The Jim Austin Computer Collection

Elliott 903


The Elliott paper tape punch (left) and reader (right) Sitting on the machine, the control panel is on the far right.


The Elliott 903 machine is a transistor based computer from the late 1960's. It is made up of a number of cards each with a few gates on. The machine came from  Hull University Psychology Department where it was operated until about 1975. The machine was removed because the room was needed. The system consists of a Main processor, Control panel, Paper tape punch and reader and a ASR 33 teletype for input and output.

You might like to look at the Elliott documerntation, Software and an Emulator provuded by Andrew Herbert at (checked 31 March 2013).

The following is from the Computer Conservation web site where an emulator for the Elliott 903 can be found (

"The Elliott 903 was manufactured by Elliott Automation Limited from 1965 as a desk-sized successor to the military computers 920B (in Nimrod Mark I) and 920M (in RAF Jaguars and in tanks). It had an 18-bit word ferrite core store with a 6 microsecond cycle time, paper tape I/O and a Teletype. Up to 64K words of store could be fitted in units of 8K. The 905 was a later faster machine which could have 128K words of store.

The machine used transistors on plug-in packages. Peripherals could include a plotter, a line printer, magnetic tapes, industrial interfaces, and displays for plant monitoring. There was no disk to act as a focus for an operating system, although one based on magnetic tape was written and several for real-time applications.

Counting the military versions probably about 1000 machines were sold. The 903 itself was used in process control, for running laboratory equipment in hospitals and elsewhere, and for teaching programming in schools. Languages available included Algol, Basic, Coral, Fortran and the SIR assembler."

Recollections of Elliott 903 machines

"This was the first machine I used whilst at school (Eton
College, England). We used it to process exam results as
well as learn about computers. One fun thing was to get it
to play tunes (it had a beep). You could get a second note
for harmonies from the paper tape reader electromagnetic
clamp. Provided you switched it on and off fast enough, it
could play bass notes. However it wore the clamp out rather
quickly. I remember the service engineer commenting that
our machine blew more tape clamps than all the others he
serviced put together. I wonder why! Eventually the school
replaced the Elliott with a pack of TRS80 micros. They cost
less than one year's service contract on the Elliott. We
used to run Fortran and a made up language called 'Sympol'
(or Symbol?) on the Elliott. Another trick we performed was
to change the Sympol compiler tape with one where we had
changed the interpretation of + for - and * for /. 
Therefore all the other students' program gave incorrect
results. Happy days.

Another thing I remember about the Elliott was that most of
my programs were reset-jumped at 8181 - I'm not even sure
what that means now, and I never knew why! You had to set
the number up on the toggles on the control panel that you
have in one of your pictures. We also had a terminal that
was connected directly to the computer, so it was possible
to interact quite well with it - however I think the Elliott
acted as if it was connected to a teletype, so that the
screen updated at about 5 cps."

Richard Gibbons, January 2005


The following recollections are from Richard Blackett (June 2007)

I worked on the RAF Elliott 920B's at RAF Wattisham in the early 80's. They were used on the Nimrod it's true. As the last person trained to service to component level I fixed the air borne version but ours were originally used to control air defense systems putting track markers up for RADAR and controlling Bloodhound ground to air missiles. The 920B was particularly resistant to electro magnetic interference. Ours would sit happily through RADAR side lobes that would reset every other computer on the base. They were replaced by the 920C but were used for programme development in a language called MNAP. I was particularly impressed as they made a different "squeek" according to what process they were doing. It meant you could run a test loop programme and sing along. You would always know when it failed as it would go off pitch. Ours could be programmed to play the flight of the bumble bee and air on a G string. Incredibly impressive at the time and must of taken someone hours to do. We also had a programme called visiprint which punched holes in paper tape as readable letters. Great for Just Married etc.%0D%0AGreat to see the pictures. I've been up to my arm pits in those boards soooo many times.


