Cooperative Research Centre for Satellite Systems

Number 83, June 1999

Australia & Korea

joint venture

Cooperation between countries in the region is a key

part of our Centres development strategy, said Dr Brian
Embleton, Executive Director of the CRCSS. Through
working together, agencies can reduce technology
development times, contain costs, and get our products to
the community more quickly.
KAIST is a diverse Korean scientific research organisation, first established in 1971. In July 1989 KAIST
merged with the Korean Institute of Technology, an elite
graduate-level scientific teaching institution. KAIST now
has dual roles of education and research; it employs over
12,000 scientists and engineers, with an additional student
population of nearly 7,000.
KAIST established its Satellite Technology Research
Centre (SaTReC) in 1989. As with the CRCSS, SaTReC aims
to promote education and training through research
programs in space science and engineering.
In 1992, SaTReC launched Koreas first satellite, the
scientific microsatellite KITSAT-1. Since then KAIST has
launched KITSAT-2, and KITSAT-3 was launched in May
1999.
KAISTSAT-4 continues the tradition of Korean scientific
mircosatellites. In addition to the CRCSS communications
payload, it will carry remote-sensing and space-science
instruments.

DACRCSS will work with the Korea Advanced

Institute of Science and Technology (KAIST) to
duplicate CRCSS-developed communications payloads on
future Korean satellites.
FedSat, the CRCSS first satellite mission and due for
launch in November 2000 [SpIN 82], will carry a sophisticated communications payload. The Centre plans to fly a
similar payload on the Korean microsatellite
KAISTSAT-4 in 2002.
The duplication will improve the communications
usefulness of both satellites. More satellites carrying the
same payload means less time between satellite overpasses,
therefore more regular communications.
The FedSat payload will receive, store and transmit
signals in three communication bands of varying power.
UHF (Ultra High Frequency) is useful for cheap, lowpower terminals. It provides low data rates but does not
require a accurate antennae. Ocean buoys use UHF to
upload salinity and temperature data to satellites. KAIST
also plans to use modified UHF ground terminals for fleet
management, traffic monitoring, and wildlife tracking.
The medium frequency S-band is used for
spacecraft control commands, and relaying
scientific data. The highest power, Ka-band,
requires a very accurate antenna and special
Earth stations, but carries data at up to 256
kbits/sec. The Ka-band system will provide
high bandwidth communications, such as
Internet and video-conferencing, to remote
areas without needing supporting infrastructure.
KAIST and the CRCSS announced the
plan at a regional meeting of space scientists
and engineers, held in Kuala Lumpur in
May. The meeting examined how AsiaPacific countries can benefit from space
applications, and share costs through using
common payloads.
The common payload concept has a
range of potential applications, and will be a
catalyst for collaboration in space science
and technology throughout the region, said Left: The Indian Space Research Organisations rocket PSLV-C2. Right: Microsatellites
Prof. Soon Dal Choi, Director-General of
KITSAT-3 (left) and DLR-TUBSAT (right) are dwarfed by IRS-P4 (top) in the payload bay,
May 1999. Images courtesy KAIST.
KAISTs satellite research centre.

SpIN 83, June 1999

In this issue ...

Australia & Korea
joint venture . . . . . . . . . . . . . . . . . 1
Come-in FedSat . . . . . . . . . . . . . . . 3
UTS in context . . . . . . . . . . . . . . . . 6
Computing on the fly . . . . . . . . . . 7
Reconfigurable computers
on the ground . . . . . . . . . . . . . . . . 9
The space multipath landscape . . 10

Auspace opens
Adelaide office

4+55partner organisation, Canberra-based spaceengineering company Auspace Limited [SpIN 80],

announced in May the opening of a new Adelaide office.
Auspace says this move will better serve the needs of its
Adelaide customers.
Dr Gerald Bolding will head the new office in
Technology Park. He has experience in design and
development of satellite modem systems. He is currently
establishing a team to develop new digitaltelecommunications products.
Auspace is a subsidiary of the European company, Matra
Marconi Space.
Further information:
Mr Peter Dingley
Business Development Manager
Auspace Limited
50 Hoskings St
Mitchell, ACT 2911
pdingley@auspace.com.au

Variable orbits . . . . . . . . . . . . 12

All feature articles in this issue were written by

Wayne Deeker, based on interviews with CRCSS
personell.

CRCSS Space Industry News

SpIN is published quarterly in March, June, September and December by the:
Cooperative Research Centre for Satellite Systems
GPO Box 1483
Canberra ACT 2601 Australia
Phone: 61 2 6216 7270
Facsimile: 61 2 6216 7272
http://www.crcss.csiro.au/
e-mail: firstname.lastname@crcss.csiro.au
ISSN 1329-4857
First issue: Number 78, March 1998
Former titles:
COSSA Space Industry News ISSN 0816-7-44, Nos. 1-35,
Nov. 1985-Aug. 1990
CSIRO Space Industry News ISSN 1037-5759, Nos. 36-77,
Oct. 1990-Dec. 1997
Editors: Wayne Deeker & Jeff Kingwell, CSIRO Telecommunications and Industrial Physics
Editorial consultant: Brian Embleton
Design & production: Wayne Deeker
Contributions and reproduction
Contributed articles should arrive at the above address by the beginning of the month before
the month of publication. Material published in SpIN may be quoted or reproduced if
accompanied by an acknowledgment to the Cooperative Research Centre for Satellite
Systems.
Disclaimer
To the extent permitted by law the Cooperative Research Centre for Satellite Systems accepts
no responsibility for the way information or opinions in CRCSS Space Industry News are
interpreted or used.
Masthead and Variable Orbits graphics design, Wayne Deeker.
Cooperative Research Centre for Satellite Systems, Canberra, 1999

