Chapter 1: Introduction

Chapter 1: Introduction

In this chapter a brief description of geostationary satellites, the motivation for the
study of reconfigurable contour beam reflector antennas and this study in particular,
and an outline of this dissertation are presented.

1.1. Geostationary satellites

Satellites have been in practical use in telecommunications since Echo 1, a 26.5 inch
magnesium sphere launched by a Thor Delta rocket on August 12, 1960 bounced a
taped message transmitted from Goldstone, California that was received by the Bell
telephone laboratory a Holmdel, N.J. Echo I stimulated a great deal of interest in the
development of active communication which lead American Telephone and Telegraph
Company (AT&T) to build Telstar, launched on July 10,1962. Telstar was an active
satellite with a microwave receiver and transmitter. It was the first satellite to transmit
live television and conversations across the Atlantic.

Geostationary satellites were proposed in 1947 by Arthur C. Clark (1917-), a British

physicist and astronomer as a means to relay radio signals from one part of the world
to another that is beyond the line of sight. Geostationary orbits are orbits occupied by
communications satellites which remain at fixed points in the sky relative to observers
on the ground. The are defined by an orbit period of one sidereal day, or about 23
hours 56 minutes 4 seconds. During one sidereal day the earth rotates about its polar
axis exactly once. To be geosynchronous, a satellite must orbit the earth in the same
period . This period defines the average orbit radius of 42155 km. This value is found
from Kepler's third law. The earth' s radius (6370 km) subtracted from the orbit radius
determines the orbit above the earth to be 35785 km. This definition doesn 't say
anything about the shape of the orbit, or the orientation of the orbit plane with respect
to the plane of the equator. The orbit can be highly elliptical, and/or it can be inclined
with respect to the plane of the equator, and still be synchronous with the earth 's
rotation. In this case, a desired class of geosynchronous orbit is the geosationary orbit

Department of Electrical and Electronic Engineering 1

University of Pretoria
 Chapter 1: Introduction

A satellite moving in a geostationary orbit remains at a fixed point in the sky at all

times. This is desirable for radio communications because it allows the use of

stationary antennas on the ground.

To be geostationary, the orbit must meet three criteria:

- The orbit must be geosynchronous.

- The orbit must be a circle.

- The orbit must lie in the earth's equatorial plane.

Individual satellites within the orbit are identified by the longitudinal position east or

west of the prime meridian.

All geostationary orbits comply with the following :
PARAMETER METRIC UNITS
Height above equator 35785 km
Average orbit radius 42155 km
Orbit circumference 264869 km
Arc length per degree 736 km
Orbital velocity 11066 kmIh
To understand the importance of these criteria, consider the result if the orbit fails to
meet them. If the orbit is not geosynchronous, the satellite does not move at the same
rate as the earth's rotation. Thus, from the point of view of an observer on earth, the
satellite appears to be in continuous motion, and it periodically disappears below the
horizon. If the orbit is not a circle, the satellite does not move at a constant velocity
(Kepler's second law). Instead, it appears to oscillate east-and-west at a rate of two
cycles per sidereal day. If the orbit does not lie in the equatorial plane, the satellite
does not remain at a fixed point in the sky. Instead, it appears to oscillate north-and­
south at a rate of one cycle per sidereal day. The terms geosynchronous and
geostationary are not synonymous: geosynchronous specifies only the orbit period, but
geostationary also specifies the shape and orientation of the orbit.
These definitions are consistent with the definitions used by the United States Federal
Communications Commission (FCC). The following definitions are quoted from the
FCC rules, as published in Title 47, Section 2.1, of the United States Code of Federal
Regulations:

Department of Electrical and Electronic Engineering 2

University of Pretoria
 Chapter 1: Introduction

Geostationary satellite: An earth satellite whose period of revolution is equal to th

eperiod of rotation of the earth about its axis .

Geostationary satellite: A geosynchronous satellite whose circular and direct orbit lies

in the plane of the earth's equator and which thus remains fixed relative to the earth; by

extension, a satellite which remains approximately fixed relative to the earth.

The circular belt containing all the geostationary satellites is called the Clarke belt.

Specific satellite positions in the Clarke belt are identified by longitude (or, more

specifically, by the longitude of the point on the equator directly beneath the satellite).

Figure 1 shows an example of a satellite in the Clarke belt.

The view from one geostationary satellite covers about 40% of the earths suface. At

the equator, a 162°-segment of the Clarke belt is visible; the visible segment decreases

as the latitude increases, and becomes zero at a latitude of 81.4°. North of 81.4° north

latitude and south of 81.4° south latitude the Clarke belt is hidden below the horizon.

Dozens of satellites have been deployed along the Clarke belt in order to accommodate

the ever-growing demand for communications capacity. In many parts of the Clarke

belt, adjacent satellites use the same frequency band and are located within 2° of each

other. A satellite intended for radio communications among fixed earth stations must

remain at a fixed point in the sky. This means that the satellite must move in a

geostationary orbit. The owners of most geostationary satellites try to maintain their

satellites in a box measuring O. 1° x 0.10. The satellite must be maintained at the proper

attitude. This term describes the orientation of the satellite within its box. If the

satellite IS not maintained at the proper attitude, its antennas will not be aimed

properly.

Department of Electrical and Electronic Engineering 3

University of Pretoria
 Chapter 1 : Introduction

Figure 1. The subtended angle of the earth as seen from a satellite in the Clarke belt is
approximately 17°.

Once a satellite is placed in proper position and attitude it tends to drift. Drift degrades
the satellite performance in two ways: the satellite may move out of position, or it may
assume an improper attitude. Drift results from external forces. While there are many
external forces acting on the satellite, the primary forces are those exerted by the sun
and other objects in the solar system.
The intensity and direction of the gravitational force exerted by the sun changes
continuously in daily, yearly and 55-year cycles. The cyclic nature of this force tends to
cancel its own effect; an easterly pull at one part of the cycle is offset by a westerly pull
half a day later; similarly a northerly pull is offset by a southerly pull but there is a net
resultant force which, over the course of several months causes the satellite to drift
away from its geostationary position.
The gravitational pull of other objects in the solar system are considerably weaker than
the sun's gravity, and their effects can be measured and predicted. Precise orbit
calculations usually account for the moon's gravity, and frequently include the
gravitational forces of other planets as well. The uneven distribution of land mass on
the surface of the earth also causes mainly east-west drift . To counteract these forces,
the satellite must be fitted with some mechanism to move the satellite back into
positions when it drifts.

