/m (f) +30 rad/m
Figure 3.16: Some selected patterns resulting from the bifocal synthesis as the initial phase derivative is
swept. The value of the initial sub-reflectarray phase derivative corresponding to each set of patterns is
given in the caption.
corresponds to beam ‘expansion’, whereby the beams to be synthesized will have a spacing greater than
the reference case, and a BCF equal to one means that the synthesized and reference beam spacings will
be the same. The synthesis parameters are tabulated in Table 3.6.
Some performance metrics of the BCF sweep are shown in Fig. 3.17. Again, due to symmetry only
the trend from the lower half of the feed cluster is observed (feeds F1 to F4 ). Looking at Fig. 3.17a
there is not much variation in the pointing error of the beams produced by the set of feeds as the BCF
is swept, apart from the edge one. There is more significant variation for the BCF = 0.40 case, but this
case also has the largest angular beam separation and so a similar percentage error as the other cases
would result in a larger BPE.
Parameter Value
FA F2
FB F6
α0 Swept from 2.40◦ to 0.45◦
S0 (z, x) = (0, 0)m
dΦs /dx0s 0 rad/m
BCF Swept from 0.4 to 2.5
Table 3.6: Initial conditions used for the bifocal synthesis procedure in Section 3.6.3.
0.8 29
F1
28
0.6 F2
Peak gain [dB]

BPE [deg.]
F3 27
F4
0.4 26 F1
F2
25
0.2 F3
24 F4
0 23
0.5 1 1.5 2 2.5 0.5 1 1.5 2 2.5
BCF BCF
(a) (b)
Maximum edge illumination [dB]
0 100
Spillover efficiency [%]
-20 80
F1
F1
-40 F2 60
F2
F3
F3
F4
-60 40 F4
-80 20
0.5 1 1.5 2 2.5 0.5 1 1.5 2 2.5
BCF BCF
(c) (d)
Figure 3.17: Some performance metrics of the bifocal synthesis procedure as applied to the configuration
of Section 3.5.2 as the BCF is swept from 0.45 to 2.5.
The other performance metrics are more interesting to look at. Apart from the BCF = 0.4 case,
it can be seen from Fig. 3.17b that except for the beam produced by the edge feed, the gain of each
beam is fairly close for BCF values in the range of 0.5 to 1.0. As the BCF increases beyond one, which
corresponds to enforcing a beam separation which is reduced compared to the reference case, the gain
of each beam begins to diverge. This means that the gain is rolling off between adjacent beams.
The gain roll-off can be partially explained by Fig. 3.17c and Fig. 3.17d. It can be observed from
these figures that as the BCF is increased beyond one, that the maximum edge illumination on the main
reflectarray begins to increase, until the peak amplitude is on its edge. This results in the trend on the
efficiency plot in which the spillover efficiency of each beam decreases monotonically after the BCF = 1
case. Additionally, the maximum edge illumination is higher and increases more quickly for the beams
produced by the feeds closest to the edge of the cluster, which explains the faster roll-off in the gain of
these beams as the BCF is increased.

The phase distributions for the range of swept BCF values are shown in Fig. 3.18. As the BCF is
swept from low to high, the shape of the sub-reflectarray phase changes from concave to convex, and the
main reflectarray focal length decreases. The concave sub-reflectarray phase distributions are analogous
to Gregorian dual-reflector antennas, and the convex ones to Cassegrain configurations. This will be
explored in more detail in the next section.
3000 2000
0.4 1.2 2.0
0.5 1.3 2.1
1000 0.6 1.4 2.2
2000 0.7 1.5 2.3
0.8 1.6 2.4
0 0.9 1.7 2.5
Phase [deg.]
Phase [deg.]
1.0 1.8
1000 1.1 1.9
-1000
0
-2000
-1000
-3000
-2000 -4000
-0.4 -0.2 0 0.2 0.4 -1 -0.5 0 0.5 1
x [m] x [m]
Figure 3.18: Phase distributions on the reflectarrays resulting form the bifocal synthesis procedure as
the BCF is swept.
Some selected patterns from the BCF sweep are shown in Fig. 3.19, where the variation in the beam
spacing can be clearly seen. It is visible that the BCF = 0.4 pattern is quite distorted, which may be
a result of phase errors and highly asymmetric amplitude tapers on the main reflectarray. Examining
the cases when the BCF is close to one, the BCF = 0.8 and BCF = 1.2 patterns, these show little
distortion and good peak gain flatness across all of the beams. As the BCF continues to be increased
beyond one, more pronounced gain roll-off can be observed due to the increasingly worse spillover shown
in Fig. 3.17d, and the patterns also begin to become distorted.
These results seem to suggest that for a given reference antenna system using the bifocal synthesis
procedure to either increase or decrease the beam spacing while holding the rest of the geometry constant
will result in gain losses due to phase errors, poor amplitude tapers, or spillover. However, small
variations around the reference beam spacing, for example the BCF = 0.8 and BCF = 1.2 cases, can be
made without incurring much performance degradation.
30 30
20 20
10 10
Gain [dB]
Gain [dB]
0 0
-10 -10
-20 -20
-30 -30
-20 -16 -12 -8 -4 0 4 8 12 16 20 -20 -16 -12 -8 -4 0 4 8 12 16 20
(a) BCF = 0.4 (b) BCF = 0.8
30 30
20 20
10 10
Gain [dB]
Gain [dB]
0 0
-10 -10
-20 -20
-30 -30
-20 -16 -12 -8 -4 0 4 8 12 16 20 -20 -16 -12 -8 -4 0 4 8 12 16 20
(c) BCF = 1.2 (d) BCF = 1.6
30 30
20 20
10 10
Gain [dB]
Gain [dB]
0 0
-10 -10
-20 -20
-30 -30
-20 -16 -12 -8 -4 0 4 8 12 16 20 -20 -16 -12 -8 -4 0 4 8 12 16 20
(e) BCF = 2.0 (f) BCF = 2.5
Figure 3.19: Some selected patterns resulting from the bifocal synthesis procedure as the BCF is swept.
3.7 Bifocal Reflectarray Synthesis with BCF = 1
In the previous section the bifocal synthesis procedure was applied with BCF = 2 to correct the beam
spacing for a given feed spacing. This indeed produced a compressed beam spacing, but with high
edge illumination on the main reflectarray, resulting in poor spillover efficiency. Let us now examine
what happens when the synthesis procedure is applied using the same angular beam spacing as the
reference case. It should be noted that the bifocal synthesis procedure has been applied to improve the
scan capabilities of reflectarrays in the past [44], but this section will offer some more detail which will
be useful to draw conclusions on this project. The initial conditions used in this section are given in
Table 3.7.
The phase derivatives resulting from the ray-tracing procedure with BCF = 1 are shown in Fig. 3.20a
Parameter Value
FA F2
FB F6
α0 1.12◦
S0 (z, x) = (0, 0)m
dΦs /dxs 0 rad/m
BCF 1
Table 3.7: Initial conditions for the bifocal synthesis procedure used in Section 3.7.
and Fig. 3.20b, which were integrated to obtain the phase distributions shown in Fig. 3.20c and Fig. 3.20d.
It is visible that the points on the main reflectarray are more densely spaced than the BCF = 2 case.
Comparing the upper halves (x > 0) of Fig. 3.20b with Fig. 3.10b, applying the synthesis procedure with
BCF = 1 gives four points within a 3 m span, whereas these four points are spaced across 6 m for the
BCF = 2 case. Additionally, the phase on the sub-reflectarray is concave rather than convex. It does
not follow a simple parabolic or hyperbolic curve, although the main reflectarray still has a parabolic
phase profile.
Looking at the amplitude distributions in Fig. 3.20e and Fig. 3.20f the illumination problem no
longer exists, and the incident field amplitude at the edge of the main reflectarray is similar to that of
the reference case. Fig. 3.21 shows an overlay of the beams from the reference single focus configuration,
and the BCF = 1 bifocal reflectarray. Although the central beams have been broadened, the coma lobe on
the edge beams has been reduced for the bifocal reflectarray. This illustrates the original purpose of the
bifocal synthesis, which is to reduce aberrations and scan loss for beams steered using feed displacement.
This characteristic was utilized in a related work using a side-fed Cassegrain geometry [55], which does
not suffer from blockage due to the sub-reflectarray as in a prime-focus configuration.
3.7.1 Comparison of BCF = 2 Case to a Cassegrain Reflectarray

As was observed earlier, the phase distributions of the BCF = 2 case shown in Fig. 3.10 appear to trace
out a parabola on the main reflectarray and a hyperbola on the sub-reflectarray. Considering classical
dual-reflector geometries, this is similar to a Cassegrain reflector antenna, where the sub-reflector is a
hyperboloid and the main reflector is a paraboloid [29].
Although (2.26) was used to synthesize the reflectarrays, for a set of rays incident on a surface with a
known or numerically determined phase derivative distribution, (2.26) allows for the angle of reflection
of each ray to be determined. If a plane wave is incident on the main reflectarray and the reflected ray
paths are traced, the focal point can be estimated as the region about which the reflected rays converge.
This procedure is shown in Fig. 3.22, where the approximate focal point of the main reflectarray is
visible. It is a region and not a single point as it would be the case for a single-focus phase distribution,
as there may be some deviations from a perfect parabolic phase distribution as well as some numerical
errors. The main reflectarray focal length was estimated to be 2.515 m. This has been reduced from
the focal length of the reference system, which was 2.770 m, corresponding to a reduction in the main
reflectarray F/D ratio from 1.54 to 1.40.
In the initial reference configuration the sub-reflectarray had a flat phase distribution, as its real
and virtual foci are equidistant from the sub-reflectarray. However, it is visible from Fig. 3.22 that this
is no longer the case, as applying the bifocal synthesis procedure has shifted the virtual focal point.
200 300

200
100
100
0 0
-100
-100
-200
-200 -300
-1.5 -1 -0.5 0 0.5 1 1.5 -3 -2 -1 0 1 2 3
x [m] x [m]
0 0
-50 -1000
Phase [deg.]
Phase [deg.]
-100
-2000
-150
-3000
-200
-0.4 -0.2 0 0.2 0.4 -1 -0.5 0 0.5 1
x [m] x [m]
0 0
-5 -5
-10 -10
-15 -15
-20 -20
-0.3 -0.2 -0.1 0 0.1 0.2 0.3 -0.5 0 0.5
x [m] x [m]
(e) Sub-reflectarray (f) Main reflectarray
Figure 3.20: Phase derivatives, and resulting phase and amplitude distributions on the reflectarrays with
BCF = 1.
Knowing the distances of the real and virtual foci from the sub-reflectarray its phase distribution can
be synthesized by applying (2.28). This yields a Cassegrain dual-reflectarray, which will be referred to
as the ‘similar Cassegrain’.
The phase distributions of the similar Cassegrain dual-reflectarray are shown in Fig. 3.23, with a
comparison to the BCF = 2 bifocal synthesis results. Using the ray-tracing method to estimate the
focal length, the main reflectarray phase distributions are almost the same for the bifocal and similar
Cassegrain cases. There are slight deviations on the sub-reflectarray, indicating that the bifocal sub-
reflectarray phase distribution is not purely hyperbolic. The differences could also be a result of the
qualitiative method which was used to estimate the focal length of the main reflectarray. Comparing
the patterns produced by the two methods in Fig. 3.24 shows that although there is a difference in the
30
20
10
Gain [dB]
-10
-20
-30
-5 -4 -3 -2 -1 0 1 2 3 4 5
Theta [deg.]
Figure 3.21: Beams produced by the bifocal reflectarray (solid) with BCF = 1 compared with the single
focus reference (dashed).
0.5
x [m]
-0.5
-1
0 1 2 3 4
z [m]
Figure 3.22: Tracing the rays from a plane wave incident at broadside to estimate the focal point of
the main reflectarray. The real focus is to the left of the sub-reflectarray, and located at F4 , and the
estimated virtual focus is to the right.
spacing of the beams, they both show a similar trend in terms of distortion on the outer beams and gain
roll-off.
From these results, it seems that applying the bifocal synthesis procedure with beam compression
results in phase distributions that are Cassegrain, or Cassegrain-like. This is perhaps not surprising,
as Rappaport used the Cassegrain geometry to extend his synthesis procedure to 3D, and also fit his
resulting reflector curves to find the ‘closest Cassegrain’ configuration [19].
The illumination of the main reflectarray for the similar Cassegrain is shown in Fig. 3.25. It exhibits
the same problems as the bifocal reflectarray with the edge taper level being too high on the main
reflectarray edges. To gain some insight into this behaviour, the antenna can be looked at from the
virtual feed perspective utilized to analyze dual-reflector antennas [29]. In this method, the curvature
of the sub-reflector is used to find an equivalent feed located at its virtual focus such that the entire
0 0
-500 -1000
Phase [deg.]
Phase [deg.]
Cassegrain
Bifocal
-1000 -2000
Cassegrain
-1500 -3000
Bifocal
-2000 -4000
-0.3 -0.2 -0.1 0 0.1 0.2 0.3 -0.5 0 0.5
x [m] x [m]
Figure 3.23: Comparison of the bifocal and similar Cassegrain reflectarray phase distributions.
30
20
10
Gain [dB]
-10
Bifocal
-20 Cassegrain
-30
-5 -4 -3 -2 -1 0 1 2 3 4 5
Theta [deg.]
Figure 3.24: Comparison of the beams produced by the bifocal and similar Cassegrain reflectarrays. The
solid lines are from the bifocal reflectarray, and the dashed from the Cassegrain.
reflector system can be analyzed as a single paraboloidal reflector. Depending on the curvature of the
sub-reflector, the beamwidth of the virtual feed can be wider, narrower, or the same as the real feed.
The original motivation behind the virtual feed method was to reduce the complexity of analyzing dual
reflector antennas, but we will only use it conceptually.
0
-5
-10
-15
-20
-0.5 0 0.5
x [m]
Figure 3.25: Amplitude distribution on the main reflectarray of the similar Cassegrain configuration.
Two dual-reflector geometries are shown in Fig. 3.26 with the real feed of the system being on the
left of the sub-reflector, and the virtual feed on the right. In Fig. 3.26a 2φR and 2φV are the subtended
angles of the real and virtual feeds, respectively. In the design of a dual-reflector antenna, the feed and
geometry would be selected to have a low illumination level at φR to maintain good spillover efficiency.
Fig. 3.26a shows that when a flat sub-reflector is used, the taper imposed by the real feed on the sub-
reflector is maintained by the virtual feed on the main reflector, due to the feeds being equidistant from
the sub-reflector. In other words, with a flat sub-reflector the real and virtual feeds have the same
beamwidth and φR = φV .
2φR 2φV
(a) Folded (b) Cassegrain
Figure 3.26: Two typical dual-reflector antenna geometries.
Now, consider the Cassegrain case shown in Fig. 3.26b. The vertex of the sub-reflector is located at
the same location as the flat reflector in Fig. 3.26a, and the real feed remains at the same position. Due
to the hyperbolic curvature of the sub-reflector the focal length of the main reflector is decreased, and
the virtual feed moves closer to the sub-reflector. This means that the virtual feed will have a larger
subtended angle than for the folded reflector, and that the virtual feed has a larger beamwidth than
the real feed (φV > φR ). Thus, to maintain the same edge taper on the main reflector as for folded
geometry, the beamwidth of the real feed must be reduced. This would imply that feed horn with a
larger aperture needs to be selected.
In Fig. 3.26b, the rays in red incident on the sub-reflector are the same as those used for the flat
sub-reflector of Fig. 3.26a, and correspond to using a feed of the same beamwidth. This shows that
if a feed with the same amplitude taper as for the flat sub-reflector is used, that these rays will be
reflected outside the area of the main reflector. The resulting edge taper on the main reflector will then
be higher than on the sub-reflector, indicating increased spillover. As the bifocal synthesis procedure has
transformed the reference dual-reflectarray into a configuration that is very similar to a Cassegrain dual-
reflectarray, without changing the characteristics of the feed, this explains the poor spillover efficiency.
If the feed beamwidth was to be narrowed to lower the edge taper on the sub-reflector, as shown by
the black rays in Fig. 3.26b, the feed sizes would typically increase. This brings us back to the original
problem whereby the minimum beam spacing is limited by the diameter of the feeds. Although in this
analysis curved reflectors were considered, the same behaviour would be observed by tracing the ray
paths for the dual-reflectarrays.
For the cases when beam compression was not enforced a Cassegrain-like phase distribution does not
result (see Fig. 3.20c and Fig. 3.18), and additionally the spillover efficiency for these cases is high. Thus,
it seems that the bifocal synthesis is not appropriate for the BCF = 2 case, as it produces solutions
with poor spillover performance. However, it is visible from Fig. 3.17d that a slight amount of beam
compression can be applied before the spillover efficiently becomes significantly degraded.
3.8 Extension of the Prime-focus Geometry to 2D Arrays

In the bifocal synthesis method proposed by Rappaport, the first iteration takes place in the xz-plane.
This is also the plane of symmetry of the antenna. To form a surface that will make up the reflector
in 3D, initial points off the plane of symmetry must be considered, and the synthesis procedure repeats
until enough points have been collected to perform interpolation and construct the surface. However,
as can be seen from the ray-tracing diagrams in the previous sections the points begin to spread in
the x-direction as the synthesis continues. For points away from the symmetry plane, the points also
spread in the y-direction. This does not provide a regular grid of points over which a surface or phase
distribution can be constructed, although interpolation could be applied. Rappaport acknowledged that
there is no guarantee that the set of points will be able to construct a physically realizable surface, and
it is a “difficult trial-and-error procedure” to do so [19], although this is not as much of an important
factor for reflectarrays as the result of the synthesis procedure is a phase distribution and not a curved
surface.
While it may be challenging to construct two-dimensional phase distributions from the one-dimensional
curves for an arbitrary geometry, things are simplified for the prime-focus case. For standard reflector
antenna geometries, the shapes making up the reflectors are typically paraboloids, hyperboloids, or el-
lipsoids. These are all surfaces of revolution formed by rotating a 1D curve about its axis of symmetry.
This is not completely restricted to symmetric prime-focus configurations, as offset geometries that are
often employed for practical reasons can be formed by only using a section of the full surface of revolu-
tion. As the prime-focus dual-reflectarray also has an axis of symmetry, we can rotate its 1D phase curve
to obtain a 2D phase curve which can be assigned to a 2D planar array. Although in this section only
prime-focus geometries will be considered, recent work on bifocal reflectarrays based on the Rappaport
synthesis procedure has extended the results of the 2D synthesis to 3D [50], and the synthesis has been
performed directly in three-dimensional space [56].
For the prime-focus case, the bifocal synthesis procedure results in a phase distribution ψ(x), and
the axis of symmetry of the antenna system is the z-axis. To rotate a point by an angle γ about the
z-axis, it can be multiplied by the rotation matrix
 
cos γ − sin γ 0
Rz (γ) =  sin γ cos γ 0 . (3.5)
 
0 0 1
To obtain the phase distribution for a 2D array, first ψ(x) was rotated about the z-axis by multiplying
each point in the distribution by (3.5) over a set of angles between 0 and π. The curves do not need
to be rotated to 2π, as the negative half of the phase distribution is already known through symmetry.
The points obtained from rotating the 1D curve at intervals of 0.1 rad are shown in Fig. 3.27.
2000 2000
1500
Phase [deg.]
Phase [deg.]
0
1000
-2000
500
0 -4000
0.2 0.5
0 0.2 0 0.5
0 0
-0.2 -0.2 -0.5 -0.5
y [m] x [m] y [m] x [m]
Figure 3.27: Points obtained by rotating the phase distributions of Fig. 3.10c and Fig. 3.10d
Rotating the 1D phase distribution forms a distribution of points arranged along a set of concentric
circles, whereas typically for a reflectarray the elements are laid out in a rectangular grid. That is, the
determined phases are not located where the reflectarray elements are to be placed. Thus, interpolation
using the MATLAB function griddata was used to determine the phase distribution over a rectangular
grid, and these interpolated values were assigned to the reflectarray. The interpolated phase curves
are shown in Fig. 3.28. The phase distributions wrapped to values between 0◦ and 360◦ are shown in
Fig. 3.29 to illustrate the large number of phase wraps that would occur if the reflectarray was fabricated.
Figure 3.28: Interpolated phase distributions.
A comparison of the beams generated by the 1D and 2D reflectarrays is shown in Fig. 3.30. Both
cases have been analyzed using the array factor method and assuming ideal phase shifters. The gain
has been normalized to the peak of the centre beam, as there will be a significant difference in absolute
gain between the two cases. For the 2D reflectarray, the elevation cut of the co-polar gain pattern
Figure 3.29: Wrapped phase distributions.
is being plotted. It is seen that the 2D case produces beams in the same location as the 1D case,
although there is a slight discrepancy in amplitude. This may be a result of different main reflectarray
illumination resulting from analyzing the 1D reflectarray in two dimensions, and the 2D reflectarray in
√
three dimensions. In 2D the amplitude of an electromagnetic wave drops with 1/ r, and in 3D 1/r,
where r is the distance from the source. For the 2D reflectarray, this would result in a lower illumination
level on the main reflectarray which would produce the slightly broader beams and lower sidelobes that
are visible in Fig. 3.30.
-10
-20
Gain [dB]
-30
-40
-50
2D reflectarray
1D reflectarray
-60
-5 -4 -3 -2 -1 0 1 2 3 4 5
Theta [deg.]
Figure 3.30: Comparison of the beams produced by the bifocal synthesis from a 1D reflectarray, and
when the phase distributions have been rotated to form a 2D reflectarray. For the 2D reflectarray, the
elevation cuts of the co-polar gain patterns are being plotted.
Contour plots of the beams from a 2D feed cluster are shown in Fig. 3.31 for the reference geometry
and the BCF = 2 case. In the 2D feed cluster, the same configuration of feeds is used as in Table 3.2,
but a second row of feeds is repeated with the same spacing along the y-axis. The contour levels shown
are at −3 dB and −5 dB from the peak of each beam. It is seen that the rotation method gives phase
distributions that provide focusing in two dimensions, forming pencil beams. The angular separation of
the beams has been reduced by applying the bifocal synthesis, but a reduction in gain is visible from
the lower levels of the contours in the bifocal case compared to the reference, as was the case for the
one-dimensional reflectarrays.
0.08 48 0.08 48
0.06 47 0.06 47
0.04 46 0.04 46
0.02 45 0.02 45
Gain [dBi]
Gain [dBi]
v [1]
v [1]
0 44 0 44
-0.02 43 -0.02 43
-0.04 42 -0.04 42
-0.06 41 -0.06 41
-0.08 40 -0.08 40
-0.05 0 0.05 -0.05 0 0.05
u [1] u [1]
(a) Bifocal (b) Folded reference
Figure 3.31: Gain contours in direction cosine space for the 2D bifocal reflectarray and its folded reference
geometry, where u = sin θ cos φ and v = sin θ sin φ.
3.8.1 Full-wave Simulation of the Sub-reflectarray

