The Leicester X-ray Astronomy Group has flown instruments on many spacecraft, but Ariel-6 (or Ariel-VI) is probably the least-remembered of all.  It was indeed something of a failure, especially contrasted with Ariel-5 which was such a great success.  But it was not a complete dud, despite its troubles in orbit, and in the end its data allowed us to publish a few papers.  Since the British taxpayers spent something like nine million pounds on the entire project, and I spent a lot of my time on it for several years, I think the story is worth telling, if only as an awful warning.  This account is mainly drawn from my own recollections, supplemented here and there from the few documents from that period which I can still find.

Ariel was a series of scientific satellites funded by the Science Research Council (a forerunner of STFC) which were designed and operated from the UK but launched on American rockets.  This reliance on US launchers was unavoidable after the only project to develop a British launcher was canceled just a few days after its first successful satellite launch.  From then on the UK sat firmly in the very last place in the world league of space-faring nations.  The table below shows what the Ariel satellites studied and when they were active:

Ariel-6 marked the end of the era: after that the UK decided to have no more space science missions unless they were led by other countries or by the European Space Agency.  The earlier spacecraft carried some US instruments in exchange for which the UK got access to US ground stations, but Ariel-6 did not.

Although Ariel-5 had not been launched when these instruments were being designed, it was known that Ariel-5 would not be able to study time variations shorter than one orbit – about 90 minutes.  There were good reasons for thinking that at least some X-ray sources might be variable on time-scales of seconds or even milli-seconds, for example pulsars or close-binaries, especially binaries involving black holes.  A study of these fluctuations would give some valuable information on the nature of the sources, so this was the main aim of the Leicester instrument.  The proportional counters also had some useful spectral resolution, enough to detect lines from Iron and perhaps a few other elements in the strongest sources.

Apart from rather serious constraints on size, mass, and power, the X-ray telescopes had to contend with many other limitations on this spacecraft.  The Cosmic Ray detector was omni-directional, but the X-ray telescopes needed to be pointed at specific targets.

The most serious constraints, however, arose from the decision to use just one ground station in order to reduce running costs.  This was at Winkfield in Berkshire.  To come within range the orbit had to be inclined at 55° to the equator, which meant that for around 30% of each orbit the spacecraft was in high-background regions near the magnetic poles and it also passed through something called the South Atlantic Anomaly, where X-ray observations were also impossible.  Ariel-5 was launched into a much more favourable equatorial orbit, indeed I think Ariel-6 is the only X-ray observatory ever launched into a low-earth orbit at such an inclination.

With only one ground station, it was also necessary to store data on board for periods of up to 16 hours.  This was done using a pair of tape recorders, but the limitations of space-worthy tape recorders meant that our telescope could only record 2 bytes/second.  In order to collect at least a limited amount of high-time resolution data despite the extremely limited data rate, the electronics module include some special modes with considerable data compression.  It could, for example, overlay data from a pulsar of known period, or accumulate autocorrelation functions or histograms of intervals between X-ray photons, both of which might allow periodiciities to be detected.

The Leicester X-ray telescope was largely constructed within the Group by John Spragg, Dave Watson, and their colleagues.  I was not involved in this at all, and indeed much had been done by the time I joined the project at the end of 1973.  The technical difficulties were considerable: the four proportional counters had titanium bodies, with thin beryllium windows to admit the X-rays, and an indium seal between these two metals.  These were filled with Xenon gas, and to cover the wide energy range there were two separate detection chambers.

The electronics module was produced by Messrs Pye of Cambridge.  This was a miracle of miniaturisation as it contained all the electronics needed to drive the proportional counters and interfaces to the spacecraft but also what amounted to a special-purpose computer with 8k bytes of memory.  The whole thing was not much bigger or heavier than a hardback book, and consumed only 2 Watts.  This at a time when our smallest mini-computer (also with 8k of memory) was the size of a small filing cabinet, weighed about 50 kg, and used a few hundred Watts of power.

