 My name is Kevin Morrell, and two various hats, as Laura mentioned. One is the National Museum of Computing, another is the Computer and Conservation Society. And I want to talk about a machine called the Harwell Computer, the Harwell Decatron Computer. It's also known as which, it's had several names. And I'm conscious when I've done this before I've gone through a chronological history and not done the reveal until the end. Now that's okay if you know what the machine actually looks like. Felly, rydyn ni'n cael ei wneud yn ymwneud yn y cyrraed mewn gwirionedd ac yn ymwneud y cwrddol mewn gwirionedd. Ac rydyn ni'n cael ei gwybod yna, sy'n mynd i gyd. Felly, yna'r gwirionedd, mae'n fod yn digwydd, mae'n gwybod... Ysbryd. Ysbryd hynny'n ymwneud o'r ddechol yn ymgynogiad. Mae'n gweithio'n gweithio'n gweithio'n gwybod unrhyw unrhyw yw'r gwirionedd. In 1949, the atomic energy research establishment at Harwell needed to improve the efficiency, accuracy and reliability of its calculations. By 1951, the Harwell Decatron was complete. Although it looks archaic, the Harwell Decatron has all the main functions that can be found in a modern computer. It has input, output, a processor and memory. Modern computers are programmed using languages, but the Decatron is programmed in pure machine code. Instructions are fed in using a numeric keypad or punched paper tape. The Harwell Decatron was designed before transistors were available, so its processor is made up of valves and relays. 827 individual Decatron tubes are used to hold data and program instructions. Each Decatron tube can store a decimal number between 0 and 9. It does give the game away slightly actually, but it's nice to see the machine as it is now. But I need to take you back to the start. I need to take you back to the end of the Second World War and explain why Harwell came about. We had shipped all of our atomic scientists, nuclear engineers and things to Los Alamos to the States. All the original work on nuclear power and atomic energy was done in Europe and the UK. We shipped everybody across to the Manhattan Project in New Mexico. But there was a special relationship between Roosevelt and Churchill that all the information that was learnt would be shared back with the UK at the end of the war. And that's the basis we worked on. Sadly, at the end of the war, we have an athlete in power in this country and Truman as President of the United States. The agreement that had been made had been lost. And in fact the Americans introduced something called the McMahon Act in 46, which forbid any sharing of all of that research and that knowledge from Los Alamos. So the UK, Europe completely abandoned by the States. It's interesting the way that history sort of pivots in a huge way on just tiny little things. Ernest Bevan, who was the Foreign Secretary had been over to the US and had a really rough meeting with the Secretary of State and came away extremely angry. Came back to the Cabinet and said to the Cabinet, oh and the Cabinet was pretty divided whether to go down this route of developing atomic weapons and atomic power or not. Bevan came back, banged the table and said we must have this thing here and it must have a bloody Union jack flying on top of it. So the whole history of the UK changed suddenly and the decisions were made in the Cabinet. Nothing was mentioned until 48 in almost a sort of side issue in Parliament and that was mentioned that we had actually started to develop atomic power in this country. Is that me? It was my shame, isn't it? Yes. Problem was all the people that had come back from the Los Alamos project had worked on individual little departments. Hardly anybody had a whole view of the whole process. And in fact a lot of fundamental research had just been skipped completely in Los Alamos. So things like half lives of certain elements were just simply not known. So they had to start again and hardware was built next to a village just south west of Oxford. It was an old RAF base, the very first of a new range of new series of RAF bases. Just as an aside you should never trust version one of anything. This is version one of an airfield that was built just in 43. The runway was over a brow of a hill, so when you were taking off you couldn't actually see the runway until you got to the point where you were committed to take off on the brow of the hill. Okay, well that's manageable, that's fine, and if you have to have bought the take off you have to have bought the take off. Sadly they built the ammunition dump at the end of the runway. So by the time you got to the brow of the hill frankly if your wheels weren't off the ground you weren't going to get anywhere and then it was extremely sharp right turn to avoid the ammunition dump. That's version one, got better still. But hardware had good water supply, lots of big hangars around that they could actually build equipment in and accommodation as well. Although there were other centres in Cheshire and Winscale, hardware was the public face