An audio file of an reminiscence interview with Paul Kelley's boss John Wheeler, who worked with LEO I. Recorded on 11 September 1978

Transcript :

John Wheeler Interview by [Interviewer Paul ...] at [:Location] on [Date]

Transcribed and edited by Jon Hales, volunteer at CCH.
[Editor comments in square brackets]

[Background: role of John Wheeler at J. Lyons Co ???].

[Abrupt start to recording .... No questions were asked. The 'interview' is a narrative by John Wheeler.]

INT: [JW]:[...] the tape it stops with a loud click so that's very reassuring.

JW: Yeah [um] [...] Joe Lyons had all these tea shops in London. They started them early in the century and they were really the first places that young ladies could go into unaccompanied by gentlemen and not be molested. Those were the days.

Anyway they had all these shops [...] and they sold cups of tea and buns and everything else in them. And they had also what they called a front shop in each one which sold over-the-counter for taking away various baked goods; bread and Lyons's cakes and pies and goodness knows what else.

And the computer [Leo 1] was first used in fact to produce the packing slips for each of the 180 odd shops in London [check: also for tea shops elsewhere in UK?] for these goods for the front shop [check: also for the restaurants]. What happened was that there was a vast [Umm] telephone room with about 30 telephones in manned by girls. And each tea shop at a prescribed time would ring this magic number and say what it was they wanted, like 200 loaves and 20 Dundee cakes and so on and so on and so on. [Umm] This would be orders for the following day. [Umm] The [uh] orders didn't cause the goods to be made because they were making a certain quantity. I don't know what they did with the rest; gave them [to] the pigs I suppose. But they ... they were ... it was not controlling the baking of the goods, it was controlling their packing and distribution. Each of these orders was then punched up on a ... on a ... on a Hollerith punch card; you know the ones with the square holes in. And fed into the computer which produced and, out at its other end on the ... on the printer, [umm] packing orders. [uh]These were on fanfold paper; we were that modern. But [Uuh] they were [um] not automatically [...] there was no bursting equipment. They were torn out by hand and guillotined into slips and given to the denizens of the packing world who then packed these things together.

The printers we were using, of course, had no alpha facility they were purely numeric. They in fact [uh] were Hollerith printers. You may have seen such things. I don't know, there must be a museum with one somewhere. [uh] Each individual character - the numeric-only characters – [um ] was mounted. They were mounted vertically one above the other 0 1 2 3 4 5 6 7 8 9. [um] Each one was spring loaded and the whole [um] print bar was raised up to the right point when it stopped and then a great hammer came along and banged all 160 characters or whatever they were set up to - all at once. [um] There was an ordinary typewriter ribbon in between the paper and the type characters and that's how we did the printing. [um] They actually used what were called pence bars; remember the old days of pounds shillings and pence. [uh] 12 pence was equal to one shilling. And so these print bars actually had nought to nine plus a character 10 and a character 11. That was the job, and that was the first job done commercially, we think, anywhere in the world. It was followed fairly closely, about 1952, by a dry run at the income tax tables. And [in] those halcyon days they only changed the [um] the taxes in this country once a year. They have budgets every five minutes now, but in those days it was every April the 5th - or something. They had a budget and inevitably changed the income tax. And for two years, at least two years, we did a parallel run with the Inland Revenue of the tax tables. I believe I mentioned this to you, as well, that policemen would stand by at the House of Commons and the moment the chancellor got to his feet would telephone Joe Lyons and say "yes the chancellor is now speaking" and then another policeman would open a sealed envelope and give us the details of his tax changes. And we were then going to work and produce the new tax tables for the long-suffering British people. It took us two days working flat out - something like 30-40 hours - and it took the Inland Revenue four months. And so in those days - again - the new tax rates were never applicable until later in the year. I think it was done retrospectively, but they didn't have the new tax tables for months after the announcement. So we saved them a bit of time there and we even made less errors than they did which was not bad, when the checking was done.

We went on to do payroll work [um] first of all Joe Lyons's payroll about 1953, I think. When I joined them in 54 [um] my ... the ... it's not my payroll - no the computer department's payroll was always done manually - what clever people there were. But everybody else's pay in Joe Lyons's Cadby Hall was done on the computer. And then, we as a bureau operation, did payroll for the Ford Motor Company.

