2019 Book CompactSlotArrayAntennasForWir
2019 Book CompactSlotArrayAntennasForWir
Alan J. Sangster
Compact Slot
Array Antennas
for Wireless
Communications
Signals and Communication Technology
The series “Signals and Communications Technology” is devoted to fundamentals
and applications of modern methods of signal processing and cutting-edge
communication technologies. The main topics are information and signal theory,
acoustical signal processing, image processing and multimedia systems, mobile and
wireless communications, and computer and communication networks. Volumes in
the series address researchers in academia and industrial R&D departments. The
series is application-oriented. The level of presentation of each individual volume,
however, depends on the subject and can range from practical to scientific.
123
Alan J. Sangster
Edinburgh, UK
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
To Hamish
Whatever you choose to be, be the very best
you can. That way lies fulfilment.
Preface
At about the time when this book was being compiled, it was hardly controversial to
suggest that the predominant evolutionary course of electronic engineering had been
characterised by an inexorable drive toward the miniaturisation of components,
devices, and systems, facilitated by the unprecedented advances in integrated circuit
(IC) technology. The obvious example is the computer, the hardware for which
could fill a modest laboratory in the 1960’s, yet by 2015 a device of comparable
computational power was small enough to be held in the palm of the hand—Moore’s
law gives this trend a quantitive basis. However, for practitioners in the discipline of
radio frequency communications the ‘fly in the oitment’, to this generally desirable
miniaturisation drift, has been the interface with ‘free-space’, namely the antenna.
This has been particularly true in the burgeoning area of handheld mobile devices.
Unlike virtually all other components in an electronic circuit the antenna is
dimensionally constrained by the unyielding laws of physics. For reasons of radi-
ation efficiency, at least one major dimension (L) of the antenna has to be, in size,
close to, or greater than, a half of the free space wavelength (ko) at the frequency of
operation (L k0 =2Þ. Needless to say, in the realms of portable and handheld
mobile communications devices, or in the domain of implantable devices for
monitoring medical conditions, circumventing this restraint has been a major focus
of recent compact antenna research. The problem can be simply stated. To create an
efficient compact antenna, at a prescribed frequency, any electrical dimension
which governs radiation strength must be preserved yet remain within the bounds of
a physical package that is required to be everywhere smaller. These contradictory
requirements are usually satisfied, with varying degrees of success, by securing
electrical length in a dimensionally restricted package through the adoption of
‘tricks’ such as meandering, corkscrewing and twisting the key dimensional ele-
ment of the shrunk antenna. However, not surprisingly, these distortions can have
deleterious effects on other important antenna properties and compensating tech-
niques are generally necessary to achieve viability.
vii
viii Preface
The purpose of this text lies in describing and providing design guidelines for
antennas which pursue compactness by recruiting the slot radiator as the funda-
mental building block. Initial research on the so called slot radiator can be traced
back in time to before the Second World War. It was strenuously propelled by
military pressures to advance the detection capabilities of radar. As a consequence
of this early ‘baptism’ the literature is comprehensive and extensive, requiring
patient time-consuming study by antenna designers, and by research engineers, with
a need to claim mastery of the topic. Recent progress has been largely triggered by
the burgeoning of modern communications techniques demanding increasingly
compact, lighter, more versatile antennas, offering large bandwidth to the user. For
the design engineer due diligence requires a thorough sifting of this literature,
which is inevitably time consuming and expensive when it is so extensive.
While the primary threads in the historical evolution of compact array antennas
are identified, the main focus is to provide an electrical engineering text which
summarises the extensive compact antenna related literature. More particularly, the
aim is to illuminate, in relation to the now ubiquitous slot radiator, how despite
significant antenna size reductions, seemingly defying physics, desirable radiation
pattern properties can be preserved. This leads to an examination of recent advances
in frequency selective surfaces or metamaterials which can, with caution, be used to
facilitate physics defying dimensional restrictions.
Slot radiators, excited from waveguide, have for some time provided basic
elements for creating array antennas most commonly for applications in radar and in
satellite communications. On the other hand, in stripline, microstrip or coplanar
waveguide formats slot radiators, either directly excited or through a patch transi-
tion, additionally form fundamental components in compact antennas and confor-
mal antennas, particularly in the realm of mobile communications. The range of
possible formats is briefly illustrated in Chap. 1.
In constructing the eleven chapters, focused both on already established and on
continuing developments in the realm of compact antennas for wireless commu-
nications, the intention has been to restrict the content to topics deemed salient to
antennas of the compact variety and particularly is this so in addressing theory,
modelling and design. This stance has been achieved by relegating some general,
but useful, background material to appendices where a copious bibliography is
provided directing the reader to standard texts which provide further elaboration.
The early chapters of the book (Chaps. 1–5) focus on the radiation mechanisms
and characteristics of the slot radiator when treated as an aperture antenna. The
progressive incorporation of the slot radiator into antenna technology developments
and in particular those directed toward compactness are both introduced and pre-
sented schematically in Chap. 1. This chapter also provides a guide to more detailed
explorations and expositions of compact antenna evolution which appear in later
chapters. In Chap. 2 the electromagnetic fundamentals of the radiation mechanism
are introduced using analogies from electrostatics and magnetostatics to construct
the key equations without undue mathematical toil. With these indispensable
equations established, Chap. 3 uses them to construct the now conventional
Preface ix
particularly like to thank the members of staff, at the Heriot-Watt University library,
who have been very helpful in ensuring that I was able to access a wide range of
written material, the contents of some of which have been germane to the realisation
of this project.
xiii
xiv Contents
A Amps
AC Alternating current
AMC Artificial magnetic conductor
BGAN Broadband global area network
BPF Bandpass filter
BW Bandwidth
BWFN Beamwidth for nulls
C Capacitance (Farads)
CBCPW Conductor backed co-planar waveguide
CEM Classical electromagnetism
CRA Cavity resonator antenna
CW Continuous wave
DC Direct current
DUT Device under test
EM Electromagnetism
EMW Electromagnetic wave
FEM Finite element method
FET Field effect transistor
FSS Frequency selective surface
HD High definition
HDMI High definition multimedia interface
HFSS High frequency structure simulator
HIS High impedance surface
HPBW Half power beamwidth
IF Intermediate frequency
IQ In-phase/quadrature-phase
L Inductance (Henries)
LEO Low earth orbit
LWA Leaky-wave antenna
MF Medium frequency
xix
xx Abbreviations
MKS Metre/kilogram/second
MMIC Monolithic microwave integrated circuit
MoM Method of Moments
PC Printed circuit
PCC Phase conjugating circuit
PMC Perfect magnetic conductor
PPL Phase locked loop
PRS Partially reflective surface
Q Quality factor (of resonance)
RF Radio frequency
RFID Radio frequency identification
RLP Ring loaded patch
SIW Substrate integrated waveguide
SLL Side lobe level
SP Square patch
SR Square ring
TE Transverse electric
TEM Transverse electromagnetic
TEN Transmission line equivalent network
TM Transverse magnetic
UWB Ultra-wideband
V Volt
VCO Voltage controlled oscillator
VHF Very high frequency
VLF Very low frequency
WG Waveguide
WLAN Wireless local area network
WPAN Wireless personal area network
List of Figures
xxi
xxii List of Figures
Fig. 5.4 E-field distributions for a shunt slot radiator in the broadwall
of WG16 waveguide as predicted by the moment method
(freq = 9.375 GHz, L = 0.361 k, x1 = 6.43 mm, w = 1.58
mm, T = 1.27 mm) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
Fig. 5.5 Relative power radiated by a broadwall waveguide slot as a
function of slot length L, with the number of basis functions
as parameter (Freq = 10 GHz, a = 22.86 mm, b = 10.16 mm,
x1 = 2.00 mm, w = 1.5 mm, T = 0 mm) . . . . . . . . . . . . . . . . 111
Fig. 5.6 Reflection coefficient magnitude (|s11|) as a function of slot
length as predicted by trigonometric basis (Lyon) and by
pulse basis (Vu Khac) moment methods (WG16,
Freq = 9.375 GHz, T = 1.27 mm, x1 = 1.43 mm,
w = 1.58 mm) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
Fig. 5.7 Transmission coefficient magnitude (|s12|) as a function of slot
length as predicted by trigonometric basis (Lyon) and by
pulse basis (Vu Khac) moment methods (WG16,
Freq = 9.375 GHz, T = 1.27 mm, x1 = 1.43 mm,
w = 1.58 mm) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
Fig. 5.8 a Transmission coefficient magnitude for a rectangular
waveguide containing a longitudinal slot of changing length
(Freq = 10.48 GHz, T = 1.32 mm, x1 = 2.00 mm,
w = 1.5 mm) b Transmission coefficient phase for a
rectangular waveguide containing a longitudinal slot of
changing length (Freq = 10.48 GHz, T = 1.32 mm,
x1 = 2.00 mm, w = 1.5 mm). . . . . . . . . . . . . . . . . . . . . . . . . . 114
Fig. 5.9 a Transverse T-slot radiator in dual-mode rectangular
waveguide. b Transverse element of T-slot showing
coordinate system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Fig. 5.10 Suggested electric field distributions in the transverse
component of the T-slot a L3 = 0, b 0 < L3 < L/2 . . . . . . . . . 117
Fig. 5.11 Transmission line based equivalent circuits for waveguide
slot radiators a conventional slot, b T-slot in bifurcated
waveguide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
Fig. 5.12 Resonant cavity set-up for T-slot scattering in dual mode
waveguide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Fig. 5.13 Relative power radiated (Prad/Pinc) versus frequency for a
transverse T- slot in bifurcated waveguide in even-mode
(WG16 wall thickness T = 1.32 mm; slot width w = 1.5 mm;
slot length L = 16 mm; septum slot L3 = 5.48 mm) . . . . . . . . 124
Fig. 5.14 T-slot transmission coefficient magnitude a and phase b as a
function of slot length L (Even mode excitation; WG16;
T = 1.32 mm; L3 = 5.48 mm; bifurcation thickness = 1.5
mm; frequency = 9.89 GHz) . . . . . . . . . . . . . . . . . . . . . . . . . . 124
List of Figures xxv
Fig. 8.7 Periodic cell models showing E-wall and H-wall inserts.
a H-wall, b E-wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
Fig. 8.8 Predicted and measured phase coefficients, as a function of
frequency, for a fin-loaded rectangular waveguide
a a = 72.14 mm, b = 12.24 mm, h = 21.8 mm, G = 4.76 mm,
t = 1.58 mm, b a = 64.5 mm, b = 12.24 mm, h = 26.5 mm,
G = 11.11 mm, t = 1.58 mm (curves = theory,
points = measured) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
Fig. 8.9 Fin-loaded feed waveguide for transverse slot radiator
in upper wall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
Fig. 8.10 Relative power radiated as a function of frequency for a
centred transverse slot in a waveguide feed (solid curves) and
a periodic waveguide feed (dashed curves) with slot length
l as parameter (l = 30, 40, 42, 46 mm: w = 4.5 mm)
Waveguide WG10: a = 72.14 mm, b = 34.04 mm
Slow-waveguide: a = 72.14 mm, b = 12.24 mm, h = 21.8 mm,
G = 4.7625 mm, t = 1.5875 mm. . . . . . . . . . . . . . . . . . . . . . . 201
Fig. 8.11 Beam scan as a function of frequency for a slow-wave
fin-loaded waveguide feed comprising six slots of the same
size, equi-spaced by 65.4 mm (solid curves = theory,
diamond markers = measurement) . . . . . . . . . . . . . . . . . . . . . 203
Fig. 8.12 Radiation from a leaky wave structure . . . . . . . . . . . . . . . . . . 204
Fig. 8.13 Guide wavelength and attenuation coefficient as a function of
slit width d in a rectangular waveguide with a = 22.86 mm
and b = 10.16 mm at a frequency of 9.4 GHz. (Dashed curves
= approximate solution: Solid curves = exact solution) . . . . . . 207
Fig. 8.14 Waveguide linear array antennas employing transverse
(series) slots. a Travelling-wave array b leaky wave array . . . 209
Fig. 8.15 Pictorial representation of the mutual coupling mechanism for
a shunt slots and b transverse or series slot in a linear
waveguide array antenna . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
Fig. 8.16 Attenuation coefficient as a function of frequency for the
dominant mode of a leaky ladder of slot radiators
(with L = a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
Fig. 8.17 Brillouin diagram for the dominant mode of a leaky ladder
of slot radiators: w/a = 0.0016, S/w = 5 . . . . . . . . . . . . . . . . . 212
Fig. 9.1 Plane wave illumination of passive periodic array of current
elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
Fig. 9.2 Electromagnetic wave scattering for FSS illuminated by a
plane wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
Fig. 9.3 a Focused reflector antenna, b unfocused slot/patch array,
c reflectarray focusing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
Fig. 9.4 Basic dimensions for 10 cm 10 cm reflectarray . . . . . . . . . 225
xxviii List of Figures
Fig. 10.17 Typical FSS formats for CRA’s with (a) and (b) forming
reflecting surfaces while (c) and (d) present transparent
surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
Fig. 10.18 Elemental scatterers for periodic surfaces described as
metamaterials fabricated on copper coated planar dielectric
substrates. a Centre connected, b loop types, c patch types,
d combinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
Fig. 10.19 Plane wave refraction at a smooth flat interface between
materials with differing refractive index. a Regular materials,
b regular material in contact with metamaterial . . . . . . . . . . . 277
Fig. 10.20 Surface reflection and refraction at (a) a conventional
interface and (b) an interface with a metamaterial. In both
qffiffiffiffiffiffiffi
cases n1=2 ¼ ll2 ee21 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
278
1
Fig. 10.21 Conventional optical lens compared with ‘flat’ lens. For
a glass lens in air (lr = 1, er = 1) and for convex lens (lr = 1,
er > 1) and for b the flat lens in air it has lr < 0, er < 0 . . . . 280
Fig. 11.1 Phase conjugation through mixing . . . . . . . . . . . . . . . . . . . . . 287
Fig. 11.2 Phase conjugation circuit employing phase comparison . . . . . 288
Fig. 11.3 Phase conjugation using PLL chip . . . . . . . . . . . . . . . . . . . . . 289
Fig. 11.4 Phase conjugation using I and Q processing . . . . . . . . . . . . . . 290
Fig. 11.5 Retro-directive transmit/receive array with orthogonally
polarised patch radiators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
Fig. 11.6 Schematic of dual-mode dual-polarised patch antenna
parasitically fed from a non-resonant X-slot by means of
microstrip lines above and below the slot . . . . . . . . . . . . . . . . 294
Fig. 11.7 Schematic of a dual-mode dual-polarised patch antenna
parasitically fed from a non-resonant X-slot by means of
balanced microstrip lines above and below the slot to improve
bandwidth and axial ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
Fig. 11.8 Schematic of dual-mode dual-polarised circular aperture array
antenna parasitically fed from balanced microstrip lines
above and below the four elements . . . . . . . . . . . . . . . . . . . . . 296
Fig. 11.9 Schematic of dual-mode dual-polarised patch antenna.
a Substrate 1—copper patch L1 L1, thickness t1, relative
permittivity er1, b substrate 2—copper patch L2 L2,
thickness t2, relative permittivity er2, c two dumbbell
coupling slots in top face, 50X feed lines in lower face (see
Gao [26] for detail), substrate thickness t3, relative
permittivity er3, d edge view showing air/polystyrene layer
between substrate 1 and substrate 2, thickness t4, relative
permittivity er4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
Fig. 11.10 Circularly polarised dual-mode square patch antenna with slot
coupling and microstrip hybrid feed arrangement . . . . . . . . . . 299
List of Figures xxxi
The aim of this chapter is to introduce to readers the idea that even quite complex
geometrical arrangements of slots or patches in compact array antennas can be
evaluated and understood, at least initially, by empirical or heuristic methods, based
on a good understanding of properties of the most basic electromagnetic wave
solution governed by the Maxwell equations. The form of this solution is generally
referred to as the plane wave or the transverse electromagnetic (TEM) wave. In fact
this understanding need not extend beyond the underlying physics of plane waves
in order to permit satisfactory comprehension of radiation sources and the influence
of boundaries and antenna geometries on radiation patterns. This empirical exercise
actually represents a very useful preliminary stage in the antenna design process,
particularly since today novel designs are generally procured through the imple-
mentation of simulations on full-wave software analysis tools such as HFSS,
COMSOL, or CONCERTO for example. Of course the efficient application of these
tools, is greatly enhanced if the user has acquired a thorough appreciation of the
mathematical procedures which are incorporated into such simulations. This rec-
ommendation is addressed in much more detail in Chaps. 4 and 5.
Readers of this text are assumed to have some background in electromagnetic
theory, at least to the level of familiarity with the formulation of electromagnetic
wave solutions from the Maxwell’s equations for passive environments such as
free-space, simple transmission lines and regular waveguides. For those requiring a
memory boost and who have limited access to alternative texts (see Bibliography),
Appendices A and B contain summaries of the requisite material.
The plane electromagnetic wave is, in words, a radio-wave, or even a wave at light
frequencies, in open space where reflecting and refracting boundaries are absent.
The essence of the wave’s field structure is depicted in Fig. 1.1.
In Fig. 1.1, E and H represent the electric and magnetic field intensities in the
region of interest. We can also define D = eE where D is the electric flux density
and B = lH where B is the magnetic flux density. Note that in vacuum, or air,
permittivity e = e0 and permeability l = l0, while the phase coefficient
pffiffiffiffiffiffiffiffiffi
k ¼ k0 ¼ x l0 e0 ¼ x=c. Fuller parameter definitions are provided in Appendix A.
Fig. 1.1 Representation of a simple sinusoidal plane electromagnetic wave showing the
fundamental orthogonal relationships between the vector electric (E) and magnetic (H) fields
and the direction of propagation (k)
1.2 Plane Electromagnetic Wave in Free-Space 3
The mathematical expression for the general free-space wave (see Appendix A) is:
Ex ¼ Af ðz ctÞ þ Bf ðz þ ctÞ ð1:1Þ
The first term in Eq. (1.1) represents a wave travelling in the +z direction while
the second allows for a reflected wave, if such exists. In this case a standing wave is
formed on the mismatched line. Given that velocity is the rate of change of z with
respect to time, it is evident that c represents velocity (actually phase velocity) of
the electromagnetic wave in ‘free space’. For vacuum it is equal to 3 108 m/sec.
The application of Maxwell’s equations also gives Hz = 0 and:
A B
Hy ¼ f ðz ctÞ þ f ðz þ ctÞ ð1:3Þ
g0 g0
Also
rffiffiffiffiffi
Ex l0
¼ ¼ g0 ð1:4Þ
Hy e0
η0 is termed the free space wave impedance, which for air or vacuum has the
value 120pX. The resultant solution is a plane electromagnetic wave, also termed a
transverse electromagnetic (TEM) wave, for which E and H are transverse to the
direction of propagation and orthogonal to each other. E and H are also in time
phase, as Eqs. (1.1) and (1.3) attest (see Fig. 1.1). If time phase between E and H
prevails, which it does in lossless free-space, the power flow density along the
direction k is given by:
jE jjH j
p¼ Watts/m2 ð1:5Þ
2
The plane wave fronts in Fig. 1.1b lie transverse to k and, in theory, cover the
entire, theoretically infinite, transverse plane. On any given plane, E and H are of
constant magnitude, represented by equally spaced vectors. Further, the magnitudes
of E and H vary in accordance with Eqs. (1.2) and (1.3) respectively, as we move
from one plane to the next.
While this plane wave solution to the Maxwell equations evidently provides a
useful guide to the nature of electromagnetic waves in free-space, the importance of
this wave structure, from an antenna perspective, is that it also describes electro-
magnetic wave propagation in a variety of transmission systems which are
important antenna feed systems. The obvious reason for this beneficial position lies
in the fact that guided electromagnetic waves can be procured by simply trapping
the TEM wave [see Bibliography] between a pair of flat conducting surfaces that lie
4 1 Evolution of Compact Slot Antennas
normal to the electric field vectors. Surface charge on the conductors instanta-
neously redistributes itself, thus ensuring that the boundary conditions on both
normal D and tangential H are satisfied. That is the charge adjustment ensures that
(using unit vector notation):
^D¼q
n and ^H¼J
n ð1:6Þ
where q and J are respectively the surface charge density (Coulomb/m2) and the
surface current density (A/m) on the metal surface. Following convention, the unit
vector is directed normal to the boundary surface.
The most common TEM transmission line alternatives are shown in Fig. 1.2.
Parallel plate waveguide (Fig. 1.2a) provides only limited guidance in directions
other than normal to the plates, but clearly shows how the insertion of smooth
conducting planes normal to the E-field has negligible effect on the propagation
conditions for the TEM mode. The resultant surface currents, dictated by the
boundary conditions, are represented by blue dashed lines. Full trapping is provided
by coaxial line (Fig. 1.2b) but at the expense of phase velocity reduction associated
with the unavoidable dielectric ‘separator’ between the inner and outer conductors.
The potential for power loss in this dielectric medium is a drawback particularly in
flexible coaxial cables. The axial current flow on the conductor surfaces are again
represented by blue dashed lines. Stripline (Fig. 1.2c) is essentially ‘flattened’
coaxial line and has the advantage of ease of fabrication of intricate transmission
networks using printed circuit board (pcb) techniques. In coaxial line and in stripline
the dielectric separator usually displays a relative permittivity (er) of between 2 and
3. By increasing this to between 6 and 10 in microstrip (Fig. 1.2d), it becomes
possible to dispense with the upper ground plane and create an open structure into
which microwave components can relatively easily be inserted. The primary dis-
advantage, as the schematic suggests, is high dielectric losses and radiation losses.
The location of surface currents in the TEM transmission lines of Fig. 1.2 has been
highlighted for antenna purposes. In order to create or enhance radiation from such
lines it is generally necessary to disrupt the current flow directions. This is illus-
trated perfectly in Fig. 1.3 for microstrip line. This is the version from which
radiation is most easily extracted.
The slot is located in the ground plane symmetrically and directly above (or
below) the microstrip line and orthogonal to it, although this need not be the case.
1.3 TEM Transmission Lines 5
For illustrative purposes, the axial TEM mode currents are assumed to be
approximately constant over a length of line which can be presumed to be less than
k/4 long, where k is the TEM wavelength. The currents on the ground plane are
strongest directly above the strip and consequently the slot exercises maximum
6 1 Evolution of Compact Slot Antennas
disruptive effect. Near the centre of the slot the current is totally impeded and
induces a build-up of charge on the slot rims, shown by the positive and negative
symbols. This is not unlike current flow into a capacitor, and as with a capacitor,
electric field is formed in the gap. On the other hand, near the slot extremities,
current paths can continue to flow by ‘bending’ around the slot ends. This increase
in current path length, as in a coil, introduces inductance into the equation. Hence
the slot can be modelled electrically as a two wire transmission line shunted by a
parallel combination of a capacitor and an inductor. If the slot length is much less
than half the free-space wavelength (<ko/2) the inductance (L) is negligible and the
slot is largely capacitive (C) in nature. However, as the slot length increases toward
ko/2 the inductance begins to match the capacitance and resonance occurs. In
addition at this point the slot radiates strongly with the E and H vectors in the slot,
at any given instant in time, aligned to produce an outwardly directed Poynting
vector (Appendix A—Eq. A.19). In modelling terms, a resistive component (Rrad)
may be included along with the L and the C for long slots.
An alternative scheme for inducing radiation from a microstrip line, which is,
perhaps, more in sympathy with its basic electrical characteristics, is the microstrip
patch antenna. In its simplest and original form the usually 50X characteristic
impedance of the feed line is significantly reduced by abruptly widening the line in
a step transition (see Fig. 1.4a). If the wide portion is then terminated in an
open-circuit, a mechanism for resonance by reflection is created between the
open-circuit and the abrupt transition. The transition, viewed from within the very
low impedance patch, also presents a high impedance reflective discontinuity.
Consequently, if the distance between the open-circuit termination and the step
junction is set to k/2, where k is the TEM wavelength within the microstrip, the first
standing wave resonance is trapped in the patch as shown in Fig. 1.4b, c. The patch
is usually chosen to be approximately ko/2 in width for optimum radiation from the
patch edges. The two ‘edges’ can be viewed as ‘patch slots’. In directions normal to
the patch face the slot pair radiate in phase developing maximum power density as
shown in Fig. 1.4b. Note that the feed line can be matched to the resonant patch by
means of rectangular inserts as shown in Fig. 1.4a. The technique is not difficult for
a radiating patch. In this case, in the absence of a reflection, the line current or
voltage amplitude is unvarying in the feed line as shown. Finally relative to the
1.3 TEM Transmission Lines 7
Fig. 1.4 Fundamental mode of operation for patch antenna, a top view of patch antenna, b side
view showing trapped E-field—note edge radiation can be viewed as radiation from a two element
slot array, c voltage and current variation in patch and line
propagation direct (x in this case) the trapped mode under the patch displays a one
half-cycle variation in x, zero in y and zero in z, so the mode is termed TE100. Note
that if the x-direction were chosen to be normal to the patch surface then the same
mode would be termed TM100.
Hopefully, it has become clear from the above discussion, that it is possible to
evaluate the radiation mechanism associated with TEM transmission line fed
compact antennas, by performing a heuristic examination of the disruption caused
to current flows and to E and H field directions in the TEM line. Examples of
applying this kind of cogitative process are becoming quite common in the liter-
ature assisted by visual field presentations constructed from electromagnetic
8 1 Evolution of Compact Slot Antennas
lower portions of the transverse slot are in antiphase so that the slot as a whole is
predicted to radiate very weakly. On the other hand, these antiphase slot fields in the
transverse slot can combine at the junction with the axial slot, producing a strong
parasitic excitation of this slot, if it is long enough. Radiation which displays y-directed
polarisation is the result. The septum slot, or clearance slot, in the microstrip ground
plane is introduced to minimise E-field suppression at the junction by proximity of the
ground plane edge. For odd-mode excitation of the balanced microstrip line the E-fields
in the upper and lower portions of the transverse slot are predicted to be in phase
(Fig. 1.6) with negligible parasitic coupling into the axial slot. Radiation is therefore
almost entirely from the transverse slot generating z-directed polarisation. In principle,
the above radiator is capable of generating circularly polarised radiations, in addition to
the linear possibilities already indicated. To achieve this the exciting TEM mode
components will require to be in quadrature phase.
Based on a moment method study [5] of the proposed T-slot geometry the
potential of this novel and compact slot to generate effective polarisation agility has
been established quantitatively, by modelling and measuring its radiation and
polarisation properties. Electromagnetic modelling tools based on the moment
method and focused on the scattering properties of slot radiators are comprehen-
sively developed in Chaps. 4 and 5.
Fig. 1.8 Interference patterns of plane electromagnetic waves reflecting off a flat perfectly
conducting surface
12 1 Evolution of Compact Slot Antennas
Figure 1.7 depicts (in two dimensions for simplicity) a pair of plane electromag-
netic waves (TEM waves) of equal magnitude travelling in different directions A
and B. The waves are represented by their wavefronts, with the wave peaks in each
case denoted by solid transverse lines (planes in 3-D) and wave troughs by dashed
lines. The distance between a wave peak and wave trough is, of course, half of the
free-space wavelength (ko/2). The waves are travelling at the velocity of light [c] in
the directions of the large arrows. On examination of this wave pattern it is not too
difficult to observe that along the horizontal chain-dotted line (or in three dimen-
sions—the y-z plane) peaks of wave A coincide with those of wave B, and troughs
coincide with troughs—and this is independent of the movement of the waves. This
line (or plane) represents a stationary (in the x-direction) field maximum ‘inde-
pendent of time’ while the waves continue to exist.
In contrast, along the green dashed line, peaks of wave A coincide with troughs
of wave B, and vice versa, resulting in a stationary field null at these positions.
Consequently, if a perfectly conducting sheet of infinite extent, orientated normal to
the x-axis, is located at the stationary null position (Z-Z) the field pattern remains
unchanged (see Fig. 1.8). For a sheet at the upper dashed line, the red direction
arrow then represents an incident wave and the blue arrow a reflected wave, which
according to Snell’s laws at a perfect mirror, reflects with a magnitude equal to the
incident wave and at an angle such that hr = hi, as is required to retain the pattern.
For perpendicularly polarised plane waves with the E-field confined to the
y-direction, the E-field pattern forms a cosine distribution between the null planes.
This pattern can be trapped or guided by introducing a second conducting sheet at
the lower null locus in Fig. 1.8 (green dashed line). The trapped pattern travels in
the z-direction with a phase velocity:
vp ¼ c= sin h ð1:7Þ
and a wavelength:
kp ¼ ko = sin h ð1:8Þ
The phase coefficient for this trapped wave bp ¼ 2p=kp and hence
mko
cos h ¼ ð1:10Þ
2a
depicted in Fig. 1.9, the free-space wavelength must be less than 2a for propagation
to occur. The corollary is that the frequency of the wave f (=1/ko) must be greater
than a certain critical value or cut-off value corresponding to the cut-off wavelength
kc ¼ 2a. Furthermore if a\ko \2a the m = 2, 3, 4, solutions all yield the
impossible requirement that cos h [ 1. This means that in the prescribed frequency
range only the m = 1 solution is possible. The solution is termed the dominant
mode for the parallel plane waveguide of separation a, and is defined as the TE10
mode—with one E-field variation in x and zero variation in y. This mode is shown
in Fig. 1.9a–c. Perfectly conducting ‘lids’ can be introduced at y = 0 and y = b to
form a rectangular waveguide, without altering the pattern, because the E-field is
normal to these walls, and the H-field induces surface current density flows
(J) which ensure magnetic field boundary conditions are satisfied. These current
flows are shown in Fig. 1.9b, c. The b-dimension is usually chosen to be approx-
imately half the a-dimension to maximise bandwidth.
Equations (1.7), (1.8) and (1.9) embody the essential features of the frequency
characteristics of the dominant TE10 mode. Equation (1.9) implies that at cut-off
(h = 0) for the TE10 mode bp = b10 = 0, while at very high frequencies well above
cut-off where h = 90° then b10 ! ko . At the same time from Eq. (1.7) we observe
that vp10 ! c. This behaviour is summarised in the Brillouin diagram presented in
Fig. 1.10, which for a waveguide of width a shows the TE10 mode cutting-on at
koa = p. Thereafter it grows in an parabolic manner (see Appendix A.2) becoming
increasingly asymptotic to the vp = c line. Close to cut-off where b10 approaches
Fig. 1.10 Brillouin diagram showing graphically the frequency characteristics of the TE10 mode
in rectangular waveguide (orange line denotes b10 = ko and thus vp = c for all frequencies: blue
solid line = b10 for a rectangular waveguide of width a)
1.4 Waveguide Feed System 15
Arguably, waveguide fed slot radiators have provided a very significant source of
ideas for recent developments in compact antenna technology. However, early
embodiments, such as those depicted in Fig. 1.11, were impelled by large antenna
arrays for military radars [6–8]. Rapid developments in the 1940s were driven by
the demands of World War II for long range radio detection systems. The fre-
quencies used were generally below or at the low end of the microwave range, so
the waveguide and the resultant antennas were inevitably large. Nevertheless,
war-time advances on slotted waveguide array antennas laid important and reliable
foundations on the electromagnetic issues associated with waveguide fed slots and
waveguide slot arrays [8]. This early work continues to inform current develop-
ments, as hinted at in Chaps. 4, 5, 6 and 7.
Traditional slot array antennas generally employ either longitudinally directed offset
‘shunt’ slots, or transverse centrally located ‘series’ slots, in the broadwall of the
rectangular waveguide feed. Inclined slots located in the sidewall of waveguide are
also common (see Fig. 1.11). Once the TE10 mode wall current flows have been
established, as shown in the Fig. 1.9, the slot excitation mechanism is not too
different from that of the TEM mode examples. Current flow interruption by the slot
induces charge build up on its long edges resulting in an instantaneous E-field
formation across the narrow dimension of the slot. This slot field changes in
sympathy with the modal pattern as it passes through the waveguide at the phase
velocity given by Eq. (1.7) (or Eq. B.15 in Appendix B). At a given incident power
level in the propagating mode, the slot radiation is maximised by operating it near
its resonant length, namely L = ko/2.
Shunt slot in waveguide
Figure 1.11 also reveals that the shunt slot is weakly excited close the waveguide
centre line where the current flow is largely axially directed, and strong near the
side wall where the current flow is almost entirely transverse. This behaviour is
used to control radiation strength. The scattering nature of waveguide fed slot
radiators in explored in detail in Chap. 5.
For the shunt slot, which is z-directed in Fig. 1.11, the induced electric field is
clearly aligned with x axis. Therefore, in the direction of the pattern maximum,
this orientation of the slot dictates that the radiation from it is polarised in a
direction transverse to the axis of the waveguide—usually referred to as trans-
versely polarised. This also true of a slot at position (B) which is separated axially
from the slot at (A) by kg10/2 and on the opposite side of the centre line. In phase
terms the slots are separated by 360° which can be interpreted as 0°. This means
that the slots radiate in phase in the y-direction. The importance of this feature is
outlined in Chap. 7.
The radiation pattern, when the slot is part of a planar array for example, is
predominantly directed into the forward half-space (y > b) since the array surface
presents a large conducting ground plane to the slot. It can be viewed as a
‘half-dough-ring’ in shape with nulls in the axial direction (± z-direction), and
1.5 Waveguide Slot Radiators 17
Fig. 1.12 Computed radiation pattern for a six element shunt slot waveguide fed array (with
acknowledgement to www.remcom.com)
Fig. 1.14 Horizontally polarised resonant waveguide planar array antenna formed from edge-slots
In principle, vertical stacking of an edge slot array should be simpler than for
broadwall slots since the waveguide height (b) is generally considerably smaller that
ko/2, the critical distance for the avoidance of grating lobes (see Chaps. 6 and 7).
However this advantage is somewhat negated by the encroachment of resonant
length edge slots into the broadwall (see Fig. 1.11), so that electromagnetic wave
leakage into the inter-guide gaps occurs. The result is radiation pattern unpre-
dictability and additional cross-polar deficiencies. With each of the slotted waveg-
uides in Fig. 1.14 depicted as terminating in a short circuit usually kg10/4 from the
last slot and the slots separated by exactly kg10/2, the sketch is representative of a
‘resonant’ array (see Chap. 7). It produces a precise high gain principal beam in the
broadside direction at the design frequency. However, high gain delivery is available
over only a very narrow bandwidth.
as a compact, robust search and rescue transponder [9, 10]. In much the same way
as is expounded in the previous section the mode of operation of the slot doublet in
waveguide (Fig. 1.15) can be explained by knowledge of the TE10 mode wall
currents and the disruption caused to these currents by the presence of the slots.
Instantaneous charge build along the long edges of the slot is the source of electric
field in the slot which always traverses the shortest distance within it (see
Fig. 1.15). For the TE10 mode, at the instant shown where the top wall currents flow
outward to the sidewalls they flow inwards on the bottom wall. This means that the
electric field in the top wall slot is polarised in the opposite direction to the bottom
slot, and this remains true for all instances in time as the mode propagates through
the waveguide. Electromagnetic wave scattering by the upper slot produces –x po-
larised radiation into the upper half-space while the lower slot radiates into the
lower half-space with +x polarised radiation. A comprehensive simulation of the
slot doublet geometry has been formulated [9, 10] using the moment method
(Chaps. 4 and 5). It exhibits good omni-directional radiation in the transverse plane
of the waveguide (Fig. 1.16) for a low height waveguide (25% less than conven-
tional waveguide). For the pattern in the figure the ratio of the maximum to min-
imum radiation intensity (at constant radius) is of the order of 1 dB. Perhaps not
surprising, for a full height waveguide it is not nearly so good, at approximately
10 dB. Unfortunately lowering the waveguide height severely restricts power
handling capability.
Polarisation agile waveguide slots [11]
An example of a significantly more complex arrangement of radiating slots in
waveguide has been proposed for a polarisation agile 3-D antenna for a long range
22 1 Evolution of Compact Slot Antennas
ground radar [11, 12]. A typical system of this kind usually comprises a parabolic
reflector providing azimuth gain and a front fed linear array providing elevation
gain and frequency scanning (see Chap. 8). To bring about polarisation agility it is
necessary to create a means of altering the current flows in the walls of the
waveguide feed structure in order to switch TE10 mode excitation from transversely
polarised slots to axially polarised slots. One way of doing this is to employ
bifurcated waveguide as suggested in Fig. 1.17.
This particular bifurcated waveguide format can be excited from a conventional
four port ‘magic-T’ (see Bibliography: D.M. Pozar). The two output ports dovetail
with the bifurcated waveguide feed while of the two input ports one sets up the even
mode of propagation, the other the odd mode alternative. Switching power between
the two input ports represents routine microwave technology. Consequently,
polarisation switching requires a broadwall slot pattern with orthogonally directed
slots which are independently excited by the even and odd propagation modes. The
slot pattern is dictated by the TE10 mode surface current flows as illustrated in
Fig. 1.17.
Given that for effective switching slot radiators must be immersed in the even
mode as a whole, or the odd mode as a whole, there are two possibilities—a
transverse slot straddling the central wall, or bifurcation, and a longitudinal slot
running along the bifurcation. The former is excited by axial wall current flows as
we have already seen. In this case only the even mode is effective in exciting the
transverse slot with an uni-directional axially polarised electric field (see
Fig. 1.17a). For the odd mode the induced electric field in the transverse slot is
weak and in anti-phase across the centre line. The resultant radiation from this slot
1.5 Waveguide Slot Radiators 23
Fig. 1.17 Dual-mode bifurcated rectangular waveguide structure showing even mode to odd
mode polarisation switching mechanism
polarised radiation pattern. If the inputs to the ‘magic-T’ were fed in quadrature
phase through a 3 dB hybrid, for example, the radiation pattern would be circularly
polarised.
A not unrelated, but subtly different, route toward a polarisation agile slot pattern in
waveguide involves using orthogonal modes which are formed in a square waveg-
uide cavity [13, 14], rather than the even and odd mode scheme outlined above.
Rectangular waveguide is readily convertible to a high-Q cavity resonator by ter-
minating a section of waveguide with a short circuit. Transmission line theory then
informs us that this zero impedance load is replicated at every half-wavelength (kg/2)
as we withdraw from the load toward the source (see Bibliography: J.D. Kraus). If
the propagating mode in the original waveguide is the TE10 mode then by locating a
second short circuit at the nearest E-field null to the load traps a TE101 cavity mode,
while the next one traps a TE102 cavity mode—and so on. For a square cavity of
dimensions a a b the resonant frequency for the cavity mode can be deduced
pffiffiffi pffiffiffi
from equation (B.15). It is fixed and given by f0 ¼ c= 2a for TE101 and f0 ¼ 2c=a
for TE102. For example for WG16 for which a = 22.86 mm a TE101 square cavity
resonates at 9.3 GHz, while a TE102 cavity resonates at 18.6 GHz.
Appropriate broadwall slot locations, commensurate with strong radiation, are
indicated in Fig. 1.18. The positions shown are determined, as with the waveguide
slots discussed earlier, by the surface current density flows (J) in the upper (or
lower) inner square faces of the cavity. For the TE101 mode depicted in the left hand
diagram of Fig. 1.18, electromagnetic coupling to the cavity fields occurs equally
strongly at four possible slot locations as shown. Four slots in a balanced or
Fig. 1.20 Evolution of ‘flat’ focused reflectors for incorporation into compact reflectarray
antennas with applications in communication systems
Fig. 1.21 Schematic representation of a Fabry-Perot resonator antenna, a end view of TE mode
structure, b top view of TE mode structure
and the orthogonal mode. However, less helpfully, at frequencies above cut-off,
waves begin to propagate in the x directions (or y) within the parallel plate
structure leading to frequency sensitive and scanning leaky-wave beams. This
operating condition is examined in detail in Chap. 10.
30 1 Evolution of Compact Slot Antennas
The principle behind the retro-directive array, which is addressed in Chap. 11, has
been known about since the 1960s. Retro-direction, whereby an incoming radar
signal, on striking a target reflector is returned back in the direction from which it
came, usefully boosts the ‘visibility’ of the original scatterer. This outcome has
historically been achieved by employing ‘corner reflectors’. These are often seen on
the masts of yachts [15]. The device increases the radar cross-section of the target
thereby enhancing its detectability against severe electromagnetic scattering ema-
nating from the sea. A schematic drawing illustrating the principle of operation of
the corner reflector is presented in Fig. 1.22. Generally it comprises three flat
conducting metal surfaces positioned orthogonally to each other—for example the
inner faces of a cube as suggested in Fig. 1.22a. In practice the corner reflector is
made less bulky and hence lighter than the ‘open cube’ by refashioning it into a
trihedral shape as shown in Fig. 1.22b.
The retrodirective nature of the corner reflector can be illustrated relatively easily
by expressing the ray directions in Fig. 1.22a in vectorial terms. For example the
incident portion of ray 1 can be defined as travelling in direction
r 1 ¼ ax x 1 þ ay y 1 þ az z 1 ð1:11Þ
where ax, ay, az are the unit vectors in the x, y and z directions respectively. If this
ray represents a plane electromagnetic wave at frequency x propagating towards
the corner reflector, then (see Appendix A) the electric field component can be
expressed as:
where a? defines a unit vector normal to r1, and k0 is the free-space propagation
coefficient. On the first reflection on the bottom x-y surface of the corner reflector,
the components of r1 in the x and y directions are unchanged while the z-component
is reversed by Snell’s law, to −azz1. The resultant reflected ray direction can be
expressed vectorially as:
r01 ¼ ax x1 þ ay y1 az z1 ð1:13Þ
At the next reflection on the rear x-z plane (Fig. 1.22a), the x and z components
of r1 are unchanged while it is the y component that is now reversed in phase.
Hence
r001 ¼ ax x1 ay y1 az z1 ð1:14Þ
Finally, at the third reflection at the y-z surface, only the x component is reversed
in phase, ending up with a ray travelling in the following direction:
r000
1 ¼ ax x1 ay y1 az z1 ð1:15Þ
For perfect conductors Eqs. (1.12) and (1.16) are essentially identical except for
the sign of the second term within the exponential argument. The spatial component
of the reflected wave is phase conjugated relative to the incident wave. Note that
this is generally true of all incident plane waves which illuminate the corner
reflector.
In mobile communication terms, and particularly in communications with
satellites, the ability to retain a communications link by retro-directing the received
signal back to its source has considerable potential, if it can be achieved with a less
limited and bulky antenna which the corner reflector represents. The key feature of
retro-directive action is the complex conjugate relationship between the received
and transmitted wave phases. It is shown in Chap. 11 that retro-directivity can be
procured using elegant antenna array formats with each element backed by a phase
conjugating circuit. Not surprisingly, retro-directive antennas, in array format and
incorporating varying levels of technical sophistication, have been evolving since
pioneering work in the 1960s [16, 17] and arguably the planar retro-directive array
offers the most cost effective option for satellite tracking. Recently, a planar
retro-directive array design has been reported for potential use in terminals
accessing the Inmarsat BGAN system [18].
32 1 Evolution of Compact Slot Antennas
References
Bibliography
1. R. Ferrari, An Introduction to Electromagnetic Fields (Van Nostrand Reinhold Co., Ltd., New
York, 1975)
2. C.T.A. Johnk, Engineering Electromagnetic Fields and Waves (Wiley, New York, 1988)
3. R.P. Feynman, R.B. Leighton, M. Sands, Lectures in Physics, vol. II (Addison-Wesley
Publishing Co., London, 1972)
4. P.M. Morse, H. Feshbach, Methods of Theoretical Physics (McGraw-Hill Book Co., New
York, 1953)
5. J.D. Kraus, Electromagnetics (McGraw-Hill Book Co., London, 1984)
Bibliography 33
6. P. Lorrain, D. Corson, Electromagnetic Field and Waves (W.H. Freeman & Co., San
Francisco, 1962)
7. A.J. Baden-Fuller, Engineering Electromagnetism (Wiley, New York, 1993)
8. P. Hammond, Applied Electromagnetism (Pergamon Press Ltd., Oxford, 1971)
9. J.A. Stratton, Electromagnetic Theory (Wiley, New Jersey, 2007)
10. D.M. Pozar, Microwave Engineering, 4th edn. (Wiley, USA, 2011). ISBN: 0470631554
Chapter 2
Fundamentals of Electromagnetic
Radiation
Today, our deep understanding of the nature of antennas, and in particular slot
antennas, originated with the study of what is now a classical problem in electro-
magnetics—namely that of the aperture in a plane conducting screen of infinite
extent illuminated by a plane electromagnetic wave onto one face (see Fig. 2.1).
The earliest significant examination of this, at the time intriguing, problem of how
to calculate the electromagnetic fields, at arbitrary location P(x, y, z), in the empty
half-space to the right of the screen, is attributed to Lord Rayleigh in 1897 [1].
The essence of a solution lies in creating a credible representation of the induced
electric field or magnetic field within the aperture which becomes the source of the
scattered fields. This is not a trivial problem and called for some considerable
ingenuity on the part of early investigators at a time when the available mathe-
matical tools were strictly analogue in character and while, contemporaneously,
only closed form solutions were practicable or workable.
Antenna radiation (or reception) is classified as a field theory problem of the
boundary value type not too unlike the electromagnetic propagation problems
outlined in Appendix A. If the field region is remote from the source, field theory,
as we have seen, identifies transverse electromagnetic waves (TEM waves) in
free-space, or trapped electromagnetic waves in confined or bounded spaces. In this
case the boundary value problem is generally termed ‘source-free’. For such
problems Maxwell’s equations, in the source free differential form presented below,
are detailed in Appendix A.
r E ¼ jxlH ð2:1Þ
r H ¼ jxeE ð2:2Þ
rD¼0 ð2:3Þ
rB¼0 ð2:4Þ
In radiation problems on the other hand, the source in no longer remote and becomes
an integral part of the field problem; radiation from a slot in a conducting surface is just
such a problem. It should, perhaps, be noted that the normal convention, in formulating
the mathematical equations associated with radiation, is to presume that the waves
emanate from the ‘source’. However, the transmit/receive process is entirely reciprocal
as indicated in Fig. 2.2, so we could equally evolve the equations on the presumption
that the focus of the waves is a ‘sink’ and the mathematics would not change.
In these source/sink present electromagnetic boundary value problems the most
general form of Maxwell’s equations are required. These are, in phasor form:
r E ¼ jxlH ð2:5Þ
r H ¼ jxeE þ J ð2:6Þ
rD¼q ð2:7Þ
rB¼0 ð2:8Þ
where J and q are the impressed ‘electric’ current density and ‘electric’ charge
density respectively, which form the source terms for the differential equations.
In radiation problems where the electromagnetic fields emanate from an aperture
(e.g. horn and slot radiators) the basic source of the radiation is by analogy a
2.1 Theoretical Formulation of Radiation from a Slot 37
‘magnetic’ current (electric field) rather than conventional ‘electric’ current (mag-
netic field), in which case it is more appropriate to use the following ‘dual’ forms
for Maxwell’s equations:
r E ¼ jxlH Jm ð2:9Þ
r H ¼ jxeE ð2:10Þ
rD¼ 0 ð2:11Þ
r B ¼ qm ð2:12Þ
where Jm and qm are the impressed ‘magnetic’ current density and ‘magnetic’
charge density respectively representing ‘aperture-source’ terms for the differential
equations. These forms are pursued in Chap. 3.
It is important, particularly in the ‘electric’ case to distinguish between the
impressed currents and charges, and currents and charges which arise because of the
presence of the fields in a medium having finite conductivity. Conduction current
density J is proportional to the electric field and is given by rE, where r is the
conductivity of the material. This current component can readily be accounted for,
in the Maxwell equations, by replacing the permittivity e by the complex form
e(1 − jtand) = e(1 − jr/xe). The density of free charge, apart from that associated
with impressed currents, may be presumed to be zero. Thus, in subsequent
discussion, the charge density q and the current density J appearing in the field
equations can be taken as comprising purely impressed charges and currents.
38 2 Fundamentals of Electromagnetic Radiation
Any other currents, if they arise, will be accounted for by the complex electric
permittivity which will be written simply as e for convenience.
The reciprocity hinted at in Fig. 2.2 can be established as follows. The two antennas
one in transmit mode (a) and one in receive mode (b) exhibit, respectively, input
impedances ZT and ZR. It is assumed that the antennas exist is the same linear,
isotropic space, which is not necessarily homogeneous, and that the electromagnetic
power radiated from antenna (a) propagates toward (b). The equivalent circuit
formations representing this antenna juxtaposition are presented in Fig. 2.3.
In the transmitter circuit (Fig. 2.3a) the generator internal impedance is represented
by Zg and to procure maximum power transfer to space it is set equal to the complex
conjugate of the antenna impedance ZT Zg ¼ ZT . In the receiver circuit, on the other
hand, optimum power transfer to the load (ZL) requires that ZL is set equal to the
complex conjugate of the receive antenna ZR; that is ZL ¼ ZR . These impedance
settings are convenient for the following reciprocity calculation, but are not essential.
The calculation proceeds by computing, firstly, the power (P1) delivered to
antenna (a) from the generator. This power is given by:
2
1 1 Vg ZT Vg Vg
P1 ¼ ReðVT IT Þ ¼ Re ¼ ð2:13Þ
2 2 ZT þ Zg ðZT þ Zg Þ 8RT
This power radiates into the surrounding space, and some of it is collected by the
receive antenna and becomes absorbed by the load. This power P2 can by computed
if we ascribe a transfer admittance Y21 to the transmit/receive network which
incorporates the generator impedance, the impedances of the antennas and the load
impedance. Hence
1
1 2
P2 ¼ Re ZL ðVg Y21 ÞðVg Y21 Þ ¼ RL Vg jY21 j2 ð2:14Þ
2 2
The division of Eq. (2.14) by Eq. (2.13) yields an expression for the
receive-to-transmit power ratio for the antenna combination—namely:
P2
¼ 4RL RT jY21 j2 ð2:15Þ
P1
If the transmit antenna and the receive antenna are interchanged with (b) trans-
mitting power P1 and (a) receiving power P2 in its load, the transfer admittance is
now Y12 and the power ratio becomes:
P2
¼ 4RL RT jY12 j2 ð2:16Þ
P1
For linear circuits it is clear that Y21 ¼ Y12 and consequently the antenna
transmit/receive combination displays the property of reciprocity.