Department of Electrical and Electronic Engineering 4

University of Pretoria
 Chapter 1: Introduction

If the satellite is allowed to drift freely two effects manifest itself. First the orbit plane
becomes inclined with respect to the earth's equatorial plane. During the course of one
sidereal day, the satellite makes one complete revolution around the earth. The orbit
plane must pass through the earth's center of gravity (Kepler' s first law) which means
the satellite must pass through the earth's equatorial plane twice each sidereal day. The
satellite is north of the equatorial plane for half of each sidereal day and south of it for
the other half. From a point on the earth's surface, the satellite appears to oscillate
along its north-south axis at a rate of one cycle per sidereal day. Next, because of the
conservation of angular momentum, the orbit assumes and elliptic shape. As a result,
the satellite no longer moves at a constant velocity (Kepler' s second law). From a
point on the earth's surface, it appears to oscillate along its east-west axis at a rate of
two cycles per sidereal day. Combining these two apparent motion the result is an
elongated 'figure-8' pattern. The satellite complete one cycle along the figure-8 pattern
each sidereal day. As the satellite continues to drift the figure-8 pattern becomes larger.
Communication satellites are fitted with small rockets called thrusters. On command
from a control station, a thruster is fired. During its firing, it ejects a gas propellant.
The ejected gas produces the force to counteract these undesired motions. A ground
control station precisely controls all the parameters involved in a firing: the position of
each thruster relative to the satellite, the timing and duration of each fire and the
pressure of the ejected propellant. If these parameters are controlled properly, the
satellite can be maintained at the proper position and attitude for years. This process is
called stationkeeping.

Every time a thruster is fired, propellant is used. Once the supply of propellant is
exhausted the satellite cannot be maintained at proper position and attitude, and the
satellite must be retired. Propellant capacity is the primary factor which determines the
useful life of a communications satellite. A primary goal of every satellite owner is the
conservation of propellant. Many studies have been done to determine the optimum
trade-off between satellite stability and propellant usage. These studies have shown
that a substantial majority of the propellant is used for just one stationkeeping function:
keeping the satellite from drifting along its north-south axis. When the propellant is
spent the satellite loses the ability of stationkeeping and becomes useless to the satellite

Department of Electrical and Electronic Engineering 5

University of Pretoria
 Chapter 1: Introduction

operator. The last of the propellant is used to decommission the satellite into graveyard
orbit, which is usually just a highly inclined geosynchronous orbit to prevent
congesting the Clarke belt.
Since 1963, approximately 400 satellites have been placed in geostationary orbit.
Conservatively assuming an average lifetime of 8 years per satellite, these satellites
have accumulated around 3200 years of in-orbit operation. As satellites are becoming
more reliable and launch vehicles get better mass into orbit capability, the expected
service life of the satellites will increase. Satellites are already achieving an expected
operating service life of 15 years. This increases the probability that the satellite service
area and/or satellite operator will change. The Canadian Anik series of satellites is an
example of this. Anik C 1 was launched on April 12, 1985 from the space shuttle
Discovery during the STS-51D mission to a position at n ow. Anik Cl was built for
Telesat Canada by Hughes Aircraft Co. , with Spar Aerospace Ltd. And other Canadian
companies as subcontractors. It was owned and operated by Telesat Canada until it
was replaced by later Anik satellites and sold to Paracomsat, an Argentine operator.
Anik C2, otherwise known as Telesat 7, like Anik Cl , was also sold to Paracomsat.
Both Cl and C2 were later bought back by Telesat, leased to the UAB and were later
used to provide coverage to the northern regions of Antartica. This example illustrate
that the application and required geographical coverage of a geostationary satellite is
very likely to change.

1.2. Antenna systems on geostationary satellites

Telecommunication services provided by satellite include television and telephone

transponders and direct broadcast television (DBS). In the case of the first two
services an operator will provide a service center or hub from where uplinks and
downlinks to the satellite are made. This will be made from ground stations with high
gain antennas. From the hub, the service is relayed into the terrestrial network. In the
case of direct broadcast television a downlink service is provided to many users each
using a lower gain antenna. For example, an 18" aperture parabolic reflector antenna is
used as a DBS receiver. The antenna on the satellite needs to provide coverage over a

Department of Electrical and Electronic Engineering 6

University of Pretoria
 Chapter 1: Introduction

geographical regIOn called the servIce area as opposed to a single beam to each
individual ground station. Shaped or contour beams are used on these satellites to
increase antenna efficiency and reduce interference in geographical areas adjacent to
the service area. The need for shaped or contour beams was a significant challenge to
antenna engineers and several methods of implementing contour beams have been
studied and used. These include arrays, array front fed paraboloids and shaped single
and dual reflector antenna systems.

From the examples in the previous paragraph, it can be seen that there will be a definite
advantage in the ability to reconfigure the contour beam to provide coverage for
different geographical service areas and from different satellite geostationary positions.
In order to comply with FCC regulations on the level of radiation allowed in areas
outside the geographical coverage area, contour beams are subject to much more
stringent specifications and this is also likely to be enforced on reconfigurable beams.

Reconfigurable contour beams can be implemented in a number of ways, including

large aperture arrays, multiple feed reflector antennas and reflector antennas with
adjustable main- and/or sub reflector surfaces. The contour beam reflector antenna
(CBRA) is widely used because of its versatility and low cost per unit aperture. The
disadvantage of using the array fed offset reflector is that the beamforming network is
heavy, lossy and expensive. The same disadvantage applies to the phased array
antenna. Both type have complex components that need to be space qualified. Space
qualification includes thermal, electromagnetic compatibility and mechanical (shock
and vibration) tests and account for a significant portion of the total cost of the
antenna subsystem [26]. The relative low cost of the shaped reflector antenna made it a
popular choice for direct broadcasting satellites. The obvious disadvantage used to be
the inability to reconfigure the contour beam. A reconfigurable CBRA has been
implemented using an adjustable mesh main reflector and a cluster feed arrangement
[4]. Another degree of freedom is added if the subreflector of a dual offset reflector
antenna can be made adjustable. Piecewise adjustable subreflectors have been used in
the past to correct for gravitational distortion in large axally symmetric dual reflector

Department of Electrical and Electronic Engineering 7

University of Pretoria
 Chapter l : Introduction

radio telescope antennas [1,2] and have also more recently been proposed as a way to
correct for main reflector distortion in dual offset reflector (DOSR) antennas [3].

1.3. The mechanical finite element diffraction synthesis technique

In this dissertation a novel way to design, synthesize and adjust the reconfigurable dual
offset contour beam reflector antenna (DCBRA) using an adjustable subreflector is
described . The DCBRA have been studied in this work using a variety of
electromagnetic and mechanical analysis techniques which will be described in this
dissertation. The reflector surfaces are treated by using a mechanical finite element
surface description in a reflector diffraction synthesis code. The mechanical FEM
module of the synthesis code was developed by the Smart Materials and Structures
Division of the Mechanical Engineering department of The Ohio State University. The
mechanical finite element code uses a shell element description and gets integrated into
diffraction synthesis software to create a unique tool for studying problems like
actuator placement, material property effects on the design and the achievable contour
beam coverage. Reflector surface adjustment is studied using a set of linear actuators
on the back of a stiff metal coated material and by bonding piezoelectric material onto
the surface and applying a controlling voltage to it to change the shape [6]. Studies are
mainly done for spaceborne applications, taking into account the fact that mechanical
actuators can be difficult to design for use in space where smearing of mechanical
components can be impossible in some cases due to the sublimation of smearing fluids
like grease and oil.