The results obtained in this chapter have so far relied on an idealized model, where the reflectarrays are
made up of ideal phase shifters which are able to provide the exact phase they are designed for with
no quantization error or generation of cross-polarization, and the feeds are able to function as perfect
point sources. It would be desirable to verify the results experimentally, but due to the shortcomings
of the bifocal synthesis procedure it was decided to not proceed with fabrication and measurement of
an experimental demonstrator, which would have to be scaled due to the large size of the reflectarrays
considered.
The next best option to experimental results is full-wave simulation. However, the large size of the
reflectarrays also makes them difficult to simulate using full-wave methods with commonly available
computational resources. It is possible to reduce the magnitude of the problem by exploiting symmetry,
but when the feed is displaced from the centre position the symmetry is lost.
While the main reflectarray is very large (240 × 240 elements), the sub-reflectarray is small enough
(80 × 80 elements) to explore full-wave simulations of it. After generating the 2D phase distributions by
rotating the 1D curves as done in the previous section, a layout was produced which consists of sets of
three co-planar dipoles on a single layer. The reflectarray unit cell is pictured in Fig 3.32, and values for
the geometrical parameters are given in Table 3.8. The dielectric used is Rogers DiClad 880B which has
r = 2.3 and tan δ = 0.005. The triple co-planar dipole unit cell is designed for wideband applications,
and variations of it have been used in several works [57, 58, 59]. Additionally, two-layer versions have
been explored for dual-polarization and dual-band applications [60, 61]. This unit cell will be revisited
in Chapter 4, where reflectarray unit cell design is discussed in detail. The layout was done by Eduardo
Parameter Value
px 7.5 mm
py 7.5 mm
lA2 0.78lA1
w 0.5 mm
sA 0.5 mm
Table 3.8: Geometrical parameters of the triple co-planar dipole reflectarray unit cell.
Martinez de Rioja of Prof. José Encinar’s group at the UPM, and the full layout and simulation mesh
of the bifocal sub-reflectarray is shown in Fig. 3.33. The meshing was performed using Gmsh [62].
px lA2 lA1 y
w
sA
py
1.524 mm
Figure 3.32: (a) Top view and (b) side view of the triple co-planar dipole reflectarray unit cell with
lA1 = 5 mm.
The sub-reflectarray mesh was used in a commercial MoM solver, ADF from Ingegneria Dei Sistemi
(IDS), to compute the incident field over the region over which the main reflectarray would be placed.
A λ0 /2 dipole feed was used due to its ease of implementation in ADF. The magnitude and phase of
the resulting x-component of the incident field, which is the co-polar component and parallel to the
dipoles, are shown in Fig. 3.34 and Fig. 3.35, respectively. For comparison, the incident fields computed
using the array factor method and assuming ideal phase shifters on the sub-reflectarray, as outlined
in Section 2.4.2, are also provided. For the array factor computations, a dipole-like excitation was
implemented by choosing q = 1 in the cosq θ feed model.
It is visible from these plots that the amplitude of the main reflectarray incident field is less uniform
for the MoM simulated case compared to the ideal phase shifter case, and there are regions where the
amplitude drops to less than 20 dB below peak. This is likely due to mutual coupling between the
Figure 3.33: (Left) The full meshed sub-reflectarray layout consisting of triple coplanar dipole unit cells,
and (right) a zoomed-in portion of the layout. The layout was done by Eduardo Martinez de Rioja from
Prof. Encinar’s group at UPM, and meshing was performed using Gmsh.
(a) F4 (b) F3 (c) F2 (d) F1
(e) F4 (f) F3 (g) F2 (h) F1
Figure 3.34: Normalized magnitude of the x-component of the incident field on the main reflectarray at
20 GHz for four different feed positions using the (a)–(d) array factor with ideal phase shifters on the
sub-reflectarray and a q = 1 feed model, (e)–(h) MoM with a λ0 /2 dipole feed and the real reflectarray
layout.
(a) F4 (b) F3 (c) F2 (d) F1
(e) F4 (f) F3 (g) F2 (h) F1
Figure 3.35: Phase of the x-component of the main reflectarray incident field at 20 GHz for four different
feed positions using (a)–(d) array factor with ideal phase shifters on sub-reflectarray and q = 1 feed
model, and (e)–(f) MoM with a λ0 /2 dipole feed.
Peak location [deg.] BPE [deg.] Directivity [dBi]

Feed
AF MoM/AF AF MoM/AF AF MoM/AF
F4 0.00 0.00 0.00 0.00 51.24 44.12
F3 0.56 0.53 0.00 0.03 51.26 47.03
F2 1.12 1.15 0.00 0.03 51.28 46.27
F1 1.76 1.76 0.06 0.08 51.17 45.83
Table 3.9: Comparison of the bifocal reflectarray performance between the ideal phase shifter AF and
MoM/AF simulation methods.
reflectarray elements, which would be captured using the MoM but not an AF-based method. For both
cases there are areas of very high illumination on the edge of the main reflectarray region. The variation
in phase is similar as the feed is displaced, although there appears to be additional distortion in the
fields from the feed furthest from centre for the real sub-reflectarray.
The complex incident field was interpolated over the main reflectarray elements, and treating the
elements as ideal phase shifters (2.55) was used to compute the reflected fields. (2.58) was then used to
determine the radiated fields, and the resulting principal plane directivity cuts are shown in Fig. 3.36,
overlaid on the patterns obtained using the array factor method with ideal phase shifters on both
reflectarrays. Directivity is being examined, as spillover losses cannot easily be quantified.
The results show a drop in directivity across the beams produced from all feed positions, although the
main beam is in the same direction for both simulation methods. Numerical values for the directivity,
and pointing direction and error are given in Table 3.9 for both simulation methods. The worst case drop
in directivity when the MoM is used to simulate the incident field on the main reflectarray is 7.12 dB,
which is for the beam produced by the centre feed. For all of the beams there are sidelobes that were
not present in the ideal case. The MoM/AF case also shows a variation in gain of around 3 dB across
all of the beams, whereas for the AF case the directivity is fairly uniform.
50 MoM/AF 50 MoM/AF
AF AF
Directivity [dBi]
Directivity [dBi]
40 40
30 30
20 20
10 10
0 0
-20 -10 0 10 20 -20 -10 0 10 20

Elevation [deg.] Azimuth [deg.]
(a) Beam from F4 , elevation cut (b) Beam from F4 , azimuth cut
50 MoM/AF 50 MoM/AF
AF AF
Directivity [dBi]
Directivity [dBi]
40 40
30 30
20 20
10 10
0 0
-20 -10 0 10 20 -20 -10 0 10 20

(c) Beam from F3 , elevation cut (d) Beam from F3 , azimuth cut
50 MoM/AF 50 MoM/AF
AF AF
Directivity [dBi]
Directivity [dBi]
40 40
30 30
20 20
10 10
0 0
-20 -10 0 10 20 -20 -10 0 10 20

(e) Beam from F2 , elevation cut (f) Beam from F2 , azimuth cut
50 MoM/AF 50 MoM/AF
AF AF
Directivity [dBi]
Directivity [dBi]
40 40
30 30
20 20
10 10
0 0
-20 -10 0 10 20 -20 -10 0 10 20

(g) Beam from F1 , elevation cut (h) Beam from F1 , azimuth cut
Figure 3.36: Principal plane cuts for half of the feeds at 20 GHz computed using AF for a main reflectarray
incident field that was simulated using MoM with a real sub-reflectarray layout (labelled MoM/AF), and
AF assuming ideal phase shifters on the sub-reflectarray (labelled AF).
The drop in directivity is possibly explained by the additional ripple in amplitude and distortion in
phase of the incident field on the main reflectarray, as is visible in Fig. 3.34 and Fig. 3.35. There are
several potential sources of these deviations. A major factor is the non-ideal effects of using real reflec-
tarray elements. In addition to factors like mutual coupling between the elements and their imperfect
scattering characteristics, which can be taken into account to some degree through unit cell simulations,
this reflectarray is fairly large and has a rapidly varying phase distribution with many phase wraps.
In the vicinity of the phase wraps, the local periodicity assumption used to design the reflectarray is
violated, and these elements will not behave as expected. This is a challenge when designing electrically
large reflectarrays. Another factor is the difference between the illumination on the sub-reflectarray from
the λ0 /2 dipole and the q = 1 source model.
The analysis of this section shows that although both simulation methods predict beams produced in
the same direction, additional modelling effort is required in order to accurately predict the performance
of the synthesis method. This would be especially true if an experimental demonstrator were to be
fabricated.
3.9 Other Work

In this section additional and related work on the bifocal reflectarray that does not affect the conclusions
reached from the prime-focus geometry will be discussed.
3.9.1 Offset Geometry

In this chapter a prime-focus antenna geometry was considered. This configuration was investigated more
as a test case, as there are less issues that arise as a result of the geometry, and it is easier to see whether
the observations are a result of the synthesis procedure itself. While this kind of configuration works fine
for folded reflectarrays where the sub-reflectarray is polarization-selective and does not cause blockage of
the main reflectarray [44, 27], it is not appropriate for our application where the antenna must generate
beams in two polarizations. Initially a dual-offset geometry was considered, where the main reflectarray,
sub-reflectarray, and feed cluster are offset by different angles. This type of geometry is more realistic
for a satellite application. However, the same conclusions were reached, which motivated the focus on
the symmetric geometry. Other work in the literature has focused on offset geometries [50, 56].
25
20
15
Gain [dB]
10
-5
-5 -4 -3 -2 -1 0 1 2 3 4 5
Theta [deg.]
Figure 3.37: Example of defining a set of masks based on a desired radiation pattern.
3.9.2 Optimization-Based Methods

Although a ray-tracing based synthesis procedure is intuitive in that it stems from the physical problem,
it is also possible perform the synthesis using optimization based methods. A problem with procedures
like the bifocal synthesis is that they are phase only and have no knowledge of the amplitude distributions
on the reflectors/reflectarrays. As was observed in the previous sections, the amplitude distributions can
have a significant impact on performance, in particular regarding spillover losses.
By optimizing for the gain of the reflectarray, spillover efficiency can be taken into account as it will
negatively impact gain. A genetic algorithm (GA)-based optimizer was developed, and some results will
be presented here. The full details of the optimizer have been previously presented [63].
40
20
Gain [dB]
-20
-40
-5 -4 -3 -2 -1 0 1 2 3 4 5
Theta [deg.]
Figure 3.38: Single point masks with BCF = 2.
Sub reflectarray Main reflectarray

4000 0
Bifocal
GA
3500 -500
Bifocal
3000 -1000 GA
Phase [deg.]
Phase [deg.]
2500 -1500
2000 -2000
1500 -2500
1000 -3000
500 -3500
0 -4000
-0.3 -0.2 -0.1 0 0.1 0.2 0.3 -0.5 0 0.5
x [m] x [m]
Figure 3.39: Comparison of the phase distributions obtained from the bifocal synthesis and GA optimizer.
The GA optimizer that was implemented is mask-based, which means that a mask is defined around
the desired gain pattern and the optimizer penalizes solutions which violate the mask. An example
case is shown in Fig. 3.37, where three beams resulting from the geometry of Fig. 3.6 with no beam
compression are used to define the masks. The dotted lines around each beam are the masks, and the
optimal solution will fit between them. It should be noted that Fig. 3.37 is simply an example which
demonstrates how the masks can be defined, and the beam results shown were not obtained through
optimization. Each beam has a set of upper and lower masks, MUi and MLi , respectively, where the
subscript i indicates the feed/beam number. A cost function was then implemented to penalize solutions
that violate the mask. However, imposing both an upper and lower mask was found to be too restrictive,
and in the end the upper mask was set to infinity and single point lower masks were implemented, such
that the function of the optimizer is the maximize the gain of each beam in its desired direction. These
simplified masks are shown in Fig. 3.38.
30
Bifocal
GA
20
10
Gain [dB]
-10
-20
-30
-5 -4 -3 -2 -1 0 1 2 3 4 5
Theta [deg.]
Figure 3.40: Comparison of the beams produced by the bifocal synthesis procedure and the GA optimizer.
The solid lines are from the bifocal reflectarray, and the dashed from the GA.
For the case presented in this section, the unknowns used in the optimization are the main-reflectarray
focal length, and half of the element phases on the sub-reflectarray due to symmetry. In Fig. 3.39 the
resulting phase distributions are compared with those resulting from applying the bifocal synthesis
procedure with BCF = 2. Although the main-reflectarray phases are very similar, the sub-reflectarray
phase is steeper and convex, but not hyperbolic. Comparing the resulting beams in Fig. 3.40, the beams
from the GA appear to roll off slower than those from the bifocal synthesis procedure. The sidelobes are
higher, but using the single-point optimization they were not constrained and so more advanced masks
might improve this.
In Fig. 3.41 the amplitude distributions on the main reflectarray are compared for the two methods.
It can be seen that although the GA illumination has more ripple, perhaps due to rapid variations in the
sub-reflectarray phase, that the edge illumination is lower. This indicates that the spillover efficiency
has been improved over the bifocal synthesis results, and shows some promise for methods that take
absolute gain into account.
Main reflectarray illumination

0
-5
|E|, normalized [dB]

-10
3 from center
2 from center
1 from center
-15 Center feed
-20
-0.5 0 0.5
x [m]
Figure 3.41: Amplitude distributions on the main reflectarray resulting from the GA optimized and
bifocal reflectarrays for the lower half of the feed cluster (F1 to F4 ). The bifocal results are shown as
dashed lines, and the GA results as solid lines.
3.10 Comments and Summary

In this chapter, a ray-tracing based synthesis method for dual-offset reflectors used to achieve wide-angle
scanning was applied to reflectarrays. This was used to correct the angular spacing of beams scanned
by feed displacement using a feed cluster for cases where the beam spacing is limited by the feed sizes.
A BCF was defined that relates the desired beam spacing of the system to a reference case for a given
feed cluster spacing. By investigating a prime-focus geometry it was found that when a BCF of two
is enforced the synthesis procedure results in a Cassegrain-like phase distribution with poor spillover
efficiency, which makes our proposed method unsuitable as a solution to our problem. However, it was
found when no beam compression is applied that the synthesis functions as it was originally intended, and
reduces scan aberrations. Additionally, minor variations in the beam spacing can be obtained without
a significant impact to performance.
The synthesis method was extended from 1D reflectarrays to 2D for a symmetric geometry by rotating
the 1D phase curve about the dual-reflectarray’s axis of symmetry. An array factor analysis showed
similar results to the 1D reflectarray. A real layout of the sub-reflectarray was performed, and an MoM
tool was used to simulate the incident field on the main reflectarray. The main reflectarray incident
field was then used as the input for an array factor computation to obtain the far-field pattern of
the dual-reflectarray system, and this was compared to results obtained using the array factor on both
reflectarrays. It was found that there were variations between the ideal array factor and MoM simulation
results, although both produced beams in roughly the same direction. The differences are likely due to
the rapidly varying phase on the sub-reflectarray, which results in a large amount of phase wraps, and
which violates the local periodicity assumption that is generally used to design reflectarrays. It is clear
that for assessing synthesis methods in the future, that more effort has to be taken in modelling the
reflectarray and feed for reflectarrays of this type. However, this was a further performance degradation
on what was already observed from the array factor method, and the same conclusions on the bifocal
synthesis procedure would have been reached.
In a related investigation using optimization methods rather than a ray-tracing based synthesis
procedure, it was found that taking absolute gain into account can be used to produce solutions with
better spillover efficiency. This suggests that in future work that a phase only synthesis cannot be relied
upon to produce efficient solutions, and that the amplitude distributions on the reflectarrays need to be
taken into account.
Chapter 4
Dual-polarization Reflectarrays
4.1 Introduction
In Chapter 3, reflectarray synthesis routines were applied to reduce the beam spacing for a multiple-beam
antenna implemented using a SFB architecture, in which the beams are produced by feeding the reflector
or reflectarray with a cluster of feeds. The feeds operate independently and are displaced relative to one
another, generating a set of independent beams at different angles, with their spacing being related to the
spacing of the feeds and the F/D of the reflective surface. This is a scalar approach, as it is independent
of the polarization of the feed. In this chapter a vector method will be explored, in which the beams are
generated independently based on the polarization of the feeds. While this does not compress the beam
spacing, compared to a conventional single polarization approach a dual-polarization reflectarray allows
for twice the number of beams to be generated by sharing the aperture between orthogonal polarizations.
Dual-polarization design techniques, from the unit cell level to the reflectarray level, will be explored and
demonstrated for both linear and circular polarization. Methods to improve reflectarray performance
will also be investigated.
4.2 Design Considerations for Reflectarray Elements

While Chapter 2 covered the basics of reflectarray operation, in this section and the next one the design
of a reflectarray element will be investigated in detail. As with any engineering problem, there are several
performance criteria and trade-offs to consider. These include bandwidth, the reflection phase response,
loss, polarization, angular stability, materials and ease of fabrication, and reconfigurability. There is a
large body of literature describing the design and characteristics of various reflectarray elements, which
will be reviewed in this section. This will form the basis for the design of the reflectarray element used
in this work, which will be described in detail in the following section.
4.2.1 Bandwidth
Much of reflectarray research has been focused on improving and studying their bandwidth [64, 65], as
this has always been noted as one of their drawbacks. The bandwidth limitation is perhaps emphasized
by the natural comparison of reflectarrays to paraboloidal reflectors, structures which ideally have infinite
bandwidth. The key attribute of a paraboloidal reflector which enables its large bandwidth is its curved
98
Chapter 4. Dual-polarization Reflectarrays 99
surface. This compensates the path length, or time delay, of rays travelling to or from the feed. A
reflectarray instead compensates the electrical path length, or phase delay, which introduces frequency
dependence and is inherently bandwidth limiting. This is especially apparent for moderate- to large-sized
reflectarrays [65, 41].
Initial reflectarray bandwidth extension techniques were based on increasing the phase range that
could be achieved by the unit cell. An increased phase range prevents phase errors which result from not
being able to cover a full 360◦ , and increases the group delay range which the element can compensate.
Additionally, a larger phase range means that the number of physical phase wraps which occur on the
reflectarray surface can be decreased. The element-to-element variation in the vicinity of these phase
wraps is rapid, which violates the local periodicity assumption and introduces phase errors which will
negatively impact a reflectarray’s radiation pattern.
The phase range of a reflectarray unit cell can be extended by designing it to have multiple resonances.
This can be done by stacking elements on multiple layers, as is commonly done with rectangular patches
[66, 67, 68], but can also be done on a single layer. Examples of multi-resonant single layer elements
include parallel co-planar dipoles [57, 59], concentric loops [69, 51], or combinations of elements such as
Jerusalem crosses and dipoles [70]. Multi-resonant elements on a single layer can also be combined with
elements on other layers, as has been demonstrated using dipole elements [60].
Although multi-resonant elements are generally used to obtain a larger phase range, the element
resonances can additionally be placed in a manner that increases the linearity of the reflection phase
response as both the control variable and frequency are swept. The phase which the reflectarray must
compensate is kd, where k is the wavenumber and d is the distance that each ray travels from the feed
to the reflectarray element to the output plane. If the reflectarray element’s phase is able to track the
linear variation of k with frequency, the reflectarray will be compensating the physical path length of
each ray in a time delay fashion similar to a paraboloidal reflector, giving clear bandwidth advantages.
Increasing the phase linearity with frequency while also increasing the phase range of the element will
increase the bandwidth over which the unit cell has a time delay rather than a phase delay response,
and increase the maximum delay that the element can compensate. Additionally, the reflectarray will
be less sensitive to fabrication errors than if an element with a sharply resonant phase curve is used.
Finally, when an element whose phase response is a linear function of its controlling dimension is placed
in the reflectarray, the variation between adjacent elements will be smooth. This will result in a better
approximation of the local periodicity assumption.
Using reflectarray elements whose phase response is linear with frequency opens up alternate reflec-
tarray design methods. In one demonstration aperture-coupled microstrip delay lines were used [71],
which can be lengthened to provide the required phase shift across the entire reflectarray with fewer
phase wraps. Microstrip transmission lines have a very linear response with frequency, although this
can be affected by resonances which occur in longer lines and the coupling behaviour of the slots. In
another work, the unit cell response was fit to an equivalent group delay, and the reflectarray was de-
signed based on its time delay profile [72]. Further efforts have used transformation optics to engineer
a material profile which is equivalent to a curved reflector [73], or an impedance surface approach to
design elements with a Bessel filter response [74]. The Bessel filter has maximally flat group delay over
a large bandwidth, resulting in a reflectarray which has very good behaviour with frequency.
Reflectarrays are typically manufactured using flat rectangular laminates, which is one of their ad-
vantages as they can be fabricated using widely available printed circuit processes. However, using a
completely flat reflectarray a large phase range is required to collimate a beam. This phase range can
be reduced by dividing the full reflectarray into a set of panels that are arranged to produce a surface
that is a piecewise linear approximation of a paraboloid [25, 52], or by using an actual paraboloidal
surface [53, 54, 75, 26]. These techniques reduce the phase range required to synthesize the reflectarray,
as a paraboloidal surface already converts the spherical waves produced by the feed into a plane wave.
The reduced phase range reduces the amount of delay that the reflectarray needs to compensate, and
correspondingly the number of phase wraps, improving the validity of the local periodicity assumption.
Additionally, as a large phase range is not required elements with smaller unit cells can be used, which
have been shown to improve reflectarray bandwidth [76, 77]. However, curved and faceted reflectarrays
are more challenging to manufacture than purely flat ones.
4.2.2 Polarization
Polarization is an important characteristic of any antenna, and polarization requirements largely depend
on the application under consideration. For example, different types of linear polarization are commonly
used in terrestrial applications. However, for space applications circular polarization may be required
to overcome polarization mismatch losses resulting from Faraday rotation or the motion of the satellite.
Different polarization requirements will require different antenna element designs. Reflectarrays offer
many possibilities for polarization control, as demonstrated by applying them to reduce cross-polarization
[47], to rotate the polarization of an incident field [27], or by designing reflectarrays to be transparent
to one polarization but reflective to another [27, 78].
Of particular interest in this thesis are reflectarray elements which are dual-polarization capable,
meaning that they can simultaneously have different responses for two orthogonal polarizations. A typi-
cal dual linear polarization element has two orthogonal dimensions which can be adjusted, each of which
is aligned with one linear polarization. By adjusting the two element dimensions the reflection phase of
both polarizations can be adjusted independently, ideally with minimal coupling between the two polar-
izations. As an example, for the rectangular patch element illustrated in Fig. 4.1 the patch dimensions lx
and ly are aligned with vertical polarization (VP) and horizontal polarization (HP), respectively. There-
fore, lx can be adjusted to change the VP reflection phase, and changing ly will adjust the HP reflection
phase. Implementations of this kind of dual-LP reflectarray have been done using stacked rectangular
patches of variable size with two [66] and three [68] layers, rectangular patches with directly connected
[79] and aperture-coupled [80] stubs, orthogonal sets of parallel dipoles in various configurations [60, 61],
and crossed dipoles [81].
For circular polarization control, the phase shift mechanism differs from linear polarization. A com-
mon method to control the phase of CP waves is the variable rotation technique (VRT). Using this
technique, rather than varying some physical dimension of the reflectarray unit cell, the element is ro-
tated by an angle ψ to produce a phase shift of 2ψ [82]. Other methods can also be used as long as
they maintain the 90◦ phase difference between the two orthogonal linearly polarized components of the
incident circularly polarized fields. However, in general the mechanism used to achieve a phase shift in
one sense of CP will also affect the orthogonal sense, making it more challenging to design a dual-CP
element than a dual-LP one.
There have been several investigations into independent dual-CP control, and some concepts from
the literature are illustrated in Fig. 4.2. Using a concentric split ring element and the VRT, a unit cell
can be designed which has independent responses at two different frequencies. One frequency can then
VP
HP lx
ly
Figure 4.1: Example rectangular patch element showing the polarizations which its dimensions are
aligned with.
be used to control RHCP and the other LHCP [83], allowing for separate RHCP and LHCP beams to
be formed in two frequency bands as shown in Fig. 4.2a. However, when the VRT is applied a rotation
which produces a phase shift of ψ in one sense of CP will produce a shift of −ψ in the orthogonal sense
[82]. As such, this method does not allow for independent control of both beams at a common frequency.
In another work circular polarization selective surface (CPSS)s, which are transparent to one sense of CP
but reflective to the other, are used rather than a traditional reflectarray [84]. This method is shown in
Fig. 4.2b. A left-handed circular polarization selective surface (LH-CPSS) which reflects LHCP is placed
on top of a right-handed circular polarization selective surface (RH-CPSS) which reflects RHCP. The
LH-CPSS imparts a phase shift on incident LHCP waves when it reflects them, but transmits RHCP
waves. These RHCP waves are delayed when they are reflected from the underlying RH-CPSS, and then
they are transmitted back through the LH-CPSS.
In an alternative approach, pictured in Fig. 4.2c, a linear-to-circular polarizer is placed in front
of a polarization-selective reflectarray which is backed by a solid reflector [78]. Incident RHCP and
LHCP fields are converted to VP and HP, respectively, by the polarizer. The reflectarray is designed
to reflect HP, but it is transparent to VP which goes on to be reflected by the curved reflector. By
using two reflective surfaces the VP and HP waves can be controlled independently, and on the second
pass through the polarizer they are converted back to RHCP and LHCP. Using a polarizer simplifies
the dual-polarization control problem by removing the circular polarization aspect from it.
Another method [53, 54, 75, 26], shown in Fig. 4.2d, differs from the others in that rather than using
a planar reflectarray, the reflectarray elements are printed on a paraboloidal surface. In this scenario the
curved surface collimates the rays from the feed in order to produce a beam in the specular direction,
and using the VRT this beam can be deflected to a specified direction. As was previously mentioned,
when applying the VRT rotating the elements delays the reflection phase phase in one sense of CP
and advances it by an equal amount in the orthogonal sense. As deflecting a plane wave requires a
linear phase gradient, when applied to a curved surface this property of the VRT produces orthogonally
circularly polarized beams which are scanned symmetrically about the direction of specular reflection,
assuming the use of a feed which radiates both RHCP and LHCP. In another variation of this technique,
Reflectarray LH-CPSS
RH-CPSS
x x
RHCP RHCP
α α
LHCP
LHCP
RHCP at f1
LHCP at f2 RHCP + LHCP at f1
z z
y y
(a) Using a dual-band reflectarray element to control (b) Using polarization selective surfaces to separately
RHCP and LHCP in different frequency bands [83]. control RHCP and LHCP [84].
Reflector Reflector
Reflectarray Printed elements
Polarizer
x x
RHCP
α RHCP
α/2
LHCP
α/2
LHCP
RHCP + LHCP at f1 RHCP + LHCP at f1
z z
y y
(c) Using two reflective surfaces to control orthogonal (d) Printing the reflectarray elements on a curved sur-
linearly polarized waves with a polarizer to convert face to produce beams symmetrically about the direc-
between linear and circular polarization [78]. tion of specular reflection [53, 54].
Figure 4.2: Dual circular polarization control concepts from the literature. α is the angular separation
between the RHCP and LHCP beams, and f1 and f2 are the operating frequencies.
rather than both beams being placed symmetrically with respect to the specular direction, the beam
from one polarization is formed in the specular direction whereas the orthogonally polarized beam is
scanned in an arbitrary direction [53]. However, when the orthogonally polarized beams are in the same
frequency band this would require the reflectarray elements to be inactive for the polarization which
produces the specular beam, making this variation more suitable for linear polarization than circular
polarization unless a polarizer is used. The use of a dual-band element allows this method to additionally
produce independent beams in two frequency bands [54, 75, 26], resulting in four independent beams
from a dual-band dual-polarization feed.
While the previous methods looked at dual-LP control using linearly polarized feeds, or dual-CP
control using circularly polarized feeds, the reflectarray elements can be designed to convert between
polarizations in addition to providing a phase shift. By designing the reflectarray elements to delay the
orthogonal components of incident linearly polarized waves such that they become 90◦ out of phase,
the scattered fields will be circularly polarized. RHCP or LHCP will be produced based on which LP
component is leading or lagging. Circularly polarized reflectarrays with linearly polarized feeds have been
demonstrated using crossed dipoles on a single layer [85], orthogonal dipoles on separate layers [85, 86],
stacked elliptical patches [87], and a combination of Jerusalem crosses and orthogonal dipoles on a single
layer [70]. In all of these examples either the linearly polarized feed or reflectarray elements have been
rotated by 45◦ to allow for the reflectarray elements to be excited by both transverse components of
the incident field, which is necessary for the polarization conversion. When the element dimensions
are assigned they must be set not only to provide a required phase shift, but to also maintain a 90◦
phase difference between the orthogonal components of the scattered field. As this will be based on the
polarization of the incident field, and both of its transverse components must be delayed appropriately,
it is challenging to produce independent beams in both senses of circular polarization using this method.
However, a dual-reflectarray configuration was recently proposed [75] where the sub-reflectarray has an
independent response to orthogonally linearly polarized fields from a dual-polarization feed. The main
reflectarray, which has a paraboloidal curvature, acts as a reflection based polarizer and converts the
incident linearly polarized fields to circular polarization. In this manner separate beams in RHCP and
LHCP are generated from a dual-LP feed.
Although in principle linear polarization can be produced from incident circular polarization by bring-
ing the orthogonal LP components to be in-phase with each other, this case is generally not considered
due to the added complexity of producing a high-performance circularly-polarized feed compared to a
linearly-polarized one.
4.2.3 Periodicity
Typical reflectarray elements use a half-wavelength unit cell period, which avoids grating lobes at all
scan angles. A half-wavelength periodicity also gives adequate space for the element dimensions to be
large enough for the element to become self-resonant, which increases its phase range. However, as
mentioned previously, unit cells with reduced periodicities can improve reflectarray bandwidth. Benefits
have also been observed with regard to loss [88, 89], and angular stability [32, 90, 89, 91]. In general,
half-wavelength elements have a high quality factor resonance which varies sharply with frequency and
the element’s phase control characteristic. Reducing the unit cell periodicity results in smoother re-
flection phase variation, and additionally smoother element-to-element variation when the elements are
placed in the reflectarray. The former factor reduces sensitivity to manufacturing tolerances and can
improve the element’s electrical characteristics, and the latter improves the validity of the local period-
icity assumption. However, reducing the unit cell size generally results in a reduction in phase range,
which can introduce negative effects such as reduced gain and increased sidelobes from not being able
to cover a 360◦ phase range. These negative effects can be mitigated through the use of multi-resonant
unit cells to increase phase range.
In applications where multiple surfaces are used in close proximity, such as some of the dual-CP
reflector designs listed in Section 4.2.2, a smaller unit cell size has an additional advantage of increasing
the attenuation constant of higher order evanescent Floquet modes. This allows for the surfaces to be
placed closer together without inducing higher order Floquet mode coupling, which can enable the profile
of the overall structure to be reduced without accounting for these effects.
4.2.4 Materials
As is generally the case for microwave circuits, printed reflectarrays are typically etched on low loss
laminates designed for high frequency use. However, additional considerations may have to be made for
space applications due to harsh conditions including large temperature variations and the presence of
cosmic rays, as well as the potentially high transmit power levels required for a downlink to a ground
station. As such, reflectarrays have been designed using space qualified materials [68], and their capability
in space applications has been demonstrated through their use as CubeSat antennas for NASA’s ISARA
and MarCO missions [92, 93]. Although the problems investigated in this thesis are based around space
applications, the materials aspect will not be a focus. In a practical design the same techniques could
be applied using materials more suited towards space use.
Non-traditional materials have been researched for use in reflectarrays for various applications. Ex-
amples include liquid crystal for electronically tunable applications [94], optically transparent indium tin
oxide for reflectarrays which can share the same area as the solar panels on a satellite [95], and graphene
for terahertz applications [96]. As substrates used for high frequency design are typically expensive,
methods to reduce the cost of reflectarrays have looked at reducing element losses. This could enable
the use of lossy, but cheaper, general purpose substrates [88, 89].
4.2.5 Reconfigurability
In a phased array in which each element is not driven separately, a beamforming network must be
developed to distribute the signal between the array elements with the correct amplitude and phase.
Phase shifters, variable attenuators, and variable gain amplifiers can then be added to electronically
control the phase and amplitude of the signals going to each element, which can be used to adaptively
change the radiation pattern. A similar approach can be applied to reflectarrays by using elements
which have a tunable reflection phase. There is a large body of work on reconfigurable reflectarrays
[97], and some methods include loading elements with varactor diodes, designing them on a liquid
crystal substrate, rotating or moving them with micro-motors, and changing their physical shape using
microelectromechanical switches. As the devices used to design high frequency phase shifters are often
expensive, methods have also been explored to enable the use of more cost-effective technologies [98].
Although reconfigurability will not be touched on in this work, it could be an interesting extension of it
in the future.
4.3 Reflectarray Element Design