I was, however, much more involved in testing the instrument and its electronics.  Chris Whitford took the lead in this, as he had been already involved in the project for some time when I joined the group, but there was easily enough work for two.  We first had to test the instrument in our own laboratories, simulating X-ray sources using a variety of small radio isotope sources.  We also spent some time at the premises of Pye in Cambridge, and then later of Marconi Space and Defence Systems (MSDS) in Portsmouth.  MSDS had been chosen as the main spacecraft contractor, and it was in a large clean-room in Portsmouth that we had to demonstrate that our instrument conformed to all the interface specifications and then to test it while connected to the spacecraft itself.

In order to test such a complex instrument and electronics package it was obviously necessary to use a computer, but at the time computers tended to be expensive things around the size of a small room.  Following on from experience with Ariel-5, two PDP-8 mini-computers were purchased for the Ariel-6 testing programme.  We had two so that there was some redundancy, and also because it was sometimes necessary to run tests on the flight model and flight spare in different locations.  The main part of each PDP-8 consisted of a box about 50 cm wide, 35 cm high, and nearly 100 cm long.  It also had a teletype console, a paper tape reader and punch; later Chris managed to get a VDU interfaced to it so we could edit and run programs and examine results on a screen.  This was luxury, compared to the noise and slowness of a teletype.  The PDP-8 processor could execute about a million instructions per second, and had a memory of 8k words of 12-bits.  Floppy discs were in their infancy, and hard discs huge and impossibly expensive, so there was no non-volatile storage at all, except for paper tape, of which we used large quantities.  An electric paper-tape rewinder was an essential part of our kit.  With care, all of these bits would just fit in the back of a large hatch-back or estate car.

The next problem was the software: with only 8k words and no disc storage, there were no high-level languages that we could use on our PDP-8s.  While we were still pondering the awful prospect of having to program everything in PDP-8 assembler code, Chris Whitford read an article about a new language called Forth invented by a couple of radio astronomers at NRAO.  They were actually selling Forth systems for various mini-computers, but at the time this did not include PDP-8s unfortunately.  All the same, the idea was sound.  The language was rather simple and designed for use on computers with limited memory: for expression evaluation the programmer had to use post-fix notation (sometimes known as reverse Polish).  This was very familiar to both of us as enthusiastic users of Hewlett-Packard calculators, which use the same notation.  Thus to evaluate an expression such as:

Incidentally, the PostScript language later widely used to drive laser printers also uses post-fix expressions (but not many people get around to writing their own PostScript programs).  Another unusual feature of Forth was that it relied upon two interpreters: a high-level interpreter took the source code written by the human and turned it into a low-level code, which was then interpreted by a very fast and much simpler interpreter which carried out the required machine-level operations. The Java programming language invented much later was based on a similar idea.

Pye versus MSDS

We had rather contrasting experiences of working with the two industrial companies involved: Pye at Cambridge who built our electronics unit, and MSDS at Portsmouth who built the spacecraft.  I think that Messrs Pye were already part of the Philips group, while MSDS were part of the GEC-Marconi empire, so in principle they were both smallish cogs in large wheels.  We always got on extremely well with the small team, maybe a dozen people, who formed the Space Electronics Division of Pye.  They were excellent electronic engineers but also very flexible and pleasant to work with, with a complete lack of formality. Things were very different at the MSDS works in Portsmouth.  Their whole system was extremely formal and bureaucratic, and the general assumption was that any tiny deviation from the specifications or laid-down procedures had to be paid for by somebody as an additional cost on the contract.  There was an SRC liaison person more or less permanently on duty at Portsmouth to try to reduce friction between MSDS and the visitors from the various universities supporting their instruments.  A couple of incidents that I still remember show the differences in attitude.

