Tracking systems for satellite communications

G.J. Hawkins
D.J. Edwards
Prof. J.P. McGeehan

Indexing terms: Satellite links and space communication, Radar and radio navigation, Radiocommunication, Antennas

lite link be maintained within defined limits. This usually

Abstract: Satellites in nominally geostationary becomes necessary when the antenna beamwidth is of the
orbits possess diurnal motion which causes the same order of magnitude as the angular motion of the
apparent position of the spacecraft to wander in satellite. The required pointing accuracy is primarily
the sky as seen by an earth station. If uncompen- determined by the beamwidth of the antenna (which is a
sated this causes a variation in the performance of function of antenna diameter and operating frequency)
the communications link. The motion of the satel- and the allowable losses on the communications path.
lite therefore has to be tracked by the earth Other factors also play a part such as propagation condi-
station antenna. During the past twenty years or tions, the mechanical performance of the antenna and the
so a considerable amount of work has gone into local weather, all of which serve to perturb most systems.
the development of tracking schemes for satellite Broadly speaking there have been three areas in which
communication systems. Here a review of the tracking techniques have been developed. The first
various techniques is presented. For each method involves ground station tracking of a subsynchronous
used the principle of operation is described and orbiting satellite. Here the satellite is in view from any
the various salient features extracted. Compari- point on the earth's surface for a limited period of time
sons are made where appropriate and practical and sweeps a path across the observer's sky in a period of
implementations of the systems discussed. The time determined by the orbital parameters of the space-
application of intelligent control algorithms to craft. Without some form of tracking, even for a fairly
tracking is also considered. The areas covered wide beamwidth antenna, the ground station would soon
include, orbit determination, optimal estimation lose contact and the communications link would suffer.
techniques, performance, and practical implemen- The second situation involves ground station tracking of
tations. Recently introduced techniques are con- a geostationary satellite. Now the satellite movement is
sidered and likely future developments projected. limited to a narrow window usually not exceeding k3".
For an antenna with a beamwidth of the same order as
this motion it is possible that no tracking will be required
1 introduction and the antenna can be fixed, perhaps requiring
occasional manual correction to account for satellite
Research in many areas has been promoted as the eco- drift. If, however, an antenna is used with a much nar-
nomics of satellite communications has evolved and rower beamwidth then tracking again becomes necessary.
emphasis is now placed on cost reduction. As satellite The third area involves satellite-to-satellite links. Here
lifetimes have been extended the operators of the system the predominant features of the tracking technique are
have looked at various schemes to reduce their costs. not only accuracy, but weight and reliability. This places
One option open to spacecraft controllers is the slacken- new constraints on the method employed.
ing of the orbital specification requirements placed upon Over the past 25 years, work has progressed so that
the satellite. This policy allows the orbit to develop four classes of tracking system have evolved to meet the
beyond the bounds that have been the norm until recent- various needs of satellite communications. These can be
ly, particularly in the north-south, or inclination plane. described as :
For the satellite operator this offers an economic advan- (a) manual/programme tracking
tage in terms of an extended satellite lifetime as on board (b) monopulse or simultaneous sensing
fuel usage is reduced. For the earth station operator the (c) sequential amplitude sensing
loosening of the orbital specification results in the space- (d) electronic beam squinting
craft wandering in the sky more than has previously been
the case. Consequently earth stations now have to The latter three techniques can be classified under the
employ some pointing control where previously none was general heading of automatic tracking or autotracking.
needed. These schemes represent closed-loop tracking systems and
Tracking has generally been required in situations once satellite acquisition has been established, tracking
where the communications link specification dictates that continues with no operator intervention. All autotracking
the received and transmitted signal levels over the satel- schemes rely on the reception of a continuous beacon
signal transmitted by the satellite. The received beacon
Paper 6208F (E9), first received 5th May 1987, and in revised form 18th
signal is used to derive pointing error information which
February1988 supplies control signals to the antenna drive servo
The authors are with the Department of Electrical and Electronic mechanisms. The antenna is driven so as to attempt to
Engineering, Bristol University, Bristol, United Kingdom minimise the pointing error and hence null the tracking
IEE PROCEEDINGS, Vol. 135, Pt. F , No. 5, OCTOBER 1988 393
error. The sequential amplitude sensing systems and In this system antenna pointing was achieved using prog-
monopulse systems are widely used in satellite communi- ramme tracking based on satellite ephemeris and the air-
cations today. Electronic beam squinting is relatively new craft inertial navigation system (INS). The most
and although still in the early stages of research is begin- prominent causes of pointing error are described as air-
ning to be deployed at several earth station facilities. craft flexing, friction and velocity lags. Although con-
Each of these four classes will now be examined in firmed flexing data was not available for the tail-fin
turn. Emphasis is to be placed upon the principle of the location of the aerial, the authors estimated angular
tracking method and the consequent effect on overall errors were well below the required overall pointing
system performance. For the autotracking schemes, this accuracy of f2".
entails a detailed description of the radiofrequency pro-
cessing necessary to derive the tracking error signals. The
effect of the servo mechanism and control systems, 3 Monopulse systems
employed in the antenna drive, are not considered since
these are relatively independent of the error detection In the examination of sequential amplitude sensing
mechanism. It must be appreciated, however, that these systems which follows later, it will be shown that a finite
elements do have a significant effect on overall system time period is required in order to extract the pointing
performance. error signal. In the time interval in which measurements
are made the beacon signal should ideally contain no
amplitude modulation (AM) components other than the
2 Manual and programme steering
modulation produced by the tracking mechanism. If this
In the absence of an autotracking system, antenna point- is not the case and additional modulation components,
ing is either undertaken manually or using programmed caused, for example, by changes in the atmospheric pro-
steering. Manual tracking involves an operator control- pagation path, are present the tracking accuracy might
ling the movement of the antenna until the received be degraded. In practice the presence of AM interference
signal strength is maximised. Programme steering uses can be sufficiently serious to form the main limit to the
prepared data describing the path of the satellite, seen achievable tracking accuracy of a sequential amplitude
from the earth station as a function of time. This infor- sensing system.
mation is fed to the antenna positioning servo system Amplitude fluctuations of the beacon signal will have
which points the antenna in the appropriate direction. no effect on tracking accuracy if the pointing error infor-
Programme tracking was used extensively at early mation can be obtained from the received beacon signal
satellite tracking installations to track the first generation at a single time instant. Monopulse tracking schemes
of orbiting satellites, such as Telstar. The Goonhilly No. offer just such a capability and consequently promise
1 antenna [l] was one such example, being initially improved operational performance compared to systems
equipped in the mid-1960s with a programme steering which take a finite time to derive the pointing error
system which remained in service until the late 1970s signals. Monopulse schemes employ either spatial ampli-
when it was replaced during the installation of a new tude comparison, phase comparison or amplitude-phase
autotracking facility. At that time the equipment required comparison techniques. Such systems have been used
to process the data tapes, perform calculations and then widely for satellite tracking and are described in a
provide steering commands was complex. Preparation of number of papers [5-211.
the data tapes was time consuming since the high angular The simplest form of monopulse technique is the
velocity of the satellite with respect to the ground station simultaneous lobing system which utilises amplitude-
required that either, the tapes were frequently updated, or comparison. Anders et al. [SI describe such a scheme, the
sophisticated interpolation techniques were used on sub- 'Precision Tracker', as employed on the tracking stations
sequent transits of the satellite. at Andover (Maine, USA) and Pleumeur-Bodou in
Most of these problems were recognised at the time France in 1963. These were used for tracking the early
and radar experience provided the basis for exploring the Telstar satellites.
possibilities of autotracking or at least initially deriving The antenna for the 'Precision Tracker' has a four-
tracking error information. Programme steering horn feed system which creates overlapping antenna pat-
decreased in popularity as these new systems were devel- terns. The beacon energy received by the four horns is
oped but was often employed as a back-up scheme in processed by a comparitor to develop three different
case the primary tracking method failed. antenna pattern response characteristics: the sum pattern
In circumstances where the tracking accuracy required corresponding to a conventional antenna pattern; the
is not too stringent, this form of satellite tracking still has elevation difference pattern, having two main lobes in the
applications. One example of modern day usage was at elevation plane with a deep null on the boresight axis;
the BT Research Laboratories at Martlesham Heath [2]. and the azimuth difference pattern, having two main
Here a 6.1 m Cassegrain aerial was used to track the lobes in a plane perpendicular to the elevation plane with
Orbital Test Satellite (OTS-2) launched by the European a deep null on the boresight axis. These patterns are
Space Agency in the spring of 1978. Tracking was shown in Fig. 1 along with the modes in the horn aper-
achieved using data provided by the satellite control tures and the angle error signal. Mathematically these
centre. With occasional manual control following satellite patterns can be expressed as :
manoeuvres, a pointing accuracy of 0.01" in both
azimuth and elevation was reported to be maintained.
sum = A +B+ C+D (1)
Programme tracking was also used for the reception of azimuth difference = (A + B) - (C+ D)
signals from the NOAA satellites at the University of elevation difference = (A + C)- ( B + D)
Dundee [3]. The wide beamwidth of the 1.8 m antenna (3)
did not require the use of an accurate autotrack tech- Since the difference signals arise from the antiphase addi-
nique. A similar argument applied to the Nimrod aero- tion of signals from the appropriate pair of horns, phase
satcom terminal described by Cummings and Wildey [4]. sensitive detectors must be used in the receiver. With two
394 IEE PROCEEDINGS, Vol. 135, Pt. F, No. 5, OCTOBER I988
 channels being required, for the sum channels and each input signal-to-noise (SNR)? of about 15 dB. The system
of the difference channels, the tracking receiver must worked with both linearly and circularly polarised
therefore be a four channel coherent device. In practice a signals, the former incurring a 3 dB effective loss since
single sum channel is frequently used. such a feed only accepts a single vector component of the
signal.
The essential characteristics of the amplitude sensing
sum azimuth elevation
channel difference difference
system are the generation of a sum pattern with
A W F l C A n n C A I q M C
maximum boresight gain and difference patterns with a
large value of slope at the crossover of the offset beams
B [ q F l D B m F l D BIS]F]D and hence a steep error voltage characteristic around the
z Aaz AeI
boresight axis. The use of the four-horn feed does not
a allow the optimisation of both of these parameters since