Applying a volume integration over V to both sides of this equation, and sub-
sequently enforcing the divergence theorem to the right hand side, leads to the
arguably more useful integral form of the Lorentz reciprocity theorem—namely:
ZZ ZZZ
ðE1 H2 E2 H1 Þ dS ¼ ðE1 J2 þ H2 Jm1 E2 J1 H1 Jm2 ÞdV
S V
ð2:18Þ
Equation (2.18) leads to some useful relationships in certain limited, but rele-
vant, conditions. In a source free region (J1 = J2 = Jm1 = Jm2 = 0) Eq. (2.18)
evidently reduces to:
ZZ
ðE1 H2 E2 H1 Þ dS ¼ 0 ð2:19Þ
S
This equation is a special case of the Lorentz reciprocity theorem and must be
satisfied on the enclosing surface S of a source free region containing the fields E1,
H1 and E2, H2 simultaneously.
From an antenna perspective it is often the case that the sources and sinks
(transmit and receive antennas) can be considered to reside in a sphere (free space)
of infinite radius. At the surface of this sphere (S) at infinity, the fields generated by
the finite sources within it (J1, J2, Jm1, Jm2) will have reduced to zero, which means
that the left hand side of Eq. (2.18) is zero. Hence:
ZZZ
ðE1 J2 þ H2 Jm1 E2 J1 H1 Jm2 ÞdV ¼ 0 ð2:20Þ
V
It is pertinent to note here that Eqs. (2.19) and (2.21) are not power or energy
expressions because complex conjugation is absent. What these equations do rep-
resent is field/source coupling or reaction resulting in further field components [1].
In antenna terms, and in particular array antennas, the process gives rise to mutual
coupling (see Chap. 6) which can present a major source of error, if handled
incorrectly, when precision pattern control is sought in such arrays. In addition, the
above reciprocity relations are particularly relevant to the electromagnetic mod-
elling of elemental antennas, such as dipoles and slots embedded within compact
antenna geometries, as we shall see in Chap. 5.
2.2 Reciprocity for Macroscopic Antennas 41
The impressed currents and charges, expressed explicitly in the source-present form
of Maxwell’s equations (Eqs. 2.5–2.8, and 2.9–2.12), must be related to each other
through the equation of continuity for current and charge. That is:
@q
rJ¼ ð2:22Þ
@t
@qm
r Jm ¼ ð2:23Þ
@t
r r E k 2 E ¼ jxlJ
ð2:24Þ
r r H k2 H ¼ r J
rr J
r2 E þ k 2 E ¼ jxlJ ð2:25Þ
jxe
r2 H þ k2 H ¼ r J ð2:26Þ
In deriving the above equations it has been presumed that l and e are invariant
with time. In free space or vacuum, l ¼ l0 ¼ 4p 107 H=m and
e ¼ e0 ¼ 36p
1
109 F=m.
42 2 Fundamentals of Electromagnetic Radiation
r2 E þ k 2 E ¼ 0
ð2:27Þ
and r2 H þ k 2 H ¼ 0
r r H k 2 H ¼ þ jxeJm
ð2:28Þ
r r E k2 E ¼ r Jm
or equivalently:
r 2 E þ k 2 E ¼ r Jm
rr Jm ð2:29Þ
and r2 H þ k 2 H ¼ jxeJm þ
jxl
These magnetic current sources, and the antenna forms which they generate, are
treated in depth in Chap. 3.
It is clear on examining their forms that the impressed current density vectors
J and Jm enter into the inhomogeneous differential Eqs. (2.25), (2.26) and (2.29) in
a particularly complicated manner, making their direct solution extremely difficult,
although with modern computer based electromagnetic solvers employing finite
difference or finite element techniques—not impossible. Nevertheless, this difficulty
is normally circumvented by resorting to the use of auxiliary potential functions,
which generate more easily solvable wave equation forms. These auxiliary potential
functions may, or may not, represent clearly definable physical entities (especially
is this so in the absence of sources), and so it is usual to adopt the view point that
these potentials are merely useful functions from which the electromagnetic fields
can be derived. The two most common potential functions which arise in radiation
problems are discussed in the next section.
In formulating auxiliary potential functions, with the aim of facilitating the solution
of the source present second order differential equations as represented by
Eqs. (2.25), (2.26), (2.28) and (2.29), it is usually advisable to choose forms that
2.3 Auxiliary Potential Functions 43
satisfy at least one or more of the basic Maxwell equations. For the ‘electric’ current
Eqs. (2.25), (2.26) such a choice is:
B¼rA ð2:30Þ
@A
E¼ þ r/ ð2:31Þ
@t
satisfies Maxwell Eq. (2.5) since r ru ¼ 0 for any scalar function. The auxil-
iary vector A is usually termed the magnetic vector potential, which has relevance
in magnetostatics, while auxiliary scalar / is termed the electric potential (some-
times denoted by V) and has its origins in electrostatics. Consequently, if we can
develop solvable source present second order differential equations for A and /,
then the associated electric and magnetic field forms can be deduced subsequently
from the above two equations.
To develop differential forms in A and / we again enrol the Maxwell equations
as expressed in Eqs. (2.5)–(2.8). On substituting D (= e0E for vacuum) in Eq. (2.7)
with the expression for E given in Eq. (2.31) we obtain:
@ q
r2 / ðr AÞ ¼ ð2:32Þ
@t e0
which hardly seems an advance, in simplification terms, on Eqs. (2.25) and (2.26).
However, Eqs. (2.30) and (2.31) only partially define A and /. Equation (2.30)
delineates only a rotational component of A (Ar say) while Eq. (2.31) defines a part
of / which has finite gradient in the space domain. Therefore, we can add to A any
component which has a zero curl (termed a lamellar component) without altering
Eq. (2.30), and we can add to / any function whose gradient is zero, leaving
Eq. (2.31) unchanged. Provided the Maxwell relationships are not disrupted, these
arbitrary introductions to A and / can be chosen in a manner which simplifies
Eq. (2.32). Such a choice is the Lorentz condition whereby:
1 @/
rA¼ ð2:33Þ
c2 @t
Actually this choice is not totally arbitrary. The Lorentz condition can be
identified with the Kirchhoff law which relates conduction current to rate of change
of charge (Eq. 2.22). It also brings the resultant radiation equations into line with
the requirements of special relativity [3].
On substituting Eq. (2.33) into Eq. (2.32) a second order differential in / alone
is formed, namely:
44 2 Fundamentals of Electromagnetic Radiation
1 @2/ q
r2 / ¼ ð2:34Þ
c @t
2 2 e0
or in phasor form:
q
r2 / þ k02 / ¼ ð2:35Þ
e0
where
x pffiffiffiffiffiffiffiffiffi
k0 ¼ ¼ x l0 e0 ð2:36Þ
c
Recalling that H ¼ B=l0 and e = e0 in free space then substituting Eqs. (2.30)
and (2.31), together with the Lorentz condition, into Eq. (2.6), yields an equivalent
second order differential equation for A. In phasor form, this equation is termed the
vector Helmholtz equation:
Equations (2.35) and (2.37) are clearly wave equations but with source functions
on the right of the equals sign which are much less convoluted than those in
Eqs. (2.25) and (2.26). These source functions have now been simplified to the
impressed charge density (divided by a constant eo) in the region of interest in
Eq. (2.35), and impressed current density (multiplied by a constant lo) in
Eq. (2.37). If q or J are known in space and time, on an antenna located in a volume
of space (say), then in principle the above equations can be solved directly for / or
A respectively. In practice, only one of the two wave equations need be solved
since / and A are related through the Lorentz condition. However, rather than
demonstrate the direct route to a solution, it is actually more instructive to construct
a solution by employing electrostatic and magnetostatic analogies.
If Eq. (2.7) is applied to a small spherical volume of stationary and isolated free
charge (q), located in a magnetic field free region, then symmetry and the circu-
lation law for E (Eq. 2.5) dictate that the electric vector D will be everywhere radial
and unvarying over an enveloping spherical surface (S) of radius r (Fig. 2.4),
enclosing a volume V. Under these circumstances volume integration on both sides
of Eq. (2.7) yields:
2.4 Electrostatics Analogy 45
ZZZ ZZ
r DdV ¼ D dA ¼ 4pr 2 Dr ¼ q ð2:38Þ
V S
Clearly Maxwell’s first equation reduces to the inverse square law, or Gauss’ law
which is usually written:
q
Er ¼ ð2:39Þ
4pr 2 e0
D ¼ e0 E ð2:40Þ
In the m.k.s. system, D has the dimension C/m2, while the constant of propor-
tionality e0 is termed the absolute permittivity of free space. Measurement reveals
that it has the value 8.84 10−12 F/m, where in the m.k.s. system a Farad in a
Coulomb/volt. More generally, D = ere0E, where er is termed the relative permit-
tivity of the material concerned. It need not be a constant, and could be a complex
tensor to represent a general non-linear, anisotropic, inhomogeneous medium.
In conjunction with the Lorentz force equation:
F ¼ qE Newtons
Equation (2.39) yields the Coulomb force law which can be expressed as:
qQ
F¼ Newtons ð2:41Þ
4pr 2 e0
for isolated point charges q and Q (in coulombs) distance r metres apart in a
vacuum.
46 2 Fundamentals of Electromagnetic Radiation
qðx; y; zÞdV 0
d/ ¼ ð2:42Þ
4pe0 jr r 0 j
The potential at P due to the entire charge volume (V) at O, is the summation of
all contributions qðx; y; zÞdV 0 at all possible positions O′ throughout the charge
volume. In the limit this summation becomes an integral, giving:
ZZZ
1 qdV 0
/¼ ð2:43Þ
4pe0 jr r 0 j
V
While this equation exhibits the correct spatial form for electric potential at
distance from an arbitrary charge distribution, it is applicable only to time invariant
charge concentrations. The discussion in relation to Fig. 3.1 suggests that if the
charge alters with time then Eq. (2.43) could justifiably be applied to fields in close
proximity to the charge, but at more distant locations as a time retardation mecha-
nism is required. So can we build time delay into Eq. (2.43)? It turns out that we can.
If the charge distribution at O′ in Fig. 2.5 alters with time, the information that
this change has occurred can be felt at point P only after a period of time jr r 0 j=c,
which is the time it takes for a TEM wave to travel distance jr r 0 j. Hence, it is not
That Eq. (2.44) is a solution to Eq. (2.35) can be established by substitution. The
solution to the phasor form of Eq. (2.35) simplifies to:
ZZZ
1 expðjko jr r0 jqðr0 ÞdV 0
/ðrÞ ¼ ð2:45Þ
4pe0 j r r0 j
V
with time expressed through the common factor exp(jxt) which can be suppressed.
Equation (2.46) is Ampere’s law, and since I is expressed in amps (A) in m.k.s.
system, the dimension of magnetic field strength H is A/m. With H in A/m and B in
Tesla consistency of the magnetic field laws requires that for vacuum:
48 2 Fundamentals of Electromagnetic Radiation
B ¼ l0 H ð2:47Þ
where the constant l0 is termed the absolute permeability of free space and has the
magnitude 4p 10−7 H/m (Henry = T m2/A). As with electrostatics we can say
that more generally, B = lrl0H, where lr is termed the relative permeability of the
material concerned. It need not be a constant, being a function of H in a ferrite
material, and it could be a complex tensor to represent a general, anisotropic,
inhomogeneous medium. Application of the circuital law for H to a current element
Id‘ rather than to a current on a long straight wire, leads to the Biot and Savart law
(see Balanis in Bibliography). Referring to Fig. 2.6 where a current element Idl0 is
located at O0 a distance r 0 from the origin at O. In determining the magnetic flux
density at P distance r from O, we can assume from Ampere’s law that the flux
forms circular loops centred on the current I. The contribution dB at P must be
proportional to I in magnitude and inversely proportional to the square of the
distance between O0 and P, namely jr r0 j. Directionally the vector dB must be
orthogonal to both Idl0 and the vector ðr r0 Þ. So mathematical reasoning suggests
that we can express dB in the following form:
l0 I dl0 ðr r0 Þ
dB ¼ ð2:48Þ
4p jr r0 j3
This the Biot-Savart law which is most commonly expressed in the form:
l0 Id‘
dB ¼ ð2:49Þ
4pr 2
for a current element at the centre of the magnetic field loop. If the straight portion
of wire carrying I is actually part of a complete electrical circuit s, then on summing
all the Idl0 contributions around s leads, in the limit of infinitesimally small dl0 , to an
integration giving:
I
l0 I dl0 ðr r0 Þ
BðrÞ ¼ ð2:50Þ
4p j r r0 j 3
s
dl0 ð r r0 Þ
r ¼ dl0
j r r0 j jr r0 j3
The time retarded form for A can be deduced in much the same way as we have
already done for / with the additional observation that current and current density
are related through:
Idl0 ¼ JdV 0 ð2:52Þ
Hence we obtain:
ZZZ
l0 Jðr0 ; t jr r0 j=cÞ 0
Aðr; tÞ ¼ dV ð2:53Þ
4p j r r0 j
V
or in phasor form:
ZZZ
l0 expðjko jr r0 jÞJðr0 Þ 0
AðrÞ ¼ dV ð2:54Þ
4p j r r0 j
V
The retarded potentials A and / represent powerful tools in the search for
solutions to complex electromagnetic radiation and reception problems, and are
widely used in various simulation packages by providing the ‘building blocks’,
such as the ‘short current element’, to construct field patterns for distributed current
50 2 Fundamentals of Electromagnetic Radiation
sources, which are known in space and time. This is particularly true of those
packages employing moment methods. The short current element, or Hertzian
dipole, is examined in the next section.
Needless to say, equivalent forms to A and / also exist, which provide expedient
routes to the solutions to Eqs. (2.28) and (2.29). These are usually defined by the
parameters Am—an auxiliary potential function associated with ‘magnetic current’
Jm, and /m—an auxiliary potential linked to ‘magnetic charge’ qm. These forms
will be considered in more detail in Chap. 3.
The Hertzian dipole and the ‘short current element’ are essentially synonymous.
The concept is presented schematically in Fig. 2.7. The short current element is
viewed as a filament of current Idl which exhibits constant magnitude along its
length dl. As such it is physically impossible to realise, but remains a helpful
mathematical construct. It functions as an exceptionally useful building block in
antenna modelling which enables the field patterns of practical antenna structures to
be determined by integration. Consequently it is instructive to examine the field
behaviour of this current filament, and this is implemented by applying the
appropriate potential function to the example shown in Fig. 2.7.
The current element is assumed to be located at the origin of a Cartesian
coordinate system and that it takes the form a constant z-directed current varying in
time at frequency x rad/sec—i.e. the current I ¼ I0 cos xt. If time retardation is
incorporated into the definition of current then this becomes:
r
I ¼ I0 cos xðt Þ ¼ I0 cosðxt k0 r Þ ð2:55Þ
c
l0 I0 dl
A¼ expðjðxt k0 rÞÞ ð2:56Þ
4pr
If we adopt the unit vector notation, for the Cartesian system, namely
^ax ; ^ay ; ^az then:
l0 I0 dl^az
A ¼ Az ^az ¼ expðjðxt k0 rÞÞ ð2:57Þ
4pr
It is evident from Eq. (2.57) that at distant point P (Fig. 2.7) the magnetic vector
potential is z-directed. However, we know from our earlier discussion on the
radiation mechanism that the important field components are orthogonal to the
direction vector r and are therefore best represented by spherical components.
Consequently we let:
where Ar ¼ Az cos h, Ah ¼ Az sin h and A/ ¼ 0. Hence the magnetic field com-
ponents of the radiated wave at P can be derived using B ¼ l0 H ¼ r A resulting
in [6]:
I0 dl jk0 1
H/ ¼ þ 2 sin h expðjk0 rÞ ð2:59Þ
4p r r
Hh ¼ Hr ¼ 0 ð2:60Þ
The electric field components of the same radiated wave can be constructed by
applying the Maxwell Eq. (2.2) to generate:
rffiffiffiffiffi
l0 I0 dl jk0 1 j
Eh ¼ þ 2 3 sin h expðjk0 rÞ ð2:61Þ
e0 4p r r k0 r
rffiffiffiffiffi
l0 I0 dl 1 j
Er ¼ cos h exp ðjk0 rÞ ð2:62Þ
e0 2p r 2 k0 r 3
52 2 Fundamentals of Electromagnetic Radiation
E/ ¼ 0 ð2:63Þ
From the discussion in Sect. 2.1 the fields radiating away from the short current
element are expected to be normal to the direction vector r, and to each other. They
are also predicted to diminish in magnitude at a rate which is inversely proportion to
distance r from the source. The only field components which meet this requirement
are the first term of H/ in Eq. (2.59) and the first term of Eh (Eq. 2.61). So in the far
field, generally defined as occupying the region given by k0r > 1:
jk0 I0 dl
H/ ¼ sin h expðjk0 rÞ ð2:64Þ
4pr
rffiffiffiffiffiffiffiffiffiffiffiffi
l0
Eh ¼ H / ¼ g0 H / ð2:65Þ
e0
instant in time, an electric dipole. The field exists only in the ‘near’ field region
(kor < 1) of the source, being negligible in the far field. It supports electrostatic or
capacitive energy storage.
The 1/r2 which exists in both the electric and magnetic field formulations
(Eqs. 2.59, 2.61 and 2.62) exhibits the characteristics of the stored inductive
magnetic field in the vicinity of a current carrying wire. In magnetostatics it is
derived using the Biot-Savart law. The electric field components arise because the
antenna current, and hence this inductive magnetic field, is time varying. The fields
are derivable using the Faraday law (Eq. 2.5). Again these fields diminish rapidly
with distance from the source and are essentially zero in the far field.
The nature of these field components can be further illuminated by examining
complex power flow in the near field of the source. Complex power flow density in
W/m2 is generally expressed in the form:
1
p ¼ E H W/m2 ð2:66Þ
2
1
p ¼ ð^a/ Er Hh ^ah Er H/ ^a/ Eh Hr þ ^ar Eh H/ þ ^
ah E/ Hr ^
ar E/ Hh Þ ð2:67Þ
2
On introducing Eqs. (2.59) to (2.63) into Eq. (2.67) the following components to
the complex power density p for the short current element are derived:
rffiffiffiffiffi
jk0 I02 dl2 l0 1 1
p ¼ ^ah þ sin h cos h
16p2 e0 r 3 k02 r 5
rffiffiffiffiffi ð2:68Þ
k0 I 2 dl2 l0 k0 j
þ ^ar 0 2 sin2 h
32p e0 r 2 k02 r 5
The real power flow density radiating in the r-direction away from the source is
given by pr ¼ Re p ¼ 12 ReðE H Þ and has the form:
rffiffiffiffiffi 2
k02 I02 dl2 l0 sin h
pr ¼ ð2:69Þ
32p2 e0 r 2
The remaining components of Eq. (2.68) are imaginary in character and repre-
sent stored energy in the near field on the source. These components can be
extracted by expressing pi ¼ Im p. This leads to an r-directed component:
54 2 Fundamentals of Electromagnetic Radiation
rffiffiffiffiffi 2
jI02 dl2 l0 sin h
pi j r ¼ ð2:70Þ
32k0 p2 e0 r 5
The real power flow density (Eq. 2.69) diminishes, as expected, in proportion to
the inverse radius squared (inverse square law), while the sine squared dependence
of the equation denotes that the radiating lobe maximises in the x-y plane (at
h = 90°) for all values of /. The stored energy density term pi jr (Eq. 2.70), which
decays in inverse proportion to radius to the power of five, also exists mainly in the
x-y plane. This electric dipole energy essentially oscillates at the frequency of the
source, between electric and magnetic stored forms, with the instantaneous power
flow density moving in a radially direction. On the other hand, the stored energy
density represented by Eq. (2.71), which displays h-directed instantaneous power
flow, is concentrate in ‘energy lobes’ located at h ¼ 45 , for all values of / (see
Fig. 2.9). Both electric dipole (1/r3) and inductive (1/r5) energy form are represented.
To complete the picture of the radiation behaviour of the short current element it
is instructive to examine to total power delivered by the element for a given current.
To determine the total power it is necessary to integrate the power density over a
spherical surface centred on the element and enclosing it. That is:
ZZ Z2p Zp rffiffiffiffiffi 2
k02 I02 dl2 l0 sin h 2
Prad ¼ pr ^ar dS ¼ r sin h dh d/
32p2 e0 r 2
S 0 0
ð2:72Þ
rffiffiffiffiffi Zp
k02 I02 dl2 l0
¼ 2p sin3 h dh
32p2 e0
0
Rp
But sin3 h dh ¼ 4=3 and k0 ¼ 2p
k0 where k0 is the free space wavelength. Hence:
0
pg0 I02 dl 2
Prad ¼ ð2:73Þ
3 k0
by using the relation η0 = 120p. Not unexpectedly, the power radiated is propor-
tional to the square of the current applied, but more importantly it is proportional to
the square of the element length normalised to the free space wavelength. This ratio
occurs repeatedly in antenna theory and underpins the unmistakeable trend that
antennas largely diminish in size as frequency rises.
In electrical circuits it is useful to express power in terms of a resistance, and so
it is convenient to make the relationship:
Prad ¼ Irms
2
Rrad ð2:75Þ
where Irms is the root mean squared current in the source, to yield:
1
Prad ¼ I02 Rrad ð2:76Þ
2
Rrad is termed the radiation resistance of the source. On combining Eqs. (2.76)
and (2.74), the following classical equation is obtained:
2
dl
Rrad ¼ 80p2 Ohms ð2:77Þ
k0
embed and match practical antennas into real transmission systems and these will
be explored in Chaps. 6, 7 and 8.
Bibliography
Fig. 3.1 Examples of aperture antennas: a H-plane horn, b E-plane horn, c sectoral horn,
d shunt-slot in waveguide, e edge-slot in waveguide, f dielectric rod, g horn fed parabolic antenna
currents Js and Jms are imposed such that the external field E, H is unaltered, then
the introduced currents are equivalent sources for the region V2. Therefore on S,
Js ¼ n^ ðH Hi Þ and Jms ¼ ^ n ðE Ei Þ where the interior of S is presumed
to contain a field distribution Ei, Hi which is specified only on S by means of the
3.1 Aperture Antenna Models 59
above boundary relations. Since the interior volume V1 is not of interest in setting
up the equivalent sources, the interior fields Ei, Hi can be arbitrarily chosen. The
most common choice is Ei = Hi = 0 which is termed Love’s equivalence principle
^ H and Jms ¼ ^
[3]. In this case, on S, Js ¼ n n E. Furthermore this means that
the medium within S can be chosen arbitrarily. For example, a conductor can be
judiciously placed there with no electromagnetic effect on the arrangement.
In Fig. 3.2a it should be noted that both electric current density Js and magnetic
current density Jms have been inserted onto the surface S. However, according to
the uniqueness theorem [3] we need only specify Js or Jms on S, not both, to ensure
that the resultant field solution is unique. For example, consider the conducting
body S with an aperture A as shown in Fig. 3.2b for which the field in the aperture
(Ea n ^) is known. The radiated field in the exterior region is unaffected if the
aperture in the conducting body is closed by a conducting wall as depicted in
Fig. 3.2b (right hand diagram) and replaced by the magnetic current Jms ¼ Ea n ^
placed closely adjacent to the position of the original aperture—essentially on the
surface of the aperture wall.
Fig. 3.2 Equivalence principles a for arbitrary surface S b for conducting body S with aperture A
60 3 Compact Aperture Antennas
Fig. 3.3 Application of equivalence principle a original horn structure b enclosing surface S
c equivalent aperture in infinite screen
1
E¼ r Am ð3:1Þ
e0
where Am is the auxiliary potential function. Unlike the magnetic potential function
A in Chap. 2, Am has little electrical significance, other than as a useful mathe-
matical generator of electric field, and hence magnetic field.
By duality, the Maxwell equations which are applicable to the problem of
radiation from a magnetic current source of density Jm and magnetic charge density
qm are:
r E ¼ jxlH Jm ð3:2Þ
r H ¼ jxeE ð3:3Þ
62 3 Compact Aperture Antennas
rD¼0 ð3:4Þ
r B ¼ qm ð3:5Þ
The magnetic current density Jm and magnetic charge density qm are linked
through the equation:
r Jm ¼ jxqm ð3:6Þ
which is suggestive of the Kirchoff relationship between electric current and charge
(see Chap. 2).
Following a derivation paralleling Chap. 2, these lead to two second order vector
differential equations for electric and magnetic fields, namely:
r 2 E þ k 2 E ¼ r Jm ð3:7Þ
and
rr Jm
r2 H þ k2 H ¼ jxeJm ð3:8Þ
jxl
Clearly the right hand side source functions in these equations are unappealingly
complex. However the auxiliary potential function choice delineated above, when
substituted into the derivation process as outline in Chap. 2, generates a much more
amenable second order vector differential equation, as follows:
Table 3.1 Duality of equations associated with electric and magnetic current sources
Electric sources (J 6¼ 0, Jm = 0) Magnetic sources (Jm 6¼ 0, J = 0)
r E ¼ jxl0 H r H ¼ jxe0 E
r H ¼ J þ jxe0 E r E ¼ Jm þ jxl0 H
H ¼ l1 r A E ¼ e10 r Am
0
Radiation mechanisms associated with antennas which fall into the category of
‘aperture antennas’ are attributable to diffraction from the electric field distribution
formed across the aperture at any given instant in time. The field distribution from
an aperture can be modelled in many ways (as we shall see in Chap. 4) but the
radiation mechanism can be understood most clearly and directly by examining the
electromagnetic penetration of a narrow one-dimensional slit as depicted in
Fig. 3.4. The arrangement comprises a perfectly flat conducting screen of infinite
extent in the x-y plane, containing a slit of width a in the x-direction and extending
to infinity at y. A plane electromagnetic wave in the left hand half-space is
incident on the screen and hence the slit and sets up, in sympathy with the
x-polarised plane wave, an x-directed electric field E0 in the aperture, which we can
presume to be constant across the narrow aperture.
Fig. 3.4 Electromagnetic wave diffraction by a parallel sided slit in a perfectly conducting
membrane (Huygen’s mechanism)
be proportional to E0 and also to the width dx of the element. Hence the field at the
remote point P in Fig. 3.4, which is located at distance r′ from the source at (x, 0, 0)
and distance r from the origin at O, can logically be represent by a function of the
following form:
AE0 dx
dEP ¼ pffiffiffiffi expðjk0 r 0 Þ ð3:11Þ
r0
where A is an amplitude constant, and k0 is the propagation coefficient for free space
electromagnetic waves in the right-hand region. If P is very remote from the
aperture such that r a, then we can reasonably presume that r 0 r in the
denominator of Eq. (3.11). However, for the much more sensitive phase shift we
use r 0 ¼ r x sin h and so Eq. (3.11) becomes:
AE0 dx
dEP ¼ pffiffi expðjk0 ðr x sin hÞÞ ð3:12Þ
r
The total diffracted field at P due to the field E0 in the aperture can now be
determined by summing all of the dx contributions at P for all values of x between
−a/2 and +a/2. This results in the following integral equation:
Z a=2
þ
AE0
EP ðr; hÞ ¼ pffiffi expðjk0 rÞ expðjk0 x sin hÞdx ð3:13Þ
r
a=2
Significantly, the form of this equation reveals that the diffracted wave at P from
the superposition of cylindrical waves, emanating from all parts of the aperture
containing the field E0, combine to form an ‘interference pattern’ which in this case
has a characteristic sin X/X distribution. Distributions of this nature are ubiquitous
in aperture antenna theory, and they will be examined in more detail in the fol-
lowing section.
Fig. 3.5 Plane wave diffraction at a narrow slit in an infinite perfectly conducting screen—
illustrating the diffraction pattern is a spectrum of plane waves
On the other hand for a radiating slit of finite width the spectrum function is by
no means constant. In this more general case, it can be evaluated be reference to
Fig. 3.5, where the slit has a width a in the x-direction. As in the Huygens example
the slit is irradiated from the left by a single plane wave of E-field amplitude E0.
The figure shows a typical plane wave component of the radiated power in the right
hand half-space. It propagates in the direction k0 which is at an angle h to the z-axis.
pffiffiffiffiffiffiffiffiffi
If the right hand region is free-space the magnitude of k0 is given by k0 ¼ x l0 e0 ,
and the direction is given by that of the constituent plane wave. The point P(x, z)
(Fig. 3.6) is expressed in the cylindrical coordinate frame by the vector r. The plane
wave propagating in the direction h subtends angle dh at the origin, and logically
the wave magnitude will be proportional to both F(sin h) and dsin h, on using sin h
rather than h to denote angle. Hence the E-field at P due to this representative plane
wave can be expressed as:
and on collecting and adding all possible contributions from all typical plane waves
the following integral function is created:
3.3 Radiation from a Linear Aperture 67
Z1
EP ðx; zÞ ¼ Fðsin hÞ expðjk0 ðx sin h þ z cos hÞÞdðsin hÞ ð3:16Þ
1
The integration limits are taken from 1 to þ 1 to ensure that all possible
contributions to the diffracted field are included—both real and imaginary.
If the field in the aperture is expressed functionally as:
Z1
f ðxÞ ¼ Fðsin hÞ expðjk0 x sin hÞdðsin hÞ ð3:18Þ
1
In electrical circuit theory Fourier transforms are regularly used to relate fre-
quency to time and vice versa through the equations:
Z1
1
f ðtÞ ¼ FðxÞ expðjxtÞdx ð3:19Þ
2p
1
68 3 Compact Aperture Antennas
Z1
FðxÞ ¼ f ðtÞ expðjxtÞdt ð3:20Þ
1
Z1
1
Fðsin hÞ ¼ f ðxÞ expðjk0 x sin hÞdx ð3:21Þ
k0
1
This transform relationship between the angular spectrum and aperture field,
admittedly expressed in a limited two dimensional form here, in Eqs. (3.18) and
(3.21), is a very important one in antenna theory and will be considered in more
detail below.
In the elementary diffraction example presented in Fig. 3.5, the electric field, and
hence f(x), in the range of x from 1 to þ 1, is non-zero only in the slit:
therefore,
E0 for j xj a=2
f ðxÞ ¼ ð3:22Þ
0 for j xj [ a=2
Z1
E0
Fðsin hÞ ¼ expðjk0 x sin hÞdx
k0
1
It has the same angular dependence in the far-field as the radiation pattern
represented by Eq. (3.14) which was deduced from the Huygen’s principle. That
the spectrum function bears a direct relationship with the far-field pattern of an
aperture antenna is more fully developed in many texts on antenna theory [3]. This
powerful relationship remains true for all aperture antennas, not just the slit example
examined here.
3.4 Spectrum Functions 69
The Fourier transform pair [Eqs. (3.18) and (3.21)] which expresses the relationship
between spectrum function (far-field) and an aperture field function (antenna source
field) provides an excellent base from which to examine the fundamental radiation
characteristics of aperture antennas. They are of course limited to furnishing pattern
information for the principal plane of the antenna only—that is the plane containing
the major axis of the aperture and the normal to the aperture (the x-z plane in
Fig. 3.5). Notwithstanding this limitation all the essential features of the antenna
pattern can be demonstrated. To assess antenna behaviour in planes other than the
principal plane the three dimensional forms of Eqs. (3.18) and (3.21) must be used.
These are developed in Sect. 3.5.
In Fig. 3.7 spectrum functions for four commonly used antenna aperture field
distributions are displayed in two-dimensional rectangular rather than polar form.
The functions themselves are listed in Table 3.2. The rectangular form is more
traditional but with the growing availability of sophisticated drafting tools within
EM solvers the polar representation is becoming increasingly common place. Of
course, for ‘real’ antennas radiation patterns are three dimensional, and modern
drafting software can generate 3-D images. However 3-D representations on a 2-D
page can lose subtle details of a given pattern, and so the 2-D depictions continue to
Fig. 3.7 Far-field patterns for apertures with the following field distributions—uniform, cosine,
cosine squared, cosine squared on a 10 dB pedestal
70 3 Compact Aperture Antennas
be used to ensure that the pattern of interest is fully recorded. Normally this means
that at least two planar patterns are required representing the major planes (or ‘cuts’)
of the antenna. These are usually chosen to be orthogonal to each other. The
traditional planes employed are predominantly the vertical and horizontal planes for
ground mounted antennas. These are also termed the elevation and azimuth planes
particularly in radar applications. The reference direction au = 0 (i.e. h = 0°) is
generally assumed to be aligned with the normal to the surface of the aperture.
These directions coincide in practice if the antenna is symmetrically distributed
relative to h = 0° axis.
stratagem introduces more nulls into the pattern, and hence moves the first null
(nearest the mainlobe peak) toward au = 0, that is h = 0, thus ‘squeezing down’ the
mainlobe width.
For example if a = 0.3 m and k0 = 3 cm the first null occurs at 5.7° for a
uniform aperture illumination (Fig. 3.7). If a is now increased to 0.6 m the first null
shifts to 2.85° thus narrowing the primary lobe. The magnitude of the first sidelobe
relative to the peak of the mainlobe, as can be seen from Fig. 3.7, has the magnitude
0.22 = −13.1 dB. For most applications this would normally be viewed as too high
and aperture field shaping would be introduced to lower it.
G ¼ gD ð3:24Þ
where η (the efficiency) is a measure of the total power transmitted by the given
antenna divided by the total power delivered to it, or in reception terms the total
power delivered to the receiver divided by the total power received by the antenna.
Directivity and gain will be examined further in Sect. 3.5.
While the examination of linear apertures has the advantage of simplifying the basic
analysis process associated with the evolution of radiation patterns and with
identifying the nature and properties of these patterns, we still need tools to
investigate practical aperture antennas. These real antennas exhibit, of course, two
dimensional apertures, and hence they generate three dimensional radiation pat-
terns. Here it will be demonstrated that two dimensional apertures can be modelled
by revisiting Eq. (3.10) and extending its applicability to two dimensions.
In the computational modelling of two dimensional aperture problems the main
difficulty lies in performing the integrations embedded in Eq. (3.10). However, if
interest is limited to the far-field, which is the normal scenario, the complexity of
the integrations can be significantly reduced. It is helpful at this stage to make some
simplifying definitions. Firstly in the far-field the following approximations can be
made with little impact on accuracy:
3.5 Rectangular Aperture 73
Here n denotes the angle between the vectors r and r′ which are depicted in
Fig. 3.8.
For far-field computations n 0 in which case R and r are essentially parallel.
Hence applying Eq. (3.10) to a radiating aperture which can be represented by a
surface S, we obtain:
ZZ
e0 expðjk0 RÞ 0
Am ¼ Jm dS ð3:27Þ
4p R
S
If we now incorporate Eqs. (3.25) and (3.26) into Eq. (3.27), Am can be written
as:
ZZ
e0 expðjk0 rÞ
Am Jm expð þ jk0 r 0 cos nÞdS0 ð3:28Þ
4pr
S
In the polar (r, h, /) coordinate system the vector Am has two components in the
far-field, namely:
where ah and a/ are unit vectors. There is no Amr since in the far-field the radiation
is transverse electromagnetic in character. At the position of the aperture surface at
r′ (x′, y′, z′) the vector magnetic current can be expressed as:
Hence combining Eqs. (3.28), (3.29), and (3.30) yields the following accessible
forms:
ZZ
e0 expðjk0 rÞ
Amh ¼ Jmx cos h cos / þ Jmy cos h sin / Jmz sin h expð þ jk0 r 0 cos nÞ dS0
4p
S
ð3:31Þ
ZZ
e0 expðjk0 rÞ
Amu ¼ Jmx sin / þ Jmy cos / expð þ jk0 r 0 cos nÞ dS0 ð3:32Þ
4p
S
The electric and magnetic fields in the far-field can be generated using the
appropriate equations in Table 3.1—noting that the divergence of Am is zero in the
far-field.
Hence we obtain, finally, the far-field solution for the general two dimensional
aperture antenna:
Er 0 ð3:33Þ
jk0 expðjk0 rÞ
Eh Amu ð3:34Þ
4pr
jk0 expðjk0 rÞ
E/ þ Amh ð3:35Þ
4pr
Hr 0 ð3:36Þ
jk0 expðjk0 rÞ Amh
Hh ð3:37Þ
4pr g0
jk0 expðjk0 rÞ Am/
H/ ð3:38Þ
4pr g0
Fig. 3.9 Rectangular aperture antenna embedded in a conducting ground plane of infinite extent
ground plane (see Fig. 3.9). The electric field in the aperture is assumed to be
constant in both the x- and y-directions and to be y-polarised. This is not a practical
field distribution but it serves the purpose of the demonstration with minimal
complexity. This field distribution can be expressed as:
Ea ¼ ay E0 ð3:39Þ
existing in the range a=2 x0 a=2 and b=2 y0 b=2. E0 is a constant in this
case.
The equivalent solvable geometry comprises a ‘closed’ half-space bounded by a
flat conducting screen extending over the entire x-y plane at z = 0. Hence applying
the boundary conditions arising from Fig. 3.3, we obtain:
76 3 Compact Aperture Antennas
a/2 x0 a=2
n ^ay Ea ¼ 2^ax E0
Jms ¼ 2^ ð3:40aÞ
b/2 y0 b=2
Er ¼ 0 ð3:41Þ
jabk0 E0 expðjk0 rÞ sin X sin Y
Eh ¼ sin / ð3:42Þ
2pr X Y
jabk0 E0 expðjk0 rÞ sin X sin Y
E/ ¼ cos h cos / ð3:43Þ
2pr X Y
Hr ¼ 0 ð3:44Þ
E/
Hh ¼ ð3:45Þ
g0
Eh
H/ ¼ ð3:46Þ
g0
k0 a k0 b
X¼ sin h cos / and Y¼ sin h sin /
2 2
The normalised E-field distributions typical of the above equations are plotted
using standard mathematical software in Fig. 3.10 for a rectangular aperture with
a = 3ko and b = 2ko, and in Fig. 3.11 for a square aperture with a = b = 3ko. Note
that in Fig. 3.10 the E-field is y-polarized (across the narrow dimension) in
Fig. 3.11 it is x-polarized.
Figures 3.10 and 3.11 provide a good pictorial representation of the penetration
of the aperture radiation pattern into the forward half space of the antenna, provided
it is fully realised that the depicted lobe surfaces are actually constant magnitude
power density contours. They could be viewed as ‘iso-density’ surfaces. Lower
power density contours would push much further out into the half-space, while
higher power densities produce surfaces closer to the antenna. Actually, experi-
mentally this form of presentation is seldom used because the volume of mea-
surement and the consequent time involved in generating such images is much too
high to be cost effective. So, as we have intimated earlier the convention is to
3.5 Rectangular Aperture 77
Fig. 3.10 Three-dimensional E-field pattern for a rectangular aperture (3ko 2ko) in a large
conducting plane—constant Ey-directed case
Fig. 3.11 Three-dimensional E-field pattern for a square aperture (3ko 3ko) in a large
conducting plane—constant Ex-directed case
present antenna patterns as two primary ‘cuts’—generally the E-plane and the
H-plane (see Fig. 3.10). As the figure demonstrates the E-plane for the aperture is
the y-z plane or the / = p/2 plane. The E-field distribution on this plane is
therefore:
Er ¼ E/ ¼ 0 ð3:47Þ
" #
jabk0 E0 expðjk0 rÞ sin k20 b sin h
Eh ¼ ð3:48Þ
2 sin h
2pr k0 b
78 3 Compact Aperture Antennas
On the H-plane (x-z plane or / = 0 plane) the E-field has the form:
Er ¼ Eh ¼ 0 ð3:49Þ
" #
jabk0 E0 expðjk0 rÞ sin k20 a sin h
Eu ¼ cos h ð3:50Þ
2 sin h
2pr k0 a
Interestingly, the term inside the square brackets in Eqs. (3.48) and (3.50) has
reduced to the spectrum functions (see Table 3.2) of the uniformly illuminated
linear apertures of length b and of length a respectively. The important conclusion
here is that the radiation patterns of two dimensional apertures can usually be
adequately represented on the two major planes, or ‘cuts’, by recruiting the spec-
trum functions. These functions generally provide good first approximations to
mainlobe beamwidths, the locations of the sidelobe zeros and maxima, as well as
the magnitudes of the sidelobes.
3.5.2 Directivity
pa 4pr 2 pa
D¼ ¼ ð3:51Þ
Prad =4pr 2 Prad
For an aperture antenna, as described above, if the electric and magnetic fields in
the far-field have already been determined, as detailed in the previous section, then
the radial power density flow at any angle in the forward half-space is given by:
1
Uðh; /Þ ¼ Re ^ah Eh þ ^a/ E/ ^ ah H h þ ^
a/ H /
2
1 2 ð3:52Þ
¼ jEh j2 þ E/
2g0
If at the ‘nose’ of the mainlobe Eh ¼ Eh0 and E/ ¼ E/0 then the power flow
density there we can computed from:
3.5 Rectangular Aperture 79
1 0 2 0 2
pa ¼ Eh þ E/ W/m2 ð3:53Þ
2g0
The total power Prad is more difficult to evaluate. It involves a surface integration
of the general power density term U(h, /) over a hemisphere in the forward
half-space of the radiating aperture (see for example Balanis [3]).
However, since the chosen aperture model is embedded in a plane conducting
surface of infinite extent a simpler procedure exists to compute the total radiated
power. If the electric field ðE0 ðx0 ; y0 ÞÞ in the aperture is available, which it usually is
for known excitation methods, the power density in the aperture can be calculated.
Thence the total power entering the forward half-space namely Prad, can be com-
puted by integrating the power density over the aperture area. That is:
ZZ
1
jE0 ðx0 ; y0 Þj dx0 dy0
2
Prad ¼ ð3:54Þ
2g0
S
where S is the aperture surface. For the uniformly illuminated rectangular aperture
in Fig. 3.9 this gives:
jE0 j2
Prad ¼ ab W ð3:55Þ
2g0
Also, for this aperture example, Eq. (3.53) in combination with Eqs. (3.48) and
(3.50), evaluated at h = 0, yields:
2
ab jE0 j2
pa ¼ W/m2 ð3:56Þ
k 2g0 r 2
where Aeff is the effective area of the aperture. It is worth stressing that, Eq. (3.57) is
actually a very useful relationship for gauging the directivity (and hence gain) of
any aperture antenna, not just for the cited example.
radiation from ‘wire’ antennas, and (b) a magnetic surface current density (Jms) on a
non-conducting interface, which is the source of radiation from an aperture.
Consequently, it becomes possible to take advantage of a suite of analytical pro-
cedures, which have been evolved to explain and simulate wire antenna charac-
teristics (see Chap. 2), and apply them to the related but separate area of
investigation presented by aperture antennas, including slots.
From the perspective of electromagnetic theory the duality between electric
current sources and magnetic current sources is emphasised in a table comparing
and contrasting the fundamental equations. Furthermore, by employing Huygen’s
principle and plane wave spectrum techniques, the precise and uncompromising
relationships between aperture size and shape, and radiation pattern characteristics,
are exposed and explored. As we shall see these seemingly unyielding physical
constraints on radiation pattern manipulation have major significance for compact
slot based antennas.
Bibliography
1. R.E. Collin, F.J. Zucker, Antenna Theory (McGraw-Hill, New York, 1969)
2. R.F. Harrington, Time Harmonic Electromagnetic Fields (McGraw-Hill, New York, 1961)
3. C.A. Balanis, Antenna Theory (Wiley, New York, 1997)
Chapter 4
Computational Modelling Techniques
for Slot Antennas
In so far as the source present Maxwell equations (Eqs. 2.5–2.12) provide a rig-
orous theoretical model of electromagnetic radiation in a properly specified source
present space, so the second order differential equations, repeated below, namely
Eqs. (4.1) and (4.2) are rigorous representations expressed in terms of the vector
potential functions A and Am defined in Chap. 2.
Although these solutions to the second order differential equations have been
constructed by analogy with statics it should be stressed that the differential
equations themselves remain rigorous provided that the parameters and the solution
space are properly defined, as we have already emphasised.
However, for incorporation into computer simulation software, where wide
applicability is a prime goal, a general approach to the solution of second order
differential equations generated by electromagnetism (and other disciplines) is
required. The required versatility can be established by noting that Eqs. (4.1) and
(4.2) can be rewritten in the forms:
Equations of this type belong, generally, within the mathematics of linear spaces
and operators (see Harrington [1]), where they take the form:
LðfÞ ¼ g ð4:7Þ
Here L denotes the scalar operator ( r2 þ k02 in the above equations) while
vector function f is the operated upon function and vector g represents the range of
the operator domain [the source term in the inhomogeneous Eqs. (4.5) and (4.6)].
Equation (4.7) can be written in the form:
Lð^ ^g
uf Þ ¼ u ð4:8Þ
where u^ denotes the unit vector, in the direction of g, which applies on both sides of
the equation for a linear operator.
In Eq. (4.8), if f and g are functions of a continuous variable x, then evidently:
Also, if L is a linear operator, as we have already supposed, then it must [1] have
an inverse L−1 which leads to an inverse form of Eq. (4.9) namely:
and recalling that L operates only on the non-primed x values we obtain in place of
Eq. (4.10):
Z
f ðxÞ ¼ gðx0 ÞL1 dðx0 xÞdx0 ð4:14Þ
Consequently, we can deduce that the solution to Eq. (4.9) takes the form:
Z
f ðxÞ ¼ gðx0 ÞGðx; x0 Þdx0 ð4:15Þ
We can make this change provided the introduced term G(x, x′) is a solution of
G(x, x′) is defined as the Green’s function and is a solution to the second order
inhomogeneous differential equation for which the source term is the d-function.
This may or may not be the whole solution. The right hand side of Eq. (4.16) could
contain a non-singular term, for example Go(x, x′), which is a solution only of the
homogeneous equation, not unlike the ‘constant’ in conventional integration:
LðGo Þ ¼ 0 ð4:17Þ
Hence, we have:
The additional term Go(x, x′) ensures that all the relevant boundary conditions,
and initial conditions are satisfied, leading to a unique solution. In electromagnetic
radiation terms this formulation can reasonably be viewed as the modelling of
radiation from a ‘point’ emitter. For a defined space containing the point source of
84 4 Computational Modelling Techniques for Slot Antennas
radio waves, G(x, x′) is the unique electromagnetic field solution which satisfies the
boundary conditions.
In physical terms, Eq. (4.9) can be interpreted as expressing the relationship
between the given source g and the field distribution which it generates. The
Green’s function expresses that part of the field at a distant point x (at some later
instant t) which is generated by a unit intensity source located at the point x′ (and
earlier time t′) within the source volume. The total field f can then be formed by
integrating (see Eq. 4.10) over all source locations x′ and over all time intervals
over which it acts.
In Chap. 2 it has been shown that in a bounded volume of space containing a source
of radiation, the solution to Maxwell’s equations leads to the formulation of an
inhomogeneous scalar Helmholtz equation in the magnetic vector potentials A, and
a corresponding one Am, with source terms representing electric current density J or
magnetic current density Jm, respectively. It should be noted that the Helmholtz
equation is not the only second order differential equation which emanates from
Maxwell’s equations. However, these other forms are encapsulated within the
vector potential formulation. So we need examine only the Green’s functions which
are applicable to the Helmholtz equation. There is actually a range of such Green’s
functions for the solution of antenna related problems. These arise to accommodate
changes in boundary conditions.
For the Helmholtz equation (Eqs. 4.5 and 4.6) we have demonstrated in the
preceding section (see Eq. 4.19) that an appropriate Green’s function is a solution
of:
where r denotes the point (x, y, z) while r0 denotes the point ðx0 ; y0 ; z0 Þ as shown in
Fig. 4.1.
In free space where the origin is naturally placed at zero (i.e. r0 ¼ 0) the Green’s
form of the scalar Helmholtz equation becomes:
This equation can be solved rigorously [4, 5] if the Sommerfeld radiation con-
dition (phasor form) at infinity is applied. It is:
4.2 Common Green’s Functions in Electromagnetic Radiation 85
jrGj [K
lim r @G ð4:22Þ
r!1 @r þ jkG ¼ 0
expðjkrÞ
GðrÞ ¼ ð4:23Þ
4pr
If the source is restored to the point r0 in Fig. 4.1 then the scalar Green’s function
can be expressed more generally as:
expðjk jr r0 jÞ
Gðr=r0 Þ ¼ ð4:24Þ
4pjr r0 j
In Eqs. (4.5) and (4.6) the right hand source term is −loJ, or −eoJm, respec-
tively, rather than the delta function source as in Eq. (4.21). The solutions to
Eqs. (4.5) and (4.6) can therefore be established by employing Eq. (4.15) to yield:
86 4 Computational Modelling Techniques for Slot Antennas
ZZZ
expðjk jr r0 jÞ 0
AðrÞ ¼ l0 Jðr ÞdV 0 ð4:25Þ
4pjr r0 j
V0
or
ZZZ
expðjk jr r0 jÞ
Am ðrÞ ¼ e0 Jm ðr0 ÞdV 0 ð4:26Þ
4pjr r0 j
V0
where V′ is the source volume. Not surprisingly Eqs. (4.25) and (4.26) replicate the
heuristically deduced forms in Chaps. 2 and 3 (see Eqs. 2.45 and 3.10). It should be
emphasised here that the advantage of the Green’s function route toward a solution
to a given second order differential equation is that it offers a very general method
for a wide range of problems. A successful outcome depends on the nature of the
integral and whether or not it can be performed over the source volume V′.
There are many situations, for example where the boundary conditions are more
easily formulated for electric field E than for the equivalent magnetic vector po-
tential. In this case it can be more convenient to proceed using Eq. (2.15) or (2.19).
Equation (2.15) can be written in the form:
To solve this equation using the Green’s function strategy we need an electric
dyadic Green’s function which is a solution of:
It is also chosen to satisfy the radiation condition. The quantity I is the unit
dyadic
and ^ax ; ^ay ; ^az ; are the unit vectors in the x, y and z directions respectively.