The diffraction synthesis procedure and methods to calculate the far-field of the DOSR
antenna efficiently is discussed in Chapter 2. This chapter includes a brief description
of methods studied as possible candidates (including the Jacobi-Bessel method) for use
in the diffraction synthesis code and a motivation for the choice of the selected method,
a FFT based method, is given. Also included in this chapter is a description of the
Gaussian beam method developed by Pathak [4]. This method was used for the first

Department of Electrical and Electronic Engineering 8

University of Pretoria
 Chapter 1: Introduction

time by Chou and Theunissen to synthesize a contour beam for an offset front fed
parabolic reflector [5] and the advantages ofthis method is shown in this chapter.

The method of optimization is described in Chapter 3. The calculation of the cost

function, that is the function that gets minimized during synthesis, is described and
techniques used to minimize this function are introduced. In this chapter the
advantages and disadvantages of global and local search techniques for this application
are discussed and two methods are compared . These methods are the genetic algorithm
and the steepest gradient solver. Both were used individually in some beam syntheses
and also used in combination in other syntheses.

In Chapter 4 the mechanical properties of thin sheets are discussed and the set of
differential equations governing their shape under different forces is shown. The
feasibility of building a reconfigurable reflector antenna is demonstrated by a practical
mechanical design using piezoelectric adjustable linear actuators. The design is based
on a mechanical finite element analysis of four prototype surfaces and a subsequent
actuator placement study. In this study, the main reflector was assumed to be fixed
and an adjustable subreflector is designed using a flexible material called Lexan. An
actuator placement study is described that was performed on materials with various
stiffnesses to determine the suitability for this application by the Smart Structures
division of the Mechanical Engineering department of The Ohio State University.

The diffraction synthesis procedure written by the author was modified to incorporate
a mechanical finite element description of the surfaces of a dual reflector antenna. The
mechanical FEM code designed by Yoon forms a unit in the contour beam synthesis
software. This enables direct synthesis in terms of the exerted actuator forces on a
surface with a predefined stiffuess matrix. This also eliminates the second step of the
design of a reconfigurable dual offset reflector antenna, the actuator placement study.
In addition, many iterations can be performed much faster and more convenient, as
opposed to the example in Chapter 4 where essentially only one mechanical iteration is
done with considerable effort. The mechanical FEM diffraction synthesis software
creates a unique and very useful tool to create a suitable design and predict the

Department of Electrical and Electronic Engineering 9

University of Pretoria
 Chapter 1: Introduction

performance of an antenna taking into consideration the mechanical properties of the

reflector surface materials and the actuators used to reconfigure the antenna. The
effect of different surface parameters and the required number of actuators and their
placement were also studied and is described in Chapter 5. The technique is illustrated
through a design of a reconfigurable DOSR antenna.

The dissertation concludes in Chapter 6 with a summary and conclusion.

Department of Electrical and Electronic Engineering 10

University of Pretoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
antennas
Chapter 2: Diffraction synthesis and radiation pattern computation
for reflector antennas

The diffraction synthesis procedure and methods to calculate the far-field of the DOSR
antenna efficiently will be discussed in Chapter 2. This chapter includes a brief
description of methods studied as possible candidates (the Jacobi-Bessel and p-series
methods) for use in the diffraction synthesis code. The selected method, the p-series
method, is compared in terms of accuracy and eft1ciency to a physical optics code
developed at the Ohio State University by Lee and Rudduck [7]. Also included in this
chapter is a description of the Gaussian beam method developed by Pathak [4]. This
method was used for the first time to synthesize a contour beam tor a front fed offset
parabolic reflector by Chou and Theunissen [5] and the advantages of this method is
shown in this chapter.

The analysis and design of reflector antennas evolved trom the early numerical
integration approaches to powerful techr..iques, such as the Jacobi-Bessel method, the
Fourier-Bessel method, and sampling methods. One of the first attempts at improving
the efficiency of the integration of the surface current density on the reflector was
made by Ludwig. This was followed by Rusch's method [8]. Both these methods make
use of asymptotic solutions to the radiation integral.
New integration techniques for the design of large, tocused reflector antennas came in
the last t'vvo decades. These include the Fourier-Bessel, Jacobi-Bessel and pseudo­
sampling techniques. More recently, a Gaussian beam analysis technique was
introduced. This method involves a closed form description of the reflected and
diffracted fields of Gaussian beams from doubly curved surfaces with edges, which
allo'.vs one to compute the far -field of reflector antennas extremely efficiently [4]. In
this analysis the observer can be in the Fresnel or tar-field region of the reflecting
surface. Tr..is techr..ique also gives the ability for diffraction synthesis of the near-field
of a reflector antenna for special applications where the reflector is used in its near

Department of Electrical and Electronic Engineering 11

University of Pretoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
antennas
field . This approach will have a very significant impact on future satellite antenna
designs.

The calculation of the radiation pattern of the sub-reflector is performed by using

techniques such as the GO/GTD approach [9] where the dim-acted field contribution is
approximated by a ray-optics field (analogous to the geometrical optics reflected field)
which is regarded as originating at a point (or points) located on the rim of the sub­
reflector.

Reflector shaping techniques for symmetrical and offset dual reflector systems also
evolved from the early geometrical optics techniques to diffraction synthesis [3,1 OJ.
The diffraction synthesis technique described by Duan and Rahmat-Samii [11] in terms
of the Jacobi polynomial surface description is relatively simple and produces smooth
reflector surfaces with continuous first and second derivatives. In the results that
follow in this chapter the main retlector of the dual offset ret1ector antenna is
synthesized using this method. It is also used to obtain the initial boundary conditions
for the subre±lector actuators as will be described later, but the surface description for
the sub reflector is ultimately made in terms of a mechanical finite element matrix. This
will be described in Chapter 5.

Paragraph 2.1 to 2.3 describe the geometry of the dual ret1ector antenna, show how
the subreflector analysis and specular points are calculated and describe three
techniques to calculate the far-field radiation pattern. Paragraph 2.4 describes a surface
series expansion description in terms of the Modified Jacobi polynomials used in [11 J
and paragraph 2.5 shows veritication of the accuracy of the p-series method and the
Gaussian beam method by comparison to an existing fhll PO-PO solver developed at
The Ohio State University. Paragraph 2.5 to 2.6 shows how the footprint is calculated
and displayed graphically on a geocentric projection. This chapter leads up to Chapter
3 where the calculation of the cost function is described and a sample synthesis is
performed.

