The Newmanry was headed by Max Newman and was responsible for the mechanised methods of codebreaking. Heath Robinson and then Colossus were used by the Newmanry to speed up the hand codebreaking methods of the Testery. The Newmanry was based in Block F with the Testery; however, when more space was needed for Colossi , Block H was built as an annex to the Newmanry.
Heath Robinson was developed to find the Lorenz chi wheel settings. However, the two tapes often stretched causing them to be out of sync and therefore unreliable.
The solution to this was Colossus, which is recognised as the world’s first electronic, semi-programmable computer. Colossus was designed and built by Tommy Flowers a GPO engineer at the Post Office Research Station in Dollis Hill. Tommy had been involved in the pioneering use of valves, which he incorporated into the design of Colossus.
This enabled Colossus to perform millions of operations per second as opposed to the thousands of operations per second performed by metal contact switches. The Colossi replaced Heath Robinsons for finding the chi wheel settings.
Colossus Mark I arrived at Bletchley Park in January 1944 but this was quickly replaced by Colossus Mark II, which contained 2500 valves, achieved a reading speed of 5,000 characters per second and a processing speed of 25,000 characters per second. Colossus Mark II had its first successful break on Lorenz in February 1944.
Colossus performed a statistical attack to find the Lorenz chi wheel settings. The message length (i.e. the number of characters) was the sample size and Colossus used this to put into context the results of its calculations.
Colossus went through all combinations of two chi wheels against the intercepted message. This was done by adding two impulses of the message to two impulses of the chi wheel combination for every position of the message tape.
Results were then counted for each position, with the current answer being compared to the previous answer. This comparison looked for any correlation based on the premise that when the chi wheels are in the correct position, more dots are produced than crosses.
For example, if your message length is 1000 characters, the average you would expect to get if the dots and crosses were evenly distributed would be 500. So, if your count was 506, this would be an expected result and the correct chi wheel settings have not been found. If your count was 493 then again this was about average so correct chi wheel settings have not been found.
If, however, your count is 658 then this would mean that the chi wheel settings for Chi Wheel 1 and Chi Wheel 2 have been found. This was the statistical method developed by Bill Tutte. The remaining settings were found by hand in the Testery.