Some standard manipulation within the rules of vector algebra [5] demonstrates
that Ge is a solution of the vector Helmholtz equation:
rr
½r2 þ k2 Ge ðr=r0 Þ ¼ I þ 2 dðr r0 Þ ð4:30Þ
k
In other words, the electric dyadic Green’s function Ge can be deduced from the
scalar Green’s function G, which satisfies Eq. (4.20) by operating on G with dyadic
operator I þ rr
k 2 . Consequently, given that the scalar Green’s function is already
known (Eq. 4.24) then we must have:
4.2 Common Green’s Functions in Electromagnetic Radiation 87
0 rr expðjk jr r0 jÞ rr
Ge ðr=r Þ ¼ I þ 2 ¼ Iþ 2 G ð4:31Þ
k 4pjr r0 j k
With the Green’s function now known for the vector Helmholtz equation we can
deduce, from Eq. (4.15), that the solution for E has the form:
ZZZ
EðrÞ ¼ jxl Jðr0 Þ Ge ðr0 =rÞdV 0 ð4:32Þ
V0
where Gm ðr=r0 Þ is termed the magnetic dyadic Green’s function. Since Ge ðr=r0 Þ
and Gm ðr=r0 Þ must generate electric and magnetic fields, respectively, which obey
the Maxwell equation, it follows that
r Ge ¼ Gm ð4:34Þ
Hence, as with the electric case, we can express the free-space magnetic dyadic
Greens function in terms of the scalar Green’s function by combining Eqs. (4.31)
and (4.34) to yield:
Gm ¼ ½rGðr=r0 I ð4:35Þ
and, analogously to Eq. (4.32), H(r) can be determined from the equation:
ZZZ
HðrÞ ¼ Jðr0 Þ Gm ðr0 =rÞdV 0 ð4:36Þ
V0
The duality between E and H in the Maxwell equations dictates that similar
definitions exist for the magnetic current source Jm. In this case the Green’s
functions Ge and Gm are solutions of the equations:
rr
½r2 þ k2 Gm ðr=r0 Þ ¼ I þ 2 dðr r0 Þ ð4:37Þ
k
Ge ¼ ½rGðr=r0 I ð4:42Þ
where dV = Adl.
The relevant Helmholtz equations for this set up are presented in rows 1, 3 or 4 of
Table 4.1. The most common approach is to first solve for the vector potential A.
So applying Eq. (4.15) we obtain:
ZZZ
expðjk jr r0 jÞ 0
AðrÞ ¼ l0 Jðr ÞdV 0
4pjr r0 j
V0
Table 4.1 Established Electromagnetic Green’s Functions
Electromagnetic Helmholtz Equation Green’s Equation Green’s function
2 2 2 0 0 0
r A þ k02 A ¼ l0 J ½r þ k Gðr=r Þ ¼ dðr rÞ jrr jÞ
Gðr=r0 Þ ¼ expðjk
4pjrr0 j
0
r2 Am þ k02 Am ¼ e0 Jm ½r2 þ k 2 Gðr/r0 Þ ¼ dðr0 rÞ jrr jÞ
Gðr=r0 Þ ¼ expðjk
4pjrr0 j
0
½r r k 2 EðrÞ ¼ jxl Jðr0 Þ ½r2 þ k 2 Ge (r=r0 Þ ¼ I þ rr
k 2 dðr r Þ Ge ðr=r0 Þ ¼ I þ rr k2 G
½r r k 2 HðrÞ ¼ r Jðr0 Þ ½r r k 2 Gm ðr=r0 Þ ¼ r ½Idðr r0 Þ Gm ¼ ½rGðr=r0 I
0
½r r k 2 HðrÞ ¼ þ jxeJm ðr0 Þ ½r2 þ k 2 Gm ðr=r0 Þ ¼ I þ rrk 2 dðr r Þ Gm ðr=r0 Þ ¼ I þ rr k2 G
½r r k 2 EðrÞ ¼ r Jm ðr0 Þ
4.2 Common Green’s Functions in Electromagnetic Radiation
expðjkrÞ
AðrÞ ¼ l0 Idl ^
az ð4:44Þ
4pr
which compares, as it should, with the equivalent result in Chap. 2 (Eq. 2.48).
Further detailed analysis of the distributions of electric and magnetic fields in the
radiation pattern, including power and energy flows, would precisely parallel the
discussion in Sect. 2.5 of Chap. 2, and need not be repeated here.
expðjk jr r0 jÞ
Gðr=r0 Þ ¼ :
4pjr r0 j
In terms of G the electric dyadic Green’s function can be written in the form (see
Eq. (4.31)):
rr
Ge ðr=r0 Þ ¼ I þ 2 Gðr=r0 Þ ð4:45Þ
k
Equation (4.45) applies to the upper half of Fig. 4.3, for a point source in V′, as
if the conducting sheet were non-existent. To accommodate its presence imaging of
the source in the ‘mirror’, which the sheet represents, is the commonly applied
technique [4]. This results in the imaged source region Vi′ in the imaged Cartesian
system x′, y′, z′. This imaged source also exhibits a Green’s function contribution at
the field point P, namely:
1 @ @ @
Ge i ðr=r0i Þ ¼ ð^ax ^ax ^ay ^ay þ ^az ^az Þ þ ^
r ^ax ay ^
þ az Gðr=r0i Þ
k2 @x @y @z
ð4:47Þ
where
expðjk
r r0i
Þ
Gðr=r0i Þ ¼ ð4:48Þ
4pjr r0i j
92 4 Computational Modelling Techniques for Slot Antennas
Fig. 4.3 Field and source regions imaged in an infinite and perfectly conducting ground plane
and
h i1=2
r r0
¼ x x0 2 þ y y0 2 þ z z 0 2 ð4:49Þ
i i i i
So, the elements in Eq. (4.47) have to be shown explicitly because of the sign
differences associated with the imaged E-fields. In the second term of Eq. (4.47), a
similar situation arises with the del operation on the field coordinates of the
image-source.
Hence the Green’s function for the half-space Ge hs ðr=r0 Þ is merely the combi-
nation of Ge ðr=r0 Þ and Ge i ðr=r0 Þ. That is:
The fields in the upper half-space of Fig. 4.3 can now be generated by applying
Eq. (4.15), which for the electric field leads to:
ZZZ
EðrÞ ¼ jxl Jðr0 Þ Gehs ðr0 =rÞdV 0 ð4:52Þ
V0
where Jðr0 Þ includes the imaged current density and V 0 incorporates Vi0 .
Consequently, on adopting Eq. (4.50) we obtain:
ZZZ ZZZ
0 0 0
EðrÞ ¼ jxl Jðr Þ Ge ðr =rÞdV jxl J r0i Gei r0i =r dV 0 ð4:53Þ
V0 Vi0
If J were chosen to represent the current density for a z-directed short current
element (current Io and length l), located distance d in front of the perfectly con-
ducting sheet, application of Eq. (4.53) leads, after some mathematical manipulation
[4, 5], to the following scalar equation for the h-component of the electric field at P:
jxl I0 l ejkr1 ejkr2
Eh ¼ sin h1 þ sin h2 ð4:54Þ
4p r1 r2
where r1 ; r2 ; h1 and h2 are defined in Fig. 4.4. Similar scalar equations can be
generated for Er and E/, using:
94 4 Computational Modelling Techniques for Slot Antennas
Fig. 4.4 Current element above a perfectly conducting ground plane and perpendicular to it
a Converging rays at point P b rays converging at infinity
1=2
r1 ¼ r 2 þ d 2 2rd cos h1
1=2 ð4:56Þ
r2 ¼ r 2 þ d 2 2rd cosðp h2 Þ
For far field radiation simulations (r > > d, Fig. 4.4b), h1 = h = h2 and
Eqs. (4.56) reduce, using binomial series, to:
r1 r d cos h
ð4:57Þ
r2 r þ d cos h
In the far field the dcos h term is negligible by comparison to r which implies
that r1 and r2 are essentially parallel vectors as depicted in Fig. 4.4b. This, in turn,
means that from an amplitude variations perspective we can, with little error,
assume that:
r1 r2 r ð4:58Þ
jxl I0 l ejkr
Eh ¼ sin h½2 cos ðkd cos hÞ ð4:59Þ
4p r
for z > 0.
The term preceding the square brackets in Eq. (4.59) represents the Eh field
distribution for a single z-directed current element located at the origin of the
coordinate system in Fig. 4.4, which is modified by an array factor (term inside
square brackets—see Chap. 5) accounting for the fact that the source is at distance
d from the perfectly conducting ground plane. By employing the modified free
space Green’s function which accommodates the presence of an infinite perfectly
conducting ground plane—the half-space Green’s function—the array factor in
Eq. (4.59) results. Equation (4.59), normalised to a 0 dB maximum field strength,
is plotted in Fig. 4.5 for four d values, namely d = 0, k/4, k/2, k. The strong
influence of the ground plane on the radiation pattern is clearly illustrated.
In summary, where a modified Green’s function exists for a space containing a
radiating source then a solution for the electromagnetic fields radiated into the
designated volume can be constructed using Eq. (4.15). Unfortunately the range of
antenna configurations which can be solved in this way is limited by the not
insignificant difficulty of generating modified Green’s functions for other than
relatively symmetrical or regular spaces. For example Tai [4] has generated dyadic
Green’s functions for the following regular spaces and for volumes containing
simple geometrical obstacles:
(1) parallel plate waveguide,
(2) rectangular waveguide,
(3) cylindrical waveguide,
96 4 Computational Modelling Techniques for Slot Antennas
Fig. 4.5 Elevation plane field patterns in dB for a vertically orientated current element located at
different heights above the perfectly conducting ground plane
provided by the finite element method (see bibliography [C]) but at the expense of
computational speed arising from a very direct computational method which simply
discretizes the space to be solved and then performs perhaps millions of field
calculations over the finite element grid until stability is secured. This approach
represents a distinct disadvantage in the context of array modelling.
For linear differential equations, represented in general terms by Eq. (4.9), the
method of moments (MoM) solution to this type of problem can be enunciated
briefly as follows. For a linear operator L in an equation (e.g. equation 4.7) for
which g is known, the determination of the unknown f can readily be computed as
we have seen if the region supporting f possesses a Green’s function (see Eq. 4.15).
However if the region for which f is the solution is (say) perturbed away from a
regular shape and does not exhibit a Green’s function, then a solution is still
possible by resorting to a discretisation of f [6]. That is we expand f in a series of
functions f1, f2, f3, …. defined in the domain of L. Therefore we express f as
follows:
X
f ¼ an f n ð4:60Þ
n
where an are multipliers and are constants to be determined. The fn terms are
generally referred to as basis functions or expansion functions. For an exact solution
to the relevant differential equation, the summation in Eq. (4.60) would normally be
infinite and the fn form a complete set (Appendix C) of basis functions. Otherwise
the summation is truncated, within defined mathematical rules, to minimise com-
putational effort.
For a linear operator, substitution of Eq. (4.45) into Eq. (4.9) yields:
X
an Lðfn Þ ¼ g ð4:61Þ
n
For a linear L, the relationship between f and g must also be linear in which case
an inner product (denoted by hf ; g i) must exist. For example a common, but not
unique, formulation is:
Z1
hf ; gi ¼ f ðxÞgðxÞdx ð4:62Þ
0
98 4 Computational Modelling Techniques for Slot Antennas
where
2 3
hw1 ; Lf1 i hw1 ; Lf2 i
6 hw2 ; Lf1 i hw2 ; Lf2 i 7
Im;n ¼ 6
4 5
7 ð4:65Þ
2 3
a1
6 a2 7
6 7
6 a3 7
½ an ¼ 6
6 : 7
7 ð4:66Þ
6 7
4 : 5
:
2 3
hw1 ; gi
6 hw2 ; gi 7
6 7
6 hw3 ; gi 7
½ gm ¼ 6
6 : 7
7 ð4:67Þ
6 7
4 : 5
:
Generally the matrix (4.59) will be non-singular (i.e. have a non-zero determi-
nant), and hence will be invertible. We can therefore set up an equation for the
unknown an constants which takes the form:
½an ¼ I1
nm ½gm ð4:68Þ
Once the an constants have been determined the solution for f is given by
Eq. (4.45). The result may be exact, or approximate, depending on the original choice
of fn and wn. It is relevant to note here that the common choice of equating fn to wn
ðfn ¼ wn Þ is referred to as the Galerkin method [1]. Commercial software packages
4.3 Concise Formulation of the Moment Method 99
which have been constructed around the moment method are listed in references [7–
11]. An antenna format which is particularly amenable to the MoM, and is used as a
textbook example, is the dipole antenna. For this antenna, the method yields rapid and
accurate solutions by employing well suited sinusoidal basis functions [1].
The moment method, which has been comprehensively evaluated in this Chapter, is
shown to be an elegant mathematical technique and that it is particularly suited to
the solution of a class of electromagnetic boundary value problem created by
antennas and radiation. Importantly, in computational terms, the method provides a
source of ‘high speed’ electromagnetic solvers, by offering a problem-specific
approach which closely reflects the underlying physics. Furthermore, the rapid
growth in the application of array antennas to new roles in mobile and satellite
communications, has created design requirements for the electromagnetic mod-
elling of arrays, and this calls for the embedding of boundary value solvers within
array simulations. For these embedded MoM solvers computational speed is
essential, since they may need to be accessed many thousands of times during the
course of a design calculation.
Expressed in words, it can be averred that the MoM provides a solution to the
problem of determining the electromagnetic fields generated by a known source
located in a volume bounded by accurately defined surfaces. The mathematical pro-
cesses required to achieve this outcome are summarised in this chapter. It is
demonstrated that, in essence, the method relies on knowledge of a simpler electro-
magnetic wave solution for a comparable bounded space when excited by a unit
intensity point source within it. It is shown that mathematically such a source is
represented by a delta function. The solution produced is termed the Green’s function
for the specified region and Green’s functions for a range of bounded spaces, which
are relevant to antenna problems, are tabulated for easy referencing. The total field
solution for a defined source distribution can then be formed by integrating the point
source solution over all source locations and over all time intervals over which it acts.
Further, it is shown that the linearity of the resultant integral equations permits dis-
cretisation of the relevant functions to form matrix representations which yield readily
to computer processing. This procedure is embellished further in Chap. 5.
References
1. R.F. Harrington, Field Computation by Moment Methods (MacMillan, New York, 1968)
2. P.H. Morse, H. Feshbach, Methods of Theoretical Physics (McGraw-Hill Book Co., New
York, 1953)
100 4 Computational Modelling Techniques for Slot Antennas
Bibliography
1. R.E. Collin, Field Theory of Guided Waves (McGraw-Hill Book Co., New York, 1960)
2. R.F. Harrington, Time-Harmonic Electromagnetic Fields (McGraw-Hill Book Co., New
York, 1961)
3. J.-M. Jin, Finite Element Methods in Electromagnetics (Wiley, New York, 2014)
Chapter 5
Moment Method Models of Compact
Slot Antennas
5.1 Introduction
In Chap. 4 it has already been noted that for open, or closed, spaces which support
time varying electromagnetic fields, excited by sources which can readily be
expressed mathematically, then the electromagnetic boundary value problem pre-
sented can be solved without undue difficulty if the Green’s function for the space in
question is available. Unfortunately, this is seldom the reality, particularly in antenna
developments. The typical problem which is encountered tends to comprise two or
more interconnected regions each capable of supporting electromagnetic waves.
A solution to such a problem requires that the fields in each region satisfy the local
boundary conditions and simultaneously the boundary conditions at interfaces. An
example of a problem of this type is illustrated in Fig. 5.1a. It comprises a narrow
and inclined slot located in the upper broadwall of the rectangular waveguide (in-
ternal dimension; width a; height b and wall thickness T). The waveguide walls are
generally assumed to be perfectly conducting. Figure 5.1b shows the slot detail
(length L; width W) and the local slot coordinate system (n, η, f).
Electromagnetically the antenna structure represented by Fig. 5.1a comprises
two regions; the free-space exterior to the waveguide (above the slot and usually
modelled as a half-space), and the waveguide region (usually modelled as infinitely
long in the z-direction). These regions are coupled electromagnetically through the
slot. The computationally most efficient, and accurate, procedure for solving this
problem, which is not untypical of a wide range of compact radiator geometries, is
the moment method. Other such geometries include slots fed from transmission
lines such as stripline, microstrip line, coplanar waveguide, but also slot fed pat-
ches, slot modified patches, cavity backed slots and combinations of all of these.
Many will be addressed in later chapters.
Other powerful electromagnetic solvers based on the finite element and finite
difference techniques can undoubtedly also be used to simulate compact slot
radiators. But these techniques, while they are in principle unlimited in their
Fig. 5.1 a Waveguide fed compact slot radiator and b slot detail
accessed and adjusted many times. This means a slot simulation has to be entered
repeatedly and clearly the computational speed offered by the moment method
generates design advantages in this scenario. Recent developments in the use of
artificial neural networks to facilitate the optimisation of compact slot antennas [2]
involves ‘training’ and ‘testing’ of the neural system by repeated reference to the
compact slot-antenna model. So again we encounter a design situation where we
require efficient electromagnetic simulations to avoid excessive, or time consuming,
computational effort.
at infinite distance from the radiator. Thirdly, it is essential that at the inner and
outer surfaces of the slot (S and S′) that field continuity conditions are applied. The
first and second of these three conditions are immediately enforced by directing the
analysis toward a solution of the relevant inhomogeneous Helmholtz equations for
E and H (Eqs. 2.19 and 2.20) and by forming the relevant Green’s functions for the
coupled regions (a) and (c). Any solution constructed around the Green’s functions
ensures that the eventual field solutions emanating from the analysis satisfy both the
electromagnetic boundary conditions on all conducting surfaces, and the radiation
condition at infinity in region (c). Consequently, the key to formulating a set of
equations which correctly represent the radiation problem depicted in Fig. 5.1a lies
in proper establishment of the field boundary conditions at those slot surfaces which
connect regions (a), (b) and (c).
The analysis begins by identifying the electromagnetic fields which must exist in
each region in Fig. 5.1a. We have already noted that if the slot is ‘closed’ by a
fictitious perfectly conducting insert a single incident wave (TE10 mode) continues
to propagate through the waveguide. The fields of this wave are defined as Ei, Hi.
When the slot is ‘open’, and capable of interfering with the wall currents of the
TE10 mode, it becomes a discontinuity in the waveguide scattering fields back into
it. These fields we will term Ea, Ha. So, in region (a) the total fields are Ei + Ea,
Hi + Ha. The discontinuity fields on the surface S also scatters fields Eb1, Hb1 into
the slot cavity thus generating further components Eb2, Hb2 on surface S′. These are
the source of radiation into region (c) expressed through Ec, Hc. In order to
complete the electromagnetic modelling process on the structure of Fig. 5.1a it is
necessary to ensure that the above fields satisfy the boundary conditions on both
surfaces S and S′. For aperture problems (see Chap. 3) the discontinuity fields on
these surfaces are the electric field components tangential to the relevant surface.
Consequently, the boundary conditions can be formulated as:
^ Ea ¼ ^
n n Eb1 on S ð5:1Þ
^ Eb2 ¼ ^
n n Ec on S0 ð5:2Þ
Fig. 5.2 Field equivalence representation of slot radiator in thick walled waveguide
Jm ¼ n^ E on S in region a
Jm ¼ ^ n E on S in region b
J m ¼ n E0 on S0 in region b
0
^0
J0 m ¼ ^
n0 E0 on S0 in region c
This theoretical model is illustrated in Fig. 5.2. Each region is now separately
excited by a magnetic current and can be represented mathematically by the inho-
mogeneous Helmholtz equation rewritten below [see Eq. (2.19) and Table 4.1].
with zero tangential electric field on the conducting surfaces, which can be
expressed through the equation:
^ r HðrÞ ¼ 0
n ð5:4Þ
on the metal surfaces. The appropriate Green’s equation for each region (see
Table 4.1) is:
106 5 Moment Method Models of Compact Slot Antennas
m ðr=r0 Þ ¼ Idðr r0 Þ
½r r k2 G ð5:5Þ
n m ðr=r0 Þ ¼ 0
^rG ð5:7Þ
Once the as yet unknown Jm (or J0 m ) is evaluated the magnetic field distribution
in each region is calculable using an equation of the form of Eq. (5.8) (see Chap. 4)
which is expressed in terms of an arbitrary surface S0 and location vector r0:
ZZ
HðrÞ ¼ j xe m ðr=r0 Þ Jm ðr0 ÞdS0
G ð5:8Þ
So
While, on surface S′ which supports the magnetic current density J0 m the con-
tinuity condition produces:
ZZ
^
0 ¼ jxe0 n G mb ðr=r0 Þ þ G
mc ðr=r0 Þ J0 m ðr0 ÞdS0
S0
ZZ ð5:10Þ
^
þ jxe0 n G mb ðr=r0 Þ Jm ðr0 ÞdS0 for r on S0
S0
in reference [5]. The relative complexity of Eqs. (5.9) and (5.10) make direct solution
unfeasible. Consequently, a numerical method designed for the task has to be adopted;
such a method is the moment method developed in the previous chapter [6, 7].
Electromagnetic radiation from a narrow rectangular slot fed from waveguide is,
as we have already intimated, not a new problem. It has been studied widely for use
as a compact radiating element in linear and planar array antennas. In today’s world
of burgeoning mobile communications, ‘compactness’ has become more critical
than array applications, and the waveguide feed arrangement has inevitably been
superseded by spatially constrained alternatives. These will be addressed in a later
Chapter. Nevertheless, the waveguide fed slot provides a cogent example of how
modelling by means of the moment method is developed in slot radiation problems.
Some of the original theoretic work on slot antennas, most notably by Oliner [3]
and Das [4], has demonstrated that the electric field distribution along the slot
length is well modelled by a single trigonometric function. Examples, generally
evolved through the application of variational techniques, are shown in Fig. 5.3.
The indicated differences are largely attributed to the focus on near resonant slots by
Oliner, and on marginally non-resonant slots by Das. This precedent suggests that
there could be potentially significant advantage, from a computational speed per-
spective, in adopting basis functions for the discretized slot field with trigonometric
properties. Mathematically this leads to the following constructions (see Eq. 4.45):
X
N
^E¼n
n ^ as fs ^ag on S ð5:11Þ
s¼1
Fig. 5.3 Slot field approximations for a narrow (w/L < 0.1), near resonant slot (L/k = 0.5) in the
broadwall (T = 0) of rectangular waveguide, as adopted in early radiation computations [Solid line
with diamond symbols = Oliner: Dashed line with square symbols = Das]
108 5 Moment Method Models of Compact Slot Antennas
X
N
n0 E0 ¼ ^
^ n0 a0s fs ^ag on S0 ð5:12Þ
s¼1
spðn þ L=2Þ
fs ðnÞ ¼ sin s ¼ 1; 2; 3; . . .; N ð5:13Þ
L
Following the procedures outlined in Chap. 4, Sect. 4.3, which are developed in
more detail in [6], we choose testing functions [Eq. (5.14)] which mirror the basis
functions, namely the Galerkin method
fi ðnÞ^ag i ¼ 1; 2; 3; . . .. . .; N ð5:14Þ
where:
ZZ ZZ
Ais ¼ jxe0 fi ^ag f^
n G ma ðr=r0 Þ þ G
mb ðr=r0 Þ fs ^
an dS0 gdS ð5:16aÞ
S S
ZZ ZZ
Bis ¼ jxe0 fi0 ^ag f^
n0 G mc ðr=r0 Þ f 0 ^
mb ðr=r0 Þ þ G
s an dS0 gdS ð5:16bÞ
S0 S0
ZZ ZZ
Cis ¼ jxe0 fi0 ^ag f^
n mb ðr=r0 Þ fs ^
G an dS0 gdS ð5:16cÞ
S S0
ZZ ZZ
Dis ¼ jxe0 n0
fi ^ag f^ mb ðr=r0 Þ f 0 ^
G s an dS0 gdS ð5:16dÞ
S0 S
ZZ
hi ¼ fi ^ag ½n
^ Hi ðrÞdS ð5:16eÞ
S
The matrix entries Ais, Bis, Cis, Dis given mathematical expression in Eqs. (5.16a,
5.16b, 5.16c, 5.16d, 5.16e) are essentially electromagnetic coupling terms repre-
senting to the mutual coupling which occurs between the assumed current distri-
butions (fi,s) on the interior and external surfaces of the radiating aperture. In the
5.2 Moment Method Solution 109
particular case of the conventional slot (in Fig. 5.1) they can be ascribed the fol-
lowing specific meanings:
Ais represents self-coupling between magnetic current elements on the inner slot
surface S.
Bis represents self-coupling between magnetic current elements on the outer slot
surface S′.
Cis represents mutual coupling to the inner slot surface (S) due to magnetic
current elements on the outer surface (S′).
Dis represents mutual coupling to the outer slot surface (S′) due to magnetic
current elements on the inner surface (S).
Once the integrals in Eqs. (5.16a, 5.16b, 5.16c, 5.16d, 5.16e) have been eval-
uated, which is mathematically routine but tedious [8, 9], the matrix Eq. (5.15) is
amenable to solution by a range of matrix manipulation tools. These can be found
in standard mathematical textbooks [10, 11] and in convenient to use computer
software packages [12]. The outcome of the process is plausible values for s, fs, a′s
and as. These can then be used to compute the scattering parameters for the compact
radiator as outlined below.
The general broadwall inclined slot radiator depicted in Fig. 5.1a can be resolved
into a longitudinal shunt slot when h = 0°, and into a transverse series slot when
h = 90°. In fact mathematically it is more sensible to do this than attempt to solve
the inclined slot example directly [9]. In practice the inclined broadwall slot has
very few applications and most studies have been directed at modelling the series
slot and the shunt slot. Below we revisit some validating results for the shunt slot,
which is by far the most commonly used in array applications.
A cursory examination of the matrix entries in Eq. (5.16a, 5.16b, 5.16c, 5.16d,
5.16e) quickly establishes that the controlling matrix in Eq. (5.15) grows in size as
the square of the number of basis functions s, i. The more basis functions selected,
therefore, the slower will be the progress toward a computed solution. For con-
ventional slots, at least, the advantage of the use of trigonometric basis functions is
that accurate solutions are available for no more than three basis functions. This is
demonstrated in Fig. 5.4 where the slot field distribution for a longitudinal slot in
rectangular waveguide as computed by the moment method employing three basis
functions (solid curve—round symbols), is compared with the predictions of Oliner
(diamond symbols) and Das (square symbols) for the same geometry and discussed
earlier. The moment method solution converges on a slot field distribution, which is
approximately cosinusoidal as predicted by Oliner, but with a rather broader, less
peaky, shape than the pure cosine. This modified shape is also predicted by Vu
Khac (solid line—triangular symbols) employing a moment method constructed
around thirty ‘pulse’ type basis functions [13]. The Das distribution with its sharp
discontinuous peak can no longer be justified.
110 5 Moment Method Models of Compact Slot Antennas
Fig. 5.4 E-field distributions for a shunt slot radiator in the broadwall of WG16 waveguide as
predicted by the moment method (freq = 9.375 GHz, L = 0.361 k, x1 = 6.43 mm, w = 1.58 mm,
T = 1.27 mm)
The key to determining slot scattering lies with Eq. (5.8) which provides the link to
the fields generated by the slot discontinuity once Jm and J0 m are known. These become
calculable [Eqs. (5.11) and (5.12)] once as and as′ are evaluated through the moment
method computation. Given that in employing the equivalence theorem, Jm and J0 m
are positioned at the slot locations, but on ‘unbroken’ waveguide walls, the required
waveguide Greens function for calculating internal scattering is the ‘sum of modes’
formulation [9] (see Appendix C). However, real power in the transmission and
reflection signals within the waveguide is carried by the TE10 mode. Consequently we
can write for the transmitted magnetic field component of the TE10 mode H10t :
ZZ
1X N
þ
H10t ðrÞ ¼ H10i as H10 ðrÞ H
10 ðrÞ fs ^
an dS0 ð5:17Þ
2 s¼1
S
where H 10 are the forward and backward propagating components of the magnetic
field for the TE10 mode in the rectangular waveguide, while H10i denotes the
incident mode in the waveguide. The reflected field component H10r can be cal-
culated from:
ZZ
1X N
þ
H10r ðrÞ ¼ as H10 ðrÞ H10 ðrÞ fs ^
an dS0 ð5:18Þ
2 s¼1
S
5.3 Scattering for Longitudinal Slot in Waveguide 111
The moment method program formulated using trigonometric basis functions has
been tested exhaustively [8, 9] to assess its sensitivity to the number of basis
functions employed. For the computation of slot radiation, and slot scattering
parameters, the sensitivity is demonstrated to be low, and it is safe to say that one
basis function is generally satisfactory for modelling conventional waveguide slot
radiators. This is demonstrated in Fig. 5.5 where the relative power radiated, as a
function of slot length, for a conventional longitudinal slot in WG16 waveguide, as
predicted by the moment method, is presented. Computations were performed for
up to five basis functions. On the scale adopted in the graph there was no dis-
cernible difference between three or more basis functions, while the single basis
Fig. 5.5 Relative power radiated by a broadwall waveguide slot as a function of slot length L,
with the number of basis functions as parameter (Freq = 10 GHz, a = 22.86 mm, b = 10.16 mm,
x1 = 2.00 mm, w = 1.5 mm, T = 0 mm)
112 5 Moment Method Models of Compact Slot Antennas
function prediction is marginally lower than the three-basis case only for below
resonance slot lengths. This is because close to resonance the slot internal field
distribution is well represented by a simple cosine, and hence one basis function is
all that is needed to generate accurate results. Needless to say this has had major
significance for the speed and efficiency of array antenna simulation incorporating
the method of moments. For example, in a multi-slot array simulation where the
MoM is embedded into an iterative program designed to perform pattern synthesis
[1], the MoM for each slot has to be accessed repeatedly, and computational speed
becomes critical.
To emphasise the point further, the trigonometric formulation with three basis
functions has been compared with the Vu Khac moment method [13] employing 30
pulse basis functions. The comparison was carried out on a longitudinal slot in
WG16 waveguide at a frequency of 9.375 GHz. The WG16 had a wall thickness of
1.27 mm and the shunt slot, with w = 1.58 mm, was off-set from the sidewall of the
waveguide by 1.43 mm. The results are presented in Figs. 5.6 and 5.7 which
display reflection coefficient magnitude |S11| and transmission coefficient magnitude
|S12| as a function of slot length L.
It is clear from Figs. 5.6 and 5.7 that the trigonometric moment method
employing three basis functions computes values for |S11| and |S12| which are in
very good agreement—largely within the computational errors—with the pulse
basis moment method. Deviations outwith the computational error level only occur
Fig. 5.6 Reflection coefficient magnitude (|s11|) as a function of slot length as predicted by
trigonometric basis (Lyon) and by pulse basis (Vu Khac) moment methods (WG16,
Freq = 9.375 GHz, T = 1.27 mm, x1 = 1.43 mm, w = 1.58 mm)
5.3 Scattering for Longitudinal Slot in Waveguide 113
Fig. 5.7 Transmission coefficient magnitude (|s12|) as a function of slot length as predicted by
trigonometric basis (Lyon) and by pulse basis (Vu Khac) moment methods (WG16,
Freq = 9.375 GHz, T = 1.27 mm, x1 = 1.43 mm, w = 1.58 mm)
well above the resonant slot length (L * 15.5 mm). At such slot lengths, reso-
nances of higher order than the half-cosine shape depicted in Fig. 5.4 begin to
influence the calculations and it becomes necessary to include more than three basis
functions used to generate Figs. 5.6 and 5.7. Phase computations, namely/S11 and
/S12, have also been compared [8], and yield largely similar levels of agreement.
To complete the validation exercise on the trigonometric moment method,
computed results have also been compared with measured characteristics for a
typical longitudinal slot radiator in WG16 waveguide. Typical results for S12
magnitude and phase are presented in Fig. 5.8a, b. The measurements were pro-
cured by the application of a cavity technique [8]. This involves resonating the
waveguide in a section containing the slot, which then becomes a detuning dis-
continuity, then measuring changes in resonant frequency and quality factor (Q) as
a function of slot length. While the level of agreement is generally good, and
certainly within the measurement errors, the displacement of the measured resonant
frequency to a longer slot length than predicted by theory, was initially puzzling.
But this was found to be almost wholly attributable to the fact that measurements
were performed on round-ended slots for practical fabrication reasons. The theory
assumed square-ended slots.
114 5 Moment Method Models of Compact Slot Antennas
A more complex slot example, comprising a polarisation agile, T-shaped, slot fed
from bifurcated waveguide has been included here to show that with insightful
application of boundary conditions, and with particular attention to the slot field
distribution and the basis function choice, the trigonometric moment method is of
sufficient flexibility to establish accurate modelling tools for a wide range of
compact slot geometries. The arrangement, which is depicted in Fig. 5.9, was
devised originally to secure polarisation agility by independently exciting a pair of
orthogonally aligned slots (see Chap. 1, Fig. 1.22) in dual-mode waveguide. Thus
the slot pair is arranged to radiate axial or transverse linear polarisations if excited
separately using mode switching, or if excited simultaneously and in phase
quadrature, circularly polarised radiation is produced [14]. While the longitudinal,
or shunt, slot portion of the orthogonal slot pair has been thoroughly examined in
the literature, as outlined in Sect. 5.3 above, a transverse slot straddling the central
Fig. 5.9 a Transverse T-slot radiator in dual-mode rectangular waveguide. b Transverse element
of T-slot showing coordinate system
116 5 Moment Method Models of Compact Slot Antennas
wall of a bifurcated wall is relatively novel [15] and requires advanced electro-
magnetic modelling to establish its radiation characteristics. We revisit this case
here to demonstrate the potential that exists in the moment method for addressing
unconventional applications such as those encountered in evolving novel compact
slot antenna geometries examined in ensuing chapters.
The matrix entries Ais, Bis, Cis, Dis defined in Eqs. (5.16a, 5.16b, 5.16c, 5.16d,
5.16e), have been evolved for a moment method formulation of the general case of
electromagnetic power transfer from a transmission system (waveguide) through a
slot cavity to a half-space where radiation occurs (Fig. 5.2). In the previous section
(Sect. 5.3) it has been applied comprehensively to the well established example of a
shunt slot in waveguide. We will now apply these Eqs. (5.16a, 5.16b, 5.16c, 5.16d,
5.16e) to the switchable T-slot radiator on dual mode waveguide, as depicted in
Fig. 5.9.
From a mathematical perspective the fundamental difference between the sym-
metrically located T-shaped slot in bifurcated waveguide (Fig. 5.9a) and the con-
ventional slot in regular waveguide shown in Fig. 5.1a, lies with the clearance slot
which attracts additional mutual coupling between the inner surface S of the
transverse slot and the septum slot surfaces Sv. The computation of these additional
coupling terms requires that a suitable trial magnetic field distribution can be for-
mulated for the T-slot which encompasses all three surfaces S, S’ and Sv. Once this
is accomplished the preliminary formulation of the moment method solution for the
T-slot proceeds much as outlined in Sect. 5.3.
Consider, firstly, the tangential electric field distribution which the waveguide
modal fields induce on the inner surface S of the transverse component of the T-slot
(Fig. 5.9b). The functional form of this field will obviously be influenced by the
septum slot and its depth L3. For example if L3 0 then the top of the waveguide
bifurcation will suppress any induced electric field (Eη) near the centre of the inner
surface S of the transverse slot, resulting in a field distribution not unlike that
depicted in Fig. 5.10a. That is, it is plausibly presumed that when the septum slot
approaches zero length the Eη field distribution on the inner surface S of the
transverse slot is essentially the same as that of two slots, of length L/2, positioned
end-to-end. Note that this is not necessarily true for the outer slot surface S’ par-
ticularly if the waveguide wall thickness is not infinitesimally small. On the other
hand, when the septum slot depth is finite, it becomes less easy to predict possible
field distributions on the surfaces S and Sv. However, from qualitative considera-
tions it seems evident that the clearance slot will be parasitically excited by the
radiating slot fields, and thus the Eη fields at the centre of the transverse slot will no
longer be suppressed on the inner surface S. So, it seems reasonable to suggest that
the Eη field on S will exhibit a functional form not too dissimilar to Fig. 5.10b.
With the above observations in mind we can proceed to the construction of
suitable basis functions for the fields on S and on Sv by noting the analogy between
the E-field forms in the T-slot to the voltage profiles on an equivalent T-network of
transmission lines [16]. Suppose that the transmission line depicted in Fig. 5.11a
represents a conventional slot positioned wholly within the broadwall of a regular
rectangular waveguide feed. The input impedance at the right hand extremity of the
5.4 T-Shaped Slot in Bifurcated Waveguide 117
Fig. 5.10 Suggested electric field distributions in the transverse component of the T-slot a L3 = 0,
b 0 < L3 < L/2
normalised line, for which Z0 = 1, can, from transmission line theory, be expressed
as:
tan bL ¼ 0 ð5:18Þ
118 5 Moment Method Models of Compact Slot Antennas
Fig. 5.11 Transmission line based equivalent circuits for waveguide slot radiators a conventional
slot, b T-slot in bifurcated waveguide
The implication of this is that the line voltage (or current) forms a standing wave
with the functionality
npx
V ¼ V0 sin ð5:20Þ
L
transverse slot portions on either side of the central bifurcation and a line, L3 in
length, representing the clearance notch. The three lines are connected in parallel at
A, and all three are terminated in short circuits. From transmission line theory the
input impedance at A, for the paralleled L/2 lines, is given by:
j bL
ZA ¼ tan ð5:21Þ
2 2
for normalised line impedances (Z0 = 1). Hence, with ZA as the load, the impedance
at B can be expressed as:
1
tanðbL=2Þ þ j tanðbL3 Þ
ZB ¼ 2
ð5:22Þ
1 12 tanðbL=2Þ tanðbL3 Þ
L L
fs ¼ sin b1 ð j xjÞ ¼ sin b1 ð jnjÞ ð5:24Þ
2 2
and preserving continuity with the conventional slot analysis we choose to employ
testing functions (fi) for the S/2 surfaces such that ðfi ¼ fs Þ. On the outer surface of
the transverse slot (S’), where the waveguide bifurcating wall is less influential with
T > 0, the basis and testing ðfs0 ¼ fi0 Þ functions are unchanged from those used in
Sect. 5.3. That is:
spðn L2Þ
fs0 ¼ sin ð5:25Þ
L
L
V ¼ sin b1 j xj ð5:26Þ
2
mirroring the trial field or basis function on S [Eq. (5.24)], while on the clearance or
septum slot surfaces Sv Fig. 5.11b suggests that here the basis function form will be:
Since these voltages (or equivalent E-fields in slot terms) must equate at the
junction A then we can write:
b1 L
sin ¼ V 0 sin b1 L3 ð5:28Þ
2
Hence, by analogy, an appropriate basis function for the field forms on the Sv
surfaces is:
sinðb21 LÞ
fsv ¼ sin b1 ðy b þ L3 Þ ð5:29Þ
sinðb1 L3 Þ
With the trial field expressions now defined through Eqs. (5.24), (5.25) and
(5.29) for the relevant surfaces of the T-slot, the analysis can proceed in a manner
not too different from the development in Sect. 5.2. Commencing with Eqs. (5.9)
and (5.10), which are generally applicable to slot radiators, a matrix equation for the
T-slot can be formed with the following entries [see Eq. (5.15)]:
ZZ ZZ
Ais ¼ 2jxe0 fi ^ag f^
n G ma ðr=r0 Þ fs ^
an dS0 gdS
S=2 S=2
ZZ ZZ
jxe0 fi ^ag f^
n mb ðr=r0 Þ fs ^
G an dS0 gdS ð5:30aÞ
S
ZZ ZZ
S
2jxe0 fi ^ag f^
n ma ðr=r0 Þ fsv ^
G an dS0 gdS
S=2 Sv
ZZ ZZ
Bis ¼ jxe0 fi0 ^ag f^0
n G mc ðr=r0 Þ f 0 ^
mb ðr=r0 Þ þ G
i an dS0 gdS ð5:30bÞ
S0 S0
ZZ ZZ
Cis ¼ jxe0 fi0 ^ag f^
n mb ðr=r0 Þ fs ^
G an dS0 gdS ð5:30cÞ
S S0
ZZ ZZ
Dis ¼ jxe0 fi ^ag f^0
n mb ðr=r0 Þ f 0 ^
G s an dS0 gdS ð5:30dÞ
S0 S
5.4 T-Shaped Slot in Bifurcated Waveguide 121
ZZ
hi ¼ fi ^ag ½n
^ Hi ðrÞdS ð5:30eÞ
S
Except for the Ais term, the above integral expressions are formally identical to
the corresponding conventional slot versions. They differ only in the assumed
functional form for the basis function fi and the testing function fs. On the other hand,
Ais contains an additional term, relative to the conventional analysis, to accommo-
date the mutual coupling which can occur between the inner surface of the transverse
slot and the septum slot surfaces. The Green’s functions G ma ; G
mb ; and G
mc are
essentially the same as for the conventional slot analysis, in Sect. 5.2, except that for
G ma which must be treated in two parts with in-phase dominant modes on either side
of the bifurcation. This is done by coordinate system adjustments. The previously
defined rectangular slot Green’s function G mb now applies to the transverse com-
ponent of the T-slot. It is important to note that in the computation of Bis, Cis, and Dis
higher order slot cavity modes will require to be included, more so than in the
conventional slot solution, to accommodate basis functions such as fs in Eq. (5.24)
which will tend to generate such modes. The integral expressions delineated in
Eqs. (5.16a, 5.16b, 5.16c, 5.16d, 5.16e) have been fully evaluated in references [8]
and [9] and some typical results have been presented in Sect. 5.3. Needless to say the
evaluation of Eqs. (5.30a, 5.30b, 5.30c, 5.30d, 5.30e) are not too different and are
detailed in [8]. It should be noted here that the moment method computations have
generally been performed for the s = 1 basis function on the inner slot surface S, and
for the first basis function on the outer surface S’ except where Lv is short, in which
case the s = 3 basis function is introduced.
Below we present a selection of representative results to demonstrate the
effectiveness of the moment method in modelling more complex and compact
antenna designs. As with the longitudinal slot example in Sect. 5.2, once the matrix
equation, with the entries detailed in Eq. (5.30a, 5.30b, 5.30c, 5.30d, 5.30e), has
been solved by moment method interaction, plausible values for s; as ; fs ; fs0 ; and fv
should become available. In a process little different to that adopted for calculating
the longitudinal slot scattering values, those for the transverse T-slot can be simi-
larly calculated. The scattering formulae are detailed in Appendix C.
The proof that a new engineering analysis, or that an analysis of a novel engineering
advance, is mathematically and technically sound lies either by showing compar-
ison with the results of other researchers, or by comparing the theory with mea-
surements on carefully prepared models, as outlined in Sect. 5.3. Here we follow
this measurement route, since at the time when the T-slot in bifurcated waveguide
was proposed [15], no earlier results by other investigators were available.
122 5 Moment Method Models of Compact Slot Antennas
Fig. 5.12 Resonant cavity set-up for T-slot scattering in dual mode waveguide
rather as expected from perusal of the conventional slot case, that radiation is weak
at low frequencies which are well below the T-slot resonance, and then grows
steadily in strength as the source frequency approaches this resonance—predicted
124 5 Moment Method Models of Compact Slot Antennas
Fig. 5.13 Relative power radiated (Prad/Pinc) versus frequency for a transverse T-slot in bifurcated
waveguide in even-mode (WG16 wall thickness T = 1.32 mm; slot width w = 1.5 mm; slot length
L = 16 mm; septum slot L3 = 5.48 mm)
Fig. 5.14 T-slot transmission coefficient magnitude a and phase b as a function of slot length L
(Even mode excitation; WG16; T = 1.32 mm; L3 = 5.48 mm; bifurcation thickness = 1.5 mm;
frequency = 9.89 GHz)
by the moment method to be at 11.1 GHz for the given dimensions. It is noticeable
that agreement between theory and measurement is good, and well within the
estimates of experimental error at frequencies below *10.5 GHz where the T-slot
is distinctly non-resonant. Close to the slot resonance, around 11 GHz, the mea-
sured results begin to deviate significantly from the moment method predictions.
This can be attributed to a basic limitation in the cavity based measurement tech-
nique. For weakly radiating non-resonant slots the amount of power extracted from
5.4 T-Shaped Slot in Bifurcated Waveguide 125
Fig. 5.15 T-slot reflection coefficient magnitude a and phase b as a function of slot length L
(Even mode excitation; WG16; T = 1.32 mm; L3 = 5.48 mm; bifurcation thickness = 1.5 mm;
frequency = 9.89 GHz)
Fig. 5.16 Relative power radiated in dB as a function of clearance slot depth (L3) for a
symmetrically located T-slot—even mode excitation (WG16; T = 1.32 mm; L3 = 5.48 mm;
bifurcation thickness = 1.5 mm; frequency = 9.89 GHz)
The above comments are reinforced in Figs. 5.14 and 5.15, which present,
respectively, plots of transmission coefficient amplitude and phase and reflection
coefficient amplitude and phase as functions of slot length. The plots are typical of a
series resonant slot in waveguide. While the computed transmission coefficient
magnitude falls from approximately unity to almost 0.94 at resonance, the reflection
coefficient magnitude mirrors this, rising from almost zero for non-resonant lengths
to almost 0.06 at resonance. The phase shift predictions establish that the slot, in
transmission line terms, could be modelled by an RLC circuit in series with the line,
with L C for short slots. At resonance L = C and the slot behaves as a series
resistance in the line.
The comparisons between theory and measurement provided in Figs. 5.14 and
5.15 fully confirm the observation made with respect to Fig. 5.13 with ‘good’ levels
of agreement for non-resonant slots (within experimental error). It is perhaps per-
tinent to note that while the measured points in Figs. 5.14a and 5.15a offer poor
confirmation of the predicted resonant behaviour of the slot, this is less true of the
phase measurements. These indicate an experimental resonant frequency of *18,
0.5 GHz below the theoretical value, and well within the measurement error
estimate.
Arguably, a feature of primary importance for potential users of the T-slot
radiator is the degree to which its radiation level is influenced by the clearance slot
length (L3). A typical result, depicting relative power radiated as a function of
septum slot length, is presented in Fig. 5.16. Not unexpectedly for L3 approaching
zero length the radiation is very weak at about −36 dB. It is then predicted to rise
rapidly as L3 is increased, to almost −20 dB when L3 = 7 mm. The measured results
confirm the theoretical prediction to within the experimental error. The experimental
errors arise from several sources one of which is cavity damping for near resonant
slots. Others are attributed to the square ended slots used in the theory. These were
difficult to replicate experimentally, and the slot used were generally round-ended.
Roughness of the machined surfaces can have a deleterious effect on measurements
as can imprecise positioning of the slots within the dual-mode waveguide test piece.
Generally the degree of agreement between theory and experiment is sufficient to
assert that the moment method is correctly simulating the T-slot radiator located in
dual-mode rectangular waveguide. Needless to say, today, we can establish the
operational integrity of any new computational analyses based on the moment
method by performing a ‘one-off’ simulation of the geometry of interest using a
modern commercially available finite element package such as Concerto [20], and
comparing results. This has been done in Fig. 5.17 for an ‘extreme’ T-slot radiator
with a very short radiating element (L = 2 mm) and a relatively long clearance
notch (L3 = 13, 14 and 15 mm) in a deep dual-mode waveguide structure. The plots
show strong resonant behaviour with the resonant frequency shifting to lower
values as the overall slot length increases—in accordance with Eq. (5.23). More
importantly the degree of agreement between the moment method (solid lines) and
the finite element technique (dashed lines) is excellent and well within modelling
errors, particularly given the limited number of basis functions adopted in the
moment method.
5.5 Chapter Summary 127
Fig. 5.17 Moment method solution (solid lines) for T-slot compared with finite element software
(dashed lines) for deep dual-mode waveguide (L = 2 mm, L3 = 13, 14, 15 mm, w = 1.5 mm,
t = 1.27 mm)
subjected to the MoM to emphasise the flexibility of the technique. Without sig-
nificant amendments to the computational steps pursued in modelling the con-
ventional shunt slot in waveguide, the key to the evolution of a MoM for this quite
different more complex scattering geometry is the careful construction of an
appropriate basis function.
It is pertinent to note here, that the geometries of many compact radiators, which
are amenable to solution by the method of moments, are undoubtedly solvable
using finite element methods (FEM) such as HFSS and COMSOL. However, while
these solvers permit model construction with minimal knowledge of electromag-
netic theory, essentially by exchanging user modelling choices for computational
intensity, they exhibit the significant downside of being relatively slow computa-
tionally in reaching a solution. As has already been intimated this rules out the use
of FEM in embedded modules within array simulation packages.
Finally, it is arguably true to say that accurate computational design of aperture
antennas, and in particular array antennas, made possible using the method of
moments, has virtually obviated the traditional need for laborious back-up mea-
surement procedures. In the not too distant past, 25–30 years ago, it was necessary
to provide precise design data on waveguide slots, ranging over a multiplicity of
sizes and orientations, in order to facilitate array synthesis calculations. At that time
measurements were the only source of such data.
References
14. J.S. Ajioka, D.M. Tang, N.W. Wong, Arbitrarily polarised slot radiators in bifurcated
waveguide. IEEE Trans. AP-22, 196–200 (1974)
15. A.J. Sangster, R.W. Lyon, Moment method analysis of a T-shaped slot radiator in bifurcated
waveguide. IEE Proc. 129(6), 299–306 (1982)
16. R.J. Chignell, J. Roberts, Compact resonant slot for waveguide arrays. IEE Proc. 125(11),
1213–1216 (1978)
17. A.E. Bailey, Microwave Measurements (Peter Peregrinus, 1989)
18. N.F. Wong, R. Tang, G.A. Okubo, Investigation of Characteristics and Practical
Implementation of Arbitrary Polarised Radiators (Hughes Aircraft Co., Ground Systems
Group, Fullerton, California, 1971)
19. D.M. Pozar, Microwave Engineering, 4th edn. (Wiley, 2011)
20. CONCERTO, Vector Fields Ltd. http://metamaterials-modelling.net/software-concerto.php
Chapter 6
Resonant and Non-resonant Linear
Slot Arrays
6.1 Introduction
A linear array is one that is formed from a single line of multiple radiating sources.
When these individual sources radiate equally in all directions the array is said to
comprise isotropic elements. In addition the array is described as uniform when the
elements are equally spaced and fed with equal signal levels. Such an array is
depicted in Fig. 6.1. Although planar array (an array of linear arrays) antennas are
more commonly used in mobile communications systems, it is advantageous to
study the radiation patterns of linear arrays which are mathematically less complex.
Nevertheless, the knowledge gained from linear array studies remains entirely
relevant to, and representative of, planar arrays.
At a distant point from the antenna the total E-field, at radius r due to n elements
can, by superposition, be expressed as:
where E0 is the field magnitude at the distant point (r) due to a single element of the
array, and very importantly:
w ¼ k0 d sin / þ a ð6:2Þ
Fig. 6.1 Geometrical relationships for linear array with element spacing d
6.2 Uniform Linear Array of Isotropic Elements 133
ET
¼ 1 þ ejw þ e2jw þ þ ejðn1Þw ð6:3Þ
E0
By replacing the exponential terms by the variable z (= ejw) the far-field pattern
of the linear array can be expressed mathematically as a polynomial, as follows:
This polynomial form has been shown [1] to be a powerful mathematical tool in
the analysis and synthesis of array antennas. However, before advancing to the
treatment of arrays in general, it is perhaps instructive to consider some special
cases.
