Page 2 of 3
Towards the Differential Analyzer
Bush wanted to solve equations that related to power line transmission. These were differential equations that took months to solve. Bush decided that a quicker route to the solution was an analog computer. The MIT electrical engineering department spent the best part of 20 years on the project!
Bush first suggested the idea in 1925 and his students worked on the problem. They jointly invented an electromechanical multiplier and integrator.
By 1928 they had a very basic machine working and this proved that is was possible. Bush had lots of help with ideas from his students and it does seem that it was a joint invention but there is no doubt that it would not have worked without Bush guiding the project. At each step he attempted to replace mechanical components with electromechanical ones.
First “thinking machine”
At the end of 1928 he had funds to build an even larger machine. This was a big machine - almost as big as a digital machine. It had servo mechanisms where possible to make the mechanical components more accurate. This is the machine that Douglas Hartree saw and copied back in Manchester - only he used Meccano on a budget of £20.
The complete Differential Analyzer
Today we think of the MIT Differential Analyser as a number cruncher of a very specific and technical nature but at the time it was hailed as the first “thinking machine”. The New York Times ran an article with the headline “Thinking Machine Does Higher Mathematics”.
In 1936 Bush delivered a paper to the American Mathematical Society entitled 'Instrumental Analysis'. It discussed the work of Charles Babbage and his attempts to build an analytical engine. Bush thought that by linking together some IBM punch card machines under the control of a central programmer he could build a close approximation to Babbage's machine. It was clear that Bush had understood Babbage’s work and the idea of programmability.
Rapid Arithmetic Machine
Soon after Bush started to work on an electronic digital machine called the Rapid Arithmetic Machine. There is evidence that he documented the design in a series of papers written between 1937 and 1938 but no trace of these can be found. The machine was based on paper tape storage. It used three paper tapes - one for the data, one for the program and one as a sort of ROM. The program tape would be read repeatedly to carry out operations on each item of data on the data tape. There was apparently no suggestion that the program would use numerical addresses nor was their any provision for conditional branching. Given the lack of detailed information, it is difficult to be sure exactly how it worked. If the machine had been finished then the need for some of these facilities would have become apparent as soon as it started work on real problems!
All the registers, logic and arithmetic units were electronic and were built using valves. It is difficult to be sure how much of the machine was built. They certainly built modules to prove the various designs needed for a working computer.
A counter - part of the Rapid Arithmetic Machine
The design was decimal and the counters and registers all used ten-state storage rings. To store a single bit in a binary machine you only need a two-state device. To store a digit in a decimal machine you needed a ring of ten valves which automatically turned on and off in sequence - so counting from 0 to 9.
A decimal ring counter for the Rapid Arithmetic Machine
In 1940 Bush estimated that the machine would be able to multiply two six-digit numbers in about .2 seconds and would run at a clock speed of .01MHz. A lot of time was spent trying to devise gas filled valves to reduce the component count and then the work just stopped. The reason was that the design team was claimed for work on the atomic bomb.