24C Page 120
Control Panel Multiple test switch to Χ1
Check depth, i.e. Χ1 = x, Χ2 = x
Rectangle switch to "Normal".
Rectangling gadget Carriage return on Χ1
Switch in appropriate depth.
Plug Panel ΔZ1 + ΔZ2 = . in all counters.
Settings Χ1 = 06, Χ2 = 02
After setting wheels return plugs to 01, 01, without resetting.
Step Χ1 (lower switch down) fast to control Χ2 slow (lower switch up).
Printer Paper of sufficient width, start at extreme left.
Final Checks Repeat first and last rows.
Unfortunately, although the rectangle is produced in its final form, it was in practice found necessary to transfer it by hand to squared paper in order to converge it, so that the advantages of this method are less than would be supposed.
24C CRUDE CONVERGENCE
(a) The general idea of convergence of a 1+2 rectangle is to find wheels ΔΧ1, ΔΧ2 which agree as well as possible with the entries in the cells of the rectangle.
The interpretation of 'agreeing as well as possible' is not obvious nor in Crude Convergence the only convergence which has been contemplated.
In a sense the evidence would be better represented, not by ordinary wheels of dots and crosses, or say ±1, but by generalised wheels in which the magnitude of a character is proportional to the evidence in its favour. It would be possible to work in terms of generalised wheels and finally convert into ordinary wheels by taking each character as dot or cross according to its sign. There is some evidence that the particular method known as 'accurate convergence' is more reliable than crude convergence. For references to other proposed method see 24W.
(b) Crude Convergence The only form of convergence used operationally is crude convergence which uses only ordinary wheels of dots, crosses, and doubts to make the bulge of ΔD12 = . a maximum.
Back to General Report on Tunny. Contents.