In this section, the design of the reflectarray element used in this work will be described. Following this,
some comparisons will be made to other element designs based on simulation results and the existing
literature.
4.3.1 Stacked Jerusalem Cross Element

Out of the dual-polarization elements overviewed in Section 4.2.2, crossed dipoles are a good candidate
element as they have a strongly decoupled response to two orthogonal linear polarizations. However,
dipole type elements must be relatively long to become resonant, requiring large unit cell periods or
high permittivity substrates. It is clear from the previous section that there are benefits to using unit
Parameter Value Parameter Value

γ 45◦ lcxl ldxl − 1.6 mm
px 3.75 mm lcyl ldyl − 1.6 mm
py 3.75 mm ldxu ldxl − 0.5 mm
wd 0.5 mm ldyu ldyl − 0.5 mm
wc 0.3 mm lcxu lcxl − 0.3 mm
lcyu lcyl − 0.3 mm
Table 4.1: Geometrical parameters for the Jerusalem cross unit cell pictured in Fig. 4.3.
cells that are smaller than λ0 /2 × λ0 /2 with regard to bandwidth, loss, and angular stability, which we
would like to take advantage of. One solution to the problem of the dipole size is to to add caps to the
ends of them, which forms a Jerusalem cross. This increases the capacitance between adjacent elements,
lowering the element’s resonant frequency and allowing its full phase range to be swept within a smaller
unit cell. Jerusalem crosses have their roots in work on frequency selective surfaces [32], and have been
previously used in reflectarrays [70] and polarizers [99].
As was previously noted, while there are advantages to reducing the unit cell size this generally results
in a reduced phase range. To recover phase range, single or multi-layered multi-resonant unit cells can
be used. While multi-resonant elements on a single layer are attractive due to their lower fabrication
complexity, which results in lower cost and less potential sources of error, for a dual-polarization element
it becomes challenging to vary the orthogonal element dimensions without increasing the unit cell size.
Thus, in this work a two-layer multi-resonant configuration is used. The addition of a second layer
does increase the fabrication complexity and cost, but this was deemed acceptable due to the potential
benefits of using smaller unit cells.
The developed unit cell is shown in Fig. 4.3, and values for the geometrical parameters are provided
in Table 4.1. It consists of two layers of Jerusalem crosses, with the top element being smaller than the
bottom one. The primed axes are aligned with the cross, and the unprimed axes have been rotated by
an angle γ from the primed ones. This rotation angle is important for the dual-CP case and will be
explained shortly, but for the dual-LP reflectarray γ = 0◦ and the two coordinate systems are aligned.
The dimensions of the unit cell are λ0 /4 × λ0 /4 at 20 GHz.
The stacked Jerusalem cross element functions in the same manner as typical dual-polarization
reflectarray elements, such as the ones previously listed, whereby adjusting the orthogonal dimensions of
the cross varies the reflection phase of two orthogonal linearly polarized waves. From Fig. 4.3, varying
ldx,l and ldx,u produces a phase shift in the reflection phase of scattered x0 -polarized plane waves, whereas
varying ldy,l and ldy,u does the same for y 0 -polarized waves.
It is visible from Fig. 4.3 that the unit cell has many geometrical parameters. It would be difficult
and time consuming to find the combination of all of these variables which yields the optimal phase
response of the cell, and so some of these parameters were constrained. These constraints are listed in
Table 4.1. The upper dipole lengths, ldx,u and ldy,u are fixed to be shorter than the ones on the lower
layer by 0.5 mm. This is to place the resonances in the reflection response such that they are closely
spaced, but not overlapping, as the element lengths are swept. This gives a fairly linear phase variation
with element length, while keeping losses low.
The lengths of the caps on the lower elements, lcx,l and lcy,l , were fixed to be 1.6 mm shorter than
their respective dipoles. This was done so that the cap lengths become shorter as the dipole length
x0
ldy,u
y x
lcx,u
γ
ldx,u lcy,u
y0 lcy,l ldx,l px
wc
wd
lcx,l
ldy,l
py
(a) Top view. The subscripts l and u indicate whether the dimension is referring to the lower or upper layer,
respectively. The smaller element is the upper one. The z direction is coming out of the page.
0.762 mm
38 µm
0.762 mm
(b) Cross-sectional side view. The bonding film and conductor thickness are exaggerated for clarity.
Figure 4.3: Stacked Jerusalem cross reflectarray unit cell geometry.
decreases. As each arm of the Jerusalem cross has its length varied from 2.0 mm to 3.6 mm, if the
cap lengths do not scale with the dipoles it becomes difficult to vary the orthogonal dipole lengths
independently through their entire range without their caps becoming overlapped and shorted together.
For some length combinations where one dipole is very long and the other short, the cap lengths were
reduced slightly from Table 4.1 to give a larger margin for manufacturing tolerance considerations. The
upper cap lengths, lcx,u and lcy,u , were fixed to be 0.3 mm shorter than their respective upper caps such
that the upper element is always smaller than the lower one. The end result of applying these constraints
is that all of the unit cell dimensions are set by the lower element lengths. For brevity the dimensions
ldx,l and ldy,l will be referred to as lx and ly , respectively, in the rest of this chapter.
The reflectarray unit cell is shown in Fig.4.4 as ly is held constant while lx is increased. It is visible
that by enforcing the constraints listed above that the upper element is always smaller than the lower
one, and the cap lengths grow with the dipoles. It should be emphasized that the constraints imposed
are not unique or necessarily optimal, and that there are likely other combinations that would produce
a similar response. Similar methods are used for other multi-resonant unit cells. For example, using
stacked rectangular patches the upper patch dimensions are typically fixed to be some ratio of the lower
z z z
y0 y0 y0
x0 x0 x0
(a) lx = 2.4 mm (b) lx = 3.0 mm (c) lx = 3.6 mm
Figure 4.4: Element variation as lx is increased with ly fixed at 2.7 mm.
ones [66, 67, 68, 76] , the lengths of the non-central dipoles in a multiple co-planar dipole unit cell
are usually usually some ratio of the central one [57, 59, 60, 61] , and the gap between the inner and
outer loop in a multiple concentric loop element is generally fixed to be a constant [51]. The authors
in several of these examples also note that geometrical constraints were chosen based on the electrical
characteristics of the element.
4.3.2 Simulation Results

The two-layer unit cell was simulated in a periodic environment using periodic boundary conditions and
Floquet ports in Ansys HFSS, as described in Section 2.3.1. The two substrate layers were composed of
a dielectric with r = 3.00 and tan δ = 0.0013, with a bonding layer in between having r = 2.28 and
tan δ = 0.0030. The dielectric properties and thickness were chosen based on readily available low loss
materials that are commonly used for high frequency applications. The conductor layers had a thickness
of 17 µm and σ = 5.8 × 107 S/m, which is the static conductivity of copper, and were simulated as
three-dimensional structures rather than using surface approximations.
The unit cell’s simulated co-polar reflection phase and magnitude responses to an x0 -polarized nor-
mally incident wave are shown in Fig. 4.5a and Fig. 4.5b, respectively, as lx is swept at several values
of ly which are given in the legend of Fig. 4.5b. The values of lx and ly were selected arbitrarily and
are meant to show a representative sample of the results. Although the reflectarray unit cell was orig-
inally designed for 20 GHz, the constant frequency results are plotted at 20.5 GHz as this was used as
the design frequency for the full reflectarray. For the dual-CP reflectarray, which will be described in
Section 4.4, a polarizer must be placed in front of the reflectarray elements. The measured performance
of the polarizer was optimal at 20.5 GHz, and so this was chosen as the design frequency.
The two resonances of the unit cell are visible in Fig. 4.5b. The placement of the resonances, which
is based on how the upper and lower dipoles are sized relative to one another, produces a fairly linear
phase variation with cross length, as shown by Fig. 4.5a. The phase variation is much smoother than
the sharp reflectarray S-curves which are typical of larger elements which use a half-wavelength cell
period. It can be seen that when lx is between 2 mm and 2.5 mm that there is no significant change
in phase. If a higher dielectric constant substrate was used the onset of the phase change would occur
earlier, potentially allowing for further unit cell miniaturization. From Fig. 4.5c the phase variation with
200 0
Magnitude [dB]
Phase [deg.] 0 -0.1
2.0 mm 2.6 mm 3.2 mm

-200 -0.2
2.2 mm 2.8 mm 3.4 mm
2.4 mm 3.0 mm 3.6 mm
-400 -0.3
-600 -0.4
2 2.5 3 3.5 2 2.5 3 3.5
x length [mm] x length [mm]
(a) (b)
200 0
Magnitude [dB]
0 -0.1
Phase [deg.]
2.0 mm 2.6 mm 3.2 mm

-200 -0.2
2.2 mm 2.8 mm 3.4 mm
2.4 mm 3.0 mm 3.6 mm
-400 -0.3
-600 -0.4
17 18 19 20 21 22 23 24 17 18 19 20 21 22 23 24
Frequency [GHz] Frequency [GHz]
(c) (d)
Figure 4.5: The simulated co-polar reflection response of the two-layer Jerusalem cross element to a
normally incident x0 -polarized plane wave at (a)–(b) 20.5 GHz for several values of ly as lx is swept, and
(c)–(d) for several values of lx with ly = 2.9 mm as frequency is swept.
frequency is also quite smooth and linear over the range of frequencies simulated. However, some unit
cell variations show a stronger deviation from a linear trend than others. This will be re-visited later on
in this chapter. It can be observed from Fig. 4.5b and Fig. 4.5d that the losses of the unit cell are low.
In Fig. 4.6 the unit cell’s response is shown at several different angles of incidence between 0◦ and
60 in its E-plane and H-plane. Referring to Fig. 4.4, the E-plane corresponds to the x0 z-plane and the
◦
H-plane the y 0 z-plane. The results in Fig. 4.6 show that the phase curves are fairly stable at the different
angles of incidence which were considered, although there is stronger variation around the resonances. If
the normal incidence response is used to characterize the reflectarray element and assign dimensions in
the full reflectarray, it can be seen from Fig. 4.6 that at each value of lx a phase error will be produced
as the angle of incidence is swept from 0◦ to 60◦ . The worst case value of this phase error is about 89◦
in the element’s E-plane, and about 63◦ in its H-plane. If the range of angles of incidence is reduced to
0◦ to 45◦ , which is a realistic range of incidence angles for a reflectarray with an F/D around one, the
worst case phase error is about 54◦ in the E-plane and 34◦ in H-plane. The loss also increases by a small
amount with an increased angle of incidence.
There is less variation in the element’s E-plane than its H-plane, possibly because in the H-plane
the tangential electric field along the element should not significantly change as the angle of incidence
is swept. Low variation with angle of incidence has also been observed in other works using reflectarray
unit cells with a sub-λ0 /2 periodicity [90, 89, 91].
The full two-dimensional reflection response as both lx and ly are swept is shown in Fig. 4.7. The
200 0
Magnitude [dB]
Phase [deg.] 0 -0.1
-200 -0.2
0 deg. 0 deg.
15 deg. 15 deg.
30 deg. 30 deg.
-400 -0.3
45 deg. 45 deg.
60 deg. 60 deg.
-600 -0.4
2 2.5 3 3.5 2 2.5 3 3.5
(a) E-plane (b) E-plane
200 0
Magnitude [dB]
0 -0.1
Phase [deg.]
-200 -0.2
0 deg. 0 deg.
15 deg. 15 deg.
30 deg. 30 deg.
-400 -0.3
45 deg. 45 deg.
60 deg. 60 deg.
-600 -0.4
2 2.5 3 3.5 2 2.5 3 3.5
(c) H-plane (d) H-plane
Figure 4.6: Simulated reflection response of the two-layer Jerusalem cross element to a TM-polarized
incident plane wave as the angle of incidence is swept in its E-plane and H-plane at 20.5 GHz with
ly = 2.9 mm.
following notation is used for the reflection matrix, ΓRA :

" #
ΓHH ΓHV
ΓRA = , (4.1)
ΓVH ΓVV
where ΓHH and ΓVV are the HP and VP co-polar reflection coefficients, respectively. ΓVH is the cross-
polar reflection coefficient for incident HP waves, and ΓHV for incident VP waves. Referring to Fig. 4.3,
at normal incidence HP corresponds to a y 0 -polarized wave, and VP to an x0 -polarized wave. From
Fig. 4.7a and Fig. 4.7b it is visible that by adjusting lx the phase of ΓVV can be varied, and the phase
of ΓHH can be controlled by adjusting ly . Although the reflection phases of both polarizations are quite
independent they are not completely decoupled, as evidenced by the undesired drift of about 34◦ in the
phase of ΓVV as lx is held constant while ly is varied. The same applies for the phase of ΓHH as ly is
held constant while lx is swept. Although the value of this phase drift is quite small, phase errors will
be introduced if it is assumed that ΓVV and ΓHH are completely independent when constructing the full
reflectarray. This will be addressed in Section 4.5.2. The phase range of the co-polar terms is fairly
large at 548◦ , showing that by using a two-layer configuration reducing the unit cell periodicity has not
resulted in a loss of phase range.
The two-dimensional magnitude response of the unit cell is shown in Fig. 4.7c for the co-polar
reflection coefficient and Fig. 4.7d for the cross-polar term. Due to the symmetry of the responses, |ΓHH |
and |ΓHV | are not plotted. The simulated losses of the unit cell are low, and the maximum co-polar
(a) (b)
(c) (d)
Figure 4.7: Simulated reflection response of the stacked Jerusalem cross element as both lx and ly are
swept at 20.5 GHz and normal incidence.
reflection loss in both polarizations is 0.12 dB. There is good isolation between the two polarizations,
and the maximum magnitude of the cross-polar reflection coefficient is −34.8 dB.
4.3.3 Comparison to Other Dual-LP Elements

