The Secrets of Station X

by Michael Smith

  The first regular transmissions intercepted by the British were on an experimental Wehrmacht link between Vienna and Athens which used the Lorenz SZ40 cypher machine. The radio-teleprinter traffic had been mentioned in the Enigma traffic as Sägefisch (sawfish), probably because of the teeth on the SZ40’s rotors, and as a result the Bletchley codebreakers gave the radio-teleprinter traffic the codename Fish, with the SZ40 traffic itself given the codename Tunny. Another teleprinter encyphering system, the Siemens and Halske T52 Geheimschreiber, was also detected in use by the Germans, mainly by the Luftwaffe, and was codenamed Sturgeon, but because of the lack of Army Enigma the codebreakers decided to concentrate on the Tunny material produced by the Lorenz machine.

  This worked on a system invented in America in 1918 by Gilbert Vernam. Teleprinter transmissions are based on the international Baudot system, a binary code in which each letter is made up of a series of five elements, or ‘bits’. Each of these ‘bits’ is either a ‘mark’ – the equivalent of the binary 1 and denoted by a cross – or a ‘space’ – the counterpart of the binary 0 and represented by a dot. Each ‘bit’ is transmitted as a separate negative or positive impulse.

  ‘The letters were in the form of five elements, always preceded by a start signal and always followed by a stop signal,’ said Ken Halton, one of the GPO teleprinter engineers who worked at Bletchley. ‘So they were basically seven units in length, the middle five being the active code elements. The start signal and the stop signal were there to start and stop the machine at the receiving end.’

  The Baudot Code as it is known was not secret, but Vernam’s cypher system combined the five elements for one letter with those of another letter ‘randomly’ selected by a cypher machine to produce a third encyphered letter. Each of the five elements that made up the letter under the Baudot Code were added to each other on the basis that like and like, either two marks or two spaces, would produce a space, while like and unlike, i.e. a mark combined with a space, produced a mark. Once the cypher machine was introduced the new combination of marks and spaces constituted a third, encyphered letter. For example, if the letter A – which in the Baudot Code is represented by ×ו•• – is added to B – which is ו•×× – the result would be •×•×× or the letter G.

  The beauty of this system lay in the decyphering process. Since, in binary mathematics, addition is the same as subtraction, an identically set cypher machine at the receiving end had only to add the same pattern of ‘random’ letters to the encyphered letters, or effectively re-encypher the message on the same setting, to come up with the original clear text. The Lorenz SZ40 took Vernam’s idea one stage further, adding not one but two separate ‘random’ letters to the original letter in an attempt to make it even more difficult to decypher.

  The Lorenz machine had twelve wheels, ten to encypher the message, paired in two separate rows of five, and two motor wheels. The movement of the wheels was very complex. Each of the encyphering wheels had a number of springed teeth equally spaced around its circumference which could be put into an active or inactive position to form either a mark or a space. The first wheels in each of the pairs, known as the ‘chi’ wheels, moved regularly one position with each letter. The ‘chi’ wheels were geared to move at different speeds and each had a different number of positions it could adopt. The second wheels, known as the ‘psi’ wheels, moved intermittently with their motion controlled by the remaining two wheels – the motor wheels. The five elements of the letter were passed through the first set of five wheels, each element through one wheel, and were either modified or left unaltered depending on the addition principle described above and whether or not the pin at that point was active or inactive. They were then passed through the second set of wheels where a similar process took place.

  The intercepted teleprinter messages were sent to Bletchley Park where they were examined by John Tiltman and his research group. Tiltman himself did the initial work, quickly identifying the messages as being encyphered using the Vernam system and began to work on a method of unravelling the messages by hand. Tiltman realised that, because of the way that binary mathematics worked, if two messages were sent using the same setting, and they could be lined up so that the starting points matched, adding them together would eliminate the ‘random’ letters that the Lorenz machine had introduced. What would be left would be a combination of the letters in the two original messages, still in their original positions, as if their two binary values had been added together to form one.