It was nice to see them (903) still is existence.
The 920 was identical as far as I can see. the controller was the 903 unit as you've got in the picture. Note the screening around the cover of the processor to block external interference.
The "Network" was connected with 212 way cables about 4" in diameter and had an enormous communications rack to connect to the Bloodhounds by dedicated phone line.
The technology was a mixture of DTL and TTL. The modules made very cute key fobs. If yours doesn't work I recommend popping the cover off the core store to have a look at the ferrites. I would also, for display, take the cover off the paper tape reperforator (punch) to show the working parts. I used to teach servicing on those. If you were good you could set it so when you ran your hand down the tape you couldn't feel the holes at all.
I think ours were removed to storage in about 1983 or 84. The store in Wales had hundreds on boxes with this stuff in them. The mobile version we had sat on specialist pallets. You just dropped four sides and a top on them and fork lifted the box into an aircraft for deployment. I never saw that.
we also had a desk top board fault analysis machine. I can't remember what it was called. When I arrived at RAF Wattisham in 1980 it didn't work because (ironically) the only way to test its boards was on it. I spent 2 years, on and off, fixing it but it was very efficient at isolating, to the module, faults on the 920 cards.
Oh, and finally, because even I can see I'm getting boring, the other user who remember it mentions an 8181 to start. This was because an 8 command was an unconditional jump and 181 was just an arbitrary start point. commands 4 and 5 were addition and subtraction and 15 was the input/output set. 15 commands -  true RISC. I didn't programme it enough to remember more. I suspect 8181 is just an easy shape to key in. Ours always started at 8 so I think 88 was our start point. This could be memory is weak on these things.

Recollections from Terry Froggatt, April 2008

I worked at "Elliotts" on the 900 series, at Borehamwood then Rochester, from 1966 to 1976.

Regarding the postings on your web site, Richard Gibbons remembers resetting and jumping to 8181, but cannot remember why. 8181 was simply the address of the entry point of the hard-wired initial instructions, which I've written up on the Computer Heritage web site.

Richard Blackett's reply is not quite right. He suggests that the first 8 means jump (and that 181 was an arbitrary start point). It is true that you could obey "8 8181" on the hand keys to jump to initial instructions, but they will only work if you are on the top interrupt level. Pressing "jump" rather than "obey" ensures this. We would normally set the B and function keys up, and set all of the address keys down by running a hand across them (giving address 8191), then subtract 10 by making a V-sign gesture under the 8 & 2 keys.

Richard Blackett also says that function codes 4 and 5 were addition and subtraction. No, function codes 4 and 5 are "read memory to accumulator" and "write accumulator to memory". Addition is function 1, and subtraction is function 2. In fact, function 2 is actually a "backwards subtract" which negates the accumulator then adds the word from memory. This is more useful than subtract, because it can be used as "negate" (with an all-zero word in memory) and as "not" (with an all-ones word in memory).

This is a nice web site which lists all the 900 series machines The page lists 144 machines made. This one could be one of the Hull machines listed. I think it was delivered new to Hull University, so it may be one of the unnamed machines. This has now been confirmed (April 2008).


Recollections of Ellott machines by John Adkins

The 'GEC road traffic teletype incident'.

I had been seconded to this division of GEC for about three weeks, commissioning traffic systems, when a colleague from our field service office turned up to service a fault logged with a Teletype on one of their systems. He came over to me and asked me where the particular system was and I pointed it out to him. This chap, Arthur had an unfortunate reputation of a very low 'correction rate' of faults and this day was not to be different from many others. The first thing a FSE should do on arriving at a site is to speak to the person who called the fault in and not to get second hand info on the supposed fault. Arthur went straight over to the teletype, turned it offline and proceeded to type some characters to see if he could see what was wrong. "Ah" he said to me, "the springs in the printhead need adjusting, look, the left hand side of the 'H' is missing". He had up to this point spoken to no one on site about this machine. He took the cover off the
 machine and got out some spanners to adjust the printhead springs, at this point someone came over and asked what did he think he was doing, Arthur explained that he was from GEC computers and was fixing the Teletype. The irate customer explained that as far as he was concerned there was nothing wrong with the mechanics and the fault was in the interface in the 903, anyway the Teletype was under contract to another company anyway. Arthur explained the slight problem he had found and said he would tweak it and then sort out the interface, I slipped away back to my commissioning job. The customer let it be known that he was not at all happy with Arthur and stormed off.

Some minutes Arthur came over to me to say that he had made the problem worse and to compound this he had dropped his flat spanner in the works of the Teletype and could I have a look to see if I could spot it. The internal workings of a Teletype were a complication of electro-mechanical workings and I could not locate the spanner among the confusion of levers etc. I suggested we tilt the whole machine on its stand gently and maybe we would be able to locate it by either the sound of it moving, or some thing like that. Increasing tilting from side to side failed to shift the spanner. Although I had worked on Teletypes for about three years up to this point, I never realised that he mechanism was not bolted to its pedastal but merely sat in tapered rubber mounts. Arthur keen to retrieve his spanner was insisting on more and more angular tilting till at last the whole mechanism came off the pedastal and crunched onto the floor. One corner of the Teletype
 mechanism looked a sorry sight of mangled metal and the cabling was torn from its connectors.