In this section some comparisons will be made between the two-layer Jerusalem cross element and
some other common dual-polarization reflectarray elements. The other elements considered are crossed-
dipoles and rectangular patches, which are pictured in Fig. 4.8. Both elements are used in two-layer
configurations for bandwidth considerations, and to give a more level comparison to the Jerusalem
cross element. A quarter-wavelength cell period is used, and for the stacked rectangular patches a
half-wavelength periodicity is also considered to offer a comparison between different periodicites. For
the rectangular patches, the dimensions of the upper element are fixed to be 0.7 times the size of the
corresponding lower patch dimensions. This is a commonly-used constraint in other works using stacked
rectangular patch elements [66, 76]. For the crossed-dipole element the same relations between the upper
and lower dipole lengths are used as for the Jerusalem cross element, which are listed in Table 4.1. Both
x0 x0
ly,u ly,u
lx,u lx,u
y0 wd lx,l px y0 lx,l px
wd
ly,l ly,l
py py
(a) Stacked crossed-dipole element (b) Stacked rectangular patch element
Figure 4.8: Other dual-polarization reflectarray elements that were considered. In (a) px = py =
3.75 mm, lx,l − lx,u = ly,l − ly,u = 0.5 mm (b) lx,l /lx,u = ly,l /ly,u = 0.7, px = py = 3.75 mm for the
quarter-wavelength cell and px = py = 7.50 mm for the half-wavelength cell.
Element 6 ΓVV range [deg.] Phase drift [deg.] |ΓVV | min. [dB] |ΓHV | max. [dB]
Stacked rectangular 242 47 −0.07 −50.0
patches, λ0 /4 × λ0 /4
Stacked rectangular 626 101 −0.20 −51.5
patches, λ0 /2 × λ0 /2
Stacked crossed 257 3 −0.08 −53.3
dipoles, λ0 /4 × λ0 /4
Stacked Jerusalem 548 34 −0.12 −34.8
crosses, λ0 /4 × λ0 /4
Table 4.2: Comparison of several two-layer dual-polarized reflectarray elements from the results presented
in Fig. 4.9 and Fig. 4.7.
elements use the same material parameters and stack-up as the Jerusalem cross element.
The reflection responses of each unit cell as both dimensions of the elements are swept are provided
in Fig. 4.9 at 20.5 GHz and normal incidence. Some performance characteristics are summarized in
Table 4.2, along with those of the Jerusalem cross element. It is visible that all of the elements can
control the reflection phase of both HP and VP with low loss and cross-polarization, but there are some
performance trade-offs which will be discussed.
Considering the phase ranges of the different elements, both the quarter-wavelength period stacked
rectangular patches and crossed dipoles have a phase range which is less than 360◦ . This is a well-known
drawback of reducing the size of reflectarray unit cells. As anticipated, the half-wavelength stacked
patches have the largest phase range. However, this element has the highest loss, which is also expected
and has motivated recent work on sub-half-wavelength unit cells [88, 89].
It is observable from the co-polar reflection phase plots in Fig. 4.7 and Fig. 4.9 that if lines of constant
(a) (b) (c) (d)
(e) (f) (g) (h)
(i) (j) (k) (l)
Figure 4.9: Simulated co-polar reflection responses of some dual-LP reflectarray unit cells to a normally
incident VP wave at 20.5 GHz. (a)–(d) λ0 /4 × λ0 /4 two-layer stacked rectangular patches, (e)–(h)
λ0 /2 × λ0 /2 two-layer stacked rectangular patches, (i)–(l) λ0 /4 × λ0 /4 two-layer stacked crossed dipoles.
phase were drawn, for the crossed-dipoles they would be virtually parallel, whereas for the other elements
there would be more of a slope to them. This shows the strong polarization independence of the crossed
dipoles, which can also be seen from their maximum phase drift in Table 4.2. Here phase drift is
taken to be the maximum change in the phase of the co-polar reflection coefficient as its controlling
dimension is fixed while the orthogonal dimension is swept. The quarter-wavelength rectangular patch
and Jerusalem cross elements perform similar to each other, and the half-wavelength patches perform
the worst. While this drift can be easily corrected for, it is undesirable. If the phase responses to both
orthogonal polarizations are completely independent then rather than tuning the two element lengths
to achieve the correct phase response for both polarizations, the designer can use the two dimensions as
an additional degree of freedom to improve performance.
Looking at the cross-polar performance, the Jerusalem cross elements have the highest cross-polar-
ization level, although it is still fairly low. This is a result of the dipole caps which are orthogonal to
the incident co-polarized field. Currents induced in this part of the element will produce radiation in
the unwanted polarization. The caps also reduce the separation between the orthogonal element arms,
which increases the coupling between them.
Some additional results are shown in Fig. 4.10 for the quarter-wavelength crossed-dipoles, and
Fig. 4.11 for the half-wavelength rectangular patches. From Fig. 4.10a and Fig. 4.10b it is visible that
the response of the crossed dipoles is virtually unchanged at different values of ly as lx is swept, again
200 0
Magnitude [dB]
Phase [deg.] 0 -0.1
-200 -0.2
2.0 mm 2.6 mm 3.2 mm
2.2 mm 2.8 mm 3.4 mm
-400 -0.3 2.4 mm 3.0 mm 3.6 mm
-600 -0.4
2 2.5 3 3.5 2 2.5 3 3.5
(a) (b)
200 0
Magnitude [dB]
0 -0.1
Phase [deg.]
-200 -0.2
2.0 mm 2.6 mm 3.2 mm
2.2 mm 2.8 mm 3.4 mm
-400 -0.3 2.4 mm 3.0 mm 3.6 mm
-600 -0.4
18 20 22 24 18 20 22 24
(c) (d)
Figure 4.10: Simulated co-polar reflection response of the stacked λ0 /4 × λ0 /4 crossed-dipole element
illustrated in Fig. 4.8a to a normally incident TM-polarized plane wave (a)–(b) at 20.5 GHz for several
values of ly as lx is swept, and (c)–(d) as a function of frequency at several values of lx with ly = 2.9 mm.
demonstrating the element’s strong phase independence. In contrast, for the half-wavelength stacked
patches it can be seen in Fig. 4.11a and Fig. 4.11b that there is a visible shift in its response. Comparing
Fig. 4.11c with Fig. 4.10c and Fig. 4.5c, the phase variation with frequency of the half-wavelength patches
is more non-linear than both the crossed-dipoles and Jerusalem crosses, particularly if the region from
18 GHz to 22 GHz is focused on. While the crossed dipoles have a very linear response with frequency,
the slopes of the curves in Fig. 4.10c corresponding to different element lengths do not appear to change
significantly. This suggests that there will be some limitations in the group delay that the element can
compensate.
The response of the quarter-wavelength crossed-dipole at several different angles of incidence in its
E-plane and H-plane is shown in Fig. 4.12a to Fig. 4.12d. The equivalent results for the half-wavelength
patches are shown in Fig. 4.13a to Fig. 4.13d. For all of these results ly is fixed at an intermediate
value while lx is swept, and the same conventions for the E-plane and H-plane as before are used. From
Fig. 4.12a to Fig. 4.12d there is little variation in the response of the crossed dipole as the angle of
incidence is swept, although there is a shift in phase for the shorter elements and some variation in
magnitude when lx is around 3 mm. The worst case variation with angle of incidence of the plotted
phase responses is about 93◦ in the element’s E-plane, and 62◦ in its H-plane. These values are in line
with those of the Jerusalem cross element, suggesting that the addition of the caps to the dipoles has
not impacted the angular stability of the element. From Fig. 4.13a it is visible that the variation in the
phase response with angle of incidence of the half-wavelength rectangular patches is quite significant in
200 0
Magnitude [dB]
Phase [deg.] 0 -0.1
-200 -0.2
2.0 mm 4.1 mm 6.2 mm
2.7 mm 4.8 mm 6.9 mm
-400 -0.3 3.4 mm 5.5 mm 7.4 mm
-600 -0.4
2 3 4 5 6 7 2 3 4 5 6 7
(a) (b)
200 0
Magnitude [dB]
0 -0.1
Phase [deg.]
-200 -0.2
2.0 mm 4.1 mm 6.2 mm
2.7 mm 4.8 mm 6.9 mm
-400 -0.3 3.4 mm 5.5 mm 7.4 mm
-600 -0.4
18 20 22 24 18 20 22 24
(c) (d)
Figure 4.11: Simulated co-polar reflection response of a two-layer λ0 /2 × λ0 /2 stacked rectangular patch
unit cell, pictured in Fig. 4.8b, to a normally incident TM-polarized plane wave (a)–(b) at 20.5 GHz for
several values of ly as lx is swept, and (c)–(d) for several values of lx with ly = 4.1 mm as frequency is
swept.
its E-plane. Additionally, from Fig. 4.13b the loss becomes quite high at some combinations of incidence
angle and length, exceeding 1 dB at points. The variations in the element’s H-plane are much less.
A comparison of the simulated unit cells is given in Table 4.3. Each one is assigned a ranking
of ‘+’, ‘−’, or ‘0’ for good, poor, or acceptable performance, respectively. For some parameters the
quarter-wavelength rectangular patches were not simulated as it is expected that they will perform
similarly to the quarter-wavelength crossed dipoles. For these categories they are assigned a rank of
‘not applicable’, or N/A. Based on the simulated results, the crossed dipole element performs very well
in every category except for phase range. Conversely, the half-wavelength rectangular patches have a
large phase range, but suffer from poor angular stability, high loss at certain angles of incidence, and
less phase independence between the element dimensions. Additionally, their behaviour with frequency
is not as linear as the quarter-wavelength elements. While the Jerusalem crosses do not perform as well
as the crossed dipoles with regard to phase drift, loss, and cross-polarization, they do not lack phase
range as the other quarter-wavelength elements do. As the performance in the categories where they are
lacking is still quite good, the Jerusalem cross will be used as the element for the reflectarray designs in
the rest of this chapter.
As a final note, all of the simulated elements could be used to design high performance reflectarrays,
and the best choice for a design would dependent on which characteristics are the most important for the
application under consideration. Additionally, the elements simulated for comparison are not necessarily
200 0
Magnitude [dB]
Phase [deg.] 0 -0.1
-200 -0.2
0 deg. 0 deg.
15 deg. 15 deg.
30 deg. 30 deg.
-400 -0.3
45 deg. 45 deg.
60 deg. 60 deg.
-600 -0.4
2 2.5 3 3.5 2 2.5 3 3.5
(a) E-plane, 20.5 GHz (b) E-plane, 20.5 GHz
200 0
Magnitude [dB]
0 -0.1
Phase [deg.]
-200 -0.2
0 deg. 0 deg.
15 deg. 15 deg.
30 deg. 30 deg.
-400 -0.3
45 deg. 45 deg.
60 deg. 60 deg.
-600 -0.4
2 2.5 3 3.5 2 2.5 3 3.5
(c) H-plane, 20.5 GHz (d) H-plane, 20.5 GHz
Figure 4.12: Simulated TM polarization reflection response of the stacked λ0 /4 × λ0 /4 crossed-dipole

element illustrated in Fig. 4.8a as lx is swept with ly = 2.9 mm at several angles of incidence.
optimal designs, and their performance could potentially be improved.
4.3.4 Comparison to Other Reported Sub-wavelength Elements

A comparison of the stacked Jerusalem cross element to other sub-λ0 /2 unit cells reported in the literature
is provided in Table 4.4. There are many different characteristics which could be considered apart from
those listed, and additionally the full picture of an element’s performance is not visible until it has been
used in a full reflectarray. However, phase range is typically lost when reducing the unit cell period, and
so it will be used as the main comparison point here. The unit cell developed in this thesis has a larger
phase range than comparable ones in the literature, apart from the multi-resonant unit cell consisting
of a loaded dipole surrounded by a loop [91]. In this case end caps are used to load the dipoles in
order which lower their resonant frequency, as is done for the Jerusalem crosses. Of note is that despite
using a multi-resonant element, a phase range of less than 360◦ was reported for the quarter-wavelength
stacked patch element [76]. While multiple resonances can be used to improve the frequency response of
an element this does not lower its resonant frequency, which is needed if the cell period is to be reduced
without sacrificing phase range.
The ‘dual-polarization’ column in Table 4.4 indicates whether or not it would be simple to use the
unit cell for dual-polarization control, not whether it was demonstrated in the work or not. Apart
from the stacked patches, the other sub-λ0 /2 unit cells in the literature use loop type elements. These
do not have an independent response to two orthogonal polarizations, and so are not suitable for the
0
0
-0.2
Magnitude [dB]
Phase [deg.]
-200
-0.4
-400
0 deg. 0 deg.
15 deg.
-0.6 15 deg.
-600 30 deg. 30 deg.
45 deg. -0.8 45 deg.
-800 60 deg. 60 deg.
-1
2 3 4 5 6 7 2 3 4 5 6 7
(a) E-plane, 20.5 GHz (b) E-plane, 20.5 GHz
0
0
-0.2
Magnitude [dB]
Phase [deg.]
-200
-0.4
-400
0 deg. 0 deg.
15 deg.
-0.6 15 deg.
-600 30 deg. 30 deg.
45 deg. -0.8 45 deg.
-800 60 deg. 60 deg.
-1
2 3 4 5 6 7 2 3 4 5 6 7
(c) H-plane, 20.5 GHz (d) H-plane, 20.5 GHz
Figure 4.13: Simulated TM polarization reflection response of the two-layer λ0 /2 × λ0 /2 stacked rect-
angular patch unit cell, pictured in Fig. 4.8b, as lx is swept with ly = 4.1 mm at several angles of
incidence. Note that the y-axis limits are different from the other angle of incidence plots on account of
the increased variation.
dual-polarization reflectarray applications considered here. This comparison shows the advantages of the
stacked Jerusalem cross unit cell: it has a large phase range and independent dual-polarization response
while fitting into a quarter-wavelength cell period. The benefits of a sub-λ0 /2 cell period are visible from
the earlier comparisons to the half-wavelength stacked patch element.
Parameter Stacked Stacked Stacked crossed Stacked

rectangular rectangular dipoles, Jerusalem
patches, patches, λ0 /4 × λ0 /4 crosses,
λ0 /4 × λ0 /4 λ0 /2 × λ0 /2 λ0 /4 × λ0 /4
Phase range − + − +
Phase drift 0 − + 0
Loss + − + 0
Cross-polarization + + + 0
Angular stability N/A − + +
Frequency dispersion N/A 0 + +
Table 4.3: Comparison of different dual-LP reflectarray unit cells based on simulation results. ‘+’, ’0’,
and ‘−’ refer to good, acceptable, and poor, respectively.
Ref. Unit cell Periodicity Phase range Dual pol.?

This Two layer stacked 0.25λ0 × 0.25λ0 548◦ Y
work Jerusalem crosses
[91] Single layer loops 0.20λ0 × 0.20λ0 314◦ N
◦
[91] Single layer loaded 0.20λ0 × 0.20λ0 563 N
dipole with loop
[88],[89] Single layer loops 0.17λ0 × 0.17λ0 > 300◦ N
[76] Two-layer stacked 0.25λ0 × 0.25λ0 > 277◦ Y
rectangular patches
Table 4.4: Comparison to reduced size reflectarray unit cells reported in the literature.
4.4 Extension to Circular Polarization

As was noted previously, it is challenging to design a reflectarray unit cell that allows for both senses of
circular polarization to be controlled independently at a common frequency. By contrast, independent
dual-LP control using a reflectarray element is a simple concept, as was demonstrated in Section 4.3.
The approach taken here will be to extend the operation of the dual-LP reflectarray element to circular
polarization by placing a linear-to-circular polarizer on top of it. By using a linear-to-circular polarizer,
the dual-CP control problem has essentially been reduced to the more straightforward dual-LP control
one. However, there are challenges with integrating the surfaces which will be explored.
The polarizer was reused from a previous work [99], in which its design and characterization are
described in detail. The polarizer unit cell is pictured in Fig. 4.14, with values for the geometrical
parameters given in Table 4.5. It consists of three layers of modified Jerusalem crosses, and has an
anisotropic response. Considering a 45◦ slant polarized electric field incident on the polarizer and
decomposing it into two components that are equal in magnitude and aligned with the x- and y-axes
in Fig. 4.14, the polarizer is designed such that the transmitted components will have a quadrature
phase difference with minimal insertion loss. Depending on whether the incident field is −45◦ or +45◦
slant polarized with respect to the x-direction, either RHCP or LHCP will be transmitted. In reverse,
when a circularly polarized wave is incident on the polarizer, slant polarization will be transmitted. The
x x
wxo
ls wcxo wcxi wxi
ws
wyo lxi gx
y lcxo lxo px y lcxi px
lyo lyi
py py
(a) Outer layers, top view (b) Inner layer, top view
1.575 mm
38 µm
1.575 mm
(c) Cross-sectional side view
Figure 4.14: Polarizer unit cell [99].
Parameter Value [mm] Parameter Value [mm]

wcxo 0.18 wcxi 0.16
lxxo 0.55 lcxi 2.06
d 0.55 gx 0.33
wd 0.50 wxi 0.59
ws 0.16 lxi 1.22
ls 1.01 wyi 1.00
wxo 0.16
lxo 2.32
wyo 0.40
Table 4.5: Polarizer unit cell dimensions[99].
polarizer has an insertion loss and axial ratio of less than 0.4 dB and 3 dB, respectively, across a 20 %
bandwidth around 20 GHz.
The full stack-up of the reflectarray and polarizer is shown in Fig. 4.15, along with a schematic view
of its operation. The height of the overall structure is 12.75 mm, which is 0.85λ0 at 20 GHz. Most of this
thickness comes from the 8 mm air gap between the structures, which was selected based on cross-polar
performance. The dependence of the cross-polar performance on the air gap size will be addressed in
Section 4.4.3.
The principle of operation of the combined structure is as follows. Considering an RHCP incident
field, as shown in Fig. 4.15, the field transmitted by the polarizer will ideally be 45◦ slant-polarized.
Taking the (x, y) coordinate systems of Fig. 4.3 and Fig. 4.14 to be common, if the reflectarray grid has
LHCP RHCP LHCP

RHCP
Polarizer
Air gap 12.75 mm

H V
Dual-LP reflectarray
Figure 4.15: Reflectarray–polarizer stack-up, with an illustration of how the two are combined to imple-
ment independent dual-CP control. The height of the air gap is 8 mm.
been rotated by 45◦ (i.e. choosing γ = 45◦ in Fig. 4.3), the transmitted field will be aligned with the
x0 -axis of the reflectarray element (VP). The transmitted field’s reflection phase can then be adjusted
by varying lx . After being reflected from the reflectarray, the VP wave will then be converted back to
RHCP by the polarizer. When an LHCP field is incident it will be converted to HP by the polarizer,
which is aligned with the y 0 -axis of the reflectarray element and can be controlled by adjusting ly . After
reflection it will be be transmitted back through the polarizer and converted to LHCP. In this manner
both the RHCP and LHCP reflection phases can be tuned independently by adjusting the dimensions of
the dual-LP reflectarray elements. It is worth emphasizing that the entire grid of the reflectarray, and
not the individual elements, has been rotated by 45◦ . This is necessary to maintain the close spacing
between adjacent elements, which gives the unit cell its large phase range.
4.4.1 Comparison to Other Dual-CP Control Methods

It is worthwhile to make some qualitative comparisons to the dual-CP control methods explored in other
works, as outlined in Section 4.2.2, to see where our proposed concept fits in.
A linear-to-circular polarizer was used to handle the CP portion of the polarization control in a dual-
surface reflector concept [78], as illustrated in Fig. 4.2c. The proposed dual-CP reflectarray operates
in a similar manner, but the two reflective surfaces are replaced by a fully planar dual-LP reflectarray.
This results in a lower profile, fully planar structure, which could have some advantages with regard to
volume, mass, and ease of deployment. Additionally, as a curved reflector is not used, array synthesis
and optimization techniques such as those explored in Chapter 3 can be easily applied without requiring
any mechanical changes.
In another work [84] two CPSSs were used instead of a combination of a polarizer and reflective
surfaces. The concept is pictured in Fig. 4.2b. The CPSS unit cells were embedded in solid metal
cavities, which are bulky and add mass and volume. Our proposed dual-CP reflectarray would have a
lower profile and be more lightweight. However, the focus of the demonstrated CPSS based dual-CP
reflector was reconfigurability, and this may have affected the design choices. In general CPSSs do not
require the use of metal cavitives, and some examples operating in our frequency range have been recently
demonstrated [100, 101]. If CPSSs similar to these examples could be modified to give an adjustable
reflection phase shift then they could potentially be used in a lower mass and volume implementation of
the concept.
Finally, independent dual-CP control was demonstrated by printing the reflectarray elements on a
paraboloidal surface [53, 54, 75, 26], as illustrated in Fig. 4.2d. While this method is elegant and does not
require the use of additional polarization control structures, printing the elements on a curved solid or
meshed reflective surfaces poses a fabrication challenge, especially compared to the low-cost and mature
processes which can be used to manufacture planar reflectarrays. Additionally, using this method the
beam locations are restricted with respect to the specular direction. Using our dual-CP reflectarray the
beam scan angles can be set arbitrarily, within reason.
4.4.2 Simulation Results

In order to design the dual-CP reflectarray and predict its performance, the response of the reflectarray
together with the polarizer must be known. As the polarizer and reflectarray unit cell have different
periodicities, and are additionally placed in lattices which are rotated with respect to one another, it is
inconvenient to apply the supercell method described in Section 2.3.2 in this case. Instead the cascaded
GSM approach, which is also described in Section 2.3.2, will be used to analyze the overall structure.
Spol SRA
Γin Z0 = 377 Ω
l = 8 mm
Figure 4.16: Transmission line model of full structure.
Using this method the structure can be analyzed as the transmission line model shown in Fig. 4.16,
where Spol and SRA are the GSMs of the polarizer and reflectarray, respectively. The air gap is modelled
as a transmission line having the characteristic impedance of free space, Z0 = 377 Ω, and this is also used
as the reference characteristic impedance for all of the blocks. Only the fundamental Floquet modes are
considered, and so the sizes of these matrices are 4 × 4. Using the Floquet modes calculator in HFSS,
the attenuation of the next set of higher order modes above the fundamental TE00 and TM00 modes
was calculated to be over 50 dB at a distance of 3.75 mm from each surface. As this was the minimum
spacing between the structures considered, which is also the periodicity of the reflectarray, the higher
order Floquet modes can be safely neglected in the analysis.
Both the reflectarray and polarizer were simulated in their own coordinate system. The polarizer
was characterized for reflection and transmission, and so its full S-matrix was known. For convenience,
the reflectarray was simulated with a ground plane and only for reflection. This allowed for the same
simulated reflectarray GSMs to be used for both the dual-LP and dual-CP reflectarrays. Thus, (2.33)
and (2.34) were used to construct the 4 × 4 GSM of the dual-LP reflectarray unit cell, and (2.40) was
used to account for the 45◦ rotation of its array grid. Finally, the S-parameters of the air gap are known
from (2.32). The full GSM of the combined structure can then be determined by multiple applications
of (2.31) to account for the two surfaces and the air gap.
The simulated characteristics of the dual-CP reflectarray unit cell are shown in Fig. 4.17 as lx and ly
are swept through their full range at 20.5 GHz and normal incidence. Similar to the dual-LP case, the
dual-CP reflection matrix is given by

" #
ΓRR ΓRL
ΓRA = , (4.2)
ΓLR ΓLL
where ΓRR and ΓLL are the co-polar reflection coefficients for incident RHCP and LHCP waves, respec-
tively. ΓLR is the cross-polar reflection coefficient when RHCP fields are incident, and ΓRL for incident
LHCP fields. It is visible that the RHCP reflection phase can be adjusted by varying lx , and the LHCP
reflection phase by changing ly . There is good independence between the two polarizations, with the
maximum drift in the RHCP co-polar reflection coefficient phase being about 28◦ as ly is swept with lx
held constant. The same applies for ΓLL as lx is swept with ly held constant. This value is slightly lower
than for the dual-LP reflectarray.
(a) (b)
(c) (d)
Figure 4.17: Simulated reflection characteristics of the cascaded polarizer–reflectarray at 20.5 GHz and
normal incidence.
The co-polar phase range of each polarization is about 526◦ , which is slightly less than the dual-LP
unit cell. From Fig. 4.17c, the maximum co-polar reflection loss is 0.84 dB, which despite being quite low
is worse than the dual-LP case. This is because the incident fields must make two passes through the
polarizer before being reflected back into free space, and so will be affected twice by its insertion loss. It
is visible from Fig. 4.17d that the cross-polarization can actually be quite high for some combinations of
element dimensions, with the worst case being about −8.1 dB. This will be addressed in the following
section.
4.4.3 Cross-polar Performance

As can be observed from the results presented in the previous section, the cross-polar performance of
the dual-CP unit cell is not as good as it was for the dual-LP unit cell. Additionally, it has a strong
dependence on the reflectarray element dimensions.
Fig. 4.18 depicts a scenario in which an RHCP wave is incident on the polarizer–reflectarray structure.
As the axial ratio of the polarizer is not 0 dB, the fields transmitted by the polarizer will have both VP
and HP components. When these components are reflected from the reflectarray and arrive again at
the polarizer both RHCP and LHCP will thus be transmitted. Due to the finite return loss of the
polarizer, instead of being fully transmitted some of the fields will be partially reflected back towards
the reflectarray and multiple reflections will continue between the two surfaces, as in a Fabry-Pérot
resonator. The magnitude and phase of these reflected waves as they reach the polarizer surface will
depend on the dimensions of the reflectarray elements and the size of the air gap, and as such the fields
transmitted by the polarizer could have a strong or weak cross-polarized component. This explains
the strong dependence on the reflectarray element dimensions that can be observed in Fig. 4.17d. The
situation is similar to the case when two microwave devices are connected together: if the return loss
of the second device is not extremely low, the performance of the chain will depend on the length of
transmission line to connect the components, as well as the reflection characteristics of each. It should
be noted that a limitation of this analysis is that both the reflectarray and polarizer are assumed to be
periodic, when in reality the reflectarray is an aperiodic, inhomogeneous structure. Additionally it is
assumed that plane waves are incident on the polarizer, when the structure of the incident wavefronts
will be more complex due to the fact that the overall structure will be fed by a spherical wave source,
and to the relatively small spacing between the reflectarray and the polarizer. Thus, sustained multiple
reflections in the manner depicted in Fig. 4.18 represents the worst-case scenario, and will not occur
exactly as shown. The actual cross-polar performance will vary with position, and the overall cross-
polar performance of the full reflectarray will depend on the aggregate scattering of all the cells which
it consists of. More advanced modelling methods would be required to capture all of these effects.
Due to its impact on the cross-polarization level, the air gap size was varied while observing the cross-
polar performance of the cascaded polarizer–reflectarray. For an incident RHCP wave, the axial ratio of
the reflected wave is shown in Fig. 4.19 at 20.5 GHz as lx is swept while ly is fixed at an intermediate
value of 2.7 mm. From these results it was found that an air gap of 8 mm resulted in a lower peak axial
ratio than the other lengths, and one that is generally lower across the range of lx , and so this spacing
was chosen for the design.
As the cross-polarization level strongly depends on both lx and ly , the cross-polar reflection coefficient
magnitude for incident RHCP is plotted in Fig. 4.20 as both of these variables are swept at 20.5 GHz.
Here, as well as in Fig. 4.19, it can be observed that changing the air gap changes the locations of the
high and low cross-polarization regions. As none of the air gap sizes produce a cross-polarization level
that is significantly lower across all of the element dimensions, the 8 mm air gap that was selected may
not be optimal. However, from the results of Fig. 4.20 the 8 mm gap does produce the lowest maximum
RHCP
+
LHCP RHCP
Polarizer
V+H
Air gap
Reflectarray
Figure 4.18: Generation of cross-polarization from polarizer mismatch.
cross-polar reflection coefficient magnitude of the air gap sizes considered.
4 mm 10 mm
10
6 mm 12 mm
8 mm 14 mm
8
Axial ratio [dB]
0
2 2.5 3 3.5
x length [mm]
Figure 4.19: Axial ratio of the reflected fields when a RHCP wave is incident on the dual-CP reflectarray
unit cell as the air gap size is varied with ly = 2.7 mm.
From the previous discussion it seems that improving the return loss of the polarizer would reduce the
dependence of the cross-polar performance on the air gap size and element dimensions. It was observed
in a previous work on combining a reflectarray and polarizer to achieve dual-CP control [102] that this
can also improve the overall cross-polar performance, and a polarizer design with better return loss was
presented. The reflection characteristics resulting from using this polarizer with the reflectarray are
shown in Fig. 4.21. From Fig. 4.21c and Fig. 4.21d it is visible that both the co-polar reflection loss
and cross-polarization magnitude have been reduced from when the polarizer design of Fig. 4.14 was
used. A comparison of some performance characteristics when each of the polarizers is used is provided
in Table 4.6. While using the polarizer with better return loss shows improvements for both phase
range and loss, the most significant improvement is in the worst case cross-polarization level which has
been decreased by over 20 dB. Despite this, the polarizer of Fig. 4.14 was used for convenience as it
had already been fabricated and characterized experimentally. While some improvements can be made
by changing the air gap size, these improvements are fairly insignificant compared to those resulting
from improving the match of the polarizer to free space. Improving the return loss also improves the
(a) 4 mm (b) 6 mm (c) 8 mm
(d) 10 mm (e) 12 mm (f) 14 mm
Figure 4.20: Cascaded polarizer–reflectarray cross-polar reflection coefficient magnitude at 20.5 GHz as
the size of the air gap between the reflectarray and polarizer is varied.
Parameter Existing polarizer [99] Higher return loss polarizer [102]