of atomic energy as far as the UK was concerned. It was operated like a university almost. Everybody that works there in those early days all they can remember is the mud everywhere, mud and chalk as they were building it. The story is that if you get to work one morning on a particular route when you left they dug the road up and several people just obliged to stay at work until they rebuilt the road. But I think it's a phenomenally exciting time. They built, I think I want to compute it as I promise, they built the first nuclear actor in Europe called Gleep which is about the size of a sort of three bedroom detached house. Gleep was so successful. It carried on to the 1990 and was used worldwide for calibrating counters and nuclear detectors and so on. It's phenomenal. There was a campaign to keep it as a national monument and unfortunately that didn't get going in time before the nuclear inspectorate started to dismantle it. So sadly Gleep has gone. Not a very powerful big reactor. Three kilowatts which is about the size of an electric kettle but absolutely reliable. Now I mentioned that all the fundamental knowledge wasn't available to the UK and in fact a lot of the fundamental research hadn't been done. So Harwell were employing every hundreds of maths graduates, metallogists, physicists, chemists and so on to actually work in this university environment. The physicists and chemists had a while of a time, really, no expense spared and so on. The mathematicians had a bit more of a rough time. They were presented with tables of things like logarithms and so on. I think some of us remember logarithms from school, huge books of mathematical tables. The only mechanical aids they had were things like Brunsvega and mechanical typewriters. So everything was absolutely manual. So if you thought you would join in the exciting world of atomic energy you were actually presented with a whole series of differential equations, instructions how to solve them and that could take three, four, five days of simply slogging through with arithmetic and the limited use of tables and calculators. Pretty mind-numbing stuff, very error prone and really I think the people that were in charge of theoretical physics and maths at Harwell were slightly embarrassed about this. But at the same time British Ericsson had developed something called the Decotron Tube and it's a Decotron tube that you could see spinning in some of those displays earlier. Now they were particularly interesting for Harwell. It's not a terribly good picture but at the end of the tube there are ten positions where a glow can show. So we have a counting tube that can count from 0 to 10. Well 0 through to 9 back to 0. Very clever functions on these tubes. We can supply pulses to the tube and that glow will move round. When we switch one on it's at random, let's say if we switch one on it's on 7. If I send two pulses into the tube it will go round to 9. So it has two functions. A is a storage device because if you leave it alone it will stay storing 9. And it's an arithmetic device because we've already added 7 and 2 and got 9. Now Harwell used them, oh the other point as well, if we're on 9 at the moment I send another two pulses in it will go 0 to 1. But it sends a signal out of a carry to say that I've passed 0. Now if you can chain a row of these together you have a counter. And if you connect a Geiger mother tube at one end you've got a counter to count pulses and disintegrations. That was perfect and Harwell used them in the thousands. As I said Harwell was slightly embarrassed really about the sort of mundane jobs that they were getting these bright graduates to do. And a chance conversation really over the garden fence between one of the theoretical physicists and electronics division said well actually I wonder if we could automate this. I wonder if we could actually build a computer to do this. And this is really 47. Three members of the electronics department, Ted Cook-Yabra, Dick Barnes and Gunny Thomas had actually been going back as opposed to Cambridge at this point when they had enough petrol to actually get there and had seen EDSAC being built. So they had an idea. Ted Cook-Yabra knew about electronics. Dick Barnes knew about control systems and Gunny Thomas knew about memory systems. So they thought they could actually do this. So they had to make a case for Harwell for spending this time and this money to actually do this. And they had to make the case to Sir John Cockroft who's in charge of Harwell and his second in command who's in charge of theoretical physics and that's Klaus Fuchs. Klaus Fuchs was very, very, very valuable to the UK project. He'd spent a lot of time in Los Alamos, had a very, very good memory, kept copious notes of everything he'd learnt and understood. Now it wasn't actually discovered till much later the reason he was keeping the copious notes, he was actually passing them to his Russian controller every evening. But that aside, he was the only person that had an overall