The Special Fish Report
Albert W. Small (December 1944)
Codes and Ciphers
TOP SECRET Special Fish Report Page 55
The 14th character of Delta-X3 was considered most likely to be
doubtful. To check delta.t, a 32-letter count was made on message SB 3037
part 1, considering the doubtful character to be a dot; and then another
considering it to be a cross. (I.E., the regular wheel-breaking type of
run.) The result was to indicate character 14 as "20. db. down,"
that is a "cross," whereas in the same message earlier, run 10,
this character 14 had been shown to be "3 db. up." The operator therefore
removed the three db. up and added in the 20 db. down, or in effect
added in 23 db. down, resulting in the final score of 11 down as
shown clearly on the Delta-X3 worksheet.
The final "good" 32-letter count used in the above determination
was then sent to Testery along with the dechi, as per custom. Therefore
no exhibit is available to show the final step in clearing up Delta-X3.
Probably the reader has "skipped through" the wheel-breaking
exhibits. Yet they are very important, because they demonstrate almost
all the practical elements of decibanning, and answer so many questions that exist
in the cryptanalysts' minds at Arlington. Take for example the problem of
combining patterns from different rectangles. It isn't done here.
They take pattern from one rectangle, strengthen the wheels, and then set
them in other messages. If they can't seem to get anywhere, they use
another rectangle and eventually try its wheels back in the first message.
The effect is almost the same. They have found out by experience that
several rectangle patterns can't be slid and merged anyway unless at least
one rectangle is significant; and if it is significant the Colossus
methods used make it unnecessary to add all the other ones to it at the