When we first tried to connect our instrument to the spacecraft it was found that a pair of pins in one of the connectors was wrongly wired up: obviously Pye’s responsibility.  After some frantic phone calls, Pye’s said they could make up a replacement set of cables that same day, but it turned out impossible to get them finished in time to catch the last possible courier service that would get them to Portsmouth by the following morning.  I was due to go to Portsmouth myself the following day to take over from Chris, and after some more hasty phone calls, the problem was solved like this.  The head of the division at Pye, Don Weighton, suggested that his wife bring forward a planned shopping trip to London and take the new cables with her first thing the following morning; I would meet her at the ticket barrier at Liverpool Street Station when the train from Cambridge arrived, before going on to Portsmouth. This was before the days of mobile phones so such a rendezvous between people who had never met before had to be made specifying an exact point in space-time.  Things might have gone smoothly except for the fact that, as I discovered when I arrived, the Cambridge train was due in on the only track at Liverpool Street where there were two platform faces, so that passengers could alight from either side.  These two flows of people were then fed to two separate sets of ticket gates, each set invisible from the other.  In the end Mrs Weighton and I did manage to find each other, and I got the cables to Portsmouth just in time to stop MSDS claiming that we had seriously delayed their integration schedule.

Somewhat later on it turned out that MSDS had made a mistake in interpreting the interface specifications so that one of their tests of our instrument incorrectly reported a failure.  They admitted the fault, but claimed that it would take many weeks to get their test software modified to run successfully, because the modified software had in turn to be tested and validated, documented, and then signed-off by a whole host of quality assurance inspectors.  The whole thing was a bureaucratic nightmare.  It might easily have delayed the launch and cost everyone a lot of money.  The agreed solution was for Pye Electronics to alter our instrument so that it produced outputs that the MSDS software was expecting.  Altering hardware to suit faulty software is not something that we were happy about, but Pye’s managed to do this in just a few days, averting the crisis.

It was not until many years later, when I read Peter Fowler’s obituary, that I discovered that he was the grandson of Ernest Rutherford.  His father, Ralph Fowler, had been a research student and later staff member of the Cavendish Laboratory in the 1920s, and had gone on to marry Rutherford’s daughter.  I wish I had known at the time, as I’m sure his recollections of this father and grandfather would have been fascinating.  But Peter was an extremely modest man and never revealed his dynastic connections.

In parallel with tall this we had to prepare suitable software for data reception, reduction, and analysis, and get hold of hardware to run it on.  The original plan had been to use the large PDP-8/e system already in use for Ariel-5 data analysis, but as time went on it became more and more likely that Ariel-5 would still be operating after Ariel-6 was launched, so that a separate computer would be needed.  Fortunately we were able to get funding to purchase a PDP-11/34, almost the smallest member of the newer PDP-11 series, complete with discs, magnetic tape drive, paper tape reader and punch, and modem (which operated at the then maximum speed of 1200 bits/second).  The really revolutionary feature was that its had a multi-user operating system, RSX-11M, which allowed several of us to use it at once (the PDP-8 could only have one interactive user at a time).  Over time we added more and more “glass teletypes” in the form of VT-52 terminals.

We then had to develop a range of programs for data reduction and analysis.  The data reduction tasks included accepting the data transmissions from the modem and extracting the scientific data from the telemetry blocks.  Analysis software included routines for spectral and time-series analysis.  Because the spectral capabilities of Ariel-6 were so different, and the time resolution much better, than on Ariel-5, we were not able to use much of the existing software so had to write much of it from scratch.  Martin Ricketts, Chris Whitford, and I wrote most of this, but of course we had help from several other members of the group.

One important program that I wrote was that for receiving data from our modem – computer networking was almost unknown at the time, and the only code we had to go on was the similar program for Ariel-5 written in PDP-8 machine code.  I re-wrote this in PDP-11 Fortran, but was able to include a few small enhancements.  When the modem was idle it was possible to telephone it from home and receive a series of beeps which told you whether that day’s data  transmission had been received and processed successfully or not.