view of what was actually going on. So it was invaluable to the UK. When our three bright chaps, Ted, Dick and Gunny, go to present this and build in his computer, they would expect three or four hours of knock-about discussion and arguments and so on. When they got there, it was a very, very quiet conversation. Cockroft didn't look very happy, Fuchs looked seriously unhappy and the whole thing was over in 15 minutes and they were told, well, if you think you can do it, you better go away and do it. Well, that's absolutely... OK, well, in one hand, yes, we've been told we can get on and do this, it's an odd situation. Many years later it was discovered that morning of that conversation, Cockroft had finally been told that his second-in-command at Hywel was a Russian spy and Fuchs had been arrested, finally arrested that morning. So I think they had probably had more on their minds than my three engineers with their song and dance show about building a computer. But they got permission to do that. And they built the machine, started in early 1950 and completely by 51 with what's called two store groups. Now, these are the store groups here, those. And it was incredibly successful. What I mentioned before, although some of the accumulators are electronic, the memory is these Decotron tubes are electronic and we saw Decotron tubes spinning around relatively fast but they will actually spin up to about 50 kHz. So they're quite fast. That centre section covered in the metal tube-like covers are all relays and it's relays that control the process of the machine and relays don't go at 50 kHz at all. They go very, very slowly. So this machine typically to do an ad, adding two floating point numbers, would take about nine seconds to multiply two floating point numbers, it would take, no, fixed point numbers, sorry, fixed point numbers, to multiply two it would take the best part of 18 seconds. Now that's desperately slow. It's desperately slow compared with edsac one that was down the road, but it was incredibly reliable. These guys knew how to build equipment that was absolutely reliable and faultless. So the argument was that, okay, edsac was very fast but probably only worked for 15 minutes before it failed again, whereas this would actually work over weekends and weeks long and do a similar amount of work. So it's very reliable. It's handed over to the mathematicians. They started to learn how to programme the machine and albeit very slow, it was very successful. Now, lots of people talked about machine code programming earlier and put their hands up. This is the only operations you've got. This is the complete machine code instruction for the machine. You can test positive or negative in the accumulator. You can jump. You can jump based on that test. So we have a conditional jump. In terms of our arithmetic, we can add, subtract, multiply and divide in hardware, which is relatively unusual and modulus and so on. That's absolutely enough. If you've got a conditional jump and frankly subtract, you've got enough as a general purpose computer. Typically, programmes are written with a bootstrap and a bootstrap programme would clear various stores, perhaps reading constants from another tape. I'll show you the paper tapes in a moment. You could call subroutines on another paper tape drive. Typically, that's how they're written. We have a subroutine on the right, which I think just squares the value that's provided in the accumulator and returns. That's the physical representation of a subroutine. It's literally a paper tape that's glued either end. So when we call that subroutine on tape reader... I can't read that. I think it is. It will read the code from that routine and stop at the return instruction. Of course, because the paper tape is glued, next time we call it, it's ready to actually run again. Now that's tiny that subroutine. Some of the routines that were run later involved having two huge bins either side. The tapes were 40, 50, 60 feet long. But it worked very well. It's a complete programming language. No high-level language. There's not enough store for anything like that. Everything's programmed in machine code. I said it was a slow machine. It was really designed to do at about the same rate as a mathematician would use with a mechanical calculator. Now there's a story of a race. Bart Fosse is a brilliant mathematician. This story grew and grew. One of the big problems on a machine like this is rounding errors. If you have a machine which is endlessly calculating based on previous results, any rounding errors, by the time you finish, are going to be absolutely pointless results. Bart sat down next to the machine to do the same job mechanically, but actually kept up for about half an hour and then retired, exhausted, and the machine carried on. Typically, it was left on for days and days and days. In a report from Ted Cook-Yarbara, running 80 hours, running 55% of the time, that point in the early 50s, that's quite phenomenal. Nothing else came