The code was superenciphered with a so called additive, which was printed on a sheet - changed daily - having ten
columns, and 30 rows with random 5-figure groups. Each row on the sheet was marked with a 3-figure number, and each column
with a 2-figure number.
The first few rows of a sheet could look something like this:
| 04 | 09 | 15 | 29 | 33 | 50 | 58 | 64 | 82 | 96 | |
| 017 | 58201 | 10035 | 57184 | 00472 | 33727 | 18483 | 10622 | 37310 | 74512 | 23295 |
| 023 | 67844 | 98944 | 84216 | 17402 | 90246 | 19774 | 97823 | 98151 | 73785 | 04471 |
| 027 | 44380 | 53478 | 05713 | 75739 | 48921 | 60935 | 99042 | 63856 | 02744 | 43826 |
The code clerk chose a group on the sheet at random as starting point, and formed a 5-figure indicator
group by combining the 3-figure row number with the 2-figure column number. Using the mock-sheet above, and starting
on row two, column two, would result in the indicator group 02309. This group was later hidden in the final telegram.
Starting at the group chosen on the sheet (=98944 following the example), the clerk then proceeded to write out one group from the sheet under each group
he had looked up in the code book when doing the basic encoding of the plaintext. Then the groups were added together, figure
by figure, using non-carrying addition ignoring the "tens" (i.e. 8+4 equals 2, not 12), the result of this operation being the
final cryptogram.
Let's say our first group from the code book is 82697 meaning Warship "Karl Marx". Adding 98944 to this
group by non-carrying addition would result in 70531.
The system was further complicated in March 1941 by performing a double superencipherment. The code clerk chose two, different starting points in the series on the sheet, and thus two numbers were added to each code group. This posed a formidable obstacle to cryptanalysis, and the new system was not broken until June 1942, more than a year later.
One final note: If you study the code groups in the scan below, you will notice that the first two figures of every group is either both odd or even, a fact which aided cryptanalysis somewhat.