Department of Electrical and Electronic Engineering l2

University of Pretoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
antennas
2.1. Geometry of the dual reflector antenna and coordinate
description

Figure 2 shows an unshaped Cassegrain dual reflector antenna configuration where the
main reflector is described by

f(x' , y') = -F (2 .1)

where Yc the paraboloid offset and F the focal distance of the paraboloid.
The subreflector is described in the x., Ys, z. coordinate system by

zs = (2.2)

,;vhere e is the eccentricity of the surface and c the half interfocal distance of the
hyperboloid . The subretlector coordinate system is translated and rotated from the
main coordinate system and is given by

ms1
n" x:-S, J -SyJ
lX' i li~"S2
YS = ms2 n s2 T
,.
Y
I z' C'
(2 .3)
Zs J t s3 ms3 ft s3 L -0 z

with Sx ,Sy ,Sz the subret1ector offset and l,m,n the direction cosines of the rotated axes
"'lith respect to the x', y', z' coordinate system.
The feed coordinate system is rotated and translated with respect to the subretlector
coordinate system and is given by

(2.4)

with Xfs, Yfs, Zrs the feed offset and l,m,n the direction cosines of the rotated axes with
respect to the X s, Ys, Zs coordinate system. The direction cosines for the feed- and
subrefiector coordinate systems in Figure 2 are determined from spherical
trigonometry.

Department of Electrical and Electronic Engineering 13

University or Pretoria
 >L . . . . ." " " and pattern computation for reflector
antennas
2.2.1. Calculation of specular points

implementation on the sub reflector consists of an evaluation of the reflected

diffracted tIeld IS calculated
specular subreflector known teed and known point
on the main evaluated, In
subreflector coordinate as shown in the specular point
on the the point on the main
retlector to calculated. direction cosines for
rp1-lp{"tp,; rays are by

Ma = ' and ,12 '

"'It)

N 0 ==

where
f
+ (Y is + ~z 8)

and

r1
Rl
(2,9-2,11)
1\1 1
RI

where

(2,12)
The normal to the hr~'+'",,{'+r'r surface is taken in the direction of the illumination and
is by

n
I 0/ +­ z] (2.13)
Lox ,<

16
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
antennas
with f as given by (2.1). Let the direction cosine of the nonnal of the subretlector
~""+"ce hO ""ive~
LJ'\..I 0
",,""I l U J1 b"J
0/
L / ini
oXs
oj ' and (2.14-2.16)
M
8ys
/ ini
of
N / ini
oZs

The following three nonlinear equations governing the reflection on the subret1ector is
derived from Snell's lav'\/

-- Z s,{t·, IIFI
'1'1
­ {r
\ '-' nm cos nliF
'I'
+ D nm (2.17)
n In

where
r.:z:-::-:z
y X.l -I- Ys
t. -- AS , and (2.18)

(2.19)

with AS as shown in Figure 10.

In addition, one finds that
Ll - Lo - 2BL, and (2.20)

(2.21)

where
~ n "f
+ lVllV11 (222)

AS is the radius of the projected expanSlOn of the surface as described above.

FoIlov'ving the procedure outlined in [12] Equations (2.17), (2.20) and (2.21) are
solved using a Nevvton tlrst order method.
The l'~evvton met,hod is implemented as follovls: Using an initial guess to a solution of
(Xu, Yo, Zo) the three nonlinear equations gives the following first order approximation
II
I~
d'i
T
,pI - y l lH'l
C~2 $2 a
01", IXY -
"0
I :::::: I ~; (223)
ux iY (k-
Yo - - 1'2
rifi'. iF3 P, IZ I
1.:::2..
L A- 0/ &J_ Zo J IF
_ 3_

This set of equations is solved to get a new approximation for (x, y, z) using LU
decomposition. The ne'vv derivative matrix and solution is calculated and the ne\v

Department of Electrical and Electronic Engineering 17

University of Pretoria
 rhaptar
l.'-"
'-.,..r'
2·.. nl·ff:ra"t;,.,...,
.1.....#' J 1 vLlVl1"::' ~yntha"l·
l. <"
"',,::,..> and radl·atl·on pattarn "on-o"utat·lo'" +'or re+1ect ,.,.r
1 L "'" '" lilt' 11.1 1 1.11 LVI

anterulas
solutions is used as the next approximation. This iteration is repeated until the

difference betvveen the nevv solution and the solution for the previous iteration falls
into the prescribed tolerance. For an initial guess the specular point is calculated for the

sub reflector the specular point is found within less than 10 iterations within a tolerance
of 10-6 as long a$ there are no inflection areas (areas \\lhere the second derivati've
changes sign) on the reflector surfaces. An advantage of using the surface series
ovp""'s;"""" I·'" tern-os of th'" morl;+'I",rI T"COhl·
\.1.1\.. U.l1 lUll 11 L 111.1 lJ .... ,.,.l y..,,.,.m;a1.I":'" is the surface produce no
.1 ' -' u \ ..U ..1 \.Iy Ju. t'Ul 11U1 11

inflection areas which result in multiple specular points.

V/ith the specular point kno\vn th·e radii of curvature of the sub reflector and the first
and second derivatives of the surface are calculated following [14]. The local
properties of the surface are deterrrtined as follo\vs:
A surface can be represented by the vector
felt, v) = (x(u, v), .y(u, v), z(u, v)) (2.24)

where (u,v) are the curvilinear coordinates of a point on the surface or alternatively

F = rex, y) = (x, y, f(x, y)). (2.25)

with tangent vectors

di ( au
ax By <1Z)
fu -a;- 8r! B;- (2.26)
or oz \
rv ~
( ox
ov , !!....
ilv ~) (2.27)

In the surface series expansion the following relationships are used

t (2.28)
v (2.29)

WI·th .t 1 ~
ana1 y; as aennea m
1· ('"
L-.! 8') ana
· ~(2 . 1'"';1)
!

The local properties of the surface are determined by a linear operator called the

curvature matrix Q . On the surface there is a normai 1-1 at each point. The variation

of if is determined by the curvature of the surface at that point. Since .N N

differentiation with respect to u and v will give

(2.30)

Department of Electrical and Electronic Engineering 18

University of Pretoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
antem1as
which means that (N u ' Nv) lies in the tangential plane to the surface at the that point.
This allows (N u' ii,,) to be expressed in terms of (Fu, f.,,) as
(2.31 )

I .....
~L..J
")2)

and the four parameters in (2.70) and (2.71) form a curvature matrix

(2.33)

Equations 2.70 and 2.71 may be rewritten as

- [ilu N" r = Q[pu I~, r (2.34)

[f" f~! ] is a 3 x 2 matrix given by

(2.35)

rilu
A

Similarly L
1'1". ,I is given by
aI! Al
[Nu N v ]
l&
u,~
air 'A
<7: . y
. x tv
air
Oy .Y .:1 12""
~
r \
.JO)
a.: cr.,
a. . Z J
A

(l; Z

A fimdamental matrix is introduced such that

-T

1
_

-
ILF FlG.J - iFJv.PuF
E _ u' u
(2.37)

Al10ther fimdamental matrix is

TT
11
(2.38)

which yields

JJ = Ql
The curvature matrix therefore given by

Q = II 1-' (2.40)

Departmeht of Electrical and Electronic Engineering 19

University of Pretoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
ELf1te!InaS
with

QII =
eO jF
EO -F2 (2.41a)
jE - eF
Q1 2 -- (2.41b)
EC _ F2

jG - gF
Q21 = EC - F2 (2.41c)

gE -./F
""' = (2.41 d)
ld22 EC - F 2

Once Q if found at a point on the surface the principal curvature and directions are

found from the eigenvalues and eigenvectors of Q . The mean and Gaussian curvatures

are determined from

(2.42)