When f(z) is multiplied by z in Eq. (6.4) we obtain:
z1ðzÞ ¼ z þ z2 þ z3 þ þ zn ð6:5Þ
ðz 1Þ1ðzÞ ¼ zn 1
giving
zn 1
1ðzÞ ¼ ð6:6Þ
z1
If this equation is recast in its exponential form, with z = ejw, while recognising
the identity:
ejh ejh
¼ sin h
2
E0 sin nw2
ET ¼ ð6:8Þ
sin w2
Element phase is referred to the array mid-point to secure this simple form.
Functionally, Eq. (6.8) is not too dissimilar to Eq. (3.23), the spectrum function for
a uniformly illuminated linear aperture. The primary difference is the oscillatory
nature of the denominator of Eq. (6.8) which strongly influences the pattern
134 6 Resonant and Non-resonant Linear Slot Arrays
distribution. The result is that the linear array pattern and the linear aperture pattern
are comparable only for a limited range of w values, which implies a limited range
of element spacings d, as illustrated in Sect. 6.2.1.
Equation (6.8) represents the functional form of the far-field radiation pattern, or
polar diagram, of the uniformly excited linear array of isotropic elements. Since the
radiation in the forward half-plane −p/2 < / < p/2 is clearly identical to the
radiation into the reverse half-plane p/2 < / < 3p/2, we need consider only the
former in the following discussion.
The ratio of sines in Eq. (6.8) implies that the pattern must be multi-lobed. The
primary or main lobe occurs, where simultaneously w ! 0 and sin ðnw=2Þ !
nw=2. For this lobe ET =E0 ¼ n; that is at the peak of the primary lobe the E-field at a
given radius for the array is n times the E-field at the same radius for a single
isotropic element. However, note that if the element spacing d is equal to or greater
than ko there may be more than one principal maximum—these secondary maxima
are termed grating lobes and appear where w = p, 2p …. It is apparent from
Eq. (6.2) that the principal maximum at w = 0 occurs where:
a
sin / ¼ ð6:9Þ
k0 d
When this principal maximum appears in a direction normal to the line of the
array, i.e. at / = 0°, the array is termed a broadside array (see Fig. 6.2). Clearly, for
this case a = 0°, which means that the element-to-element phase shift is set to zero
degrees. In the diagram, /1 denotes the angle to the first null measured from the
array normal, while d1 provides essentially the same information, but measured
from the array base line - d1 ¼ 90 /1 .
When / = 90° the principal maximum lies along the line of the array (Fig. 6.3)
and the array, in this case, is termed end-fire. The angle from the lobe maximum to
the first null is denoted by d2 in the figure. From Eq. (6.9) it is clear that for an
end-fire array we require a = −kod.
For a broadside array of equally space and equally excited radiators (uniform array)
the nulls of the radiation pattern occur where the sine ratio in Eq. (6.8) exhibits zero
values, i.e. where:
6.2 Uniform Linear Array of Isotropic Elements 135
nw
¼ mp ð6:10Þ
2
Hence substituting for w using Eq. (6.2) and noting that for a broadside array
a = 0,
nk0 d sin /
¼ mp ð6:11Þ
2
136 6 Resonant and Non-resonant Linear Slot Arrays
2mp
sin / ¼ ð6:12Þ
nk0 d
If the array is large, such that nd mk0 then /m is small for the first few nulls
nearest to the main lobe, and is given by:
mk0 m
/m ¼ ð6:14Þ
nd L=k0
2 114:6
BWFN ¼ 2/1 radians ¼ ð6:15Þ
L=k0 L=k0
For reasons intimated earlier it is much more usual for antenna engineers to
employ half-power beamwidth (HPBW) in defining the radiated main beam and for
the broadside uniform array it is not difficult to show that
2 0:44 50:4
HPBW radians ¼ ð6:17Þ
L=k0 L=k0
For this uniform broadside array of isotropic elements with d < ko/2 the radiated
power density in the principal lobe can be approximated by:
PT
parray W=m2 ð6:18Þ
2pr 2 h3dB
PT
piso ¼ W=m2 ð6:19Þ
4pr 2
Hence the directivity (D) of the array has the simple form:
parray 2
D¼ ð6:20Þ
piso h3dB
The only difference between a uniform end-fire array and a uniform broadside array
is in the value of the progressive phase shift a, so the form of Eq. (6.8) is unaltered
by the change. Consequently, the condition for pattern nulls remains as in
Eq. (6.10). However, in the end-fire case, with a = −k0d, the equation relating w
and / becomes:
w ¼ k0 d sin / k0 d ð6:21Þ
w ¼ k0 d cos d k0 d ð6:22Þ
or
2mp mk0
cos d 1 ¼ ¼ ð6:23Þ
nk0 d nd
Once again, directing our attention to a long array with nd mko , the nearest
pattern nulls occur at angles given by:
rffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffi
dm mk0 m
¼
2 2nd 2L=k0
i.e.
sffiffiffiffiffiffiffiffiffiffi
2m
dm ð6:23Þ
L=k0
For the end-fire array it is not difficult to show that HPBW = 0.66 BWFN,
and therefore we can write:
sffiffiffiffiffiffiffiffiffiffi
2
HPBW 76:3 degrees ð6:25Þ
L=k0
Note that the half-power beamwidth for the end-fire array is very much larger
than for the corresponding broadside array. For example, when the ratio L/ko = 50
the end-fire array beamwidth is approximately 15 times greater than that of the
equivalent broadside format. Some improvements in end-fire gain can be achieved
without increasing the number of radiating elements by judicious adjustment of the
element spacings d. The optimum spacing is termed the Hansen-Woodyard con-
dition, which will be examined below.
The above examination and discussion of broadside and end-fire arrays underlines a
major property of array antennas, namely their ability to provide electronic beam
scanning. The two cases represent the scanning extremes of zero degrees (broad-
side) and ninety degrees (end-fire) when the progressive phase shift a is set to,
respectively, zero and −kod. At the / = 0° (d = 90°) scan condition the principal
beam exhibits minimum beamwidth and optimum directivity. As the beam is
scanned away from broadside towards the end-fire condition both the beamwidth
and the directivity deteriorate, slowly at first, and then increasingly rapidly as
end-fire is approached. This behaviour is illustrated graphically in Fig. 6.4 which
6.2 Uniform Linear Array of Isotropic Elements 139
sin nw2
¼ 0:707 ð6:26Þ
sin w2
for uniform linear arrays with d = ko/2. The curves in Fig. 6.4 represent varying
scan angles (/) and the scan limit defines the angle to which the principal beam can
be scanned unmodified by the proximity of end-fire. For example, a very long array
can be scanned much closer to end-fire than a short one. This figure makes it very
clear that array length is the major determinant of the principal lobe beamwidth and
hence of the array gain.
It is pertinent to ask here whether or not, for any given number of elements n,
there is an element spacing d and an array length L = (n − 1)d which procures
optimum array gain and minimum beamwidth. For a broadside array calculation
based on Eq. (6.8) it can be demonstrated that minimum beamwidth and maximum
gain (allowing for power losses in sidelobes and grating lobes) occur where the
element spacing is just less than one wavelength. This is shown clearly in Fig. 6.5
where gain is presented as a function of d/ko for a range of array options with
values of n ranging from 2 to 24. The pronounced dip in the value of the gain at
140 6 Resonant and Non-resonant Linear Slot Arrays
Fig. 6.5 Gain as a function of d/k0 for a uniform broadside array (adapted from [3])
d=k0 0:95 particularly where n is large, is produced by the appearance of the first
pair of grating lobes in the end fire directions.
In this short section the intention is to introduce the reader to antenna synthesis,
rather than analysis, where we have been directing our attention up to this juncture.
The method which will be outlined applies strictly to uniform arrays. Nevertheless,
the procedures developed here provide a good general introduction to antenna
synthesis concepts without embarking on an overly complex mathematical journey.
Actually, in practice, antenna design engineers generally use very sophisticated
commercial software to perform the complex operations involved in synthesis, but
to do this efficiently some awareness of the basic process is essential.
It has already been shown (Eq. 6.4) that the far-field pattern of an n-element
uniform linear array can be expressed in the polynomial form:
6.3 Array Design Using Theory of Polynomials 141
1ðzÞ ¼ 1 þ z þ z2 þ z3 þ þ zn1
where z ¼ ejw and w is the difference in phase of waves radiated from any pair of
adjacent elements. By a fundamental theory of algebra a polynomial of order
(n − 1) has (n − 1) zeros (some of which may be multiple zeros) and can be
factored into (n − 1) binomials [1]. Thus
The quantities r1, r2, …, rn−1 are termed the roots of the (n − 1)th order poly-
nomial. Since z ¼ ejw has magnitude unity, these roots can be represented graph-
ically as located on a unit circle in the complex z-plane. The magnitude of f
(essentially the far-field pattern) can then be determined by multiplication of the
vector magnitudes (z − r) as illustrated in Fig. 6.6, which shows z ¼ ejw with w
measured clockwise from the real axis denoted by Re. The imaginary axis is
denoted by Im. Complex vector r1 ¼ ejw1 is also represented on the diagram, and
the vector (z − r1) is just the vector difference between z and r1.
For arrays formed from isotropic elements, radiation will be symmetrically
distributed around the axis of the array, and consequently only radiation angles / in
the range −p/2 to p/2 need be considered. From Eq. (6.2), this means that w, the
argument of z, increases from −kod + a to kod + a as / varies from −p/2 to p/2.
Thus the range of w is 2kod. For example, for a broadside array for which a = 0, if
the element spacing is set at half the free-space wavelength (d = k0/2) then w varies
from −p to +p, as shown in Fig. 6.7a. Note that z is presumed to travel in the
clockwise direction on the unit circle. For d = ko/4 (Fig. 6.7b) the range of w is
from −p/2 to +p/2, while the range is from +p/2 to −p/2 when d = 3ko/4, with
z traversing the unit circle 1.5 times (Fig. 6.7c). Given that mainbeams occur when
w = 0°, 360°, …, 2 mp, arrays with large element spacings ðd k0 Þ generate a
grating lobe each time the locus of z passes zero on the circle diagram.
For a uniform linear array antenna the roots of its polynomial representation
(Eq. 6.27) are not restricted as to where on the unit circle they lie, and embedded in
this observation resides the basis of array synthesis. The roots rn can be located at
any angle w except w = 0 where the primary beam is always located. For the
uniform array case the roots, plus the pole at the w = 0 point (black dot), divide the
circle into n equal arcs. This is illustrated in Fig. 6.8a for a four element array.
The far-field pattern magnitude is obtained by forming the product of the vectors
(z − ri). This can be done graphically by multiplying the lengths of the blue vectors
in Fig. 6.8b for all positions of z (red dot) as it moves around the circle. It is not
difficult to observe that the principal maximum occurs at w = 0 where all three
vectors exhibit their maximum length. Subsidiary, less pronounced maxima, also
occur mid-way between the zeros, where the pattern nulls occur. These features can
be located in ‘real space’ (denoted by /) by replacing w with /, using Eq. (6.2) For
example, the nulls of the four element array occur at the w values −p/2, −p, −3p/2,
or more generally at wm = −2 mp/n with m = 1, 2, 3. Therefore the nulls in the
‘real space’ radiation pattern occur at angles /m given by:
6.3 Array Design Using Theory of Polynomials 143
Fig. 6.8 Location of roots on circle diagram for four element uniform array
a 2mp
sin /m ¼ ð6:28Þ
k0 d nk0 d
The ‘real space’ far-field pattern of the four element array can now be con-
structed either graphically, or by plotting the function f(z), which has the form:
By making the relation z1 ¼ zejp=3 Eq. (6.30) can be written in the more
convenient form:
p p 5p
a¼ ¼ ð6:32Þ
2 3 6
This modified end-fire array pattern is sketched in Fig. 6.12 (curve B). Note the
improvement in directivity.
If the overall length of the array is fixed but the number of elements it contains is
allowed to increase, it becomes possible to improve the directivity still further if the
nulls are optimally spaced within the range of w on the circle diagram. Curve C in
Fig. 6.12 shows the pattern that results when the number of elements is increased to
6.3 Array Design Using Theory of Polynomials 145
Fig. 6.11 Root locations for four element end-fire array with improved far field pattern
146 6 Resonant and Non-resonant Linear Slot Arrays
Having discovered that the location of the roots, on the circle diagram for the array
polynomial, has a major influence on the radiated pattern, it becomes possible to use
this graphical technique to determine the optimum element spacing for a uniform
array. For the broadside case the optimum pattern results from including the
maximum number of nulls in the range of w. Therefore, for a large number of
elements n, the maximum number of nulls are active when the range of w is 4p, as
shown in Fig. 6.13a. Since the range of w for a broadside array is 2kod this implies
that 4pd/ko = 4p and hence the optimum element spacing is d = ko.
In the case of the end-fire array the maximum number of nulls become active
when the range of w approaches 2p, as suggested in Fig. 6.13b, which means that in
this case the optimum element spacing d is almost ko/2. Note that in the limit where
d = ko/2 there are two opposed end-fire mainbeams, and the array is bidirectional.
Also, for the end fire array, if it is uniform, the roots are dispersed in an equally
spaced format around the unit circle, and so their spacing is 2p/n, except on either
side of the pole at z = 1. Actually to further improve directivity the width of the
principal lobe can be reduced further, than suggested in Fig. 6.13b, by simply
adjusting the range of w so that for d = 0 the pole lies midway between w = 0 and
the first zero of the polynomial. This is termed the Hansen-Woodyard condition [4].
To achieve this adjustment an extra inter-element phase shift −p/n has to be
introduced, or equivalently a phase shift of almost −p distributed along the length
of the array.
Fig. 6.13 Diagrams showing the range of w for optimum element spacing. a Uniform broadside
array, b uniform endfire array. (The dotted arrows imply that the zeros occupy the entire circle)
148 6 Resonant and Non-resonant Linear Slot Arrays
With fully adjustable arrays, in both element phase and in element power level,
detailed radiation pattern control, encompassing even far-out sidelobes, becomes
feasible. To emphasise this capability, here we will consider the formation of
radiation patterns with no sidelobes.
If the roots of the polynomial f(z) are all equal and can be co-located at the p
position on the unit circle, as suggest in Fig. 6.14, then the resultant pattern has
only one null or zero at w = p and consequently should possess no sidelobes. So
does this root location scheme lead to a sensible polynomial which can be trans-
formed into a realisable array? For the endfire array with the unit circle represen-
tation shown in Fig. 6.14, the polynomial must have the form:
The levels to which the individual elements of the array must be driven to secure
this solution can be deduced by expanding Eq. (6.33) as a binomial series. For
example for a four element array the relative magnitudes are in the ratios 1: 2: 2: 1.
This element excitation scheme will result in no sidelobes for the four element array
provided the element spacing d and the progressive phase shift a are also properly
imposed.
References
1. S.A. Schelkunoff, W.T. Friis, Antennas, Theory and Practice (Wiley, New York, 1952)
2. R.S. Elliott, Beamwidth and directivity of large scanning arrays. Microwave J. (1963)
3. The Microwave Engineers Handbook, Page T-136 (1963)
4. M.M. Dawoud et al., Realization of superdirectivity from active and passive array antennas, in
IEE Conference Publication No. 169 (1978)
Bibliography
1. R.E. Collin, F.J. Zucker, Antenna Theory (McGraw-Hill Book Co., New York, 1969)
2. R.F. Harrington, Time Harmonic Electromagnetic Fields (McGraw-Hill Book Co., New
York, 1961)
3. C.A. Balanis, Antenna Theory (Wiley, New York, 1997)
4. R.S. Elliott, Antenna Theory and Design (Prentice-Hall Ltd., Englewood Cliffs, N.J., 1981)
5. R.C. Hansen, Microwave Scanning Antennas (Academic Press Ltd., New York, 1964)
6. E.C. Jordan, Electromagnetic Wave and Radiating Systems (Prentice-Hall Ltd., Englewood
Cliffs, N.J., 1968)
7. A.W. Rudge, K. Milne, A.D. Olver, P. Knight (eds.), The Handbook of Antenna Design:
Volumes I and II (Peter Peregrinus Ltd., London, 1986)
8. B.A. Munk, Metamaterials: Critique and Alternatives (Wiley, New York, 2009)
Chapter 7
Conventional Waveguide-Fed
Travelling-Wave Slot Arrays
7.1 Introduction
have seen in Chap. 6, desirably focused and directed radiation patterns depend on
the phase and radiation strength of the exciting electromagnetic wave at the position
of each slot. In the case of excitement by a travelling wave, this means slots
equi-spaced in the axial direction of the waveguide, generating *360° progressive
phase increases to secure an approximately broadside main radiation lobe. Lobe
location is a key issue with travelling wave arrays as we shall see.
Radiation from an aperture or slot machined or etched into a conducting wall of
a rectangular waveguide, or any other shape of waveguide for that matter, is
maximised by orientating the slot into alignment with the TE10 mode magnetic field
(H10) adjacent to the inner surface, or equivalently normal to the accompanying
surface current density (Js10). This behaviour is illustrated in Chap. 1 and
demonstrated theoretically in Chap. 5. For consistency, given that the longitudinal
shunt slot is examined in detail in Chap. 5, we will direct our attention here toward
the longitudinal slot array. Generally, for such an array resonant slots are employed
to maximise array bandwidth. In Chap. 5, Figs. 5.6, 5.7 and 5.8 inform us that at
resonance a longitudinal broadwall slot presents a real impedance to the waveguide
—so in transmission line terms it can be modelled with a shunt resistance or
conductance as shown in Fig. 7.1. A shunt, rather than a series, connection is
dictated by the phase behaviour of S11 and S12 for the longitudinal slot.
The travelling wave array antenna formed from resonant longitudinal slots
inserted periodically into a broadwall of waveguide has been examined widely in
the literature. This interest has been encouraged by the early observation [2] that the
off-set resonant slot, as depicted in Fig. 7.1a, can be modelled in transmission line
terms by a simple shunt conductance (Fig. 7.1b). In fact Stevenson [2] has gen-
erated an elegant closed form equation for the normalised shunt conductance g of a
narrow slot radiating into a half-space, namely
Ys kg a 2 pk0 pd
g¼ ¼ 2:09 cos sin2 ð7:1Þ
Y0 k0 b 2kg a
where d = a/2 − X1, Ys is the slot admittance in the waveguide (=G when real), Y0
is the waveguide characteristic admittance, while kg = kg10 is the TE10 mode
wavelength in the waveguide at the operating frequency. The validity of Eq. (7.1)
as a ‘trial’ value in any travelling wave array synthesis procedure is confirmed by
the moment method computations presented in Chap. 5. If the power entering the
waveguide in the TE10 mode is known then the guide voltage and hence the
equivalent transmission line value V (Volts) is available. Hence the power radiated
by the slot, normalised to the input power, for the scenario in Fig. 7.1b is simply:
V 2g
pr ¼ ð7:2Þ
2
In words, the slot excitation level is proportional to g, which in turn [Eq. (7.1)] is
proportional to sin2 pda : The primary mechanism for controlling the radiation
pattern for a shunt slot travelling wave array thus suggests itself. Namely, from
7.1 Introduction 153
Fig. 7.1 Longitudinal (shunt) slot in TE10 mode rectangular waveguide: a waveguide geometry,
b equivalent circuit for resonant slot with normalised g = Ys/Yo
Fig. 7.2 Resonant slot travelling-wave array antenna (a) and equivalent transmission line model (b)
7.2 Travelling-Wave Array—Design Principles 155
combined power being returned to the array input, and hence the microwave source.
Operation of a travelling wave array exactly at broadside will be examined as a
special case at the end of this section.
The main lobe of the travelling wave array radiation pattern is located in the
angular range p=2 / p=2 in the y-z plane. In the transverse x-y plane the
pattern is essentially omni-directional in the forward half-space (+y-direction).
Furthermore, the antenna is generally assumed to be terminated in a perfectly
matched waveguide load represented by admittance YL in the equivalent trans-
mission line.
For radiation pattern calculations in the ensuing section, we note here, referring
to Fig. 7.2, that geometrically the centre of any given slot (n) can be defined as
located at rn where
^y
rn ¼ nd^az þ xn ^ax þ b ð7:3Þ
For a fixed height waveguide, b is common to all slots in the array, and hence it
has no influence on pattern calculations and can therefore be omitted. Thus
In Chap. 6 it has been noted that the radiation pattern for an array antenna can, in
the absence of mutual coupling, be elegantly formed from the product of the array
factor and the element factor. The former applies to the radiation pattern of the
equivalent array comprising isotropic radiators, while the latter alludes the radiation
pattern of the isolated element. For a uniform array of isotropic elements it is shown
that the radiation pattern can be simply represented mathematically [see Chap. 6,
Eq. (6.3)] as:
E
¼ 1 þ ejw þ e2jw þ þ ejðn1Þw ð7:5Þ
E0
where Eo is the field at a given radius for a single element, while E is the radiated
field strength for the array at the same radius in the plane containing the array (y-z
in Fig. 7.2). Also, for unstaggered elements, w the propagation phase difference
between adjacent elements is given by (see Eq. 6.2) namely:
w ¼ k0 d sin / þ a ð7:6Þ
However, for the travelling wave array, the slots are seldom equally excited and
are seldom unstaggered, so Eqs. (7.5) and (7.6) have to be adjusted slightly to
represent a non-uniform structure. Hence we obtain:
156 7 Conventional Waveguide-Fed Travelling-Wave …
E
¼ a1 ejk0^rr1 þ jpjb10 d þ a2 ejk0^rr2 þ 2jp2jb10 d þ þ an ejk0^rrn þ jnpjnb10 d ð7:7Þ
E0
where d is the element separation in the z-direction. It is also the distance from the
origin to the first element. an is the field excitation amplitude at the nth slot while
b10 is the propagation phase for the TE10 mode in the waveguide. The unit vector ^r
is given by:
with D ¼ p2 / in the y-z plane, and h the angular rotation from the x-axis in the
x-y plane. Consequently
Generally we are interested in the radiation pattern in the y-z plane (h = p/2
plane) in which case the most relevant and useful form of Eq. (7.9) is:
Evidently, and perhaps not surprisingly, Eq. (7.10) closely matches Eq. (7.6)
except that the a phase term is replaced by the slot-to-slot phase delay (−b10d) for
the TE10 mode in the waveguide. The additional p phase shift term accounts for slot
staggering. It disappears for an array with shunts slot located only on one side of the
centre line.
Referring back to Chap. 6, we have observed there that the main radiating beam
in the plane of the array (y-z plane) occurs where w = 0, while secondary beams or
grating lobes occur at w = 2 mp (m = 1, 2, … M). That is, for primary lobes we
must have:
Hence primary lobes occur at values of the slot spacing d given by:
Equation (7.12) is plotted in Fig. 7.3, for low values m, for a resonant shunt slot
travelling-wave array located in 22.86 mm 10.16 mm (WG16) rectangular
waveguide at an operating frequency of 10.0 GHz. Equation (7.12) reveals that the
curves in Fig. 7.3 intercept the / = −90° axis at
ð1 2mÞk0 kg10
d¼ ð7:13Þ
2ðk0 þ kg10 Þ
7.2 Travelling-Wave Array—Design Principles 157
Fig. 7.3 Influence of slot spacing d on the principal beam locations (/ degrees) for a staggered
shunt slot travelling wave array in WG16 waveguide at 10 GHz
k0
sin / ¼ ð7:15Þ
kg10
The figure thus provides a clear and rapid indication of the number of principal
beams generated by the array for a particular value of d. At a given value of
d (60 mm say), if a horizontal line (blue dashed line) is inserted onto the figure at
this value then the possible main beams, or grating lobes, in /-space will be
predicted to occur at approximately −20°, 7°, 33° and 75°. To secure a single main
beam in the y-z plane the slot spacing d must be located between the m = 0 curve
and the m = −1 curve at / = −90°. That is, we must have:
For WG16 operating at 10 GHz this implies that d should be in the range 8.55–
25.65 mm. The m = 1 solution between / = 75° and / = 90° is not significant in
158 7 Conventional Waveguide-Fed Travelling-Wave …
the above range of d if kg < 2k0. This condition lies in the normal operating range
for standard rectangular waveguides.
Although seldom adopted because of grating lobe issues a brief examination of the
unstaggered shunt slot array provides further insight into the nature of the
travelling-wave array antenna. As we have already noted, the fundamental phase
relationship governing primary beam locations, namely Eq. (7.11), is modified by
the absence of the p term when applied to the unstaggered case. Hence the
equivalent d-equation [to Eq. (7.12)] locating main beams becomes
mk0 kg10
d¼ ð7:17Þ
ðkg10 sin / k0 Þ
Fig. 7.4 Influence of slot spacing d on the principal beam locations (/ degrees) for a unstaggered
shunt slot travelling wave array in WG16 waveguide at 10 GHz
7.2 Travelling-Wave Array—Design Principles 159
the blue solid line in Fig. 7.4. It is solid because it now denotes a real solution,
usually termed the d-independent solution. Physically this operating condition
occurs when the phase velocity of the TE10 mode in the waveguide vp10 = x/bg10
matches the z-component of the velocity of the radiated beam along the outer
surface of the waveguide. For this solution to be a single beam operating mode, the
element spacing d must be less than for the m = −1 solution at w = −90°, namely
k0 kg10
d¼ ð7:18Þ
ðkg10 þ k0 Þ
Fig. 7.5 Influence of slot spacing d on the principal beam locations (/ degrees) for a unstaggered
shunt slot travelling wave array in WG16 waveguide at a 8 GHz and b 12 GHz
yn ¼ gn þ jbn ð7:19Þ
The normalised admittance ‘seen’ by the nth slot ‘looking’ toward the termi-
nation is also defined as:
7.2 Travelling-Wave Array—Design Principles 161
Fig. 7.6 Shunt slot travelling wave array antenna—design example a array configuration,
b general transmission line model
Consequently, ‘looking’ toward the nth slot from the (n − 1)th slot, the com-
bined admittance becomes:
Hence, using transmission line theory, the reflection coefficient seen by the nth
slot can be expressed as:
1 yTn
qn ¼ ð7:22Þ
1 þ yTn
Now the full range of power contributions at the nth slot is delineated in
Fig. 7.6b, and hence applying power conservation there, we obtain:
where, Pn is the power radiated by slot n, Pfn is the power in the waveguide
reaching slot n, Psn is the forward power in the waveguide on the load side of the
nth slot, Pqn is the backward power on the source side of the nth slot and Pbn is the
‘backward’ flowing power in the waveguide on the load side of the nth slot.
Equation (7.24) can be simplified into (Fig. 7.6b):
where Pn jin and Pn jout represent the net powers in the waveguide flowing into and
out of the nth slot. If the voltage on the transmission line representation of the slot
array (Fig. 7.6b) at the nth slot is Vn then the following definitions apply:
1
Pn ¼ jVn j2 gn ð7:26Þ
2
and
1
Pfn ¼ jVn j2 gTn ð7:27Þ
2
1
PT ¼ jVT j2 gT ð7:28Þ
2
Pn ¼ Cg a2n ð7:29Þ
The constant Cg can be estimated by introducing a trial value for that proportion
of the source power reaching the load after passing through the array (r say). Thus
7.2 Travelling-Wave Array—Design Principles 163
PT
r¼ ¼ PT ð7:30Þ
Pin
if power from the source Pin is normalised to unity. With these simplifications we
can define [9] a continuous conductance per unit length g(z) such that:
gn
gðzÞ ¼ gðndÞ ¼ ð7:31Þ
d
PðzÞ
gðzÞ ¼ ð7:32Þ
Pf ðzÞ
Zz Zz
Pf ðzÞ ¼ 1 PðzÞdz ¼ 1 Cg ðaðzÞÞ2 dz ð7:33Þ
0 0
with a(z) = an/d and using the continuous version of Eq. (7.29). Furthermore at the
load end of the array (z = L) we must have, from Eqs. (7.30) and (7.33):
ZL
r ¼1 Cg ðaðzÞÞ2 dz ð7:34Þ
0
On combining Eqs. (7.32), (7.33) and (7.34) an elegant expression for the
continuous conductance is generated—namely:
ðað1ÞÞ2
Lgð1Þ ¼ ð7:35Þ
RL RL
1
1r ðað1ÞÞ2 d1 ðað1ÞÞ2 d1
0 0
where f = z/L. Equation (7.35) is plotted in Fig. 7.7 for a(f) distributions with
Taylor functionality in order to produce Chebyschev radiation patterns [10]—one
which produces −20 dB flat sidelobes in the array radiation pattern, while the
second case develops −25 dB sidelobes.
Four curves are presented in Fig. 7.7, two associated with −20 dB sidelobe
designs (blue traces with circle markers) and two associated with −25 dB sidelobe
designs (red traces with square markers). In both cases the curves are differentiated
by the power (r), as a percentage of the source power, reaching the load. It is
noticeable that for the low power cases (r = 5%) g(f) as a function f is much more
sensitive to increasing f than for the cases where r = 35%. The reason for this
164 7 Conventional Waveguide-Fed Travelling-Wave …
Fig. 7.7 Continuous conductance profile for lossless, perfectly matched, shunt-slot
travelling-wave array antennas (with acknowledgement to [9])
difference in behaviour lies with the fact that with only 5% of the input power
reaching the load the last few slots in the array are exposed to a very weak level of
pffiffiffiffiffiffiffiffiffi
TE10 mode excitation. However, given that að1Þ / Pð1Þ , it becomes clear from
Eq. (7.31) that a low power level (Pf(f)) requires a large g(f) value to secure the
desired slot excitations a(f). This difficulty is exacerbated for the −20 dB sidelobe
array which requires higher a(f) levels at the slots near the array extremities to
procure the higher sidelobes. Note that this option equates to higher gain. In fact the
way in which the magnitude of g(f) escalates as f approaches unity highlights a
significant design limit for the shunt slot travelling-wave array antenna. Eventually,
as r is reduced, the level of g(f) cannot be realised by a shunt slot in waveguide
where the maximum g(f) is restricted by the optimum off-set (d = a/2) of a slot
from the centre line of the waveguide [see Fig. 7.1 and Eq. (7.1)].
A compromise between the magnitude of r, array efficiency and the desired gain
becomes inevitable. High efficiency (low r) is easier to procure with low sidelobe
levels which equate to less than optimum gain. In general this balancing process
becomes less pronounced for long arrays with large N. With r at the 35% level in
Fig. 7.7 it is interesting to note that the g(f) distribution along the array follows
closely the Taylor distribution applied to a(f). This is simply because Pf(f) remains
relatively constant along the array in these cases. Finally, low sidelobe designs
require care in implementing a terminating load which must be well matched to the
feed waveguide forming the array. An imperfect match results in a reflected TE10
mode which becomes a source of a secondary radiation pattern with a ‘main’ beam
pointing in a direction equal and opposite, relative to the array normal, to the
7.2 Travelling-Wave Array—Design Principles 165
designed beam. This ‘reflection’ beam together with the original sidelobes in the
same direction should ideally combine to form a lobe which is higher than the
designed sidelobe level. Calculations performed by Dion [10] suggest that at the
−30 dB sidelobe level in an array designed on the basis of ‘dumping’ 5% of the
source power in the termination, a load with a generous VSWR of *1.34 would be
satisfactory. On the other hand, at r = 35% a high quality load with a VSWR 1.1
is required to maintain the sidelobe specification. At the −20 dB sidelobe level the
load requirements are, not unexpectedly, considerably more relaxed with a load
with a VSWR 1.4 being acceptable, even at r = 35%.
For a slot array of predetermined length L, the Dion continuous conductance
technique permits the allocation of trial values for the discrete conductances gn in an
ensuing practical design. For such an N element array design with L = Nd we
choose gn as defined in Eq. (7.31) with z = nd, on the assumption that the first slot
is located a distance d from the array input. An estimate for the parameter r is also
furnished by the Dion analysis and hence the power PT transmitted into the load is
calculable.
For an array more accurately modelled by admittances (Fig. 7.6b) embedded in a
waveguide, which is not loss free, Kaminow and Stegen [4] provide an iterative
procedure to procure a design. This commences with the trial conductances from
the above procedure and the known conditions at the load, then progresses slot by
slot back toward the array input, by employing power conservation principles and
transmission line theory. The power conservation requirement is defined in
Eq. (7.24), with power definitions in Eqs. (7.26), (7.27) and (7.28). Since it is
unlikely that the termination will be other than well matched we can reasonably set
pffiffiffiffiffiffi
gT = 1, in which case VT ¼ PT .
At slot N, which we assume is distance d from the load, the power conservation
requires:
PN PN
gN ¼ ð7:37Þ
PfN PN jin
Now, for a lossy line, transmission line theory gives the voltage at slot N as:
using Eq. (7.21). At the Nth slot transmission line theory also yields:
yT þ tanhðc10 dÞ
ysN ¼ ð7:40Þ
1 þ yT tanhðc10 dÞ
jVn yn
\ ¼ \ expðjb10 dÞ ð7:42Þ
jVn þ 1 yn þ 1
This equation ensures that the slot to slot excitation phase shifts in accordance
with array beam formation in the specified design direction. It leads to the following
iterative relationship for phase, namely
bn bn þ 1 Vn þ 1
tan1 ¼ tan1 þ b10 d þ \ ð7:43Þ
gn gn þ 1 Vn
For the Nth slot, ‘looking’ into a matched termination, this reduces to:
bN
tan1 ¼ b10 d ð7:43bÞ
gN
With PN, gN and bN at the final slot now established the iteration becomes
focused on the (N − 1)th slot. Firstly the power levels at slot N can be transferred,
in accordance with transmission line theory and power conservation, back to slot
N − 1 by applying power conservation at a reference plane on the load side of slot
N − 1 (see Fig. 7.6). This leads to
And
h i
PN jin ¼ 1 jqN j2 Pf ;N ð7:45Þ
The following useful equation [10] relating the N − 1th slot to the Nth slot then
results:
7.2 Travelling-Wave Array—Design Principles 167
( )
2 j qN j 2
PN1 jout ¼ PN jin expð2a10 dÞ þ sinhð2a10 dÞ ð7:46Þ
1 jqN j2
where qn is defined in Eq. (7.22). Hence with PN known at the Nth slot and PN jin
available from Eq. (7.37) the power leaving the N − 1th slot can be calculated
using Eq. (7.46). The power on the source side of slot N − 1 is computed using
Eq. (7.25) with gn−1 derived from the transmission line Eq. (7.40). VN−1 is deter-
mined from transmission line theory [Eq. (7.39)] which then permits the evaluation
of bN−1. The process is then repeated for slot N − 2, N − 3, etc. back to slot 1.
Eventually, with all of the gn’s and bn’s evaluated the slot offsets dn, the slot
spacings dn and the slot lengths Ln can be deduced from Eqs. (7.1) and (7.12) with
the slot resonant lengths ‘tweeked’ to accommodate the susceptance values.
Actually, the sophisticated antenna simulation software packages which are avail-
able to today’s designers, allows the electrical data generated above for the
travelling-wave array antenna to be inserted into a waveguide slot analysis program
(see Chap. 5) to ascertain accurate dimensional design parameters including slot
widths and waveguide wall thicknesses.
Fig. 7.8 Resonant form of the waveguide shunt slot array a schematic representation, b circuit
representation
Fig. 7.9 Pictorial representation of the slot excitation mechanism in the resonant array
7.2 Travelling-Wave Array—Design Principles 169
X
N
ge ¼ gn ð7:47Þ
n¼1
where N is the number of slots and for a good array matched to the source, normally
we would choose ge = 1. Note that the kg10/4 short circuited line presents an open
circuit across the final slot.
The matching condition alone is insufficient to determine all of the gn values, but
we have seen from Chap. 3, and from Chap. 6, that the field distribution across the
aperture of the antenna, or the element excitation level in the case of an array, can
be shaped to procure high directivity or low side lobe radiation patterns (see
Fig. 3.7). Consequently, for the resonant array let the excitation level, or coefficient,
for the nth slot be an. For a uniform array (see Chap. 6) the an’s have a fixed level,
but may be set to a cosine pattern (for example) if lower sidelobes are desired. Since
the slot conductance is proportional to the radiated power [see Eq. (7.2)] we can
therefore write:
gn ¼ Cg a2n ð7:48Þ
where Cg is a constant yet to be defined. Hence combining Eqs. (7.47) and (7.48)
yields:
X
N
Cg a2n ¼ 1 ð7:49Þ
n¼1
for a matched array. If the an’s are established to meet pattern requirements then the
constant Cg becomes calculable. Once Cg is known the gn’s can be determined and
hence the remaining unknowns required to complete the array design, namely the
slot off-sets d, can be computed by resorting to Eq. (7.1).
With their relative simplicity, from a geometrical perspective, their robustness, their
reliability and efficiency, slotted waveguide array antennas have found particular
favour in airborne radar, radiometry and satellite communications roles.
Furthermore, in both resonant and travelling-wave forms, control of slot excitation
levels and phase relationships can be sufficiently precise, as we have seen, to permit
array designers to procure antenna radiation patterns with tightly specified gains,
and sidelobe levels. This leads to high operating efficiencies. It is fair to say that
modern precision manufacturing techniques now permit relatively faithful repli-
cation of computer simulations, and this has resulted in array designs displaying
sidelobe levels well below −20 dB. A typical ‘flat plate’ waveguide slot array is
170 7 Conventional Waveguide-Fed Travelling-Wave …
shown in Fig. 1.13. The flat plate diameter, in the case of a circular embodiment,
can range from well over a metre at S and C band frequencies to little more than a
few centimetres at mm-wave frequencies (>60 GHz). Applications for such
antennas range widely from airborne tracking and seeking roles to ground based
mobile phone systems which predominantly employ linear array versions of the
slotted waveguide antenna (see Johnson [6] and Rudge [11]).
Following pioneering effort in the 1940s and 1950s [1–3], slot array antennas in
radar, and in communications applications flourished, arguably most fruitfully and
determinedly during the ‘cold war’. However by the 1990s these developments had
seemed to stall. The advent of ever more precise manufacturing methods, such as
laser machining, electron beam technology, photographic etching and 3D printing,
have recently begun to push applications into the mm-wave frequency range (circa
60 GHz). The low-loss advantage of waveguide, over alternative feed mechanisms,
remains a major attraction for antenna designers, and unquestioningly helps to
maintain interest in slotted waveguide arrays. At mm-wave frequencies relatively
small arrays can procure the tightly controlled radiation patterns, which are
instrumental in the realisation of atmospheric sensors, automotive radar sensors and
gigabit/second wireless interfaces [12]. In the ‘wireless’ category, a number of
applications are being planned such as high definition multimedia interfaces
(HDMI) enabling (1) uncompressed high definition (HD) video, streaming wire-
lessly rather than by inconvenient cable technology, (2) computer mobility in
distributed environments, (3) wireless docking stations, (4) wireless gigabit
Ethernet operation, (5) fast transfers of bulky files, (6) wireless gaming [13]. Other
applications will undoubtedly emerge in the near future.
Wireless systems at 60 GHz have materialised as one of the most promising
candidates for multi-gigabit indoor communication links. Evidently, a deciding
factor that encouraged this recent interest, lies with the huge unlicensed bandwidth
(up to 7 GHz) available worldwide. While this is comparable to the unlicensed
bandwidth allocated for ultra-wideband (UWB) systems, at 60 GHz the avail-
able bandwidth is continuous and power limits are much less restrictive. This
difference is governed by the fact that an UWB system must operate in concert with
other systems, and is thus subject to very much more demanding regulations [5].
In communication terms, the bandwidth available at 60 GHz band is one of the
largest unlicensed bandwidths which has ever been allocated. The huge 7 GHz
bandwidth has the potential to enable systems demanding very high levels of
channel capacity and flexibility thus making 60 GHz technology particularly
attractive for gigabit wireless applications. Furthermore, 60 GHz regulation allows
much higher transmit power compared to other existing wireless local area net-
works (WLANs) and wireless personal area networks (WPANs) operating at
microwave frequencies, so accommodating and counteracting the higher transmit
power levels, which are necessary to overcome the inevitably higher path losses
encountered at 60 GHz.
At the time of writing, it is true to say that a large number, and a wide range, of
planar antennas have been studied for millimeter-wave radio and radar applications.
The most obvious alternative to the waveguide slot array is the substrate mounted
7.3 Some Slot Array Applications 171
patch array [6]. While high gain operations have been demonstrated with the latter,
microstrip substrates suffer from serious loss mechanisms at mm-wave frequencies.
Consequently, the efficiency diminishes as the gain and/or frequency is pushed
upwards. Despite the maturity of patch array technology these limitations are
unavoidable. It has been estimated that the efficiency of microstrip arrays in the
60 GHz band, with gains of the order of 35 dBi, could be 20% lower than an
equivalent waveguide slot array [16]. On-chip antennas have also been mooted for
mm-wave wireless roles but these exhibit their own particular drawbacks. On
conductive high-permittivity silicon their radiation efficiency is inevitably low. As a
result, in a WPAN environment, where wireless communication distances of the
order of 10 m need to be accomplished, multi-element arrays are necessary to
achieve the 15–20 dBi gain levels required for reliable transmission. If imple-
mented in MMIC chip technology such systems attract high cost [16].
It is, therefore, difficult not to conclude that, for wireless applications, waveguide
slot antenna arrays represent the most attractive candidates where high-gain planar
antennas, with high efficiency, are sought. Nevertheless waveguide fed slot arrays
present their own difficulties where low cost is important. Primarily, the compli-
cated three dimensional feed arrangements in early designs tended to limit their use
to cost-insensitive military or professional applications. An array configuration with
manufacturing costs potentially comparable to that of microstrip counterparts was
required. The solution lay with single-layer waveguide slot array designs which
lend themselves to mass production [13–18].
The impetus for this development was to note that, in principle, planar
waveguide fed slot arrays can be reduced to two parts. These are the slotted top
plate and the grooved bottom plate containing the waveguide channels (Fig. 7.10).
Several versions of these single-layer waveguide arrays, intended for high efficiency
over 12 and 20 GHz bands, were initially developed [13–17]. Subsequently,
manufacturing methods improved sufficiently to permit the extension of single layer
waveguide arrays to higher frequencies up to and including 60 GHz [18–20].
Slotted waveguide, single-layer, planar array antennas, aimed at satisfying
mm-wave developments in communications, evolved rapidly after 1995, under the
promptings of research scientists at the Tokyo Institute of Technology in particular.
Initially these revolved around the conceptual arrangements depicted in Fig. 7.10
and in particular the alternate phase arrangement in Fig. 7.10b [13]. This anti-phase
relationship between adjacent waveguides is counteracted by the ‘mirrored’ slot
staggering, which is clearly seen by comparing Fig. 7.10a, b. So while the phase
reversals have no influence on the beam forming, it has significant implications for
the wave propagation in the waveguide grooves. With no net current flow in the
internal walls the degree of electrical contact between the slot plate and the
channelled plate could be greatly relaxed. This meant that more cost effective array
manufacturing techniques could be explored [15, 16].
In Ref. [15] it was shown that while anti-phase operation of a waveguide fed
planar slot array eliminated the need for ‘good’ electrical contact between the
common internal waveguide walls, and the slotted top plate, the same did not apply
to the peripheral walls. The solution was to introduce traditional ‘choke’ technology
172 7 Conventional Waveguide-Fed Travelling-Wave …
Fig. 7.10 Single layer planar slotted waveguide array antennas with a co-phase corporate feed
b anti-phase transverse linear waveguide feed
7.3 Some Slot Array Applications 173
(see Fig. 1.13), long used to suppress leakage between minimal contact waveguide
flanges. With the addition of choke suppression, Ref. [16] demonstrated that mass
production of the waveguide antenna at low cost was possible. Die-casting was
adopted for the corrugated base plate of the antenna to circumvent high precision
metal cutting which was considered to be too time-consuming and expensive. The
dimensional tolerance associated with die-casting is typically only 50lm and
normally some further machining is necessary in the 26 GHz band. To avoid this, a
high precision casting technique was developed (see Ref. [16]).
To confirm the fabrication process, a travelling-wave array antenna with a
specified gain of over 30 dBi was designed, for operation at approximately
25 GHz. The aperture area occupied by slots was set to be 16 cm 15 cm. It
comprised 16 parallel waveguides, with a = 8 mm and b = 3 mm, each feeding 19
staggered shunt slots. For the die-cast antenna the gain exceeded 31 dBi over the
frequency range and the corresponding efficiency was more than 65%. The return
loss was a creditable −20 dB. At that juncture in the evolution of arrays for
mm-wave systems (circa 2005), the planar slotted waveguide array antenna could
justifiably be claimed to surpass competitive offerings in any given frequency and
gain range.
More recent developments [17] acknowledged the ‘beam-squint’ problem
encountered with travelling wave arrays (see Figs. 7.4, 7.5 and 7.6) which led to the
evolution of matched shunt slots incorporated into a planar array of centre fed
slotted waveguides. By feeding each linear sub-array, not from their ends as in
Fig. 7.10, but from their mid-points, the long line effect was halved with con-
comitant bandwidth enhancement. But more significantly, this symmetrical feed
scheme enforced the main beam of the array as a whole to remain on the boresight.
Furthermore, the necessity for the sub-arrays to be designed for non-boresight
operation to minimise return loss was greatly mitigated by employing matched
slots. Reflection cancellation can be procured in a number of ways, usually by
introducing an insert in the vicinity of the slot, which exhibits a reflection phase
counter to that of the slot. The inductive post was found to be the most effective
mechanism for suppressing the cumulative slot reflection at the input port, thus
allowing the beam tilting technique to be discarded [17]. The resultant antenna
displayed wide bandwidth and a main beam directed steadily at boresight. However
sidelobes were higher than equivalent end-fed examples of the same antenna due to
the blockage incurred by the centred feed geometry.
Later developments in the slotted waveguide array scenario were very much
driven by poor transmission compatibility with microwave integrated circuits. This
disadvantage does not arise to the same extent with patch arrays fed from mi-
crostripline or coplanar waveguide. Needless to say the realisation of single-layer
waveguide arrays offering well designed well matched transition structures for
efficient integration with other planar circuits became a high priority [18]. This so
called substrate integrated waveguide (SIW), also called post-wall waveguide or
laminated waveguide, was realised in a metal-clad dielectric substrate by intro-
ducing two rows of metallised via-holes by a routine printed circuit-board fabri-
cation technique (see Fig. 7.11). Importantly this could be accomplished at low
174 7 Conventional Waveguide-Fed Travelling-Wave …
Fig. 7.11 Millimetre-wave slotted waveguide array antenna in substrate based post-wall or SIW
waveguide [18]
cost. It has been demonstrated [18] that the SIW slot array antenna could easily be
integrated with other circuits, resulting in cost-effective subsystems. A range SIW
slot antenna arrays and beam forming networks have been developed [19]. In Ref.
[19] in addition to the extension of the SIW concept from 60 to 79 GHz the
designed antenna was formed on sufficiently thin substrate sheets (flexible PC foil)
to limit waveguide leakage problems in the dielectric and at the post-wall, which
can arise at mm-wave frequencies. In the polyimide foil the post wall is replaced by
a closely packed ‘forest’ of nickel wires formed by a process termed ion track
technology. Physical flexibility was a useful bonus.
The loss of gain and efficiency which is unavoidable with substrate based slot
array solutions, led some researchers to persist with hollow waveguide options in
order to retain its intrinsic low loss advantages [20–23]. To procure such a hollow
waveguide antenna design, at a cost level compatible with substrate based alter-
natives, development gravitated toward a structure described as a double-layer,
corporate-feed, waveguide slot array [20]. The typical antenna consisted of a cor-
porate waveguide feed (see Fig. 7.10a) arranged in the lower layer, and 2 2 small
subarray units in the upper layer. The 4 elements in the subarray are almost equally
spaced from the coupling aperture between layers at the end of each branch of the
corporate feed so that the array is substantially fed by a full corporate feed
(Fig. 7.12). With constant element spacing of less than the free-space wavelength,
grating lobes were effectively controlled.
7.3 Some Slot Array Applications 175
Fig. 7.12 Corporate feed and laminated waveguide slot array antenna
While in the 12 GHz band, a cavity-backed planar slot array antenna with
16 16-elements had demonstrated a gain of 33.7 dBi with 85% efficiency and
12% bandwidth (VSWR < 2) using conventional fabrication techniques, it was
difficult to repeat these characteristics in the millimetre wave band. Greater fabri-
cation accuracy was needed. This was further illustrated [24] in a 40 GHz band,
two-layer, slotted-waveguide antenna array with 24 24-elements. While it dis-
played a creditable gain of 33.8 dBi, efficiency was appreciably lower at 51%, and a
narrower bandwidth of 5.5% was measured, even when the array was carefully
assembled and secured with screws. In contrast, later attempts [20, 21] which have
employed the laminated fabrication method, achieved superior efficiency and
bandwidth outcomes from a corporate feed hollow waveguide array. The figures
compared very favourably with lower frequency examples, even in the high fre-
quency band of 60 GHz. An illustrative photograph of the 60 GHz corporate feed
hollow waveguide array antenna (16 16 slots), showing the laminated con-
struction is presented in Fig. 7.13. The robust yet compact form for of the manu-
factured array is clearly seen particularly when compared with a coin similar in size
to a UK sterling 10p piece. The inclination of the slots by 45o relative to the main
axis of the almost square antenna, represents a simple yet effective scheme to
minimise grating lobes [21].