anywhere near that sort of level. Ammitting a five-week pause due to damage, that's quite fun. Ted Cook-Yarbara is in charge of the whole department at that point. Late at night wants to change one of the memory units, so it gets a step ladder up behind the machine, reaches over and knocks it over, literally the whole lot, and that's the five weeks. The general feeling was that, thank God, it was the boss that had done it and not anybody else, but it was the boss that did it. It's not mentioned in Ted's report about why the five weeks wasn't available, but that was the reason. We talked about a charmed life. By 1957, Harwell had actually moved on. They had built something called Cadet, which is a transesterised computer, the first completely transesterised computer. Ted Cook-Yarbara had been over to Bell Laboratories in the US to see the first point contact germanium transistors. They were told originally, you cannot take any out of the lab at all, none can be removed, the supply is so short. Ted Cook-Yarbara smogled half a dozen back in his socks, literally suffered them in his socks, and came back. By the time they came back, within six months Harwell had produced a book on transistor theory and suggested circuits, much to the annoyance of Bell. By 1955, they were available generally, and they built the Cadet. They had access to other machines as well. By that stage, you could buy commercial machines not quite off the shelf, but at least to order. But such was really the love of this machine. Everybody used it, but still trying to use it. They organised a competition to see if we could find a winner, to see if we could find a further education establishment, university or a college, that put together a bid to have the machine. I was organised by the Oxford Mathematical Institute, and 30 submissions came in, the short list of nine. Our winners, who have just come on to the moment, turned up with the Lord Mayor in his chains, the whole of the department, and put together their case for winning it. The winners were the Wolverhampton College of Technology. They made a huge trust, and this is very valuable to us as a museum. The chaplain charge of the department, Cecil Ramsbottom, loved publicity. Any excuse to get the college in the papers, the TV, anything he would do. So we have a wealth of sort of information about this. Whether they understood what they got is another matter. One of the things they sold it for, it can for instance work out wage calculations much more quickly than a human being. It seriously couldn't. Absolutely seriously couldn't at all. But they promised local industry that they would actually have access to the machine as well. And they'd raised funds for this. But they're very chuffed. They organised, ah, yes, point point. They renamed the machine, rather than calling it the Harwell machine. Now, unusual with acronyms, which came first is not exactly clear, but they renamed it as the Wolverhampton instrument for teaching computing from Harwell. And again got that in the button, in the local papers, the Birmingham papers, the Wolverhampton Express and Star, but started with the machine, the first undergraduate course in 65. And we have a picture of, I think there are 12 undergraduates. They're all 18 and 19 starting on that course. That same group of now just all retired came back to the museum last year. So we have a matching photograph in front of the machine as they are now. Interestingly, IBM had heard about Wolverhampton in teaching computing. IBM turned up in the third year of this undergraduate course and offered them all jobs. And all but one actually then worked for IBM for the rest of their lives. So it was useful that they all kept in touch about this. They were, ah, used the machine with local schools, local grammar schools, and again these are all publications from local papers and so on. They promised help for local industry, and one of the industries in Wolverhampton is key making, and the Chubb company are making Chubb Blocks. One of the problems with making certain, mortise keys, certain keys is the pins need to be a certain width. There are only certain, otherwise they'll actually snap off. There are only certain designs and the which computer was used by these two gentlemen to actually solve the problem and produce key patterns for Chubb company. They made a few hardware changes as well. The chap on the left is Chapo Peter Burden who was a sixth former, was waiting to go to Cambridge to read maths, so he used the machine in his summer holidays. We'll come on to Peter later because he reappears in our story. Harwell used it until 73. This is the second retirement of the machine. And again Cecil Ramsbothan is a self-publicist as well. What machine listed in the Guinness Book of Records was the world's most durable computer. There was endless press coverage. ATV, the local independent TV station, did some recording there as well, which is sadly lost, which would be absolutely wonderful to see. But there's a whole champagne