6 ΓRR range [deg.] 527 548
|ΓRR | minimum [dB] −0.84 −0.18
|ΓLR | maximum [dB] −8.14 −29.89
Table 4.6: Comparison of the simulated dual-CP reflectarray unit cell performance at 20.5 GHz when
using the existing polarizer or one designed for higher return loss.
cross-polar performance across all of the reflectarray element dimensions, whereas adjusting the air gap
does not. In future work the performance of the polarizer with the reflectarray should be taken into
account earlier on in the design process, as designing the surfaces to perform well separately does not
guarantee that they will work well when combined.
(a) (b)
(c) (d)
Figure 4.21: Simulated reflection characteristics of the cascaded polarizer–reflectarray at 20.5 GHz and
normal incidence when a polarizer unit cell with higher return loss is used [102].
4.5 Reflectarray Design
4.5.1 Antenna System Geometry

To experimentally demonstrate the concepts explored in the previous sections, two prototypes were
designed. One prototype operates in linear polarization and produces separate beams in horizontal and
vertical polarization when fed with a dual-LP feed. The other prototype operates in circular polarization
using the method described in Section 4.4, and produces independent beams in RHCP and LHCP when
fed by a dual-CP feed.
As reflectarrays are often proposed for applications where paraboloidal reflectors are used, the geom-
etry was chosen based on a paraboloidal reflector where the reflectarray could potentially replace it. A
cross-section of the offset configuration in the xz-plane is shown in Fig. 4.22, along with an equivalent
paraboloidal reflector which has the same F/D ratio. The reflectarray tilt angle was chosen so that it lies
tangent to the equivalent curved reflector. The feed offset angle was selected to give an approximately
even illumination level on both edges of the reflectarray. The focus of the reflectarray is located on the
reflector axis, which reduces the effects of blockage from the feed and its supporting structure compared
to a focus at a higher vertical position.
xR
Equivalent reflector
Reflectarray
29.1 cm Pol. 2
5◦
Pol. 1
Dual pol. feed

◦ Pol. 1 + Pol. 2
76
29◦
z
y zR
29.0 cm
Figure 4.22: The antenna system geometry with the reflectarray’s equivalent paraboloidal reflector.
(x, y, z) is the global coordinate system and (xR , yR , zR ) is the reflectarray coordinate system, where y
and yR are the same. The transverse dimension of the reflectarray, in the y-direction, is 19.5 cm.
The desired beam directions for both polarizations are shown in Fig. 4.22. ‘Pol. 1’ is vertical po-
larization for the LP case, and RHCP for the CP case. ‘Pol. 2’ is then the orthogonal polarization,
which is horizontal polarization for the LP design and LHCP for the CP design. The desired Pol. 1
beam direction is along the z-axis of the global coordinate system, which is the specular direction in
this case, and the angular separation of the two beams is 5◦ . As the reflectarray design in this work is
a proof of concept, the beam scan angles were chosen arbitrarily to demonstrate that the reflectarray
can produce a different beam for each feed polarization. However, as reflectarrays typically produce a
parasitic sidelobe in the specular direction, for a dual-polarization reflectarray such as the one designed
here there may be benefits with regard to cross-polarization to having both beams scanned away from
the specular direction [78]. For example, in the case shown in Fig. 4.22 if the feed is simultaneously
operating in RHCP and LHCP, the parasitic lobe of the LHCP beam will be in the direction of the
RHCP beam. This will increase its cross-polarization level.
The reflectarray aperture, when projected onto the xR y-plane, is elliptical with major and minor axes
of 30 cm and 19.5 cm, respectively. These dimensions were chosen to match those of the polarizer that is
placed on top of the reflectarray to obtain CP functionality. As a result of the reflectarray size and the
3.75 mm periodicity, the dual-LP reflectarray consists of 3268 elements, and the dual-CP reflectarray 3272
elements. The different numbers of elements for the two designs comes from the dual-CP reflectarray
lattice being rotated 45◦ , as a different number of elements fits into the aperture.
For an offset geometry, the angles of incidence at the reflectarray surface can become quite large.
Thus, the F/D ratio was chosen to be close to one to give a maximum angle of incidence that is less than
40◦ . Restricting the angle of incidence negatively impacts spillover efficiency, as the 10 dB beamwidth
of the horn that will be used in the experimental work is specified to be around 75◦ in its E-plane and
52.5◦ in its H-plane. The E-plane of the horn when operating in VP illuminates the major axis of the
reflectarray, and the H-plane its minor axis. These beamwidths were used with (2.49) to model the
feed as a cosq θ tapered point source with qE = 5.0 and qH = 10.5. This results in a computed edge
illumination of about −5.6 dB relative to the peak illumination level along the reflectarray’s major axis
and −5.1 dB along its minor axis. When the feed is operated in HP qE and qH in the feed model are
interchanged, as it is assumed in this section that the same illumination on the reflectarray is maintained.
This taper level is higher than desired, as typical reflector/reflectarray designs aim for an edge
illumination level of around −10 dB as a compromise between spillover and taper efficiency [21]. As
the goal of this work was not to design an antenna with high aperture efficiency, the reduced spillover
efficiency was accepted. However, the reflectarrays designed in this work can be considered to be one
panel of a larger reflectarray, which would have lower edge illumination. Spillover efficiency could also
be improved through the use of a more directive feed.
Selecting a geometry to limit the angles of arrival of the rays from the feed seems at odds with
the earlier statements regarding the angular stability of the reflectarray unit cell. However, while the
reflectarray unit cell shows good angular stability, the polarizer used for dual-CP operation was designed
for low insertion loss and axial ratio at normal incidence and its cross-polarization performance varies at
larger angles of incidence. While the maximum allowable angle of incidence could probably be increased
for the dual-LP reflectarray without a significant degradation in performance, it was desired for both
prototypes to have the same geometry in order to separate the reflectarray performance by itself from
effects resulting from integrating the polarizer.
4.5.2 Reflectarray Synthesis
The reflectarray was synthesized to produce pencil beams when the feed is operating in Pol. 1 or Pol. 2 as
shown in Fig. 4.22. The required phase distributions at 20.5 GHz, which have been shifted such that the
maximum phase is 0◦ , are shown in Fig. 4.23. The coordinate system (xR , yR , zR ), shown in Fig. 4.22,
in which the reflectarray is not tilted is used to plot the phase distributions at zR = 0. The design
frequency was chosen to be 20.5 GHz, as this was the frequency at which the minimum axial ratio of
the polarizer was measured [99]. The full phase range required to produce the Pol. 1 beam is 839◦ , and
for the Pol. 2 beam 1133◦ , both of which are not possible to implement without phase wraps using the
designed unit cell.
The phase distributions of Fig. 4.23 were synthesized using (2.22). As there are separate beams
in each polarization, ψ1 (xR , yR ) will be used to denote the required phase distribution to produce the
Pol. 1 beam, and ψ2 (xR , yR ) the phase distribution for the Pol. 2 beam. These can both be implemented
by adjusting the orthogonal element dimensions lx and ly to produce the required reflection phase for
both polarizations at each element in the reflectarray. However, it is visible from Fig. 4.7 and Fig. 4.17
that the reflection phases of the orthogonal polarizations depend on both lx and ly . For example,
considering an incident VP wave, as lx is adjusted there is an undesired phase drift in the HP reflection
phase, and the responses for both polarizations are not completely independent. Thus, a search must be
done to determine the best pair of element dimensions to satisfy the required reflection phase for both
polarizations.
(a) Pol. 1 (b) Pol. 2
Figure 4.23: Required phase distributions for both beams. The beam directions in the global coordinate
system are (a) θ = 0◦ and (b) θ = 5◦ .
For each polarization a position dependent phase error can be defined as
∆ψ (xR , yR ) = ψreq (xR , yR ) − ψLUT (lx , ly ) , (4.3)
where ψreq (xR , yR ) is the required phase shift to be produced by the element, and ψLUT (lx , ly ) are the
phase shifts in the element look-up table (LUT). The element LUT simply consists of the reflection
phase as a function of all simulated values of lx and ly , as shown in Fig. 4.7a and Fig. 4.7b for LP, and
Fig. 4.17a and Fig. 4.17b for CP. Denoting the phase errors for both polarizations as ∆ψ1 and ∆ψ2 , the
phase errors of both polarizations can be taken into account using the cost function
q
2 2
C (lx , ly ) = |∆ψ1 | + |∆ψ2 | . (4.4)
For a given element, the combination of lx and ly which minimizes C (lx , ly ) was assigned, and this
process was repeated for every element in the array.
From the simulated unit cell results, it is visible that performance is strongly dependent on the
element dimensions. This can be observed from the variation in the dispersive behaviour of the dual-
LP unit cell in Fig. 4.5c for different element lengths, or the spread in the cross-polarization level of
the dual-CP unit cell in Fig. 4.17d as lx and ly are adjusted. To maximize the performance of the
reflectarray it would be desirable to avoid using the poorer performing elements. Noting that the range
of the required phase distribution to produce each beam extends well beyond what can be achieved with
the unit cell, the phase distributions must be wrapped before assigning elements. The phase wrapping
routine can then be used to shift the distributions of elements that are assigned, potentially improving
the performance of the reflectarray.
The phase wrapping routine is as follows. First, the required phase distribution, ψ (xR , yR ), and
ψLUT (lx , ly ) are shifted such that their maximum values are 0◦ . This is shown for ψ (xR , yR ) in Fig. 4.23.
Next, recalling that for a pencil beam reflectarray we are free to add a constant ψ0 without affecting
its operation, an arbitrary value of ψ0 is added to the phase distribution. Where there are values of
ψ (xR , yR ) which extend beyond the phase range of the unit cell, multiples of 360◦ are added until this
is not the case. As adding multiples of 360◦ may cause ψ (xR , yR ) to have values that are greater than
0◦ in some locations, at these locations multiples of 360◦ are subtracted to bring them below 0◦ . The
wrapping procedure is complete when the entire required phase distribution is within the normalized
phase range of the unit cell, allowing for element dimensions to be assigned. The phase wrapping routine
is illustrated through an example in Fig. 4.24 where it has been applied to the phase distribution required
to produce the dual-LP reflectarray’s HP beam with ψ0 = 500◦ .
600 600
400 400
200 200
Phase [deg.]
Phase [deg.]
0 0
-200 -200
-400 -400
-600 -600
-800 -800
0 0.05 0.1 0.15 0.2 0.25 0.3 0 0.05 0.1 0.15 0.2 0.25 0.3
x [m] x [m]
(a) Iteration 1 (b) Iteration 2
600 600
400 400
200 200
Phase [deg.]
Phase [deg.]
0 0
-200 -200
-400 -400
-600 -600
-800 -800
0 0.05 0.1 0.15 0.2 0.25 0.3 0 0.05 0.1 0.15 0.2 0.25 0.3
x [m] x [m]
(c) Iteration 3 (d) Iteration 4
Figure 4.24: Phase wrapping routine applied to the required phase distribution to produce the HP beam
for the dual-LP reflectarray with ψ0 = 500◦ .
As is visible from Fig. 4.24, adding different values of ψ0 shifts the placements of the phase wraps
in the reflectarray phase distribution. This will affect the distributions of element dimensions which
are assigned, which will impact performance as different distributions will have higher concentrations of
better or worse performing elements. Additionally, these elements will be shifted into locations on the
reflectarray with higher or lower illumination levels. As in its simplest form the radiation pattern of a
reflectarray is the sum of the contribution of each element weighted by its illumination level, this will
also affect performance.
To assess the impact of modifying the element distributions, a parametric study was done in which
the phase offsets for both polarizations were swept from 0◦ to 1150◦ and some performance criteria were
observed. These criteria are fractional bandwidth (FBW) and XPD. Such a large range of phase offsets
was considered because of the phase wrapping implementation. For example, phase offsets of 0◦ , 360◦ ,
and 720◦ will all produce different element distributions due to the phase wraps being placed differently.
FBW is computed using
fh − fl
FBW = , (4.5)
fc
where fh is the upper edge of the band, fl is the lower edge, and fc is the centre frequency. Here
fl and fh are taken to be the frequencies at which the co-polar gain in the design direction drops by
1 dB from its peak value. fc is taken to be the frequency of peak co-polar gain. Defining bandwidth
by the frequencies at which gain has rolled off by a certain amount is commonplace in the reflectarray
community [71, 69, 76, 88, 89, 91, 60], although different authors use different roll-off amounts. As the
beam design directions are taken as the reference location, and in computing the radiation patterns the
simulated element GSMs are used, the effects of dispersion and beam squinting are taken into account
in the simulated bandwidth. XPD is taken to be the point-by-point ratio between the co-polar and
cross-polar gain patterns, and will be quoted in the design direction of each beam unless otherwise
specified.
Based on the results of the parameteric study, the combination of phase offsets which produced the
largest performance improvement for the parameter of interest was used for the final reflectarray design.
Due to their different simulated unit cell characteristics there were different performance considerations
made for the dual-LP and dual-CP reflectarrays, which will be discussed in the following sections.
Previous investigations have used optimization techniques to minimize phase errors across frequency
when assigning elements from the LUT [67], or to select ψ0 to adjust the dispersion of the phase distri-
bution [51]. The method used here, apart from not using optimization techniques, differs in that rather
than minimizing frequency dependent errors from the unit cell level the radiation pattern of the entire
reflectarray is taken into account. Although this requires more computational resources, ultimately the
radiation performance of the reflectarray is what we want to improve, and the method used here takes
this performance directly into account. It is also simple to implement, whereas framing the problem as
an optimization one can be challenging.
While the array factor method used to compute the reflectarray patterns is good for initial analysis,
it does not take into account mutual coupling between the elements in an aperiodic environment, and the
characteristics of a real feed. The effects of the latter include the fields incident on the reflectarray having
a higher cross-polarization level, and non-ideal phase and amplitude. As such, the results presented in
the next two sections are perhaps a bit optimistic. However, to more accurately predict performance
a more limited parametric study could be carried out in conjunction with full-wave simulations of the
finite reflectarray layout and the feed antenna.
Dual-LP Reflectarray
As mentioned previously, different performance criteria were considered for the dual-LP and dual-CP
reflectarrays. As is visible from Fig. 4.7d, the cross-polarization level of the dual-LP unit cell is very
low. Thus, for the dual-LP reflectarray bandwidth was the main parameter of interest. The results of
the parametric study for the dual-LP reflectarray are shown in Fig. 4.25, where regions of better and
worse performance with regard to both cross-polarization and bandwidth can be seen.
(a) (b)
(c) (d)
Figure 4.25: Results of the phase offset parametric study for the dual-LP reflectarray.
Out of the phase offsets that were studied, values of 500◦ for VP and 750◦ for HP were found to
give the highest FBW in both polarizations, and so these were used for the final dual-LP reflectarray
design. The variation of the co-polar gain and XPD with frequency before and after adding these phase
offsets are shown in Fig. 4.26a and Fig. 4.26b. ‘Before offset’ refers to the case when the phase wrapping
routine is applied with ψ0 = 0◦ for both polarizations, and ‘after offset’ to when ψ0 is selected according
to the parametric study results. The improved co-polar gain bandwidth resulting from adding the phase
offsets is visible from the increased co-polar gain at the higher frequencies for both polarizations.
The quantities of the element lengths that were assigned are shown in Fig. 4.26c and Fig. 4.26d. From
these plots it is clearly visible that adding the phase offset has shifted the distributions of lx and ly in the
reflectarray, which explains the difference in performance due to the different cross-polar and dispersive
behaviour of different elements. The dual-LP reflectarray performance before and after adding the phase
offsets is summarized in Table 4.7. The results show an increase in FBW from 19.4 % to 22.1 % for HP
and 20.2 % to 22.0 % for VP, along with a large improvement in XPD for both polarizations. However,
33 -50
Co-polar gain [dBi] 32

-60
31
XPD [dB]
30 -70
VP, no offset VP, no offset
29
VP, with offset VP, with offset
-80
HP, no offset HP, no offset
28
HP, with offset HP, with offset
27 -90
17 18 19 20 21 22 23 24 17 18 19 20 21 22 23 24
(a) (b)
(c) (d)
Figure 4.26: (a)–(b) Computed performance of the dual-LP reflectarray, and (c)–(d) distributions of
assigned element lengths before and after adding the phase offsets.
Case Gco [dBi] XPD [dB] FBW [%]

◦
HP, ψ0 =0 31.5 −56.0 19.4
HP, ψ0 = 750◦ 31.6 −69.7 22.1
VP, ψ0 = 0◦ 31.6 −56.5 20.2
VP, ψ0 = 500◦ 31.8 −70.1 22.0
Table 4.7: Simulated dual-LP reflectarray performance before and after choosing its phase offset accord-
ing to the parametric study of Fig. 4.25. The values for co-polar gain and cross-polar discrimination are
provided at 20.5 GHz.
the cross-polarization level was already very low, and is a bit idealistic as the cross-polarization of the
feed was not taken into account in this analysis. It should be noted that the highest frequency for which
the response of the unit cell was simulated was 24 GHz, and the reflectarray patterns were not computed
beyond this. As the gain of the reflectarray has not rolled off by 1 dB at this frequency, the simulated
bandwidth will actually be slightly higher than what is quoted in Table 4.7.
A comparison to some other works on wideband reflectarrays is provided in Table 4.8. It is difficult to
provide a one-to-one comparison across all of the cases shown, as there is a lot of variation in the design
parameters such as reflectarray size, F/D, frequency range, and whether simulation or measurement
results are presented. Additionally, some works faced additional challenges as they focused on covering
multiple frequency bands. Nonetheless, a qualitiative comparison is still useful to provide the simulated
performance of our dual-LP reflectarray with some context. The main criteria of comparison for wideband
Ref. Delay mechanism Design frequency/ Electrical Bandwidth remarks

band [GHz] size [λ0 ]
This Two layer stacked 20.5 13.3 × 19.9 Simulated 1 dB gain BW
work Jerusalem crosses, > 20.0 %
sub-λ0 /2
[74] Sub-wavelength 5.0–10.0 9.1 × 9.8 Measured upward trending
impedance surface with gain with frequency across
Bessel filter response band (66.7 %)
[72] Sub-wavelength stacked 8.0–12.0 8.5 × 8.5 Measured upward trending
rectangular patches gain with frequency across
designed from time delays band (40.0 %)
[91] Single layer loops, 12 8.0 × 8.0 Simulated 1 dB gain BW of
sub-λ0 /2 13.3 %
[91] Single layer loaded dipole 12 8.0 × 8.0 Value for gain BW not
with loop, sub-λ0 /2 quoted, but noted that it is
worse than the above loops
[60] Two-layer parallel 11.95 15.9 × 15.9 Measured 2 dB gain BW of
dipoles, λ0 /2 27.0 %
[88, 89] Single layer loops, 12.5 16.7 × 16.7 Measured 1 dB gain BW of
sub-λ0 /2 18.5 %
[76] Two-layer stacked 32 17.0 × 17.0 Simulated 1 dB gain BW of
rectangular patches, λ0 /4 19.1 %
[51] Single layer concentric 11.4–12.8 48.0 × 48.0 Measured 1 dB gain BW of
loops, λ0 /2 > 12.0 %, values quoted for
receive band
[71] Four layer aperture 9.65 23.0 × 22.1 Measured 1.5 dB gain BW
coupled delay lines, λ0 /2 of 26.7 %
Table 4.8: Comparison to other wideband reflectarray works.
reflectarrays is generally fractional bandwidth, for which different authors use different figures of merit.
As such, values have been provided as they were quoted in the relevant papers. The reflectarray sizes
are provided for reference, as for larger reflectarrays the dispersive behaviour of the elements becomes
more important [65], and more care must be taken to design an element phase response which is linear
with frequency. Where it was not explicitly quoted in the source, the reflectarray electrical size was
computed based on the design frequency and reflectarray dimensions.
Based on simulation results, the performance of the stacked Jerusalem cross unit cell is in line with
or better than other works using sub-λ0 /2 unit cells [76, 88, 89, 91]. Additionally, all of these works
but one [76] use loop type elements which would not be suitable for applications requiring independent
dual-polarization control.
For reflectarrays using traditional phase delay methods with λ0 /2 unit cells, the works using parallel
co-planar dipoles [60] and aperture coupled delay lines [71] show excellent performance. The co-planar
dipoles are similar to the Jerusalem cross unit cell used in this thesis, in that the unit cell is a two layer
multi-resonant structure, although the orthogonally polarized elements are not electrically connected.
The electrical size of the reflectarray is also similar to ours. However, the authors used a different
bandwidth metric, and so a direct bandwidth comparison cannot be made. For the work using delay
lines, the lines were made long enough such that the required phase distribution of the reflectarray
could be fully compensated without any phase wrapping. This was demonstrated for one polarization.
Although a second polarization could be controlled by adding an orthogonal coupling slot and second
delay line, it may be difficult to fit the two lines within the same size unit cell while avoiding coupling
between them, which would contribute to cross-polarization and reduce the polarization independence of
the element. Additionally, the structure requires a minimum of three layers to accommodate the delay
line, coupling slot, and radiating element, whereas the element used in this thesis requires two layers.
The state of the art bandwidth performance has been achieved using non-traditional design tech-
niques. One example is an impedance surface reflector designed to produce a Bessel filter response [74],
which functioned over a 66.7 % bandwidth. However, to realize the elements required to implement
the filter response, meandered lines and inter-digitated capacitors were used which respond to only one
linear polarization. In another work the time-delay profile of the reflectarray was compensated by as-
signing elements based on their group delay, as determined from the linear trend of their reflection phase
variation with frequency [72]. Sub-wavelength unit cells were used to produce the linear phase variation
with frequency, and as such there will be limits on the maximum group delay that the unit cell can
compensate. This could potentially be improved by increasing the number of layers or applying unit cell
miniaturization techniques, while taking care to maintain the linearity of the element’s phase variation.
As the phase variation with frequency of the Jerusalem cross element used here is also quite linear, this
method could potentially be applied to our reflectarray.
Dual-CP Reflectarray
Comparing Fig. 4.7d with Fig. 4.17d, it is visible that the dual-CP unit cell has much a higher cross-
polarization level than the dual-LP one. Thus, rather than focusing on bandwidth, cross-polarization
was the performance metric of interest for the dual-CP reflectarray. The results of sweeping the phase
offsets of both the RHCP and LHCP phase distributions are shown in Fig. 4.27. As was the case for
the dual-LP reflectarray, there regions of better and worse performance with regard to both bandwidth
and cross-polarization. Based on the results of Fig. 4.27c and Fig. 4.27d, phase offsets of 650◦ and 300◦
were chosen for RHCP and LHCP, respectively.
The co-polar gain and XPD of the dual-CP reflectarray as frequency is swept are shown in Fig. 4.28a
and Fig. 4.28b. The results are summarized in Table 4.9. The ‘no offset’ and ‘with offset’ cases have
the same meaning as in the previous section. It is visible that by adjusting the phase offset the shape
of the simulated XPD response with frequency has been changed, resulting in an improvement of 4.6 dB
for RHCP and 11.7 dB for LHCP at 20 GHz. Although it was not the focus of the parametric study,
the FBW has also improved slightly. For the RHCP beam, the gain roll-off at the lower frequencies
has become steeper, but closer to the design frequency there is not as much of an impact. Fig. 4.28c
and Fig. 4.28d show that the element length distributions have been changed, as was the case for the
dual-LP reflectarray.
It should again be emphasized that the simulated results here are quite optimistic, as they do not
take into the effects of using a real feed, or take into account angle of incidence effects on the combined
polarizer–reflectarray unit cell. These non-ideal effects will be more pronounced for the dual-CP reflec-
tarray than the dual-LP reflectarray, as any phase errors which cause the phase difference between the
orthogonal linear components of the fields radiated by the reflectarray to differ from 90◦ will produce
(a) (b)
(c) (d)
Figure 4.27: Results of sweeping the phase offsets of both polarizations for the dual-CP reflectarray.
Case Gco [dBi] XPD [dB] FBW [%]

◦
RHCP, ψ0 = 0 31.3 −26.5 16.2
RHCP, ψ0 = 650◦ 31.3 −31.1 16.9
LHCP, ψ0 = 0◦ 31.1 −24.8 16.3
LHCP, ψ0 = 300◦ 31.1 −36.5 16.3
Table 4.9: Simulated dual-CP reflectarray performance before and after choosing phase offset according
to parametric study of Fig. 4.27 . The values for co-polar gain and cross-polarization ratio are given at
20.5 GHz.
32 0
Co-polar gain [dBi]

30 -10
XPD [dB]
28 -20
RHCP, no offset
RHCP, with offset RHCP, no offset
26 -30 RHCP, with offset
LHCP, no offset
LHCP, with offset LHCP, no offset
LHCP, with offset
24 -40
17 18 19 20 21 22 23 24 17 18 19 20 21 22 23 24
(a) (b)
(c) (d)
Figure 4.28: (a)–(b) Computed dual-CP reflectarray performance, and (c)–(d) distributions of assigned
elements before and after adding the phase offset.
cross-polarization.
4.6 Experimental Results

