25E Page 167
if there were many weak characters.
The following paragraphs enumerate methods which can be used in difficult cases to make wheels complete and certain.
1. Set all messages, with flogging.
2. Make sure that all wheel-breaking runs for each chi not yet certain have been done using the latest wheels for the other four chis, and that the decibanning is on the basis of these wheels: if the wheels are nearly correct, crude decibanning is permissible.
3. Do a 32-letter count against each doubtful character, and deciban it on the 32-letter count for the whole wheel. This is equivalent to doing every possible short wheel-breaking run separately, but saves time by considering only uncertain characters. It is done easily on Colossus by putting a single pin in the special pattern trigger, and plugging special pattern = cross.
4. Span all messages, looking for slides [25D(c)] and changes in ΔP characteristics [25D(d)]
5. Make a temperate use of wheel characteristics.
6. Make a provisional de-chi on uncertain wheels for Room 41 where it can be treated by non-statistical methods. In an extreme case de-chi on four wheels only.
7. Span /'s on ΔD on a hundred letters immediately before each antopause with faint hope that ΔZ is really Δ key, the P tape of the German Tunny machine having broken. (R5 pp 70, 80)
8. In one instance wheel-breaking was completed because there was a crib into a message already set on four chis: the ordinary crib run failed because of a slide, but running ΔP1,2,4,5 against ΔD1,2,4,5 and looking for/'s in ΔΨ'1,2,4,5 succeeded.
25E SPECIAL METHODS FOR Χ2 LIMITATION
(a) Running against Χ2 crosses.
Because the bulges of runs against Χ2 = X ae so much greater than against Χ2 = ., these are made separately (as in setting), and indeed it is rarely worth while to do runs against Χ2 = ., and then only for good motor cross letters (R3 p 101).
Back to General Report on Tunny. Contents.