Fig. 7.13 Photograph of laminated hollow waveguide slot array a top view, b side view [21]
References
1. H.A. Bethe, Theory of diffraction by small holes. Phys. Rev. 66(7–8), 168–183 (1941)
2. A.F. Stevenson, Theory of slots in rectangular waveguide. J. Appl. Phys. 19, 24–38 (1948)
3. A.A. Oliner, The impedance properties of narrow radiating slots in the broad face of
rectangular waveguide. IRE Trans. AP-5, 4–20 (1957)
4. I.P. Kaminow, R.J. Stegen, Waveguide Slot Array Design (Hughes Aircraft Co. Technical
Memorandum, 348, Culver City, California, 1954)
5. A. Derneryd, Linear polarised microstrip antennas. IEEE Trans. Antennas Propag. AP-24,
846–850 (1976)
6. R.C. Johnson, Antenna Engineering Handbook (McGraw-Hill Inc., New York, 1993)
7. S. Silver, Microwave Antenna Theory and Design (McGraw-Hill Book Co., New York, 1949)
8. R.E. Collin, F.J. Zucker, Antenna Theory (McGraw-Hill Book Co., New York, 1969)
178 7 Conventional Waveguide-Fed Travelling-Wave …
9. R.S. Elliott, Antenna Theory and Design (Prentice-Hall Ltd., Englewood Cliffs, N.J., 1981)
10. A. Dion, Non-resonant slotted arrays. IRE Trans. Antennas Propag. AP-6, 360–365 (1958)
11. A.W. Rudge, K. Milne, A.D. Olver, P. Knight (eds.), The Handbook of Antenna Design:
Volumes I and II (Peter Peregrinus Ltd., London, 1986)
12. S.K. Yong, C.-C. Chong, An overview of multigigabit wireless through millimeter wave
technology: potentials and technical challenges. EURASIP J. Wirel. Commun. Netw. 2007,
1–10 (2007)
13. K. Sakakibara, Y. Kimura, A. Akiyama, J. Hirokawa, M. Ando, N. Goto, Alternating
phase-fed waveguide slot arrays with a singlelayer multiple-way power divider. Proc. Inst.
Electr. Eng. Microwaves Antennas Propag. 144, 425–430 (1997)
14. M. Ando, J. Hirokawa, in High Gain and High Efficiency Single Layer Slotted Waveguide
Array in 60 GHz Band. Proceedings of the 11th International Conference on Antennas and
Propagation, Publication No. 436, vol. 1 (Edinburgh, UK, 1997), pp 464–468
15. Y. Kimura, T. Hirano, J. Hirokawa, M. Ando, Alternating-phase fed single-layer slotted
waveguide arrays with chokes dispensing with narrow wall contacts. Proc. Inst. Electr. Eng.
Microwaves Antennas Propag. 148, 295–301 (2001)
16. Y. Kimura, Y. Miura, T. Shirosaki, T. Taniguchi, Y. Kazama, J. Hirokawa, M. Ando, A
low-cost and very compact wireless terminal integrated on the back of a waveguide planar
array for 26 GHz band fixed wireless access (FWA) systems. IEEE Trans. Antennas Propag.
53(8), 2456–2463 (2005)
17. S. Park, Y. Tsunemitsu, J. Hirokawa, M. Ando, Centre feed single layer slotted waveguide
array. IEEE Trans. Antennas Propag. 54(5), 1474–1480 (2006)
18. X.-P. Chen, K. Wu, L. Han, F. He, Low cost high gain planar array antennas for 60 GHz band
applications. IEEE Trans. Antennas Propag. 58(6), 2126–2129 (2010)
19. S. Cheng, H. Yousef, H. Kratz, 79 GHz slot antennas based on substrate integrated
waveguides (SIW) in a flexible printed circuit board. IEEE Trans. Antennas Propag. 57(1),
64–71 (2009)
20. Y. Miura, J. Hirokawa, M. Ando, Y. Shibuya, G. Yoshida, Double-layer full-corporate-feed
hollow-waveguide slot array antenna in the 60 GHz-band. IEEE Trans. Antennas Propag. 59
(8), 2844–2851 (2011)
21. T. Tomura, Y. Miura, M. Zhang, J. Hirokawa, M. Ando, A 45o linearly polarised hollow
waveguide corporate-feed slot array antenna in the 60 GHz band. IEEE Trans. Antennas
Propag. 60(8), 3640–3646 (2012)
22. M. Zhang, J. Hirokawa, M. Ando, Fabrication of a slotted waveguide array at 94 GHz by
diffusion bonding of laminated thin plates. IEICE Trans. Commun. E93-B(10), 2538–2544
(Oct 2010)
23. D. Kim, M. Zhang, J. Hirokawa, M. Ando, in Design of a Dual-Polarisation Waveguide Slot
Array Antenna Using Diffusion Bonding of Laminated Thin Plates for 60 GHz Band. IEEE
Antennas and Propagation Society International Symposium, 2012. [https://doi.org/10.1109/
aps.2012.6348506]
24. S.S. Oh, J.W. Lee, M.S. Song, Y.S. Kim, Two-layer slotted waveguide antenna array with
broad reflection/gain bandwidth at millimetre-wave frequencies. IEE Proc. Microwave
Antennas Propag. 51(5), 393–398 (2004)
Chapter 8
Frequency Scanned and Leaky-Wave
Linear Slot Arrays
8.1 Introduction
The aim of this chapter is to provide the reader with a thorough introduction to the
nature of the transition from ‘simple’ slot arrays in conventional waveguides to
often quite complex configurations, generated using advanced simulation tech-
niques. Such antennas cannot readily be analysed using the methods outlined in
Chaps. 5, 6, and 7. Additionally it is intended that the following sections will
provide a ‘bridge’ to the exploration in Chap. 9 of open periodic structures in so far
as they are employed to realise compact array forms, and thereby radiation pattern
optimisation. As we shall see, these antenna forms can be categorized as frequency
selective surfaces and are sometimes referred to as meta-structures.
In the previous Chapter (Sects. 7.2.1 and 7.2.2) we have already discovered that
for linear slot arrays implemented in conventional rectangular waveguide a beam
squinting propensity is exhibited with changing frequency. This is usually con-
sidered to be problematic, particularly in bandwidth terms, and certainly not
helpful. Nevertheless, a linear slot array which offers main beam positioning or
scanning that is highly sensitive to the antenna operating frequency can have very
definite and positive applications.
The evolution of the frequency scanned antenna was arguably propelled by the
desire to achieve range, height and azimuth information on a target from a single
radar platform. Unlike the active phased array antenna which can achieve this role
readily, the frequency scanned antenna option, by being wholly passive, is very
reliable and cost effective [1]. It is well known that slotted waveguide linear arrays
are troubled by beam squinting if operated over a significant bandwidth.
The essence of the problem is illustrated in Figs. 7.4 and 7.5. Normally radar
engineers would wish to minimise this difficulty, but in the frequency scanned
antenna the beam squinting is purposefully enhanced. This can be done in several
ways, given that scanning of the primary beam for a slot array in air filled
waveguide is governed by Eq. (7.11). That is:
b10 d 2mp
sin / ¼ ð8:1Þ
k0 d
for operation well above cut-off. Obviously the larger we make the relative per-
mittivity of the waveguide filling, the greater is the influence of b10d in Eq. (8.2)
thus augmenting the influence of frequency on /. Unfortunately, at typical fre-
quencies for long range radar operating in the L, S and C bands, this solution results
in feed structures which are unacceptably heavy and inefficient, when conventional
dielectric materials are employed.
The logical alternative to raising the relative permittivity of the guide filling to boost
the phase shift term in Eq. (8.1), is to simply increase the interior distance between
radiating apertures to lg (>d) so that Eq. (9.1) becomes:
b10 lg 2mp
sin / ¼ ð8:3Þ
k0 d
Fig. 8.1 Schematic demonstrating frequency scanning using a sinuous waveguide feed
which is generally termed the frequency scan equation. To scan the primary beam
over all of ‘real’ space from −90° to +90° then we must have −1 < sin / < 1.
Clearly this range can be achieved with low frequency deviation if lg d. Note
that d and lg can be adjusted independently thus permitting separate optimisation of
scan sensitivity and of grating lobe suppression.
For grating lobes (see circle diagrams in Chap. 6) at angles /n and /n+1
(say) Eq. (8.3) can be manipulated into the following grating lobe equation:
k0
sin /n sin /n þ 1 ¼ ð8:5Þ
d
182 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays
Clearly this equation is independent of lg which implies that the grating lobes of
the linear sinuous array antenna simply obey the strictures for linear arrays as
presented in Chap. 6.
In practice, frequency scanned array antennas based on the sinuous feed seldom
exist as stand-alone linear arrays. It is much more common to see the sinuous feed
combined with a parabolic reflector as shown in Fig. 8.2, or as an end feed to a
planar array of parallel slotted waveguides (see Chap. 7) as presented in Fig. 8.3.
These antennas provide 3-D scanning capability with azimuth information devel-
oped from the rotation of the structure as a whole while height information is
generated from the frequency scanned feed. Early versions were by no means
compact.
As we have seen in the preceding sections the key to enhanced frequency scanning
in a waveguide feed is the b10d term in Eq. (8.1) which needs to be much larger
than the kod term. Enhancement of b10 is possible by dielectric loading as
demonstrated in Sect. 8.2.1, while by differentiating the internal distance d ! lg
between apertures from the external spacing d (see Sect. 8.2.2) it is possible with
the resultant sinuous structure to improve scanning sensitivity by making lg d.
Unfortunately dielectric loading presents considerable antenna weight and poor
electrical efficiency, while serpentine feeds are bulky, incur aperture blockage
difficulties in reflector antennas (see Fig. 8.2), and the long sinuous waveguide
structure attracts power loss issues.
8.2 Frequency Scanned Slotted Waveguide Array 183
The feed structure for a slotted waveguide linear array possessing enhanced scan-
ning ability usually comprises a fin loaded rectangular waveguide which ‘slows’ the
propagating mode relative to empty waveguide. Hence the term ‘slow-wave’ feed.
184 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays
So, in order to progress, we need to be able to assess the influence of the period-
ically spaced metal fins or irises on the waveguide propagation characteristics. We
will attempt to do this in a manner which is physically intuitive. The literature on
the topic of slow-wave structures is very extensive and the early contributions, in
the 1950s and 1960s, are highly mathematical. So the subject can be somewhat
impenetrable without deep study.
For the original empty, perfectly conducting, rectangular waveguide of width a, and
height b, we discovered in Chap. 1 that the waveguide supports a dominant TE10
mode, which cuts off at a frequency given by koa = p, that is fc = c/2a Hz. This is
depicted in Fig. 7.14a (generally referred to as the x–b diagram) where the modal
frequency characteristics are presented diagrammatically, with the parabolic shape of
the curve representing the characteristic equation for the TE10 mode (see Appendix B):
p2
ðb10 aÞ2 ¼ ðk0 aÞ2 ð8:6Þ
a
Actually, in general the x–b diagram (Brillouin diagram) for empty rectangular
waveguide can exhibit many higher order solutions, with cut-offs reaching
increasingly higher values on the koa axis, representing all of the possible TEmn and
TMmn modes as encapsulated by Eq. (A2.13) in Appendix B.
Interestingly, when the rectangular waveguide is modified by the insertion of a
periodic array of metal fins (period = p, height = h) as suggested in Fig. 8.4, the
structure can still support an electromagnetic wave [6]. Not surprisingly its prop-
agation characteristics change significantly to accommodate the influence of the
fins. In short any valid modal solution for the periodic waveguide must satisfy the
electromagnetic boundary conditions on E and H over the fin surfaces as well as
over the waveguide inner surfaces, as encountered in the empty waveguide case.
Note, that if the TE10 mode waveguide contained only one metal iris or fin,
reflection and transmission of this mode at the iris would occur. However, while
these modal constituents would continue to satisfy the electromagnetic boundary
conditions on the waveguide walls they can only partially satisfy the boundary
conditions in the vicinity of the fin. Complete satisfaction of the boundary condi-
tions for the structure as a whole requires a suitable combination of higher order TE
and TM modes [6, 7]. These modes will be evanescent (see Chap. 1) and contribute
to capacitive stored energy fields in the immediate vicinity of the iris [4]. Now,
when a multiplicity of fins in a periodic array are present in the waveguide, as
hinted at in Chap. 7, the boundary requirements at each fin does not change and
multiple TE and TM waveguide modes will continue to be necessary to meet them.
However, the fact that the fins are actually closely spaced, in a periodic manner, sets
a concomitant requirement of periodicity on those modal fields which are in close
proximity to the surface (see Fig. 8.5) of the fin-loaded structure.
8.3 Transition from Uniform Waveguide to Slow-Waveguide Feed 185
Fig. 8.4 Frequency/phase diagrams (x–b diagrams) for a uniform rectangular waveguide—TE10
mode = red curve: TE20 mode = mauve curve
structure, has the property that the fields in adjacent cells, distance p apart, are
related by a complex constant. For a lossless structure, it can be expressed in the
form:
But, since f ðx; y; z þ pÞ f ðx; y; zÞ for a periodic function, then Eq. (8.9)
becomes:
Hence, on combining Eqs. (8.10) and (8.8) we observe that the Floquet rela-
tionship [Eq. (8.7)] is confirmed for the electromagnetic solution of a closed
periodic structure.
That the Floquet formulation is consistent with the Maxwell equations can be
deduced by considering a parallel problem in signal processing theory, namely the
representation of square or digital waveforms in the time domain by electrically
more convenient sinusoidal waveforms of varying frequencies. The mechanism for
generating this relationship is provided by Fourier series. The theory simply states
that any periodic function that is finite, single-valued and continuous can be rep-
resented by a series of trigonometric functions. Good pictorial representations of the
process are to be found on the internet [12]. Consequently, the periodic field
function E0(x, y, z) can by expanded into a Fourier series in the space domain
z (rather than time t as in signal processing) such that:
X
1
j2prz
E0 ðx; y; zÞ ¼ Er ðx; yÞ exp expðjb0 zÞ
r¼1
p
ð8:11Þ
X1
¼ En ðx; yÞ expðjbr zÞ
r¼1
8.3 Transition from Uniform Waveguide to Slow-Waveguide Feed 187
Fourier theory dictates that the amplitudes of the r space harmonics Er(x, y) of
the zero’th mode are given by,
Zp
1 j2prz
Er ðx; yÞ ¼ E0 ðx; y; zÞ exp dz ð8:12Þ
p p
0
and
2pr
br ¼ b0 þ ð8:13Þ
p
is the phase coefficient for the rth space harmonic ð1\r\1Þ. Note that in the
context of a fin loaded structure based on rectangular waveguide, as depicted in
Fig. 8.5, the zero’th mode is essentially an evolution of the TE10 mode of the empty
waveguide.
For higher order modes Emn(x, y) (m = 1, 2, 3,…, n = 0, 1, 2,…) the corre-
sponding possible solution forms are
X
1
j2prz
Em:n ðx; y; zÞ ¼ Er ðx; yÞ exp expðjb0 zÞ
r¼1
p
ð8:14Þ
X1
¼ Er ðx; yÞ expðjbr zÞ
r¼1
where b0 becomes the fundamental phase coefficient for the mnth mode, and the Er
represent the relevant harmonic magnitudes for this mode.
The question then remains as to whether or not the field forms represented by
Eqs. (8.13) and (8.14) satisfy the electromagnetic wave equation as they must? For
the electric field in z-directed waveguide this has the form (see Appendix B):
r 2 E c2 E ¼ 0 ð8:15Þ
A similar requirement applies to Em,n (x, y, z). Hence combining Eq. (8.16) with
Eq. (8.11) we obtain:
X
1 X
1
r2 Er ðx; yÞ expðjbr zÞ c2 Er ðx; yÞ expðjbr zÞ ¼ 0 ð8:17Þ
r¼1 r¼1
X
1
Therefore, the postulated field solution E0(x, y, z) for the periodic structure will
be a solution to the wave equation, and hence Maxwell’s equations, if each space
harmonic is also a solution. This will normally be the case since the space har-
monics are generated by solving Maxwell’s equations for individual periodic cells.
The cells are generally represented by resonant structures exhibiting much simpler
boundary conditions.
From an engineering perspective, the seemingly intractable complexity of
potential solutions to the electromagnetic boundary value problem, as presented by
a periodic structure, can be ameliorated significantly by noting that the structure is
composed of a linear assembly of identical cells (Fig. 8.5). Actually, the resonant
frequencies of these cells can be determined with little difficulty on an electro-
magnetic wave simulator such as HFSS [13] or COMSOL [14], and once known
the formation of the phase (x–b) diagram for the structure as a whole becomes
possible. Furthermore, when the cells are at resonance, only two logical possibilities
exist for neighbouring units. The fields either assume a symmetrical mode distri-
bution with zero phase shift (bp = 0) from cell-to-cell, or they assume an
anti-symmetrical form with p-phase shift (bp = p) from cell-to-cell. This behaviour
can be observed in low frequency LCR filter circuits [5, 15]. In filter terms the first
option represents a zero phase shift frequency cut-off condition while the second
equates to a cut-off condition exhibiting p-phase shift.
For the fin-loaded rectangular waveguide the lowest possible operating frequency is
determined by the fundamental (TE10) mode cut-off in the empty waveguide, given
by Eq. (8.6) with b10a = 0. At this frequency kg10 ) 1 which means that the gaps
between the fins have effectively zero depth electrically and the modal fields are
restricted to the empty region above the fins. That is, the periodic structure behaves
electromagnetically like an empty rectangular waveguide of height b and replicates
its frequency behaviour at cut-off (see Figs. 8.4 and 8.6) at bp = 0 and ba = 0
respectively). In cell modelling terms this resonance is generated when magnetic
walls (H-walls) are located above neighbouring irises, as suggested in Fig. 8.7. The
H-walls reflect the cell-to-cell zero phase shift requirement at this cut-off. The finite
element simulation predicts, not unexpectedly, 6.55 GHz (koa = p) in accordance
with Fig. 8.4 and with Eq. (8.6).
When an EM simulation is performed on the cell model with E-walls (Fig. 8.7b)
the principal bp = p frequency is predicted to be 9.87 GHz (koa = 4.726) with a full
half cycle of TE10 mode field pattern formed between the base of the fin-gap and the
top wall of the periodic structure. It is plotted on Fig. 8.6 as the first cut-off
8.3 Transition from Uniform Waveguide to Slow-Waveguide Feed 189
Fig. 8.6 Frequency/phase diagram (x–b diagram) for rectangular waveguide periodically loaded
with metal fins—showing fundamental plus first higher space harmonic (NB slow-wave solutions
to right of v = c line: fast-wave solutions occur in region to the left of v = c line—note in an open
structure (see Chap. 10) waves in the fast region become leaky)
maximum. Needless to say given the nature of the Floquet theorem, the bp = 0
frequency is repeated at bp ¼ 2p; 4p; . . .1, while the bp = p frequency is
repeated at bp ¼ 3p; 5p; . . .1.
The precise shape of the curve in Fig. 8.6 between the locations of key cut-off
frequencies, which have just been identified, requires the solution of the Maxwell
equations for the periodic structure as a whole. An investigator possessing elec-
tromagnetic simulation software such as HFSS [13] or COMSOL [14] could
attempt a direct application of a finite element, or finite difference technique,
although this route would demand considerable computational power to accom-
modate a model of the complexity presented by a periodically loaded waveguide.
The computationally more efficient alternative would to set up a problem specific
solution based on the moment method as outlined in Chap. 5, or perhaps on a
variational method as advocated by Cairo and Kahan [16] or an alternative MoM
[17]. Software such as MATLAB [18] can greatly facilitate the evolution of a
suitable algorithm to achieve this, particularly in the case of the moment method.
190 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays
Fig. 8.7 Periodic cell models showing E-wall and H-wall inserts. a H-wall, b E-wall
Here we wish to generate potential functions which are solutions of the electro-
magnetic wave equations within a fin loaded periodic waveguide (see Fig. 8.5). It is
evident that Fig. 8.5 represents a rectangular waveguide, as discussed in Chap. 1,
except that the lower broadwall presents an impedance other than zero to the
electromagnetic waves. Consequently it as appropriate to evaluate waveguide
propagation in this modified set-up.
In Chap. 1, it is demonstrated that the fundamental frequency characteristics of
rectangular waveguide are contained in the propagation equation [Eq. (1.68)],
which is also presented graphically in Fig. 1.6. For air filled waveguide with per-
fectly conducting walls this equation is:
mp2 np2
b2wmn ¼ k02 kcmn
2
¼ k02 ð8:19Þ
a b
for both TE and TM modes. If these are propagating solutions it is necessary that
the propagation coefficient for the mnth mode in the waveguide, bwmn is real, in
which case:
k0 [ kcmn and bmn \k0
Since bwmn ¼ x=vpmn and k0 ¼ x=c then clearly vpmn [ c. That is, the
solutions are ‘fast’ with the phase velocity of the mnth mode travelling at a speed
greater than that of light axially down the waveguide. So what conditions would be
required for the propagating modes to be ‘slow’ with vpm \c? Mathematically it
would require:
2 mp2
b2m ¼ k02 þ kcm
2
¼ k02 kcy ð8:20Þ
a
For slow-wave solutions we have already stipulated that bm > k0 and given that
all the terms on the right side of Eq. (8.21) are real, kcy must be imaginary (say jam).
This implies that in the fin loaded waveguide, in the space (0 < y < b) above the
fins, valid fields solutions must display exponential decay in the −y direction.
Embracing the evident analogy between the fin loaded rectangular waveguide
and its empty counterpart (Chap. 1) the following potential functions in phasor
form can be constructed for TE and TM mode solutions. For the mth such mode:
192 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays
rX
¼1
1 mp
zm ¼
ATE 2
½Ar sinhðamr yÞ þ Br coshðamr yÞ sin exp jbmr z
k
r¼1 cu
a
rX
¼1
ð8:22Þ
1 mp
ATM
zm ¼ 2
½Cr sinhðamr yÞ þ Dr coshðamr yÞ cos exp jbmr z
k
r¼1 cu
a
summed over an infinite number of Floquet space harmonics. Note that for higher
2 2
order space harmonics a2cmr ¼ b2mr þ mp a k0 . Also kcu2
¼ k02 þ a2cmr and
bmr ¼ bm0 þ 2rp p , while p = G + t.
On applying the Maxwell equations the following field expressions are gener-
ated for the mth mode in the rectangular space (a b) above the fins (denoted by
subscript ‘u’):
X1
1 n mp
Eux ¼ j b A mr þ jxl a cmr Dmr sinh acmr y
k2
r¼1 cu
a mr 0
mp o mpx
þ j bmr Bmr þ jxl0 acmr Cmr cosh acmr y cos exp jbmr z
a a
ð8:23Þ
X1
1 n mp
Euy ¼ ja cmr b mr A mr þ jxl 0 Dmr cosh acmr y
k2
r¼1 cu
a ð8:24Þ
mp o mpx
þ jacmr bmr Bmr jxl0 Cmr sinh acmr y sin exp jbmr z
a a
X
1
mpx
Euz ¼ ðAmr sinh acmr y þ Bmr cosh acmr yÞ sin exp jbmr z ð8:25Þ
r¼1
a
X1
1 n mp
Hux ¼ jxe 0 a cmr Amr þ j b Dmr cosh acnr y
k2
r¼1 cu
a mr ð8:26Þ
mp o mpx
þ jxe0 acmr Bmr þ j b Cmr sinh acmr y sin exp jbmr z
a mr a
X1
1 n mp
Huy ¼ 2
jxe0 Amr þ jbmr acmr Dmr sinh acmr y
k
r¼1 cu
a
mp o mpx
þ jxe0 Bmr þ jbmr acmr Cmr cosh acmr y cos exp jbmr z
a a
ð8:27Þ
X
1
mpx
Huz ¼ ðCmr sinh acmr y þ Dmr cosh acmr yÞ cos exp jbmr z ð8:28Þ
r¼1
a
8.3 Transition from Uniform Waveguide to Slow-Waveguide Feed 193
X1
1 mp np mpx npzi
Efx ¼ 2
K mn þ jxl 0 bgmn L mn cos sin sin bgmn ðy b hÞ
k
n¼0 cf
a G a G
ð8:29Þ
where it is presumed that the Floquet periodicity is built into the constants Kmn and
Lmn, for the mnth gap mode. This ensures that the gap modes are phase shifted from
one gap to the next in sympathy with the eventual mnth modal solution (bm0) for the
periodic structure as a whole. Thus Kmn (or Lmn) at the ith slot (say) is related to Kmn
in the first slot by,
ðiÞ
Kmn ¼ Kmn expðjibm0 ðG þ tÞÞ:
Also
X1
1 np mp mpx npzi
Efy ¼ 2
b gmn K mn þ jxl0 Lmn sin sin cos bgmn ðy b hÞ
k
n¼0 cf
G a a G
ð8:30Þ
X
1
mpx npzi
Efz ¼ Kmn sin cos sin bgmn ðy b hÞ ð8:31Þ
n¼0
a G
X1
1 mp np mpx npzi
Hfx ¼ 2
jxe b K
0 gmn mn Lmn sin cos cos bgmn ðy b hÞ
k
n¼0 cf
a G a G
ð8:32Þ
X1
1 mp np mpx npzi
Hfy ¼ 2
jxe 0 Kmn þ b gmn Lmn cos cos sin bgmn ðy b hÞ
k
n¼0 cf
a G a G
ð8:33Þ
X
1
mpx npzi
Hfz ¼ Lmn cos sin cos bgmn ðy b hÞ ð8:34Þ
n¼0
a G
194 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays
The phase coefficient bgmn applies to the low height gap-waveguide and is given
by the following equation (see Chap. 1):
mp2 np2
b2gmn ¼ k02 ð8:35Þ
a G
and
np 2
kcf2 ¼ k02 ð Þ ð8:36Þ
G
Note that for higher order modes in the gaps, the right-hand side of Eq. (8.35)
can become negative, in which case the bgmn terms in Eqs. (8.29) to (8.34) are
replaced by jagmn, when appropriate in the ensuing formulation. Also for n = 0,
bgm0 = bwm0.
The above partial field solutions, have been constructed by employing conventional
waveguide analogies in the upper space (u) and the fin gap (f), by applying the
electromagnetic field boundary conditions on the side walls at x = 0 and x = a, and
on the fin surfaces within the gaps, and at the bottom of the gaps (y = b + h). Thus
all are automatically satisfied. We are therefore left to ensure the boundary con-
dition Etan = 0 is satisfied at y = 0, that is:
(1) Ex = 0 and Ez = 0 at y = 0
Furthermore the boundary conditions at the periodic surface at y = b must be
observed. This implies that:
(2) Etan and Htan are continuous across the y = b interface. It is sufficient to take
care of the tangential field components only—since the Maxwell equations
ensure that the normal components fall into line.
The application of condition (1) immediately requires that:
Condition (2) requires that Eux ¼ Efx and Euz ¼ Efz at y = b in the range
0 < z < G + t. This is automatically the case at the fin tips (G < z < G + t) where
Eux ¼ Efx ¼ Euz ¼ Efz ¼ 0.
Continuity of Ez is enforced mathematically by performing an averaging process
over one period of the periodic surface. That is:
8.3 Transition from Uniform Waveguide to Slow-Waveguide Feed 195
Z
Gþt ZG
Euz expðjbmr zÞ dz ¼ Efz expðjbmn zÞ dz ð8:38Þ
0 0
On substituting Eqs. (8.25) and (8.31) into Eq. (8.38) and setting y = b yields:
ZG X
1
npz
ðG þ tÞAm0 sinh acm0 b ¼ Kmn sin bgmn h cos expðjbgmn zÞ dz ð8:39Þ
n¼0
G
0
Note that on the left of the above equation the space harmonic summation
disappears since the higher harmonics integrate to zero over the period p of the
corrugated surface. Am0 represents the fundamental term in the space harmonic
series for the mth slow-waveguide mode. This mode is neither TE nor TM. It is
generally referred to as the HE11 hybrid mode. Equation (8.39) can be reconfigured
to give:
X1
G jbm0 G
Am0 ¼ Kmn bm0 sinðbgmn hÞ exp Uðbm0 nÞ ð8:40Þ
ðG þ tÞ sinh acmr b n¼0 2
ðG þ tÞ h mp i
2
jbm0 Am0 þ jxl0 acm0 Dm0 sinh acm0 b
kcu G a
X1
np=Ghmp np i jb G
¼ 2
Kmn þ jxl0 bgmn Lmn sinðbgmn hÞ expð m0 ÞUðbm0 nÞ
n¼0
k cf a G 2
ð8:41Þ
Continuity of Hz at y = b generates:
X1 np
jbmr G n Lmn
j Dmr coshðacmr bÞ exp Vðbmr nÞ ¼ n cosðbgmn hÞ ð8:42Þ
n¼1
G 2 2
196 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays
X1
bmr h mp i b G
2
jxe 0 a cmr Amr þ j bmr D mr coshðacmr bÞ expðj mr ÞVðbmr nÞ
k
r¼1 cu
a 2
n h mp np i
¼ n2 jxe0 bgmn Kmn Lmn cosðbgmn hÞ
2kcf a G
ð8:43Þ
8
bmr < sinww m even
Vðbmr nÞ ¼
bmr ðnpG Þ2 : w
2 jcosw ð8:45Þ
m odd
2 ðn ¼ 0Þ
where wr ¼ bmr2 G and nn ¼ .
1 ðn [ 0Þ
Relatively routine algebraic manipulation of Eqs. (8.40), (8.41), 8.42) and (8.43),
leads to a pair of simultaneous equations for Kn and Ln (for the mth mode), which
can be summarised as outlined below. Note that in terms involving n, which occur
outside of the summations over n, it is convenient to replace the integer n with
integer s.
X
1
fWnrs Kns þ Xnrs Lns g ¼ 0
r¼1
ð8:46Þ
X1
fYnrs Kns þ Znrs Lns g ¼ 0
r¼1
8.3 Transition from Uniform Waveguide to Slow-Waveguide Feed 197
where
" #
coth acmr b X1 mpnp np2 k 2
Wnrs ¼ Vðbmr sÞ a G
b sin bgmn h
2 cu
Uðbmr nÞ
acmr n¼0
jxl0 mr G kcf2
1
Gþt cos bgms h coth acmr b X np2 k2
Xnrs ¼ dns ns þ Vðbmr sÞ bgmn cu Uðbmr nÞ
G 2 acmr n¼0
G kcf2
Gþt cos bgms h
Ynrs ¼ dns ns jxe0 bgms
G 2kcf2
" #
X1
ðmpa Þ2 ðnpG Þ2
2 coth acmr b jxe0 2
þ bmr Vðbmr sÞ bw10 sin bgmn h Uðbmr nÞ
acmr n¼0
k02 kcf2
Gþt mp np cos bgms h
Znrs ¼ dns ns
G a G 2kcf2
coth acmr b X1 mpnp
j a G bmr bgmn
þ b2m0 Vðbmr sÞ Uðbmr nÞ
acmr n¼0
kcf2
ð8:47Þ
The accuracy of any solution for the propagation behaviour of periodic waveg-
uide, based on Eq. (8.48), is dependent on the number of terms in the infinite series
that is retained in computations. Obviously there is a limit. Infinite series of similar
character are reported in Ref. [19], which provides useful pointers to possible sum-
mation techniques based on the calculus of residues. However, in the present context
we can test the validity of the heuristically deduced solution by noting that only the
first passband for the periodic structure with a single field variation in the x-direction
(m = 1) is significant. In this scenario drastic approximation can legitimately be made.
In Eq. (8.48) the only term of any significance is Ynrs, since this is the only term which
does not equate to zero when n = 0. Consequently, Eqs. (8.48) can justifiably be
approximated by setting YnrsKns = 0, to produce the following identity. It is pertinent
to note that the W and Z terms in Eq. (8.48) emanate from the equations for Ex and Hz
respectively. Thus equating them to zero to form (8.49) implies that Ex = 0 and Hz = 0
at y = b. In conjunction with the boundary conditions for E the implication is that Ex =
0 everywhere inside the fin-loaded guide. While this is approximately in accord with
reality for the fundamental mode, it is unlikely to remain valid for most high order
modes in higher passbands.
198 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays
ð8:49Þ
Equation (8.49) has been programmed for three space harmonics above the
corrugated surface and for three waveguide modes in the inter-fin gaps below the
corrugated surface. The results for the phase coefficient (b11) of the fundamental
mode HE11 in the guide, as a function of frequency, are presented in Fig. 8.8. Two
fin-loaded waveguide examples are examined, and when compared with mea-
surements, performed carefully on precisely machined prototypes, the level of
agreement between calculation and experiment is clearly extremely good. For the
fundamental mode it is fair to conclude that the heuristically deduced solution,
despite the evident approximations involved, is capable of providing a reliable
source of design information for fin loaded slow-waveguide structures. Of course
today, structures of this type can be comprehensively and accurately modelled by
resorting to full-wave electromagnetic solvers [12–14]. Nevertheless such solvers
have to be set up properly—a process which is open to errors—and the availability
of accessible equations such as Eq. (8.49) can provide a very useful check on the
modelling process.
Fig. 8.8 Predicted and measured phase coefficients, as a function of frequency, for a fin-loaded
rectangular waveguide a a = 72.14 mm, b = 12.24 mm, h = 21.8 mm, G = 4.76 mm,
t = 1.58 mm, b a = 64.5 mm, b = 12.24 mm, h = 26.5 mm, G = 11.11 mm, t = 1.58 mm
(curves = theory, points = measured)
8.4 Slot Radiators in a Periodic Waveguide Feed 199
The design of a scanned linear array antenna based on fin-loaded waveguide fed slot
radiators requires knowledge of the radiation properties of such slots, which may be
transversely or longitudinally directed, or inclined to the guide axis. Here we will
limit our attention to the transverse slot radiator, as shown schematically in Fig. 8.9,
which is fully representative of the computational process for slots in general. The
application of a moment method to this problem leads, as we have seen for the
conventional waveguide fed slot in Chap. 5, to a matrix equation of the form:
½ A ½C ½ a ½ h
¼ ð8:50Þ
½D ½B ½ a0 ½ 0
The same equation applies to the periodic waveguide example in Fig. 8.9 but the
matrix entries on the left of the equation require slight modification to reflect the
waveguide change. The entries are:
8 9
ZZ < ZZ =
Ais ¼ jxe0 fi ^ag n^ Gma ðr=r0 Þ þ Gmb ðr=r0 Þ fs ^
an dS0 dS ð8:51aÞ
: ;
S S
8 9
ZZ < ZZ =
0
Bis ¼ jxe0 fi0 ^ag ^0
n Gmb ðr=r0 Þ þ Gmc ðr=r0 Þ fs ^
an dS0 dS
: ;
S0 S0
ð8:51bÞ
8 9
ZZ < ZZ =
Cis ¼ jxe0 fi0 ^ag ^
n Gmb ðr=r0 Þ fs ^
an dS0 dS ð8:51cÞ
: ;
S S0
Fig. 8.9 Fin-loaded feed waveguide for transverse slot radiator in upper wall
200 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays
8 9
ZZ < ZZ =
0
Dis ¼ jxe0 fi ^ag ^0
n Gmb ðr=r0 Þ fs ^
an dS0 dS ð8:51dÞ
: ;
S0 S
ZZ
hi ¼ fi ^ag ½n
^ Hi ðrÞ dS ð8:51eÞ
S
The matrix entries Ais, Bis, Cis, Dis given mathematical expression in
Eqs. (8.51a) are essentially electromagnetic coupling terms giving mathematical
expression to the mutual coupling which occurs between the assumed current
distributions (fi,s) on the interior and external surfaces of the radiating aperture. For
the ‘slot-cavity’ and for the exterior half-space the magnetic dyadic Green’s
functions Gmb and Gmc respectively, remain unchanged from the conventional
waveguide moment method derivation in Chap. 5. On the other hand, the magnetic
dyadic Green’s function Gma for the periodic feed waveguide clearly differs from
Gma . Nevertheless it can readily be deduced from the plain waveguide form, as
elaborated in Appendix C. For the transverse slot shown in Fig. 8.9 the matrix
entries have the same meanings as for the slot radiator in conventional waveguide.
These are:
Ais represents self-coupling between magnetic current elements on the inner slot
surface S.
Bis represents self-coupling between magnetic current elements on the outer slot
surface S′.
Cis represents mutual coupling to the inner slot surface (S) due to magnetic current
elements on the outer surface (S′).
Dis represents mutual coupling to the outer slot surface (S′) due to magnetic current
elements on the inner surface (S).
Once the integrals in Eqs. (8.51a) have been evaluated, which is mathematically
routine but tedious [18, 20], the matrix Eq. (8.50) is amenable to solution by a
range of matrix manipulation tools. These can be found in standard mathematical
textbooks [21, 22] and in convenient to use computer software packages [12–14,
23]. The outcome of the process is plausible values for s, fs, a′s and as. These can
then be used to compute the scattering parameters for the compact radiator much as
outlined in Chap. 5.
Moment method computations have been performed on a fin-loaded waveguide
fabricated in WG10 (a = 72.14 mm, b = 34.04 mm) designed to provide a feed for
a frequency scanned linear array of transverse slots. These are summarised in
Fig. 8.10, where the results are compared with an equivalent set-up in conventional
waveguide (also WG10). In the figure the normalised power radiated by a centred
transverse slot is presented as a function of frequency, for four different slot lengths
of 30, 40, 42 and 46 mm, all at constant width (w = 4.5 mm). The solid (green)
curves denote the conventional waveguide case, while the dashed (blue) traces
8.4 Slot Radiators in a Periodic Waveguide Feed 201
Fig. 8.10 Relative power radiated as a function of frequency for a centred transverse slot in a
waveguide feed (solid curves) and a periodic waveguide feed (dashed curves) with slot length l as
parameter (l = 30, 40, 42, 46 mm: w = 4.5 mm) Waveguide WG10: a = 72.14 mm,
b = 34.04 mm Slow-waveguide: a = 72.14 mm, b = 12.24 mm, h = 21.8 mm, G = 4.7625 mm,
t = 1.5875 mm
The fin-loaded periodic waveguide has been shown to exhibit phase shift charac-
teristics, as a function of frequency (see Fig. 8.8), which are comparable with the
phase sensitivities manifested by dielectric loaded waveguide or sinuous waveg-
uide. Furthermore slots located in the top wall opposite to the periodic surface
(Fig. 8.9) can be excited to a level not unlike slots in a conventional waveguide
feed (Fig. 8.10). Consequently, as a feed for a frequency scanned linear array
antenna [10, 11, 24] the fin-loaded waveguide structure offers the advantages of
much less weight than the dielectric loaded option and much less bulk than the
serpentine waveguide alternative. Early attempts at forming fin-loaded waveguide
feeds for a frequency scanning radar were derailed by heavy propagation losses
[24]. However, it was eventually discovered that by adopting the highest accuracy
and control of the manufacturing process the periodic feeder loss can be made
comparable with a sinuous waveguide exhibiting similar scanning capability.
A slow-waveguide feed with the dimensions detailed in Fig. 8.8 [curve (a)] was
used to create a scanned linear array comprising six slots on the broadface. The
slots were of equal size, and with relatively low radiation per slot the field distri-
bution across the aperture was close to uniform. The measured beam width and
sidelobe levels for the array were consistent with this excitation set-up. The devi-
ation of the peak of the main radiated beam has also been investigated. The pre-
dicted squint based of the theoretical phase coefficient change with frequency
shown in Fig. 8.8 [curve (a)] was obtained using Eq. (8.1). It is plotted as a solid
blue curve in Fig. 8.11, and compared with measured results (red diamonds). The
degree of agreement further confirms the reliability of the developed formula
[Eq. (8.49)] for the phase coefficient of the slow-wave feed.
Careful and perceptive readers will have noted by now that electromagnetic
waveguides display a degree of commonality in their frequency characteristics. This
can largely be ascribed to the propagation equation which for conventional rect-
angular waveguide is given by Eq. (8.6). Standard waveguide text books [6, 7]
confirm this by demonstrating that the same equation applies to empty waveguides
of other cross-sections, such as circular, coaxial and elliptical. The only difference is
that the transverse resonance terms, (mp/a) and (np/b) in the rectangular case, must
be replaced by Bessel function roots or elliptic function roots. Consequently, it is
convenient to express Eq. (8.6) in a more general form, namely:
8.5 Uniform Leaky-Wave Antennas 203
Fig. 8.11 Beam scan as a function of frequency for a slow-wave fin-loaded waveguide feed
comprising six slots of the same size, equi-spaced by 65.4 mm (solid curves = theory, diamond
markers = measurement)
kz ¼ bm jam ð8:53Þ
for the mth mode for which kx ¼ mp a . bm denotes the axial phase shift per unit
distance in rad/m while am denotes axial attenuation rate in nepers/m. In this case
Eq. (8.52) requires that:
204 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays
if the space into which the electromagnetic waves leak is defined as y > 0 in
Fig. 8.12. In this direction the leaky wave field strength has the functional form:
The amy term if positive implies, rather curiously that in the y-direction the leaky
wave fields grow without limit. The consequent singular behaviour at infinity,
means that mathematically, while it is a solution of the homogeneous field equa-
tions, the leaky wave cannot be a proper mode of the source free Maxwell equa-
tions. Nevertheless, despite this ‘improperness’ leaky waves can still be valid
representations of the fields generated by an open structure in certain restricted
regions provided a source is present [25].
The situation can perhaps be clarified by reference to Fig. 8.12 which depicts a
leaky waveguide containing a long axially directed slit (commonly in the side wall
[25]) fed from the left from a closed waveguide and an intervening matching
transition. To the right, in the z-direction, the structure is assumed to be infinite. The
slit is located at y = 0 and extends to infinity in the z-direction along with the
waveguide. From Fig. 8.12 it is readily deduced that if the wave solution within the
waveguide exhibits a dominant ‘mode’ phase coefficient b0 then the wave com-
ponent parallel to the waveguide and travelling along the outer surface must match
it. This implies that:
8.5 Uniform Leaky-Wave Antennas 205
b0 k0
/ cos1 ¼ cos1 ð8:56Þ
k0 kg0
0:88k0
D/ ð8:57Þ
L sin /
In Eq. (8.56) kg0 is the wavelength of the dominant field solution in the guide.
Note that it demands that b0 \k0 . Consequently, leaky behaviour is linked to ‘fast’
waves and the radiation is launched into the forward quadrant. If the attenuation
coefficient for this solution is a0, then the field magnitude in the waveguide must
diminish as z increases, which in turn implies that the power available for radiation
decays. The resultant higher power density in the radiation fields at the input to the
slit is represented in the diagram by closely spaced rays, with the spacing growing
with increasing z. Consequently, it is not difficult to appreciate that if leaky radi-
ation occurs the radiation intensity, at any position z′, grows in the y-direction up to
the point y ¼ z0 tan /. This growth is in accordance with Eq. (8.55). In practical
terms the leaky wave radiation is confined to the region y\z tan /.
The key to long-slot leaky-wave antenna design in which the rectangular
waveguide feed is modified by a long axial slit, through which the modal energy in
the waveguide dissipates, lies in determining the propagation and attenuation
coefficients for the adapted waveguide. For long leaky slits which can be presumed,
with little error mathematically, to be of infinite extent, an electromagnetic
boundary problem is formed. The resultant boundary value problem is readily
amenable to solution by an electromagnetic solver either based on the moment
method, or on the finite element method [13, 14]. As we have noted previously,
while such solvers can be very accurate, if set up properly, it is generally good
engineering practice to anticipate their use by developing simple analogues which
are open to approximate field solutions. In this way a thorough insight into the
antenna characteristics becomes available, while at the same time furnishing a
method of checking the predictions of an electromagnetic solver if adopted.
For the long slot leaky-wave antenna with the aperture located in the sidewall of
the rectangular feed waveguide, an elegant and surprisingly accurate field solution,
using approximate methods, is provided by Goldstone and Oliner [25]. The analysis
employs a solidly established transverse resonance technique which recognises that
the non-axial propagation coefficients (kx, ky) in Eq. (8.52) represent the transverse
resonances of all possible modes. Transmission line techniques are used to model
the resultant field interactions in the transverse directions, including the slit dis-
continuity, represented as a complex impedance, between the transmission line
analogue and a half-space. For this class of leaky-wave antenna it is recognised that
the leaky-wave solution is essentially a perturbation of the electromagnetic mode
existing in the waveguide in the absence of the slit. Consequently, perturbation
206 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays
theory is used to establish ‘closed’ form equations to represent the real and
imaginary parts of the propagation coefficient in the leaky-wave structure.
For a rectangular waveguide perturbed by a slit running along the length of a
sidewall Eq. (8.52) can be expressed as:
where Dj is the perturbation on the closed waveguide value for the x-directed mode
number, and is assumed to be complex, which means that kz is complex and given by:
kz ¼ b ja ð8:59Þ
Here YT0 denotes the normalised terminating admittance of the transmission line
analogue. Expressions for the terminating impedance or admittance, where the slit
as ‘seen’ from the transverse propagation perspective can be viewed as a capacitive
iris, are available in many standard texts on electromagnetism [6, 7, 26]. Not
surprisingly they confirm that this admittance is strongly dependent on slit width
d. Reference [25] gives:
" #
GT p2 bð1 ðdbÞ2 Þ2 2p
G0T ¼ 0:285
Y0 16d cos2 pd2b
jd
" # ð8:62Þ
BT p2 bð1 ðdbÞ2 Þ2 2p
B0T ¼ 0:156
Y0 16d cos2 pd 2b
jd
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
with j ¼ k02 ðpaÞ2 b2 .
The phase shift constant b expressed as k0/kg, and the attenuation coefficient aa
are plotted in Fig. 8.13 as functions of the slit width d. The plots show clearly the
8.5 Uniform Leaky-Wave Antennas 207
Fig. 8.13 Guide wavelength and attenuation coefficient as a function of slit width d in a
rectangular waveguide with a = 22.86 mm and b = 10.16 mm at a frequency of 9.4 GHz. (Dashed
curves = approximate solution: Solid curves = exact solution)
dependence of a and b on the slit width, with the influence on the attenuation rate
more evident. However, more importantly the curves show minimal deviation from
the approximate closed form solutions (Eq. (8.54); dashed curves) from
super-positions which are nominally ‘exact’. From an engineering view-point it
cannot be too often stated that the availability of easily understood and easily
implemented solutions for a given class of antenna is immensely helpful to the
design process.
Once a and b have been ascertained to a requisite degree of accuracy, it becomes
a relatively straightforward matter to determine the radiation characteristics of the
leaky-wave antenna. Such characteristics include beam direction, beamwidth,
radiation efficiency and scan sensitivity to frequency. Since, in this section scanning
antennas are the focus of our attention we will limit ourselves to the scan sensitivity
issue. For long-slot antennas which are largely uniform in cross-section, and
unchanging geometrically in the axial direction, the scan behaviour usually depends
on whether or not the feed waveguide is air filled or partially dielectric filled.
Generally, the air filled option provides a superior scanning antenna. This is
explicable by directing one’s attention toward transverse wave numbers (kx, ky)
which in the air filled case are independent of frequency [see Eq. (8.52)]. The
implication is that the beamwidth of the radiation pattern of the leaky-wave antenna
does not change as the beam is scanned with frequency. This scan behaviour is
summarised in Eq. (8.56), which dictates that the beam scan angle / progresses
from broadside ðb=k0 ) 0:/ ¼ 90
Þ near cut-off, to endfire at high frequencies in
208 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays
the direction of propagation within the feed (+z-direction: b=k0 ) 1). Furthermore,
the application of elementary trigonometry to (8.56) yields:
b2
sin2 / ¼ 1 cos2 / ¼ 1
k02
b 2 k 2
But b2 ¼ k02 kc2 , consequently for an air filled feed 1 k0 ¼ c
k0 , and
therefore:
kc
/ ¼ sin1 ð8:63Þ
k0
5:53
D/ ð8:64Þ
kc L
Note that (8.64) confirms that the beamwidth is independent of frequency for the
air filled feed, as indicated above.
In the partially dielectric loaded feed increasing the frequency causes the field
distribution of the internal mode to become more and more confined to the
dielectric, weakening the slit excitation level. This process significantly degrades
the beamwidth which is inevitably considered to be a disadvantage. On the other
hand, it is well known that the phase shift coefficient in a dielectric loaded
waveguide exhibits an increasing magnitude with frequency, which is significantly
larger than for an equivalent empty waveguide. The consequence is, that with
dielectric loading, the uniform leaky waveguide antenna can be designed to scan
over a larger range of angles than the air filled alternative, for a prescribed fre-
quency range.
Fig. 8.14 Waveguide linear array antennas employing transverse (series) slots. a Travelling-wave
array, b leaky wave array
Fig. 8.15 Pictorial representation of the mutual coupling mechanism for a shunt slots and
b transverse or series slot in a linear waveguide array antenna
have been tried, most commonly in the form of inclined slots and transverse (series)
slots. The transverse slot in rectangular waveguide as shown in Fig. 8.14 most
clearly illustrates the difficulty of adopting slot forms other than the shunt variety,
so the following exposition is focused upon it.
The overwhelming reason for the low adoption of non-longitudinal slots in
waveguide arrays of this type is mutual coupling between neighbouring slots, which
is strongly influenced by slot orientation and positioning. The electromagnetic field
mechanism underpinning divergent levels of mutual coupling between notionally
similar slots in waveguide is readily explained by reference to Fig. 8.15. The figure
depicts a y-z plane cut of a rectangular waveguide feeding a pair of longitudinal
slots in sketch (a) and a pair of transverse slots in sketch (b). For narrow slots it is
now well established that the electric field induced in the slot by the exciting TE10
mode is almost wholly aligned in a direction normal to the major axis of the slot.
So, for the longitudinal slots in Fig. 8.15a the E-field in each slot is x-directed. This
means that in the y-z plane containing the slot the radiated E-field is also x-directed,
exhibiting the typical (dipole) pattern shape with nulls at the conducting surface of
the waveguide where the x-directed electric field is ‘shorted’. This pattern dictates
that at the neighbouring slot located d * ko/2 away the E-field is weak and mutual
coupling is very low. Similar observations apply to the internally scattered fields.
As a consequence shunt slots in waveguide represent almost ideal array elements
and shunt slot arrays can be investigated easily, as we have seen, using the con-
ventional array theory presented in Chap. 6.
On the other hand, the electromagnetic interaction between a neighbouring pair
of transverse (series) slots in the broadwall of a rectangular waveguide is by no
means insignificant, as Fig. 8.15b purports to demonstrate. In this case, with the
major axis of each slot aligned in the x-direction the induced E-field in both is
z-directed. Thus, in any y-z plane intercepting the slots the radiated E-field forms
omni-directional, semi-circular, equi-field contours as shown. This results in the
radiation power density propagating over the guide surface (z-direction) matching
the normally directed radiation (y-directed). Furthermore, the E-field for this
8.6 Periodic Leaky-Wave Antenna Arrays 211
z-directed surface radiation is normal to the waveguide wall. This behaviour means
that the electromagnetic coupling between slots is no longer weak unless the slots
are separated by more than the free space wavelength. Actually to form a broadside
mainbeam with an array of series slots, with no possibility of a p-phase shift
adjustment, as with shunt slots, slot separation is required to be *kg10. For an air
filled waveguide feed this equates to a slot interval d > ko. While this is advanta-
geous in lowering mutual coupling, it has the effect of permitting the formation of
undesirable grating lobes (see Chap. 6). It can be solved by dielectric loading of the
waveguide which has the effect of shrinking kg10. However, such a change simply
negates the attempt to reduce mutual coupling. Actually, dielectric loading is sel-
dom an acceptable solution due to weight and power loss penalties. It is perhaps
pertinent to note that conventional array theory (Chap. 6) can continue to be applied
to transverse slots afflicted by mutual coupling provided the radiation pattern of
each slot element in the array is not modelled as an isolated slot as in Chap. 5.