&')
independent control of the sum and difference patterns is
unavailable. Less compromise and better performance
may be obtained with five-horn feed in which the central
horn, used for the sum channel, is surrounded by four
other horns whose functions are to provide the azimuth
A a z = ( A + B ) - ( k C D and elevation difference patterns. Feeds of this type are
described by Pratt [6] and a number of other researchers
[7-91.
Z=A+B+C+D The other major form of monopulse system relies
A- B
upon detection of higher order modes excited in a micro-
wave feed system, to determine the discrepancy between
the pointing direction of the antenna and the satellite.
b
Such systems are generally referred to as multimode
monopulse systems.
Y
indicates
One of the first systems of this type is described by
Cook and Lowell [lo]. This scheme was also employed
on one of the tracking antennas at Andover (Maine,
hybrid
junction USA) and the early experiments used much of the same
equipment employed in the amplitude-comparison track-
--@- /e- ing arrangement described by Anders et al. [ibid].
With this system a circular feed was employed. The
waveguide had a large enough diameter to support, at
the beacon frequency, the propagation of the two lowest
-e- -0- order modes, namely the TE,,, or dominant mode, and
C e
the TM,, mode. With such an arrangement the magni-
tude and phase of the TM,, mode component can be
used to determine the pointing error information.
Analysis of the TM,, mode component reveals its magni-
tude is dependent upon the radial pointing error, 0 (see
Fig. 2). When 0 is small the amplitude proves to be

d f

Fig. 1 Simultaneous lobing system

IIModes in horn apertures
b Comparator bridge network
c Overlapping antenna patterns
d Sum pattern
e Difference pattern
f Error signal voltage

Examination of the angular error signal, for either the

elevation or azimuth planes, shows its magnitude and
sign depend upon the angle off boresight for that plane.
Further, the scheme offers two important features; firstly
it is clear from Fig. If the error function is approximately
linear with respect to off-axis angle, and secondly, it can
be shown that the output scale factor for the error signal
is rendered insensitive to changes in absolute received
signal level within the frequency response of the AGC
system and the dynamic range of the receiver.
Results of tracking experiments carried out with 'Pre- Fig. 2 Definition of space co-ordinates for an elliptically polarised
cision Tracker' which employed a 2.5 m Cassegrain signal
reflector, having a 2" beamwidth, showed that a tracking
accuracy of approximately 0.01" was obtainable with an t All quoted SNR ratios measured at input of tracking receiver
IEE PROCEEDINGS, Vol. 135, Pt. F, N o . 5, OCTOBER 1988 395
directly proportional to the pointing error. In addition, struction of the feed arrangement used, however, the
the circumferential direction, 4, of the pointing error can scheme will only cope with the particular beacon fre-
be shown to be proportional to the phase difference quency for which it was designed. This feed arrangement
between the TE,, and TM,, signals. Recovery of the is currently used on the Goonhilly No. 4 antenna.
azimuth and elevation tracking errors is then possible Similar schemes have been used throughout the world
using the relationships : and are still popular some twenty-five years after the first
operational systems were constructed. Hayashi et al. [131
azimuth pointing error = k,8 cos 4 (4) describes such a system employed at the Kashima Branch
elevation pointing error = k, 8 sin 4 (5)
of the Radio Research Laboratories in Japan. Here a
10 m antenna system is used for reception of signals from
where k, and k, are constants. the ETS-I1 satellite. The antenna is notable for its multi-
The signal processing arrangement chosen for the ple frequency operation. Beacon signals of 1.7 GHz,
system, described by Cook and Lowell, relied upon pro- 11.5 GHz and 34.5 GHz can be received and tracked.
cessing three signals to determine 6 and cp. A horizontal Another Japanese tracking station, employing a TM,,
sum signal corresponding only to the horizontally pol- mode feed, was reported in 1981 by Inoue and Kaitsuka
arised component of the incident wave, the vertical sum [14]. This was a K-band facility for use with the Japa-
component corresponding only to the vertically polarised nese domestic communication satellite. The system con-
component and the difference signal representing the sists of an 11.5 m earth station antenna where the
cp-plane component. The sum and difference patterns, half-power beamwidth of the antenna is 0.08" at 20 GHz
resemble in form and function, those of a simultaneous and 0.05" at 30 GHz,. Operation with these high beacon
lobing system and correspond to the TE,, and TM,, frequencies presents special problems, one of these being
mode patterns respectively. The extraction of 8 and 4 induced crosspolarisation caused by rainfall along the
consisted of a stiaightforward comparison of amplitude propagation path. This leads to cross coupling between
and phase between the sum and difference signals. The the antenna drive axes error signals and hence a wors-
original system achieved this using a four-channel coher- ening in tracking performance. The results of experiments
ent receiver in which the individual sum signals were using the stations at Yokosuka and Sendai show that the
multiplied with the difference signal to yield the two acquisition range of the system was 0.08"in all directions.
required output error voltages. For a large SNR, the tracking error depends mainly on
Analysis of the far-field distributions produced by the wind gusts and backlash. For a SNR of 60 dB, the track-
two modes, on a plane normal to the feed boresight, ing error is less than 0.005" for an average wind velocity
show that the field for the TE,, mode is parallel to that of 10 m/s. In strong winds the tracking accuracy deterio-
in the feed while for the TM,, mode a radial distribution rated to 0.01" with average wind velocity of 25 m/s.
is observed. In consequence, it is an essential requirement Operation of the system in light wind conditions results
of this tracking scheme that the satellite beacon signal is in a tracking error of less than 0.003' even with 20 dB
circularly polarised so that the TM,, mode is excited rain attenuation.
with an efficiency that is independent of rp. Antennas in the United Kingdom employing the
Simplification of the processing can be achieved by TM,, monopulse technique include Goonhilly No. 3
receiving only one of the orthogonal components of the (notable for its rectangular feed horn), Goonhilly No. 5
TE,, mode, for example, the horizontal component, and (used for INMARSAT communications), Goonhilly No.
then generating the other with a 4 2 phase shift. Now a 6 (a dual frequency antenna working at 4/6 GHz or
two-channel coherent receiver can be used but the sense 11/14 GHz), and Madley Nos. 1-3 situated in Hereford-
of the circular polarisation must be known. Experimental shire, all of which are operated by British Telecom Inter-
work at the Andover aerial showed the use of either the national.
two- or four-channel system led to similar operational To track satellites transmitting beacon signals with
results. either circular or linear polarisation using the multimode
The closed-loop tracking performance of the Andover monopulse technique, it is necessary to use a feed which
system was measured by Lozier et al. [ll]. The 20.4 m excites either orthogonal TE,, modes or both the TM,,
horn-reflector antenna had a 3 dB beamwidth of 0.225' and TE,, mode. Some of the possible approaches are
at the beacon frequency. The acquisition angle was found described below.
to be 0.2" in all directions. Measurements of the tracking Choung et al. [l5] discusses a tracking feed for use in
errors showed that the autotrack system maintained the the Ku-band that utilises the generation of orthogonal
null axis of the antenna within 0.005" of the actual satel- ,
higher-order TE, modes. With this arrangement when
lite direction. The practical threshold limit for successful the main beam axis is not aligned with the boresight axis,
operation was determined to be about - 130 dBm at the TE,, modes are excited in the feed, the amplitude of the
input to the sampling coupler for the horizontal and ver- higher TE,, mode being proportional to the angle of
tical dominant mode signals. misalignment. Orthogonal TE,, modes are sensed in a
Simplification of the multimode monopulse system has coupler and used to derive the required azimuth and ele-
resulted in some interesting work. In 1975 Chivers and vational error signals. The overall layout of the coupler is
Brain [12] described a tracking system utilising the exci- shown in Fig. 3. Four coupling arms are summed
tation of the TE,, and TM,, modes but this time using through a hybrid network to obtain a single TE,, mode
only one' receiver amplifier for both the beacon and com- coupler. The two mode couplers are then summed
munication signals. This was accomplished by injecting together through a final 90" hybrid to separate the
the microwave pointing error signal, derived from com- azimuth and elevation error signals. Mode rejection
parison of the TE,, and TM,, mode signals, into a between the TE,, and TE,, modes is claimed to be
common receive waveguide channel. This eliminates the approximately 40 dB. The authors do not state if this
need for separate amplifiers in the tracking receiver feed was incorporated in a tracking facility. A similar
system and, according to the authors, produces a more system is reported by Savini [16]. The scheme uses four
stable and reliable system. Due to the mechanical con- instead of eight coupling positions and relies upon rota-
396 IEE PROCEEDINGS, Vol. 135, P t . F, No. 5, OCTOBER 1988
tion of the coupler, by the servo system, to obtain intended for use on-board the ESAs large satellite
orthogonal pointing error measurements from the TE,, (OLYMPUS) to provide a direct TV broadcast beam
mode signal. covering the Italian region. The angular accuracy of the