As we surveyed the sorry sight, Arthur standing there in disbelief, in came the customer on cue, I slipped away quietly, the customer launched into Arthur unquietly (doesn't seem like a correct word).

Arthur phoned the office and a replacement Teletype was installed until the customers one was fixed.

Arthur was distraught and I sympathised as best I could, inside I was bursting to laugh out loud but I think I managed to muster an acceptable look of concern externally.

I cannot remember much more about the incident but I know Arthur was given a few days off and returned to work to perform an office job. I would often just burst out laughing as I recalled the incident as the Teletype mechanism performed its lazy arc from pedastal to concrete.


The 905 on board ship incident.

This occured when I was working out of the Bristol office and I had gone down to a naval dockyard somewhere with my boss Bob to perform a routine service on a 905 system. As there were dozens of small printed circuit cards in those days, with each card having about 70 edge connectors, intermittent contacts on these edge connectors formed quite a chunk of faults. The way to find a dodgy connection, whether edge or solder joint was to run on of the test programs whilst running say a screwdriver handle along the faces edges of the cards in the racks. As the program would make a noise from the speaker, the program halting could be used to locate the dodgy card as it was struck (gently of course!). The gold plating on our PCB's was very thick compared to other manufacturers and all engineers carried a green ink rubber to clean up the gold contacts if grubby. This was frowned upon by management, quite correctly as it depleted the gold layer, but continued in

This day at the dockyard the naval engineer commented that one of the system power supplies was playing up and it would trip out with an over voltage error. Bob slid out the offending power supply and pulled out the PS control card, the edge connector was very dirty and Bob got out his green ink rubber. "Here's a good tip" he advised the naval engineer, "get yourself a good rubber and buff up the edge connectors like this", Bob started a furious rubbing with the rubbber, the edge connector turned from its intial grubby appearance to a fleeting burnished gold, quickly then to a shiny copper colour. This was greeted with Bob stopping abruptly mid rub and a stoney silence, "hmmm we'll have to get you a new power supply ASAP I think", was Bobs comment after a long pause.

The whole cabinet was in an awful state with a thick layer of fine dust covering everything, "have you got a vacuum cleaner" asked Bob, "no, but we do have a blower" replied to NE. The NE went out of the room and returned with a battered blower that resembled a massive hair dryer, "careful with it, there's no guard on the impeller" warned the NE. The next few moments were quite eventful, firstly the dust cloud formed by this blower was in itself comical, but when Bob tie was sucked into the impeller and the blower very quickly pulled up to his throat, the laughs soon stopped. The knot of the tie had jammed the impeller and Bob was croaking out attempted screams, we quickly cut the tie from round his neck and Bob sat down to recover. Luckily the tie didn't get very tight and he was uninjured physically, mentally he took about half an hour before he could start work again. We teased what remained out of the impellor to reveal a severely chewed up tie.

The rest of the service went without incident.


The Portugese road traffic system interface.

I wasn't involved in this one, it was our boss at the time, Pete who was called out to a newly installed 905 traffic control system that just wouln't run. My memory is a little hazy when it comes to the details on this, but you may be able to sort out any discrepancies in the details.

The 905 control switch could be put into 'Auto', 'Test' or 'Manual' positions and systems that were used for only one program, like a traffic control system would be run in 'auto' mode as the program would be permanently resident in memory. The customer had had two engineers trained up and the system had worked for a few days and then had stopped and refused to start up again. Their engineers had struggled to rectify the fault without sucsess and had finally contacted GEC to send out an engineer.

Pete arrived on site and was ushered into the control room where all the diverse customer groups were huddled individually. Pete went over to the machine, now surrounded by the whole crowd of customer folk, switched it into manual, set the switches to 8181 and pressed the 'reset' the 'start' buttons (is that what we did?) and the system burst into life. The crowd of people moved away from the machine into their bemused separate groups again and Pete stood all alone, himself bemused. He couldn't remember where the switch had been set before, nor could anyone else. He stopped the program ran all the tests and could find no fault. He tried the 'auto' start key position, I think auto start works from power on, but I'm not sure.

Someone had apparently set the switch to manual or test and power cyclying the system had not intiated a program start and that was all that was wrong.


The BR rail indicating system problems.

My first ever solo fault to try to sort out was an 'in house' fault on a machine that was being built for British rail and was a 905 based system with a number of cabinets each with dozens of mini CRT screens, each displaying four character indicating train positions, or something like that.