The machine ... well [um] it was a serial computer - great thing serial computers - very clever. An 18 or 36 bit word length you could please yourself a short or long word and the total instruction code was 32 different instructions. [um] I think you would now call it to a 'two address system'. That is to say, you could if you wished address a source and a destination location in the main store with the same instruct... - with appropriate instructions. The storage, of course, was by means of mercury delay lines; long bits of gas pipe full of mercury. [um] And the total storage for all purposes, that is data, [um] programming [uh] program and everything else was 2048 16 bit words [this appears inconsistent with the point above about word length], which wasn't a lot. The programmers in those days were extremely clever to get anything done at all. There was no backing storage, no discs, no tapes and nothing else at all. [um] The mercury delay lines were interesting. They were, as I said, basically [um] long bits of gas pipe full of mercury with a piezoelectric crystal in each end. The problem was initially to get [um] an intimate enough contact between the mercury and the crystals. The crystals were not optically polished flat, which was the trick the National Physical Laboratory got up to with theirs, and English Electric. Joe Lyons, [um] for reasons known only to themselves, elected not to polish the crystal surfaces and so they were full of cracks - very minute ones but cracks nevertheless - and the problem was to get the mercury into close enough mechanical contact for the thing to work. The original way of doing this was to fill these tubes initially right to the brim. They had filler holes in the top surface somewhere ... fill them with methylated spirits - I don't know what you call methylated spirits – 'wood alcohol' I suppose. It's methyl alcohol anyway and so you had the 64 tubes [um] all filled with [uh] methylated spirits and you then poured in the mercury which was great because it ... it was much heavier and fell to the bottom and you got a good contact between the crystal and the mercury. But, of course, all this [um] methylated spirits bubbled out of the top and before the end of the day everybody was drunk on the fumes. It was a perilous business. We later discovered that a couple of drops of another wetting agent - sodium benzoate - on the crystal faces before they were installed in the tubes, would do just as well. So we were able to abandon the methylated spirits, much to some people's annoyance, I think.

[um] This had its own problems of course because we then used to get an interesting phenomenon called third reflection, whereby a pulse sent down this mercury delay - an acoustic pulse sent down this tube - would bounce off the crystal surface at the other end. It would go out and it would be picked up and turned back into an electronic pulse of course but it would also bounce mechanically - reflect back off the crystal surface back to the other end of the tube - reflect off that transmitting surface and back to the to ... the other end again. And you'd get what was called a third reflection which produced a little spike of grass [?] [um] just after the pulse you wanted. And we cured these by stuffing things like knitting needles and bits of [um] piano wire down the filler holes of the mercury tubes.

[um] The peripheral equipment was two paper tape readers - about 100 rows a second I suppose for anti-photoelectric readers, three card readers, [um] entirely mechanical Hollerith card readers, two card punches and the two line printers I've described. And they went about 80 lines a minute. I think we were the first people to fit an alpha printer to a computer. This would be about 1956, 57. We bought it in France. It was made by Compagnie de Machine Bull who later became part of [uh] de La Rue Bull, but at the time they were just making an alpha printer for [uh] uh card punch purposes. [um] It had a separate print wheel for each print position with the [uh] letter of the alphabet and numbers around the periphery of the wheel. And [um] it was rather like stopping a bicycle by sticking a piece of wood between the spokes, because all these wheels rotated all the time and when you call ... called on a certain character, it used to stop the bottom drive wheel dead and the top one with the characters on literally rolled against the paper. And if you were unlucky, as we often were, [it] flew right out the top of the machine as well. So we put an alpha printer on which worked very well.

[um] The total electronics it [Leo 1 or perhaps Leo 2] had - 6 500 valves - it had ordinary [um] electronic delay lines you know 'lump' delay lines; some inductance and capacity. It had potentiometers all over the place which was not a good idea. The thing was so loosely designed that you had to go around every day and twiddle the pots [potentiometers] to make it work at all. Forced ventilation, I suppose several kilowatts of heat went out into the Hammersmith Road. All the programming was in machine code, that is in binary. We had no high level languages whatsoever. We tried to add magnetic tape about 1952-53 in collaboration with STC [Standard T[...] and Cables?] and ITT and it failed miserably. [um] Not their fault I'm sure. It just was beyond the state of the magnetic tape art at that time. We actually got magnetic tape as you probably remember onto Leo 2 using Decca tape for the purpose.

Well Paul, that's all I know about Leo 1. All that comes to mind apart from, [uh] you know, odd little [um] bits and pieces of [uh] technicality, which I'm not sure you're interested in, or that I can even explain adequately at this time.

[End of interview] 

Provenance :
Collected by Paul Kelley and donated to the LEO Society via John Daines

Copyright
The video above may be subject to copyright of the original author and is present only on this website for non-profit educational purposes only. If you are the original copyright owner and wish to have the media removed from this website please contact us using the relevant contact email address listed on our contact page.