Fig. 3 Layout of T E , , mode coupler using two orthogonal modes

The use of feed systems utilising the excitation of both complete tracking system was specified as 0.1".A similar
TMol and TE,, modes to derive tracking information arrangement is described by Reinders and Oei [20] for
are widespread. Such systems difference the individual earth station antennas employed to track the Orbital
higher order mode components to recover the required Test Satellite.
orthogonal tracking error signals. One such scheme is Other examples of stations employing monopulse
described by Watson et al. [17]. The feed, based on work techniques include the maritime terminals used with the
previously carried out by Molker and Watson [lS] SKYNET system [21] and many large Intelsat earth sta-
separates out sum and difference (x and y in orthogonal tions, especially those working at 11/14 GHz [22,23].
planes) signals which are multiplexed into a four-channel
tracking receiver to produce normalised error signals to
control satellite pointing. Utilising these inputs, it was 4 Sequential amplitude sensing systems
shown [19] that the tracking system is rendered insensi-
tive to the deleterious effects of atmospheric depolar- The monopulse techniques, although highly successful in
isation. operation, require relatively complex wavefront pro-
The key component in the feed chain is a planar cessing to derive the tracking error information. This
mode-extractor (Fig. 4), which has four symmetrically processing inevitably requires at least a two-channel
positioned transverse coupling apertures between the coherent receiver and often a four-channel system is
main central circular waveguide and individual rectangu- necessary. This presents two major disadvantages; first,
lar waveguides. This mode coupler is interposed between the tracking system is expensive and secondly, owing to
the corrugated horn and a high performance circular its complexity, it is mechanically large. For an earth
waveguide polariser. The size of the central circular station, mechanical size is relatively unimportant but in
waveguide is such that both the TE,, and TMol mode space this presents a serious problem since ultimately the
can propagate. Results of laboratory tests presented in usable satellite communication payload will be reduced if
the paper show cross coupling to be less than 32 dB the mass of the control system is large.
between the difference channels with null depths for the Until recently sequential amplitude sensing offered the
difference patterns exceeding 50 dB. The system was only alternative to the monopulse technique. Two

TEZ1/TM derived tracking signals

O?
t TEll derived sum signal
mode
ext roc t ion
t v down-link
J
, up-link
/ / /
TEZI/TMO1derived tracking signals
+TEI1 derived sum signal
H

'4
TEzI/TMo~ derived tracking signals
tTE11 derived sum signal
H

18.lGHz 1 TEZl/TMOlderived tracking signals

+TEI1 derived tracking signal
V
Fig. 4 Tracking mode couplerfor T E , , , T M , , and T E , , mode extraction

IEE PROCEEDINGS, Vol. 135, Pi. F,No. 5, OCTOBER I988 397

 methods are available; conical scanning and the step- nical complexity and was only used on a limited number
track or hill-climbing technique. Conical scanning of tracking stations. The most notable of these being the
evolved from the conical scanning radar developed after Apollo Ascension station commissioned in September
the Second World War for the detection of airborne 1966 [l].
targets. The step-track technique was introduced in an Both of the previously described techniques result in
attempt to significantly reduce the overall cost of satellite modulation of the up-link transmission as both the trans-
tracking facilities. Both use a single channel receiver and mission and reception paths through the antenna are the
employ time-division multiplexed, spatially distributed same. For the Goonhilly No. 2, Bahrain and Hong Kong
measurement of the beacon signal to derive the tracking 1 antennas, mode conversion techniques were applied to
information. the conical scanning principle in order to overcome this
problem. The scheme was patented in 1968 by the
4. I Conical scanning Marconi Company Ltd. [28]. Lockett and Shinn [29]
The principle of conical scanning is shown diagram- described the operation of such a feed mechanism, shown
matically in Fig. 5 and described by Skolnik [24]. The in cross-section in Fig. 6, as though it were transmitting.
target axis

U
radar
Fig. 5 Conical scan tracking

concept relies on continuously rotating an offset beam This simplifies the explanation of the scanning mecha-
about the boresight axis of the antenna. When the satel- nism and is valid since the principle of reciprocity will
lite is off this axis the received signal in the beacon apply when the antenna is operating in the receive sense.
channel is modulated at a frequency equal to the rotation The feed used on these antennas has a circular cross-
rate of the beam. The amplitude of the modulation will section throughout. At its smallest diameter only the
depend upon the shape of the antenna pattern, the squint TE,, mode can exist in the guide. Next there is a conical
angle and the angle between the satellite line of sight and transition followed by the cylindrical tracking section in
the rotation axis. The phase of the modulation depends which TM,, mode can also exist. An unsymmetric mis-
on the circumferential angle between the satellite and the match is introduced into this section and has the effect of
rotation axis. The conical scan modulation is detected in converting some of the TE,, mode to TM,, mode.
the tracking receiver and applied to the servo-control From the tracking section, both TE,, and TM,,
system which continually positions the antenna on the modes reach the aperture of the feed and are radiated
satellite. When the antenna points directly at the satellite, towards the subreflector. The secondary pattern (due to
the line of sight to the satellite and the rotation axis coin- the TE,, mode) has a maximum on the axis of the
cide and the conical scan modulation is zero. It is clear antenna while that due to the TM,, mode has a zero on
this method constitutes a very similar situation to that the axis. If these two patterns are in phase then their sum
which exists for a monopulse system utilising TM,, has a maximum which is off the axis and hence the main
mode excitation. Again the amplitude and phase of a antenna beam has been offset relative to the beam axis.
derived tracking error signal is being used to determine With the TE,, mode circularly polarised, it can be con-
the azimuth and elevational tracking errors. It therefore sidered to consist of two components of mutually perpen-
follows that these errors can be derived using relation- dicular polarisations, one of which is in phase with the
ships similar to those in eqns. 4 and 5. TM,, component, the other being in quadrature with the
The simplest form of conical scanning can be achieved TM,, component. Thus the beam is offset in the direc-
by rotating the complete antenna about the boresight tion of the polarisation of the first of these components.
axis. It is obvious such a method is only suitable for The direction of the offset, therefore, bears a definite
small aerials but such systems have been used. The later relationship to the orientation of the tracking section and
SCOTT military shipborne satellite terminals are one if this is rotated, the antenna beam is seen to carry out a
such example [25]. Here the autotrack system rotated conical scan.
the whole antenna using servo motors at a frequency of When receiving, the application of reciprocity shows
1 Hz. Earlier terminals had no autotrack capability and that amplitude modulation of the beacon signal will
relied on a stabilised platform and manual positioning occur when the satellite lies off the boresight axis and the
C261. tracking section is rotated. Since with this arrangement
For a large Cassegrain antenna conical scanning can the feed is no longer offset, up-link modulation is
be achieved either by rotating an offset feed, or by using a removed. The technique is only effective when the beacon
fixed feed and rotating an offset sub-reflector. The offset signal from the satellite is circularly polarised. If a lin-
feed was used on the Goonhilly No. 1 antenna and was early polarised signal is sent through the tracking
described by Knox and Doble in 1970 [27]. Use of the arrangement the radiation pattern, instead of carrying
offset sub-reflector was unpopular because of its mecha- out a conical scan, would move with simple harmonic
398 I E E PROCEEDINGS, Vol. 135, Pt. F, No. 5, OCTOBER 1988
motion in the direction of the polarisation. This results Kong but also many American systems utilised this tech-
from the radiation pattern of the TM,, mode having a nique, including the 64 m dish at Goldstone [30]. After
polarisation which is radial in every direction. this time, however, as monopulse and step-track systems

diplexer polariser rotating U 1

joint tracking section horn --+
d=5hg/4 for TEll
Fig. 6 Layout offeed components =3hg/4 for TMOl