  A number of messages sent on the same setting, or as the codebreakers described it ‘in depth’, were recovered. But Tiltman was having difficulty separating the clear texts out. Then a lazy German operator came to his assistance. On 30 August 1941, the operator sent a message 3,976 characters long. When asked to repeat it, he sent it again with the exact same settings.

  It should have produced exactly the same message which would have been no help to Tiltman at all. But although the codebreakers knew from the operator chat that the two messages were the same and they certainly began identically, within a few letters they had become different. The vital clue that allowed Tiltman to work out what had happened was the fact that this message had fewer characters. The operator had left something out of the original message.

  Anxious to cut down the length of time the job would take, he had abbreviated a number of parts of the message, beginning with its introduction. He had cut the word Spruchnummer, message number, using the German abbreviation for number to make it Spruchnr. Tiltman now knew that if he lined the subsequent apparently different parts of the two messages up and added them together, it would strip off the keys, leaving him with two identical clear texts added together.

  Because each character of the combined text represented only one letter, albeit it added to itself, the stripped text would have similar characteristics to the German language and could be recovered relatively easily by exploiting basic cryptanalytical tools such as letter frequency.

  Sadly it was not quite as easy as that because the message had been abbreviated in a number of places. But Tiltman, who preferred to work alone, standing at a custom-built high desk, was a brilliant codebreaker. Gradually he worked his way through the message recovering the clear text up to the next abbreviation, working out what that was, realigning the two texts and reconstructing the next piece of plain language.

  Eventually he managed to recover the complete text. This was in itself an amazing feat. But it was to be followed by one that was perhaps even more remarkable. Once Tiltman had completely decyphered the message all he had to do was add the clear text to the encyphered version to find the elements that had been added by the Lorenz machine. He gave these to the research section so they could try to reconstruct the machine that would have produced those 4,000 letters of key.

  ‘They worked hard, guided by an ingenious theory, but to no avail,’ recalled Bill Tutte, one of the other members of the research section. He later wondered if it was ‘a gesture of despair’ that led Captain Gerry Morgan, his immediate superior, to hand him the key strip and some other documents and say: ‘See what you can do with this.’ Tutte, a young Cambridge chemistry graduate who had subsequently become interested in mathematics, wrote out a stream of the first of the five individual ‘bits’ or ‘impulses’ that made up each of the encyphering characters, looking for some form of pattern. If only one wheel had been used then a repeating pattern would be found in which every repeat matched precisely with the pattern on the wheel. However if two wheels had been used the precise pattern repeated in each row produced by the first wheel would to some degree be altered by the effects of the second wheel in the pair. Tutte detected a pattern suggesting that one of the wheels might be able to produce twenty-three different positions, or possibly twenty-five. So he decided to test both out, multiplying the two together to form 575 and using that as a basis for his work. In doing so he realised that 574 was an even better fit.

  I wrote the first impulse on a per
iod of 574 and marvelled at the many repetitions down the columns from row to row. But surely the Germans would not use a wheel of that length? Perhaps the true period was forty-one, this being a prime factor of 574? So I wrote the first impulse a third time, now over a period of forty-one.

  It clearly worked. One of the wheels on the machine did have forty-one positions. At this stage of the process, the rest of the section joined in the attack. It later transpired that Tutte’s initial belief that one of the wheels had twenty-three positions was correct, but it was the last of the chi-wheels so it would have taken a lot longer for his attack to succeed, although it would have done so in the end. ‘Then I suppose my success would have been attributed entirely to close logical reasoning,’ Tutte later mused, suggesting that other members of the research section, frustrated in their own efforts, might not have been as generous in their praise as he deserved. ‘As things were, I was supposed to have had a stroke of undeserved luck,’ Tutte wrote. ‘Think twice, O Gentle reader, before thou takest an unexpected and opportune short cut.’