Cooperative Research Centre

for Satellite Systems
Mission
The mission of the Cooperative Research Centre for Satellite Systems (CRCSS) is to deliver
sustainable advantage for Australian industries, universities and government agencies
involved in services based on the applications of small satellites.
The Centre commenced on 1 January 1998, under the Cooperative Research Centres Program
of the Commonwealth of Australia. Objectives of this program include:
to contribute to national objectives, including economic and social development;
to stimulate broader education and training experience;
to capture the benefits of research and strengthen links between research and its
commercial and other applications;
to promote cooperation in research.
The core participants of the Centre comprise:
University of South Australia
CSIRO
Queensland University of Technology
University of Technology, Sydney
Auspace Limited
Vipac Engineers & Scientists Limited
University of Newcastle
The Centre also includes as supporting participants:
Defence Science & Technology Organisation
La Trobe University
Codan Qld. Ltd.
DSpace Pty Ltd
Curtin University of Technology
SpIN issues to #68 are online and cross-referenced at:
http://www.crcss.csiro.au/spin/spinmain.html

SpIN 83, June 1999

Come-in FedSat

ovember 2000: FedSat is safely in orbit.

But hold the champagne; first you have to
find it. A small error could mean searching a
billion cubic kilometers of space for a box small
enough to fit in a car boot. And once youve
found it, youll have to keep the antenna exactly
on it as it whizzes overhead. CRCSS partner
UTS, the University of Technology Sydney, is
triumphing over these engineering challenges.
Theyre finding novel solutions, which will allow
unprecedented communication benefits to
people in remote areas.

A@5=J\I payloads include experimental Ka-band

communications equipment. UTS is responsible for
designing and building a transportable Ka-band Earth
Station to communicate with FedSat.
Ka-band is a relatively new area of high-frequency
telecommunications technology with exciting possibilities.
It offers the availability of wider bandwidths allowing
faster data-transfer rates, not available using lower
frequency bands. But because of its newness, and because
of problems unique to that band, CRCSS engineers have
many unusual problems to overcome.
Keith Willey, engineering PhD student, is part of the
UTS team. The big problem with Ka-band, he explained
to SpIN, is the high frequencies. The higher the frequency
you use, the smaller the beam width is for a particular size
antenna. Because FedSat transmits less power than a
small torch, it takes a big antenna (1.2 m) to receive its
data. Since Ka-band gives a very narrow beam width, that
big dish has to point very accurately.
Another problem with Ka-band is signal interference.
The signal fades, and experiences multipathing (reflections) at low elevation angles. This causes problems similar
to ghosting on your TV. Even rain weakens it. In commercial systems currently available, says Keith, they use
a much higher link margin; they give themselves extra
power to overcome these things when they happen.
FedSats an experimental satellite, so We havent got that
extra power.
Normally in tracking if the signal drops off, it means the
antennas gone too far off target. But at Ka-band, fadeouts might mean that ... weve hit a cloud, or weve gone
through a more humid part of the atmosphere, or theres
some scintillation going on [winds in the upper atmosphere which distort the signal]. These must be separated
from genuine signal loss due to pointing; the weak power
and narrow beam width make this difficult.
But there are some big benefits too. As a general rule,
said Keith, the higher the frequency, the more bandwidth
you can get. The reason is, lets say most electronic
components have a workable range of 1% of their design
frequencies; well, as you go up into Ka-band, because

SpIN 83, June 1999

CRCSS engineering PhD student, Keith Willey. Photo courtesy UTS.

youre at 20-30 gigahertz, 1% is 200-300 megahertz, and

that gives you a huge operating range of your components.
So you have the ability to use higher bandwidths for your
signal. Wider bandwidth means sending more data. Its
like having a wider freeway; instead of being limited to
two lanes, weve got ten lanes. So you can transmit more,
and thats one of your big advantages.
Also, as the spectrum fills up, its harder to get wide
bandwidths at the lower frequencies (UHF, C-band, Lband). As communications increase, weve got less and
less spectrum available, and Ka-bands the next step, said
Keith.

Acquisition

The first thing we have to do is acquire the satellite when

it comes over the horizon, said Keith. The main information the UTS team will use are Two Line Elements
(TLEs), a matrix of data containing the six parameters
needed to define an orbit. NORAD, the North American
Aerospace Defence Command, create and publish on the
Internet TLEs for all objects in near-earth space, down to
about the size of a tennis ball. The TLEs are used in
conjunction with an SGP (Simplified General Perturbation) orbital model to predict where and when to find
objects, including FedSat. There are other models
available, said Keith, but the advantage of using SGP
and TLEs is that theyre readily available, and the

UTS tower block

in central Sydney.
Photo courtesy
UTS.

to a much higher elevation to actually pick up the

satellite. FedSat has an innovative isoflux Ka-band
antenna, which compensates by providing more power at
low elevation angles. But nothings perfect, says Keith.
Information from the first pass will go back into the
orbital model, making the satellite easier to find on
subsequent passes. The UTS team will use the NORAD TLE
the first time, then generate their own, more accurate
versions. Orbital perturbations arent well modelled, so if
observers didnt constantly update the orbital data, the
satellites would soon be lost.

Tracking

information is updated about every week.