This difficulty can be overcome by using TE,, mode were fully developed it became less popular. The conical
generation and is also discussed by Lockett and Shinn scan system was not as accurate as the monopulse tech-
[29]. It is possible with the appropriate choice of the nique because of its sensitivity to AM interference.
tracking section diameter, to transform part of the TE, , Monopulse systems were also mechanically less complex
mode into a TE,, mode while at the same time trans- and hence more reliable. The step-track techniques, as we
forming very little energy into the TM,, mode. Rotation will see shortly, offer a low cost solution to the tracking
of the tracking section provides modulation of the problem. With an achievable accuracy approaching that
received signal which can be used for direction finding in of a conical scan scheme, these became the most viable
the same way as has been described for circular polarisa- solution for the smaller earth station installation.
tion and the TM,, mode. Unfortunately, this cannot be
described in terms of the simple conical scan relation-
ships. Analysis shows the degree of modulation now 4.2 Step -track systems
obeys the relationship: In 1970 the step-track (or hill-climbing) tracking tech-
nique was proposed by Tom [31]. The scheme aimed to
depth of modulation =6 cos (4 - 2a) cos a (6) provide a cheaper alternative to the satellite tracking
problem by considerably simplifying the microwave and
where 6 is the off axis angle of the satellite, 4 is the satel- electronic elements of the tracking equipment.
lite circumferential angle, ct is the coupling slot circum- The step-track technique resulted from the develop-
ferential angle in the tracking section. ment of the Comptrack experiments carried out in 1967
Isolation of the satellite off axis and circumferential [32]. The operation of Comptrack was similar to the
angles is then only possible by processing the second and operation of the manual tracking scheme except that a
third harmonic components of the conical scan modula- small computer was used to perform the functions of the
tion. This requires the addition of specialised matched operator. The experimental results of Comptrack demon-
filters to the tracking receiver and the use of a more strated the conceptual credibility of a peak-seeking tech-
complex feed in order to generate these harmonics. In nique.
consequence such an arrangement is generally tuned to The step-track scheme represents a further simplifica-
operate at one particular spot frequency and so although tion of the Comptrack system. The name ‘step-track’
linearly polarised signals can now be received the feed originates because the antenna is made to step towards
constraints mean only one beacon frequency can be the energy peak of the received signal. After signal acqui-
worked. sition the antenna is commanded to make an initial
Application of either of these mode conversion tech- angular move. By comparison of the received signal level
niques to the conical scan principle results in an improve- before and after the move, the direction of the next move
ment in tracking accuracy of approximately five-fold is determined. That is, if the signal has increased, the
when compared to a system in which the up-link trans- antenna continues to move in the same direction. If the
mission suffers modulation. signal level has decreased, the direction of movement is
From the description of the conical scan mechanism it reversed. This process is made continuous and alternates
is clear that unwanted amplitude modulation of the between two orthogonal axes to give the capability of
beacon signal will cause erroneous tracking error infor- tracking.
mation, leading to a degradation in the achievable track- The simplicity of a step-track system can be seen by
ing accuracy. All the forms of conical scanning described the general block diagram shown in Fig. 7. The system
so far are susceptible to such AM interference especially only requires a signal strength detector with associated
if the frequency components of the fluctuations are at or timing generator, and stepping motors. Operation would
near the conical scan frequency. In general, filtering of involve the following sequence of events:
the beacon signal is required to remove the majority of (a) sample and measure signal strength
this unwanted modulation. Such a technique was (b) move x-axis
employed at the Goonhilly No. 2 tracking station. Infor- (c) sample and measure signal strength
mation relating to false modulation caused by polarisa- (d) move y-axis
tion of the beacon signal and effects of satellite spin were (e) wait
utilised to derive matched filters to eliminate all fre- cf) go back to (a)
quencies except the fundamental at the rate of rotation of
the tracking section. It is clear that the antenna’s boresight axis is contin-
By the early 1970s conical scanning was well estab- uously being stepped, therefore, an average pointing error
lished as a method of satellite tracking. Not only was it exists even when the system is operating under perfect
used on the antennas at Goonhilly, Bahrain and Hong conditions. With noise perturbations erroneous decisions
IEE PROCEEDINGS, Vol. 135, Pt. F, No. 5, OCTOBER 1988 399
will occur which will serve to increase the average point- the Symphonie satellite. The stepping motion was initi-
ing error. Tom analysed the step-track performance by ated by incrementing the antenna axes with a step size of
calculating the effective antenna gain loss caused by the 0.025". Experimental results presented in the paper claim

L NA
- down- - envelope - lowpass - decision
converter detector filter circuit
>antenna

u
beacon
J
Fig. 7 Possible step-tracking subsystem conzguration

stepping motion. This analysis showed optimum param- the achieved rms tracking accuracy in each axis was
eters exist for the post-detection filter bandwidth, the better than 0.025" with a beacon SNR >12 dB. It is
time between samples and the step size for a given satel- further stated that tracking could be maintained at a
lite drift rate. Tom claimed even for the lowest practical reduced accuracy with an input SNR ratio as low as
input signal/noise ratio (SNR = 3 dB), the gain loss from 3 dB. As the Symphonie satellite is a low-drift geostation-
tracking would be moderately small if the satellite drift ary satellite the authors propose that once the step-track
rate is not excessively large. system has maximized both axes, and hunting has begun,
The step-track scheme has a number of limitations. the tracking should then be disengaged. At regular inter-
Firstly it is almost axiomatic that locating a beam vals tracking would then be reinitiated to realign the
maximum can never be as accurate as finding a sharp antenna. This eliminates unwanted AM modulation on
null. Secondly, its susceptibility to AM interference the communication signals and reduces wear on the
means an input SNR some 15 dB higher, than that antenna drives. It does mean however the antenna
required by a monopulse system, is required if two cannot respond to wind gusting and in such circum-
similar systems are to maintain a given pointing accu- stances tracking accuracy would be degraded, depending
racy. A penalty is therefore suffered in terms of either on the stiffness of the antenna structure.
requiring a proportionally higher satellite EIRP or larger An even larger earth station employing the step-track
antenna. Thirdly, the tracking mechanism does not work technique is Goonhilly No. 1. This was converted from
in real time. A dynamic lag always exists between the conical scan to step-track in 1983. Step-tracking is
satellite position and the antenna position and this achieved using a hydraulically driven nodding sub-
degrades the achievable accuracy to around 0.05" if no reflector since movement of the antenna structure would
optimisation techniques are undertaken. Fourthly, the have been difficult due to its physical size. This scheme
transmitted and received communications signals are utilises optimisation techniques, in the form of a control
amplitude modulated (at the stepping rate) due to the algorithm known as the smoothed step track technique,
stepping motion. Although it is technically possible to in order to achieve an accuracy better than k0.01".
cancel this by 'antimodulating', the solution is compli- Step-track schemes have also found extensive use for
cated and its reliability questionable [33]. Since modula- maritime satellite communication tracking systems. The
tion of the uplink is often undesirable, the tracking MARISAT civil communications terminals discussed by
technique is more applicable to receive-only stations. Johnson [36] are one example. The sampling time at
Finally, the system has a slow dynamic response since the each of the sampling angles is made long enough to
tracking operation involves either stepping the main average the system noise. Noise in the marine
antenna or mechanically nodding the antenna sub- environment results from two main sources; the residual
reflector. Stepping the main antenna causes substantial dynamic angular pointing error due to the ship's motion
wear and tear on the steering motors during the execu- after stabilisation, and the apparent low-frequency modu-
tion of the search patterns. Consequently on larger lation of the received carrier due to multipath fading. For
antennas a nodding subreflector may be preferred. an antenna beamwidth of lo", a step amplitude of
Despite these disadvantages the step-track technique approximately 0.5" is required with an integration time of
has found wide application because of its simplicity and 30 seconds. The tracking accuracy under these conditions
hence low cost at many of the tracking stations through- is estimated at _+ 1.5". Another maritime terminal
out the world. It is particularly suitable for use with geo- employing the step-track system is the Standard-B ship
stationary satellites where the link budget does not place earth stations used with the INMARSAT satellite. One
undue demands on the antenna pointing system. Ohashi such system, described by Shiokawa et al. [37] employs a
et al. [34] discuss such a scheme employed on the 2 m 4-axis mount allowing El/Az/Y/X movement. Pointing
antenna Earth station of the Radio Research Labor- errors are kept below 3" under 'normal' sea state condi-
atories, Japan. The operational frequency of the unit is tions.
20/30 GHz and programme tracking back-up is essential
for successful operation in rainy conditions when the
beacon SNR falls below the required operational thresh- 5 Electronic beam squinting
old.
Richharia and Verma [35] discuss the application of The appearance of the electronic beam squinting, EBS,
step-track to a larger 10.7 m parabolic antenna devel- technique has only occurred in recent years. EBS rep-
oped at the Indian Space Applications Centre to track resents a pseudosimultaneous amplitude sensing system
400 IEE PROCEEDINGS, Vol. 135, Pt. F,No. 5, OCTOBER I988
and is a logical continuation in the development of the axis (x in Fig. 8) each one of the four dipoles is short-
sequential lobing technique. It employs electronic switch- ircuited in turn to steer the beam to the positions 1,2, 3,
ing techniques to effectively achieve simultaneous spatial 4 on the contour plot. The received beacon signal
measurement of the beacon signal, and hence single time strength at these positions is measured in the receiver
stage and stored in conjunction with its co-ordinate
elevat ion direction.
(UP) For the satellite position given, the measured signal
1 will be stronger with the beam in position 1 or 3 com-
pared to positions 2 or 4. Furthermore, beam position 1
will produce a stronger signal than beam position 3.
Using the data accumulated about the off-axis per-
formance of each direction during a single time frame, the
co-ordinate gosition x is computed and this provides an
error signal for the feedback loop operating the steering.
To establish the position estimates, if a parabolic main
beam shape is assumed it can be shown that the tracking
error is given by:
a = 6G/(K8) (volts/deg.) (7)
where 6G is the difference in gain measured at the two
squinted positions in a common plane, 0 is the beam
I elevation deflection, and K is a constant.
t (down) The Rude Skov. I1 system operating at 1.7 GHz, is
Fig. 8 Polar diagram showing directional location of secondary beam used to track low orbiting weather satellites, such as the
peak levels ( I , 2,3,4) relative to boresight (0)and incoming beacon (x) American TIROS-N, and geostationary weather satellites
including GEOS, METEOSAT and the Japanese GMS.
controller
(microprocessor)
mode generators to /
squint the secondary