With the reflectarray geometry finalized and the phase distributions and elements assigned, the next
step was to generate layouts for the reflectarrays and fabricate them. They could then be characterized
through measurements in order to test the concepts discussed in the previous sections experimentally.
As the previous simulations were performed using an array factor method, prior to measurements a
logical next step would be to simulate the finite reflectarrays with the real feed using full-wave methods.
However, the reflectarrays are electrically large and contain large numbers of elements with fine features.
It was found that full-wave simulations were not feasible using the available computational resources and
simulation tools. This is especially true for the dual-CP reflectarray, in which the three-layer polarizer
and air gap would have to be simulated in addition to the two layer reflectarray. Thus, it was decided
to proceed directly to measurements.
The proof of concept prototype reflectarrays were fabricated on two layers of Rogers RO3003 lam-
inates (r = 3.00, tan δ = 0.0010) that were bonded together using Rogers 3001 (r = 2.28, tan δ =
0.0030). Two separate prototypes were designed and built, corresponding to the different designs for
dual-LP and dual-CP operation discussed in Section 4.5. The two fabricated reflectarrays are pictured
in Fig. 4.29. Note that the dual-CP reflectarray is pictured without the polarizer that must be placed on
top of it to facilitate CP operation. The polarizer was printed on two Rogers RT/Duroid 5880 laminates
(r = 2.20, tan δ = 0.0009) that were bonded together using the same bonding film [99]. For the dual-CP
reflectarray the air gap between the two structures was implemented using 8 mm nylon spacers.
Although in previous sections the reflectarray designs were elliptical, from Fig. 4.29 it is visible that
the fabricated reflectarrays do not form full ellipses. This is because the full area of the panels that the
reflectarrays were etched on could not be used as some room needed to be left for fabrication processes.
This truncation reduced the number of elements by 256 for the dual-LP reflectarray, and by 271 for
the dual-CP reflectarray. The difference in the numbers of elements again comes from the dual-CP
reflectarray grid being rotated by 45◦ . While this truncation was not taken into account in the previous
sections, in this section it has been modelled by appropriately adjusting the reflectarray sizes in the
array factor simulations.
In Fig. 4.30 the dual-LP reflectarray is pictured in the NSI near-field scanner used for all of the
measurements, along with the feed horn and the aluminum mounting structure. Absorbers were attached
to the mounting structure to reduce scattering from it which would influence the measurement results.
The reflectarray is mounted on a fibreglass panel on a sheet of acrylic using nylon screws and nuts. The
fibreglass panel was added as the reflectarray is relatively thin and so it is prone to bending. As the
mounting holes in the acrylic were drilled by hand using a drill press, the placements of the holes in the
acrylic sheet and reflectarray do not exactly line up, which resulted in bending of the reflectarray. This
was exacerbated for the dual-CP reflectarray, where additional mounting holes needed to be drilled in
the polarizer. Fibreglass is not fully rigid, and so this allowed the reflectarray surface to be flattened
out more by adjusting the screws around its border.
It is visible from Fig. 4.30 that the reflectarray geometry is not set up as in Fig. 4.22. Due to
the frequencies of interest and high directivity of the reflectarray, the beam locations and quality are
very sensitive to positioning errors. It was found that using the geometry of Fig. 4.22 made it quite
challenging to position the various components correctly. Using the geometry shown in Fig. 4.30 where
the reflectarray is not rotated with respect to the vertical axis of the measurement setup made it easier
to position everything up in a repeatable manner, as only the feed offset angle needs to be set properly,
(a) Dual-LP reflectarray (b) Dual-CP reflectarray, without polarizer
Figure 4.29: Fabricated reflectarray prototypes.
Figure 4.30: Dual-LP reflectarray with mounting structure and feed horn, set up in near-field scanner.
rather than the offset angles of both the reflectarray and feed. As a consequence, for the measured
results the Pol. 1 and Pol. 2 beams will be located at 14◦ and 19◦ , respectively, instead of 0◦ and 5◦ .
The reflectarrays were fed by an A-INFO dual-polarization horn antenna, also pictured in Fig. 4.30,
which is specified to operate from 15 GHz to 22 GHz. The feed horn contains an ortho-mode transducer
(OMT), and so it can support both HP and VP excitations using a separate port for each. However,
during measurements rather than using the HP and VP ports on the horn, the VP port was used and
the horn was simply rotated by 90◦ to operate it in HP. This is because it was found experimentally
that there are differences in the phase centre locations of both ports, which would influence the dual-CP
reflectarray performance as the measured CP patterns are obtained from post-processing the results of
two orthogonally polarized LP measurements. However, rotating the horn means that the illumination
of the reflectarray will be different for the two polarizations. Referring to Fig. 4.22, when operating in
VP the broader plane of the horn’s main lobe illuminates the major axis of the reflectarray, and when
rotated by 90◦ it will instead illuminate the reflectarray’s minor axis. This was taken into account in
the computational results presented in this section by swapping the qE and qH parameters of the feed
model when it is horizontally polarized. The phase centre of the feed was found through measurements
to be located roughly 2 cm inside its mouth, and so this was placed at the design focal point of the
reflectarray.
In order to determine the absolute gain of the measurements the gain comparison method was used.
A reference standard gain horn (SGH), pictured in Fig. 4.31, was measured using the same cables and
adaptors that were used to connect the reflectarray feed horn. The difference in magnitude between
the measured SGH patterns and its known gain was then computed as a function of frequency, and the
resulting offsets were added to the reflectarray measurements in post-processing.
Figure 4.31: Standard gain horn used to determine absolute gain, set up in near-field scanner.
In the following sections, the measurement results will be presented, along with a comparison to the
expected results from array factor computations. For the array factor results, the feed model used is the
same cosq θ tapered point source that was used in Section 4.5. The feed model parameters are qE = 5.0
and qH = 10.5 when operating in VP, and as previously stated the two parameters are interchanged for
HP operation. Referring to Fig. 4.22, the simulated elevation cuts were taken in the xR zR -plane. The
azimuth cut plane was formed by rotating the yzR -plane to the design direction of each beam.
4.6.1 Feed Horn

Before delving into the reflectarray experimental results, some measurements of the feed that were taken
using the near-field scanner will be covered. The normalized measured magnitude patterns of the feed
at the design frequency of 20.5 GHz are shown in Fig. 4.32. For the co-polar pattern cuts in Fig. 4.32a
and Fig. 4.32b a comparison to the cosq θ feed model that was used to obtain the array factor results
is provided. It is visible that the model patterns track the envelope of the measured patterns, although
the measured results shown more variation over space. The agreement is better in the elevation plane
when the feed is operated in HP, and the azimuth plane when it is operated in VP, using the two ports
of the horn. When feeding the reflectarray only the VP port of the horn was used. The variation in the
beamwidth of the feed was not taken into account in simulations.
0 0
Magnitude [dB]
Magnitude [dB]
-5 -5
VP, measured VP, measured

-10 VP, cosq 3
-10 VP, cosq 3
HP, measured HP, measured
HP, cos q 3 HP, cos q 3
-15 -15
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30
(a) Co-polarized elevation patterns at 20.5 GHz (b) Co-polarized azimuth patterns at 20.5 GHz
(c) Cross-polarized pattern at 20.5 GHz when the VP (d) Cross-polarized pattern at 20.5 GHz when the HP
port is excited port is excited
Figure 4.32: Measured normalized magnitude patterns of the dual-polarization feed horn with a com-
parison to the cosq θ model used.
The measured cross-polar patterns of the feed are shown in Fig. 4.32c and Fig. 4.32d, where the
values have been normalized to their respective co-polar peaks. Across the measured space, the worst-
case cross-polarization level is −11.2 dB when the VP port is excited, and −9.9 dB when the HP port
is used. These values occur away from the principal planes of the feed. When only the principal plane
patterns are considered, the worst-case cross-polarization is −26.4 dB when the feed is operating in VP,
and −20.6 dB for HP. While these values are quite low they are much higher than what the cosq θ model
predicts, which assumes there is no cross-polarization in the principal plane patterns of the feed.
Two sets of measured co-polar phase patterns of the feed are shown in Fig. 4.33 and Fig. 4.34. Using
the NSI software, which is used to control the near-field scanner and process measurement data, once
the near-field of the antenna under test has been measured it is possible to change the origin of the
200 200
VP, co-pol VP, co-pol
HP, co-pol HP, co-pol
Phase [deg.] 100 100
Phase [deg.]
0 0
-100 -100
-200 -200
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30
(a) Elevation cuts at 20.5 GHz (b) Azimuth cuts at 20.5 GHz
(c) 20.5 GHz with VP port excited (d) 20.5 GHz with HP port excited
Figure 4.33: Measured co-polarized phase patterns of the dual-polarization feed when the origin of the
measurement coordinate system is located at the mouth of the horn.
coordinate system used to perform the near-field to far-field transformation. The location of the origin
was swept and the phase patterns were observed. In Fig. 4.33 this origin was set to be at the mouth of
the feed, whereas for Fig. 4.34 it was set to be about 2 cm inside it. It was determined that the 2 cm
location produced the flattest phase patterns, and so this was used as the effective phase centre of the
feed and positioned at the design location of the reflectarray’s focus. The same location of the feed’s
effective phase centre was also obtained using scripting tools provided by NSI.
For a reflectarray phase distribution synthesized according to (2.22), in order for phase errors to
not be introduced by the feed it must act as a spherical source over the region of its pattern which
illuminates the reflectarray. Referring to Fig. 4.22, this means that the feed should have a flat phase
pattern across the region −23◦ ≤ El ≤ 22◦ and −19◦ ≤ Az ≤ 19◦ , where the angles are expressed in
the coordinate system of the feed. From the results of Fig. 4.34, when the horn is fed from its VP
port the measured phase variation over these regions in elevation and azimuth are about 41◦ and 36◦ ,
respectively. While not extremely large, this still represents a deviation from the perfect spherical source
assumed for the design and computational results, and will have an impact on the experimental results.
The phase flatness could also be impacted by the positioning of the feed, as if it is rotated away from
the measurement plane in azimuth or elevation this would introduce additional phase variation.
200 200
VP, co-pol
HP, co-pol
Phase [deg.] 100 100
Phase [deg.]
0 0
-100 -100 VP, co-pol

HP, co-pol
-200 -200
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30
(a) Elevation cuts at 20.5 GHz (b) Azimuth cuts at 20.5 GHz
(c) 20.5 GHz with VP port excited (d) 20.5 GHz with HP port excited
Figure 4.34: Measured co-polarized phase patterns of the dual-polarization feed when the origin of the
measurement coordinate system is located about 2 cm inside the horn.
4.6.2 Dual-LP Reflectarray

The measured co-polar and cross-polar principal plane gain patterns of the dual-LP reflectarray at some
selected frequencies between 18 GHz and 22 GHz are shown in Fig. 4.35 for the case when the feed is
vertically polarized, and Fig. 4.36 when it is horizontally polarized. Although this was noted previously,
for clarification the horizontally polarized feed refers to when its VP port is excited, but it has been
rotated by 90◦ . The radiation patterns across the entire measurement space are shown in Fig. 4.37.
The selected frequencies correspond to the edges of the measurement band, the design frequency, and
an intermediate one. Frequencies below 18 GHz were not measured as from the simulated results the
reflectarray gain drops off sharply at the lower frequencies. Frequencies above 22 GHz were not measured
as this is beyond the specified operational bandwidth of the feed, and a higher frequency feed with a
similar form factor was not available at the time when the measurements were taken.
The measured and design beam peak locations are given in Table 4.10, showing that there is a slight
error in the positions of the measured beams. In elevation the worst-case pointing error is 0.5◦ , whereas
in azimuth this is 1.7◦ . These pointing errors are likely a result of the positioning of the reflectarray
and feed in the experimental setup, and the higher error in azimuth could be a result of the reflectarray
mount being slightly tilted in the horizontal direction, or the feed being displaced horizontally by a
small amount relative to the reflectarray. Due to the shifting of the beam locations due to measurement
Measured location [◦ ] Design location [◦ ]

Polarization
Az. El. Az. El.
VP 1.5 13.9 0.0 14.0
HP 1.7 18.5 0.0 19.0
Table 4.10: Measured co-polar beam peak directions at 20.5 GHz and their corresponding design direc-
tions for the dual-LP reflectarray.
errors, rather than using the design directions for the principal pattern cut planes, the measured peak
directions were used. The elevation pattern cuts were taken through the location in azimuth of the VP
co-polar gain peak at 20.5 GHz. The azimuth cuts of each beam were taken through the location in
elevation of their respective co-polar peaks at 20.5 GHz. By taking the cuts through a fixed position
across all frequencies the effects of beam squinting are taken into account. 20.5 GHz was taken to be the
reference for the pattern cuts as it is the design frequency.
Apart from the small pointing errors, the measured radiation patterns show that the beams are
approximately in the design directions and separated by 5◦ , and the reflectarray maintains good radiation
characteristics in both polarizations over the measured frequency range. The measurement results clearly
show the generation of independent beams in VP and HP. There is relatively good correlation between
the simulated and measured performance of the reflectarray, although there are some deviations, the
most notable of which are a drop in gain, beam aberrations, and an increased cross-polarization level.
These characteristics are likely due to the non-idealities of the feed that were not accounted for, such
as the phase pattern and increased cross-polarization level, as discussed in Section 4.6.1. Additional
errors could be due to positioning, as the experimental setup was found to be very sensitive to this.
Another contribution could be manufacturing errors, such as over-etching or under-etching the elements,
or bending of the reflectarray surface. The reduction gain could also be due to increased losses in
the reflectarray compared to simulations. The unit cells were not experimentally verified, such as by
measuring the reflection coefficient of a sample placed at the end of a rectangular waveguide, and as such
the dielectric and conductor losses could be higher than expected. Changes in the dielectric constant
of the substrate with frequency were not modelled, which could potentially shift the resonances of the
unit cell and introduce phase errors. Finally, blockage from the feed and mounting structure were not
accounted for, although based on the offset configuration and scan angles of the beam these effects should
not be very significant. More investigations and characterization would be required to determine the
root source of the discrepancies between measurement and simulation. However, the feed characteristics
and positioning errors are likely the most significant contributors, and their impact was very noticeable
when performing the measurements.
In Fig. 4.38a and Fig. 4.38b the co-polar and cross-polar gain variation of the dual-LP reflectarray
with frequency are plotted, respectively, where a comparison between simulation and measurement
results is provided. Two cases are shown for the computational results: one case where the simulated
element GSMs are used, and another where the elements are treated as ideal phase shifters. The
ideal phase shifter refers to an element which provides the exact phase required at each position on the
reflectarray across all frequencies without introducing any cross-polarization or loss. As such it represents
the maximum theoretical gain that this reflectarray could potentially achieve with the feed model that
was used, and the only reductions in efficiency in this case would be due to spillover and non-uniform
illumination of the reflectarray. As the ideal phase shifters do not generate any cross-polarization, they
are not included in the cross-polar performance comparison.
It is visible from Fig. 4.38a that all three of the curves track upwards with frequency over the
measurement band, although the measured results show a reduction in gain that was seen in the pattern
cuts. However, the measured gain does not roll off enough by the highest measured frequency to be able
to quantify the fractional bandwidth. From Fig. 4.38b the measured XPD of the reflectarray is less than
−30 dB in the direction of the main beam across all of the measured frequencies, which is quite low,
although it is a lot higher than the simulated results. However, the simulated results can be considered
quite idealistic, as they did not take the feed cross-polarization into account.
In Fig. 4.38c the total directivity of the reflectarray is plotted against frequency, with a comparison to
the computational results using the simulated element GSMs. It is again visible that both curves track
upwards with frequency, but as the directivity is not separated into co-polar and cross-polar components
the difference is not as severe as in Fig. 4.38a. The drop in directivity again suggests that phase errors
are present, and could also result from a different amplitude taper than what was assumed for the
computational results. The performance as a function of frequency will also differ from the simulations
because of the frequency dependent characteristics of the feed which were not accounted for.
VP co-pol, measured
30 VP co-pol, measured 30 VP co-pol, AF
VP co-pol, AF VP x-pol, measured
20 VP x-pol, measured 20 VP x-pol, AF
Gain [dBi]
Gain [dBi]
VP x-pol, AF
10 10
0 0
-10 -10
-20 -20
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30

(a) 18.0 GHz elevation cut (b) 18.0 GHz azimuth cut
VP co-pol, measured
30 VP co-pol, measured 30 VP co-pol, AF
VP co-pol, AF VP x-pol, measured
20 VP x-pol, measured 20 VP x-pol, AF
Gain [dBi]
Gain [dBi]
VP x-pol, AF
10 10
0 0
-10 -10
-20 -20
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30

(c) 19.0 GHz elevation cut (d) 19.0 GHz azimuth cut
VP co-pol, measured
30 30 VP co-pol, AF
VP co-pol, measured VP x-pol, measured
VP co-pol, AF VP x-pol, AF
20 VP x-pol, measured 20
Gain [dBi]
Gain [dBi]
VP x-pol, AF
10 10
0 0
-10 -10
-20 -20
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30

(e) 20.5 GHz elevation cut (f) 20.5 GHz azimuth cut
VP co-pol, measured
VP co-pol, AF
30 VP co-pol, measured 30
VP x-pol, measured
VP co-pol, AF VP x-pol, AF
20 VP x-pol, measured 20
Gain [dBi]
Gain [dBi]
VP x-pol, AF
10 10
0 0
-10 -10
-20 -20
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30

(g) 22.0 GHz elevation cut (h) 22.0 GHz azimuth cut
Figure 4.35: Measured patterns cuts of the dual-LP reflectarray at some selected frequencies when the
feed is vertically polarized, with a comparison to AF predictions. The black dash-dot line shows the
design direction of the beams in the elevation plane.
HP co-pol, measured
30 HP co-pol, measured 30 HP co-pol, AF
HP co-pol, AF HP x-pol, measured
20 HP x-pol, measured 20 HP x-pol, AF
Gain [dBi]
Gain [dBi]
HP x-pol, AF
10 10
0 0
-10 -10
-20 -20
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30

HP co-pol, measured
30 30 HP co-pol, AF
HP co-pol, measured
HP x-pol, measured
HP co-pol, AF
20 20 HP x-pol, AF
HP x-pol, measured
Gain [dBi]
Gain [dBi]
HP x-pol, AF
10 10
0 0
-10 -10
-20 -20
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30

HP co-pol, measured
30 30 HP co-pol, AF
HP co-pol, measured
HP x-pol, measured
HP co-pol, AF
20 20 HP x-pol, AF
HP x-pol, measured
Gain [dBi]
Gain [dBi]
HP x-pol, AF
10 10
0 0
-10 -10
-20 -20
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30

HP co-pol, measured
HP co-pol, AF
30 HP co-pol, measured 30
HP x-pol, measured
HP co-pol, AF HP x-pol, AF
20 HP x-pol, measured 20
Gain [dBi]
Gain [dBi]
HP x-pol, AF
10 10
0 0
-10 -10
-20 -20
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30

Figure 4.36: Measured patterns cuts of the dual-LP reflectarray at some selected frequencies when the
feed is horizontally polarized, with a comparison to AF predictions. The black dash-dot line shows the
design direction of the beams in the elevation plane.
(a) 18.0 GHz (b) 19.0 GHz (c) 20.5 GHz (d) 22.0 GHz
(e) 18.0 GHz (f) 19.0 GHz (g) 20.5 GHz (h) 22.0 GHz
(i) 18.0 GHz (j) 19.0 GHz (k) 20.5 GHz (l) 22.0 GHz
(m) 18.0 GHz (n) 19.0 GHz (o) 20.5 GHz (p) 22.0 GHz
Figure 4.37: Dual-LP reflectarray measured patterns at some selected frequencies across two-dimensional
space. In (a)–(h) the feed is vertically polarized, and in (i)–(p) it is horizontally polarized.
VP, measured
36 HP, measured
-30
VP, AF
Co-polar gain [dBi]
34 HP, AF -40
VP, ideal, AF
XPD [dB]
HP, ideal, AF
32 -50 VP, measured
HP, measured
30 -60 VP, AF
HP, AF
28 -70
26 -80
18 19 20 21 22 18 19 20 21 22
(a) (b)
34
Directivity [dBi]
33
32
VP, measured
VP, AF
31
HP, measured
HP, AF
30
18 19 20 21 22
Frequency [GHz]
(c)
Figure 4.38: Measured (a) co-polar gain, (b) cross-polar discrimination, and (c) total directivity variation
with frequency of the dual-LP reflectarray with a comparison to array factor predictions. The values are
taken at the intersections of the principal plane cuts.
4.6.3 Dual-CP Reflectarray

The dual-CP reflectarray was measured using the same setup as the dual-LP reflectarray. As a dual-CP
feed was not available, post-processing was used to synthesize the CP excitation numerically. Operating
the dual-LP feed in VP, the HP and VP fields of the reflectarray were measured. This will be termed
the VP measurement. The same measurement was repeated with the feed operating in HP, by rotating
it 90◦ as for the dual-LP reflectarray, to obtain the HP measurement. The rotation of the feed rather
than using the HP port is important for the dual-CP case, as there are phase differences between the two
ports which would contribute to cross-polarization. Taking the two LP measurements, either the HP
or VP measurement was delayed by 90◦ , and the two were then summed vectorially to simulate either
an RHCP or LHCP excitation. Which of the LP measurements is delayed by 90◦ determines whether
RHCP or LHCP is produced. The resulting fields were then projected onto a CP basis in order to obtain
the CP radiation patterns. The procedure is similar to that which was used to obtain the CP radiation
patterns for the computational results, except in the measurements only the fields resulting from the
reflectarray are accessible, and the delay cannot be directly added to the feed. A similar approach was
taken in a recent transmitarray work [103]. As the feed is dual-polarized and has an OMT, if the delay
differences between the two ports are compensated for it could be fed by a quadrature hybrid coupler
to operate in CP. However, such a coupler was not available and would not be trivial to design for the
frequency range under consideration.
Some measured pattern cuts with comparisons to simulations are shown in Fig. 4.39 and Fig. 4.40
for the synthesized RHCP and LHCP feeds, respectively. The two-dimensional radiation patterns across
the measurement space are plotted in Fig. 4.41. The same frequencies are presented as were shown for
the dual-LP reflectarray, although wideband performance was not expected from this reflectarray due to
the effects of the polarizer. As was the case for the dual-LP reflectarray, the beams are slightly shifted
in azimuth and elevation from the design directions due to positioning errors in the experimental setup.
The locations of the peaks are provided in Table 4.11, and the same method of choosing the locations
of the principal pattern cut planes as discussed for the dual-LP reflectarray was used.
From the measured results, from 18 GHz up to the design frequency the generation of two independent
beams in RHCP and LHCP is visible, and the beams are separated by about 5◦ as intended. However,
there is a marked increase in cross-polarization compared to the computational results, and the corre-
lation between measurement and simulation is not as strong as it was for the dual-LP reflectarray. At
the higher end of the measured frequencies it can be seen from the 22 GHz patterns that, although good
performance was not expected at these frequencies, the reflectarray loses its dual-polarization function-
ality and the co-polar and cross-polar patterns are at a similar level. The increase in cross-polarization
also results in a reduction in co-polar gain, although it is possible that another contribution to this is
increased losses compared to the simulation results.
The dual-CP reflectarray would suffer from the same effects which produced the aberrations seen in
the dual-LP reflectarray patterns, such as phase errors resulting from both the feed and positioning, as
well as manufacturing defects. Additionally, the polarizer incurred some damage when drilling mounting
holes in it to attach it to the reflectarray. One of the corners became caught on the drill bit, causing the
layers in one corner to separate slightly. Under a highly tapered illumination, as in a typical reflectarray
setup, the amplitude of the incident field at this section would be quite low and its contribution to
the overall radiation pattern would not be very significant. However, in this case the edge taper of the
reflectarray is quite high and so this damage may have a greater effect. Regarding phase errors resulting
Measured location [◦ ] Design location [◦ ]

