There had to be some variability in the circuits, or else the generator would produce the same noise over and over again.
This was allowed for by making the interconnections required for the combination of the eight multivibrator outputs to pass through the wirings rather like those of an Enigma, with rotors and a plugboard.
So a setting of this 'Enigma' would serve to define a particular sequence of key, in a way that both sender and receiver could agree upon in advance.
With the rotors fixed in position, the key would not repeat itself for about seven minutes.
In practical operation, speech in one direction of transmission could be limited to this time and a new key sequence started off on reversing the direction of transmission.
This could be done simply by stepping on the rotors.
There were enough rotor and plugboard positions for the resulting system to be as safe—according to his theory—as a genuinely random one-time key.
Getting the Delilah system to work as a whole was a job which pushed their resources to the limit.
The system was useless unless sender and receiver could keep their multivibrators in time to the microsecond.
They spent most of the first half of 1945 in achieving the necessary precision.
They also had to test the output of the Delilah key generator, when they had built it, for the evenness over the frequency range which the calculations predicted for it.
It was typical of the conditions in which they worked that they had no frequency analyser.
Alan would have seen one at Bell Labs, and there was one known to exist at Dollis Hill, but at Hanslope they had to make one for themselves.
It provided a challenge of the desert island sort, which Alan as usual enjoyed.
After a lot of work they had a device, but when they first tried it out, Alan had to confess 'It's a bit of an abortion, isn't it!'—so they called it the ABORT Mark I.