Ariel-6 was launched on 2nd June 1979 on a Scout Rocket from a NASA launch facility at Wallops Island.  We were pleased that the launch went almost flawlessly, and that Ariel-6 appeared to be working well once it reached orbit.  It was necessary to check it out carefully, and the high-voltage supplies to the scientific instruments could not be switched on for a few days to allow for out-gassing of the hardware now in a near-perfect vacuum.  Almost at once, however, the spacecraft controllers found that a problem: the spacecraft modes were changing from one orbit to the next without an explicit command.  I have a copy of a press cutting dated June 4th referring to this as possible interference and a “minor problem”.  It was, unfortunately, a sign of worse to come.  The next few sub-sections list all the various problems with which we struggled to cope over the next two years.  It is not easy to put them in order of importance, but their combined effect was to make it extremely hard to make useful scientific observations with the X-ray telescopes.

The spurious command phenomenon was intermittent but usually affected Ariel-6 at least once or twice a day.  The tape recorders were not generally affected by the problems so the times at which the instrument changed state could generally be determined after the data for a day were sent back to the ground station.  As time went on it was possible to map the places in orbit, or relative to the Earth beneath, where these spurious events were most commonly encountered.  It became clear that problem was not related to the position of the satellite in space, or the passage through the zones of charged particles around the magnetic poles, or to the sunlight/shadow boundary, but only to the position of the satellite over the Earth.  The areas where we received most of these spurious commands were over the Soviet far east, and over Eastern Europe.  A few enquiries revealed these as known locations of high-powered Russian radar systems known to NATO as “woodpecker” because the pulse repetition rate made the signal sound, on an AM radio, something like the noise of a bird drumming on a tree trunk.

The parts of the spacecraft most frequently affected by this were the high-voltage supplies to the proportional counters.  This is probably because for safety reasons these were controlled by what was effectively three switches in series: all had to be on to allow the high voltage through.  This mean that if switches were toggled at random, seven out of eight times the effect would be for the high voltage to be disabled.  It turned out that the Bristol Cosmic Ray instrument was least affected, and the MSSL telescope most affected by this.  Perhaps it was something in the bit-pattern of the individual commands which made random pulses generate some commands more often than others.  Eventually there was an investigation of the spacecraft electronics and an attempt to simulate the problem on the ground: this gave us a possible explanation: the solar panels were probably acting as antennas of the right size to pick up these radar signals and inject pulsed interference directly into the spacecraft DC supplies.  There was, of course, very little that we could do about this.  The orbital altitude and inclination meant that there were about 15 orbits per day, only 5 of which passed within range of the ground station at Winkfield.  The spacecraft controllers were instructed to check the state of the instruments whenever possible and switch on the high voltage supplies every time there were found to be off, but this was of limited value.

The MSSL group, with their telescope suffering most, went further.  They developed a small portable radio transmitter working on the appropriate frequency which would send off “high voltage on” commands whenever necessary (I forget whether it worked all the time, or only at times when the spacecraft was actually passing overhead).  They worked out that most orbits that passed over Russia also passed over Australia later on the same orbit.  Through their contacts with the Anglo-Australian Observatory they arranged for this transmitter box to be shipped out there and plugged in.  This worked for the remainder of the operational life of Ariel-6 and significantly alleviated the problem, but obviously did not solve it.

Besides identifying the sites of Soviet radar stations, we had one other insight into the secret world of the military.  In late September 1979 we noticed a sudden increase in the charged-particle rate experienced above the southern ocean, roughly off the coast of South Africa.  This declined over the next few weeks and we didn’t take a huge amount of notice of it as we were still learning how the background rates varied around the orbit.   It was not until later that we came to associate our South African background anomaly with what is now known as the Vela Incident.  On 22nd September 1979, one of the Vela satellites run by the US Defense Department detected what was interpreted as the flash from an atomic explosion: this was thought to have been the first (and I think only) atomic test by South Africa, perhaps with the help of Israel.  But Vela flashes could also come from natural phenomena such as lightning and the Wikipedia account of the incident seems to suggest that there is still some doubt about whether this atomic test actually took place.  I think our Ariel-6 data confirm it pretty well.

Another serious problem with Ariel-6 was that the thermal design was defective and the whole spacecraft was  in continual danger of over-heating.  This almost certainly led to the reduced life of the tape recorders, and the reduced capacity of the batteries.  In an attempt to reduce the solar gain, the spacecraft controllers reduced the maximum offset of the telescopes from the anti-sun direction was reduced, further restricting the area of sky that the X-ray telescopes could access.