Rather the presence of coupled neighbouring slots is modelled by assuming that the
active slot is embedded in a ground plane of passive slots arranged in a periodic
grid representative of the array antenna design. Note that in planar slot arrays
transverse slot mutual coupling is largely unavoidable. Not surprisingly, for
inclined slots mutual coupling strength is a function of the inclination angle [1].
The consequence of the above is that transverse slot arrays can be realised only if
the mutual coupling between slots is viewed as instrumental to the array operation
rather than as disruptive to it. This means that the feed is no longer just a support
structure whereby the TE10 mode illuminates a sequence of slot radiators
(Fig. 8.14a), but forms an integral part of a slow-wave structure created by the
periodic array of slots. In addition to the TE/TM modes of rectangular waveguide
the ladder structure (Fig. 8.14b) can also support surface wave and leaky-wave
modes, which significantly contribute to the radiation characteristics of the antenna.
For the waveguide periodically loaded with slot radiators (Fig. 8.14b) the leaky
phenomenon, as we have seen, results in complex propagation coefficients for the
fundamental space harmonic of each mode—a and b (say) for the dominant mode
[27]. For a typical structure with b/a = 0.466, L/a = 1.0 theoretical estimates of
these parameters, normalised to a/ko and b/ko, are presented in Figs. 8.16 and 8.17.
In a manner reminiscent of the uniform leaky-wave antenna, radiation beamwidth
and radiation efficiency are influenced by the magnitude of a, so that operation near
the waveguide cut-off, where a is large, results in a wide primary beam, which
narrows as the frequency is raised (see Fig. 8.16). Above cut-off a rises rapidly to a
peak once finite power is able to enter the guide. It then falls away with increasing
frequency as the slot radiators, with L = a, become increasingly non-resonant. The
relationship between attenuation constant and frequency can obviously be tailored
by dimensional adjustments as is hinted at in Fig. 8.16.
Generally the ‘leaky’ slotted ladder in the broadwall of waveguide, as suggested
in Fig. 8.14, does not impose heavy loading on the waveguide mode. Consequently
the phase coefficient for the modified guide deviates little from the closed
waveguide prediction, as indicated in Fig. 8.17. The noteworthy observation, which
can be extracted from Fig. 8.17, is that the closed waveguide b-curve represents the
212 8 Frequency Scanned and Leaky-Wave Linear Slot Arrays
Fig. 8.16 Attenuation coefficient as a function of frequency for the dominant mode of a leaky
ladder of slot radiators (with L = a)
Fig. 8.17 Brillouin diagram for the dominant mode of a leaky ladder of slot radiators:
w/a = 0.0016, S/w = 5
8.6 Periodic Leaky-Wave Antenna Arrays 213
border between fast wave and slow wave solutions. Where the leaky wave curve
falls below this line, at approximately b = 0.6ko, the solution ceases to be leaky and
reverts to a trapped surface wave. Therefore, for the specified antenna geometry,
leaky wave performance is available only between cut-off at koa = p and koa = 4.
Finally, it is pertinent to note here that an important difference between the
uniform and the periodic leaky-wave antenna is that the dominant mode in the
former is a fast wave so that the structure radiates whenever it is coupled to
free-space. On the other hand, the dominant mode for a periodic leaky-wave
antenna is a slow-wave that is intrinsically non-radiating. However, ‘fast’ space
harmonics of a higher order mode can exist, and one of these can be a source of
leaky-waves in an unclosed structure such as Fig. 8.14b. Note that if a harmonic is
‘leaky’ the mode as a whole is ‘leaky’. Since the engineering aim is to create an
antenna that forms a single radiating beam the structure must be designed to favour
the mode which is potentially leaky. Sophisticated design of a periodic leaky-wave
antenna, combining a slow-wave structure with a leaky ladder of slots, can achieve
very large scan range, potentially over the angular range −90° < / < 90° by
operating over a frequency range which transitions from a backward space har-
monic to a forward space harmonic. This is not possible with the uniform slit based
alternative.
References
9.1 Introduction
As we have observed in earlier chapters for wire, dipole and patch antenna
elements on which an electric current density J exists (Chap. 2), the starting point
for a radiation calculation is Eq. (9.1), while for antenna elements of the aperture
type (slots, holes, slits and patch/slots) where radiation emanates from an E-field
usually referred to, by analogy, as the magnetic current density (Jm), Eq. (9.2)
represents the appropriate commencement of any calculation. It is, hopefully, also
clear from Chaps. 2, 3 and 4 that these equations lead to the powerful observation
that for any wire antenna or aperture antenna, if the current densities on a specified
antenna geometry are known and integrable, and if the dyadic Green’s functions
(see Table 4.1) for the coupled spaces are available, then near and far field radiation
patterns can be computed.
The introduction of passive arrays, such as reflectarrays and frequency selective
surfaces, into the analysis mix requires the application of Eqs. (9.1) and (9.2) to a
space, or spaces, which are not accommodated by the traditional antenna analysis
tools. A procedure for accommodating radiators, which form elements of a periodic
surface, is required [7]. We commence the process by replacing Eq. (9.1) by its
arguably more familiar usage as it applies to the Hertzian current element (see
Chap. 2). It has the form:
l0 I0 dl
A¼ expðjðxt k0 rÞÞ ð9:3Þ
4pr
We could have commenced with Eq. (9.2) and proceeded in a similar manner by
applying it to an isolated magnetic current element (i.e. an elemental slot).
However, we shall persevere with Eq. (9.3) since periodic surfaces used in antenna
pattern control are more commonly fabricated from multiple scatterers of the ‘wire’
based variety (see Fig. 9.1). Note that in Eq. (9.3) the term
218 9 Compact Planar Reflectarrays
Fig. 9.1 Plane wave illumination of passive periodic array of current elements
1
expðjðxt k0 rÞÞ
4pr
represents in essence the scalar Green’s function for free-space in phasor form. The
vector form is given in Table 4.1.
Generally speaking, the incorporation of frequency selective surfaces or periodic
surfaces into antenna technology aims to influence radiation characteristics by
making use of the controlled and adjustable scattering behaviour of the surface,
when illuminated by a plane electromagnetic wave. There are few antenna situa-
tions where plane wave illumination cannot be applied and therefore we concentrate
on this aspect. In Fig. 9.1 a plane electromagnetic wave is presumed to propagate
from the left, in direction kð¼ ^ak k0 Þ, where ^ak is the unit vector and k0 = x/c, and
illuminates an inclined flat surface supporting a regular lattice of dipole like scat-
terers. The periodic surface lies in the x-y plane of the x, y, z Cartesian system, with
x-directed metallic dipoles spaced Dx apart in the x-direction and Dy apart in the
y-direction. A more general system would have the dipoles directed at angle w to
the x-axis, but in such a case the dipole can be resolved into x-directed and
y-directed portions, so it is not too restrictive to limit consideration to the less
complex x-directed case. In the x-direction the elements are numbered from
1 \ p \ 1 while in the y-direction they are numbered from 1 \ q \ 1,
so we can identify the current on a typical element as ^ ax Ipq .
9.2 Passive Arrays 219
At the reference element Ioo the plane wave arrives in the direction.
Thus, if the incident plane wave k induces a current Ioo at the reference element,
then at a typical element located at p, q, where p is the element count in the
x-direction and q is the count in the y-direction, we must have, on applying the
Floquet theorem:
The exponential factors express the plane wave phase shift between element 0, 0
and element p, q. Note that the currents Ipq are also secondary sources of radiation,
and if the periodic surface is immersed in an incident plane wave of infinite extent,
the elements will radiate equally strongly.
To assess the scattering from the passive array as a whole, we first need to
evaluate the radiation at an observation point r(x, y, z) exterior to the array surface
from each current element p, q. Relative to this observation point the distance to the
element p, q is given by (see Fig. 9.1):
Consequently, by analogy with Eq. (9.3) it is possible to expect that the mag-
netic vector potential at r, due to Ipq, is Apq and that it can be expressed as:
l0 I0 dl
Apq ¼ expðjk0 rpq Þ ð9:7Þ
4prpq
Given that all of the current elements are aligned in the direction dlð¼ ^
ax dlÞ then
the magnetic vector potential Ap for the entire linear array at p, with 1\q\1
can be expressed as:
l0 I0 dl X
1
Ipq expðjk0 rpq Þ
Ap ¼ ð9:8Þ
4p q¼1 rpq
This can be formulated in terms of Ioo by substituting Eq. (9.5) into Eq. (9.8), to
give:
l0 I00 dl X1
expðjqky Dy Þ expðjk0 rpq Þ
Ap ¼ expðjpkx Dx Þ ð9:9Þ
4p q¼1
rpq
Generally the infinite sum in the equation is convergent, except where kyDy is a
multiple integer of p—a grating lobe condition. Unfortunately, the summing con-
vergence rate is very slow [6]. Relatively routine mathematical manipulation of the
220 9 Compact Planar Reflectarrays
summation in Eq. (9.9) can generate a more amenable version [6] which can be
expressed as:
X1
l0 I00 dl ky nk0 ð2Þ
Ap ¼ expðjpkx Dx Þ exp jk0 y þ H0 ðk0 asn Þ ð9:10Þ
4jDy n¼1
k o D y
where
a2 ¼ z2 þ ðpDx xÞ2 ð9:11Þ
and
" 2 #1=2
ky nk0
sn ¼ 1 þ ð9:12Þ
k0 Dy
ð2Þ
In (9.10), H0 is the Hankel function of order zero and of the second kind. While
Eq. (9.10) is useful, here it represents a step toward a solution for the infinite planar
array, since it is the adoption of planar periodic surfaces, rather than passive linear
arrays, that provide routes toward developing compact, electronically controllable
antennas. For the infinite planar array formed from an array of y-directed linear
arrays (q), it is clear that:
X
1
A¼ Aq ð9:13Þ
p¼1
P
1 h i P
1 h i
A ¼ l4jD
0 I00 dl k
y
exp jk0 y koy þ nk Dy
0 2p
Dx exp jk 0 x k0 þ
kx ik0
Dx
n¼1
i¼1
2 1=2 ð9:15Þ
expðjk0 z 1ðkk0x þ ikD0x Þ ðkk0y þ nkDy0 Þ
2
Þ
1=2
1ðk0 þ Dx Þ ðk0 þ Dy Þ
kx ik0 2 ky nk0 2
l0 I00 dl X 1 X1
expðjk0 r ^
ar Þ
A¼ ð9:16Þ
2jk0 Dx Dy n¼1 i¼1 sz
9.2 Passive Arrays 221
where r(x, y, z) is the vector denoting the reference position (Fig. 9.1) while the unit
vector is given by:
kx ik0 ky nk0 kz
^ar ¼ ^ax þ þ ^ay þ ^
az ð9:17Þ
k0 Dx ko Dy k0
Also
" 2 #1=2
kz kx ik0 2 ky nk0
¼ 1 þ þ ð9:18Þ
k0 k0 Dx k0 Dy
By analogy with Eq. (9.3) it is evident that the Eq. (9.16) represents, in effect,
the free-space Green’s function for a space bisected by a plane periodic surface.
Furthermore, relatively straightforward applications of the Maxwell equations (see
Chap. 4) to Eq. (9.16) generates the corresponding expressions for electric field and
magnetic field, namely:
I00 dl ^ar X
1 X1
expðjk0 r ^
ar Þ
H¼ ð9:20Þ
2Dx Dy n¼1 i¼1 sz
pffiffiffiffiffiffiffiffiffiffiffi
where g0 ¼ l0 =e0 . Note that Eqs. (9.16), (9.19) and (9.20) apply to both the
forward and backward scattering into the neighbouring half-spaces. A thorough
assessment of the infinite array Green’s function issue is presented in Ref. [7].
The nature of the solutions contained in Eqs. (9.16), (9.19) and (9.20) is depicted
schematically in Fig. 9.2. The plane passive periodic surface created by an array of
current elements (dipoles, patches) is supported by a dielectric sheet (yellow) which
lies in the x-y plane of the Cartesian coordinate system. The sheet is illuminated by
an incident plane electromagnetic wave propagating in the direction ^ ak illustrated
by the large blue arrow.
The aligned scatterers in the uniform passive array are equally excited by the
incoming plane wave and they adopt its phase profile over the array surface. The
initial response electromagnetically is specular scattering of a single reflected wave
into the incidence half-space (z < 0), and a single transmitted wave into the reverse
space (z > 0). This behaviour matches that of plane wave incidence on a smooth flat
semi-transparent screen. The reflected wave (red arrow) is formed in the direction
^ar , while the transmitted wave (green arrow) is directed along ^ ar þ . For a plane
screen this is the only scattering which occurs, whereas the periodic surface
generates additional outcomes depending on the magnitudes of the ratios Dx/ko and
Dy/ko.
222 9 Compact Planar Reflectarrays
Fig. 9.2 Electromagnetic wave scattering for FSS illuminated by a plane wave
The key equations are (9.17) and (9.18). When i = n = 0 in these equations, the
fundamental non-periodic scattering is being implemented. The equations reduce
to:
kx ky kz
^ar ¼ ^ax þ ^ay ^
az ð9:21Þ
k0 k0 k0
indicating, as anticipated, that two plane waves are scatterered in the directions
^ar , namely the transmitted wave ð^ar þ k0 Þ and the reflected wave ð^
ar k0 Þ. Note that
these directions are independent of the nature of the surface periodicity. On the
other hand, for large enough values for Dx and Dy it is possible for other values of
i and n to generate more real solutions for ^ar in which case additional scattered
waves can appear. These are the grating lobes (see Chap. 6) as shown schematically
by the orange coloured arrows in Fig. 9.2, approximating the particular case of
i = 0 and n = −1. Furthermore, it is clear from Eq. (9.18) that for high enough
values of i and n the propagation ratio kz/k0 can be imaginary (= jkz/k0), whereas kx
and ky remain real. This implies that above (+z) and below (−z) the periodic surface,
waves can exist which propagate freely in the x and y directions, but decay
exponentially in the z-directions. These solutions are termed surface waves, slow
9.2 Passive Arrays 223
waves or trapped waves. They are depicted in Fig. 9.2 as blue bands above and
below the array which fade monotonically with distance jzj. While in radiation
terms these surface waves are insignificant in magnitude a finite distance away from
the array, they represent vital stored energy in the array and are very influential to
the radiation control which the passive array surface provides.
9.3 Reflectarrays
The reflectarray provides a good example of a flat periodic surface being used to
control the scattering of incident electromagnetic waves. There are three basic
options (1) all metallic, (2) patch, (3) printed. The design principles underpinning
the reflectarray concept are fully explored in Refs. [1, 2] and are well rehearsed in
[3–6, 8–11]. Obviously, the essence of focusing electromagnetic waves with a
reflector is well understood and can readily be explained by reference to Fig. 9.3a.
In antenna terms, the requirement of the reflector is to reflect the spreading (un-
focused) rays from the source (usually a horn) so that waves emanating from the
source arrive at a plane surface ahead of the reflector with the same phase, i.e.
focused. This surface is termed the equiphase surface (Chap. 3). It is usually chosen
to coincide with the extremities of the reflector and is then viewed as the radiating
aperture of the antenna as a whole. For curved reflector systems the route to
forming an equiphase surface is to ensure that all rays from the horn or feed to this
‘aperture’ surface are equal in electrical length. This in turn dictates that the
reflector is parabolic in shape.
For the reflector in Fig. 9.3a, at the equiphase front, the total phase delay from
the horn phase centre is
for all possible ro and r1 values, where ko = 2p/ko is the free space propagation
coefficient. So provided the reflector has a parabolic profile a focused beam is
formed in the direction normal to the reflector (z-direction).
The reflectarray has been proposed as a means of producing focused reflection
using flat, low cost microstrip or stripline substrates. More recently reflectarrays
formed using all metallic grooved plates or 3-D dielectric printed protrusions on a
metal base have been advocated. In the microstrip option, it usually comprises a
periodic array of square resonant patches or narrow resonant slots, as suggested in
Fig. 9.3b. This patch/slot dual interpretation is possible since the radiation mech-
anism for a resonant patch is associated with the fields in the ‘open-circuited’ edges
(edge slots) of the patch (see Fig. 1.4). However if the patches are all the same each
patch reflects in mirror fashion so that the array as a whole acts like a flat mirror,
reflecting the radiated wave from the horn back towards it without modifying the
spreading unfocused pattern as Eqs. (9.17) and (9.21) predict.
224 9 Compact Planar Reflectarrays
Fig. 9.3 a Focused reflector antenna, b unfocused slot/patch array, c reflectarray focusing
where /mn counteracts the physical length differences for the rays from the horn to
the equiphase surface. Note that the 2Np factor incorporates the arbitrary and
electrically meaningless full cycle phase differences between well separated rays.
These are irrelevant to the design calculation since electrically 2p ¼ 360 0 .
The additional −p accommodates the reflection phase shift for a resonant patch
which is 180°.
To get a better appreciation of the dimensional constraints it is generally quite
helpful to consider a real example such as a 12 GHz design [3]. In this design the
9.3 Reflectarrays 225
flat array is required to subtend an angle of 72° at the horn phase centre to achieve a
compromise between excessive aperture blockage (high sidelobes) and excessive
distance (roo) which results in an unwieldy array and lowered bandwidth. The
square array dimensions are set at 10 cm 10 cm. The angle choice immediately
leads to the dimensions for roo and rnn as shown in Fig. 9.4. For an array with a
period of 1.22 cm the reflectarray in this case will comprise a 9 9 grid of patches.
When Eq. (9.2) is applied to this geometry we obtain, at 12.25 GHz, the fol-
lowing reflection phase requirements for a central patch and an edge patch. The
free-space wavelength at 12.25 GHz is ko = 2.45 cm. Therefore, for the 6.7 cm ray
there are 6.7/2.45 = 2.75 wavelengths between the horn phase centre and the central
patch of the array. The 2 full wavelengths can be ignored (2N term). Thus for the
central patch a reflection phase of /oo = 0.75ko = 0.75 360° = 270°. From this
we subtract 180° to allow for the reflection phase of a resonant patch. For the 8.2 cm
long ray to an edge-patch a similar calculation yields /mn = 0.347ko − 180° =
−55.1°. The establishment of the reflection phase settings for all patches, consistent
with forming a focused beam, is relatively routine (see Fig. 9.5) requiring only a
simple algorithm to step through all ray possibilities. This procedure is particularly
required for much larger arrays.
The phasing over the surface of a 30 30 element front fed reflectarray of
square patches, as computed by a commercial mathematics package, is depicted in
Fig. 9.5. The square elements of the array are colour coded to display the reflection
phases required by every element in order to produce a focused beam at broadside
to the reflectarray surface. The right hand scale demonstrates the link between the
colour range and reflection phase, with dark blue representing 0° and dark red/
brown representing 360°. A typical rippling colour pattern is depicted as the
226 9 Compact Planar Reflectarrays
Fig. 9.5 Array phasing to form a broadside beam for a 30 30 square reflectarray (0° = dark
blue, 360° = brown)
element phases compensate for increasing ray path length between the centre and
the periphery of the reflectarray surface. Note that the proximity of deep red and
dark blue patches does not indicate a large phase jump, since 360° equates to 0°.
Once the element phase requirements are known it becomes necessary to implement
these phases within the hardware. In the case of a patch array the patch size is used
with a reflection phase of 180° at resonance. In all metal arrays the groove depth is
tailored to the phase requirement, while in 3D printed arrays the protrusion heights
provide phase control. Note that the computation of the reflection phase shifts of
elements in an array cannot be determined from calculations on an isolated element,
because the mutual interactions from neighbouring elements in the array contribute
significantly to the result. However, for arrays of the order of 10 10 elements, or
more, the finite array can generally be replaced by an infinite array with seriously
influencing the behaviour of individual elements. This, in turn, means that periodic
array techniques can be adopted to determine reflection phase. The most efficient
route to a solution is by means of the moment method [6, 7] applied to an integral
equation for the unknown currents in the patch element, or the unknown fields at
the groove surface in the case of all metal arrays (see Chap. 5). Of course the
9.3 Reflectarrays 227
Fig. 9.6 Typical reflection phase versus patch size for 28 GHz reflectarray [3]
Fig. 9.7 COMSOL model to determine the patch reflection phase when embedded in an array
specification. Given that the relative position of the feed horn and the reflect array
are defined in order to compute the element phases, the precise illumination of the
array elements by the chosen feed, entails a determination of its radiation pattern.
The techniques for doing this are provided in Chap. 3. With this information,
backed by geometric optics, it becomes possible to compute the magnitude and
phase of the incident radiation at every element. Each element, of the now finite
reflectarray, can be viewed as a radiator with known excitation phase and signal
level. Hence, the radiation pattern of the reflectarray antenna as a whole can be
constructed from the array theory developed in Chap. 6. If the pattern fails to meet
the desired specification on an initial pass then further iterations will be required.
This computational procedure will generally lead to acceptable design [3].
Computed E-plane and H-plane patterns of just such an exercise are presented in
Fig. 9.8. These were calculated, as described above, for a 77 GHz reflectarray,
which is defined in Ref. [3]. The differences between the traces can be attributed to,
firstly, pattern differences in the E and H planes for the feed and, secondly, to the
patch elements which radiate differently in the E and H planes near endfire. For the
front mounted feed, blockage can differ between the two principal planes and may
be reflected, in addition, in the overall patterns. Comparisons with measurements,
which are generally excellent, are provided in the original paper.
At the time of compiling this text, in 2017, with the primary aim of collecting, in a
single volume, a compendium of design techniques which are generally now
available for creating antennas employing passive arrays of slots, the reflectarray
was featuring strongly in the literature. Some versions of the reflectarray naturally
emphasise the use of slots, while others based on patches, for example, can be
viewed as slot dependent insofar as patch radiation can be attributed to fields in the
edge ‘slots’ (see Chap. 1, Fig. 1.4). Consequently, it seems apposite at this point to
9.4 Reflectarray—State of the Art 229
Fig. 9.8 Typical H-plane a and E-plane b radiation patterns for a 77 GHz, 6 in. square
reflectarray, formed from 0.005 in. thick Duroid substrate with 5776 linearly polarized patch
elements
review recent publications on this topic while placing these contributions in the
context of earlier work on slot array antennas described in previous chapters.
The design principles underpinning the reflectarray concept are fully explored in
Ref. [3] and are well rehearsed in Refs. [8–11], which are summarized in Sect. 9.2.
230 9 Compact Planar Reflectarrays
The Pozar paper [3] provides three examples of antennas designed for use as high
frequency space-based sensors. These would in the past have been implemented
using parabolic dish technology, but have instead been designed around planar
reflectarrays, formed from microstrip patches. Such arrays are potentially more
convenient to transport and to install. Simple square patches are used in all three
designs with the required reflection phase adjustment at every patch being achieved
by dimensional modifications. The operational conditions for individual patches
may thus move from below resonance to above resonance, thus providing the
required reflection phase ranging from −180° to +180°. The examples presented in
[3] are of millimetre-wave designs—three operating at 28 GHz and two at 77 GHz.
Of the 28 GHz designs two are directed at similar 6 in. square reflectarrays but
fabricated onto different substrates—0.01 in. thick Taconic, and 0.02 in. thick RT
Duroid. The former required 528, 0.609 cm 0.667 cm, off-square patches, while
the latter was formed from 784, 0.544 0.544 square patches. In reported antenna
modelling outcomes, and in comparisons with measurement, the reflectarrays are
illuminated from the boresight direction from suitably designed circular corrugated
horns. The differences in performance between the two antennas are shown to be
small, and measurements are demonstrated to be in very good agreement with
theory, confirming the accuracy of the analysis method outlined in Sect. 9.2.
Comparisons are reported to have been made with an equivalent parabolic
reflector antenna, and generally indicate that any differences in performance are
slight, except for bandwidth, which is at best 5% for the reflectarray, much less than
for a parabolic reflector antenna which is limited mainly by the feed. The source of
this limited bandwidth is the resonant patches themselves which display impedance
bandwidths in the 5% range. Nevertheless, for antennas operating in the
millimetre-wave bands, a 5% bandwidth still provides considerable gross frequency
range, in GHz terms, and for many applications, it seems to be not too restrictive.
On the other hand, in the 12–14 GHz frequency band, for antennas in satellite
communications roles 5% is likely to represent a distinct disincentive to the
adoption of reflectarray technology.
The 77 GHz reflectarray [3] is presented both in prime focus, and in Cassegrain,
formats. The microstrip reflectarray itself is 6 in. square, fabricated on 5 thou-
sandth’s of an inch thick RT Duroid substrate, and required 5776 patches. Both
antenna types are shown to exhibit high gain, but the Cassegrain version is less
efficient largely because of higher ‘blockage’ generated by the sub-reflector. The
antennas are reported to exhibit −1 dB gain bandwidths of 2.5%. A typical example
of a prototype microstrip reflectarray in a Cassegrain geometry, and aimed at
operation in W-band (94 GHz), is shown in Fig. 9.9.
The bandwidth limitation inherent to the single layer microstrip patch reflec-
tarray is tackled in three ways in the literature. Firstly, non-linear changes in
reflection phase when presented as a function of patch size (see Sect. 9.2) is con-
sidered to be a major factor [3] in narrowing the bandwidth in patch based
reflectarrays, and it has been suggested that much more linear phase change could
be secured by using line length [8]. The proposed X-band array in this submission
[8] employs equi-size square patches and each patch is coupled through ground
9.4 Reflectarray—State of the Art 231
plane slots to a microstrip circuit on the reverse of the ground plane. This circuit
contains couplers and junctions linked to tailored line-lengths which establish the
required reflection phases at each patch in accordance with beam focusing
requirements. Needless to say, with the introduction of a microstrip circuit to
control reflection phases it is a small step to consider introducing active elements
[9] to enhance antenna performance. Such solutions are inevitably complex. From
the perspective of securing low costs in satellite communications and sensing
systems, there seems, at least in 2017, little advantage in taking this route.
Secondly, multiple substrate layers have been employed for bandwidth
enhancement in other patch related applications with adjacent patches on different
layers tuned to slightly different frequencies. The approach is advocated by Encinar
and Zornoza [11] who describe two-layer and three-layer printed reflectarrays. In
Ref. [10] two 40 cm circular reflect arrays with slightly different patch size patterns
are described. The paper observes that for a single layer reflectarray a reflection
phase range >300° can be secured by adopting as thin a substrate as possible, but
that this means that the phase, as a function of patch size, becomes highly
non-linear, thus limiting bandwidth. Modelled and measured results for patch
reflectarrays fabricated on two layers, show that with the outer patch of the
co-located patch pair is set smaller than the inner patch by 30%. It is observed that
reflection phase as a function of patch size can be linearised without degrading
phase range. A designed 40 cm reflectarray, front fed from an 11.95 GHz corru-
gated horn used in TV receivers, displays a creditable −1.5 dB gain bandwidth of
16.7%. However, when the two layer technology is applied to a 150 cm off-circular
reflectarray the bandwidth decreases significantly to *4% [9]. This is largely
attributed to insufficient reflection phase variation in order to accommodate the
much larger range of ray lengths from horn to patch. By introducing a third layer [9,
10] the reflection phase range can be increased from 400° to 800° [9]. With the
introduction of a three layer patch array the 1.5 m diameter reflectarray antenna is
shown to exhibit a credible 10% operating bandwidth [10] but obviously at the
expense of fabrication complexity.
232 9 Compact Planar Reflectarrays
Conceptually, the all-metal reflectarray is arguably a less complex entity than the
patch versions described in the previous section [14–19]. As with the patch based
reflectarray, formed from a flat microstrip substrate, the all-metal alternative is
configured from a flat metal sheet. In order to replicate the beam forming charac-
teristics of a parabolic reflector it must adhere to the reflection phase requirements
at the elements as represented by Eq. (9.23). Inevitably these adjustments at the
element level are achieved rather differently in the all metal alternative. There are
actually three reflection phase altering mechanisms which have been thoroughly
explored in the literature. These are (a) a flat metal plate with (usually) square
grooves of varying depth, (b) a flat conducting sheet with raised metallic
‘pill-boxes’ hollowed to form elemental shorted waveguides, and (c) a pixelated or
stepped structure without the hollowing. These alternatives are illustrated
schematically in Fig. 9.10.
9.4 Reflectarray—State of the Art 233
Fig. 9.10 2-D schematics of possible all-metal reflectarray antennas a grooved metal plate,
b hollowed ‘pill-boxes’ raised from surface of metal backing plate, c stepped or pixelated metal
plate
Case (a) The periodic surface in this option is strictly flat and comprises an array
of small rectangular recesses of varying depth. Radiation from the individual ele-
ments of the array emanates from a magnetic current density Jm at the outer surface
of the groove created by the incident illumination reflecting off its base. That is, the
periodic surface can be viewed as a conducting plane supporting a lattice of
magnetic current radiators. The arrangement is in electromagnetic terms the dual of
the patch reflectarray studied in Sect. 9.2, where the dielectric surface of the sub-
strate supports a lattice of electric current radiators. A solution paralleling the
procedure outlined for the microstrip reflectarray in Sect. 9.2 is evidently available.
A Green’s function which is in effect the dual of Eq. (9.15) has been formulated
elsewhere for the grooved flat metal plate (see Eqs. (9.9) and (9.10) in Ref. [20]).
This Green’s function is defined as applicable to the ‘semi-infinite’ array which
means an infinite array curtailed in size to the required reflectarray area but without
accommodating edge effect. This is not unreasonable for large arrays where edge
elements represent a very insignificant contribution to the overall array behaviour.
The availability of a suitable Green’s function means that it is not too difficult to
formulate an integral equation for the unknown aperture magnetic current, and the
moment method can, as described in Chap. 5 and in Sect. 9.2, be employed to
234 9 Compact Planar Reflectarrays
develop depth profiles for the recess elements, which acknowledges that they are
embedded within a periodic surface.
Actually, much of the literature directed at the all metal reflectarray, in the case
(a) configuration, is noticeably exercised by the limitations of the semi-infinite array
model [15–17]. The alternative solution accommodates edge effect by adopting a
modelling technique (T-block Method) which is largely borrowed from a decon-
struction procedure widely used in dispersion calculations for ridged metal
waveguide [5]. It is claimed to offer more rapid computations than earlier solutions.
The introduction of the T-block concept into the reflectarray analysis is depicted
schematically in Fig. 9.11. In the central region, elemental recesses in the metal
ground plane can be viewed as residing in a semi-infinite periodic surface, for
which the half-space Green’s function exists. Array scattering computations for this
portion of the array can proceed as outlined above. For the edge portions of the
periodic array an alternative procedure is required. In the T-block method each
groove on this part of the array is viewed as a separate element in a finite con-
ducting ground plane. Scattering solutions can be formulated for this geometry by
using diffraction theory for an aperture in a finite ground plane [21]. This is per-
formed for all edge grooves with their relative positions in the periodic lattice are
guaranteed by the T-block process [15–17]. T-block superposition is used subse-
quently to determine the scattering for the periodic surface as a whole. In some
contributions, this problem specific analysis is backed up by a full wave electro-
magnetic simulation based on the finite element technique, mainly as a check on
accuracy.
Beam focusing by means of the plane all-metal reflectarray antenna requires
reflection phase adjustment from individual grooves paralleling the microstrip
based reflectarray summarised in Eq. (9.23). The desired phase condition is illus-
trated in Fig. 9.12 where the reflection phase from each groove is adjusted using its
depth (d(n)) to achieve a condition which matches that of the equivalent parabolic
reflector antenna. Relative to the reflectarray coordinate system at (0, 0) the focus of
Fig. 9.12 Phase matching between all-metal reflect array and equivalent parabolic reflector
the parabolic curve is at (0, fo − do) where do is the distance from (0, 0) to the
lowest point on the parabola represented mathematically by the relation:
y ¼ a0 x 2 d0 ð9:24Þ
where b10 is the phase coefficient for the lowest mode in the groove waveguide, k0
is the free-space phase coefficient, d(n) is the depth of the nth groove, while f0 and d0
are defined on the diagram. The nth groove, in T-block notation, is defined as being
located distance T(n) from the origin. Consequently, if we take a0 = 1/(4f0), which
is not untypical of reflectors for very high frequency sensor systems, then the
groove depth requirement for a focused reflectarray is given by [5, 15]:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
kg10
d ðnÞ ¼ d0 þ f0 ðT ðnÞ Þ2 þ ðf0 d0 Þ2 ð9:26Þ
2k0
Fig. 9.13 Reflection phase computations for three all-metal reflectarray options
described by Chou et al. [14] where the variable height metal protrusions essentially
enclose shorted cylindrical waveguides, which operate not unlike the recesses in
[5]. The claimed advantage of this approach is that the reflectarray can be manu-
factured using a low cost die-cast process. However, this is off-set by the fact that
the array surface is no longer planar and consequently the semi-infinite array
approach to reflectarray analysis, employed earlier, is less valid.
A 48 cm 48 cm square reflectarray operating at 12.45 GHz and fed from an
off-set circular corrugated horn is described in [14]. A measured gain of 33 dB
together with a −3 dB gain/bandwidth of *3% is demonstrated.
Case (c) The elements for an all-metal reflectarray described by Lee et al. [18,
19] take the form of non-resonant square cross-section metal posts or cells rising
above the flat metal base plate of the reflector. The lattice of protrusions of varying
height results in a quasi-curved reflecting surface which generates a high gain beam
when illuminated from a horn feed. In effect what is created here is a parabolic
reflector with a ‘pixellated surface’ (Fig. 9.10). It can be designed using geometric
optics. It is not really a reflectarray in the conventional ‘flat plate’ sense. The
design, fabrication and measurement of a precision machined 7.2 cm 7.2 cm
square reflector operating at 94 GHz is described in the published papers.
Reflection phase calculations emanating from the surface of each step are simple.
For a step of height d above a reference ground level it is given by:
4pd
\q ¼ þp ð9:27Þ
k0
Clearly the reflection phase is a linear function of height d, and this is shown in
Fig. 9.13 (green dashed trace). However, it is also a linear function of frequency,
which inevitably curtails the available bandwidth, when compared with the
equivalent parabolic reflector.
A gain of *35 dB for the front fed antenna is reported in the literature [18]. The
−3 dB bandwidth is not specified but it is likely to closely match that of an
equivalent parabolic dish version of this mm-wave antenna. In fact, a Cassegrain
version of this antenna [19], which displays a slightly higher gain of 36 dB, has
been fabricated and measured and shown to have a creditable −1 dB gain/
bandwidth of 10.7%. A photograph of this reflectarray is reproduced in Fig. 9.15.
Fig. 9.15 Cassegrain illuminated all-metal reflectarray with pixelated or stepped periodic surface
(from [19])
Fig. 9.16 Plane wave illumination of 3D-printed reflectarray in off-set receiver geometry
dimensions and element sizes of reflectarray solutions, was deemed to favour this
evolving technology [6]. The paper describes the printing of array elements in the
form of variable height dielectric blocks onto a flat conducting surface. With the
blocks occupying the whole periodic cell—an option rather than a necessity—the
process results in a stepped surface formed from dielectric blocks of square
cross-section and varying height. Each step in the non-flat periodic surface acts as a
dielectric resonator if tuned to the array operating frequency with height used to
procure reflection phase adjustments, as suggested in Fig. 9.16. The reference plane
for phase determination is adjusted to coincide with the tallest block, and all blocks
are then adjusted by the same height in order to ensure that the reflection phase from
the tallest element at the centre of the array is 360°.
9.4 Reflectarray—State of the Art 239
Fig. 9.17 Phase and magnitude of reflection coefficient (q) for a dielectric block in 3D-printed
reflectarray
A drawback for the printed reflectarray is that unless the dielectric used is very
dense, and therefore heavy, only partial reflection occurs at the top surface of each
block. This is illustrated in Fig. 9.17. The transmitted component reflects off the
conducting base plate creating a resonant condition within the bloc. The
re-transmitted portion then interferes with the direct reflection from the block
resulting in the typical oscillatory reflection pattern (see Fig. 9.17) for the reflection
magnitude (green curve, triangles) and phase (blue curve, circles) as block height
changes. Interestingly, to correct this difficulty, the paper suggests that in the future
dielectric constant variations, rather than cell height, could represent a better
method of phase correction.
Three 100 GHz reflectarrays are reported to have been designed and fabricated
using different criteria for determining the dielectric resonator patterns. These arrays
are horn fed from a 30° off-set position. Gains of *25 dB and encouragingly high
−1 dB gain/bandwidth levels ranging from 14 to 20% are claimed.
interest in atmospheric monitoring from space at mm-wave frequencies and this has
inevitably influenced developments.
The relevant technologies which have evolved to date are as follows.
1. Double layer microstrip array as described in [22].
2. Variably shaped reflecting patch elements in single layer microstrip reflectarray
[4].
3. Die-cast metal reflectarray as described in [6].
The analysis of arrays of this description is dependent of the formulation of a
general mathematical representation of radiation in the presence of a passive
periodic surface. The basic steps towards establishing such a formulation are pre-
sented in Sect. 9.1. By commencing with the Hertzian potential function for radi-
ation from a current element, derived in Chap. 2, and by assuming, not
unreasonably, that the non-edge elements of a finite multi-element passive array are
effectively embedded in an infinite array, it is demonstrated that a Green’s function
for an element of the periodic array radiating into the surrounding space can be
formed. It then becomes possible to synthesise radiation patterns for typical passive
arrays such as the reflect-array. The phase compensation calculations required to
procure parabolic reflector type focusing from a ‘flat’ array are also studied.
The chapter is completed with a resume of the state-of-the-art in reflect-arrays.
References
1. D.G. Berry, R.G. Malech, W.A. Kennedy, The reflectarray antenna. IEEE Trans. Antennas
Propag. 11(10), 645–651 (1963)
2. J. Huang, J.A. Encinar, Reflectarray Antennas (Wiley-IEEE, Hoboken, NJ, 2008)
3. D.M. Pozar, S.D. Targonski, H.D. Syrigos, Design of millimetre wave microstrip
reflectarrays. IEEE Trans. Antennas Propag. 45(2), 287–296 (1997)
4. Hu Wenfei et al., 94 GHz dual-reflector antenna with reflectarray subreflector. IEEE Trans.
Antennas Propag. 57(10), 3043–3050 (2009)
5. Y.H. Cho, W.J. Byun, M.S. Song, High gain metal-only reflectarray antenna composed of
multiple rectangular grooves. IEEE Trans. Antennas Propag. 59(12), 4559–4568 (2011)
6. P. Nayeri et al., 3D printed dielectric reflectarrays: low cost high-gain antennas at
sub-millimetre waves. IEEE Trans. Antennas Propag. 62(4), 2000–2008 (2014)
7. B.A. Munk, Frequency Selective Surfaces (Wiley, New York, 2000)
8. A.W. Robinson, M.E. Bialkowski, H.J. Song, A passive reflectarray with dual-feed microstrip
patch elements. Microw. Opt. Tech. Lett. 23(7), 295–299 (1999)
9. A.W. Robinson, M.E. Bialkowski, H.J. Song, A 137-element active reflectarray concept.
Microw. Opt. Tech. Lett. 26(3), 147–151 (2000)
10. J.A. Encinar, Design of two layer printed reflectarrays using patches of variable size. IEEE
Trans. Antennas Propag. 49(10), 1403–1410 (2001)
11. J.A. Encinar, J.A. Zornoza, Broadband design of three-layer printed reflectarrays. IEEE Trans.
Antennas Propag. 51(7), 1662–1664 (2003)
12. J.H. Yoon et al., Single-layer reflectarray with combination of element types. Electron. Lett.
50(8), 574–576 (2014)
13. J.H. Yoon et al., Axially symmetric dual-reflectarray antennas. Electron. Lett. 50(13), 908–
910 (2014)
References 241
14. H.-T. Chou, C.-Y. Lin, M.-H. Wu, A high efficiency reflectarray antenna consisted of periodic
all-metallic elements for the Ku-band DTV applications. IEEE Antennas Wirel. Propag. Lett.
14, 1542–1545 (2015)
15. Y.H. Cho, W.J. Byun, M.S. Song, Metallic-rectangular-grooves based 2-D reflectarray
antenna excited by an open-ended parallel plate waveguide. IEEE Trans. Antennas Propag. 58
(5), 1788–1792 (2010)
16. Y.H. Cho, Transverse magnetic plane-wave scattering equations for infinite and semi-infinite
rectangular grooves in a conducting plane. IET Microw. Antennas Propag. 2(7), 704–710
(2008)
17. Y.H. Cho, TM plane-wave scattering from finite rectangular grooves in a conducting plane
using overlapping T-block method. IEEE Trans. Antennas Propag. 54(2), 746–749 (2006)
18. W. Lee et al., Non-resonant conductor reflectarray element for linear reflection phase.
Electron. Lett. 51(9), 669–671 (2015)
19. W. Lee, Y.J. Yoon, A broadband dual-metallic reflectarray antenna for millimetre wave
applications. IEEE Antennas Wirel. Propag. Lett. 15, 1–4 (2016)
20. J.P. Skinner, P.J. Collins, A one-sided version of the Poisson sum formula for semi-infinite
array Green’s function. IEEE Trans. Antennas Propag. 45(4), 601–607 (1997)
21. R.E. Collin, F.J. Zucker, Antenna Theory (McGraw-Hill Book Co., New York, 1969)
22. A.P. Feresidis, J.C. Vardaxoglou, High gain planar antenna using optimised partially
reflective surfaces. IEE Proceedings-Microwave, Antennas Propag. 148(6), 345–350 (Dec.
2001)
Chapter 10
Compact Planar Resonator Arrays
10.1 Introduction
Arguably the search for low cost antenna solutions, which embody aspects of
phased array technology, has underpinned recent interest in compact array antennas
(CRA) excited by the modal fields in a planar resonant cavity. This technology also
offers the intriguing possibility of securing what has been termed ‘configurability’,
which simply implies an ability to achieve some level of electronic positioning or
adjustment of the radiated beam. Electronic beam control is a technology which has
traditionally been restricted to full phased arrays. Intrinsically high gain, and effi-
ciency, represent a further attraction for CRA’s, as does the possibility of low
fabrication costs for antennas in the millimetre-wave range of the electromagnetic
spectrum.
In its original form, as described by Trentini [1], the planar resonator array was
referred to as a partially reflective surface (PRS) antenna and is founded on an
essentially rather simple concept. It comprises a Fabry-Pérot cavity resonator cre-
ated by the sandwich formed by the electrically conducting (perfect electric con-
ductor—PEC) antenna ground plane and an electromagnetically (‘porous’) periodic
surface (PRS) located approximately half of a wavelength above it. In principle the
structure, which can realistically be viewed as a parallel plate waveguide, may be
unconfined in the transverse directions; that is, parallel to the reflecting planes (x-
y plane in Fig. 10.1). An elementary feed antenna, which may be a microstrip patch
antenna or a ground plane slot, excites the system [1–4]. A weakly resonant
electromagnetic mode (usually TE10) is set up in the cavity due to the multiple
reflections between the reflecting layers. Curiously, antennas of this type have been
simultaneously and independently studied from the viewpoint of the leaky-wave
antenna (LWA), rather as if an alternative technology was being pursued [5–8].
However, this is strictly not true since the antennas in question are also based on the
Fabry-Perot resonator. They differ from the PRS class by trapping the electro-
magnetic resonance between a high impedance surface (HIS) or an artificial
Fig. 10.1 Radiation mechanism for a partially reflective surface on an infinite ground plane
a E-plane (y-z plane); b H-plane (x-z plane)
magnetic conductor (AMC), replacing the PEC in the PRS version. Electrically the
operating principles are interchangeable so that the PRS antenna is actually a
special case of the LWA. Consequently simplicity is served by viewing all antennas
of this type as cavity resonance antennas (CRA).
In order to realise a CRA it is essential that the PRS is only partly reflective to the
cavity mode fields, resulting in electromagnetic coupling to the external space
through periodically spaced apertures, or radiating elements, embedded in the PRS.
Depending in the rate of power leakage the finite volume containing meaningful
10.2 Cavity Resonator Arrays 245
stored energy in the Fabry-Perot resonator will rise or shrink to fulfil the dictates of
power conservation. This is depicted by the shaded areas in Fig. 10.1a, b. High
leakage will result in minimal or weak mode formation and hence low directivity
with only a small area of the PRS illuminated by the minimally established mode.
On the other hand a highly reflective surface, offering a low level of PRS trans-
mission, permits high energy accumulation before the power radiated from a now
larger area of the PRS balances the input power. This condition equates to high
directivity for the antenna as a whole [1, 3].
The original parallel plate cavity resonator, on which the PRS antenna is based, is
essentially a 2-D resonator trapping a TE10 mode (one half-cycle E-field variation in
the z-direction, invariant in the y-direction). If the waveguide is cut-off at the
resonant frequency the fields decay monotonically and rapidly in the x-direction
and contribute little to the resonance condition (c.f. Figure 10.1a). For the PRS
therefore, it is clear from Fig. 10.1 that the essentially two dimensional Fabry-Perot
cavity will support quasi-TEm0 modes which could be either y-polarised as shown,
or x-polarised. With a preselected reflector spacing of d * k/2 the quasi-TE10
mode, as depicted, will be formed. Note that while the quasi-TE10 mode produces a
largely monotonic field variation (in y) due to power loss through the PRS, as
suggested in Fig. 10.1a, for an infinite structure the x-directed propagation must
display a travelling wave behaviour [1] as shown in Fig. 10.1b. The same obser-
vation is true for the orthogonal TE10 mode if excited by an omni-directional
source. However, for d * k/2 the mode is very close to cut-off and the guide
wavelength for the fundamental mode, ky = kg10, will be very large. Thus from the
slowly varying excitation field, exhibiting a very long wavelength, a broadside
pencil beam is formed by the PRS for both modes. Clearly, in the Fig. 10.1a
representation, coupling apertures on the PRS will be excited in phase, resulting in
an intrinsically broadside mode of operation for the antenna in both planes. One the
other hand, the aperture field distribution forming the radiation pattern is not a
uniform one, since the mode, as explained above is concentrated in the central
region of the resonator. This is also true in the H-plane (Fig. 10.1b), if kg10 is long
enough, in which case essentially only one cycle is effective.
For a Fabry-Perot resonator operating in the n = 1 mode (quasi-TE10 mode) as
depicted in Fig. 10.2 the plate separation d dictates the internal reflection angle h in
accordance with the relationship:
pffiffiffiffi
2pd 2p er
¼ d cos h ¼ p ð10:1Þ
kz k0
246 10 Compact Planar Resonator Arrays
0:5k0
d ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð10:2Þ
er sin2 hr
Clearly for a broadside beam we require d k=2. The dimensions of the cavity
(d) and the relative permittivity of the dielectric spacer (er) require to be adjusted to
ensure higher order mode radiation lobes (secondary lobes) are suppressed partic-
ularly where scanning of the main beam is anticipated. Generally the scanning
requirement is that the main (n = 1) beam can be located anywhere between hr = 0
and hr = p/2 without the n = 2 mode forming a lobe in real space. The n = 1 mode
can be located at endfire when:
0:5k0
d ¼ pffiffiffiffiffiffiffiffiffiffiffiffi ð10:3Þ
er 1
and the n = 2 mode will be present, with a lobe located at broadside, if:
k0
d ¼ pffiffiffiffi ð10:4Þ
er
Equating (10.3) and (10.4) leads to the restriction that for beam scanning without
secondary lobes the relative permittivity must be greater than 1.333. The maximum
n = 1 beam scan angle, which can be secured without degrading the radiation
pattern with a higher order mode lobe, is then given by:
10.3 PRS Antenna 247
pffiffiffiffiffiffi
1 3er
hr ¼ sin ð10:5Þ
2
This condition should not be confused with the occurrence of grating lobes (see
Chap. 6) which is dictated by the element separation in the periodic surface of the
PRS antenna.
Scanning in the z-x plane is possible [5, 6] for a PRS antenna by frequency
adjustment. For a periodic element separation, or pitch, in the x-direction of
magnitude p, array theory in Chap. 6 informs us the when the main (n = 1) beam is
scanned to endfire in the forward direction a grating lobe at backward endfire is
avoided if p/k0 < 0.5. This generally represents the normal mode of operation of a
PRS.
The excitation of the PRS array is mainly shaped by the modal concentration, and
significantly it is only very slightly influenced by the feed. A radiation pattern with
a tendency toward low sidelobes is the result [1]. This can be illustrated very
effectively by performing a simple ray analysis of the parallel plate geometry. It is a
generally accepted principle in electromagnetic wave systems, as we have already
observed, that any legitimate field pattern can be resolved into a spectrum of plane
waves (see Chap. 3). In a rectangular box, such as a conventional waveguide, it is
shown in Chap. 1 that the waveguide mode patterns can be broken down into plane
wave components. For the parallel plate waveguide, which is just a rectangular
waveguide minus the top and bottom walls, the TE10 mode (see Fig. 10.1) is
formed from a pair of y-polarised plane waves repeatedly reflecting on the side
walls at z = 0 and z = d. However, for the PRS antenna the wall at z = d is only
partially reflective with a reflection coefficient qðhÞ\/ðhÞ which is a function of the
angle of incidence h of the constituent plane waves. This is illustrated in Fig. 10.2.
Transmission through the wall at z = d can therefore be expressed as:
1=2
s o ¼ 1 q2 ð10:6Þ
Hence, the radiated field pattern for the PRS antenna can be crudely expressed
as:
X
1
Erad ¼ E0 gðhÞqn ð1 q2 Þ1=2 expðjUn Þ ð10:8Þ
n¼0
where the angles Un represent the range of possible reflection angles at the PRS for
the n possible rays. In this equation g(h) is the element pattern for the source at (0,
0, 0) in Fig. 10.2, and
4pd
Un ¼ nH ¼ n / p cos h ð10:9Þ
ko
Since the reflection coefficient magnitude for the PRS is by definition less than
unity it follows that:
X
1
1
ðq exp jUn Þn ¼ ð10:10Þ
n¼0
1 q exp jH
On inserting Eq. (10.10) into Eq. (10.8) an equation for the radiated power
density (P(h)) emerges [1]:
ð1 q2 ðhÞÞ
PðhÞ ¼ g2 ðhÞ ð10:11Þ
1 þ q2 ðhÞ 2qðhÞ cos½/ðhÞ p 4pd
k0 cos h
Equation (10.11) emphasises the earlier qualitative observation that the larger
the magnitude of q the sharper is the primary beam resulting in high directivity.