Polarization
Az. El. Az. El.
RHCP 0.9 13.7 0.0 14.0
LHCP 0.6 18.7 0.0 19.0
Table 4.11: Measured co-polar beam peak directions at 20.5 GHz and their corresponding design direc-
tions for the dual-CP reflectarray.
from the phase pattern and positioning of the feed, for the dual-LP reflectarray this would simply cause
aberrations in the radiated beams. However, as the dual-CP reflectarray relies on measurement from
both linear polarizations, deviations in phase from a spherical source will result in a position dependent
phase difference in the co-polarized incident field depending on whether the feed is operated in VP or
HP. When the LP measurements are post-processed by delaying either the HP or VP measurement by
90◦ , these phase differences will contribute to an increased cross-polarization level.
The variation in co-polar gain and cross-polar discrimination with frequency are plotted in Fig. 4.42a
and Fig. 4.42b, respectively. From these plots, the reduced co-polar gain is again visible, and it rolls
off quite steeply beyond the design frequency. As was observed from the pattern cuts, there is a visibly
higher cross-polarization level in the measured results compared to the computational ones. This is
particularly true at the higher frequencies, and for the LHCP beam at 22 GHz the cross-polar gain
becomes higher than the co-polar gain. However, at the lower frequencies the cross-polarization level
is lower. From Fig. 4.39 and Fig. 4.40 there is generally better agreement between the measured and
simulated radiation patterns at the lower frequencies.
Unlike the dual-CP reflectarray, the dual-LP reflectarray showed fairly low levels of cross-polarization
in its measured patterns. One difference between the two cases is that the dual-CP reflectarray requires
the use of a polarizer. In Fig. 4.43a the simulated transmitted cross-polarization magnitude of the
polarizer is plotted as a function of frequency as the angle of incidence is swept from 0◦ to 40◦ , which
accounts for the range of incidence angles across the reflectarray. For the case considered, RHCP is
incident and so the transmitted cross-polar component is HP. It is visible that the point of lowest cross-
polarization shifts downward in frequency as the angle of incidence increases. At the design frequency
of 20.5 GHz, the magnitude of the transmitted cross-polarized component increases by about 10 dB as
the angle of incidence increases from normal incidence to 40◦ . The variation at the lower frequencies
is not as strong. For example, at 19 GHz where less discrepancy was observed between the measured
and computed patterns, the transmitted cross-polarization varies by about 3.4 dB. Additionally at the
lower frequencies the cross-polarization transmitted by the polarizer improves instead of worsening as
the angle of incidence increases.
Considering the rays from a point source feed placed at the focal point of the reflectarray, the
angles of incidence across its surface are plotted in Fig. 4.43b. It is visible that although for the unit
cell characterization and reflectarray design normal incidence was assumed, that a large portion of the
reflectarray does not meet this condition. This will result in errors which contribute to the negative
effects observed in the measured results. Angle of incidence effects could potentially be mitigated by
designing the polarizer for oblique incidence instead of normal incidence, as was done in the previously
mentioned dual-surface reflector work [78]. For example, in this case if the polarizer was designed for an
incidence angle of 20◦ rather than normal incidence, the variation in angle of incidence from the polarizer
design angle would be ±20◦ instead of +40◦ , which may be more tolerable. As the rays reflected from the
reflectarray will ideally be parallel, due to the offset angle and beam directions a 20◦ angle of incidence
is additionally closer to what the rays will experience on their second pass through the polarizer.
Although to get a full picture of the cross-polar performance both the dual-LP reflectarray and
polarizer must be considered together, the return loss characteristics of the polarizer mentioned earlier
coupled with the angle of incidence variation of both structures represent position dependent errors
which were not accounted for in the simulation and design phase. Due to these effects, and the non-
ideal characteristics of the feed, the discrepancies between the computational and experimental results
are perhaps not surprising. Further investigations would be required to determine the root cause of
the errors with certainty. However, using the measured incident fields from the real feed to synthesize
the reflectarray phase distributions would likely improve results for both the dual-LP and dual-CP
reflectarrays, and should be investigated in future work.
RHCP co-pol, measured

RHCP co-pol, measured RHCP co-pol, AF
30 RHCP co-pol, AF 30
RHCP x-pol, measured
RHCP x-pol, measured RHCP x-pol, AF
20 RHCP x-pol, AF 20
Gain [dBi]
Gain [dBi]
10 10
0 0
-10 -10
-20 -20
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30

30 30 RHCP co-pol, AF
RHCP co-pol, measured RHCP x-pol, measured
RHCP co-pol, AF RHCP x-pol, AF
20 RHCP x-pol, measured 20
Gain [dBi]
Gain [dBi]
RHCP x-pol, AF
10 10
0 0
-10 -10
-20 -20
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30

30 RHCP co-pol, measured 30 RHCP co-pol, AF
RHCP co-pol, AF RHCP x-pol, measured
20 RHCP x-pol, measured 20 RHCP x-pol, AF
Gain [dBi]
Gain [dBi]
RHCP x-pol, AF
10 10
0 0
-10 -10
-20 -20
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30


30 RHCP co-pol, measured 30 RHCP co-pol, AF
RHCP co-pol, AF RHCP x-pol, measured
20 RHCP x-pol, measured 20 RHCP x-pol, AF
Gain [dBi]
Gain [dBi]
RHCP x-pol, AF
10 10
0 0
-10 -10
-20 -20
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30

Figure 4.39: Measured pattern cuts of the dual-CP reflectarray resulting from the synthesized RHCP
feed at some selected frequencies, with a comparison to AF predictions. The black dash-dot line shows
the design direction of the beams in the elevation plane.
LHCP co-pol, measured

30 30 LHCP co-pol, AF
LHCP x-pol, measured
LHCP co-pol, AF
20 20 LHCP x-pol, AF
Gain [dBi]
Gain [dBi]
LHCP x-pol, AF
10 10
0 0
-10 -10
-20 -20
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30

LHCP co-pol, AF
30 LHCP co-pol, measured 30
LHCP co-pol, AF LHCP x-pol, AF
20 LHCP x-pol, measured 20
Gain [dBi]
Gain [dBi]
LHCP x-pol, AF
10 10
0 0
-10 -10
-20 -20
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30

30 30 LHCP co-pol, AF
LHCP co-pol, AF
20 20 LHCP x-pol, AF
Gain [dBi]
Gain [dBi]
LHCP x-pol, AF
10 10
0 0
-10 -10
-20 -20
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30


30 LHCP co-pol, measured 30 LHCP co-pol, AF
LHCP co-pol, AF LHCP x-pol, measured
20 LHCP x-pol, measured 20 LHCP x-pol, AF
Gain [dBi]
Gain [dBi]
LHCP x-pol, AF
10 10
0 0
-10 -10
-20 -20
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30

Figure 4.40: Measured pattern cuts of the dual-CP reflectarray resulting from the synthesized LHCP
feed at some selected frequencies, with a comparison to AF predictions. The black dash-dot line shows
the design direction of the beams in the elevation plane.
(a) 18.0 GHz (b) 19.0 GHz (c) 20.5 GHz (d) 22.0 GHz
(e) 18.0 GHz (f) 19.0 GHz (g) 20.5 GHz (h) 22.0 GHz
(i) 18.0 GHz (j) 19.0 GHz (k) 20.5 GHz (l) 22.0 GHz
(m) 18.0 GHz (n) 19.0 GHz (o) 20.5 GHz (p) 22.0 GHz
Figure 4.41: The measured patterns of the dual-CP reflectarray in direction cosine space at some selected
frequencies. By post-processing the LP measurements, in (a)–(h) the feed is operating in RHCP and in
(i)–(p) it is operating in LHCP.
35 10 RHCP, measured
LHCP, measured
RHCP, AF
Co-polar gain [dBi]
0 LHCP, AF
30
XPD [dB]
-10
25
RHCP, measured
LHCP, measured
-20
RHCP, AF
20
LHCP, AF -30
15 -40
18 19 20 21 22 18 19 20 21 22
(a) (b)
Figure 4.42: Measured co-polar gain and cross-polar discrimination variation with frequency of the
dual-CP reflectarray in the principal planes of each beam with a comparison to array factor predictions.
-5
-10
-15
| [dB]
-20
|EH
t
-25
0 deg.
-30 10 deg.
20 deg.
-35 30 deg.
40 deg.
-40
17 18 19 20 21 22 23 24
Frequency [GHz]
(a) (b)
Figure 4.43: (a) Simulated transmitted cross-polarization of the polarizer for incident RHCP as the angle
of incidence is varied in the φ = 0 plane. (b) The angle of incidence across the truncated reflectarray
surface.
4.7 Comments and summary

In this chapter, a reflectarray unit cell was developed for applications requiring large bandwidths and
independent control of two linear polarizations. This unit cell was then combined with a linear-to-circular
polarizer to facilitate dual circular polarization control. Two prototype reflectarrays were designed, one
operating in linear polarization and the other in circular polarization, and techniques to extend the
performance of both were explored. The prototypes were fabricated and characterized experimentally.
The measured results show that the fabricated dual-LP reflectarray is able to generate separate beams
in two orthogonal polarizations, while demonstrating good radiation characteristics over the measured
frequencies, and strong cross-polar performance. However, it suffered from some aberrations in the
radiation patterns. These aberrations likely stem from designing the reflectarray to collimate an ideal
point source and not accounting for the actual phase pattern of the feed which was used, as well as
positioning errors to which the experimental setup was highly sensitive.
The dual-CP reflectarray was also able to generate a separate beam in each sense of circular po-
larization at 20.5 GHz and below. However, it had much poorer cross-polarization characteristics than
anticipated, although this improved at the lower frequencies that were measured. The discrepancies
between the measured and computational results are likely due to not taking into account non-ideal
effects in the design phase, such as the characteristics of a real feed horn, and angle of incidence effects.
Nonetheless, from the performance of the dual-LP reflectarray, the method of cascading a linear-to-
circular polarizer and dual-LP reflectarray to implement dual-CP control seems promising. In future
work improved performance should be achievable if more effort is put into characterization of the dual-CP
unit cell and the feed.
Chapter 5
Conclusions
In this thesis design methods were investigated for multiple-beam reflectarrays where the beams are gen-
erated via feed displacement. The primary goal of this work was to find ways to potentially reduce the
number of reflectors required for multiple-beam coverages used in high-throughput satellite communica-
tions applications which are implemented using a SFB architecture. In this closing chapter the findings
and implications of this work will be summarized, and avenues through which it could be continued will
be suggested.
5.1 Summary
In Chapter 3 a synthesis procedure for designing bifocal shaped dual-reflector antennas was applied
to reflectarrays. The bifocal synthesis procedure allows for two focal points, and the angles at which
beams are produced from them, to be specified. This gives more degrees of freedom than a paraboloidal
reflector, which has a single focal point, and for which the scanning characteristics are determined by its
F/D ratio. In our application of the bifocal synthesis procedure to reflectarrays, the two focal points were
set to be feeds in a SFB feed cluster, and the scan angles were designed to be smaller than what would
be produced from a single focus reflector/reflectarray system. In this manner the angular separation of
the beams could be compressed without modifying the feeds or their arrangement, potentially reducing
the number of reflectors needed to generate a required multiple-beam pattern. It was found that this
could be achieved, albeit at the cost of increased spillover efficiency. This would make this technique
of limited use for the application under consideration, as part of the reason that multiple reflectors are
used in SFB implementations is to reduce spillover losses.
It was found that the bifocal synthesis procedure does not result in degraded spillover efficiency if the
beam spacing is not compressed from that of a reference geometry which has good spillover efficiency.
Furthermore, if the desired beam spacing used in the synthesis is kept the same as the reference case
the solution produces beams which show less scan aberrations as the feeds are displaced, which was the
original motivation for multi-focal reflectors. As the bifocal synthesis procedure neglects spillover, or
the amplitude distributions on the reflectarrays, a GA-based optimizer was also explored to compress
the beam spacing. The cost function which was used for optimization is based on the gain pattern of
the reflectarray, and so spillover efficiency is taken into account. Preliminary results showed that the
compressed beam spacing could be produced with better spillover efficiency than the solutions from
157
Chapter 5. Conclusions 158
the bifocal synthesis procedure, indicating that there is room for improvement by using methods which
account for spillover.
An alternative approach to reducing the number of reflectors was explored in Chapter 4. Rather
than reducing the beam spacing resulting from a cluster of feeds, design methods were investigated
to generate independent orthogonally linearly- or circularly-polarized beams. In a four-reflector SFB
system each reflector generates one frequency/polarization combination, and if one reflector can generate
the required beams in both polarizations then the number of reflectors required could potentially be
halved. A dual-linear polarization reflectarray element consisting of two layers of Jerusalem crosses
with a quarter-wavelength unit cell period was developed. The quarter-wavelength cell period differs
from the half-wavelength period which reflectarray unit cells typically use, and was considered mainly
due to bandwidth improvements which have been noted in the literature. Simulations in a periodic
environment showed that the designed element has a phase range well beyond 360◦ and a strongly
independent reflection phase response to incident vertically- and horizontally-polarized waves, both of
which are advantages over published sub-half-wavelength reflectarray unit cells. This unit cell was then
combined with a linear-to-circular polarizer to extend the dual-linear polarization control concept to
dual-circular polarization operation.
Two prototypes were designed to demonstrate the proposed concepts, one for dual-linear polarization
operation, and another for dual-circular polarization. Considerations were made when synthesizing
the reflectarray phase and element distributions to improve performance. The two prototypes were
fabricated and measured. For the dual-CP reflectarray, the measured results showed the generation
of independent beams in RHCP and LHCP at the design frequency of 20.5 GHz and below. However,
the cross-polarization level was much higher than anticipated, which was attributed to using an over-
simplified analysis method, and not accounting for non-idealities such as the real amplitude and phase
pattern of the feed which was used. The measured patterns of the dual-LP reflectarray showed stronger
correlation with the computational results, although there were some aberrations in the patterns and an
increased cross-polarization level. Both of these factors are also likely due to the non-ideal characteristics
of the feed, as well as positioning errors to which the measurement setup was highly sensitive.
In conclusion, while the results of both projects leave a lot of room for improvement, they can serve
as a good starting point for future work. In the next section some ways in which the results of this work
could be improved will be described, and suggestions for future research will be proposed.
5.2 Future Work

While the work in this thesis does demonstrate some concepts that could be used to reduce the number
of reflectors in SFB systems, some clear drawbacks with the approaches used were identified. In future
work if these issues are addressed, solutions could be developed which perform better are more well suited
for practical applications. Additionally, there are several ways in which the work can be extended.
5.2.1 Bifocal Reflectarray

The main drawback of the bifocal synthesis is that while it is able to reduce the angular separation of
the beams, it results in solutions with poor spillover efficiency. Preliminary results using the GA-based
optimizer showed some promise, as spillover efficiency can be directly taken into account. One direction
that could be pursued in the future would be to further develop the ray-tracing synthesis procedure such
that it is able to account for the illumination level at the edges of the reflectors. The addition of a third
element, such as a lens before the sub-reflectarray, could potentially give more degrees of freedom for the
synthesis routine. It would also be worthwhile to pursue optimization-based methods further, perhaps
using a framework which converges more quickly than a GA.
While spillover efficiency became worse as the desired beam spacing of the bifocal reflectarray was
reduced, the beam spacing can be compressed slightly before the onset of efficiency degradation. Thus,
a slight decrease in the beam spacing can be obtained without incurring too much spillover loss. This
could perhaps be used to obtain a minor decrease in the number of reflectors required for a multiple-
beam coverage pattern implemented using an SFB architecture, but some system level analysis may be
required. The concept of synthesizing reflectarrays to produce beam compression or expansion could
also potentially be useful in non-satellite applications where multiple-beam or steerable antennas are
required, for example RADAR.
Finally, for the bifocal reflectarray problem, a full-sized reflectarray design with realistic requirements
was considered. In future work it may be better to start from a scaled version of the problem to
first demonstrate the concepts. A smaller version of the problem would lend itself better to full-wave
simulations and fabrication.
5.2.2 Dual-Polarization Reflectarray

The main problem with the developed dual-polarization reflectarrays is that there were aberrations
in the measured results that were not visible in the computational results, as well as an increased
cross-polarization level, particularly for the dual-CP reflectarray. This likely stems from modelling the
reflectarrays in a simplified way, in which the feed is treated as an ideal spherical wave source and the
behaviour of the reflectarray elements in the finite array is not accounted for. In future work more effort
should be put into characterization of both the surfaces and the feed. Full-wave simulations of the finite
reflectarray when fed by a real feed should have been done prior to proceeding with fabrication. The
performance of the dual-CP reflectarray unit cells, which consist of both the dual-LP reflectarray elements
and the polarizer, should be characterized experimentally and compared to the periodic simulations.
This could be done by placing samples inside array simulators based on rectangular waveguide, or more
preferably by using a quasi-optical measurement system. The latter would allow for the elements to be
characterized at different angles of incidence.
There are also some changes that should be made from a design perspective. First of all, the
polarizer should be designed to have higher return loss. It was noted from simulations that this can
greatly improve the cross-polar performance of the dual-CP reflectarray unit cell. Polarizers which are
designed for oblique incidence should also be investigated, and angle of incidence effects should be taken
into account when assigning element dimensions rather than only using the normal incidence response.
This would reduce phase errors resulting from variations in the angle of incidence. While designing for
the angle of incidence at each element would result in a large number of simulations which would become
very time consuming, a simpler method would be to divide the reflectarray into zones of different ranges
of incidence angles and to use one set of simulation results for each zone.
During the design phase attention should be paid to the combined polarizer–reflectarray performance.
In this thesis the dual-LP reflectarray and polarizer were designed completely independently, and so there
was no opportunity to tune either surface to improve the overall performance beyond adjusting the air
gap size. Finally, the pattern of the real feed should be taken into account rather than assuming that the
reflectarray is fed by an ideal point source, as measurements of the feed showed that its characteristics
deviate from those which the cosq θ model assumes. This could be done either experimentally by mea-
suring the incident field on the reflectarray directly using the near-field scanner. Another method would
be to apply a spherical wave expansion to the measured feed pattern and to then use this to compute
the incident field.
In addition to design changes that should be made, there are some ways that this work could be
furthered. In this thesis only one frequency band was considered. However, in a four-colour coverage
scheme overlapping beams are to be produced in both the transmit and receive bands. Four-reflector
implementations of SFB actually use the same reflectors for transmit and receive operation, otherwise
additional reflectors would be required. An interesting extension would be to design the polarizer–
reflectarray to not only have independent responses in orthogonal polarizations, but also in two frequency
bands. For the application under consideration one frequency band would be centred near 20 GHz, and
the other around 30 GHz. A dual-CP dual-band reflectarray could then produce the full set of beams
for one frequency sub-band in transmit and receive using an appropriate feed. This work is currently in
progress [104].
Antennas which have electronically reconfigurable patterns, such as a steerable main beam, are
attractive as they can quickly adapt to changes in environmental conditions, track a moving receiver, or
correct for pointing errors. This could be accomplished by loading the Jerusalem cross elements with
electronically tunable components, for example varactor diodes, such that the reflection phase varies as
a function of the varactor bias voltage. Although the quarter-wavelength cell period is perhaps denser
than typical reconfigurable reflectarrays, recent work has demonstrated tunable surfaces with a denser
cell period [105]. Additionally, the low loss of the unit cell could lend itself to the use of cheaper tunable
components in order to reduce cost, as was the subject of a recent work [98].
While the SFB application under consideration requires mechanical scanning, this was not actually
tested for the developed reflectarray prototypes. It may be worthwhile in future work to displace the
reflectarray feed to look at the scanning performance. Additionally, it was noted that the bifocal reflec-
tarray is able to correct beam aberrations resulting from feed displacement if the beam spacing is not
changed from that of the starting single focus geometry. It would be interesting to apply this to the
dual-polarization reflectarray. The required sub-reflectarray could perhaps be kept the same for both
polarizations, reducing the number of polarizers required for the dual-CP case.
5.3 Contributions
The work on the bifocal and GA reflectarray synthesis methods, and the dual-CP reflectarray concept
were presented at the following conferences:
• C. Geaney, J. Sun, E. M. de Rioja, et al., “Synthesis of a multi-beam dual reflectarray antenna using
genetic algorithms,” in 2017 IEEE Antennas and Propagation Society International Symposium
(APSURSI), 2017
• E. M. de Rioja, J. A. Encinar, C. Geaney, et al., “Study of bifocal dual reflectarray configu-

rations for multi-beam antennas in Ka-band,” in 2017 IEEE Antennas and Propagation Society
International Symposium (APSURSI), 2017
• C. S. Geaney, M. Hosseini, and S. V. Hum, “A cascaded polarizer–reflectarray for independent

dual circular polarization control,” in Proceedings of the 38th ESA Antenna Workshop, 2017
Additionally, a journal paper on the dual polarization reflectarray work has been accepted:
• C. S. Geaney, M. Hosseini, and S. V. Hum, “Reflectarray antennas for independent dual linear and
circular polarization control (in press),” IEEE Transactions on Antennas and Propagation, 2019
Bibliography
[1] “Satellite frequency bands,” https://www.esa.int/Our Activities/Telecommunications Integrated

Applications/Satellite frequency bands, accessed: 2019-04-23.
[2] “CSO atmospheric transmission interactive plotter,” http://www.submm.caltech.edu/cso/

weather/, accessed: 2019-04-23.
[3] M. Schneider, C. Hartwanger, and H. Wolf, “Antennas for multiple spot beam satellites,” CEAS
Space Journal, vol. 2, no. 1, pp. 59–66, 2011.
[4] J. Angevain, N. Fonseca, D. Schobert, G. Toso, P. de Maagt, and C. Mangenot, “Multibeam

reflector antennas for space applications: Current trends and future perspectives in europe,” in
12th European Conference on Antennas and Propagation (EuCAP 2018), Apr. 2018, pp. 1–5.
[5] J. M. Montero, A. M. Ocampo, and N. J. G. Fonseca, “C-band multiple beam antennas for
communication satellites,” IEEE Transactions on Antennas and Propagation, vol. 63, no. 4, pp.
1263–1275, Apr. 2015.
[6] “Blank SVG map of Canada,” https://commons.wikimedia.org/wiki/File:Canada blank map.svg,

accessed: 2019-04-23.
[7] N. Llombart, A. Neto, G. Gerini, M. Bonnedal, and P. De Maagt, “Leaky wave enhanced feed
arrays for the improvement of the edge of coverage gain in multibeam reflector antennas,” IEEE
Transactions on Antennas and Propagation, vol. 56, no. 5, pp. 1280–1291, May 2008.
[8] W. Imbriale, S. Gao, and L. Boccia, Space Antenna Handbook. West Sussex, UK: Wiley, 2012.
[9] D. L. Doan, E. Amyotte, C. Mok, and J. Uher, “Anik-F2 Ka-band transmit multibeam antenna,”
in 2004 10th International Symposium on Antenna Technology and Applied Electromagnetics and
URSI Conference, July 2004, pp. 1–4.
[10] H. Wolf, M. Schneider, S. Stirland, and D. Scouarnec, “Satellite multibeam antennas at Airbus
Defence and Space: State of the art and trends,” in The 8th European Conference on Antennas
and Propagation (EuCAP 2014), Apr. 2014, pp. 182–185.
[11] É. Amyotte, Y. Demers, L. Hildebrand, S. Richard, and S. Mousseau, “A review of multibeam
antenna solutions and their applications,” in The 8th European Conference on Antennas and Prop-
agation (EuCAP 2014), Apr. 2014, pp. 191–195.
[12] P. Angeletti and M. Lisi, “Multimode beamforming networks for space applications,” IEEE An-
tennas and Propagation Magazine, vol. 56, no. 1, pp. 62–78, Feb. 2014.
162
Bibliography 163
[13] J. C. Lafond, E. Vourch, F. Delepaux, P. Lepeltier, P. Bosshard, F. Dubos, C. Feat, C. Labourdette,

G. Navarre, and J. M. Bassaler, “Thales Alenia Space multiple beam antennas for telecommuni-
cation satellites,” in The 8th European Conference on Antennas and Propagation (EuCAP 2014),
April 2014, pp. 186–190.
[14] C. Leclerc, M. Romier, H. Aubert, and A. Annabi, “Ka -band multiple feed per beam focal array
using interleaved couplers,” vol. 62, no. 6, pp. 1322–1329, June 2014.
[15] Y. Demers, E. Amyotte, K. Glatre, M. A. Godin, J. Hill, A. Liang, and M. Riel, “Ka-band user
antennas for VHTS GEO applications,” in Proc. 11th Eur. Conf. Antennas Propag. (EuCAP),
Mar. 2017, pp. 2418–2422.
[16] J. Ruze, “Wide-angle metal-plate optics,” Proceedings of the IRE, vol. 38, no. 1, pp. 53–59, Jan.
1950.
[17] F. Holt and A. Mayer, “A design procedure for dielectric microwave lenses of large aperture ratio
and large scanning angle,” IRE Transactions on Antennas and Propagation, vol. 5, no. 1, pp.
25–30, Jan. 1957.
[18] B. L. Rao, “Bifocal dual reflector antenna,” IEEE Transactions on Antennas and Propagation,
vol. 22, no. 5, pp. 711–714, Sep. 1974.
[19] C. Rappaport, “An offset bifocal reflector antenna design for wide-angle beam scanning,” vol. 32,
no. 11, pp. 1196–1204, Nov. 1984.
[20] S. K. Rao, “Advanced antenna technologies for satellite communications payloads,” IEEE Trans-
actions on Antennas and Propagation, vol. 63, no. 4, pp. 1205–1217, Apr. 2015.
[21] C. A. Balanis, Antenna Theory: Analysis and Design. Hoboken, NJ: Wiley, 2005.
[22] D. Berry, R. Malech, and W. Kennedy, “The reflectarray antenna,” IEEE Transactions on An-
tennas and Propagation, vol. 11, no. 6, pp. 645–651, Nov. 1963.
[23] M. Abd-Elhady, W. Hong, and Y. Zhang, “A Ka-band reflectarray implemented with a single-
layer perforated dielectric substrate,” IEEE Antennas and Wireless Propagation Letters, vol. 11,
pp. 600–603, May 2012.
[24] R. Deng, F. Yang, S. Xu, and M. Li, “A 100-GHz metal-only reflectarray for high-gain antenna
applications,” IEEE Antennas and Wireless Propagation Letters, vol. 15, pp. 178–181, May 2016.
[25] H. Legay, D. Bresciani, E. Labiole, R. Chiniard, E. Girard, G. Caille, D. Calas, R. Gillard, and
G. Toso, “A 1.3 m facetted reflectarray in Ku band,” in 2012 15 International Symposium on
Antenna Technology and Applied Electromagnetics, June 2012, pp. 1–4.
[26] M. Zhou, S. B. Sørensen, N. Vesterdal, M. F. Palvig, Y. Brand, S. Maltais, J. Bellemore, and