But TLEs can be inaccurate. Theyre only guaranteed
with 90% confidence to find the satellite within a 5 km
radius. For FedSat that represents an error of about 0.36
degrees. Our tests have shown that it is possible to get
much larger errors. But using a TLE does put us in the ball
park. Plus there might be other positioning errors from
setting up the portable Earth station further increasing the
chances that FedSat wont be where we expect.
The biggest error lies with environmental effects on the
satellite. The sun and moon tug all satellites off course;
this, combined with the Earths irregular shape, means
even a circular orbit will eventually degrade into a chaotic
one [SpIN 70]. Unlike some other low Earth orbit satellites,
FedSat has no station-keeping thrusters, so theres nothing
to stop it drifting. FedSat will have an 800km orbit; at this
altitude the primary orbital perturbation forces are due to
atmospheric drag and the non-uniform mass of the Earth.
As the satellites orbital height decreases due to drag, so its
orbital velocity increases, making it move faster across the
sky. There are also third order effects, like solar radiation
pressure, but drag and the non-uniform mass of the Earth
have the biggest effect on FedSats orbit, said Keith.
So, when the satellite comes over the horizon, the very
first time you set up your Earth station, youve got an idea
where it is, but its unlikely that youre going to pick up a
signal straight away. So you move the antenna around
until you find the satellite.
Even under ideal circumstances, FedSat will be visible
above the horizon for no more than 15 minutes; it could be
far less, even as little as 30 seconds. Everything involving
the satellite has to occur within that visible time, so the
longer it takes to find means less communication time.
How fast the UTS team finds it that first time will
depend on the error. If the TLEs are pretty accurate, well
find it within a couple of seconds, depending on the
weather. But the lower the elevation angle, the more
attenuation because of the extra atmospheric thickness.
If its a cloudy day, we might have to wait until it gets

Once youve found the satellite, the next step is to keep the
antenna pointing right at it as it zooms across the sky.
With FedSat, the narrow Ka-band beam-width means
theres very small margin for error in tracking. Even slight
deviations from dead-on will limit communications ability.
Its even more complicated because normal tracking
involves moving slightly off then back on the beam, to
gauge where the signal is strongest. We can only move
such a small amount before we lose communications that
theres really insufficient movement to get a good difference between the amplitudes, to determine the exact
position of the satellite. The problem wouldnt occur so
much at lower frequency bands. If you were using say
UHF frequencies to track FedSat, you would not suffer the
same attenuation or narrow beamwidth problems. The
pointing requirements for communications would not be
as critical. The wider beam widths mean that you can
tolerate a larger antenna pointing error, before effective
communications are lost, said Keith.
Commercial systems deal with this problem by giving
the satellite enough power and sufficient beam width. This
enables quality reception to be maintained by moving
slightly ahead of and behind the satellite. But these
options arent available for Ka band communications with
FedSat, needing a whole new Earth station concept.
I see this as a really useful part, probably an unexpected part of the project, says Keith. We didnt
envisage that wed have to design a pedestal system. But I
see it as a very interesting and useful part of our research
and development. It may also have commercial applications for the CRC.

FedSat pedestal & Earth station

So FedSat presents special tracking problems. It will be

transmitting at very low power, in a narrow beam, and
moving quickly across the sky. It needs a relatively large
dish that can move quickly in any direction, and accurately
follow an orbital pass. Most conventional satellite dishes
dont need so much agility; those dishes either point at
geostationary satellites, which hardly move relative to the
ground, or they point at satellites with much wider beam
widths.
Most satellite dishes employ azimuth over elevation
movement; one motor moves the antenna 360 degrees
around the horizon, and another points it between horizontal and vertical. But this design cannot track a satellite
directly overhead; it must spin around and re-acquire

SpIN 83, June 1999

 zenith passes on the way down. Thats okay for widebeam satellites, but for FedSat would cause critical delays.
The UTS ground station employs a newer and more agile
design. It has an X-Y axis forming a cross, says Keith.
This can accurately and quickly move the dish in all
directions.
And since the use of the Ka-band is relatively new,
suitable components are not yet widely available. So the
UTS team have to build much of the equipment themselves. The biggest thing were manufacturing is the
pedestal itself to move the antenna; this includes the
algorithms to control its movement and the required signal
processing, says Keith.
To get a pedestal capable of overhead passes and which
could keep up with FedSat is quite expensive, says Keith.
Suitable commercially available pedestals alone begin at
$100,000. Our whole Earth station is going to cost about
that, including all the electronics, the antenna, the signal
processing, everything. So all up, to commercially buy
what were building here, I think you wouldnt get much
change out of half a million dollars, maybe more.
One of the main goals of the UTS project is a transportable Ka-band Earth station capable of being set up
anywhere. Ka-band components are small because of the
high frequency, so the prototype Earth station, including
the Ka-band antenna, would easily fit inside a utility
vehicle tray. As more Ka-band manufacturers come online, subsequent units will be even smaller. Certain offthe-shelf components can be eliminated, in favour of
smaller, better integrated products which include several
components in the one box.

Applications

But what can all this bandwidth actually do? Wed really
like to communicate at a minimum of a 128 kilobits per

second, which gives us the equivalent of 2 ISDN lines,

each of which is 64 kbits/sec, said Keith. Which means
itd be possible to have say a videoconference between
Earth stations in different locations. Obviously, both
stations would have to see the satellite at the same time.
There are two modes of operation. The regenerative
mode uses an on-board computer to process the signal
then re-transmit it. Now that gives you extra capacity to
send data because it takes out some of the transmission
errors. This should give transfer rates of 250 kbits/sec. But
signal relay without reprocessing, called bent-pipe mode
because the signal goes from the ground to the satellite
and down again, should give rates of 128 kbit/sec. This is
still more than twice as fast as the fastest modems
available in 1999.
An Earth station capable of being set up nearly anywhere gives unprecedented remote-area communications
capability. FedSats processor can hold a message, then
relay it later (called store-and-forward). We might, for
instance, want to stick one out in the middle of Australia,
so doctors in remote areas can uplink particular data, or
information, or even X-ray images; then it can go to a
main centre where specialists look at it and make a
decision. And when the satellite comes over again, these
people download the information they want. So theyre the
sort of things were developing, and fast acquisitons
imperative to that.
One of the huge advantages of this system is its
autonomy. Some satellite communications or tracking
systems require a phone line or Internet connection to
provide tracking data. We might, in the long run, want to
market these technologies. They might want to use it in
the highlands of New Guinea, where its difficult to get a
telephone line. Needing Internet or phone lines for a

groundstation is a limitation in remote-area uses, plus

Name:

Keith Willey

Position:

PhD student, Faculty of Engineering, University of Technology, Sydney.

