In the spring of 1944 a visitor to a hut at Hanslope Park, a country house in Buckinghamshire used by the Radio Security Service, would have found two men surrounded by valves, resistors and a mass of wiring. One was a young electronic engineer named Donald Bayley. The other was Alan Turing, who by that point had done more than any individual to break the German naval Enigma, and who was now learning to solder.
The machine they were building was called Delilah. It encrypted speech, it was roughly the size of three shoeboxes, it worked, and it was never used for anything.
Why Speech Was the Hard Case
Enciphering text is a problem about discrete symbols. Speech is a continuous, rapidly varying electrical signal, and there is far more of it: a few seconds of talking carries more raw information than a long telegram.
The scramblers in use at the start of the war worked by rearranging the signal. The A-3 system used for transatlantic telephone calls between Churchill and Roosevelt split the audio into bands, shuffled them, and inverted some of them. It sounds like nonsense to a casual listener, and it is not encryption. The German Post Office had a station at Eindhoven that unscrambled those calls routinely, sometimes within minutes, and transcripts went to Berlin.
Turing knew this because he had been sent to the United States in late 1942 and spent two months at Bell Laboratories in New York evaluating exactly that problem. It was there that he saw the American answer, and where he took tea most days with Claude Shannon, who was working on the same system and thinking about the mathematics of information. Shannon’s theory of information and Turing’s ideas about computation were being discussed across a canteen table by two men who could not tell each other what they were actually working on.
What the Americans Built
Bell Labs produced SIGSALY, the first genuinely secure voice system. It sampled speech, measured the energy in ten frequency bands plus pitch information, digitised those measurements, and added a random key. The key was the crucial part: a one time pad of thermal noise recorded onto matched pairs of vinyl phonograph records, one for each end of the link, played in synchrony and destroyed after use.
It was unbreakable and it was enormous. A SIGSALY terminal weighed around fifty tonnes, filled a large room, required its own air conditioning plant, and needed a crew of specialists. Twelve of them were built. The one in London sat in the basement of Selfridges.
Turing’s assessment was that the principle was right and the engineering was absurd for any use outside a fixed headquarters. A commander in the field, a ship, or an aircraft could not carry fifty tonnes of vacuum tubes and a turntable.
The Design
Delilah was his answer, and the specification he set himself was portability and strength at the same time.
The approach was to treat the speech signal as a sequence of samples and add a key to each sample, which is the same logic as a one time pad applied to sound rather than to letters. The receiving end subtracts the identical key and recovers the speech.
- Sampling. The speech was sampled at 4,000 times a second, each sample becoming a pulse whose height carried the signal
- The key. Rather than a phonograph record, the key stream was generated inside the machine by a set of multivibrator circuits, driven by a starting configuration set on plugs and switches. Both machines set to the same configuration generate the same stream. This is a key generator rather than true random noise, and it is the ancestor of the stream ciphers used in digital communications
- Synchronisation. The two ends had to stay locked together to within a fraction of a sample, and getting that to work over a real circuit was the hardest engineering problem in the project
The result fitted into three units of about a foot and a half each. Compared with fifty tonnes, that is a different category of machine.
Turing With a Soldering Iron
The collaboration is documented mainly through Donald Bayley, who was in his early twenties and had a proper engineering degree, and who left an account of it decades later.
Turing’s theoretical work was, as Bayley put it, faultless, and his practical work was another matter. He had a habit of designing circuits on paper with an idealised view of components, and Bayley spent a good deal of time explaining why real valves and real resistors did not behave that way. Turing wired his own circuits, badly at first, and improved. He also insisted on doing his multiplication in an eccentric way and refused to use standard engineering notation.
Bayley also recorded that Turing arrived at Hanslope with a bicycle whose chain came off at a predictable number of revolutions, which he counted rather than repaired, and that he had buried two bars of silver in the woods against post war inflation and was never able to find them again.
The demonstration piece was a recording of a Churchill speech. Passed through Delilah, the encrypted output was noise. Decrypted at the other end, the speech was recognisable and intelligible, with some distortion.
Why It Was Never Used
The machine worked by early 1945, and by early 1945 the war in Europe was ending. There was no procurement process eager to adopt an unfamiliar system from a small research station, no established requirement it had been built to satisfy, and the people who might have championed it had other priorities.
The quality was also marginal for military voice traffic. It was intelligible rather than clear, and a commander shouting over a bad line has little patience for distortion.
So Delilah was written up and shelved. The report Turing and Bayley produced stayed classified for decades and only became publicly available in this century, which is why the machine is absent from most accounts of his career.
What It Fed
The Delilah years are usually treated as an interruption between the codebreaking and the computers. They are better read as the bridge.
Turing arrived at Hanslope as a mathematician who had designed the Bombe as a logical machine and left it to others to build. He left as someone who had personally designed, wired and debugged pulse circuits at high speed. Within a year he was writing the proposal for the Automatic Computing Engine at the National Physical Laboratory, a stored program computer specified down to the circuit level, with a fanatical concern for efficient hardware.
The thread also runs the other way, into the theory. Sampling a continuous signal, quantising it, adding a key, transmitting it as pulses and reconstructing it at the far end is the digital treatment of an analogue quantity, three or four years before anyone spoke about digital communication as a field. The 1950 paper on machine intelligence is the work of a man who had spent the previous decade making abstract machinery physical.
Reading the Original Documents
Turing’s wartime writing is not what most people expect. It is direct, occasionally funny, impatient with ceremony, and concerned throughout with getting something to work rather than with presenting a finished result.
The clearest example is his own treatise on the Enigma, written to explain the machine and the methods to newcomers at Bletchley Park, which circulated internally under a nickname taken from the author. We publish it as The Prof’s Book, in an edition made for reading rather than for archives: the document as he wrote it, with the working shown.
Delilah has no equivalent published edition. It exists as a technical report, a set of circuit diagrams, and the memory of a young engineer who spent two years in a hut with a man he described afterwards as the most original thinker he ever met, and the worst at soldering.
No products in the cart. 