The radiation mechanism could be viewed as electromagnetic seepage from a
standing wave (in the z-direction) within the parallel plate housing. For a patch
based PRS this mechanism has been illustrated in reference [4] by modelling the
near field 2 mm above the PRS (see Fig. 10.3). It is important to note that with a
dual polarised feed (square waveguide or square patch the modal patterns in
Fig. 10.1 will occur with Fig. 10.1a representing the y-z plane, and Fig. 10.1b
representing the x-z plane (TE01 mode replacing TE10). If these orthogonal modes
are excited simultaneously and in phase then a high gain, linearly polarised and
symmetrical pencil beam is formed at broadside. The leaky energy dissipation rate
results in an aperture field shape which is virtually free of sidelobes (see Chap. 6,
Sect. 6.3.2). If the feed modes are in quadrature phase the radiation will be cir-
cularly polarised for a PRS with dual-polarisation capability—for example a peri-
odic surface of square patches [4] or square apertures [3] or cross-slots [6].
10.4 Leaky-Wave Fabry-Perot Antenna 249
Fig. 10.3 PRS antenna near-field 2 mm above periodic surface. Note the concentration of energy
towards the mid-region (gridded) for a surface modelled as unrestricted in x and y (from [4])
It is apparent from just a cursory look at Fig. 10.4b that any serious theoretical
treatment of the cavity resonator antenna has to accommodate the presence of one
or more periodic surfaces. As we have noted in earlier chapters, while it is possible
with today’s ‘full-wave electromagnetic solvers’ based on either finite element or
finite difference techniques to ‘crunch out’ a solution to any given CRA geometry,
250 10 Compact Planar Resonator Arrays
Fig. 10.4 Sketch depicting the fundamental difference between a PRS antenna and a leaky-wave
antenna a PRS antenna with Fabry-Perot cavity formed by a PRS in parallel with a perfect
electrical conductor (PEC) or ground plane and separated by a distance of k/2 b LWA employing a
Fabry-Perot cavity formed from a PRS above a high impedance surface (HIS) and separated by a
distance < k/2 (h = k/4 if the HIS functions as a perfect magnetic conductor (PMC))
insight and understanding of the electromagnetic wave interactions within the CRA
usually entails modelling, typically based on the moment method. In Chap. 9 the
issue of electromagnetic scattering, and the formulation of a moment method model
in the presence of a periodic surface, also arose in connection with planar reflec-
tarray antennas. However, from a periodic surface perspective, while superficially
similarities exist between a CRA and a reflectarray, electromagnetically they call
for rather different solutions. With the passive reflectarray, scattering at the periodic
surface is made accessible to the moment method, by adopting the notion of the
infinite array. This then permits a degree of mathematical simplification by enabling
the imposition of Floquet’s theorem to the electromagnetic field representations and
hence the formulation of a Green’s function for an exterior space bounded by a
periodic surface. For relatively large (in wavelength terms) reflectarray surface
areas this is actually a rather insignificant approximation, as we have observed.
In contrast to the reflectarray, the CRA is an active antenna, for which the
Fabry-Perot resonator acts as an electromagnetic field transition between the ele-
mentary source, whether slot, patch, dipole or probe, and the antenna aperture
formed by the PRS (see Fig. 10.4a). Consequently, modelling for the CRA can
follow more conventional antenna array and moment method techniques as outlined
in Chaps. 3, 5 and 6. The essence of such a moment method analysis is detailed in
Chap. 5 Sect. 5.2. In simplistic terms it entails the computation of the coupling
between two electromagnetic systems such as a feed system (waveguide in Chap. 5
) and free space usually modelled as a half-space. The unknown quantity, which
needs to be established, is the field distribution in the coupling mechanism,
10.4 Leaky-Wave Fabry-Perot Antenna 251
kx ¼ k0 sin hr ð10:12Þ
The incident electric field and magnetic field components associated with the ko
directed plane wave are Einc and Hinc. In the left hand diagram Hinc is entirely
transverse to kx and is labelled as TM incidence, while in the right hand diagram
Einc is entirely transverse to kx generating TE incidence.
252 10 Compact Planar Resonator Arrays
Fig. 10.5 Fabry-Perot resonator antenna a illuminated by a plane wave from a distant test dipole
at r, h, / b PRS comprising a periodic array of slots
For a closed PRS, that is a surface with no slots, the plane wave reflects totally
on the outer surface and hence for TE incidence (see Fig. 10.6):
Fig. 10.6 Electromagnetic wave field components at CRA surface when illuminated by a plane
wave and transmission-line equivalent circuits
where Hxpþ denotes the x-directed magnetic field tangential to the outer surface of
the plate. Hxinc is the x-directed magnetic field component of the incident plane
wave. Furthermore for this surface wave directed along kx, E and H are related
through:
A similar relationship exists for the TM case. Given that the plane wave at the
CRA emanates from a test dipole at distance r from the antenna, the field Einc can
be expressed as:
254 10 Compact Planar Resonator Arrays
jxl0
Einc ¼ a^h expðjk0 rÞ expðjk0 ^
ar rÞ ð10:16Þ
4pr
The term inside the square brackets represents radiation from a point source, and
the second term ensures that, of all possible plane waves, the plane wave directed
along ^ar is mathematically selected. A similar equation can be formed for the TM
case with Einc directed along ^a/ . It is also helpful here to define Z0TE and Z0TM :
Hence:
For narrow x-directed slots as suggested in Fig. 10.5b the Hxinc field component,
generated by the TE illumination of the PRS, can in turn be associated with trial
magnetic current densities (unknown at this stage) in the antenna slots. These are
denoted by Jmx s
ðxp ; yq Þ on the slot array surfaces at z = 0 in real space, and by
~Jmx
s
ðkxp ; kyq Þ in the plane wave spectral domain. The integers p and q define slot
locations. For mathematical convenience, rather as outlined in Chap. 9, Sect. 9.1,
p and q are allowed to range in value from 1 to þ 1 since those slots outside
the limited number of active examples at the heart of the array contribute nothing to
the results.
The adoption of the infinite array concept in Sect. 9.1 yielded the analytical
benefit of making it possible to model the reflectarray with minimal complexity and
tedious repetition. When the array is of infinite extent its periodicity can be
expressed mathematically through the Floquet Theorem. Furthermore, considerable
mathematical simplification accrues to the CRA analysis presented below by per-
forming the computation in the spectral domain. When all relevant functions are
Fourier transformed into the spectral domain the array can be modelled quite
straightforwardly as summations over identically excited elements. The Floquet
modes are accommodated automatically within the Fourier transforming process. It
is well established in signal theory that an infinite sequence of identical ‘pulses’
transforms into a frequency spectrum of sinusoidal waves. By analogy the periodic
array of identical aperture fields is transformed into Floquet modes, which satisfy
the periodic boundary conditions at the substrate surface.
The x-directed magnetic current density Jmx s
ðxp ; yq Þ in the slots of the PRS
þ
generate a magnetic field Hxs adjacent to the outer surface (z > 0) of the PRS, and
simultaneously a magnetic field Hxs adjacent to the inner surface (z < 0). If the
spectral domain magnetic Green’s function ðG ~ ðkxp ; kyq ÞÞ for the periodic surface is
m
known, then we can write:
10.4 Leaky-Wave Fabry-Perot Antenna 255
1 X 1 X1
~ ðkxp ; kyq Þ~J s ðkxp ; kyq Þ exp jðkxp x þ kyq yÞ
Hxs ¼ G ð10:19Þ
p1 p2 p¼1 q¼1 m mx
where p1 and p2 are the element to element pitches in respectively the x and
y directions.
In the spectral domain, the above Green’s function can be deduced relatively
simply by employing a transmission line equivalent network (TEN) [10]. The
analogous circuits show downward wave propagation in the negative z-direction in
the radiation space z > 0, and transmission into the substrate z < 0 (Fig. 10.6). The
figure presents TEN’s for both the TM orientated incident plane wave and the TM
alternative. The more general case of oblique incidence is presented in reference
[10]. The transmission lines are presumed to extend to infinity in the +z direction
where they are terminated in Rrad the radiation resistance. In the air region the
transmission line characteristic impedances are identified as Z0TM and Z0TE , while in
the substrate we use Z1TM and Z1TE . At z = 0 on the PRS the scatterers, in our case
slots, are excited by the Hn (TM) and Hη (TE) of the obliquely incident plane wave,
where n is rotated angle d from the x axis. Consequently they can induce magnetic
current densities ~Jmx
TM
and ~Jmx
TE
in the x-aligned slots. Operation in the spectral do-
main is presumed. For slots ~Jmy is usually negligible so we can concentrate on ~ Jmx .
For z-directed propagation, the TEN relates the fields in the interior and the exterior
of the substrate and across the interface at z = 0 [10]. In this representation the
current density excitations at z = 0 can be replaced by the currents ~Imx TM
and ~Imx
TE
,
where ~Imx ¼ ~Jmx L and ~Imx ¼ ~Jmx L. Hence, on applying network theory to the
TM TM TE TE
equivalent circuits in Fig. 10.7, the voltage at the junction at z = 0 is given by:
~inTM ¼ ~Imx
V TM TM
Zin
ð10:20Þ
~in ¼ ~Imx ZinTE
V TE TE
where ZinTE;TM are the parallel impedances of the lines above and below z = 0. These
are respectively Z0TE;TM (see Eqs. (10.17), (10.18)) and
TE;TM
Zsub ¼ jZ1TE;TM tan b1 d ð10:21Þ
b1 xlr l0 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Z1TM ¼ and Z1TE ¼ and b1 ¼ x er e0 lr l0 ð10:22Þ
xer e0 b1
On combining Eqs. (10.20), (10.21) and (10.22), in the spectral domain, this
yields the relation:
256 10 Compact Planar Resonator Arrays
Fig. 10.7 Scanned E-plane (a), and H-plane (b) radiation patterns for a slot based CRA with a
dielectric filled resonator (er = 2.2) for which L = 0.6 cm, w = 0.05 cm, p1 = 1 cm and
p2 = 0.3 cm. The CRA is resonant at 11 GHz for d = 0.85 mm. Over the frequency range
depicted the height (d) varies from d k=2 to d k=3 (from [6])
10.4 Leaky-Wave Fabry-Perot Antenna 257
h i
~ xs ðkx ; ky Þ ¼ 1 kx2 IxTE þ ky2 IxTM ~
H J ðk ; k Þ ð10:23Þ
0 kt2 0 0 ms x y
where Vin is the voltage at the surface (Fig. 10.6). To form a Green’s function in the
spectral domain we choose Vin = 1 to represent the ‘point’ source. Hence
1
Ix j0 ¼ Yin ¼ Y0TE;TM ¼ ð10:25Þ
Z0TE;TM
Z0TE and Z0TM and hence Y0TE and Y0TM are defined in Eqs. (10.17) and (10.18).
Also
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
kz0 ¼ ðk02 kt2 Þ ¼ ðk02 kx2 ky2 Þ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð10:27Þ
kz1 ¼ ðk12 kt2 Þ ¼ ðk12 kx2 ky2 Þ
pffiffiffiffi
where k1 ¼ b1 ¼ er k0 .
Similarly, for the magnetic field just below the PRS and within the substrate
(region 1 with relative permittivity = er) the appropriate Green’s function is:
h i
~ ðkx ; ky Þ ¼ 1 jk 2 Y TE cotðkz1 dÞ þ jk 2 Y TM cotðkz1 dÞ
G ð10:28Þ
m
kt2 x 1 y 1
with
xlr l0
Z0TE ¼ ð10:29Þ
kz0
kz0
Z0TM ¼ ð10:30Þ
xer e0
The magnetic fields above and below the PRS can now be expressed in terms of
the unknown slot magnetic current densities. When the boundary conditions are
applied at a reference slot (usually the p = 0, q = 0 slot), that is:
258 10 Compact Planar Resonator Arrays
where Hxpþ is the tangential magnetic field on the ‘closed’ metal plate, Hxsþ is the
slot magnetic current induced field just above the slotted surface and Hxs is the
equivalent field just below the surface. At this stage we choose to follow the
Galerkin approach (see Chap. 5) to the implementation of the moment method and
employ entire domain basis functions in the trial form of the ‘unknown’ slot
magnetic current density Jms. Hence
X
N
Jms ðx; yÞ ¼ An Fn ðx; yÞ ð10:32Þ
n¼1
In the spectral domain the slot magnetic current density can therefore be written
as:
X
N X
N
~Jmx ¼ ~n ðkx ; ky Þ ¼
An F An~fn ðkx Þ~
gn ðky Þ ð10:34Þ
n¼1 n¼1
where:
Hence on applying the Galerkin method the following matrix equation for the
unknown An’s is generated:
2 ~m ðkx0 ; ky0 Þ
Rm ¼ Eyinc F ð10:38Þ
Z0TE;TM
10.4 Leaky-Wave Fabry-Perot Antenna 259
1 X 1 X1
~ þ ðkxp ; kyq ÞF
~n ðkxp ; kyq ÞF
~m ðkxp ; kyq Þ
Zmn ¼ G
p1 p2 p¼1 q¼1 m
ð10:39Þ
1 X 1 X1
~ ðkxp ; kyq ÞF
~n ðkxp ; kyq ÞF
~m ðkxp ; kyq Þ
þ G
p1 p2 p¼1 q¼1 m
Once the unknown An’s are determined by solving Eq. (10.37) the field Ey inside
the substrate at the position of the source can be calculated, and by reciprocity,
which involves sampling the cavity field, the far field becomes calculable.
Typical E-plane and H-plane far-field computations for slot based CRA’s at fre-
quencies near resonance (f = 11 GHz, d = 0.85 mm) have been generated and are
depicted in Fig. 10.7. The calculations apply to a substrate with er = 2.2 and for a PRS
for which L = 0.6 cm, w = 0.05 cm, p1 = 1 cm and p2 = 0.3 cm. For an infinite
parallel plate waveguide with a dielectric filling (er = 2.2), and a plate separation of
0.85 mm, propagation for the TE10 mode cuts-on at 11.89 GHz in the absence of slots.
The slot array lowers the resonance to 11.0 GHz, represented by the blue solid line in
Fig. 10.7a, b. At this frequency the radiation peak is precisely at broadside and is
sidelobe free for both the E-plane and the H-plane, as predicted by Trentini [1]. The
symmetry of Eq. (10.36) suggests that the E-plane and H-plane patterns for the CRA
should be the same. That they are not is due to the adoption of narrow slots to form the
PRS. As we have seen in earlier chapters the E-plane (y-z plane in Fig. 10.5) pattern
for a narrow slot in a half-space is semi-circular, while in the H-lane (x-z plane in
Fig. 10.5) it is a half ‘doughring’ with nulls along the major axis of the slot.
Consequently, in Fig. 10.7 the E-plane pattern exhibits (Fig. 10.7a) much stronger
radiation close to +90° and −90° than occurs in the H-plane (Fig. 10.7b).
More importantly Fig. 10.7 highlights the issue of beam splitting with a CRA, or
leaky wave antenna, and how the split beam maxima are influenced by antenna
thickness (d) and operating frequency. It is evident that the major influence on beam
split angles is the ratio of separation height (d) between the PRS and the ground
plane to the wavelength (k) within the substrate (i.e. d/k) and that in the H-plane it
is possible to locate a maximum beam anywhere between broadside and 75°. In the
E-plane it is predicted that the peaks can scan almost to end-fire, as indicated in
Fig. 10.7a. What is unexpected is that for 2-D antennas of this type so called
‘scanning’ is fundamentally associated with the formation of a split beam, partic-
ularly at scan angles well away from broadside. The mechanism for this behaviour
is suggested in Fig. 10.1. In the H-plane (Fig. 10.1b) a single beam, at broadside, is
formed only at resonance since this ensures that the fundamental parallel plate mode
is cut-off. This in turn implies that the TE mode in the positive and the negative
260 10 Compact Planar Resonator Arrays
x-directions in Fig. 10.1b exhibits zero phase shift. The radiating elements of the
PRS are thus excited in phase so forming a single broadside beam in both the x-
z plane and the y-z plane. At higher frequencies the parallel plate TE mode is no
longer cut-off in the +x and −x directions, for mid-point excitation as shown in
Fig. 10.1. Leaky-wave beams are thus possible in these directions resulting in the
2-D split beam patterns shown in Fig. 10.7. Note that since d and k are interlinked,
and essentially interchangeable in the basic mode of operation of the Fabry-Perot
resonator, beam splitting behaviour similar to Fig. 10.7 can be observed when
thickness adjustments (d) replace frequency [5, 6]. The beam-splitting is particu-
larly associated with the mid-point excitation commonly adopted with the CRA.
One directional scanning occurs if the array can be end fed.
This split beam characteristic represents an awkward feature of the leaky-wave
CRA, which is highlighted in references [5, 6], namely that the desired broadside
pencil-shaped radiating beam degrades rapidly with small changes to cavity height
from the resonant condition. Given that in theory the CRA periodic surface is
presumed to be of infinite extent the PRS could take the form of a circular plate
with radiating elements spreading radially outwards from the centre. Propagation
from a centrally located feed will also spread radially outwards at above resonance
frequencies. In three dimensions, at above cut-off, the split beam maxima of
Fig. 10.7 will occur along radii at all possible azimuth angles thus forming a hollow
conical beam with a circular rim. While some form of frequency scanning seems
possible, the obvious practical implication of conical beam formation, is that the
basic centre fed CRA antenna provides a broadside pencil beam radiation pattern
only over a relatively narrow bandwidth.
The narrowness of the bandwidth of the basic CRA is studied in reference [9] where
design guidelines are presented to tailor the CRA’s directivity bandwidth. This is
predicated on a thorough understanding of the role of higher order modes in the
leaky-wave CRA which was achieved by applying reciprocity, and the full-wave
spectral domain periodic analysis, as outlined above, to a 2-D geometry. Improved
bandwidth is possible, not by replacing the conducting ground plane by an HIS as
in Fig. 10.4b, but by introducing a second PRS inserted between the original PRS
and the ground plane forming in effect a pair of coupled Fabry-Perot cavities (c.f.
Fig. 10.8). The technique is not unlike patch antenna bandwidth enhancement
262 10 Compact Planar Resonator Arrays
which involves the stacking of two or perhaps three patch radiators in close
proximity to form a coupled resonator system. Bandwidth improvement follows
separate adjustments of the resonant frequencies of each patch. In the case of the
CRA, bandwidth enhancements of six times that of the single layer LWA, at the
same directivity level, can be accomplished.
High directivity, high gain, beam formation in a two dimensional leak-wave
antenna (2D-LWA) of the Fabry-Perot type is described in references [5–9]. The
basic format of the studied LWA is shown schematically in Fig. 10.8 with a PRS of
square patches on the underside of the upper substrate, separated by a distance
d from an HIS formed from a square array which has twice as many much smaller
square patches distance hd2 above a perfectly conducting ground plane. Actually,
the periodic surfaces representing the PRS and the HIS need not be square patches
but could be crossed dipoles, or crossed slots, or round holes, in a conducting
screen. In [5] the PRS is formed from printed dipoles, in [6] it is constructed from
slots etched into a plane conducting screen, while the antennas in [8, 9] are
developments based on square patches. Bandwidth enhancement, as indicated
above, is linked to the reduced thickness (d) which the HIS enforces.
High gain with improved bandwidth potential can be procured in theory by
omni-directional excitation of a circular plate shaped CRA. In Sect. 10.3, where the
basic mechanism of operation of the PRS antenna is examined (c.f. Fig. 10.1), the
explanation given there for leaky-wave radiation, presumes that the parallel plate
waveguide forming the antenna as a rectangular structure, which is best represented
by the Cartesian system. However, given that the Fabry-Perot cavity is theoretically
infinite in extent [1] this need not be the case. Cylindrical symmetry could equally
be assumed as is done in [5]. In this case the antenna is much more obviously a
travelling-wave device with the field pattern of Fig. 10.1b travelling radially out-
wards from the source. If the source is omni-directional (e.g. a coaxial probe) then
the dominant TM mode propagates radially outwards with equal magnitude in all
directions exciting the PRS as it does so. This operational model is tentatively
hinted at in Fig. 10.3. For an air dielectric the fundamental mode, as in the PRS
antenna case, cuts-on at d * ko/2, and a pencil beam at broadside is formed.
into the mix. The route to this goal has largely been through the introduction of
‘reconfigurable’ array elements into the PRS [11–16] and/or the HIS [17].
The methodology of electronic control in a CRA for beam scanning is hinted at
in the fundamental Eq. (10.11). However, this equation should be treated with
caution insofar as it is two dimensional and has been formulated using a simple ray
theory and multiple reflections. With q(h) denoting the internal PRS reflection
coefficient at incident angle h (from x) in Fig. 10.1a, it is clear that control of q is
key to pattern modification [7]. Note that the g(h) term is a magnitude factor only—
influencing energy distribution in cavity. It has no influence on phase at the radi-
ating surface. However, in practice control of q has been implemented by the
introduction of a second periodic surface (HIS) either replacing the ground plane or
inserted between the ground plane and the PRS as indicated in Fig. 10.9a.
Reconfigurability has generally been advanced by introducing phase agile cells into
this additional periodic surface rather than the PRS.
Arguably, the introduction of reconfigurable components into PRS antennas
appeared first as a frequency compensation technique [12], whereby PRS directivity
could be maintained over an enhanced tunable frequency range. An antenna
comprising a PRS formed from an array of square patches is described, and fre-
quency adjustment is achieved by introducing a phase-agile surface on a substrate
positioned close to, and just above the ground plane. The geometry is shown
schematically in Fig. 10.9, where the lower periodic surface is represented as being
‘phase agile’. This is achieved by employing split patches, termed reflection cells,
with the gap bridged by one or more varactor diodes (c.f. Fig. 10.9b). The
reflectivity of the reflection cells, and hence the phase, is controlled by the suitably
located varactor diodes. It is reported [12] that CRA gain can be maintained con-
sistently above 10 dBi over a tuned frequency range of 13.5% from 5.2 to
5.95 GHz. The bandwidth is six time larger than for the standard PRS antenna. By
further exploiting the principle enunciated in [7] and encapsulated in Eq. (10.11), it
has been suggested that dynamic beamwidth reconfiguration in PRS antennas is
feasible and a range of techniques are explored in references [13–15].
While the reconfigurable PRS has presented the primary route toward controlling
and optimising beamwidth and directivity, it has also been proposed [15] that PRS
beam positioning or scanning can, in principle, be procured by phase control of a
two element feed antenna embedded in the Fabry-Perot cavity as shown in
Fig. 10.10.
Referring back to Eq. (10.11), it has already been noted that, in the E-plane of
the Fabry-Perot resonator antenna the g(h) term, representing the feed, only indi-
rectly influences the radiation pattern P(h) and consequently, in this plane, phase
changes at the two element feed cannot modify P(h) and the antenna’s broadside
radiation characteristic. Consequently beam scanning in this case must be restricted
to the H-plane with the PRS excited from a linearly polarised feed. While the
simulated beam scanning predictions are presented for the H-plane, that scanning is
restricted to this plane is not made clear in references [14, 15]. It is fairly evident
from Fig. 10.1b that excitement of the PRS array elements is provided by the axial
(z-directed) magnetic field of the close-to-cut-off quasi-TE10 mode in the parallel
264 10 Compact Planar Resonator Arrays
Fig. 10.9 CRA with a phase agile HIS and b typical phase agile cell incorporating varactor diode
switches
10.7 Reconfigurable PRS for Fixed Frequency Scanning 265
Fig. 10.10 Schematic of a high directivity PRS antenna with additional beam scanning in the
H-plane (from [15])
plate waveguide. If the PRS is reconfigurable as suggested in Fig. 10.11, this allows
the surface reflection coefficient to be adjusted by controlling the capacitances of
the varactor diodes. The introduction of a two-element feed can be made to favour
the propagation of this mode in the forward (+z) or backward (−z) directions, thus
providing a scanning mechanism which results in three beam positions in the
H-plane given by the basic leaky-wave equation.
266 10 Compact Planar Resonator Arrays
Fig. 10.11 Schematic depicting diode modified patch element for typical reconfigurable patch
based PRS antenna (from [15])
k
H ¼ cos1 ð10:40Þ
kg
Electronic beam scanning in a CRA represents active research at the time of writing
in 2017 but solutions are proving to be elusive. While a range of modifications,
which have been introduced into the original Fabry-Perot resonator, such as addi-
tional periodic layers variously referred to as an HIS, AMC, FSS or metamaterial,
have achieved improved bandwidths, gain and efficiency by controlling
beam-splitting, frequency independent scanning has seemingly remained out of
reach. Nevertheless, comprehensive modelling of the CRA points to the leaky-wave
mechanism as the most promising route to effective non-phased-array beam scan-
ning at a fixed frequency.
Leaky-wave scanning performance of a cavity antenna, based on the application
of reconfiguration to a PRS layer, is partly addressed in [16]. Careful inspection of
Fig. 10.1b reveals that for a non-directional feed, an in-phase, close-to-cut-off,
quasi-TE10 mode field propagates in the PRS antenna to the left and right of the
source; that is in the −z and +z directions. The PRS thus splits into two ‘half-arrays’
whose in-phase radiations add constructively into the free-space region contributing
in the H-plane to the predicted [1] broadside pencil beam. In reference [16] this split
operational mode is emphasised by introducing slightly different array geometries
in the left and right regions of the PRS antenna. In summary, the arrangement has
arguably produced a limited 1-D leaky-wave scan mechanism in the H-plane of the
cavity resonator antenna which was originally presented as a PRS antenna. The
investigation outlined in [16] suggests that more significant and more practical
levels of scanning in cavity resonator antennas require more emphasis on the
leaky-wave mechanism. This is addressed in the next section.
Fig. 10.12 a Schematic of 1-D reconfigurable leaky-wave antenna b detail of HIS cell (from [17])
replaced by a perfect conductor, then the TE10 mode as depicted in Fig. 10.1 would
be set up in the 1-D leaky-wave antenna. However, the HIS in its passive state will
likely exhibit a high impedance approaching that of an open circuit or a perfect
magnetic wall, in which case the TE10 mode becomes a TE1/2,0 at resonance, with
the tangential electric field progressing from a maximum at the HIS to approxi-
mately zero at the PRS.
The situation is represented schematically in the accompanying Brillouin dia-
gram (Fig. 10.13). The lowest modes in the parallel plate waveguide, namely the
TEM mode and the TE10 mode, are required to meet the electromagnetic boundary
conditions presented by the periodic inserts which introduces space harmonics and
slow-wave modes to the right of the ‘light’ line (dashed green trace) and
leaky-wave modes to the left. The original TEM mode can generally be identified
with a slow TM mode (yellow dashed traces in Fig. 10.13), while the empty
waveguide TE10 mode transmutes into the lowest TE slow mode, retaining the
predictable cut-off frequencies associated with the conventional waveguide. So, a
leaky wave antenna array supporting a TE mode with a low cut-off frequency at x1
will radiate at broadside at this frequency, and at endfire just above x4, as dictated
by Eq. (10.40). Thus frequency dependent beam scanning proceeds away from
broadside toward endfire. Any further increase in frequency moves the operational
mode into the slow-wave region where the propagation mechanism becomes a
surface wave, and radiation is suppressed. Actually, in a reconfigurable geometry,
leaky wave radiation is recoverable at higher frequencies by capacitance changes in
10.8 Beam Scanning in a Leaky-Wave CRA 269
the varactor diodes shown in Fig. 10.12. These modify the HIS impedance such
that the operational mode is shifted to a higher resonance or cut-off frequency.
But, more importantly, the Brillouin diagram shows how frequency independent
beam scanning can be realised with a reconfigurable HIS. At the frequency xs for
example, if the varactor diode adjustment is capable of shifting the TE mode
frequency characteristics to the positions shown, then the propagation coefficient
for the leaky wave mode is moved from bp1 to bp2 to bp3. The beam scans from
endfire to broadside. In this 1-D format the antenna is largely omni-directional in
the y-z plane (E-plane) above the parallel plates. The main-beam, which is scanned
in the H-plane (x-z plane), is therefore ‘fan’ shaped. In an LWA designed to operate
at 5.6 GHz, a beam scan range in the H-plane of 8°–30° from broadside is
demonstrated without loss of gain (Fig. 10.14). The influence of HIS changes on
the radiation characteristics of similar leaky-wave CRA’s are also reported in [11].
To avoid beam splitting (see Fig. 10.9) the 1-D structure is end fed from rectangular
waveguide. This represents a complete departure from the normally centre-fed CRA
geometry.
That the 1D leaky wave antenna depicted in Fig. 10.12 possesses the range of
adjustments required to generate effective H-plane beam scanning is demonstrated
in reference [18]. While the evaluation did not include diode switching within the
studied device, it is shown that predictable and reliable positioning of the primary
beam can be produced at a fixed frequency by simple adjustments in the length of
the dipoles forming the PRS (LPRS), and the length of those forming the HIS (LHIS).
It is anticipated that these length modifications could be implemented electronically
by embedded varactor diodes as illustrated in Fig. 10.13. Moment method
Fig. 10.13 Schematic representation of the Brillouin diagram for a parallel plate waveguide
loaded with transverse periodic surfaces
270 10 Compact Planar Resonator Arrays
Fig. 10.14 Computed beam scan angles for 15 GHz LWA’s design to provide 10° beamwidths at
scanned angles of 5° (blue long dashed trace), 20° (green solid trace), 40° (brown dashed trace)
and 60° (red short dashed trace)
computations, which have been performed on several versions of the LWA, are
summarised in Fig. 10.14. The analysis has been confirmed by measurements [18].
Largely in parallel with the linear LWA developments described in the previous
section, the introduction of a reconfigurable HIS into a planar antenna geometry has
been reported [19–22] with the aim of forming a reflecting surface for use as a
reflectarray, or as a compact conformal scanned LWA. However, while competing
developments in CRA technology have tended to separate the PRS from the HIS,
within the parallel plate structure, and apply reconfigurability by means of the HIS,
in the cited articles varactor diode control is incorporated into a periodic structure
which notionally combines the two roles. The arrangement is described variously as
a tunable textured ground plane, a tunable impedance surface and an artificial
magnetic conductor. It is depicted schematically in Fig. 10.15.
The periodic surface comprises a regular array of square patches above a con-
ducting ground plane, not unlike CRA’s described earlier in this chapter. However,
the inclusion of shorting posts between the patches and the ground plane makes the
notion of a parallel plate operating mechanism less sustainable. Nevertheless, ref-
erences [21, 22] suggest that the Brillouin diagram for this periodic surface exhibits
10.8 Beam Scanning in a Leaky-Wave CRA 271
jxL
Zs ¼ ð10:41Þ
ð1 x2 LC Þ
Close to resonance, the surface impedance Zs tends towards infinity and the
periodic structure behaves as an HIS, or artificial magnetic conductor, with a zero
reflection phase shift. The term textured surface has been introduced to reflect the
non-uniformity of the Zs for this structural format.
As with the linear leaky-wave antenna described in the previous section, scan-
ning can be either frequency dependent, or stepped at fixed frequency by varactor
272 10 Compact Planar Resonator Arrays
diode adjustment. In the former case, increasing frequency from the dominant TE
mode cut-off results in growing bp (Fig. 10.13) until the chosen curve crosses the
‘light’ line (vp = c). In this frequency range the TE mode is leaky and according to
Eq. (10.40) the primary radiating beam scans from broadside to end-fire. However,
the inclusion of varactor diodes as suggested in Fig. 10.15, confers reconfigura-
bility on the periodic surface, by changing patch-to-patch capacitance and hence
resonance (see Eq. (10.41)). This manifests itself as repositioning of TE mode
cut-off frequencies from x1 to x4 in Fig. 10.13. So, in much the same manner as for
the linear leaky-wave structure, at a fixed frequency xs (say) diode adjustment can
move the phase shift coefficient bp from bp1 to bp2 to bp3, thus scanning the
leaky-wave beam. Scanning in the forward direction of the incident mode is clearly
demonstrated in reference [21]. Furthermore, it is also suggested there that the
textured surface of Fig. 10.15 also supports a backward TE mode which enables
scanning into the backward sector from 0° to −50°.
In practical examples of this textured surface leaky-wave antenna, excitation of
the surface relies on either an off-set flat notch shaped feed located just above it
(by *ko/50) [21], or on a flat bow-tie dipole antenna [22]. The scanning is 1-D, in
either the x-z or y-z plane depending on the location of the feed. A successful
graduation to 2-D scanning had not been reported by the end of 2017.
Arguably, without being too disingenuous, there was a period over the past ten to
twenty years when the route to virtually guaranteed paper publication on any
antenna topic was to include the word ‘metamaterial’ in the title. Actually many
such papers [23–27] did not include metamaterials as strictly defined—namely a
material, or medium, displaying both negative permittivity and negative perme-
ability, sometimes referred to as a negative refractive index material (see below).
Most of these contributions describe the employment of periodic surfaces in novel
antenna forms, using the unique features of such surfaces, in particular frequency
selectivity and reconfigurability, as we shall see in the following sections.
Before we consider the metamaterial role it is important that we fully appreciate the
EM characteristics of a regular passive periodic surface from an antenna perspec-
tive. As has already been observed, the primary difference, electromagnetically,
between an array antenna and passive periodic surface used as a scatterer, such as a
frequency selective surface (FSS/metamaterial), is in the nature of the energy
source. For the array antenna, as we have seen, each element of the array is an
individual source of EM radiation (Chap. 6, Fig. 6.1), whereas the passive FSS is
10.9 The Role of Metamaterials in Compact Slot Arrays 273
swept by an incoming plane wave (Fig. 10.16) which excites simultaneously the
periodically spaced elements. This is, of course, closely analogous to the leaky
wave array discussed in the preceding section of this chapter. Hence, as elements of
an array, the FSS apertures combine following antenna array rules, thus generating
scattered waves in the forward (transmitted) and backward (reflected) directions. In
practice the periodic array would normally be planar. However, since the planar
array is merely a combination of parallel linear arrays, we can deduce the essence of
the behaviour of a passive periodic surface by limiting our attention to a linear
geometry, as we have done for the array antenna analysis in Chap. 6.
With reference to Fig. 10.16, it is useful initially to consider the limiting cases of
EM wave incidence on the conducting screen containing a periodic array of
apertures of arbitrary size, usually ranging from zero to half-wavelength long in
their major dimension. For infinitesimal slots the screen is perfectly conducting
everywhere, and by Snell’s laws leads to a single reflected wave equal in magnitude
to the incident wave and reflected in the direction −/ when the incident angle, to
the surface normal, is +/. The opposite limit, which is rather meaningless practi-
cally, involves apertures which are large enough to totally replace the screen. In this
case, there is a single transmitted wave at +/ and no reflected wave.
The more practical large iris limit is the half-wavelength option, in which case a
transmitted wave and a reflected wave is formed with essentially equal magnitude.
At resonance each iris radiates equally into the forward and backward half-spaces.
If we limit our attention to a two dimensional scenario, these re-radiated directions
can be deduced from the linear array theory presented in Chap. 6. For the incident
Fig. 10.16 Plane wave scattering at a passive metallic screen supporting a periodic array of
apertures separated by distance d
274 10 Compact Planar Resonator Arrays
wavefront direction depicted in Fig. 10.16 (dark blue solid trace below the array),
the element to element phase shift w (Eq. 6.2) is now dictated by the direction of
the incoming phase front. For a passive surface there is no independent phase delay
or advance and so a ¼ 0 . Equation (6.2), therefore, takes the following form for a
uniform periodic array excited by an incident plane wave at incident angle /:
w ¼ k0 d sin / ð10:42Þ
In preceding sections of this chapter it has generally been assumed that the cavity
resonator antennas, which have been chosen for study, have incorporated periodic
surfaces formed from scattering elements of relatively regular shape (Fig. 10.17).
For passive arrays of such patches or apertures at microwave and millimetre-wave
frequencies it is evident, as demonstrated above, that conventional antenna array
theory, including Snell’s laws comprehensively define the surface characteristics.
However the passive surface need not necessarily be composed of regularly shaped
scatterers and an indication of the range of possible shapes which could be used in
appropriate circumstances are illustrated in Fig. 10.18.
10.9 The Role of Metamaterials in Compact Slot Arrays 275
Fig. 10.17 Typical FSS formats for CRA’s with (a) and (b) forming reflecting surfaces while
(c) and (d) present transparent surfaces
Fig. 10.18 Elemental scatterers for periodic surfaces described as metamaterials fabricated on
copper coated planar dielectric substrates. a Centre connected, b loop types, c patch types,
d combinations
For case (b), which replaces the regular solid with a metamaterial we obtain:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
vair c 1 ðlrm l0 Þðerm e0 Þ pffiffiffiffiffiffiffiffiffiffiffiffiffi
nmeta ¼ ¼ ¼ pffiffiffiffiffiffiffiffiffi
¼ lrm erm ð10:46Þ
vmeta vmeta l0 e 0 1
if the negative root is chosen as in reference [33]. For the metamaterial the relative
permittivity and permeability are respectively −erm and −lrm. The positive root
would imply that Eqs. (10.45) and (10.46) are identical and that EM refraction at
the metamaterial surface is no different to the regular material, which is hardly an
advance. If substantiated, the electrical ramifications of the negative choice in
Eq. (10.46) are not insignificant, and hence the Veselago paper caused quite a ‘stir’.
For plane wave refraction at a plane interface between two regular materials, the
angle of refraction (ht), which is intrinsic to the design of optical devices, is given
by Snell’s law, namely:
sin hi v1 c pffiffiffiffiffiffiffiffi
¼ ¼ ¼ lr er ¼ nsolid ð10:47Þ
sin ht v2 vsolid
sin hi v1 c pffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ ¼ ¼ lrm erm ¼ nmeta ð10:48Þ
sin ht v2 vmeta
In other words, for a metamaterial interface, the refraction angle (ht) is the
opposite sign to the corresponding angle in a regular set up (see Fig. 10.19).
Since the Snell’s laws arise by enforcing the electric and magnetic field
boundary conditions for the plane waves at the regular interface (Fig. 10.20a) then
Fig. 10.20 Surface reflection and refraction at (a) a conventional interface and (b) an interface
qffiffiffiffiffiffiffi
with a metamaterial. In both cases n1=2 ¼ ll2 ee21
1
10.9 The Role of Metamaterials in Compact Slot Arrays 279
the same must be true for the metamaterial interface. The situation is illustrated in
Fig. 10.20.
Figure 10.20a depicts a mirror flat interface between two conventional materials
(1) and (2) for which the permittivities are respectively e1 and e2, and the perme-
abilities are respectively l1 and l2. Satisfying the electromagnetic field boundary
conditions at the interface at z = 0 leads to Snell’s laws (Appendix A), which are
represented in vector form in the figure. The directional vectors ki1, kr1 and kt2 must
obey the specified angular relationships while the field vectors Ei, Hi, Er, Hr, Et and
Ht automatically satisfy the boundary conditions. Power flow requirements dictate
that for the incident, reflected and transmitted plane waves E, H and k obey the
clockwise rule (Appendix A) with the Poynting vector, and hence power flow,
directed along the relevant k vector. Note that the boundary conditions must be met
everywhere over the z = 0 interface at any given instant in time. This is only
possible if
It is not difficult to show that adherence to the Snell’s laws ensures that this is the
case.
So, does case (b) in Fig. 10.20 also meet all of these fundamental physical
constraints? The negative refractive index for the metamaterial requires that the
refractive wave is directed along kt2 at an angle –ht below the z-axis. However, in
addition a negative refraction index implies that, assuming for simplicity that region
(1) is air or vacuum:
x pffiffiffiffiffiffiffiffiffiffiffiffiffi xn1=2
kt2 ¼ lmr emr ¼ ð10:50Þ
c c
The significance of Eq. (10.50) is that kt2 is directed towards the interface. So
the field boundary conditions are in accord with case (a) with the tangentially
directed wave components on each side of the interface meeting the dictates of
Eq. (10.50). However, while the boundary conditions are seemingly satisfied for the
negative index metamaterial, power flow directions are puzzling. The directions of
Et and Ht in case (b) dictate that the Poynting vector S ¼ E H and hence power
flow is in the direction opposite to kt2. While for power flow normal to the interface
energy conservation is satisfied, for tangential power flow in the z = 0 plane the
power flow in the regular medium (1) opposes the flow in the metamaterial (2),
which suggests that the negative index material is not physically possible. This
embodiment of the refracted wave has been compared to a backward wave which is
a possible solution to the Maxwell equations in periodic structures. But in backward
wave modes, such as those that occur in backward wave microwave oscillators, for
example, power flow everywhere within the modal field structure propagates with
the same group velocity, which is in opposition to the phase velocity direction –
hence the description ‘backward wave’. The electromagnetic field solution in case
(b) of Fig. 10.20 contravenes this behaviour.
280 10 Compact Planar Resonator Arrays
The ‘coup de gras’ for artificial metamaterials displaying negative refractive index
is arguably the ‘flat lens’ proposed by Veselago [33], although many other practical
anomalies have been explored and discarded [32]. These intensive studies suggest
that most metamaterial claims are identifiable with established properties of peri-
odic structures.
The Veselago flat lens is viewed in conjunction with a conventional lens per-
forming the same focusing role in Fig. 10.21. For the conventional lens the
focusing mechanism is easily explained. A source at point (A) in Fig. 10.21a is
imaged at point (B) by ensuring the electrical path lengths between (A) and (B) are
the same, balancing electrical path lengths in the air with path lengths in the lens of
lower phase velocity. In addition the lens surface is parabolically shaped to ensure
that the glass/air refractive exit angles match the entry angles. The conventional
lens is predominantly used at optical frequencies and consequently it is usually
made of hard transparent material such as glass with lr = 1. For the flat lens this
restriction does not necessarily apply. It is generally imagined to operate at below
light frequencies as a 3-D periodic structure for which lr and er are negative and
greater than unity. The operational principles of the flat lens are not too different to
those of the conventional lens. The electrical path lengths from the source at (A) to
the image at (B) must be the same for all rays, while the negative refraction angles
must be arranged to secure a focused image on the opposite side of the lens as
suggested in Fig. 10.21b. If the path between (A) and the secondary focus at (C) is
examined for both an edge ray, and a straight through ray, then to arrive at point
(C) in phase requires that the additional phase delay in the air for the edge ray must
be negated by an extra phase advance in the metamaterial. A similar requirement
applies to all other rays. Equations (10.48) and (10.50) suggest that this is possible
for phase. The difficulty with this argument is that it omits the time element. Phase
coherence at (C), and hence at (B), must occur at the same instant of time, at all
times. This implies that the additional time delays in the air for the longer paths
must be balanced by time advances in the metamaterial, and this is synonymous
with negative time or time reversal! In real physical structures and systems, at least
in this corner of the universe, cause and effect is sacrosanct, and hence the flat
metamaterial lens is not possible.
The Fabry-Perot resonator antenna or cavity resonator antenna in its most basic
form, is shown to be a leaky parallel plate waveguide operated close to the cut-off
of the fundamental TE mode. It is examined initially from a plane wave ray per-
spective and to advance the analysis the periodic waveguide wall is modelled as a
uniform and partially reflective surface (PRS). This leads to simple but actually
rather useful estimates of directivity and in particular how it relates to the reflec-
tivity of the periodic surface. This type of antenna is usually realised as a copper
clad microwave substrate with one surface made periodic by the presence of a
suitably designed array of radiating patches or radiating slots.
More generally, the CRA is shown to represent a class of compact planar
antenna which includes the partially reflective surface (PRS) antenna, the
Fabry-Perot cavity antenna, and the leaky wave CRA. All of these forms can be
modelled by the reciprocal periodic moment method which is fully developed in
Sect. 10.4. Supported by evidence from the literature, modelling demonstrates that
high gain and/or directivity is available from a PRS antenna, but with very restricted
bandwidth. Inconvenient beam splitting and scanning occur as the operating fre-
quency is raised above cut-off. Nevertheless bandwidth enhancement is an active
pursuit.
Splitting of the primary radiated beam also occurs with centre fed leaky-wave
antennas if the applied frequency is above the fundamental mode cut-off value.
Consequently, as with the PRS antenna, beam scanning using frequency is prob-
lematic. However, beam scanning remains a goal of CRA development and it is
shown how this can be achieved, in 1-D embodiments, by using end-fed arrays
incorporating one or more additional reconfigurable periodic layers in the form of
high impedance surfaces (HIS). The planar leaky-wave antenna incorporating a
textured surface has not yet demonstrated 2-D scanning although it possesses the
potential to do so. The conundrum then remains of applying leaky-wave methods to
secure 2-D beam scanning from a compact planar CRA. To do so will require novel
feed arrangements which probably means abandonment of conventional centre fed
geometries associated with the CRA. With 1-D scanning the complex diode
282 10 Compact Planar Resonator Arrays
References
1. G. von Trentini, Partially reflective sheet arrays. IRE Trans. Antennas Propag. 4(4), 666–671
(1956)
2. A.P. Feresidis, G. Goussetis, S. Wang, J.C. Vardaxoglou, Artificial magnetic conductor
surfaces and their application to low-profile high-gain planar antennas. IEEE Trans. Antennas
Propag. 53(1), 209–215 (2005)
3. N. Guerin, S. Enoch, G. Tayeb, P. Sabouroux, P. Vincent, H. Legay, A metallic Fabry-Perot
directive antenna. IEEE Trans. Antennas Propag. 54(1), 220–224 (2006)
4. A. Foroozesh, L. Shafai, Investigation into the effects of the patch-type FSS superstrate on the
high-gain cavity resonance antenna design. IEEE Trans. Antennas Propag. 58(2), 258–270
(2010)
5. T. Zhao, D.R. Jackson, J.T. Williams, H.-Y.D. Yang, A.A. Oliner, 2-D periodic leaky-wave
antennas-part I: metal patch design. IEEE Trans. Antennas Propag. 53(11), 3505–3514 (2005)
6. T. Zhao, D.R. Jackson, J.T. Williams, 2-D periodic leaky-wave antennas-part II: slot design.
IEEE Trans. Antennas Propag. 53(11), 3515–3524 (2005)
7. A.P. Feresidis, J.C. Vardaxoglou, High gain planar antenna using optimised partially
reflective surfaces. IEEE Proc. Microwaves Antennas Propag. 148(6), 345–350 (2001)
8. C. Mateo-Segura, G. Goussetis, A.P. Feresidis, Sub-wavelength profile 2-D leaky-wave
antennas with two periodic layers. IEEE Trans. Antennas Propag. 59(2), 416–424 (2011)
9. C. Mateo-Segura, A.P. Feresidis, G. Goussetis, Bandwidth enhancement of 2-D leaky-wave
antennas using double-layer periodic surfaces. IEEE Trans. Antennas Propag. 62(2), 586–593
(2014)
10. T. Itoh, W. Menzel, A full-wave analysis method for open microstrip structures. IEEE Trans.
Antennas Propag. AP-29(1), 63–68 (1981)
11. L. Zhou, H. Li, Y. Qin, Z. Wei, C.T. Chan, Directive emissions from subwavelength
metamaterial-based cavities. Appl. Phys. Lett. 86, 101101-1–101101-3 (2005)
12. A.R. Weily, T.S. Bird, Y.J. Guo, A reconfigurable high–gain partially reflecting surface
antenna. IEEE Trans. Antennas Propag. 56(11), 3382–3390 (2008)
13. T. Debogovic, J. Perruisseau-Carrier, J. Bartolic, Partially reflective surface antenna with
dynamic beamwidth control. IEEE Antennas Wirel. Propag. Lett. 9, 1157–1160 (2010)
14. T. Debogovic, J. Bartolic, J. Perruisseau-Carrier, Array-fed partially reflective surface antenna
with dynamic beamwidth control and beam-steering. Paper presented at the 6th European
Conference on Antennas and Propagation (EuCAP 2012, Prague, 2012)
References 283
There are essentially three routes to tracking a moving signal source, such as a
satellite, with array technology. Firstly, in a full blown phased array antenna with a
transmit/receive unit behind each element of the array, feeding into a high speed
digital signal processor. In this case the antenna radiation pattern can be configured
almost instantaneously into any form required. This is the expensive option. An
undoubtedly cheaper alternative is to use a form of monopulse tracking commonly
seen in radar and in parabolic reflector tracking systems [1, 2]. But this is also
technically challenging for commercial applications aimed at satellite/mobile/
wireless communications.
A third form of tracking, which has been given impetus by recent reconfigurable
developments in array antennas, is provided by the ‘retro-directive’ array [3, 4]. The
principle is not new. Retro-direction, whereby an incoming radar signal, on striking
a target is returned back in the direction from which it came, is achieved (see
Chap. 1) by ‘corner reflectors’, often seen on the masts of yachts [5]. The device
increases the radar cross-section of the target thereby enhancing its detectability
against the background of the sea. Arguably the retro-direction in an array offers the
most cost effective option for emerging wireless systems.
The retro-directive array topology can be further sub-divided into two basic
classifications, depending on whether or not the front-end transmit/receive devices
are ‘mixer’ based, or ‘phase-lock-loop (PPL)’ based. The essence of the two
alternatives is outlined below.
The existence of retro-directive antenna action lies in the nature of travelling waves
in free space and on how this behaviour is expressed mathematically. An electro-
magnetic wave travelling away from a source in the direction r varies with time and
with distance from the source. At a single frequency x the variation will be
sinusoidal and this can be expressed as:
which has the functional form of a travelling wave moving in the direction r at
velocity c. Now for a wave of the same frequency x, travelling in the opposite
direction (−r) with amplitude A0 the electric field can be expressed as:
r
E 0 ¼ A0 expðjxtÞ expðjk0 rÞ ¼ A0 exp jxðt þ Þ ð3Þ
c
/ ¼ k0 d sin h ð4Þ
arrangement, apart from the requirement for a local oscillator at double the oper-
ating frequency, is the presence of the circulator. This has to be a high quality
device, to minimise tracking errors, and at microwave frequencies such a device is
very expensive. Furthermore, this cost is multiplied by the number of elements in
the array.