  Once Tutte began to make progress, other members of the research section joined in and they managed to work out its complete internal structure and how it operated right down to the intermittent movement of the second row of wheels.

  Given that no one at Bletchley had any idea what a Lorenz machine looked like, Tutte had achieved a near miracle, but he remained unassuming and modest about his feat, recalled Shaun Wylie, who later moved from Hut 8 to work on Tunny. ‘You could hardly get anything out of him,’ Wylie said. ‘I once wanted to hear from him the saga about how he’d done his astonishing bit of work and I think we got interrupted after about half an hour but I really hadn’t got much out of him.’

  But the importance of the breakthrough was not lost on those in charge. ‘That the Research Section was in fact able to achieve this feat within a matter of a few months was one of the outstanding successes of the war,’ said Nigel de Grey. With Tiltman and Tutte having shown that it was possible to break the Tunny traffic, it was decided to set up a section to exploit it, de Grey added. ‘The system was being fairly rapidly extended by the Germans over their high command networks and such messages as could be decyphered by “depth” reading left little doubt that their contents would have considerable intelligence value.’

  The new section was run by Ralph Tester and therefore became known as ‘the Testery’. Ralph Tester was a 39-year-old accountant who, having spent much of his working life in Germany, had an exceptionally deep knowledge both of the country itself and the language. He had been working with Tiltman on police cyphers having only recently transferred to Bletchley from the BBC Monitoring Service at Caversham, near Reading, which intercepted German public radio broadcasts. A small site at Knockholt, near Sevenoaks in Kent, which was owned by Section VIII, the radio section of MI6, was used to intercept the Tunny material.

  The station, which was staffed mainly by ATS operators, came on line in mid-1942. The German transmissions were turned into a perforated teleprinter tape. The teleprinter tape could then be fed through a teleprinter to send it to the Testery but the tapes were themselves also sent to Bletchley by dispatch rider.

  Everything was done twice to ensure there were no mistakes, Kenworthy recalled.

  An error in one character in several thousand was enough to cause trouble. A system was introduced to overcome this on the principle that two separate people would hardly be likely to make the same mistake. All tape was therefore measured by two girls before being read up.

  The Fish link between Kesselring’s headquarters and Berlin, codenamed Bream at Bletchley, was the first of many such links that now began to produce extremely high-grade intelligence on German dispositions and intentions, and because it was at such a high level this intelligence was not just limited to the specific area covered by that particular link.

  As the Germans increased the number of links and the amount of Tunny traffic grew, new outstations would be opened at Forest Moor, near Harrogate; Wincombe, near Shaftesbury in Dorset; Kingask, near Cupar in Scotland; and Kedlestone Hall, near Derby. The codebreakers recruited to work in the Testery included Roy Jenkins, who subsequently became Chancellor of the Exchequer and, as Lord Jenkins of Hillhead, Chancellor of Oxford University; Peter Benenson, the founder of Amnesty International; and Donald Michie, a Classics scholar from Balliol College, Oxford, who subsequently became Professor of Machine Intelligence at Edinburgh University.

  Peter Hilton, later a distinguished Professor of Mathematics at the State University of New York, but brought into Bletchley Park as a 21-year-old student, was one of those working in the Testery. I was recruited by a team looking for a mathematician with a knowledge of German. I wasn’t a mathematician at the time.

  I was in my fourth year at Oxford. My knowledge of German was what I had taught myself in one year so I wasn’t what they were looking for at all really. But I was the only person who turned up at the interview and they jumped at me and said: ‘Yes, you must come.’ I loved it. There is this enormous excitement in codebreaking that what appears to be utter gibberish really makes sense if only you have the key and I could do that sort of thing for thirty hours at a stretch and never feel tired.

  Just as with the Enigma, German errors were helpful in breaking the system.