rol signal

pin dipole)

Fig. 9 RF feed chain employing electronic beam squint tracking for circularly polarised beacon signals

frame tracking error determination. It requires a rela- A programme track facility is available if the EBS system
tively simple single channel tracking receiver (similar to fails. Orbit prediction, for the programme tracking
that employed in the step-track scheme) and offers the scheme, is based on the ephemeris data; time and longi-
potential of tracking accuracies approaching that of tude where the satellite’s footprint will cross the equation
monopulse because of the near simultaneous nature of and known orbital data.
tracking error derivation. A scheme utilising the same principle, but employing a
One of the earliest schemes employing the beam squint different feed technology, was patented by Edwards and
technique was the Rude Skov. I1 satellite receiver in the Watson [40] in 1984. This new technique, also described
Netherlands [38, 391. The tracking system antenna com- by Dang et al. [41] and Watson and Hart [42], takes the
prises a central dipole around which are located four form of generating proportions of higher-order wave-
equally positioned parasitic dipoles. These are mounted guide modes within the antenna feed horn to elec-
on a common ground plate, installed in the focal plane of tronically squint the secondary antenna pattern in
a 3 m parabolic reflector. Individual parasitic dipoles can azimuth and elevation. The technique uses a high per-
be made to idle (not working) or can be short-circuited formance primary feed configuration intended for either
(working). This has the effect of squinting the copolar on-board satellite or earth station applications. As with
beam of the antenna to four cardinal axes. other methods of tracking, gross changes of antenna
Fig. 8 shows the squinting action in contour plot form, pointing are undertaken by incremental drive of the main
the centre axis (4) of the circle represents the direction of axis drive motors.
the boresight, while the positions 1, 2, 3, 4 represent the A feed network was designed which produces the
squint position achieved by activating each of the four desired beam shift necessary to define the magnitude and
parasitic dipoles. To locate a satellite off the boresight direction of an antenna pointing error based on deriving
IEE PROCEEDINGS, Vol. 135, Pt. F, No. 5, OCTOBER 1988 401
 error signals in azimuth and elevation using the excita- The system clearly offers many advantages over the
tion of multimode signals. The feed (Fig. 9) comprises the standard approaches of either monopulse or step-track
serial connecton of a conical corrugated horn, mode con- techniques. The principle of these are:
verter, corrugated polariser and conventional orthomode (i) no need for a separate or expensive tracking recei-
transducer. ver
With a beacon signal off the boresight of the antenna (ii) relatively simple feed arrangement
axis the feed is dimensioned so that TMol and TE,, (iii) fast electronic acquisition, which implies the
modes are excited in the horn in the same way as system can cope with all the effects encountered in track-
described for the multimode monopulse systems. Two ing movement, thus maintaining EIRP stability
discrete mode converter sections are used. A TE,, mode (iv) improved noise performance due to high data sam-
unit, consisting of two diametrically opposed auxiliary pling rate
rectangular waveguides, is coupled longitudinally to the (v) achievable pointing accuracy should compare with
periphery of the central circular waveguide and provides the traditional monopulse techniques
azimuth plane deviation. The TM,, mode unit, also con- (vi) employment of this method leads to a reduction of
sisting of two diametrically opposed auxiliary wave- the mechanical motion during acquisition, placing less
guides, is transversely coupled and provides elevation demand and wear on mechanical parts and hence mini-
plane deviation. Each of the auxiliary waveguides is ter- mising maintenance times and cost
minated with a pin diode. Activation of the pin diode,
effected by reverse biasing it, causes the appropriate Research on the beam-squint tracking technique is pre-
incoming higher-order mode to be converted to the fun- sently being undertaken at the University of Bristol
under the guidance of McGeehan and Edwards. It is the
damental mode by introducing asymmetry into the circu-
aim of the research programme to develop the method
lar waveguide. This, as for the previous scheme, has the
effect of squinting the copolar beam to four cardinal axes. for use at frequencies up to 60 GHz and investigate the
use of novel phase-locked loop and automatic gain
The tracking error signals are similarly obtained by
sequential activation of the mode converters to squint the control techniques to further enhance the operational
beam to the four positions (Fig. 8) defined for the pre- performance.
vious system. The complete tracking system is shown in
Fig. 10. The received beacon signal is detected by a com-
munication receiver, transformed by an A/D converter 6 Control algorithms
into a digital measurement, which is then passed to a
microprocessor. Based on the amplitude of the tracking With the advent of inexpensive and powerful digital com-
error voltage and the control technique employed, this puting, the application of control algorithms to tracking
then instructs the steering control mechanism. The rapid systems has become possible. Such techniques allow opti-
switch-and-measure sequence enables the whole search misation methods and orbit prediction to be accom-
pattern to be completed in a fraction of a second and can plished which can yield improvements in the dynamic
be regarded as simultaneous. performance of the system and hence lower the minimum
Results presented by Dang et al. [41] indicate the required signal/noise ratio. This section does not aim to
beam-squint tracking system appears to be fulfilling the examine in detail all the possible approaches but should
accuracy claims made of it by its designers. A feed chain give a flavour of the work that has been, or is being
has been installed in a 5.5 m diameter offset Gregorian undertaken in this area.
antenna system (3 dB beamwidth 0.32"). The effective
beam offset was measured at 0.063'. The beacon receiver
was a commercial coherent receiver with a phase locked 6.1 Optimal control techniques
loop bandwidth of 3 Hz and an output time constant of There has been little published material addressing
1 ms. The nominal scan frequency was set at 512 Hz. The optimal control techniques applied to satellite tracking
figures provided indicate an acquisition range greater systems. The earliest example of work is that undertaken
than 0.3' and steady state tracking error less than 0.003". by Prime [43] who considered the subject in 1966. Mod-
This corresponds to a beacon variation of less than elling the geometrical form of the radiation lobe as a two
0.1 dB and compares very favourably to the results that dimensional 'hill', he examined
could be obtained for a monopulse scheme installed on a climbing strategies, and identifie
similar sized antenna. as :

traffic
T poiariser
;sator -(Rx)
-(Tx)

antenna diode
control unit interface
(includes encoders
microprocessor
elevation
and A/D converter - motor