reception and there's access to what to actually do with the machine. Absolutely nobody would let it be dismantled. So it went to Birmingham Science Museum. Now I grew up in Birmingham. It's not a terribly good picture of the Science Museum. It was a huge ramshackle place. Absolutely fantastic. Unbelievable. I spent every Saturday afternoon at Birmingham Science Museum for many, many years. This next picture, my mom reckons it's me. I'm not sure it's not, but still. There's me with my mother standing next to me looking at the Science Museum. I'm pretty sure it's not me. Anyhow, it was on display at Birmingham Science Museum until 1997. Let's get a move on. When it was taken off display, the third retirement and went to a sort of collection centre in central Birmingham. Now, by this stage, I'm involved in the Computer Conservation Society and was going to Birmingham to do some sort of audit of the equipment that they got. And I saw right at the back of that those three odd shaped things. I thought, I'm lucky how I recognise that. That's from the witch computer. It took quite a while and we put together, let's go back a bit, jump in again, to dig all of that equipment out and find it. We eventually found what we thought was a good 60-70% of the machine. Yes, I realised I'm done. Sorry. We had to make decisions about what we were going to do with the machine. We found in the Birmingham Collection Centre this store of components, which is fantastic. The bottom of that box includes all the circuit diagrams as well. It's unbelievable. We also got back in touch with Dick Barnes and Take Out Y Arbra. Got some information from them. And then set up the working party, persuading Birmingham Museums and County Council that we could have the machine and the machine was brought to TNMOC. I'm going to skip that. That's the unloading of the van that was actually on BBC News one evening. Then the restoration process started. It's been quite a sorry state. It had suffered really at Wolverhampton, not being looked after as well as it might have been. But by, we had help from Dick Barnes again. And of course we've got the benefit of modern equipment to restore this. So in 2000, I don't remember, about five years, exactly five years ago, the machine was restored. We had the reboot event. The reboot event had the original designers and the users of the machine brought back in. That was televised worldwide. The YouTube video went to a million hits over the same weekend. Quite a phenomenal event. The machines now, at this time of day, will be in use. There will be an education party at the museum and they will have brought code in and be running code on the machine. It's incredibly reliable. Quite a wonder. There are future plans. There are various books. It's useful when you've been a huge amount of research, quite a book, because then you can actually gradually forget it all, knowing it's in the book. It's very useful. We've built the simulation and the emulation at the moment as well. And there's a web accessible emulator coming up soon. There's some thought that a few people want to make a smaller replica of it. And there have been electronic replicas of it using LEDs arranged like Dekatrons. So there's a lot going on with the machine. I'll show you this video because it recaps the last... I didn't work, did it? Oh, here we go. This is the Harwell Dekatron computer. It is the oldest working example of a first generation computer anywhere in the world. In 1949, the atomic energy research establishment at Harwell needed to improve the efficiency, accuracy, and reliability of its calculations. By 1951, the Harwell Dekatron was conquered. Although it looks archaic, the Harwell Dekatron has all the functions that can be found in a modern computer. It has input, output, a processor, and memory. That printer is being wound on manually. It's a con. 0 and 9. Rather than showing the results on a screen, the Harwell Dekatron would either print down or punch holes in paper tape. In 1957, transistorisation made the Harwell Dekatron obsolete. The atomic energy research establishment launched a competition to give away their machine. The competition was won by Wolverhampton Staffordshire College of Technology who renamed it The Witch. This stood for the Wolverhampton instrument for teaching computation from Harwell. The college used The Witch for 16 years between 1957 and 1973. The Witch helped teach scores of students about computing. In 1973, it took another journey to the Birmingham Museum of Science and Industry where it was on display for many years. By 1983, The Witch was no longer working. The artist John Eather painted the computer out of use and called this proposition the Portrait of a Dead Witch. The painting is now on display in the Jamstreet Café in Manchester. In 2009, the Harwell Dekatron was brought to the National Museum of Computing to be restored to working order. In 2014, it was recognised by Guinness World Records as the world's oldest working original computer. Thank you.