G. Toso, “Design of dual-band dual-polarized reflectarray for future multiple spot beam appli-
cations in Ka-band,” in The 13th European Conference on Antennas and Propagation (EuCAP
2019), Apr. 2019.
Bibliography 164
[27] R. Leberer and W. Menzel, “A dual planar reflectarray with synthesized phase and amplitude
distribution,” IEEE Transactions on Antennas and Propagation, vol. 53, no. 11, pp. 3534–3539,
Nov. 2005.
[28] N. Yu, P. Genevet, M. A. Kats, F. Aieta, J.-P. Tetienne, F. Capasso, and Z. Gaburro, “Light
propagation with phase discontinuities: Generalized laws of reflection and refraction,” Science,
2011.
[29] P. Hannan, “Microwave antennas derived from the Cassegrain telescope,” IRE Transactions on
Antennas and Propagation, vol. 9, no. 2, pp. 140–153, Mar. 1961.
[30] E. Almajali, D. McNamara, J. Shaker, and M. R. Chaharmir, “Derivation and validation of the
basic design equations for symmetric sub-reflectarrays,” IEEE Transactions on Antennas and Prop-
agation, vol. 60, no. 5, pp. 2336–2346, May 2012.
[31] M. Zhou, S. B. Sorensen, E. Jorgensen, P. Meincke, O. S. Kim, and O. Breinbjerg, “An accurate
technique for calculation of radiation from printed reflectarrays,” IEEE Antennas and Wireless
Propagation Letters, vol. 10, pp. 1081–1084, 2011.
[32] B. A. Munk, Frequency Selective Surfaces: Theory and Design. Hoboken, NJ: Wiley, 2000.
[33] A. Bhattacharyya, Phased Array Antennas: Floquet Analysis, Synthesis, BFNs, and Active Array
Systems. Hoboken, NJ: Wiley, 2006.
[34] T. Smith, “Hybrid maritime satellite communication antenna,” Ph.D. dissertation, 2013.
[35] C. Tienda, M. Arrebola, J. A. Encinar, and G. Toso, “Analysis of a dual-reflect array antenna,”
IET Microwaves, Antennas Propagation, vol. 5, no. 13, pp. 1636–1645, Oct. 2011.
[36] W. Stutzman and G. Thiele, Antenna Theory and Design. Hoboken, NJ: Wiley, 2013.
[37] C. A. Balanis, Advanced Engineering Electromagnetics. Hoboken, NJ: Wiley, 2012.
[38] R. F. Harrington, Time-Harmonic Electromagnetic Fields. New York, NY: Wiley, 2001.
[39] W. C. Gibson, The Method of Moments in Electromagnetics. Boca Raton, FL: CRC Press, 2015.
[40] A. Ludwig, “The definition of cross polarization,” IEEE Transactions on Antennas and Propaga-
tion, vol. 21, no. 1, pp. 116–119, Jan. 1973.
[41] J. Huang and J. A. Encinar, Reflectarray Antennas. Hoboken, NJ: Wiley, 2008.
[42] P. J. B. Clarricoats, “IEE electronics division: Chairman’s address. some recent advances in mi-
crowave reflector antennas,” Electrical Engineers, Proceedings of the Institution of, vol. 126, no. 1,
pp. 9–, Jan. 1979.
[43] A. N. Plastikov, “A high-gain multibeam bifocal reflector antenna with 40◦ field of view for satellite
ground station applications,” IEEE Transactions on Antennas and Propagation, vol. 64, no. 7, pp.
3251–3254, July 2016.
[44] W. Menzel, M. Al-Tikriti, and R. Leberer, “A 76 GHz multiple-beam planar reflector antenna,”
in 2002 32nd European Microwave Conference, Sep. 2002.
Bibliography 165
[45] W. Hu, M. Arrebola, R. Cahill, J. A. Encinar, V. Fusco, H. S. Gamble, Y. Alvarez, and F. Las-
Heras, “94 GHz dual-reflector antenna with reflectarray subreflector,” IEEE Transactions on An-
tennas and Propagation, vol. 57, no. 10, pp. 3043–3050, Oct. 2009.
[46] C. Kocia and S. V. Hum, “Analysis of reflectors fed by reconfigurable array lenses,” in The 8th
European Conference on Antennas and Propagation (EuCAP 2014), Apr. 2014, pp. 18–21.
[47] C. Tienda, J. A. Encinar, M. Arrebola, M. Barba, and E. Carrasco, “Design, manufacturing and
test of a dual-reflectarray antenna with improved bandwidth and reduced cross-polarization,” IEEE
Transactions on Antennas and Propagation, vol. 61, no. 3, pp. 1180–1190, Mar. 2013.
[48] J. Rodriguez-Alvarez, M. Arrebola, C. Tienda, J. A. Encinar, and F. Las-Heras, “Bifocal an-

tenna based on dual-reflectarray dual-offset configuration,” in 2012 6th European Conference on
Antennas and Propagation (EUCAP), Mar. 2012, pp. 2348–2352.
[49] E. M. de Rioja, J. A. Encinar, A. Pino, B. Gonzalez-Valdes, C. Tienda, S. V. Hum, and G. Toso,

“Application of bifocal concept to dual reflectarray configurations for multi-beam satellite antennas
in Ka-band,” in 2017 11th European Conference on Antennas and Propagation (EUCAP), Mar.
2017, pp. 2427–2430.
[50] E. M. de Rioja, J. Encinar, A. G. Pino, B. Gonzalez-Valdes, S. V. Hum, and C. T. Herrero,

“Bifocal design procedure for dual reflectarray antennas in offset configurations,” IEEE Antennas
and Wireless Propagation Letters, pp. 1–1, 2018.
[51] M. R. Chaharmir, J. Shaker, N. Gagnon, and D. Lee, “Design of broadband, single layer dual-
band large reflectarray using multi open loop elements,” IEEE Transactions on Antennas and
Propagation, vol. 58, no. 9, pp. 2875–2883, Sep. 2010.
[52] M. Zhou, S. B. Sørensen, P. Meincke, and E. Jørgensen, “Design and optimization of multi-
faceted reflectarrays for satellite applications,” in The 8th European Conference on Antennas and
Propagation (EuCAP 2014), Apr. 2014, pp. 1423–1427.
[53] M. Zhou and S. B. Sørensen, “Multi-spot beam reflectarrays for satellite telecommunication appli-
cations in Ka-band,” in 2016 10th European Conference on Antennas and Propagation (EuCAP),
Apr. 2016.
[54] M. Zhou, S. B. Sørensen, J. R. de Lasson, N. Vesterdal, R. Jørgensen, and E. Jørgensen, “Advances

on high-performance curved reflectarrays for telecommunication applications,” in 2017 IEEE In-
ternational Symposium on Antennas and Propagation (APSURSI), July 2017.
[55] E. M. de Rioja, J. A. Encinar, C. Geaney, and S. V. Hum, “Study of bifocal dual reflectarray
configurations for multi-beam antennas in Ka-band,” in 2017 IEEE Antennas and Propag. Soc.
Int. Symp. (APSURSI), July 2017.
[56] E. Martinez-de-Rioja, J. A. Encinar, R. Florencio, and C. Tienda, “3-D bifocal design method
for dual-reflectarray configurations with application to multibeam satellite antennas in Ka-band,”
IEEE Transactions on Antennas and Propagation, vol. 67, no. 1, pp. 450–460, Jan. 2019.
Bibliography 166
[57] E. Carrasco, M. Barba, J. A. Encinar, M. Arrebola, F. Rossi, and A. Freni, “Design, manufacture
and test of a low-cost shaped-beam reflectarray using a single layer of varying-sized printed dipoles,”
IEEE Transactions on Antennas and Propagation, vol. 61, no. 6, pp. 3077–3085, June 2013.
[58] F. Rafael, B. R. R., C. Eduardo, E. J. A., B. Mariano, and P.-P. Gerardo, “Broadband reflectarrays
made of cells with three coplanar parallel dipoles,” Microwave and Optical Technology Letters,
vol. 56, no. 3, pp. 748–753, Mar. 2014.
[59] J. H. Yoon, Y. J. Yoon, W. S. Lee, and J. H. So, “Broadband microstrip reflectarray with five
parallel dipole elements,” IEEE Antennas and Wireless Propagation Letters, vol. 14, pp. 1109–
1112, Jan. 2015.
[60] R. Florencio, J. A. Encinar, R. R. Boix, V. Losada, and G. Toso, “Reflectarray antennas for dual
polarization and broadband telecom satellite applications,” IEEE Transactions on Antennas and
Propagation, vol. 63, no. 4, pp. 1234–1246, Apr. 2015.
[61] E. M. de Rioja, J. A. Encinar, M. Barba, R. Florencio, R. R. Boix, and V. Losada, “Dual polarized
reflectarray transmit antenna for operation in Ku- and Ka-bands with independent feeds,” IEEE
Transactions on Antennas and Propagation, vol. 65, no. 6, pp. 3241–3246, June 2017.
[62] C. Geuzaine and J.-F. Remacle, “Gmsh, a three-dimensional finite element mesh generator with
built-in pre- and post-processing facilities,” http://gmsh.info/.
[63] J. Sun, “Synthesis of a multibeam dual reflectarray beam pattern using genetic algorithms,” Un-
dergraduate Thesis, University of Toronto, Toronto, ON, 2017.
[64] J. Huang, “Bandwidth study of microstrip reflectarray and a novel phased reflectarray concept,”
in IEEE Antennas and Propagation Society International Symposium. 1995 Digest, vol. 1, Jun.
1995, pp. 582–585 vol.1.
[65] D. M. Pozar, “Bandwidth of reflectarrays,” Electronics Letters, vol. 39, no. 21, pp. 1490–1491,
Oct. 2003.
[66] J. A. Encinar, “Design of two-layer printed reflectarrays using patches of variable size,” IEEE
Transactions on Antennas and Propagation, vol. 49, no. 10, pp. 1403–1410, Oct. 2001.
[67] J. A. Encinar and J. A. Zornoza, “Broadband design of three-layer printed reflectarrays,” IEEE
Transactions on Antennas and Propagation, vol. 51, no. 7, pp. 1662–1664, July 2003.
[68] J. A. Encinar, L. S. Datashvili, J. A. Zornoza, M. Arrebola, M. Sierra-Castaner, J. L. Besada-

Sanmartin, H. Baier, and H. Legay, “Dual-polarization dual-coverage reflectarray for space appli-
cations,” IEEE Transactions on Antennas and Propagation, vol. 54, no. 10, pp. 2827–2837, Oct.
2006.
[69] M. R. Chaharmir, J. Shaker, and H. Legay, “Broadband design of a single layer large reflectarray
using multi cross loop elements,” IEEE Transactions on Antennas and Propagation, vol. 57, no. 10,
pp. 3363–3366, Oct. 2009.
[70] G. B. Wu, S. W. Qu, S. Yang, and C. H. Chan, “Broadband, single-layer dual circularly polar-
ized reflectarrays with linearly polarized feed,” IEEE Transactions on Antennas and Propagation,
vol. 64, no. 10, pp. 4235–4241, Oct. 2016.
Bibliography 167
[71] E. Carrasco, J. A. Encinar, and M. Barba, “Bandwidth improvement in large reflectarrays by using
true-time delay,” IEEE Transactions on Antennas and Propagation, vol. 56, no. 8, pp. 2496–2503,
Aug. 2008.
[72] S. M. A. M. H. Abadi, K. Ghaemi, and N. Behdad, “Ultra-wideband, true-time-delay reflectarray

antennas using ground-plane-backed, miniaturized-element frequency selective surfaces,” IEEE
Transactions on Antennas and Propagation, vol. 63, no. 2, pp. 534–542, Feb 2015.
[73] L. Liang and S. V. Hum, “Realizing a flat UWB 2-D reflector designed using transformation
optics,” IEEE Transactions on Antennas and Propagation, vol. 62, no. 5, pp. 2481–2487, May
2014.
[74] ——, “Design of a UWB reflectarray as an impedance surface using Bessel filters,” IEEE Trans-
actions on Antennas and Propagation, vol. 64, no. 10, pp. 4242–4255, Oct. 2016.
[75] D. M. de Rioja, E. M. de Rioja, J. A. Encinar, A. Pino, Y. Rodriguez-Vaquiero, B. Gonzalez-

Valdes, O. Rubiños, and M. Arias, “Single and dual reflectarray configurations for multibeam
satellite antennas in Ka-band,” in 2019 13th European Conference on Antennas and Propagation
(EUCAP), Apr. 2019.
[76] P. Nayeri, F. Yang, and A. Z. Elsherbeni, “Broadband reflectarray antennas using double-layer
subwavelength patch elements,” IEEE Antennas and Wireless Propagation Letters, vol. 9, pp.
1139–1142, Nov. 2010.
[77] D. M. Pozar, “Wideband reflectarrays using artificial impedance surfaces,” Electronics Letters,
vol. 43, no. 3, pp. 148–149, Feb. 2007.
[78] M. A. Joyal, R. E. Hani, M. Riel, Y. Demers, and J. J. Laurin, “A reflectarray-based dual-surface re-
flector working in circular polarization,” IEEE Transactions on Antennas and Propagation, vol. 63,
no. 4, pp. 1306–1313, Apr. 2015.
[79] D.-C. Chang and M.-C. Huang, “Multiple-polarization microstrip reflectarray antenna with high
efficiency and low cross-polarization,” IEEE Transactions on Antennas and Propagation, vol. 43,
no. 8, pp. 829–834, Aug. 1995.
[80] M. E. Bialkowski and H. J. Song, “Dual linearly polarized reflectarray using aperture coupled
microstrip patches,” in IEEE Antennas and Propagation Society International Symposium. 2001
Digest, vol. 3, July 2001, pp. 486–489 vol.3.
[81] D. M. Pozar and S. D. Targonski, “A microstrip reflectarray using crossed dipoles,” in IEEE
Antennas and Propagation Society International Symposium. 1998 Digest, vol. 2, June 1998, pp.
1008–1011 vol.2.
[82] J. Huang and R. J. Pogorzelski, “A Ka-band microstrip reflectarray with elements having variable
rotation angles,” IEEE Transactions on Antennas and Propagation, vol. 46, no. 5, pp. 650–656,
May 1998.
[83] T. Smith, U. Gothelf, O. S. Kim, and O. Breinbjerg, “An FSS-backed 20/30 GHz circularly
polarized reflectarray for a shared aperture L- and Ka-band satellite communication antenna,”
IEEE Transactions on Antennas and Propagation, vol. 62, no. 2, pp. 661–668, Feb. 2014.
Bibliography 168
[84] S. Mener, R. Gillard, R. Sauleau, A. Bellion, and P. Potier, “Dual circularly polarized reflectarray
with independent control of polarizations,” IEEE Transactions on Antennas and Propagation,
vol. 63, no. 4, pp. 1877–1881, Apr. 2015.
[85] R. Chaharmir, J. Shaker, and M. Cuhaci, “Development of dual-band circularly polarised reflec-
tarray,” IEE Proceedings - Microwaves, Antennas and Propagation, vol. 153, no. 1, pp. 49–54, Feb.
2006.
[86] Y. Y. Chen, Y. Ge, and T. S. Bird, “An offset reflectarray antenna for multipolarization applica-
tions,” IEEE Antennas and Wireless Propagation Letters, vol. 15, pp. 1353–1356, Dec. 2015.
[87] S. R. Lee, E. H. Lim, F. L. Lo, and W. H. Ng, “Circularly polarized elliptical microstrip patch
reflectarray,” IEEE Transactions on Antennas and Propagation, vol. 65, no. 8, pp. 4322–4327,
Aug. 2017.
[88] J. Ethier, M. R. Chaharmir, and J. Shaker, “Novel approach for low-loss reflectarray designs,” in
2011 IEEE International Symposium on Antennas and Propagation (APSURSI), July 2011, pp.
373–376.
[89] ——, “Loss reduction in reflectarray designs using sub-wavelength coupled-resonant elements,”
IEEE Transactions on Antennas and Propagation, vol. 60, no. 11, pp. 5456–5459, Nov. 2012.
[90] E. Almajali, D. A. McNamara, J. Shaker, and M. R. Chaharmir, “Observations on the performance

of reflectarrays with reduced inter-element spacings,” in 2011 IEEE International Symposium on
Antennas and Propagation (APSURSI), July 2011, pp. 369–372.
[91] M. Niroo-Jazi, M. R. Chaharmir, J. Shaker, and A. R. Sebak, “On radiation performances of

reflectarray antennas constructed with subwavelength unit cells,” IEEE Antennas and Wireless
Propagation Letters, vol. 14, pp. 1258–1262, Feb. 2015.
[92] R. E. Hodges, M. J. Radway, A. Toorian, D. J. Hoppe, B. Shah, and A. E. Kalman, “ISARA -

integrated solar array and reflectarray CubeSat deployable Ka-band antenna,” in 2015 IEEE Inter-
national Symposium on Antennas and Propagation USNC/URSI National Radio Science Meeting,
July 2015, pp. 2141–2142.
[93] R. E. Hodges, N. Chahat, D. J. Hoppe, and J. D. Vacchione, “A deployable high-gain antenna

bound for Mars: Developing a new folded-panel reflectarray for the first CubeSat mission to Mars.”
IEEE Antennas and Propagation Magazine, vol. 59, no. 2, pp. 39–49, Apr. 2017.
[94] G. Perez-Palomino, M. Barba, J. A. Encinar, R. Cahill, R. Dickie, P. Baine, and M. Bain, “Design
and demonstration of an electronically scanned reflectarray antenna at 100 GHz using multires-
onant cells based on liquid crystals,” IEEE Transactions on Antennas and Propagation, vol. 63,
no. 8, pp. 3722–3727, Aug. 2015.
[95] C. Kocia and S. V. Hum, “Design of an optically transparent reflectarray for solar applications
using indium tin oxide,” IEEE Transactions on Antennas and Propagation, vol. 64, no. 7, pp.
2884–2893, July 2016.
[96] E. Carrasco and J. Perruisseau-Carrier, “Reflectarray antenna at terahertz using graphene,” IEEE
Antennas and Wireless Propagation Letters, vol. 12, pp. 253–256, 2013.
Bibliography 169
[97] S. V. Hum and J. Perruisseau-Carrier, “Reconfigurable reflectarrays and array lenses for dynamic
antenna beam control: A review,” IEEE Transactions on Antennas and Propagation, vol. 62, no. 1,
pp. 183–198, Jan. 2014.
[98] B. Du, “Analysis and design of simple, low loss and low cost reconfigurable reflectarrays,” Masters
Thesis, University of Toronto, Toronto, ON, 2017.
[99] M. Hosseini and S. V. Hum, “A circuit-driven design methodology for a circular polarizer based on
modified Jerusalem cross grids,” IEEE Transactions on Antennas and Propagation, vol. 65, no. 10,
pp. 5322–5331, Oct. 2017.
[100] M. Hosseini and S. V. Hum, “A circular polarization selective surface employing Jerusalem
cross-based polarizers,” in 2017 IEEE International Symposium on Antennas and Propagation
USNC/URSI National Radio Science Meeting, July 2017, pp. 1501–1502.
[101] M. Kim and G. V. Eleftheriades, “Design and demonstration of impedance-matched dual-band

chiral metasurfaces,” Scientific Reports, vol. 8, no. 1, pp. 2045–2322, Feb. 2018.
[102] M. Hosseini and S. V. Hum, “A dual-CP reflectarray unit cell for realizing independently controlled
beams for space applications,” in 2017 11th European Conference on Antennas and Propagation
(EuCAP), Mar. 2017, pp. 66–70.
[103] P. Naseri, S. A. Matos, J. R. Costa, and C. A. Fernandes, “Phase-delay versus phase-rotation cells
for circular polarization transmit arrays—application to satellite Ka-band beam steering,” vol. 66,
no. 3, pp. 1236–1247, Mar. 2018.
[104] P. Naseri and S. V. Hum, “A dual-band dual-circularly polarized reflectarray for K/Ka-band space
applications,” in The 13th European Conference on Antennas and Propagation (EuCAP 2019),
Apr. 2019.
[105] M. Y. Xu and S. V. Hum, “Realization of low-complexity reconfigurable Huygens’ metasurfaces,”

in The 13th European Conference on Antennas and Propagation (EuCAP 2019), Apr. 2019.

Geaney Ciaran S 201911 MAS Thesis

Uploaded by

Copyright:

Available Formats

Geaney Ciaran S 201911 MAS Thesis

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Geaney Ciaran S 201911 MAS Thesis

Uploaded by

Copyright:

Available Formats

Reflectarray Antennas for Multiple-Beam Satellite Applications

Ciaran Sean Geaney

A thesis submitted in conformity with the requirements

c Copyright 2019 by Ciaran Sean Geaney

Ciaran Sean Geaney

is especially true in satellite communications applications.

1.1 Usage of radio and microwave communication bands. . . . . . . . . . . . . . . . . . . . . . 2

2.1 A planar array of antenna elements arranged in a rectangular grid. . . . . . . . . . . . . 17

3.1 Examples of early bifocal spatially-fed antennas. . . . . . . . . . . . . . . . . . . . . . . . 56

Figure 1.1: Usage of radio and microwave communication bands [1].

1.1 Antennas for Satellite Communication Applications

1.1.1 Phased Array Antennas

Power splitting network

Figure 1.3: Schematic drawing of a transmit phased array.

1.1.2 Reflector Antennas

1.1.3 Reflectarray Antennas

1.2 Multiple Spot Beam Coverages

(c) Eurostar 3000 Ka-band multi-beam mission reflectors [10]

Figure 1.9: Examples of SFB antenna systems.

1.2.3 Direct Radiating Arrays

Figure 1.11: Examples of prototype MFB feed systems.

1.3 Motivation and Thesis Goals

1.3.1 Bifocal Reflectarray

1.3.2 Dual-polarization Reflectarrays

1.3.3 Thesis Goals

1. Apply a ray-tracing procedure for synthesizing dual-reflector antennas [19] to dual-reflectarrays,

For the dual-polarization reflectarrays we aim to:

1.4 Thesis Outline

2.1 Antenna Array Background

Figure 2.1: A planar array of antenna elements arranged in a rectangular grid.

E (θ, φ) = EF (θ, φ) · AF (θ, φ) , (2.8)

ψi = −k (x0i sin θ0 cos φ0 + yi0 sin θ0 sin φ0 + zi0 cos θ0 ) . (2.11)

kx = k sin θ cos φ (2.13a)

AF (kx , ky ) = M N · IDFT2 (Imn ) , (2.16)

2.2 Reflectarray Background and Design Principles

2.2.1 Phase Shift Approach

−k (|r0 − F|) + ψi − k |r0 | = C + 2πN, (2.18)

a = (r00 − r0 ) · r̂0 = r00 · r̂0 − r0 · r̂0 , (2.20)

−k (|r0 − F|) + ψi − k (r00 · r̂0 − r0 · r̂0 + b) = C + 2πN. (2.21)

ψi = k (|r0 − F| − r0 · r̂0 ) + ψ0 + 2πN, (2.22)

different path lengths from the feed to the reflectarray.

2.2.2 Phase Derivative Approach

which when re-arranged yields

2.2.3 Dual Reflectarrays

ψs,i = k (|r0s − FR | − |r0s − FV |) . (2.28)

Figure 2.4: A dual-reflectarray antenna transforming a spherical wave to a plane wave.

2.3 Analysis of Reflectarray Elements and Periodic Structures

2.3.1 Element Modelling and Phase Shift Mechanism

Figure 2.5: Analysis of a unit cell using periodic boundary conditions.

2.3.2 Multi-layer Structures

Figure 2.8: Analysis of a multi-layer periodic structure.

2.3.3 Rotation of One-port Scattering Parameters

âTE = âTE0 cos γ − âTM0 sin γ

TE,TE TE0 ,TE0 TM0 ,TM0 1 

TE,TE 1  TE0 ,TE0 TM ,TM 0 0

2.3.4 Circularly Polarized S-parameters

EiR = Ei · âR,i (2.42a)

the reflected fields, then gives:

where the one-port CP reflection matrix takes the form:

TE,TE TE0 ,TE0 TM0 ,TM0 1

TE,TE 1 TE0 ,TE0 TM ,TM 0 0