Fig. 10 Simplified block diagram for beacon and receiver channels

402 I E E PROCEEDINGS, Vol. 135, Pt. F , No. 5, OCTOBER 1988

 (a) that the optimising process has to be adapted to a account and determines an orbit that provides the best fit
system controlled in terms of axial co-ordinates, rather to the collected data subject to the known dynamics of
than one characterised by conventional Cartesian co- the orbital motion of the body. Such a process can
ordinates greatly enhance the achievable tracking performance of a
(b) the mode of operation is subjected to a per- tracking facility in the case where:
formance criterion involving minimisation of mean (a) the orbital estimator is integrated in the autotrack-
square error in the presence of constraints ing facility and forms part of the closed loop control
mechanism
Prime considered a control strategy whereby the beacon (b) the orbital estimator, using data from a number of
signal strength is made to ascend the quadratic hill tracking facilities, predicts the satellite orbit. This data
(preferably along the maximum slope) as the antenna being used by programme track facilities to control
aligns itself with the target. The method relies upon a antenna positioning
priori knowledge of the lobe geometry and under perfect
conditions the procedure described was exact and solu- The major constituents of an orbit determination
tions could readily be obtained. The introduction of process are the system model, the method of measure-
signal noise, however, creates a nonlinear function with ment and the estimation technique. Raol and Sinha, in
consequent interaction of the control signals. This probably the only paper of its kind [47], examine in
problem was addressed later by Stacey and Prime [44] detail the topic of orbit determination and review the
from which developed the gradient algorithm. various techniques which have been employed in these
This and other early work laid the foundations for the three areas.
development of optimal control techniques applied to Many system models have been used to describe the
satellite tracking. Many optimisation methods have been motion of a body in space. These include the classical
applied to other control systems (aircraft stabilisation, orbit elements (C6) model, the unified state model
radar, fire-control units) and these offer the opportunity (USM), the position-space measurement model and com-
to enhance the operational characteristics of the satellite binations of the above. Organisations predicting sateIlite
tracking system. Techniques which could be employed orbital data, for use at programme tracking stations, will
included regression, Fourier transform methods, poly- choose a method dependent upon the accuracy required
nomial filters and estimation techniques for signal trends. and the measurement data available. Where orbit predic-
Of the work that has been published, that described by tion is being used as part of an autotracking system, such
Choi [45] and Raghavan and Satyanarayana [46] is models are often too accurate and implementation in
typical. Choi examines three types of filter, the standard real-time impractical. In these circumstances simplified
Kalman filter, the time-varying a-B filter, and the adapt- satellite motion models will be used. Richharia [48]
ive Kalman filter. Details of algorithms for each type of examining an optimal strategy for tracking of geostation-
filter are given and performance characteristics examined ary satellites proposed a simplified model based on an
using tracking data obtained from the Eastern Test elliptical orbit perturbated with sinusoidal and linear
Range at the Eastern Space and Missile Centre. The functions. A computer simulation for the model showed if
results presented show all three filters offer good per- the tracking was based on the model prediction alone,
formance when tracking a slowly moving satellite. For a the 3a error would only be (the half-power beamwidth of
manoeuvring target, however, the time-varying a-j? offers the antenna)/100. Edwards and Terrell [49] also use a
superior tracking performance. It also has the advantages simplified model in their smoothed step-track antenna
of a lower processing requirement and easier system controller. This model, originally conceived by Slabinski
implementation. [SO], consists of a polynomial part, quadratic in time, a
Raghavan and Satyanarayana [46] discuss the use of a simple trigonometric part whose period is the siderial
recursive filter in a microprocessor-based digital antenna day, and two Poisson terms. Slabinski showed the error
controller. Mathematically this filter can be represented in such a point expression to be less than 0.015' over a
as : 28-day interval.
N N
The type of measurements used in the orbit prediction
Y(nT)= 1 akX(nT - kT)- 1 bk Y(nT - kT) process again depend upon the problem in hand. A
(8) ground station employing autotracking will generally just
k=l k=l
have azimuth and elevation data derived from the track-
where Y(nT)is the predicted response, this being a func- ing process whether it be monopulse, step-track etc. A
tion of the present response and past N values of the satellite employing satellite-to-satellite tracking tech-
response. The authors do not discuss the calculation of niques might also have inputs available from on-board
the filter coefficients or give experimental results of its instruments such as star-trackers, horizon trackers and
use. gyro units. Highly accurate orbital models will require
possibly range, range rate, spacecraft-based observation
data, as well as azimuth and elevation data from one or
6.2 Orbit determination more ground stations at various locations throughout the
The problem of orbit determination, utilising least world.
squares estimation techniques, was first addressed by The orbital estimation problem can be expressed as:
Gauss in 1795. Much work has been undertaken from
Gauss via Kalman to present-day numerical stable and f =f(x(t), t ) + n(t) (9)
accurate estimators for precise orbit determination.
Orbit determination is the process of describing the z(t) = W t ) , t ) + 4 t ) (10)
orbit of a satellite or other body, based on a set of obser- where eqns. 9 and 10 represent the system and measure-
vations. Consideration of the possible tracking scenarios ment models respectively, associated with the orbiting
shows these observations will either be made at a ground satellite and data tracking system. Here n is the vector of
station or from sensors on-board a satellite. A high- n-states of the chosen co-ordinate system and other aug-
precision orbit estimator takes the system noise into mented parameters; z is the vector of observab1es;f; h are
IEE PROCEEDINGS, Vol. I35, Pt. F, No. 5, OCTOBER I988 403
known, nonlinear function relationships; and n, v are from an individual step-cycle measurement tends to get
process and measurement noise vectors. Given these greatly reduced by the smoothing effect of the tracking
models and some a priori information on x(O), n, v and algorithm. This allows the system to work at a much
the 'noisy' measurements available, a filtered estimate of reduced SNR compared to the standard step-track
x is required. This can be achieved using an estimator. scheme.
orbit
e infor mat ion S

measurement antenna
satellite -- system estimator con troller dynamics

Fig. 11
t1
Estimatorlcontroller tracking system
xa

Raol and Sinha discuss more than ten types of estima- Edwards and Terrell used a least squares algorithm in
tor that have been used, or considered, for orbit predic- order to fit the tracking data to the orbital model.
tion applications. These range from the well known Analysis, it is claimed, of the mathematical properties of
techniques of Gaussian least squares differential correc- the algorithm result in the RMS pointing error being
tion (GLSDC), Kalman filtering and Adaptive filtering to reduced by a factor of typically 5.5. The factor of 5.5
the less well known methods, including the SBDC tech- improvement in allowable beacon level fluctuations
nique and Pugahev filtering. The reader is directed to translates to a corresponding improvement in oper-
Reference 47 for a comprehensive list and more complete ational SNR. That is random Gaussian beacon noise
description. may be:
The type of estimator employed again depends upon
20 log,, 5.5 = 15 dB (11)
the application. Factors including the computational
power available, the type of measurement data, noise times as great for SST as step-track for a given pointing
properties and the required accuracy, all have to be con- performance. Therefore whereas step-track typically
sidered. Possibly the best known and most widely used requires SNR 35-40 dB, SST will tolerate less than 25 dB
estimator is the Kalman filter. Dressler and Tabak [51] tracking signal SNR.
reported its use in a high-performance satellite tracking Computer simulation of the tracking algorithm with a
system in 1971. This system (Fig. 11) was the first which 0.035" step-cycle suggested an RMS tracking error of
combined the use of optimal estimation and control tech- about 0.003" was possible. Practical trials were carried
niques. The estimator, an extended Kalman filter (EKF), out using the 32 m Madley No. 3 Cassegrain antenna
generates an estimate P of the present system state x and an 8 m commercial Gregorian system. For the 32 m
based on all the noise-corrupted measurements z up to Madley antenna, using a 0.005" step-cycle, the tracking
that time. This estimate is then employed in the control- error did not exceed f0.007" over a 3 day period. For
ler to compute a control signal u to drive the antenna so the 8 m system, using a 0.05" step-cycle, the communica-
as to optimise the tracking performance criterion. Exten- tion channel fluctuation is kept below 0.25 dB. The SST
sive simulations of the tracking system utilising the system, following this experimental work, was installed
estimator/controller algorithms were performed in con- on the Goonhilly No. 1 antenna in 1983.
junction with models of a 10 m and 25 m antenna. Dress- The use of a tracking system employing orbital estima-
ler and Tabak reported that the simulations tion and optimal control techniques is presented by
demonstrated that a real-time tracking accuracy of James and Maney [53] as a way of overcoming many of
0.0005" could be achieved when tracking orbiting satel- the special problems associated with shipboard satellite
lites. Similar systems are described by Orlando and Net0 terminals. Of these, one of the most serious is the effect of
[52] and Richharia [48]. Richharia's scheme, mentioned the ship motion. This causes misalignments to couple
earlier, utilises a scalar Kalman filter and much simplified into the antenna position as time-varying components
orbital model. Computational requirements are therefore with roll, pitch and heading frequencies and their cross
much reduced but tracking errors of up to 0.01" can be harmonics. Standard autotrack techniques are adequate
expected under worst case conditions. to track the satellite, but usually lack sufficient band-
Tracking facilities employing both orbit prediction width to null these time varying errors.
and optimal control techniques are now becoming more James and Maney's approach for tracking the satellite
common-place. Two such systems will be briefly exam- and solving the misalignments is shown in simplified
ined, the smoothed step-track antenna controller [49] form in Fig. 12. The scheme utilises the Kalman filter
and a shipboard satellite terminal described by James operating on a nonlinear representation of the two
and Maney [53]. orthogonal pointing errors and multiple pointing error
The smoothed step-track (SST) controller, described measurements to solve for the multiple error sources. In
by Edwards and Terrell [49] has already been mentioned these circumstances the pointing error model becomes a
with regard to the method of orbit prediction used. The weighted sum of the misalignment variables, with time
SST technique exploits the concept of step-track mea- varying coefficients.
surement, in combination with orbital prediction, to form The results of a number of computer simulations are
an accurate estimate of the satellite track, which is used presented by the authors. These show the necessary con-
for continual prediction of the present satellite position. vergence in the alignment corrections and their depen-
At any time the accuracy of the position estimate will dence on the measurement noise. Further it is clear a
depend on a host of data spanning the previous days of trade-off exists between the steady-state noise jitter and
tracking. Therefore the error in estimating the position dynamic errors. In all, the results indicate that tracking
404 IEE PROCEEDINGS, Vol. 135, Pt. F, No. 5, OCTOBER 1988
sate I I ite
calculated?
posit ion