Background:

My background for about 20 years prior to joining the CRC had been in broadcasting. I
started off as a technician working with the Australian Broadcasting Corporation. After some
experience with commercial stations in northern NSW, he became a specialist broadcast
engineering officer and technical manager. His interest in satellite communications came from
its use in broadcasting. He completed his undergraduate engineering degree at UTS while
working in the ABC engineering department.
I saw the CRCSS as an opportunity to get into working with satellites, especially in Ka-band,
which I believe will play a big part in future communications systems. Keith believes Australia
can make a strong contribution to this field. If it werent for the CRCSS I wouldnt be doing a
PhD, Id be off working somewhere, but I saw it as an opportunity ... to work on something that
was actually going into orbit, and have some application. I found that particularly appealing;
its a way of doing research and attacking a project at the same time.

What next?:

Id like to think this project is viable, and will attract some sort of funding, whether from
private industry or government, to keep the research side of this going. ... If that was the case,
there could be quite a good industry or business surrounding that, so Id be interested to keep
working in that area.

SpIN 83, June 1999

 those connections already provide many

communications options. We want to have a system that
can be self-contained, stuck out in the middle of nowhere,
and go for it.
The CRCSS is a leader in this field due to the necessity of
Australian conditions and distances. But the CRCSS is also
able to investigate other Ka-band problems. One big plus
for us in Sydney, is weve got the ability to do Ka-band
studies in a real urban environment. Weve got water, he
gestures towards his 23rd-floor window, and panoramic
harbour view, which gives us reflections. Weve got
bridges, weve got buildings. So as a stage two part of our
process, after FedSats launched, we think theres a lot of
research to be done in overcoming multipathing in urban

environments. These reflections cause problems in

tracking Ka-band satellites in cities. Therell be plenty of
further research to do, and were ideally located to do
that.
The research work of UTS and their CSIRO colleagues
puts the CRCSS at the forefront of Ka-band communications technology. FedSat will be among the first
demonstrations of high-capacity, low power, low Earth
orbit satellite Ka-band systems. Later, similar Earth
stations will enable people in remote areas, all over the
world, to quickly transfer life-saving or other vital
information, without requiring any other infrastructure.
Australian research will be making a big difference to
peoples lives.

UTS in context

Auspace has no
capability to test the
antenna pointing and so
forth, so the radio
frequency signal will be
fed by wire into the test
points in the communications payload. On the
receive side, the signals
will be extracted from the
communications payload
and go back into our
Earth station.

UTS team leader, Associate Professor Sam Reisenfeld,

explained to SpIN how the UTS work fits in the
CRCSS.
The UTS groundstation will communicate with Kaband equipment on the satellite. The Ka-band
hardware is being developed by CSIRO Telecommunications and Industrial Physics [also in Sydney], so we
have a very tight working connection with them, said
Professor Reisenfeld. In the overall design, the
space-borne equipment for Ka-band is partially
CSIRO, consisting of all the Ka-band electronics,
Monolithic Microwave Integrated Circuits. But at
lower frequencies, theres a handover of electronics
on board FedSat to equipment thats being designed
and developed by the University of South Australia.
The UniSA is providing the UHF payload and the
baseband processor. In the simplest case, the signal
path will be through the UHF subsystem. However, in
an alternative regenerative mode of the Ka-band
transponder, the signal can go through the baseband
processor.
For this to work, The UTS must provide the signal
processing, and all the analogue and digital electronics and system engineering for the whole Kaband link.

Timescales

Its a fairly complicated schedule, continued

Professor Reisenfeld. Were looking at the end of
September to integrate an Earth station for the first
time, totally. Currently, theyre working on testing
the components separately before larger integration.
Then Professor Reisenfelds team will remove the
antenna-system electronics, which will be tested
at Auspace in Canberra and possibly the UniSA, and
replicate the electronics on a duplicate Earth station.

Sam Reisenfeld, courtesy UTS.

In the SeptemberOctober timeframe, well be testing various spatial

pointing algorithms, said Professor Reisenfeld.
Before the launch in November 2000, theyll continue
software improvements in both spatial and Doppler
tracking algorithms. Weve implemented these on a
very versatile, flexible, digital signal processing
board, so that during the additional year until launch,
and then after launch, we can continue to make
improvements and adjustments through software
modifications, said Professor Reisenfeld.

Value of the CRC

When I originally came to Australia in 1988, he

said, there was the high potential of a Cape York
launch complex. And at that point I took the job at
UTS jointly with working satellite research at the-then
OTC research and development lab in Sydney. Until
now there has not been the realisation of the
potential Australian space program.
Without the CRCSS, I very well may have gone back
to the US or to Europe, where satellite communications are flourishing, and there are numerous
business opportunities. But with the CRC there is
potential that has opened up in this country, and Im
very hopeful that some of those will be realised in a
major business enterprise in Australia.

SpIN 83, June 1999

Computing on
the fly

hat if new computer hardware could be

put in a satellite, after launch, as easily as
installing software? It could shorten
development times, enabling decisions on
computer requirements to wait until after
launch. And it could allow re-customisation of
old satellites for new purposes. Sounds
impossible, but FedSat will be the worlds first
test of space-based reconfigurable logic
computers. It may begin a new species of
adaptable, maybe even self-repairing
spacecraft.