Nothing would work! A quick simple operation on the control panel showed loads of errors, I put a card on an extending card and diagnosed about 10% of the chips blown. We were told on the 905 Field Service course that 99% of the time there would be only one fault on the system at a time, so I was totally bemused by this situation and called the office for help. An experinced engineer came out and came to the same conclusions as me. He conatacted the engineer that I had seen and started digging deeper than I had. It transpired that the large earth straps linking all theses cabinets had been removed to allow the cabinets to me moved around and they had decided to reconnect the straps with the power on. When they finally connected the last strap to the computer cabinet there had been a massive spark and the computer ha powerd down. We saw the braided earth strap and there was a blackened melted part to it, this spike had blown dozens of chips inside the

The two of us took two days to replace all the blown chips. I often wonder if that system was any good after that and if other chips were damaged and would have a shortened life.

Here is some interesting footage on an Elliott 905 from Nigel Williams

He says

"A 10-second video clip featuring the Elliott 503 which was operating
at the University of Tasmania as part of the joint Hydro-Electric
Commission / University of Tasmania Computing Centre (HUCC). The film
"Life in Australia: Hobart" is from 1966, the Elliott 503 was
commissioned in 1964, it was likely only two years old at the time of
the film.

The clip opens with a view of University students in a physics
lecture, a close-up view of the oscilloscope, then transitions to the
Elliott 503 sequence. The Elliott clip starts with a close-up of the
8-channel paper tape reader, then the camera draws back to view the
operators console with female operator standing nearby. In the
foreground are the two tape readers, followed (obliquely from the
viewer) by the console switches and keyboard, console typewriter (an
IBM Model B modified electric typewriter) sitting behind (and
partially obscured by the console), tape punches to the rear (closest
to the window blinds) and the engineers' display panel in the far back
left. One of the Elliott main cabinets is partly visible at the far
left rear. Tape is running through both tape readers at high speed
while the operator watches. Next is a close up view of the console
switches, where the bit 39 switch is depressed and the clip finishes
with a close-up of the tape reader reading a paper tape.

A 1970s (possibly 1974) view of the same machine is here:"

Some user history of 903's from Andrew Herbert (June 2012):

I used the 903 in three ways:

  1. As a sixth former, my school's computer club has use of the 903 at a local technical college via a postal service.  Initially we were taught FORTRAN II and I then progressed to Algol 60 and SIR.  I was privileged to be allowed to deliver the club's coding sheets to the college and collect the previous week's output on games afternoons - my contribution to the school's sporting prowess being dismal.  The college let me operate the machine and pour over the documentation much of which I laboriously copied by hand and still have.
  2. My father worked for Marconi Elliott Avionics in Rochester and on the strength of my programming skills got me summer jobs there.  The first year I was an operator for the central computing service run on an Elliott 503 and also got to experience an 803 plotting engineering calculations.  When one of the engineers discovered I could program, I was moved to work on the 920B (military version of the 903) and the 902, the 903's 12-bit cousin.  In subsequent years I also used GEC 2050' s and 4080's to drive test rigs for avionics with various kinds of 920 in them.
  3. I did my undergraduate degree at Leeds where they still had a 903 that had previously been used for student teaching: that job had moved to a KDF9, the previous central mainframe, which had just been replaced by a 1906A.  I essentially had the use of the 903 as a personal computer but didn't do much with it, being seduced by the novelty of the KDF9 machine architecture and the chance to explore Algol 68 on the 1906A.

It being such a simple machine, I've retained a surprising amount of knowledge about it over the intervening 40 years.  I got to know Terry Froggatt through playing with the 903 simulator he and Don Hunter put on the BCS Computer Conservation Society (CCS) web site - I re-programmed this in F# as an exercise in learning this new functional language developed by colleagues at Microsoft Research.  It's grown into a full simulator of the entire range with a lot of additional support for handling the various Elliott paper tape formats and low-level monitoring and debugging.  I had encountered Terry's name when I worked for Marconi Avionics, as a lot of the utilities I used were authored by him.  Terry has been feeding me digital copies of tapes from his archive and I know have just about all the issued 900 software catalogued and demonstration programs / examples running on the simulator.  When Don intimated to Terry that he wanted to dispose of his machines, Terry kindly offered me first refusal.  The machine fortunately has been reasonably reliable although I've had a few issues arise that have taken time to deal with, and Terry's help has ben invaluable and the 903 at TNMoC is near enough for me to swap cards etc to confirm problems when the do arise (and in return I've been feeding them demonstrations they can run to amuse visitors).

The strangest thing about returning to the machine after 40 years if the satisfaction of filling gaps in my knowledge: there were things in the documentation I didn't understand or get to try as a student, and now, happily retired I've been able to pick up the threads again and resolve a lot of unfinished business.


More pictures

Picture of the processor cards


Picture of the front panel