satellite
position
{THTHTl
conversion
coordinate con
system
tro I antenna

antenna
position

Q
platform
attitude,
antenna
alignment orientation

I corrections 1
I
1
1

Kalman tracking
tracking error
filter error

Fig. 12 Maritime tracking system

Table 1: Summary of satellite t r a c k i n g systems

Tracking Subcategory Remarks Performance Usage
category
Manual non-autotracking; tracking accuracy many stations can
simple; dependent on revert to manual
requires operator operator - generally tracking if other
low tracking methods fail
Programme non-autotracking; tracking accuracies often employed as back-up
steering simplistic with approaching 0.01' in case primary
modern technology; possible tracking system fails
requires operator
intervention;
accuracy reliant on
orbit prediction
Monopulse Simultaneous autotracking tracking accuracy widespread in many of
lobing [l] tracking information very good, the larger earth
Multimode obtained in a single typically 0.005"; stations;
-TMo, 121 time frame; operational SNR also shipborne
- TM~,/TE,, [31 2,4 or 8 channel -15 dB; terminals, sat-to-sat
- orthogonal coherent receiver fast dynamic comms
--E,, [41 required ; response
expensive
Conical Rotation of antenna autotracking; tracking accuracy widespread up until
scan VI; sequential amplitude typical 0.01'; mid 70s;
rotation of offset sensing system; operational SNR monopulse and step-
feed [2]; sensitive to AM -30 dB; track then preferred
rotation of sub- interference; medium dynamic
reflector [3]; mechanically complex; response
TMol mode [4]; uses single channel
TMo,/TE,, mode [51; receiver;
Conopulse [6] methods 1-3 result in
modulation of uplink
transmission
Step-track step antenna [l]; autotracking; tracking accuracies widespread where
nodding sub- sequential amplitude of 0.01' have been lower accuracy acceptable
reflector [2] sensing system; obtained, typically and cost important;
sensitive to AM <0.1 O; many of the smaller
interference; operational SNR earth stations, ship-
simple low cost -30-45 dB; borne terminals
system ; slow dynamic response
single channel
receiver;
method [l]results in
high wear of
mechanical drives
Electronic dipole feed [l3 ; autotracking; tracking accuracy use limited to
beam TMo,/TE2, mode; pseudoamplitude approaching monopuIse, mainly experimental
steering conversion sensing system; 0.005"; tracking systems;
techniques [2] much reduced sensitivity operational SNR shows good overall
to A M interference; 15-30 dB; tracking performance;
uses single channel fast dynamic well suited to sat-to- ,

receiver; response sat comms;