HNeil Bergmann, researcher at Queensland

University of Technology, leads the CRCSS computing
project.
He told SpIN that while reconfigurable logic has been
around for twenty years or so, it hasnt reached widespread
use. The idea of reconfigurable logic chips, said Dr
Bergmann, is that you have some sort of generic logic
structure which, when you buy the chip, you can configure
to do a particular application. New circuits can be
specified by electronically setting switches.
These chips come in several versions. Certain kinds can
be programmed once by setting fuses. Another kind uses a
switch which will stay on even when the power is
removed; these can be erased up to a thousand times using
Ultra-Violet light.
The final type, Field Programmable Gate Arrays
(FPGAs), store their programming information in a RAM
cell which sets a programming switch one way or the
other. These do not hold the program without power, but
thats not a problem for the FedSat experiment. Each time
the computer boots, one can install new circuit configurations, to do whatever is necessary.
FPGAs have mainly been used as a research tool, but
now, it looks like space applications really are in many
ways more useful than the terrestrial ones, said Dr
Bergmann. Formation of the CRCSS allowed him to
redirect his research into this new and promising area.

High performance

The argument for why FPGAs are higher performance is

quite convoluted, said Dr Bergmann. Firstly you have to
realise that in comparison to a full custom integrated
circuit that does the same function, its certainly not
higher performance. Customisable circuits pay a size
penalty; it takes a hundred logic gates to build one whose
function you can program. So FPGAs are bigger, and
theyre about five times slower.
But FPGAs zoom ahead because of their parallel
processing. Despite their speed advantage, conventional
circuits can do just one calculation at a time. These chips

SpIN 83, June 1999

spend as much time moving data around as actual calculations, so up to 99% of the logic gates are idle most of the
time. With an FPGA, you can design the circuit exactly
for the algorithm you want, and get much higher numbers
of transistors doing useful things at the same time, said
Dr Bergmann. If you want to do ten operations, rather
than taking 1 piece of data and doing the 10 operations for
that, and taking the next one and doing ten operations for
that, an FPGA can have 10 separate operators. The first
piece of data comes into operator number 1, the when it
moves to operator 2 you can bring another piece in, and
operator 1 can start moving on that. These custom
circuits can be repeated almost infinitely, so although
they may only be doing clock cycles 1/10th as fast, if you
can fit 100 of those operations on your FPGA youve got a
performance factor of ten, and if youve got 100 FPGAs,
well youre up to a performance factor of 1000. FPGAs
offer much higher performance if the calculation is simple
and repetitive.
Big applications that show promise are things like
image and video processing, where you have very regular
algorithms repeated over perhaps some million pixels,
said Dr Bergmann. These calculations can be done in
parallel. For example, the standards for studio-quality
digital video are 27 million cycles per second. Theres no
way microprocessors can handle that sort of rate, whereas
thats just about the limit of FGPAs comfortable clock
speed. Similarly, data encryption and text-searching use
simple, repetitive algorithms, which benefit from the
stackable FPGAs. But so far few products use FPGAs,
partly because theyre so hard to program.

FPGAs and space

A lot of the things FPGAs are good at seem especially

applicable in the space environment, said Dr Bergmann.
On Earth, faulty equipment can easily be replaced. But not
so in space. Engineers expend so much effort making
spacecraft reliable [SpIN 80], because theres no second
chances if something breaks. It could end the mission.
Reconfigurable logic enables repair from the ground.
It also allows different satellite instruments or systems
to all use the same brain, with different circuit programs
as needed. That saves weight and space.
Even more intriguing, FPGAs might enable spacecraft to
repair themselves. Because FPGAs can change their
function, this sort of goes back to some of the early ideas
about computers as autonomous devices, said Dr
Bergmann. Computers can already modify their own
programs; chess computers, for instance, adapt their
behaviour from experience in a form of artificial intelligence. But if the programs specify hardware circuits,
FPGA computers could rewire themselves, learning and
adapting in an almost organic way. Equipped with banks
of FGPAs, and certain diagnostic algorithms, Dr Bergmann
says FPGA spacecraft could just work out what was wrong
and fix themselves. Its clearly a long way off, but its one
of the reasons people are interested in FPGAs.
The reconfigurable computer on FedSat represents a
proof-of-concept experiment. In our case, were not going
to have lots and lots of FPGAs, so were not providing a

 very high performance supercomputer in a spacecraft.

The aim is to test its reconfigurability in a space environment, and test different circuit configurations.
Eventually this will lead to a plug-in computer card with
a generic interface, for use on later spacecraft. Most
satellites carry several instruments, each with their own
separate connection to the central computer. The generic
interface Dr Bergmann and team hope to develop will
allow instruments to use the same interface, called an
Adaptive Instrument Module,
which changes on a case-bycase basis. If you want to use
the camera, you load the
interface configuration which
processes that data; when you
want to use the magnetometer or some other
instrument, you load the
configuration for that. A
standardised generic interface
could eliminate the expensive
individualised interfaces now
Dr Neil Bergmann, QUT.
used for each satellite
Photo, Wayne Deeker.
instrument.
The FedSat payload will help test some of Dr Bergmanns
ideas about configuring the FPGAs, uploading information, and error-correction. Theres a whole lot of
techniques we think we need to understand before we can
really use these things, so the FedSat payload gets to help
us understand how to use FPGAs in space. Currently it
doesnt perform a useful function for other payloads on
the satellite, for example it doesnt process anyone elses
data, so its a fairly autonomous payload.
Relying on the FPGA computer to process instrument
data would be too dangerous at this experimental stage.
We havent done one of these payloads before, so its
certainly possible that ours would fail. If that happened,
and the FPGA computer were integrated with the
satellites main computer, itd bring down everything. But
FedSats separate arrangement means the other payloads
are safe if the FPGA failed. It could just be turned off.
Dr Bergmann and colleague Dr Anwar Dawood are
designing the FedSat computing payload in conjunction
with Johns Hopkins university in the US. They have some
funding from NASA, because NASA and Johns Hopkins are
interested in using FPGAs for space applications, said Dr
Bergmann. Johns Hopkins university were able to use
some seeding NASA money to help develop the FedSat
prototype. We also have a proposal at NASA for Johns
Hopkins university to build the hardware for our flight
model, but we dont know whether thatll be successful.
If its not successful, the payload will be built in Australia.