(2.43)

where KI and K2 are the two eigenvalues of Q . The matrix Q can be diagonalized
following the following procedure

Let the two eigenvectors of Q be denoted by

(2.44a)

-
d~ = fdId Z l (2.44b)
L . 2L -.l

which satisfy

Odn
~
= K nd n n = 1,2 (2.45)

The solutions of (2.45) are given by

(2.46a)

(2.46b)

A matrix D is formed such that

(2.47)

Then the matrix

Department of Electrical and Electronic Engineering 20

University of Pretoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
ante!1_flaS

(2.48)

is the diagona!ized curvature matrix. The first and second rows of the 2 x 3 matrix

..
1T
r" I (2.49)

give the principal directions. After normalization the unit principal directions in (2.49)
are given by
e1 = ...L
)'1 (T" + a rvJ\ (2.S0a)

e2 ...L
)'2
({3 r !I
T
, - \
1'" ) (2.S0b)

where
d12 Q22 -K2 Q! 2
a dn. ~ QI- K2 (2.S1a)
d21 QI- K2 Q 21
f3 = dn. Q I2 Gz -K2 (2.S1b)

YI =
{\E -}- 2aF -}- a 2 :-;')
t~_
I 12
(2.S1c)
2 \I I ?
Y2 = ( .fJ E -}- 2fJF -}- G) .- (2.51 d)

The four vectors ~! ' r", el , e2 lie in the tangent plane to the point on the surface and

once Q is determined it is a straightforward matter to determine the principal

dirt:ctions and curvature fi'om (2.51). A simple test to verify the accuracy of the code
can be made for circularly symmetric surfaces by ensuring that radius of curvature
(inverse of principal direction curvrature) for the dependent variable in the normal
direction touches the symmetry axis .

2.2.2. Edge diffraction

The diffracted field is calculated following [9] usmg the diffraction terms tor a
perfectly conducting 'vvedge such as sho\vn in Figure 4 a.nd is given by

Department of Electrical and Electronic Engineering 21

University of Pretoria
, 57bo"S)C
b\5 '4'2...61 ~S
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
antennas

D,(rp, rp' , 4, n) =

(2 .52)

with

_,-;-S--,(_P-=.:_+_,"'-r·),-p-,,--~p--,-i_""' . 2 f3 (2.53)
Ti
'-' = ; ( i \/ ; \ SIn o·
P ~\ Pi + S)'J' i + S)

Note that p; the principal radii of curvature for the incident wavefront and p~ is the
2

incident radius of curvature in the plane containing the incident ray and the edge and

LI"
s(p : + S
\
)P 2 PI
r r
~ ;" 2 (2.54)
r ( r \( y \ ., 111
flo
,oe ,Pi + S ),,0 2 + S)

p; are the principal radii of curvature tor the reflected wavefront and p:is the
2

retlected radius of curvature in the plane containing the reflected ray and the edge such
that

-1..._~ 2(.01. neX.? .n)

p; d Qr,: sin
2
Po (2.55)

where
fi = normal to the surface at the diffraction point
n~ = normal to the edge curvature

s' = incident ray direction, and

a =e radius of curvature of the edge.

Finally the remainder of the terms are given by

F(X) (256)

a ± (13) = 2 cos 2 2nnN; - fJ ,and (2 .5 7)

2... rn)
/2 . .)<;
p ~

Department of Electrical and Electronic Engineering 22

University of Pretoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
ante!1J1as

source

Figure 4. Perfectly conducting wedge for calculation of diffraction from sub reflector
edge.

Department of Electrical and Electronic Engineering 23

University of Pretoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
ante!l!1aS
For this particular problem only the diffracted tield associated with the reflected

subtended by the main reflector was sufficiently far out of the incident shadow
boundary regiorl. i\. rim search procedure is used en the subreflector to find the tV~l O

extremes in the path length (points where ~o = n/2) and points along the rim in these
t\VO areas are used to find the required diffracted ray path to calculate the diffracted
tield .

2.3. Calculation of the far-field radiation pattern

As will be demonstrated in Chapter 3 a very efficient method to calculate the far-field

radiation pattern of the main reflector is needed since -it needs to be calculated so man)T
times at many observation points during optimization. This paragraph discusses three
methods that \vere e'valuated for the purpose namely the p-series method, the Jacobi-

Bessel method and the Gaussian beam method.

2.3.1. The p-series method

The far-field is calculated following [8] and [15]. In the coordinate systems of Figure
2, the main reflector surface is described by

z' = f (x' , y' ) (2.59)

and its unit normal is given by

A

n = NI INI (2.60)

where

N- = [Of
- ~. '
x- ~ I ~ Il .
-of y+z (2.61)
ox oy ' __
The induced PO current on the reflector is given by
T
J
-
-
2n
~
'
..
/\
LT {;:'
11 S , I
\
} . (2.62)

Department of Electrical and Electronic Engineering 24

University of Pretoria
 Chapter 2: Diffraction and radiation computation reflector
antennas
incident magnetic Held is found by GTD on teed/subretlector
described
main reflector is by

(2.

v x 7 and (2.64)
1
= -- V x (2.65)
jl1Jf.l

the approximation 1'1 ~ R . i' into (2.63) allows

the calculation of the required field quantities.

e jkR , ~
- - TO (2.67)
4JlR \ II +

T If (2.68)

The integration IS on with the help of the

<::r\2

OJ

+ 1
~v,) (2.69)

so (2.68) is

H]Cf)e 1kli
' f" J sd.,)' (2.70)

70) can be simplified by introducing following definitions

'\
) - '\
).j s
T ...,I ')
1

and
z' ""Ar'­
,",V" e "on,'
Y.Y (2.72)

where
?!
v, = 9ccs¢

Department Engineering 25
University
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
antennas
v = sin e
sin ¢ (2.74)

When (2.72) is substituted into (2 .71) one obtains

T( u, v) = If J (x' , y' )e jkZ' dx' dy'

cos () ejk(ux'+VY') (2.75)
s

Equation (2 .75) is rewritten as

T( u, v) = Sf J(x' , y' )e1lc' [ e- jkz' (1- wdJ) ]e Jk (UX'+'1") dr' dy' (2.76)
s
so that a Taylor series expansion can be made for small values of 8
P -? '" 1
T(u, v) = L -, [-
p=o p .
jk(l - cos eVTp]
­
(2 .77)

where

If z'P J(x; , y' )eJ~""'e;1c( ux'''Y')dx' dy' (2.78)

s

Equation (2.77) is now expressed as a sum of Fourier transforms. The hjgher order
terms only become significant for wide-angle observations. This form of the equation
\-vas implemented using an FFT a!gorithln and resulted in 'very fast evaluation of the
radiation pattern with reasonable accuracy for the main beam and first few sidelobes.