As with practically all spacecraft, during the sunlit part of each orbit the solar cells charged batteries so power was maintained all round the orbit.  These were Nickel-Cadmium batteries.  It was by then well-known this type requires careful design of charging circuits otherwise they exhibit something called the “memory effect” in which the appear to lose capacity. In fact their capacity is not changed much although their terminal voltage may be depressed after a sequence of partial charges and discharges; if the equipment is not designed to cope with the reduced voltage it appears as a loss of capacity. Unfortunately the charge/discharge circuitry on Ariel-6 was not designed to take account of this effect and the minimum voltage threshold was set too high, so that the power was often disconnected from the instruments quite unnecessarily as the cells discharged normally.  Power could, of course, be restored during the next passage over the ground station, but all the same we lost a considerable amount of observing time from this defect.

Ariel-6 did not carry star trackers to measure its absolute pointing position as did virtually all subsequent X-ray observatories, only Sun sensors, but the Sun is much larger in the sky than any other star, so the measurement accuracy was not all that good. This meant that it was difficult to determine how far off-axis the target of each observation was, and this prevented us from determining absolute fluxes of X-ray sources.

The spacecraft was spin-stabilised with an initial spin rate of around 60 revolutions/minute, though this slowly decreased.  The idea was that the X-ray telescopes pointed along the axis of the spacecraft, so that the rotation would be no problem.  Unfortunately attempts to balance the spacecraft on the ground proved to have been inadequate, as when it reached orbit the spin axis was found to be more than a degree away from the telescope axes (one degree was the maximum value allowed in the spacecraft specification).  Our telescope relied upon a honeycomb collimator which had a fairly flat response over the central area of the field-of-view, but it decreased sharply to zero further out.  As a result of the telescope axis described a small cone about the centre of rotation and all the flux coming from an X-ray source was strongly modulated with the spin period.  We expected some modulation and had software to detect it and remove its effects, but the effects of high-coning angle and poor attitude measurement together meant that in many observations the collimator transmission went to zero for part of each spin cycle, so that no correction was possible.  This seriously decreased our ability to perform high-time resolution observations of strong X-ray sources.

Ariel-6 was equipped with two tape recorders for partial redundancy.  These were continuous loop machines so that new data was simply recorded over the top of the oldest data that had been recorded earlier.  Each recorder had a capacity to record the standard data stream for around 18 hours.  After only a few months in orbit one of the recorders gradually became unreliable and then failed completely.  This meant that the one remaining tape recorder was only just capable of recording for the period when the spacecraft was out of touch with the ground station, and then its contents had to be transmitted to the ground at once to avoid useful data being overwritten.  This seriously complicated spacecraft operations, and often some sections of recorded data were lost because of these problems.  Unscrambling the data blocks that were recorded, with partial overlaps, was also much more complicated, and a some of our software had to be re-written to cope.

Control Centre Move to Chilton

After about a year of operations from Appleton Laboratory in Ditton Park, using the ground station nearby at Winkfield, the authorities decided to merge Appleton Laboratory with Rutherford Laboratory at Chilton just south of Harwell.  Part of the deal was to stop using the Winkfield ground station in favour of a dish that was already sited at Chilton. The aim was, I suppose, to reduce costs. The control centre staff (Alan Rogers, Tony Lowe, Jock Gourlay, John Wright, and others whose names I have forgotten) coped remarkably well with this transfer, and operations were hardly affected. Having the ground station under their direct control made for increased flexibility, but also a heavier workload.

Up to that point the regular method of scheduling observations with our instrument was to write down lists of commands on paper, specifying the times for observations, calibrations, and so on, and send these lists to the control centre by telex, or fax, or even through the post.  The control centre staff translated our wishes into the sequences of binary commands and transmitted them to the spacecraft during a convenient ground pass.  During a visit to the new control centre I noticed that these sequences were generated on one of their computers and then transferred to 5-track paper tape, which was then fed through a reader which sent telecommands to the spacecraft.  It turned out that they would be very happy for us to do this command-generation ourselves.