Working with FedSat

Dr Bergman needs to test the effects of radiation on the

hardware-program error rate. FedSat provides the
experiment with just the right amount. Earths magnetosphere protects low-Earth orbit satellites from the worst of
the space radiation. So FedSat doesnt need the most
expensive radiation-proof circuits, as it would if it were a

deep-space probe. We need

to check that our
algorithms for correcting
soft-errors actually work, so
we need a bit of radiation.
Further out, FedSat might
experience radiation errors
at 100 times a second, too
much to cope with. On
Earth, an FPGA computer
might experience radiation
errors once a century: not
worth fixing. FedSat should
Dr Anwar Dawood, QUT.
experience soft-errors once
Photo, Wayne Deeker.
an hour, a comfortable rate
to see if the correction algorithms work.
But FedSat has its challenges too. Low-Earth orbit
satellites are only visible for up to 15 minutes at a time,
maybe 3 times a day, and when they are visible they zoom
across the sky [page 3]. That makes it very hard to
communicate with low-Earth orbit satellites. Dr Bergmann
says the payloads can collect far more data than ever
possible to download.
Dr Bergmann will upload the new FPGA configuration to
the satellites central computer, using the S-band
communications link [page 1]. The central computer will
check for errors, then forward the program onto the FPGA
computer. The FPGA computer can store three or four
separate configurations. If the computing payload uses one
large FPGA, it will require up to 200 kilobytes of programming data. But S-band can only do 100 kilobits per
second, maybe a megabit on a really good day. So
transferring the computing data will really challenge the
satellites communication systems.
Then theres the power problems. FedSat only has solar
panels on certain cube faces, unlike other satellites which
have huge fold-out wings. We only have about 20 watts
average power over an orbit. All the payloads chew up
power pretty quickly, and when youve only got 20 watts to
start witha mouse running in a little cage could generate
almost that muchthats a tiny amount of power
compared to what computer cards might normally use.
And were trying to run the whole satellite off that. The
other payloads, and satellite systems including navigation,
attitude correction, must all share the available power.
Theres no reason why we couldnt run our payload 24
hours a day if the satellite had the power to do that.
Its a tricky balance keeping all the payloads working
with only 20 watts. You can have more experiments on
when the thing is in sunlight ... but you cant put everything on and run the batteries flat. Batteries keep everything alive when the satellites in shadow, but if they get
too flat they may not come back on. Hence scheduling
experiment time is very important.
The schedule needs of the computing payload are very
flexible, just the more the better. Unlike the other
payloads which must come on at certain specific times
for instance the communications payload must come on
when the satellites over a groundstation, and the GPS

occultation experiment needs to catch GPS satellites

SpIN 83, June 1999

 setting over the horizonthe computing payload

requires no specific time. If it happens that no-one else is
doing anything, were happy to just switch on and run our
experiment for a while.

Milestones & future

So far weve done a design review on our prototype board,

and have decided what we expect to be in the payload.
The project is now split into hardware and software halves;
people at Johns Hopkins university have finished initial
design of a prototype circuit board. And weve just
purchased a real-time operating system and software
development tools, and are starting our software design.
Johns Hopkins university will deliver a prototype
computer, functionally equivalent to but not resembling
the flight payload, around July 1999. Success or otherwise
of the NASA funding proposal will determine options from
there.
If we had to build within Australia we might end up
with the same payload, but clearly theres other things that
we wouldnt be doing over that period. The payload
design is ready, so first comes the prototype then the flight
model. Early next year well have all that ready to deliver
for integration into the satellite.
After launch, when the payload has proven itself,
theres also the opportunity for us to offer computing
services to other onboard payloads. The magnetometer is

Reconfigurable computers
on the ground

the most obvious of those. It will be collecting far more

data than it can download, so if we can do some on-board
processing for them, to look at doing data compression
etc., then that might improve the performance of those
payloads.

Pioneer research

This is very timely research, said Dr Bergmann. Space

technology is usually behind terrestrial applications
because, you dont want do a new experimental technology in space. Normally you prove it on the ground
first. FPGAs are a fairly old technology, but FedSat will be
their first use in space.
FedSat gave a new direction to Dr Bergmanns research,
and now theres a worldwide spark of interest. It all
happened coincidentally, but its very useful for us. Thats
why NASA people are interested in this, because theyre
just starting to get into this area as well. Theres a lot of
basic research to do on how to work with FPGA computers
in space.
A lot of people work in this area, but were probably the
first mission that can upload the FPGA configurations to a
satellite in orbit.
Further information:
http://www.scsn.bee.qut.edu.au/~Reconfig/
FPGA.html#Intro

logic over the last 10 years, but without much systematic examination of the best approaches for designing
these circuits. Her two main research questions were
when is it useful to to use reconfigurable logic, so we
tried to establish video computing parameters about
when it would be useful; and secondly, if its useful to
use it for a particular application, whats the best way
to design those circuits for that application.
Dr Chung compared standard video compression
algorithms and some of her own new methods, and
examined which were the most important components
of video compression algorithms. She was able to
measure the huge performance increase FPGA
computers give over conventional chips in video
compression.

Photo courtesy QUT.

