15C Page 35
arrived in November, just in time for the move to more adequate quarters in Block F. The original Pilot model, which was by this time completely worn out, was thereupon abandoned. Some of the later Robinsons had four 'bedsteads' enabling complicated runs to be done without special tape-making.
Meanwhile Colossus I was delivered in February, 1944, and immediately sent up the output to more than twice its previous level. Colossus was entirely the idea of Mr. Flowers of Dollis Hill. His original scheme was to set up the message, as well as the wheels, on valves but this was given up when it was realised that messages of 5000 or more would be wanted. The combination of one tape, carrying the message, with wheel patterns set up electrically, gave nearly all the advantages of the pure valve machine with a great saving in valves and in setting time. The advantages of this machine over Robinson were (1) Its speed, a factor of 25/2 when 5 counters were available on all chi-runs: (2) The absence of inertia which enabled a run to be stopped at any moment and the wheels switched to assigned settings. (3) The great reliability resulting from the use of valves throughout, instead of relays and the abolition of synchronised tapes. A preliminary order for four further Colossi was placed in March, 1944, increased to twelve at the end of April. The order for Robinsons was curtailed. Great pressure was put on Dollis Hill to deliver the Colossi quickly and they promised on the 14th March to have Colossus 2 (i.e. the first production model) working by 1st June. This promise they fulfilled. Colossus 2 came into action on 1st June at 0800. The remaining Colossi followed at the rate of about 1 a month. A new building (Block H) was erected to house Colossi 5 to 11. Its plans were approved on 25th May, 1944 and it was ready for occupation on 17th September. Work on assembly of Colossus 11 had started on 8th May, 1945 and was stopped (before completion) a few days later.
(c) Staff expansion.
The machine expansion was accompanied by an expansion of Newmanry staff which finally amounted to 272 Wrens and 27 men. The organisation had to be correspondingly elaborated, mainly by the multiplication of Registrars to keep track of tapes and jobs in their travels round the Newmanry, and to keep in touch with Major Tester's Section.
(d) Reallocation of work between the two sections.
The original paper schemes for machines processes proposed the setting of all 12 wheels by statistical methods carried out on Robinson. The Motor was to be set by running the motor-pattern against strokes of ΔD, and psi wheels could then be run against ΔD, 'contracted' by missing out letters opposite motor dots. The tunny machine had a special contrivance for making this contracted version.
This programme was actually carried out for some months, until it was realised that, given a de-chi, it is possible to set the motor and psis by 'language' methods. This work was done in Major Tester's section, and a convenient division of work and utilisation of available resources resulted. With the introduction of the P5 limitation this division became a necessity, since, on the one hand, chis could no longer be set on depths, and on the other, de-chis could no longer be 'contracted' on Tunny. The division of work on chis and psis necessitated a close co-operation between the two sections, and an important step was the setting up of the joint registry. With the switch over to Colossus, complete setting on all 12 wheels by machine again became possible, and when at the end of 1944 Colossi began to be plentiful a large proportion of messages were completely set by machine methods.
With the introduction of the P5 limitation it became necessary to break the chi-wheel-patterns statistically from ΔD. As long as the wheels changed only once a month this could be done without seriously interfering with the normal setting organisation, and with the use of only about two Wrens a shift