Electronics give birth to the computer

Programmable computers had been at least a century in the making before they finally became a reality in the 1940s.

The 19th-century engineer Charles Babbage grasped the concept, but lacked the technology to do it justice. His contemporary, Lincolnshire-born George Boole, developed the Boolean algebra that underlies computer logic.

Alan Turing invented the modern computer in abstract in a famous paper in 1936, by which time technology was just about up to the task. Many elements of the computer were already used in advanced telephone exchanges, where electro-mechanical switches implemented Boolean logic and functions such as counting.

By the end of the 1930s, it was realised these switches could be replaced by much faster electronic valves. One of the few experts in this field was doing work for Bletchley Park. Tom Flowers, head of switching research at the Post Office Lab in London’s Dollis Hill, believed a valve-based machine should be used to crack the Nazi high command’s Lorenz cipher. But Bletchley had doubts because valves had a tendency to blow.

Flowers persevered because he knew they had a much lower failure rate when the heater power was turned down and they were never switched off. For a full explanation of how codebreaking worked, see section entitled 'Find out more'.

The initial task of Colossus involved performing XOR operations on the data stream of an encrypted message and a loop of the masking data generated by the 12 cogwheels of the Nazi’s Lorenz enciphering machine. This masking stream, which was changed relatively infrequently, had been obtained by other methods, but later versions of Colossus could deduce this. Colossus was built to establish the cogwheel start positions, usually changed with each message.

Photoelectric sensors read the encrypted message from a looped paper tape at 5,000 five-bit characters per second. The tape also carried a continuous row of holes that provided a clocking pulse for the entire machine.

In Colossus’ largely mechanical predecessor, the masking characters were also fed from a second paper tape. Flowers’ first great innovation was to generate them electronically using chains of special valves called Thyratrons that switched on and off in sequence, avoiding the considerable problem of keeping two long paper tapes in sync.

There was a separate path for each of the five bits of the standard Baudot teleprinter code used to define characters. These bit streams were dealt with separately, because each was affected by a different set of Lorenz wheels. First the Bit 1 stream was XOR-ed with itself, shifted by one character, to produce what was called the delta because it tracked changes. This stream was then XOR-ed with the masking Bit 1 stream and the number of ones or noughts counted and printed out.

The process could be repeated for all possible starting points in the masking sequence, though this was not usually necessary. The distribution of noughts and ones in these streams was more or less random for all positions except the one used for the actual message. The same procedure could be applied to the other bit streams.