Vera Chung, Dr Bergmanns former student, is

among the first PhD graduates from the CRCSS.
Vera was working with me on reconfigurable
computing prior to the establishment of the CRC, so
her work is not in space applications, said Dr
Bergmann. Dr Chung worked on video compression.
There has been quite a lot of work on reconfigurable

SpIN 83, June 1999

The overall aim, said Dr Bergmann, is hardly

anyone uses reconfigurable computers except
experts. Thats clearly not so for conventional chips;
very few who use them need to know how they work.
That suggests theres a problem with reconfigurable
computing in terms of the ease of designing. So
eventually were trying to make them easier to use,
but the first step is to figure out how we should use
FPGA, and how could we use it efficiently.
Dr Vera Chung graduated from the CRCSS in April,
and is now a computing lecturer at La Trobe university
in Melbourne.

The space multipath

landscape

Above: GPS
multipathing.
Reflected GPS
signals cause
inaccuracies in GPS
position calculations.
Centre: GPS
multipath signals
give an intial peak,
since the nonreflected signal takes
the shortest time to
arrive, followed by
subsequent reflected
signals. Ignoring the
subsequent signals
can help mitigate the
affects of GPS
multipathing.

ignals from the US Global Positioning

System, a satellite network introduced in the
1980s, provide GPS receivers with accurate
position information. With special processing,
civilian users can approach military-level
accuracies. One of the few remaining problems
is multipath, echoes in the GPS signal. Its
complex to solve on the ground, but space might
hold the answers. SpIN examines the CRCSS
GPS multipath project.
Multipath errors are one
of the last to be removed
from GPS positioning, Dr
Rod Walker, Queensland
University of Technology
research fellow, told SpIN.
Multipathing is echoes in
the GPS signals due to
landscape features; uncertainty about the exact path
of the GPS signal creates
Dr Rod Walker, QUT.
errors in calculating
Photo, Wayne Deeker.
distance to the satellites,
and subsequent incorrect position calculations.
Dr Walker spent several years helping the mining
industry, where multipathing is a big problem. Part of this
work included predicting landscape effects, but you have
to get the terrain information very accurately before the
prediction means anything to you, he said. In mines the
landscape changes all the time, so its been difficult to
isolate the landscape influences on multipathing.
Cracking the GPS multipath problem will have major
rewards, so its a competitive field.
The CRCSS project represents a new tack, applying the
techniques to spacecraft. The beauty of that, said Dr
Walker, is you know exactly what the spacecraft looks
like, to the last millimetre, and theres nothing else out
there for the GPS signal to reflect off. So any reflections
you get at the GPS antenna are reflections from other
objects on the spacecraft. Space-based multipath research
may lead to a reduction of the problem on Earth.
GPS multipath also affects spacecraft.Despite the
reduced magnitude of multipath errors in space, theyre
still important because of the ultra-high position accuracy
certain remote sensing spacecraft need. When youre
working with science missions needing to measure the
height of the ocean to the centimetre level, you need to
know where your spacecraft is to the same level of
magnitude. GPS is a convenient way of doing that, but
because of the spacecraft multipath, you cant determine
the position of the spacecraft to that level.

10

Bottom: modelling
GPS multipath from
known surfaces.
Images courtesy Dr
Rod Walker.

Novel research

Were one of the only groups trying to solve this problem

through software simulations, and by using electromagnetic models, said Dr Walker. Other groups have tried
taking GPS receivers into the field, and constructing
reliable models from the mass of received signals, with
limited success. Its a function of so many parameters. Its
a function of satellite position at a particular instant, and
the location of the local reflectors. Its just too complex to
derive multipath-correction models from complex real
landscapes.
Dr Walkers group is trying a different angle. Weve
started by trying to predict exactly what well see in certain
environments. The aim is to develop a model of a very
simple landscape, FedSat, which can predict, and correct in
real time, the GPS multipath seen on FedSat.
Although we could possibly get away with using data
off Microlab-1, said Dr Walker, or any other mission
thats got a quality GPS receiver on board, FedSat
provides a unique opportunity to customise the receiver,
and the shape of the satellite, to exactly match the
research needs. Being involved at this much lower level
allows us to tailor where we put the antenna, tailor our
data rate, and weve got first go at the GPS receiver data.
Other missions might not transmit all the observable
parameters needed for the multipath experiment.

SpIN 83, June 1999

 But even so, FedSats not the perfect setup. The power
constraint provides a challenge. We aim to achieve the
same level of performance as a continously operating
receiver, but by using the spacecraft for short periods
only, he said.

FedSat GPS payload

After the QUT bid for a FedSat payload was accepted, Dr

Walker and colleagues investigated space-qualified GPS
receivers options within budget. NASA offered one of
theirs , which we had identified as one of the best-case
scenarios, but out of our price range. NASA basically
agreed to provide a receiver under a collaborative
agreement, so we were over the moon in terms of what
sort of receiver we were going to have access to.
NASA is interested in GPS orbit determination mainly
for the reduced infrastructure needed compared to
traditional ground-based tracking methods (such as
Doppler tracking and laser ranging, SpIN 70). Many
satellites require high-precision orbit determination from
several tracking stations around the world. But one GPS
receiver might effectively do the same job. So although
GPS is gaining acceptance in the space community as a
Precise Orbit Determination instrument, therere still some
problems associated with it, said Dr Walker. The major
advantage is that its a multipurpose instrument, you just
put one box on and it can provide a wealth of information.
The NASA receiver alleviated a major concern about
finding a good one, but introduced new engineering
challenges of how to interface it with the rest of FedSat.
Would the power connections be compatible, and would
FedSats on-board clock work with the precise GPS timing

information from the payload?

