The Colossus

its purpose and operation

by Tony Sale

| Back to Fish Index |

Tony Sale's
Codes and Ciphers


The machine age comes to Fish code breaking

The mathematician Max Newman now came on the scene. He thought that it would be possible to automate some parts of the process for finding the settings used for each message. He approached TRE at Malvern to design an electronic machine to implement the double-delta method of finding wheel start positions which Bill Tutte had devised. The machine was built at Dollis Hill and was known as Heath Robinson after the cartoonist designer of fantastic machines.


Max Newman

    The Wynn-Williams proposal.


The Logic Circuits
When Wynn-Williams was asked to produce electronic circuits to implement the double delta algorithm he chose to use a phase modulated carrier from a master oscillator at 25kc/s to perform the XOR logic.

He decided to use 0 and 180 degrees of phase to represent 0 and 1. The elegance of this is that if a "1" causes 180 degrees phase shift, then another 1 returns the phase to zero and thus this implements an XOR function (0 + 0 = 0, 1 + 1 = 0, 0 + 1 = 1, 1 + 0 = 1).

The 180 degrees phase shift was achieved via a diode bridge circuit and a balanced transformer. The biasing of the bridge, + - 10 volts, determined whether the input carrier went straight through (no phase change) of shifted 180 degrees. A triode valve amplifier was included with each bridge circuit to compensate for the losses in the bridge and to give unity gain from input to output.
The output phase at the end of a series of logic circuits was compared with the phase input to the logic circuits in a detector circuit. This gave a voltage output of nearly zero if the input and output are in anti phase or some, much larger, positive voltage if they were in phase. The output voltage from the detector was sampled by a pulse derived from the sprocket hole signal from the tape reader. The result of this sampling, either a pulse if the detector output was positive, or no pulse if the output was zero was then passed to the decade counters to accumulate a count down the whole length of the tapes.


The Decade Counters

There were four decimal decade counters in series giving a 9999 maximum count. The first stage of the decade counters consisted of a ring of ten thyratrons (gas filled thermionic triode valves). The circuit for this was designed by Wynn-Williams before the war for counting in nuclear particle experiments. A thyratron valve will strike and hold an internal arc discharge when there is a positive voltage on its anode and the grid voltage is raised towards the cathode voltage allowing current to start flowing. Once the discharge is started the grid voltage has no further influence over the anode current. Thus the thyratron "remembers" it has been struck and thus acts as a one bit store. Unfortunately the only way to stop the discharge in a thyratron is to drive the anode negative with respect to its cathode. In the decade thyratron ring the thyratron which had been struck had to prepare the next thyratron in the ring to be struck on the next input pulse, but at the same time the next thyratron struck had to cause the pervious thyratron to be extinguished. In the Wynn-Williams circuit this was achieved by coupling successive thyratron's cathodes together with a large capacitor. The thyratron ring was the fast, least significant, decimal counter. The next two counters, the tens and hundreds, used high speed relays with slow speed relays in the thousands counter. The count was displayed on a lamp panel. There were four sets of counters, each of 9999 capacity. The output from the logic circuits was switched alternately into one of two counter sets, the changeover occurring at the end of reading the data on the two tapes. Each tape were joined end to end in a continuous loop. Special holes were punched into the tapes to signify end of data and start of data. The remaining two counters sets were used to count sprocket holes. These counts allowed the calculation of Chi wheel positions for a particular score. Initially all counts had to be read off the lamp panel and written down, a great source of error. Later a special printer known as a "Gifford" printer was added. This was not a great success.


Heath Robinson


Heath Robinson consisted of three parts, the frame on which the teleprinter paper tapes were mounted and read optically, known as the Bedstead, a wide short rack containing the counters, a lamp output panel and later the Gifford printer on a front table, and a tall 19 inch rack known as the valve rack which contained the logic circuits and a jack field panel for plugging up the algorithms. The short counters rack was produced at TRE and the Bedstead and valve rack at the GPO research labs at Dollis Hill to Wynn-Williams circuit designs. The cover name for the project was "Apparatus Telegraph Transmitting", case number 11951. The Bedstead was designed by Arnold Lynch and Eric Speight. Harry Fensom and Alan Bruce worked on commissioning the system at Dollis Hill. There were difficulties in getting the ring modulator logic to work due to extra phase shifts in the circuits when more than six circuits were connected together one after the other. Allen Coombs relates this problem and tells how he went to Tommy Flowers for advice. Tommy Flowers said "change the frequency" which Allen Coombs did. It solved the problem but neither he nor Tommy Flowers knew why.
Eventually it all worked together and Heath Robinson was moved to Bletchley Park.


Heath Robinson was delivered to Bletchley Park in June 1943 and was first installed in Hut 11 which had been the original Bombe room for Turing Bombes, the machines used to break Enigma.

Harry Fensom and Alan Bruce were the two GPO maintenance engineers assigned to Heath Robinson. Two WRNS(Womens Royal Naval Service) ladies at a time were the operators and Jack Good and Donald Michie were the code breakers.

The first problem was teleprinter tape preparation. At least 2000 characters of cipher text was required, joined end to end to make a continuous loop. Then a similar length of Chi wheel patterns had to be punched up and arranged to be just one character longer than the cipher tape. This was to automatically change the relative wheel patterns by one position after each complete run through the tapes.

Then it was found that the optical readers in the Bedstead gave errors if a long stretch of adjacent holes or no holes occurred on the tapes. This meant adjustments to both texts to compensate for this.

A major problem was keeping the two tapes in synchronism at over 1000 characters per second. Originally the sprocket drive cogs were motorised but this proved impossible to sustain without tearing the tapes and a friction drive was used from the paper tape pulleys with the sprocket shaft just idling to keep synchronisation. This proved to be better but there was still a problem with tape stretching in the distance between the sprocket cogs and the optical reader aperture.


Heath Robinson worked well enough to show that Max Newman's concept was correct. Newman then went to Dollis Hill where he was put in touch with Tommy Flowers, the brilliant Post Office electronics engineer. Flowers went on to design and build Colossus to meet Max Newman's requirements for a machine to speed up the breaking of the Lorenz cipher.



This page was originally created by the late Tony Sale
the original curator of the Bletchley Park Museum