relatively simple should be widely used
in future

IEE PROCEEDINGS, Vol. 135, Pt. F , NO. 5, OCTOBER 1988 405

 ~ -

errors of less than 0.01” could be expected if using this 3 BRUSH, R.J.H., and BAYLIS, P.E.: ‘An automatic high-resolution
system. picture transmission station’, ESA Journal, Vol. 8, 1984, pp. 425-435
4 CUMMINGS, D., and WILDEY, C.G.: ‘Military aeronautical satel-
lite communications’, IEE Proc. F, 1986,133, pp. 4 1 1 4 1 9
5 ANDERS, J.V., HIGGENS, E.F., MURPHY, J.L., and SCHAE-
FER, F.J.: ‘The precision tracker’, Bell Syst. Tech. J., 1963, pp.
7 Conclusions 1309-1356
6 PRA’TT, T.: ‘High efficiency feeds with tracking capability’. IEE
Conference on Earth Station Technology, 1970, pp. 250-256
Over the past 25 years or so a number of error sensing 7 MILNE, K., and RAAB, A.M.: ‘Optimum illumination tapers for
techniques have been used in satellite tracking systems. four-horn and five-horn monopulse aerial systems’. IEE Conference
These are summarised in Table 1. Publication 21, 1966, pp. 12-16
Of these, two have become prevalent, they may be 8 SCIAMBI, A.F. : ‘Five-horn tracking feeds for large antennas’. IEE
described as monopulse and step-track. The monopulse Conference Publication 21, 1966, pp. 158-162
9 SCIAMBI, A.F.: ‘Five-horn feed improves monopulse performance’,
technique relies on sensing the received wavefront and Microwaoes, June 1972,11, pp. 56-58
processing the signal to derive the pointing information. 10 COOK, J.S., and LOWELL, R.: ‘The Autotrack system’, Bell Syst.
By tracking the amplitude and phase of a higher-order Tech. J., 1963, pp. 1283-1307
wavefront, tracking errors less than 0.005” can be 11 LOZIER, J.C., NORTON, J.A., and IWAMA, M.: ‘The servo
system for antenna positioning’, Bell Syst. Tech. J., 1963, pp. 1253-
achieved. Unfortunately, the system relies upon at least a 1281
two-channel coherent receiver for its operation which 12 CHIVERS, M., and BRAIN, J.R.: ‘A simple multimode tracking
incurs the penalties of mass and expense. Whilst such system for microwave aerials’. IEE Conference Publication on Satel-
factors can be absorbed in a typical ground station, they lite Communication System Technology, 1975, pp. 116-121
represent a high penalty in terms of a spacecraft and the 13 HAYASHI, R,, TESHIROGI, T., KIMURA, S., KOMURO,
H., YOSHIMURA, K., SUZIKI, Y., and AOYAMA, S.: ‘ETS-I1
mass of the monopulse system leads to a reduction in the experiments part I1 : Earth station facilities’, lEEE Trans., 1980,
effective communications payload. In the case of step- AB16
track systems, a relatively simple noncoherent receiver 14 INOUE, T., and KAITSUKA, T.: ‘K-band tracking system for
can be used to sense the amplitude of the signal. Whilst domestic satellite communication system’, IEEE Trans., 1981,
A B 1 7 , pp. 561-570
this leads to a significantly cheaper implementation than 15 CHOUNG, Y.H., GOUDEY, K.R., and BRYANS, L.G.: ‘Theory
the monopulse technique, the performance of the com- and design of a Ku-band TE,, mode coupler’, IEEE Trans., 1982,
munications link is sacrificed and a minimum signal/ WIT-30, pp. 1862-1866
noise ratio of some 30 dB is required for satisfactory 16 SAVINI, D.: ‘An improved automatic tracking system for linear and
circular polarization’, CSELT Rapp. Tec. (Italy), 1980, 8, pp.
operation. This is to be compared with a figure of 15 dB 239-292
for the monopulse technique. Furthermore, as error 17 WATSON, B.K., DANG, N.D., and GHOSH, S.: ‘A mode extrac-
detection of the system is not real-time it suffers from tion network for rf sensing in satellite reflector antenna’. IEE Second
dynamic lag, the overall accuracy is degraded and the International Conference on Antennas and Propagation, 1981, pp.
systems have significant dynamic limitations. 323-327
18 MOLKER, A., WATSON, B.K., and ERICSSON, L.M.: ‘A tracking
In the early eighties electronic beam squint techniques mode coupler for a circular waveguide feed system to provide off-
have emerged. These techniques, while still very much in boresight tracking capabilities’. Technical Note Ml/KV8066, 5th
the research stage, appear to have overcome many of the July 1978
problems associated with the early systems. They offer a 19 MOLKER, A., and ERICSSON, L.M.: ‘Cross-polarisation effects
tracking accuracy and dynamic performance similar to on different rf sensing systems’. Technical Note Ml/KV8099, 25th
September 1978
that of the monopulse system while at the same time ,
20 REINDERS, M.A., and OEI, W.S.: ‘Antenna feed system for a satel-
being relatively simple. Only a single-channel receiver is lite station with linearly and circularly polarized beacon tracking
required, resulting in a relatively cheap and low mass capabilities’. NTG-Fachber (Germany), Vol. 57, June 1977, pp.
system. In addition to this, the acquisition range of these 97-101
21 BROOKS, C.W.: ‘Operational shipborne earth stations’. IEE Con-
techniques is considerably beyond that of monopulse ference Proceeding on Satellite Systems for Mobile Communications
systems as the error signal is derived from the primary and Surveillance, 1973, pp. 52-55
beam of the antenna. The electronic beam squint system 22 REED, A.G.: ‘A purchaser’s view of some developments in earth
seems ideally suited for use in the next generation of station technology’. IEE Conference Proceedings on Satellite Com-
satellite tracking applications as the trend towards digital munications System Technology, 1975, pp. 102-109
23 WOOSTER, C.B., and MILLER, B.J.: ‘Advances in Intelsat earth
processing continues. station technology’. Communication ’78. IEEE Conference on Com-
Furthermore, the availability of cheap, small and munications Equipment and Systems, Birmingham, 4-7th April
powerful microprocessors offers the opportunity of 1978, pp. 220-224
applying signal processing techniques to satellite track- 24 SKOLNIK, M.I.: ‘Introduction to radar systems’(McGraw-Hill)
25 HARRIES, G., and HEAVISIDE, J.W.: ‘Naval satellite communica-
ing. Theoretical and practical work in the areas of tion terminals’. IEE Conference Proceedings on Satellite Systems for
optimal estimation and orbital prediction show signifi- Mobile Communications and Surveillance, 1973, pp. 48-51
cant improvements in system performance can be 26 HARRIES, G., and STANNARD, E.J.: ‘Design of a small shipborne
achieved beyond that of simple error sensing. In many terminal (“SCOT)’. IEE Conference Proceedings on Earth Station
future systems it is clear that, such techniques, will form Technology, 1970, pp. 63-69
27 KNOX, D.M., and DOBLE, J.E.: ‘The modified primary feed and
as much an integral part of the tracking facility as the tracking system for Goonhilly No. 1 aerial’. IEE-Conference Pro-
type of radio-frequency sensing technique employed. ceedings on Earth Station Technology, 1970, pp. 201-207
28 MARCONI CO. LTD.: ‘Satellite tracking system’. Patent UK
1171401,24th January 1968, published 19th November 1969
29 LOCKET”, N.J.L., and SHINN, D.H.: ‘Frequency selective conical
scanning, using waveguide mode conversion’. IEE Conference Pro-
8 References ceedings on Earth Station Technology, 1970, pp. 257-262
30 OHISON, J.E., and REID, M.S.: ‘Conical scan tracking with the 64
1 CHAPIN, J.C.: ‘Antenna steering systems - past, present and m diameter antenna at Goldstone’. Contract NAS7-100, Jet Propul-
future’. IEE Conference on Earth Station Technology, 1970, pp. sion Lab. Calif. Inst. of Technology, Pasadena, October 1976
420-424 31 TOM, N.N.: ‘Autotracking of communication satellites by the step-
2 BLAKE, R.G.: ‘Down-to-earth measure for space communication’, track technique’. IEE Conference Proceedings on Earth Station
Post Ofice Telecom. J., Winter 1978/79, pp. 8-10 Technology, 1970, pp. 121-126

406 IEE PROCEEDINGS, Vol. 135, Pt. F, No. 5, OCTOBER 1988

32 LARKIN, R.J. : Philco-Ford Invention Disclosure Memo, 16th 43 PRIME, H.A.: ‘Application of optimal control techniques to large
October 1967 steerable aerials’. IEE Symposium on Design and Construction of
33 ELEY, R.G.: ‘Auto-tracking receivers’. IEE Conference Proceedings Large Steerable Aerials, June 1966, pp. 3 9 4
on Earth Station Technology, 1970, pp. 302-308 44 STACEY, B.A., and PRIME, H.A.: ‘The application of search tech-
34 OHASHI, H., IGUCHI, and HASHIMOTO, Y.: ‘A geostationary niques to autotracking’. IEE Conference Proceedings on Earth
satellite tracking system for small-sized earth stations’, Reo. Radio Station Technology, 1970, pp. 133-139
Res. Lab.(Japan), December 1983,29, (153), pp. 527-538 45 CHOI, J.: ‘Comparison of real-time tracking algorithm for airborne
35 RICHHARIA, M., and VERMA, N.N.: ‘Design of a step-track for target’. Conference Proceedings IEEE Southeastcon ’85, USA, 31st
low cost earth stations’, IETE (India), 1982,28, pp. 533-538 March-3rd April, 1985, pp. 139-143
36 JOHNSON, M.B.: ‘Antenna control system for a ship terminal for 46 RAGHAVAN, S.V., and SATYANARAYANA,T. : ‘Microprocessor-
MARISAT’. IEE Conference Proceedings on Maritime and Aero- based digital controller for antenna pointing systems’, IETE (India),
nautical Satellite Communications and Navigation, 1978, pp. 1983,29, pp. 57-61
123-126 47 RAOL, J.R., and SINHA, N.K.: ‘On the orbit determination
37 SHIOKAWA, T., KARASAWA, Y., and YAMADA, M.: ‘Compact problem’, IEEE Trans., 1985, AES-21, pp. 274-290
antenna systems for INMARSAT Standard-B ship earth stations’. 48 RICHHARIA, M.: ‘An optimal strategy for tracking of geostation-
IEE Conference Proceedings on Maritime and Aeronautical Satel- ary satellites’,JIETE (India), 1984,30, pp. 103-108
lite Communications and Navigation, 1978, pp. 95-99 49 EDWARDS, D.J., and TERRELL, P.M.: ‘The smoothed step-track
38 VOLDBY, J.E.: ‘A new satellite receiver at Rude Skov. II’, Elektronik antenna controller’, Int. J . Sat. Commun., 1983,1, pp. 133-139
(Denmark), January 1978,1, pp. 1.&14 50 SLABINSKI, V.J.: ‘Expressions for the time-varying topocentric
39 HUELE, H.T.: ‘An earth station for weather satellites’, Philips direction of a geostationary satellite’, Comsat Tech. Rev., 1975, 5, pp.
Telecom. Rev., September 1981,39, pp. 118-129 1-14
40 EDWARDS, D.J., and WATSON, B.K.: ‘Electronic tracking 51 DRESSLER, R.M., and TABAK, D.: ‘Satellite tracking by com-
systems for microwave antennas’. British Patent Application No. bined optimal estimation and control techniques’, IEEE Trans.,
8414963,12th June 1984 1971, AC-16, pp. 83%840
41 DANG, R., WATSON, B.K., DAVIS, I., and EDWARDS, D.J.: 52 ORLANDO, V., and NETO, O.M.: ‘Development of an algorithm
‘Electronic tracking systems for satellite ground stations’. 15th Euro- for real-time control of a satellite tracking system’. 13th Internation-
pean Microwave Conference, Pans, France, September 1985, pp. al Symp. on Space Technology and Science, Tokyo, 27th July 1982
681487 53 JAMES, M.B., and MANEY, J.J.: ‘Adaptive alignment of a ship-
42 WATSON, B.K., and HART, M.: ‘A primary-feed for electronic board satellite terminal’. Military Communications Conference,
tracking with circularly-polarised beacons’. Proceedings of Military MILCOM ’85, Boston, USA, October 1985, Vol. 1, pp. 300-305
Microwaves, Brighton, June 1986, pp. 261-266

IEE PROCEEDINGS, Vol. 135, Pt. F, No. 5, OCTOBER 1988 407