  Sometimes the German operator made the mistake of encyphering two successive messages using the same wheel setting. When he did this, we could combine the two encyphered texts and what we got was a combination of the two German messages. So you had one length of gibberish which was, in a certain sense, the sum of two pieces of German text. So you were tearing this thing apart to make the two pieces of text. And it’s absolutely a marvellous process because you would guess some word, I remember once I guessed the word ‘Abwehr’. So that means you have a space and then ‘Abwehr’, eight symbols of one of the two messages.

  By subtracting the Baudot elements for those letters from the characters in the combined text, Hilton would then be left with eight letters from the other message.

  But the eight letters of the other message would have a space in the middle followed by ‘Flug’. So then you would guess, well that’s going to be ‘Flugzeug’ – aircraft. So you get ‘zeug’ followed by a space and that gives you five more letters of the other message. So you keep extending and going backwards as well. You break in different places and try to join up but then you’re not sure if top goes with top, or top goes with bottom.

  Then of course when you’ve got two messages like that, as a codebreaker, you have to take the encyphered message and the original text and add them together to get the key and then you have the wheel patterns. But for me the real excitement was this business of getting these two texts out of one sequence of gibberish. It was marvellous. I never met anything that was quite as exciting, especially since you knew that these were vital messages.

  Throughout 1942, the work on the Tunny material had to be done by hand and, although some useful material was gained on the German campaign against Russia and from the links between Italy and North Africa, many of the messages took several weeks to decypher.

  Max Newman, one of the mathematicians working in the Testery, was a thin, bald academic from Manchester University, who like Tutte had worked in Tiltman’s research section. He had been Turing’s tutor at one stage. It was Newman’s suggestion that machines might be able to prove mathematical statements that had led Turing to write his ground-breaking paper ‘On Computable Numbers, with an Application to the Entscheidungsproblem’, and Newman who had ensured that it was published.

  Newman became convinced that, using similar principles to those advocated by Turing, it would be possible to build a machine that, once the patterns of the wheels had been worked out in the Testery, would find the settings of the first row of wheels, thereby making the codebreakers’ task immeasurably easier.

  ‘Newman judged that much of the purely non-linguistic work done in the Testery could and should be mechanised and that el
ectronic machinery would be essential,’ said Jack Good, who worked with Newman in what became known as the Newmanry. ‘He convinced Commander Edward Travis, by then the head of BP, that work on such machinery should be begun, and so the Newmanry was born.’

  Newman went to Wynn-Williams at the Telecommunications Research Establishment in Malvern and asked him to design the machine. It was known as Robinson, after Heath Robinson, the cartoonist designer of fantastic machines, and the first version was delivered to Bletchley Park in May 1943. It worked on the principle that although the encyphering letters were supposed to be random, they were not. No machine can generate a truly random sequence of letters. Robinson compared a piece of teleprinter tape carrying the encyphered text with a piece of tape on which the wheel patterns had been punched to look for statistical evidence that would indicate what the wheelsettings were.

  But while Robinson could clearly do its job, there were problems, Travis told the weekly meeting of senior staff.

  Although Mr Newman’s new research machinery is still going through teething troubles, it is likely to prove better than anything they have yet produced in the USA. The only snag is that it needs a lot of personnel. It should be able to handle 28–30 Tunny messages a day which would be invaluable.

  Robinson was designed to keep the two paper tapes in synchronisation at 1,000 characters a second but at that speed the sprocket wheels kept ripping the tapes. Turing, who, while working on the Bombe, had been impressed by the abilities of a bright young telephone engineer at Dollis Hill called Tommy Flowers, suggested to Newman that he might be just the man to get Robinson to work.

  I came into the project when the Robinson machine didn’t work properly, because it was made almost entirely of telephone parts, telephone switching parts, which was my area. I was brought in to make it work, but I very soon came to the conclusion that it would never work. It was dependent on paper tape being driven at very high speed by means of spiked wheels and the paper wouldn’t stand up to it.