2.3.2. The Jacobi-Bessel series expansion method

For a well focussed reflector system Equation (2.77) can be rearranged in a form for
reflectors vvith elliptical apertures (circular apertures are a special case of these) that
allows the Jacobi-Bessel series to be used the evaluate the Fourier transforms very

Tp := ff J(x' , y' )ejkr'COS B'cos () {e Jkp'sin B cos(¢' ¢) }p' dp' d¢' (2.79)
s

The factor in brackets in Equation (2.79) is the polar form of the Fourier kernel. This
l(ernel is recast in the fcllc\>ving form to have its centerccincide vvith the pencil beam
direction (8 B , <PB ):

Department of Electrical and Electronic Engineering 26

University of Pretoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
antennas

eikp sin e CO'ii,.¢'-¢) elkp (C u cos ¢'-tCy sin ¢')

e- jkp (Cu cos if -tC y sirt ¢' ) (2.80)

where eu and Cv are constants yet to be determined.

The following f ..mctionals are obtained by combining the first two tenus on the right­
hand side of (2. 80), such that
B cos <:t> == sin () cos ¢ + Cu ' and (2.81)

B sin Q = sin G sin ¢ + Cv ' (2.82)

B and ell can be solved and the shifted kernel is obtained as

(2.83)
Again using a Taylor expansion, the radiation pattern can be written as

DI 2
f f (J(p , rj)
27!

)e-1kP(C" coo ¢'+C, sin ¢') e JkLo rJ(p , rj} ) - f(P w , rj} w )]P (2 .84)
o 0

where Lo ,Lw ,Lwo f •.mctionals given by

Lo ==
T( I
Lp, if/ ; BJ = f(p' , tp ) cos /1 '\ 1"\
17(} == z' cos B (}

Lw = L(p'w , ¢'w , eJ = f(p'w , d/

, w ) cos eo == z' w cos e (2.85 a b c)
"
Lwo == L(pw , rpw
!
" ) cos Go
, G) == JcP'w , rpw =
z' w cos Go

The Fourier series basis functions can then be integrated against the Fourier transform

k.erne! such that one obtains

2 7! (cos ndll {cos nCl>']
2~/. lJ (~)
fo 1J.\. 'le R
"'In nrf."r
V'" >If')
COS( ¢'- o:t»d¢' =
<"1n •
l"'" nw )
..d>r 11
(2.86)

where In is the nth. order Bessel function.

The rnodified Jacobi polynomials are defined as

(2.87)

where a is a real number and P is the Jacobi polynomial. This can be calculated from
using the recurrence relationship. Figure 4 shov·/s the modified Jacobi polynomials for
m = 0, 1 and 2.

Department of Electrical and Electronic Engineering 27

University of Pretoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
antennas
The first term under the integral in Equation (2.84) before the square brackets can be
expanded as
N -HI'M~oo

- p2~
Q (D . . , ,,{/d.) l: l: [p C
n=O 1/1=0
nm cos n¢' +pDnm sin n¢' ]1';: (S' ) (2.88)

where

p = if s' (2.89)

and pC nm and pDnm are constant vector coefficients constructed by using the
h
crtuogonal'1ty 0...f tHe
h expanSIon
. h .
A.~nctlons.

These constants are obtained from

{ C'}
n_.nm
p = -,.~'~ f J Q-
2" 1
(q.
p .­
s' , 'I'
do') sIn
{cos n¢'1
. n.!.' JFn(S'
m
)s' ddo'
. 'I'
ds (2.90)
l p - nm) 0 0 \. If'

where Gn takes on the value of 1 for n=O and 2 for n:;t:O.

The expression for the radiation integral is finally gi'ven by

T(e, ¢)

N 4 00 j\-J -}> 00

I I r [/'nm
n=() m=()
cos n<D+ pD"m Sill n<D] (2.91)

-J2(n + 2m + 1)

where a=D/2.

Department of Electrical and Electronic Engineering 28

Upjversity of Pretoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector .
antennas

Modified Jacobi polynomials Fm(a.~)(t)

m=O; p=O ; a=O,1 ,2,3
6
! I i
5 1 T I 1

4
: 1 ! j I j 1
3 ~~.
1 I ~ I !
1 ;
I
2
I I i I 1 i //
1"-. -.......;: 0........:...
1
'-'
u.. 0

, '" , i ~i/ i Y\
-1 i

, !I "" 'K : /;~ i

l ........... _~_<

\
-2 I ! , ~\
, i I I I I

"
i i
-3
,, i , !
-4 i ! 1
;
; , ,
:
, i : ;
-5
I. , : i ,
-6 I
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Figure 5 (a) Modified Jacobi polynomial for m=O.

Modified Jacobi polynomials Fm(a,~)(t)

m=1; p=O; a=O,1 ,2,3
6
5
4
3
2
1
u.. 0
-1
-2
-3
-4
-5
-6
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Figure 5 (b) Modified Jacobi polynomials for m= 1

29
Department of Electrical and Electronic Engineering
University of Pretoria
 Chapter 2: Diffi"action synthesis and radiation pattern computation for reflector
antennas

Modified Jacobi polynomials Fm(a,~)(t)

m=2; ~=O; a.--Q,1 ,2,3
6

4
3
2
1 ""

u.. 0
-1
-2
-3

-4

-5

-6

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
t

Figure 5 (c). Modified Jacobi polynomial for m=2 .

Department of Electrical and Electronic Engineering 30

University of Pretoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
antennas
exploiting the efficiency of the FFT was eventually used [8]. This involves using (2.77)
and sampling the surface current density distribution to calculate the far-field at the

required vectors. The sampling interval determines the u-v space sampling interval as
sho\vn in Figure 9 'vvhere

(2.92)

The sampling intervals have to be chosen keeping the k-space band limit in mind . To
get the u-\! space data at any other interval a resampling technique described by

Papoulis [16] is used .

\
\
Is{
I

\ \ Visible

'~
Invisible

Figure 9. Diagram to determine sampling interval effect on the visible k-space

spectrum.

Department of Electrical and Electronic Engineering 35

University of Pretoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
antennas
2.3.3. The Gaussian beam technique

A closed form asymptotic solution was developed at the ElectroScience Laboratory,

OSU by Chou and Pathak [4] for predicting the reflection and diffraction of an
arbitrary Gaussian beam (GB) by a general finite curved surface. This solution can be
seen as analogous to the ray solution based on the urjform asymptotic theory of
diffraction (UAT) but the uniform solution for the GB's remain valid in the regions of

and memory requirements and inefficient for vep} large antennas because the
integratioli has to be pertormed numerically. The GB method a\loids this integration
because the GB reflection and diffraction solution are acquired in closed form and offer
a significant ad'vantage in terms of the synthesis time required for very large antennas.
Analyses were made on synthesized surfaces and a version of the synthesis code using

The general astigmatic GB is a projection in real space of a ray field with the source
located in complex space. In real space the GB is a field vJhose amplitude tapers a\.vay
exponentially with a Gaussian taper in the transverse direction to the propagation
direction. Numerical results ha'Je sho\tvn that a GB is produced in the paraxial region
upon reflection from an infinite surface of an incident GB and in particular the reflected
field of a G-B incident en a p<lrabolic surface behaves as a GB \vithin the paraxial
region. The reflection of a GB from an infinite three dimensional surface will be
discussed briefly. i\ method to de an expansion of a feed radiation pattern in terms of
GB's will be shown and some results demonstrating the use of the GB method to
synthesize an offset parabolic reflector ~vill be sho\vn.
A rotationally symmetric scalar GB is used for the basis function and can be written as

(293)

where the coordinate system of the beam aligns with the beam axis in the Zg direction.

