Ask The Experts | Detailing Colossus

hello i'm gavin clark i'm host of ask an expert the national museums of computing's program answers your questions on computers in in in from history and in the museum's collection itself colossus is one of the museum's most popular topics and we recently tackled a question about this code breaking machine on the topic topic of block diagrams um you can of course see a block diagram uh which outlines the architecture of colossus at the museum um which is from uh the website of tony sale tony sale of course was the uh the mastermind behind the rebuild of colossus that you can see at the museum today we had an email following on from our twitter q a session from alan mcintosh who followed up with our reference to the block diagram which we discussed and he said i think i recognize the one that you showed as being from dc hallways a technical description of colossus one your correspondent might be interested in the slightly more detailed one that flowers gave in his art his article the design of colossus in the annals of the history of computing volume five number three july 1983 which i could give as an attachment and he sent us a copy there um the inclusion of a shift registers presumably means that this diagram is for a mark ii colossus now there's a lot to unpack in that um the fact there is two types of colossus not just one two model variants peter hope is with us today and he took us through the answering the original question that sparked allen's uh reference that he sent us via email peter welcome back thanks for joining us again today i thought maybe we could have a chat and just go through this good to have you again sure um now tell us a little bit where where would this what what's the block difference between the block diagram you see in the museum today versus the one that alan has sent us and where would the diagram that um alison has come from okay well the diagram on display in the museum is of tony sales or origin and it's very very high level and it applies in concept equally to the what we now know is the mach 1 colossus effectively the kind of prototype if you like but also that particular block diagram will also apply to the mark ii um it's able to apply to both because it is it hasn't had much detail in it it satisfies the curiosity i guess of most people who would pass by that particular part of the display um but to those wanting to know a little more about the internal workings of the machine i think it would leave them wanting um and it's it's it's it's great that alan's picked this diagram up which incidentally i've had a copy of myself for donkey's ages and kind until this query came up i completely forgot about it myself um and the the diagram he refers to is very definitely of a mark ii colossus and there is a number of reasons why we can say that we'll get into it in a second i mean what is interesting i think is the um is the fact that there's folks on the diagram the schematic itself because the one he's showing us here which you had in your collections very very detailed um this of course came from tommy flowers who people may not know was the architect the man who created colossus in the first place and he worked with gpo in dollars hill during the 1930s and the second world war which was kind of i suppose the outsourced r d operation for the government and and for bletchley park and he built this thing but what i think a lot of people would find incredible is that this diagram was actually came 40 years after the fact didn't it he came for this came from memory he didn't this he didn't this is not contemporaneous it were this came after the fact didn't it yes he did yeah the the origin of it actually goes back to 1981 um when tommy was uh preparing a lecture um and in preparation for the lecture of course he had to rely on first of all his own memory and the memory of the colleagues of his who were still around at the time in 1981 um because of course they were under orders at the end of the war to destroy all the documentation they are that they held and flowers was was assiduous in doing that he tells us he there was a bonfire of diagrams and the like some of the engineers did squirrel bits and bobs away but there were only fragments um they're on display in the colossus gallery at the national museum of computing so you can have a look at those um you can't tell an engineer to throw anything away i think that's a fact of life yeah my my loft is proof of that so um he was memory only really that to put all this together and it's remarkable uh the level of detail that is in there and the way that it parallels um what we now have as our rebuild at the museum the the so that was the the lecture was in 1981 it was formalized even more in 1983 um in a document that alan was referring to the annals of the history of computing volume five number three july 1983 um which was um the annuals of computing were operated or created if you like by the american federation of information processing societies otherwise known as afips um and i mean as part of that that that framework the ieee is part of that overall framework and that's how i got my copy being a senior member of the ieee for some time and having completely forgotten about it once i got it you square it away and it was only alan's query that brought it to mind again um and he's right there is a there is a very more detailed schematic diagram embedded in that document and we've got to take tommy's word and tommy's memory as being i guess the most accurate um source of information we've got and certainly the most authorizative because he did design it well tell us a little bit about let's come back to this difference between a mark one and mark two now when we talk about colossus i suppose in a high level we just say colossus we don't get into the variants but there were two distinct types there was the mark one which was very much the prototype and then there was mark two which there were many there were so many of what nine or ten i think roughly and what was the why do we we look at this diagram why do we think what is why are the clues in it tell us this is a mark one okay just a little bit about mark ones and mark two the mark one was you could you could you could look at that as being a proof of concept flowers initial um ideas about replacing the heath robinson and and overcoming its unreliability um didn't meet with the with the level of approval that he was looking for um and so um he he and the post office had to create that really in in the early part of 1943 around february 1943 onwards um under their own steam and so they couldn't really um as a proof of concept they needed to prove that it could execute effectively bill tut's algorithm for finding machine settings on the lorentz cipher and by december 1983 they'd done that that machine was 1500 vacuum tubes it was quite crude and it only took you part way through the process there was a manual input required to complete the the decoding process um but it did prove that flowers in idea of using electronics to speed the machine up and make it more reliable was viable having having demonstrated that the next phase was to try to look to means