One way around the circulator difficulty is to provide separate transmit and
receive arrays, as advocated by Myamoto [9, 10] which allows the circulator to be
replaced by a ring hybrid coupler, or rat-race. However, the requirement to
duplicate the transmitter and receiver array hardware is clearly a disadvantage. The
Myamoto solution requires up and down links at different frequencies which, in
turn, demands frequency doubling and frequency shifting circuitry in the device
being tracked, such as a satellite. This could present unwelcome hurdles in certain
applications. Such difficulties have led to the development of phase locked loop
(PLL) based conjugating circuits originally proposed by Ghose [4], and later pur-
sued in Refs. [11–16]. These are discussed in the next section.
288 11 Retro-directive Compact Array Antennas
V ¼ VA expðx1 t þ / þ x2 t þ UÞ ð5Þ
where VA is an arbitrary amplitude factor, and / and U are transmit and receive
phases associated with a main beam at an angle relative to the array normal. If we
assume (Fig. 11.2) that for transmit and receive magnitudes F1 F2 then when the
loop is phase locked we must have U = −/, i.e. phase conjugation. Note that this
phase condition for phase conjugation requires that phase locking is performed on
the sum signal from the mixer at A. Physically, this means that at phase lock, the
signal transmitted by the array is forced to return in the direction of the incoming
received signal—or expressed another way the primary beam of the array is forced
to be pointing at the source. The second mixer at the lower left of the diagram
The phase conjugation scheme represented by Fig. 11.3 assumes that a double array
exists with the receive array separate from the transmit array. This configuration, as
already observed, could obviously be a disadvantage in many situations, as could
the employment of differing transmit and receive frequencies. Furthermore at
microwave frequencies the very high sum frequency fed back to the PLL requires a
costly BPF and a high degree of division. An ‘in-phase/quadrature phase’
(IQ) modification has been proposed [15, 16] to circumvent these difficulties and is
shown diagrammatically in Fig. 11.4. The IQ modulator on the right of the circuit is
pivotal to the approach. The circuit as a whole (one to each array element) con-
jugates the signal (xRF) received at the left port, from the appropriate array element,
before dispatching it at the right (xout) to the transmit radiating element. As will be
demonstrated below, the transmit and receive elements need not be separate radi-
ators, as Fig. 11.4 seems to suggest. The IQ modulator, in analogue form, primarily
comprises a pair of up-converting mixers, a 90° phase shift and a summing
junction.
Phase conjugation is procured as follows. A received signal (xRF) at frequency
2400.5 MHz (say), and phase delay /, is mixed with a 2400 MHz signal from the
frequency controlled local oscillator (VCO1) to form a 0.5 MHz IF signal at A, i.e.
and at C:
which is phase conjugated relative to the input as required. The output frequency is
equal to the input frequency if xLO1 þ xLO2 ¼ 2xRF .
It is clear from the literature that the performance of retro-directive array antennas is
measurably susceptible to direct leakage between the transmission and receiver
sections of the system, particularly if transmit and receive operations are performed
at the same frequency. The main source of this leakage is mutual coupling between
closely spaced array elements, and to avoid this, the transmit and receive functions
are often totally separated, by adopting two distinct arrays [6, 12, 16]. Leakage
between the transmission and reception circuits both reduces the sensitivity of the
system and diminishes tracking accuracy.
It is not difficult to appreciate that doubling up on array hardware to counteract
leakage seems like technological ‘over kill’. It certainly adds significantly to build
costs. The usual way around this is to employ a duplexer behind each array element
which allows the transmit and receive channels to use the same antenna element.
This is suggested in Fig. 11.1 where the duplexer takes the form of a circulator.
Unfortunately, as indicated previously, duplexers, which display high enough
isolation between transmit and receive ports to permit reliable and accurate
retro-directive behaviour in an array, tend to be very expensive.
Given advances in planar microwave circuit fabrication techniques, a potentially
much more cost effective solution is to employ array radiator elements which have
dual-mode capability [17]. An example is depicted in Fig. 11.5, where the
retrodirective array is formed from square patches capable of receiving/transmitting
292 11 Retro-directive Compact Array Antennas
While retro-directive array antennas for mobile applications are, on the basis of
current technology, likely to be optimally realised using patch radiating elements,
needless to say, non-patch alternatives, providing circularly polarised and/or
dual-mode capability, are also scattered through the literature [29]. Generally, these
competing possibilities comprise either microstrip fed cross-slot radiators, or
microstrip fed circular apertures. An example which gives a flavour of non-patch
possibilities is depicted in Fig. 11.8. The four holes etched into the metallic ground
plane of the middle substrate (b) are sandwiched between substrates (a) and
294 11 Retro-directive Compact Array Antennas
Fig. 11.6 Schematic of dual-mode dual-polarised patch antenna parasitically fed from a
non-resonant X-slot by means of microstrip lines above and below the slot
11.7 Dual-Polarised Radiation from Apertures 295
Fig. 11.7 Schematic of a dual-mode dual-polarised patch antenna parasitically fed from a
non-resonant X-slot by means of balanced microstrip lines above and below the slot to improve
bandwidth and axial ratio
296 11 Retro-directive Compact Array Antennas
Fig. 11.8 Schematic of dual-mode dual-polarised circular aperture array antenna parasitically fed
from balanced microstrip lines above and below the four elements
(c) which carry branching microstrip lines. The elliptic shapes, incorporated into the
branch-lines, should register with the circular apertures, and are designed to
improve the matching process. The upper substrate er1 excites vertically polarised
radiations from the four apertures while the lower substrate er3 generates
11.7 Dual-Polarised Radiation from Apertures 297
Fig. 11.9 Schematic of dual-mode dual-polarised patch antenna. a Substrate 1—copper patch
L1 L1, thickness t1, relative permittivity er1, b substrate 2—copper patch L2 L2, thickness t2,
relative permittivity er2, c two dumbbell coupling slots in top face, 50X feed lines in lower face
(see Gao [26] for detail), substrate thickness t3, relative permittivity er3, d edge view showing air/
polystyrene layer between substrate 1 and substrate 2, thickness t4, relative permittivity er4
Fig. 11.10 Circularly polarised dual-mode square patch antenna with slot coupling and microstrip
hybrid feed arrangement
[31]. A typical manifestation of this approach is depicted in Fig. 11.11. The circular
symmetry of the round patch confers sufficient marginal improvement in polari-
sation circularity, at wide angles in the radiation pattern, to acquire enhanced
ellipticity bandwidth for the circular radiator, when compared with square patch
alternatives.
oscillator signal at twice the RF frequency is introduced at the left of the circuit. It is
also split into two equal components with a differential 180° phase shift added by
line length. The advantage of the RF frequency signals on the microstrip lines, in
the upper and lower portions of the circuit, being in antiphase is that leakage or
cross-coupling across the channels, which can degrade the retro-directive operation,
is minimised. The FET gate mixers provide both phase conjugation and gain, with
11.9 Phase Conjugated Active Antenna Element 301
Fig. 11.12 Phase conjugating active antenna element incorporating a balanced quasi-optical
mixer
the difference frequency IF signals adding at the T-junction before passing to the
antenna for transmission. Note that at the patch the transmit and receive signals are
isolated from each other by employing the orthogonal linear polarisations of the
square patch. For retro-directive communication between a mobile platform and a
satellite, which depends on circularly polarised radiations, the Fig. 11.12 circuit is
inappropriate. While the concept is promising, clearly further research is needed to
achieve both a dual-polarised and a circularly polarised active antenna element for a
mobile communications system with fully operational retro-directive capability.
It is tentatively suggested that the future may reside with active antenna ele-
ments, but this remains some distance away, until dual-polarised, circularly
polarised capability can be confidently demonstrated in commercial products.
References
1. R.C. Johnson, Antenna Engineering Handbook, 3rd edn. (McGraw-Hill Inc., New York,
1993)
2. A.W. Rudge, K. Milne, A.D. Olver, P. Knight (eds.), The Handbook of Antenna Design:
volumes I and II (Peter Peregrinus Ltd., London, 1986)
3. C.Y. Pon, Retrodirective array using heterodyne technique. IEEE Trans. Antennas Propag.
AP-12, 176–180 (1966)
4. R.N. Ghose, Electronically adaptive antenna systems. IEEE Trans. Antennas Propag. AP-12
(2), 161–169 (1964)
5. L.G. Van Atta, Electromagnetic Reflector. US Patent 2,908,002, 6 Oct 1959
6. R.C. Chernoff, Large active retrodirective arrays for space applications. IEEE Trans.
Antennas Propag. AP-27(4), 489–496 (1979)
7. K. Leong, Y. Wang, T. Itoh, A full duplex capable retrodirective array system for high speed
beam tracking and pointing applications. IEEE Trans. Microwave Theory Tech. MTT-52(5),
1479–1489 (2004)
8. G.S. Shiroma, R.Y. Miyamoto, W.A. Shiroma, A full-duplex dual-frequency self-steering
array using phase detection and phase shifting. IEEE Trans. Microwave Theory Tech.
MTT-54(1), 128–134 (2006)
9. R.Y. Myamoto, Y. Qian, T. Itoh, An active integrated retrodirective transponder for remote
information retrieval. IEEE MTT-S, Int. Microwave Symp. 3, 1431–1434 (2000, June)
10. R.Y. Myamoto et al., Digital wireless sensor server using an adaptive smart-antenna/
retrodirective array. IEEE Trans. Vehicular Technol. 52(5), 1181–1189 (2003)
11. S.L. Karode, V.F. Fusco, Frequency offset retrodirective antenna array. Electron. Lett. 33(16),
1350–1351 (1997)
12. S.L. Karode, V.F. Fusco, Multiple target tracking using retrodirective antenna arrays. Nat.
Conf. Antennas Propag. Publication No. 461, 178–181 (1999)
13. V.F. Fusco, C.B. Soo, N. Buchanan, Constant phase IF self-tracking receiver analysis and
characterisation. IEEE Proc. (MAP) 151(6), 507–513 (2004)
14. N.B. Buchanan, V.F. Fusco, Triple mode PLL antenna array. IEEE MTT-S Int. Microwave
Symp. Digest, 1691–1694 (2004)
15. V.F. Fusco, C.B. Soo, N. Buchanan, Analysis and characterisation of PLL-based retrodirec-
tive array. IEEE Trans. Microwave Theory Tech. MTT-53(2), 730–738 (2005)
16. V.F. Fusco, N. Buchanan, High performance IQ modulator based phase conjugator for
modular retrodirective antenna array implementation. IEEE Trans. Microwave Theory Tech.
MTT-57(10), 2301–2306 (2009)
17. V.F. Fusco, N. Buchanan, Dual mode retrodirective phased array. Electron. Lett. 45(3), 139–
141 (2009)
18. T.D. Huynh, Dual polarised aperture coupled microstrip patch antenna system. US Patent
No. 5,896,107, 1999
19. L. Desclos et al., Single and dual mode patch antennas. US Patent No. 6181281, 30 Jan 2001
20. N. McDonald, Dual polarised patch antenna. European Patent No. EP 0901185, 14 Nov 2001
References 303
21. S.M. Sanzgiri et al., Dual polarised microstrip patch antenna array for PCS base stations. US
Patent No. 5,923,296, 13 July 1999
22. A. Adrian, D.H. Schaubert, Dual aperture-coupled microstrip antenna for dual or circular
polarisation. Electron. Lett. 23(23), 1226–1228 (1987)
23. K. Gosalia, G. Lazz, Reduced size dual-polarised microstrip patch antenna for wireless
communications. IEEE Trans. AP-51(9), 2182–2186 (2003)
24. S.L. Karode, V.F. Fusco, Dual-polarisation microstrip patch antenna using feedforward
isolation enhancement for simultaneous transmit/receive applications. Nat. Conf. Antennas
Propag. Conference Publication No. 461, 49–50 (1999)
25. M. Edimo, A. Sharaiha, C. Terret, Optimised feeding of a dual-polarised broadband
aperture-coupled printed antenna. Electron. Lett. 28(19), 1785–1787 (1992)
26. S.C. Gao, L.W. Li, P. Gardner, P.S. Hall, Wideband dual-polarised microstrip patch antenna.
Electron. Lett. 37(20), 1213–1214 (2001)
27. A. Uz Zaman, L. Manholm, D. Derneryd, Dual-polarised microstrip patch antenna with high
port isolation. Electron. Lett. 43(10), 551–552 (2007)
28. J.R. Sandford, A. Tengs, A two substrate dual-polarised aperture-coupled patch. IEEE
Antennas Propag. Soc. Int. Symp. AP-S Digest 3, 1544–1547 (1996)
29. E. Rammos, A new wideband high-gain stripline planar array antenna for 12 GHz TV.
Electron. Lett. 18(6), 252–253 (1982)
30. A.J. Sangster et al., A dual-polarised stripline fed slot antenna incorporating signal
cancellation. IEE Proc. (MAP) 148(6), 357–362 (2001)
31. N.C. Karmakar, M.E. Bialkowski, Circularly polarised aperture coupled circular microstrip
patch antennas for L-band applications. IEEE Trans. AP-47(5), 933–940 (1999)
32. N.C. Karmakar, M.E. Bialkowski, A beam forming network for a circular switched-beam
phased array antenna. IEEE Microwave Wireless Lett. 11(1), 1–9 (2001)
33. M.E. Bialkowski, N.C. Karmakar, A two ring circular phased array antenna for satellite
communications. IEEE Antennas Propag. Mag. 41(3), 14–23 (1999)
34. K. Chang, R.A. York, P.S. Hall, T. Itoh, Active integrated antennas. IEEE Trans. MTT-50(3),
937–944 (2002)
35. R.Y. Miyamoto, Y. Qian, T. Itoh, Active retrodirective array for remote tagging and wireless
sensor applications. IEEE MTT-S Int. Microwave Symp. Dig., Boston, MA, 1431–1434
(2000, June)
Appendix A
Free-Space Electromagnetic Waves
All materials contain electric charges bound loosely or otherwise within atoms and
molecules. If these materials exist in an environment which naturally or artificially
causes agitation of the charge, and hence changes in the associated electric and
magnetic fields, then electromagnetic waves are unavoidable. These waves can
appear in quite complex forms such as trapped wave, fast waves, slow waves,
surface waves, leaky waves, evanescent fields and radiating waves. To describe
these various and quite diverse formats a set of equations is required which rep-
resent the field components of the wave at a point in space. The point forms
(differential forms) of Maxwell’s equations are developed in a host of text books on
the topic (see Bibliography) and essentially entails the recruitment of well known
vector-differential theorems such as the divergence theorem and Stokes’ law to
accomplish the transitions.
Many electromagnetic wave guiding and manipulating problems are of the
source-free variety, which implies that the source is so far distant that the waves of
interest are freely propagating and in an unbounded region (free-space) are plane
waves. These waves, also termed transverse electromagnetic (TEM) waves, are
described as ‘plane’ because the radius of curvature of the wave-front is very large,
and thus the natural rate of curvature of the front, can be deemed mathematically
insignificant allowing it to be fully described by means of Cartesian coordinates. In
this scenario, the EM problem reduces to a boundary value problem, for which
Maxwell’s equations, in differential form, become:
rD¼0 ðA:1Þ
rB¼0 ðA:2Þ
@B
rE¼ ðA:3Þ
@t
@D
rH¼ ðA:4Þ
@t
where E and H represent the electric and magnetic field intensities in the region
of interest. As always, D = eE is the electric flux density while B = lH is the
magnetic flux density. The ‘del’ operator ðrÞ expresses directional derivatives in
the three space directions. It is a vector, which in the Cartesian system (for example)
has the form:
@ @ @
r ¼ ^ax þ ^ay þ ^az ðA:5Þ
@x @y @z
where ^ax ; ^ay and ^az are unit vectors directed along x, y, and z respectively.
When the del operator is multiplied to a scalar (/(x, y, z) say) the result is a vector
which expresses the gradient or slope of the function / in all three space directions.
i.e.:
@/ @/ @/
r/ ¼ ^ax þ ^ay þ^
az ðA:6Þ
@x @y @z
Cross multiplication of del with a vector produces the operation of ‘curl’, while
dot multiplication produces the operation of ‘divergence’ (‘div’). Crudely, curl is
circulation at a point, while divergence is flux at a point.
@
rrE¼ lr H
@t
@ @
¼ ½ le E ðA:7Þ
@t @t
@2E
¼ le 2
@t
Appendix A: Free-Space Electromagnetic Waves 307
Hence, on using a convenient vector identity, which states that for any vector F
r r F ¼ rr F r2 F ðA:8Þ
@2E
rr E r2 E ¼ le ðA:9Þ
@t2
But, from Eq. (A.1), r E ¼ 0, for a linear, homogeneous medium for which
permeability l and permittivity are constants. Therefore
@2E
r2 E ¼ le ðA:10Þ
@t2
and by analogy:
@2H
r2 H ¼ le ðA:11Þ
@t2
Equations (A.10) and (A.11) are wave equations. Equations of this nature, with
appropriate variables, appear in most branches of science and engineering and their
solutions have been studied widely. Solutions depend very much on the boundary
conditions—namely the conditions imposed on the variables at the periphery or
containing surface of the solution region. They can fix the magnitude of the variable
(Dirichlet condition) or the rate of change of the variable (Newman condition) or a
mixture of both. A unique solution depends on the conditions being neither under
specified or over specified.
For example, let us consider formulating a solution to Eq. (A.10), and inevitably
Eq. (A.11) because of the Maxwell linkages, for a region of free space (l = l0:
e = e0) which is large enough to presume that all boundaries are effectively at
infinity. In this case we can choose to represent the region mathematically using
Cartesian coordinates, and furthermore since we anticipate that the solution is a
waveform we can arbitrarily determine that the waves travel in the z-direction. This
implies that the rates of change of the E-field in x and y are zero, and using (A.1) it
follows that Ez = 0. The equation to be solved, therefore, is:
@2E @2E
¼ l0 e0 2 ðA:12Þ
@z 2 @t
@ 2 Ex 1 @ 2 Ex
¼ ðA:13Þ
@z2 c2 @t2
This is easily demonstrated by substitution back into the equation. The first term
represents a wave travelling in the +z direction while the second allows for a
reflected wave, if such exists. Given that velocity is the rate of change of z with
respect to time, it is evident that c represents velocity (actually phase velocity) of
the electromagnetic wave in ‘free space’. For vacuum it is equal to 3 108 m/sec.
The application of Maxwell’s equations also gives Hz = 0 and:
A B
Hy ¼ f ðz ctÞ þ f ðz þ ctÞ ðA:15Þ
g g
Also
rffiffiffiffiffi
Ex l0
¼ ¼ g ðA:16Þ
Hy e0
η is termed the free space wave impedance which for air or vacuum has the value
120p X. The resultant solution is a plane electromagnetic wave, also termed a
transverse electromagnetic (TEM) wave, for which E and H are transverse to the
direction of propagation and orthogonal to each other. E and H are also in time
phase, as Eq. (A.16) attests (see Fig. A.1).
Electrical engineers are generally very familiar with the relationship between
power (P), voltage (V) and current (I) in the form:
1
P ¼ VI Watts ðA:17Þ
2
where V and I are defined in peak, rather than in the more common r.m.s.,
format. But, voltage is simply integrated electric field E (V/m), and from Ampere,
current is integrated magnetic field intensity H (A/m), so by analogy we can suggest
that for the plane wave:
Appendix A: Free-Space Electromagnetic Waves 309
1 1
p ¼ EH ¼ ce0 E2 Watts=m2 ðA:18Þ
2 2
This means that p is the real power flow density in the TEM wave. In general,
complex power flow density in an electromagnetic wave is given by the Poynting
vector S, where:
1
S ¼ E H Watts=m2 ðA:19Þ
2
When a plane electromagnetic wave at the frequency of light, or in fact any radio
frequency, is incident upon a smooth interface (by ‘smooth’ it is meant that any
surface undulations or protuberances are in size very much less than the wavelength
of the impinging waves) between two extended propagating media, part of the wave
is reflected back into the incident medium while part is transmitted or refracted into
the second medium, usually with a change of direction.
Analytically, the relationships between the incident and reflected waves can be
developed by considering a plane electromagnetic wave, incident at a physically
real angle hi to the normal, at the interface between two semi-infinite regions of
space, as suggested in Fig. A.2. Each region is presumed to comprise linear
homogeneous medium with a different index of refraction (n). The index of
refraction is defined as:
c
n¼ ðA:22Þ
v
where c is the speed of light in vacuum, or free space, while v is its speed within
the specified medium. Also, with reference to Fig. A.2, the following definitions
apply:
1
c ¼ pffiffiffiffiffiffiffiffiffi ðA:23Þ
l0 e 0
and
1
v1 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðA:24Þ
l0 e0 er1
1
v2 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðA:25Þ
l0 e0 er2
Here, e0 and l0 are the free space permittivity and permeability respectively,
while er1 is the relative permittivity of medium 1 and er2 is the relative permittivity
of medium 2. Both media are assumed to be loss-less and non-magnetic in which
case l1 = l2 = l0. The indices of refraction for the two media then become:
pffiffiffiffiffiffi pffiffiffiffiffiffi
n1 ¼ er1 ; n2 ¼ er2 ðA:26Þ
Maxwell’s equations in the semi-infinite regions remote from the interface are,
as we have seen above, fully satisfied by TEM plane waves. It remains then to
satisfy the Maxwell boundary conditions at the interface. If this can be done the
resultant solutions represent complete EM solutions for the specified boundary
value problem. For an incident TEM electromagnetic wave, as depicted in Fig. A.2,
the E-field vector and the H-field vector must be mutually orthogonal to each other
^ directed in
and to the direction of propagation, usually defined by a unit vector k,
the direction of the relevant ray. In this case we can write:
1 ^
H ¼ ðk EÞ ðA:27Þ
g
where η is the wave impedance for the medium containing the wave. Hence for
regions 1 and 2 respectively:
rffiffiffiffiffiffiffiffiffi
l0
g1 ¼ ðA:28Þ
er1 e0
rffiffiffiffiffiffiffiffiffi
l0
g2 ¼ ðA:29Þ
er2 e0
However, this condition does not fully establish the polarisation direction, which
must also be specified. There are two basic choices from which any other polari-
sation possibilities can be deduced. We can choose the E-field vector of the incident
wave to be either normal to the yz-plane, or parallel to it. The yz-plane in Fig. A.2 is
generally termed the plane of incidence for the incoming wave, being the plane that
312 Appendix A: Free-Space Electromagnetic Waves
contains both the direction vector k ^ and the unit normal to the interface ð^ nÞ. When
the electric field in the incident TEM wave is normal to the plane of incidence, the
wave is said to be perpendicularly polarised, and when it is parallel to this plane it is
described as parallel polarised. Note that in relation to the surface of the earth, while
parallel polarisation equates to horizontal polarisation, perpendicular polarisation
can be termed vertical polarisation only if hi approaches 90°. Perpendicular polar-
isation is often termed transverse electric (TE) propagation while parallel polarised
waves get the complementary description of transverse magnetic (TM) waves.
Now that we know the electromagnetic field forms (TEM waves) remote from
the interface between regions 1 and 2 in Fig. A.2, we can examine the field con-
ditions (boundary conditions) precisely at the interface. For the diffraction set up
depicted in Fig. A.2 with a perpendicularly polarised TEM wave incident at hi the
field directions at a given instant in time can be represented vectorially as shown.
Just at the interface a typical ray of the incident TEM wave is both reflected off the
surface and transmitted through it. Also, for a ‘smooth’ surface ‘common sense’
suggests that it is safe to presume that the reflected and transmitted waves retain the
polarisation of the incident wave. Furthermore, there will be a single reflected ray
and a single transmitted ray. Actually, this latter assumption is not strictly necessary
as we will show presently.
When the TEM wave direction (or ray) lies in paths other than along the
coordinate axes, it is usual to define the ray direction by the vector k which is
chosen to be equal in magnitude to the wave coefficient k. That is k ¼ kk. ^ Hence
we can express mathematically the wave component in any other direction (r say).
For the case shown in Fig. A.3, where the electric field is x-directed, the expression
has the form:
Fig. A.3 Representation of TEM wave with E, H, and k in mutually orthogonal directions
Appendix A: Free-Space Electromagnetic Waves 313
Also
ky ¼ k sin h ðA:34Þ
and
kz ¼ k cos h ðA:35Þ
Hence, employing these relationships, Eq. (A.30) can be expanded into the
non-vectorial form:
where
x x
k1 ¼ ¼ n1 ðA:37Þ
v1 c
For a transverse electromagnetic wave the electric and magnetic fields are related
through Eq. (A.27). Hence on combining Eqs. (A.36) and (A.27), and observing the
field directions in Fig. A.2 we obtain for magnetic fields:
Also, we note that if these field components represent a TEM wave, then we
must have:
Ei g
¼ g1 ¼ 0 ðA:40Þ
Hi n1
314 Appendix A: Free-Space Electromagnetic Waves
Similar constructions lead to the following equations for the reflected and
transmitted field components:
with
Er
¼ g1 ðA:44Þ
Hr
and
where
Et
¼ g2 ðA:48Þ
Ht
and
x
k2 ¼ n2 ðA:49Þ
c
The above field expressions for the incident and reflected waves in region 1 and
the transmitted waves in region 2 each separately satisfy Maxwell’s equations in
these regions. A solution that satisfies Maxwell’s equations for the entire volume
including the interface is achieved by enforcing the electromagnetic field boundary
conditions, given in equations at the interface. That is, at z = 0, we require that
across the divide between regions 1 and 2:
Ex is continuous ðA:50Þ
Hy is continuous ðA:51Þ
Bz is continuous ðA:52Þ
Appendix A: Free-Space Electromagnetic Waves 315
On combining Eq. (A.50) with the field expressions (A.36), (A.41) and (A.45)
we obtain with little difficulty:
This equation must remain true over the entire z = 0 boundary, from
1 y þ 1. This is only possible if:
and
These equations clearly dictate that hr1 = hr2 = hr3 = …, and ht1 = ht2 = ht3 =
… In other words, an ‘optically smooth’ surface produces only one reflected wave
and one transmitted wave.
Equation (A.55) is the source of Snell’s Laws which state that at an optically
smooth interface between two lossless media:
hr ¼ h i ðA:58Þ
sin ht k1 n1
¼ ¼ ðA:59Þ
sin hi k2 n2
However, these laws govern only the reflection and refraction angles. We also
need to have knowledge of the relative magnitudes of the reflected and transmitted
wave, and how these are influenced by material properties.
When Eqs. (A.50), (A.51) and (A.52) are applied to the TEM field components
at the boundary, while also applying Snell’s laws, the following relations are
generated:
316 Appendix A: Free-Space Electromagnetic Waves
Ei þ Er ¼ Et ðA:60Þ
g1 ðHi þ Hr Þ ¼ g2 Ht ðA:63Þ
Hr
qTE ¼ ðA:64Þ
Hi
then making use of Eqs. (A.61) and (A.63) the following useful relationship is
deduced:
g2 cos hi g1 cos ht
qTE ¼ ðA:65Þ
g2 cos hi þ g1 cos ht
This can also be expressed in a slightly more familiar form, which explicitly
incorporates the indices of refraction, namely:
n1 cos hi n2 cos ht
qTE ¼ ðA:66Þ
n1 cos hi þ n2 cos ht
Ht
sTE ¼ ðA:67Þ
Hi
then:
2n2 cos hi
sTE ¼ ðA:68Þ
n1 cos hi þ n2 cos ht
Hr Er
qTE ¼ ¼ ðA:69Þ
Hi Ei
Appendix A: Free-Space Electromagnetic Waves 317
and
Et n1
¼ sTE ðA:70Þ
E i n2
An analogous derivation can also be followed through for the parallel polari-
sation case (TM case). If this is done we obtain:
Er Hr n2 cos hi n1 cos ht
jqTM j ¼ ¼ ¼ ðA:71Þ
Ei Hi n2 cos hi þ n1 cos ht
and
2n1 cos hi
sTM ¼ ðA:72Þ
n2 cos hi þ n1 cos ht
The reflection coefficient, as a function of incident angle for both the TE and TM
cases, is plotted in Fig. A.4. Clearly, for an interface between lossless dielectrics of
differing refractive indices the reflection behaviours are distinct. While for the TE
case it increases monotonically from a magnitude of 0.33 (n1 = 1, n2 = 2) at hi = 0,
to unity at hi = 90°, it drops to zero close to 60° in the TM case. At the zero
reflection angle the two surfaces are said to be ‘matched’ for surface normal wave
components. It is termed the Brewster angle, a physical property which underpins
the design of light polarizers.
Bibliography
B.1 Introduction
As we have seen the TEM solution occurs in unbounded free space. It also forms
in two wire transmission systems such as coaxial line. In hollow pipes, solutions are
of the TE or TM type. A preliminary insight into the nature of propagation in a
hollow rectangular waveguide is provided in Chap. 1. It is evident that the solutions
to Maxwell’s equations in this type of structure divide into two classes, namely
transverse electric (TE) modes and transverse magnetic (TM) modes. The modes
Ph ¼ ^az Ph ðB:1Þ
and
Pe ¼ ^az Pe ðB:2Þ
Other potential function choices are possible depending on the dictates of the
problem to be solved. For example in antenna analyses the functions A (vector
magnetic potential function) and / (electric potential function) are solutions
commonly used (see Chap. 2). Here we concentrate on the TE mode solutions to
illustrate the typical waveguide analysis procedure. In terms of Ph the electric and
magnetic fields in the waveguide are given by:
Appendix B: Basic Waveguide Theory 321
E ¼ jxl0 r Ph ðB:3Þ
H ¼ k02 Ph þ rr Ph ðB:4Þ
r2 Ph þ k02 Ph ¼ 0 ðB:5Þ
pffiffiffiffiffiffiffiffiffi
For an empty waveguide k0 ¼ x l0 e0 ¼ 2p=k0 where k0 is the free-space
wavelength. Since the solution to Eq. (B.5) must represent z-directed propagating
waves in the waveguide, then we can assume that:
In the Cartesian coordinate frame (Fig. B.1), a wave equation such as Eq. (B.7)
can be solved by the method of separation of variables, which means that wh can be
expressed as Hz = X(x)Y(y). Consequently, on applying the boundary conditions
on E and H, and hence wh, at the perfectly conducting walls of the waveguide (see
Fig. B.1), the solution to Eq. (B.7) is easily found to be:
mpx npy
wh ¼ Cðcos cos Þ exp jbmn z ðB:9Þ
a b
With wh now known it is not difficult to establish the fields within the waveguide.
From Eq. (B.4) we obtain:
cmn @Hzmn
Hxmn ¼ 2
kcmn @x
cmn @Hzmn
Hymn ¼ 2
kcmn @y
ðB:14Þ
jxl @Hzmn
Exmn ¼ 2 0
kcmn @y
jxl @Hzmn
Eymn ¼ 2 0
kcmn @x
It is important to observe, that these modal field expressions form (in mathe-
matical terms) a complete set. This means that any arbitrary field distribution in the
waveguide, due for example to the presence of a discontinuity such as a post or an
iris, can always be represented mathematically as a combination or summation of
the known modes of the empty waveguide by a process not unlike the use of
Fourier series to represent arbitrary periodic waveforms with sinusoids.
Appendix B: Basic Waveguide Theory 323
For fast wave solutions in uniform metallic waveguides the propagation equation
has the general form:
c2mn ¼ kcmn
2
k02 ðB:15Þ
where kcmn is a constant whose magnitude is governed by the size and shape of
the waveguide, and k=k0 is the plane wave phase constant for the interior medium.
For rectangular waveguide, we have already seen that:
mp2 np2
2
kcmn ¼ þ ðB:16Þ
a b
It has other dimensionally dependant values for other waveguides, such as cir-
cular waveguide and elliptical waveguide. However, for all such waveguides the
propagation equation has the same form, as expressed in Eq. (B.16).
In a lossless waveguide for which kcmn is real and constant, and
pffiffiffiffiffiffiffiffiffi
k0 ¼ x l0 e0 ðB:17Þ
is real and linearly related to frequency, two types of modal solution exist in
waveguide.
These are:
1. Those for which ko > kcmn giving
1=2
c ¼ jbmn where bmn ¼ k02 kcmn
2
ðB:18Þ
bmn is termed the phase coefficient in rad/m for the mnth mode. The exponential
multiplier in the field expression now has the form:
expðjðbmn z xtÞÞ
2 1=2
cmn ¼ amn ¼ kcmn k02 ðB:19Þ
expðamn zÞ expðjxtÞ
and this implies that the mode produces a field which attenuates in z. It does not
propagate. Such a mode is said to be evanescent.
The frequency at which ko = kcmn is termed the cut-off frequency. It is the
frequency at which, in a loss-less waveguide, a given mode changes abruptly from
propagating to non-propagating. For any waveguide there is in general a separate
and distinct cut-off frequency for every TE mode, and every TM mode, in the mode
set. On the other hand it is not uncommon for a TE mode and a TM mode to have
the same cut-off frequency. Such modes are said to be degenerate. For a TE mode
and a TM mode in rectangular waveguide the cut-off frequency is evaluated as
follows:
Let xc denote the frequency at which ko = kcmn. Then:
r ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffi mp2 np2
xc l0 e0 ¼ þ
a b
r ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
mp 2 np2
i:e: xc ¼ c þ
a b
s ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðB:20Þ
c m 2 n2
Hence fcmn ¼ þ 2
2 a2 b
c pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi
¼ ðm2 b2 þ n2 a2 Þ
2ab
2ab
kcmn ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðB:21Þ
ð m b2 þ n2 a2 Þ
2
The mode with the lowest cut-off frequency, which means that it can exist alone in a
smooth empty waveguide, is the TE10 mode. The field expressions for this mode (in
magnitude), using Eqs. (B.11) and (B.14), are:
Appendix B: Basic Waveguide Theory 325
px
Hz ¼ cos
a
Hy ¼ 0
b a px
Hx ¼ j 10 sin ðB:22Þ
p a
Ex ¼ 0
ka px
Ey ¼ j g0 sin
p a
expðjðb10 z xtÞÞ
The functional forms in Eq. (B.22) are sufficiently simple to permit the con-
struction of field distributions in pictorial form, as shown in Fig. B.2. This is also
true of the next dozen or so TE and TM modes as many texts on electromagnetic
theory demonstrate [see Bibliography].
For this TE10 mode the cut-off frequency is given by:
c
fc10 ¼ and
2a
kc10 ¼ 2a
Fig. B.2 TE10 mode field patterns, a E and H-fields in the y-z plane, b E-field in the x-y plane, c E
and H-fields in the x-z plane
solutions for uniform waveguide are termed ‘fast’ waves. Below cut-off, k0a = p
for the TE10 mode, the solutions are evanescent as we have already seen and these
cut-off solutions are represented by the quadrant of a circle (blue dotted curve for
TE10 mode—it is not quite circular because of the slightly different axis scales). The
attenuation coefficient a in nep/m can be deduced from the horizontal distance from
the frequency of interest on the k0a axis to the dotted curve of the mode of interest.
Appendix B: Basic Waveguide Theory 327
Fig. B.3 Brillouin diagram for TE10 mode showing frequency (k0a) as a function of phase (b1oa)
for rectangular waveguide (dark blue curve: a = 2 cm, b = 1 cm), and as a function of a10a (green
dotted curve). Orange line represents the speed of light line v = c
Bibliography
1. R.E. Collin, Field Theory of Guided Waves (McGraw-Hill Book Co., New York, 1960)
2. R. Ferrari, An Introduction to Electromagnetic Fields (Van Nostrand Reinhold Co. Ltd., New
York, 1975)
3. C.T.A. Johnk, Engineering Electromagnetic Fields and Waves (Wiley, New York, 1988)
4. R.P. Feynman, R.B. Leighton, M. Sands, Lectures in Physics, Vol. II (Addison-Wesley
Publishing Co., London, 1972)
5. P.M. Morse, H. Feshbach, Methods of Theoretical Physics (McGraw-Hill Book Co., New
York, 1953)
6. J.D. Kraus, Electromagnetics (McGraw-Hill Book Co., London, 1984)
Appendix C
Waveguide Green’s Function
and Scattering Parameters
For the moment method solution of broadwall slot radiators in empty rectangular
waveguide, the following are the relevant magnetic dyadic Green’s functions for the
geometry of Fig. 5.1.
(1) The exterior region usually treated as a half-space is (see Morse and Feshbach
and Collin in the Bibliography):
(2) The appropriate Green’s function for an arbitrary slot radiator in waveguide
[see Chap. 5, Eq. (5.5)] is the magnetic dyadic Green’s function. However the
problem can be simplified by noting that an inclined slot radiator in the
broadwall of waveguide as depicted in Fig. 5.1 can generally be resolved into
an x-directed (transverse) element and a z-directed (longitudinal) element.
Actually in practice broadwall waveguide slot radiators are either purely
transverse or wholly longitudinal. Inclined slots are hardly ever seen. This
means that the magnetic dyadic Green’s function for the feed waveguide can be
separated into a part that is relevant to the transverse slot computation and a part
that applies to the longitudinal slot.
As has already been hinted at above the magnetic dyadic Green’s function for
waveguide, see Collin in Bibliography, can be constructed by a summation of the
complete set of modes for the rectangular box forming the waveguide. That is:
8
> P
1
> 1
< 2jxe0 Hnþ ðrÞH
n ðrÞ z [ z0
ma ¼
G n¼1
ðC:2Þ
> P1
>
: 2jxe
1
H þ
n ðrÞHn ðrÞ z\z0
0
n¼1
where
H
n ¼ ðhn þ hzn Þ expð Cmn zÞ ðC:3Þ
E
n ¼ ðen ezn Þ expð Cmn zÞ ðC:4Þ
These vector fields, representing the normal mode field solutions for the
waveguide, are normalised to be in accord with:
ZZ
½en hm ^az dS ¼ dnm ðC:5Þ
Sx
while
mp2 np2
Cmn ¼ þ k02 ¼ kc2 k02 ðC:6Þ
a b
By combining Eq. (C.2) through (C.9) the Green’s function component relevant
to transverse slot modelling is as follows
Appendix C: Waveguide Green’s Function and Scattering Parameters 331
8 1 1 9
> P P npy >
>
> ðxe Þ2 np 2 2
ð Þ j jÞ mpx0 npy >
b >
mpx
> 0 B expðC mn z z0 sin sin cos cos >
>
< m¼1 n¼1
b mn a a b >
=
1
xx
Gma ðr=r0 Þ ¼ P1 P1
mp 2 2
2jxe0 >
> ð a Þ Cmn Amn expðCmn jz z0 jÞ sin a sin a cos b cos b >
2 mpx mpx0 npy npy
>
>
> >
>
>
:
m¼0 n¼0 >
;
n 6¼ m ¼ 0
ðC:10Þ
The factors Amn and Bmn arise from the normalisation process represented by
Eq. (C.10) and are given by:
eom eon
A2mn ¼ ðC:11Þ
jxl0 Cmn kc2 ab
4
B2mn ¼ ðC:12Þ
jxe0 Cmn kc2 ab
where eom and eon are equal to zero if m ¼ 0 or n ¼ 0 otherwise they are equal to
unity.
For the longitudinal slot radiator in rectangular waveguide it is shown in Collin
that the sum of modes formulation produces a function for which non-convergence
issues arise. However a direct solution to Eq. (5.5) in Chap. 5 together with the
appropriate waveguide boundary conditions has been determined by Rahmat-Samii.
The result is the following:
X1 X 1
eom eon mpx mpx0 npy npy0
zz ðr=r0 Þ ¼
G expðCmn jz z0 jÞ cos cos cos cos
ma
m¼0 n¼0
2abC mn a a b b
ðC:13Þ
(3) For a waveguide with finite wall thickness (T) the slot is treated as a rectangular
cavity of dimensions L W T, with a local coordinate system n; g; 1 as
indicated in Fig. 5.1. For this cavity the relevant magnetic dyadic Green’s
function (consistent with n-directed magnetic currents at open ends) can be
formulated heuristically following the summation of modes approach in Collin
(see Bibliography). The slot is viewed as a very short (length = T) of rectan-
gular waveguide (a replaced by L: b replaced by W). This yields:
332 Appendix C: Waveguide Green’s Function and Scattering Parameters
X
1 X
1
mb ¼ el0 en0 lpðg þ W2 lpðg0 þ W2 Þ npðn þ L2Þ npðn0 þ L2Þ
G cos cos sin sin
l¼0 n¼0
WL W W L L
" #
1ðkn0pLÞ2 cosðkg 1Þ cosðkg ð10 þ TÞÞ
1 6¼ 10
kg sinðkg TÞ cosðkg ð1 þ TÞ cosðkg 10 Þ
ðC:14Þ
where
np 2 lp
kg2 ¼ k02 ð Þ ð Þ2 ðC:15Þ
L W
For a slot in waveguide, or in any metallic surface, excitation of the slot is asso-
ciated with magnetic field components aligned with its major axis, and hence
current density flows normal to this axis (see Chap. 1). Consequently for the
longitudinal slot in waveguide as shown in Fig. 5.1, with h = 0°, the component of
H10 which is effective in exciting the slot is the z-component H10z. From
Appendix B, this has the form:
p2 px
H10z ¼ A10 cos expð C10 zÞ ðC:16Þ
a a
with
ðC:17Þ
On combining Eqs. (C.16) and (C.17) the reflection coefficient (S11) for back-
ward scattering into the TE10 mode in the waveguide we obtain:
Appendix C: Waveguide Green’s Function and Scattering Parameters 333
N " #
A10 p pðx1 þ w=2Þ pðx1 w=2Þ X sp
S11 ¼ pffiffiffi sin sin as L
The transmission coefficient (S12) directly from the TE10 mode magnetic field, as
above, or from the relation:
By conservation of energy the power radiated (prad) by the longitudinal slot can
be deduced to be:
As far as the calculation of the waveguide scattering and power radiated is con-
cerned the T-slot radiator differs little from the conventional longitudinal slot
presented above but using moment method results presented in Chap. 5, Section 5.4
. However the derivations are slightly complicated by the fact that the T-slot scatters
into both halves of the bifurcated waveguide. The internal scattering coefficients
þ
have therefore been considered to comprise two components S11 and S11 repre-
senting backward scattering into the left hand (x-positive) and right hand
þ
(x-negative) waveguides respectively, while S12 and S
12 are the corresponding
forward scattered components. With these definitions established it is not difficult to
show that:
b10 w L p
S
11 ¼ 2jA10 sin a1 r ; ðC:22Þ
2 2 a
for s = 1 since, as we have noted in Chap. 5, the first basis function is enough to
secure good results, and
334 Appendix C: Waveguide Green’s Function and Scattering Parameters
where b is the phase coefficient for the slot transmission line. In this case the
transmission coefficient can be calculated from:
S
12 ¼ 1 S11 ðC:24Þ
Bibliography
1. P.M. Morse, H. Feshbach, Methods of Theoretical Physics (McGraw-Hill Book Co., 1953)
2. R.E. Collin, Field Theory of Guided Waves (McGraw-Hill Book Co., 1960)
3. R.E. Collin, On the incompleteness of E and H modes in waveguides. Can. J. Phys. 51, 1135–
1140 (1973)
4. Y. Rahmatt-Samii, On the question of computation of Dyadic Green’s function at the source
region in waveguides and cavities. IEEE Trans. Microwave Theory Tech. 23, 762–765 (1975)
Appendix D
Green’s Function for Fin Loaded
Waveguide
The Green’s function for slow-waveguide periodically loaded with metallic fins can
be developed heuristically from the well-established empty rectangular waveguide
formulation (see Appendix C). The rectangular waveguide Green’s function in
Appendix C can be separated into three elements.
(1) The real part of the dominant mode term.
(2) The imaginary part of the dominant mode term.
(3) The imaginary part formed from the higher order all of which can generally be
assumed to be evanescent modes.
By analogy it is not unreasonable to suggest that a Green’s function for the
fin-loaded rectangular waveguide will display similar components. In relation to
item (3), the effect of periodic loading on the form of a potential Green’s functions
for a slow-wave structure has been studied elsewhere and clearly indicates that in
the space a b above the corrugated surface the evanescent modes are essentially
the same as for the equivalent empty waveguide. Thus referring to Eq. (C.10) in
Appendix C we can deduce that the evanescent mode contribution to the require
Green’s function is:
8P 1 P 1 2 2 9
>
> ðxe0 Þ2 np npy >
>
b Bmn expðCmn jz z0 jÞ sin a sin a cos b cos b >
mpx mpx0 npy
>
> >
>
> m¼1 n¼1 >
>
>
< 1
>
=
1 P1 P
G
xx
ðr=r Þ ¼ mp 2 2
C A2
expðC mn j z z 0 jÞ sin mpx
sin mpx0
cos npy
cos npy
=ma 0 mn mn
2jxe0 >
> m¼0 n¼0
a a a b b
>
>
>
> >
>
>
> n 6¼ m ¼ 0 >
>
>
: >
;
m ¼ 1 : n 6¼ 0
ðD:1Þ
ph ¼ ^ax wh ðD:2Þ
ðD:4Þ
As noted above the third term in Eq. (D.4) does not contribute to radiation from
a transverse slot in the y = b wall and can be ignored. Finally
xx ðr=r0 Þ ¼ G
G xx ðr=r0 Þ þ G
xx ðr=r0 Þ ðD:5Þ
ma <ma =ma
Index
A novel, 272
Admittance reciprocity, 261
transfer, 39 reflector, 57
Ammeter, 39 retro-directive, 31, 286, 299
Ampere’s law, 47, 48 simulation, 167
Amplifier synthesis, 133
buffer, 289 thinning, 260
voltage, 288 Aperture
Antenna antenna, 57, 69, 78
3-D, 21 blockage, 182, 225
1-D format, 269 circular, 293
active, 250, 267, 302 dual-mode, 298
aperture, 57, 75 effective area, 79
broadside, 248 elliptic, 296
cavity resonator, 216, 249, 267, 274 field distribution, 251
compact, 8, 28, 40, 151, 243 linear, 72
configurable, 263 radiating, 251
conformal, 270 rectangular, 72, 76, 79
dielectric rod, 57 source, 21, 35, 37, 62, 69, 80
directional, 71 theory, 216
directivity, 78 two dimensional, 72, 74, 78
efficiency, 72 Array
Fabry-Perot, 262 analysis, 140
flat plate, 169 antenna, 140
gain, 79 apeture, 293
horn, 60 beam squint, 173, 179
impedance, 38 bidirectional, 147
infrared, 216 binomial, 148
integral equation, 226 boresight, 173
leaky-wave, 19, 57, 205, 208, 213, 215, broadside, 134, 140
243, 260, 267, 272, 281 cavity resonator, 249
lens, 57 colour coded, 225
long-slot, 207 compact, 179, 220, 270, 285
macroscopic, 38, 39 control, 223
mm-wave, 216, 237 design, 154
Hankel K
function, 220 Kernel, 82
Hansen-Woodyard, 138 Kirchoff, 62
Helmholtz, 41, 84 Kirchoff law, 43
equation, 87 Kronecker delta, 197
vector equation, 41, 44
Helmholtz equation, 105 L
Hemisphere, 79 Laser
Hertzian machining, 170
dipole, 88, 90, 93 Law
element, 217 inverse square, 45, 54
Hertzian dipole, 50 Leaky
Heterodyne, 299 ladder, 212
Heuristic, 7, 81 Leaky-wave
Horizontal plane, 70 1-D scan, 269
Horn antenna pattern, 249
circular, 230 array antenna, 249
corrugated, 230 attenuation, 207, 212
phase centre, 223 linear array, 209, 267
pyramidal, 236 long slot antenna, 207
Huygen phase shift, 272
method, 89 phenomenon, 211
principle, 63 radiation, 262
Hybrid slit, 217
coupler, 25 slit width, 217
quadrature, 297 Lens
flat, 280
I optical, 280
Illumination, 71 parabolic, 280
uniform, 71 Line
Image current, 146
focused, 280 parallel, 151
Impedance, 38 shorted, 119
characteristic, 55 voltage, 118
complex, 205 Linear
input, 116, 119 operator, 82
load, 119 polarisation, 8, 115, 301
mutual, 216 space, 82
Inductor, 8 Linear accelerator, 183
Inhomogeneous, 41, 84 Load
Inmarsat, 31 matched, 154
Interference Specification, 165
pattern, 65 Lobe
Iris grating, 176
capacitive, 183, 206 grating condition, 219
Isotropic primary, 156
elements, 155 Longitudinal slot, 210
Iteration, 166 Lord Rayleigh, 35
Lorentz
J condition, 43
Junction reciprocity, 39
summing, 290 Loves principle, 59
Index 343
M T-junction, 301
Macroscopic Microstripline, 173
values, 275 Microwave, 4
Magic-tee, 122 circuit, 25
Magnet frequency band, 215
permanent, 47 integrated circuit, 173
Magnetic Mirror, 12
charge, 37 flat, 223
charge density, 61 Mixer, 289, 301
conductor, 271 conjugating, 286
current, 37, 200 gate, 300
current density, 59 heterodyne, 299
current element, 200 quasi-optical, 301
flux density, 47, 48 up-converting, 290
source, 63 Mobile, 10
vector potential, 43, 51 communications, 301
Magnetic field, 14 platform, 301
Magnetostatic Mode
analogy, 44, 47 anti-symmetrical, 188
Magnetostatics, 43 backward TE, 272
Mainbeam backward wave, 279
Mainlobe cut-off, 188
nose, 78 degenerate, 293, 324
width, 71 dominant, 204
Material evanescent, 15
double negative, 275 even, 9, 22
MATLAB, 189 fast, 213
Matrix, 200 hybrid, 195
entries, 199 improper, 204
equation, 258 leaky-wave, 268
impedance, 271 odd, 9, 22
manipulation, 200 orthogonal, 15
Matrix form orthogonality, 197
Maximum parallel plate, 259, 262, 265, 268
principal, 134 perturbation, 206
secondary, 134 quasi-TE10, 245
Maxwell, 35, 84 receive, 38
equations, 81 slow-wave, 195
Measurement sum of, 331
error, 113, 126 surface wave, 253
Medium symmetrical, 188
anisotropic, 45 TE, 10
homogeneous, 45 TE10, 13
inhomogeneous, 45 TEM, 9
isotropic, 39 transmit, 38
non-linear, 45 transverse resonance, 202, 205
Metallodielectric, 275 Model
Metamaterial, 267, 272 ComsolL, 216
frequency selective, 272 T-block, 235
Meta-structures, 179 Modulator
Metasurface, 275 IQ, 290, 291
Microstrip, 28 Moment method, 10, 21, 96, 99, 189, 190, 199,
feed, 293, 297 200, 205, 226, 227, 233
patch, 300 concise, 97
344 Index