»b) Equation (2.93) can be written in spherical coordinates as

Department of Electrical and Electronic Engineering 36

University of Pretoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
antennas
The field expansion can be made in terms of the electric or magnetic tlelds. The
magnetic far-field radiated by the feed antenna is denoted by

(2.96)

The magnetic field of a feed pattern can be written in terms of a set of rotated GB' s as

(2.97)

where em the expansion coefficient oftne m1h GB basis function H~ denoted by

_ ;k f z!' + .L (xi" zt)2 .. (Jim )2 )

I 2
+
jb
jb
= e m zt e L (2 .98)
+ jb

with the polarization vector emchosen to be

(2 .99)

In this form the GB is normalized to the feed pattern in the beam axis direction. Figure
10 shows a representation of the feed pattern expanded in a set of GB ' s in using the
beam\valst values as indicated. Figure 11 sho\vs the real coordinate system used fer a
few launched beams and the real part of the saddle point for a 10° offset angle from
bores:ght going around in 36° increments.

The ret1ection of a three-dimensional Gaussian beam from a slowly varying surface will
be sUIT.u1'J1arlzed. This also follows [4] \vhere a full treatment of GB reflection from a
double curved surface contai!ling an edge is made using an asymptotic technique on the
PO radiation integral oftne scattered GB.

Department of Electrical and Electronic Engineering 38

University ofP etor'a
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
antennas

H i (0) [~i ] [
Xi l
y , --'

where H (0) is the incident magnetic field at the origin O.

The curvature matrix Qi is complex for a GB and is given by

0'(0) = I Q;1
~ IQ'
L 21
(2.102)

and the GB propagation rule is given by

(2.103)

where 1 is the unit matrix.

The transformation relating parabolic surtace coordinates to incident GB
coordinates is given by

r~: 1
L~i-'
=
rc"
Ic~:
Ir
L-' .~T
c ::
C zz

C:n ::J]
Depending on the terms in the curvature matrix the GB can be a rotationally symmetric
(2.104)

- ,elliptical - or a general astigmatic ·GB . The incident GB strongly il1urr.inates the

reflecting surface in the spot area around O. The PO radiation integral for the electric
and magnetic scattered fields from the surface are respectively gi'\len by
r ~

E ' (r) - J:~. JJ lR x R x J eq (r' )] e -;'" dS (2.105)

s

H ' (r) -
jk
4;;- If rR L
x ] eq (;;-' \] ~
I ) R dS (2.106)
S

where S denotes the part of the surface which is directly illuminated by the GB and Zo
is the impedance of free space. R is given by the usual paraxial approximation namely

R = P - P' ; R = INI (2.107)

with
P = xX + yjl + zz and r' = x' x + y' y + z' Z (2.108)

where r is the vector to the observation point and r' the vector to any point on the
surface.

Department of Electrical and Electronic Engineering 40

University of Pretoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
antennas
The PO base equivalent current is given by the expression

Jeq (f') = 2n x l-r Cf' ) (L,. . . ,l OA'

7)

where the unit vector normal to the surface is given by

(2110)

The integration is difficult to perform over a surface coordinate system and the surface
Jacobian is used to obtain the integration o'ver a projectioD. on the z=o plane. The
surface Jacobian is given by

dS' = Jl + (;;f + (;~ rdx,. dy' (2.111)

and the differential current element can be written as

-
J eq (r' )dS' = 2lzr + x
,
~, + y JXf< 2
y' l -.
H I (r' )dx' dy' (2.112)

where it follows from Equation (2.109) that the incident magnetic field in the
neighborhood of the reflectio!1 point can be approximated by
HI (r') ::::
-
H mi (r' )e jkq( r ') (2.113)

with

H~ (r' ) (2.114)

and

q(r' ) (2.115)

The scattered electric and magnetic field can now be written as

_ j~~o H2[R
s
x R x (£ + x ~: + y ~) x H ~ (1" )]

and

H (r-)
S -
=
L4" J rf fRA "v (Az +
• X
A L
R, + Y LRz )
A

X
H mi \r
r-' )] e- JkiR - Q( T' ) ]
R

dS . (2.117)
S

Equation (2 .116) and (2.117) have the same general form as

P S(r) = Jf F(x' , y ' ) e 1J.i(x' , y') dx' dy' (2.118)

s

where f(x ' ,y' ) denotes the phase term of the integrand and is defined by
fe x ' , y ' ) = -H, + q(x' , y' ) (2.119)

Department of Electrical and Electronic Engineering 41

University of Pretoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
-anre!LTlas
PS(r) represents either the scattered electric or the scattered magnetic field .
In this analysis the observer can be in the Fresnel region of the reflecting surface In the
Fresnel region R can be expanded in terms of a Taylor series about the GB reflection
point 0 retaining up to the quadratic terms in x' and y' . Using the approximation the
phase term f{;·~' ,y') can be expressed as

(2.120)

where the coefficients Qo, Ql , bo, b1 and c are defined as

1
2" \~ -; -
'1 (z

bo = (~ - C 3 1)

(2.122)

a1 = t {~2 (7 -
b1 -- (~
(2.124)

(2.125)

Note that f(x',y') can be written as

1'1 •• 1 . .. \ _ .r {X •. \ C' { ... I X \2
- u 1(y'_yJ2 - 2c(x'-xJ(y'_yJ (2 .1 26)
j I....\. , Y } - j I. s' Y s } - 10 I."\' - s}

where
a,bo - b,e
2
a( aoQ! - c )

(2.127)

(2.128)

l(x s , yJ = -;' (2.129)

It is shown in [4] that (xs ,ys ) constitutes the complex stationary phase point in (x' ,y')
and f (x~ ,Ys ) is the value of the complex exponeiitial term iii the integrand of (2.118)
when it is evaluated at this stationary point. For large k an asymptotic approximation

Department of Electrical and Electronic Engineering 42

University of Pretoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
antennas
for (2.118) is derived in [4] leading to a closed form result for the double integral. In

by

r(sJ + ~ (f) (2 .130)

where
- 21r e Jkj, (x' ,[ )
"if" (x' , [ ) (2 .131)

1 ~') a )
rro / ry \
-
1 _ ­ 21 rl1 _ erj'(Sa)\ ] + 2J;s
e- ?;
so (2 .132)

H (X 'I 1 ---ikj~ (x. )

.. x\- e J )'k/' x (xe ) t;""

(2.133)

Pr (n is the contributioJ1 from the sadd1e point which is the asymptotic contribution

from the integral without the edge so that Pr (n is referred to as the retlected field of

the GB. The other term Pd (n is the contribution from the edge and is determined by

the complex distance of the edge point from the saddle point so that Pa (n is referred

to as the edge diffracted field. T is lhe lransitiull function and depends on the cOIH[Jlex
phase difference of between the saddle point and the edge point.