whereby you could speed the process up one way of speeding the process up is to create more machines um and but given the limited uh manufacturing capability and limited number of valves available then that was going to be problematic they were overcome those by putting a whole load of people on it and dedicating the work of the the post office factories division in birmingham almost um well 100 involved in in creating the uh parts to assemble them colossus from linda lisa also made available valves from the uh from the states um so it was all looking a doable prospect the mark ii colossus thousand four hundred vowels actually are rebuilt two thousand four hundred and twenty because i've counted them all um that um was the next logical development um and in an attempt to get the speed up experiments at to actually after experiments of bletchley they realized they could push the speed of the tape to speed the processing up the nominal speed of the tape reads at 5 000 characters a second they found they could push that to 9 700 and then all they'll let loose because the tape would fracture into tiny pieces it's doing nearly 60 miles per hour at that speed um and it fractures into small pieces another finding bits of it all over the machine just embedded everywhere the stable speed was 5 000 characters a second and that was the limit of processing for a mark one so the only logical way that flowers could design in more speed was to make the thing parallel um in other words to do more than one process bear in mind the processes in colossus are quite repetitive and are very similar the ability to to parallel up was the next logical step that tommy flowers took in the design and the next machine to be delivered on the end of may 1944 was indeed a mark ii colossus and it embodies parallel processing it's five times parallel and so your effective throughput for no speed increase on the tape is five times you're now reading the tape effector at 25 000 characters a second or your processing should i say processing data at that speed um looking at the the block diagram we can actually see where those shift registers doors are um on the left hand side of the dock that the the message tape loop and the reader itself and that's producing five bit binary output and to the right of that block actually incidentally in modern day computing terms that will be temporary temporary data storage effectively a one bit store in the tape reader itself you then have a shift register store it's six characters of shift register um why is it six characters it's because this is all doing modulo 2 processing the the whole process of the machine looks at the current character and and process it by looking at the character before so it needs to to process five characters it actually has to store six yeah so there are six shift register stores um and they are truly shift register swords they are recognizable if you look at the circuit diagrams they're recognizable as flip flops and what we now know is a flip flop it's a one bit data store um and it's not new even in 1944 when flowers is designing this the design of it concept of it goes back to 1919 was known as the eccles jordan trigger we now know that as a one bit store or a flip flop um so you can see the shift register source is five of those um the five times parallel and they're feeding their output um to the pro to a program store program switches which is a block kind of in the middle to the right of the diagram um there's only one of those shown there are five um there are five of these and they're all processing they're doing the same process in parallel at the same time there are functional units in that block diagram as well that's your arithmetic logic parallel those functions are fairly basic logic functions there's modulo 2 edition which is key to both the enciphement and the decipherment phase of code breaking there are boolean adders and there are inversions inverters so changing the state of a boolean condition and there is there are also lots more one-bit stores lots more um not sure franchises but flip-flops in the arithmetic logic unit so the program switches themselves this is altering the behavior of the machine they're your program store it's not as we wouldn't know a program store these days because that would be in memory effectively um but but it's there nonetheless it's paralleled in this machine even though it's a single purpose not a general purpose machine there are some some quite interesting parallels to modern day architecture in there so it's more of a this is a more kind of a four shot that mark two is more of a foreshadowing of things to come whereas the one that was absent these kind of features yeah the mark one was was very much an incremental uh between the the the bare bones of the heath robinson and the rather more sophisticated mark ii they were in fact you mentioned how many they were they were pro they were planned to be 12 um there were ten in operation when when hostilities finished in 1945 and the 11th was in the in the state of partly being partly constructed um so h block itself was designed to hold well there were four in f block originally the nh block was designed to hold the rest of them so another eight would have gone in there um we like to call it the words for the world's first purpose-built computer center because it only exists because they needed to stick colossus somewhere um so there may have been 12 in in total but in fact there were only 10 running at the end of the war and they were 10 mark there were 10 mark twos or was it no they were all mark twos by that time the original mark one had been been brought up to mark two um and they're all different mark twos as well interestingly enough um the programmability of this because it was switches and and and patch cords um it was found that some of the machines were left in a particular configuration for a special for particular kinds of jobs so if a job required a certain configuration of the machine um there will be let's say colossus number five would have been left in that configuration effectively hardwired into that configuration it would have been reprogrammable if they needed to um but for efficiency they could steer work towards the the most appropriate machine the whole operation was a factory of course and they were they were just 24 24 24-hour operation okay great that's great thanks very much for taking us through that kind of fact that amazing history and fascinating uh understanding of the of this very simple and yet complicated diagram it's incredible how these these uh these artifacts show up and come and bubble up to the surface um and thanks to alan for submitting the email and giving us the opportunity to discuss this subject definitely yeah and the best way to experience it is to come and see it in real life and in operation of course and of course if you've got if you've got any other questions anybody has any questions just tweet us with the hashtag asktnmoc thanks very much you

Channel: TNMoC

Views: 4,565

Keywords: #askTNMOC, Colossus, block diagram, Tommy Flowers, computer history, TNMOC

Id: LqwKfY0vErw

Length: 16min 18sec (978 seconds)

Published: Wed Nov 04 2020