The system I designed worked as follows.  I wrote some software for our PDP-11 to translate mnemonic instructions into the sequences of binary commands that the spacecraft understood, and then we loaded 5-track paper tape into our paper-tape punch, and punched off the results.  The program preceded this command sequence with a standard header and trailer to be read by the human telex operators using 5-track Telex code, the middle section was intelligible only to Ariel-6.  We had promised that our software would only generate commands that affected the Leicester instrument on the spacecraft and they trusted us. So each day we dialled up Rutherford Appleton Laboratory on our Telex machine, and transmitted the commands on the paper tape. The operators there were instructed to switch on their own tape punch, and pass the resulting tape to the Ariel-6 Control Centre.  The central section of this tape could then be used to send commands directly to the spacecraft when it passed over.  This worked remarkably well, reducing the risk of transcription error, and saving effort at RAL.  There was only one problem: the Telex operators at RAL complained that our telex transmissions looked like gibberish – indeed expressed in Telex characters they were, and that it was against the Wireless Telegraphy Acts to send Telex messages in code.  I think this provision dates from the first World War and as far as I know is still in force.  We eventually managed to convince them that our messages were not in code, in the sense of being encyphered to ensure secrecy, but merely in a sequence of characters that only Ariel-6 happened to understand.  After the situation was explained, we got their reluctant cooperation, but whether they really understood what we were doing I don’t know.  If only we had been able to use computer-to-computer networking, but software to do this was in its infancy and it would have taken too long to develop such a solution ourselves.

Science

Despite all the problems, we did eventually get a bit of observing time on a few objects of interest, and managed to analyse the data and publish the results.  A list of published scientific papers is given in the appendix.   One of the few significant results that I can remember from that time was that our observations of millisecond fluctuations in the X-ray flux from Cygnus X-1 showed a small lag between the low energy and high energy bands, as predicted by the model produced by Martin Rees.  On it own, given the difficulty of making the measurement and the rather modest statistical significance,  this was hardly enough to prove that it contained a black hole, but was one more piece of evidence that eventually led to that conclusion.

Conclusions

In retrospect we can see that Ariel-6 was a seriously limited platform for an X-ray telescope, but we had no reason to expect that it would have so many operational problems that it would not be able to do much science.  It is hard to avoid the conclusion that this was the result of a number of separate design failures.  The main contractor, MSDS, had been chosen by an old-established technique generally known as “Buggin’s Turn” – the only other large company in the UK capable of building spacecraft was British Aerospace, and they were out of the running because they had been chosen for Ariel-5.  Given the complete absence of competition it is hardly surprising that MSDS didn’t have much incentive to do a good job but I expect the failures also arose from the general lack of experience in building spacecraft at that time.

As it happens I had an earlier encounter with another company in the GEC-Marconi Group. Before coming to Leicester I worked for a couple of years at the Royal Aircraft Establishment in Farnborough.  One of the responsibilities of the department that I joined was to specify and procure a new avionics system for a type of RAF plane (because of the Official Secrets Act I am probably not allowed to reveal more, even though it was a long time ago).  The manufacturing contract was put out to tender and three companies bid for it, after which my colleagues and I did the technical assessment.  I think it was a fixed-price contract, so the decision ought to have been made purely on technical merit.  Company A we rated as very good, and company B was nearly as good. Company C we thought was extremely poor: their central development was around a computer series that was many years out-of-date.  It would be very hard to program more or less guaranteeing inflexibility and lack of performance.  There was no doubt about this – every one of us rated company C as providing an unacceptable solution. Naturally this was the company that was chosen.  We later heard a rumour that the Minister had over-ruled the civil service advice because the company had said that if they were not chosen there would be many redundancies in two government-held marginal constituencies.  You may not find it hard to guess which industrial group company C belonged.

Things must be very different now – after all we have ESA’s system of juste retour which isn’t at all the same thing as Buggin’s Turn, is it?