Department of Electrical and Electronic Engineering 43

UI1ive s'ty efP retoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
anten..llas
2.4. Surface expansion in terms of the Modified Jacobi polynomials

The sutface of the main reflector is given by

z(t, If/) = LL in
(C nm cos n Ij/ + D l1m sin n If/)Fm (t)
n
(2.! 34)

where Cnm and Drun are the expansion coefficients and Fmn(t) are the modified Jacobi
'~ollY11
1-' "- o m
1 It; als . 'T'h ~"'''
1 lC:'C polynom'Ia
11
1s are
1 1-el~ted
a to "he
t i L'7
. erru'l r ~
1.1\.0;:; c l11,.,i
: ~~ le l'
~oly_"_: ,, lS
1 I1UIIII(l.I .LII<lL
. 1..,,+ ,, -- ­
<lIe;

often used in the study of optical aberrations [11]. Combinations of the modified
Jacobi polynomials and the Fourier harmonics fcrm a complete set of orthogonal basis
functions in the unit circle. Figure 12 shows a projection of the unit area used tor the
surface series expansicn for the subreflectcr.
The main ret1ector surface expansion coefficients and those of the sub reflector surface
are adjusted during synthesis. The synthesis process produces a continuous surface
with continuous first and second derivatives. The first two terms of the infinite set of
Fourier transforms \vill be used and ha'/e been shovvn to be sufficient \vhen the
synthesized pattern for the CONUS beam case was compared using an accurate PO
reflector analysis soft\\;'are package developed at OS-U [7]. This method has the
disadvantage that the tar-field spectrum is calculated in intervals determined by the
sampling interval of the surface current der~ ~ity . To find the gain at a point not falling
on the interval resampling of the u-v space must be perlOrmed. The far-Held template
for different geographical regions of coverage can then be set up and ccmpar d.

Following [11] the main and the sub reflector surfaces can be described by a senes

superquadric function:

f_\rzb J)21~
U U

[( .E.\2'.I +
a! 1
-
­
1 (2.135)

where a and b are the semi-major axes of the projected apenure.

Department of Electrical and Electronic Engineering 44

University of Pretor' a
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
anterl.f.1aS
These aperture functions can be used to described apertures with circular, elliptical and
rounded corner boundaries using the parametric representation
Xl (t, 1fI) at cos If! r(lf!) , and (2 .136)
/1 \ at sin If r( y/) (2 .137)
Y 1
~, !f/)

where r(Ij)) is given by

(2 .138)

The superquadric boundary is exactly represented by the parametric curve t=1. This is
important \.vhen diffraction frem the edge of the reflector surfaces are calculated using
UTD.

Department of Electrical and Electronic Engineering 46

Univers·ty 0fPr-etoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
antennas
2.5. Verification of the accuracy of the developed codes
2.5.1. The p-series method

Several tests were performed to establish the accuracy of the reflector synthesis code.

the main and the subretlector was developed by Lee and Ruddick [7]. This code was

synthesized surfaces and feed illumination functions were used in the OSU NECREF
\'ersion 3.0 cede. Comparisons \vere also made \vith the Gaussian beam techIlique for
detocussed reflector antennas. These radiation patterns were overlayed to determine
the ,,,; rl'" ,.,ngla ""·erf.'r-rm"nf"'e
1. \I" J.uv P
u..l.l V .au.
.I
'"'t' tha c"d o Ev,-.al1en t a"'rO"'m°.,t
J..V.& J. J.V V 6L..lJ.V V V. hv,",.l .It. V,",.l.l V .d . ;~
.h.' "htal"n"'d
VVl. J. v nu p
.., to ",,,ro
J.J..lVJ. \.I

than 30° oft' boresight for a sampling interval of 0.9 wavelengths even though only two

Figure 13(a) and (b) show the co-polarized and cross polarized radiation pattern for a
CONTJS beam synthesized \vith the developed cede. The same synthesized surfaces
were analyzed using the OSU NECREF Version 3.0 code and the co-polarized and
'-'ro"~
v ...,.;> ""'"'l'ar;z"'d
fJV J.......
.I. rarl;at;o"
u.....1 t-'U"" rn a ra
,..."tte
1.1 J..l.l .I V Sh""'D
.IV \"\1 J. ;n J;'\" O""re
1.l ..L b'-4-.l 1..I. L1r l'l"
\, U} 'lnrl
U.l.l.U "V 11o r
Ih) .c: 1 f"'o'""'"'arl'
v "H.p "'"'D
.:.>V ..1...

The code shows excellent agreement in the main beam area for both the co-polarized
field In the cress-polarized field there is an expected difierence but the field shape and
maximum cross-polarized signal level in the main beam region shows good agreement.
These results establish confidence in the accuracy of the far-field prediction of the
diffraction synthesis software.

Department of Electrical and Electronic Engineering 47

TJnive s'ty ofPr-etoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
antennas
2.5.2. The Gaussian beam technique

The synthesis procedure using the GB method is implemented using only the ret1ected

over which the synthesis will be performed. In the latest implementation only single
offs et examp1es are synthesized but the feed-subreflector combination radiation pattern
can be calculated using a GB expansion such that one can do main reflector synthesis.

although this has not been done as part of the work reported on in this dissertation.
Figure 15 (a) shmvs the far~ field calculated using PO and should be compared with the
calculation using the GB technique in Figure 15(b).

Department of Electrical and Electronic Engineering 52

Univ rs·ty e ~ Pretoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
antennas

Figure 15 (a) and (b). Far-field calculated using PO and the GB technique for a front
fed offset paraboiic reflector shaped using the modified Jacobi poiynomiai expansion.

Department of Electrical and Electronic Engineering 53

University of Pretoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
ant enna
2.6. Calculating the antenna footprint

The various radiation footprints of the antennas are calculated by using 4 coordinate

Translation and rotation of the coordinate systems are handled using matrix operations

algorithm is implemented by calculating the points where vectors in the antenna

coordinate system intersect the geocentric surface and using linear interpolation to
draw the required contour levels on each quadrilateral. This is described in [27].
Figure 16 sho\vs h -Ol"V the u ~,v spa-ce test gtid maps onto the geocentric surface for the
case of the CONUS beam synthesis. The values of the illuminating electric tield are
calculated on the corner points of each quadrangle taking into account the range

spread factor. These values are used in a linear interpojating algorithm that draws the
contours at preselected intervals between preselected bounds.

Department of Electrical and Electronic Engineering 54

U .versity of Pretoria
 Chapter 2: Diffraction synthesis and radiation pattern computation for reflector
antennas

Satell Height 36000 Antenna theta 6

Satell i theta 0.0 phi o
phi -100.0

Figure 16. The u-v-space test grid is shown mapped onto geocentric sUlTace. These

test points are used to caicuiate the cost function during synthesis

Department of Electrical and Electronic Engineering 55

U iversity of Preto .a