Next, one of the issues that arose was where should the
antenna be mounted. In the centre pointing upward, or
on a side face? The FedSat GPS receiver will also provide
data for another CRCSS study; Dr Elizabeth Essex of La
Trobe university will use it to measure GPS signals through
the ionosphere to construct a 3 dimensional map of that
region [next issue]. A side-mounted antenna provides
better data for that. We had to ensure the performance of
the receiver would not be diminished by the fact that we
were thinking of mounting the antenna on the side, said
Dr Walker.
Then, what sort of antenna? Elizabeth Essex would
prefer we had a large gain on the antenna, so she can
measure the satellites as they go through more and more
of the atmosphere, since GPS signals weaken with
passage through the atmosphere. Weve been given the
opportunity by JPL [NASA Jet Propulsion Laboratory,
California] to go to a higher gain antenna. Now were
talking to the platform designers in the UK at the moment,
saying this is our new antenna spec, can it fit?

CRC

Dr Walker says the CRCSS program has allowed him to

continue his specialist research into GPS multipath effects.
Its actually given that some new life by applying these
techniques to space, where theyre very well suited.
Ultimately, Dr Walker would like to develop a novel
receiver architecture based on all the studies well be doing
on a receiver in space, and implement that in collaboration
with a receiver manufacturer and test it out.
It could mean satisfying benefits for the CRCSS.

Eliminating military error

Military users can obtain centimetre-level position
accuracy, but typically the majority of other users
accuracy is usually limited to a few hundred metres.
For many civilian uses, such as GPS locators on taxis,
a high degree of error is impractical. However, a
receiver on a fixed base station can be used to
reduce the error. Surveyors accurately locate the fixed
GPS receiver; that then compares the initial GPS
reading against its actual location. The base-station
receiver constantly calculates and broadcasts the
current error-factor (which always changes); mobile
GPS receivers thus know the error at any given
moment, so can compensate to provide their true
position, within five or ten metres. The further away
the mobile receivers go from the base station, the
greater the error, since the system depends on the
base-station and mobile receivers all viewing the
same set of satellites at once. The signal paths from
each satellite must also be highly correlated.

SpIN 83, June 1999

Surveyors using other differential techniques can

reduce the error to 2 cm.
Other techniques measure the GPS carrier phase.
Phase is the arrangement of peaks and troughs
along an electromagnetic wave cycle. Continuous
observation of one satellite can measure the phasechange, thus monitor range with millimetre accuracy.
Several of these observations combined give a
position accurate to a few centimetres.
Multipathing is a major source of error in carrierphase processing. It can be reduced with a chokering angenna, which reduces reflections from the
ground and low angles. Other special antennae,
phased-array antennae, form a narrow beam at a
certain satellite; theyre complicated and expensive,
and require accurate attitude and position
information. Although useful for limiting multipath,
their cost and complexity have prevented widespead
use.

11

Space endeavour competition

The 1999 Space Endeavour Competition is now accepting
entries from school students up to year 12. First prize is a
trip to Woomera, South Australia, with one parent, to
watch a small rocket launch in October 1999.
The categories for entry are:
1. Design a payload to fly briefly in a sounding rocket
[SpIN 80], able to withstand takeoff stresses 50 times the
force of gravity. The payload must be less than 30 kg,
able to fit in a cylinder 50 cm long by 11.5 cm wide, and
retrievable intact from the crashed rocket. Credit will be
given to entires demonstrating understanding of the
opportunities small rockets provide. The design must be
realistic, because you may have to actually build it.
Entries can be in any formatessay, poster, drawings, or
modelprovided its mailable.
2. Or, write an essay on the topic: Why is it difficult to
launch satellites into space? Recommended length: 1000
words. The essay, like the payload, can be any format
sendable through the post. Judges will credit originality,
accuracy, neatness, plus an understanding of the science
involved.
Entries close 16 July 1999.
The April/May 1999 issue of The Helix (number 65)
contains more information, a list of experts willing to give

New view of Australia

CSIRO has unveiled a new view of Australia from space.
The image is a crystal-clear composite of 100 separate
False-colour composite ATSR image of Australia. ATSR data are used
courtesy of the European Space Agency. ATSR data supplied by the
Daresbury Rutherford Appleton Laboratory, UK, and the image
constructed by CSIRO Atmospheric Research.

Sounding rockets at Woomera. Courtesy Insitution of Engineers.

advice, and an entry form. Reprints of The Helix feature can

also be obtained from:
Mrs Trish Grice
The Institution of Engineers
Engineering House
11 National Circuit
Barton, ACT 2600
Ph: 02 6270 6548
Fax: 02 6273 2358
Email: tgrice@eol.ieaust.org.au
The 1999 Kistler Space Endeavour Competition is
supported by The Institution of Engineers, Kistler
Aerospace Corporation, AUSLIG, the Australia Space
Research Institute [SpIN 82], and CSIROs Double Helix
Science Club.

satellite images, each depicting approximately 500 km

square, said Dr Peter Turner of CSIRO Atmospheric
Research.
A sophisticated sensor, known as the Along Track
Scanning Radiometer, on board the European Remote
Sensing Satellite creates the images. ATSR achieves great
precision through probing each surface location twice as
the satellite passes. The first view is ahead of the sensor
and the second directly underneath. This duplication
eliminates most of the atmospheric distortions common in
conventional satellite images.
The ATSR measures sunlight reflected from the ground
and gives us really detailed information about Australias
terrain and vegetation, said Dr Turner.
CSIRO designed the instrument in conjunction with the
Rutherford Appleton Laboratory in UK, and CRCSS partner
Auspace Limited contributed to instrument components.
The new satellite measurements will help scientists
monitor the environment. The sensor measures ocean
temperatures accurate to a quarter degree. ATSR images
also allow scientists to assess cloud distribution and
vegetation health and coverage.
The composite false-colour image resolution is 4000 x
5000 pixels of 1 km square. Variations in intensity across
the image result from changing illumination and viewing
conditions between component images.

12

SpIN 83, June 1999