Bell Telephone Laboratories relay computing devices

From Computer History Wiki
Jump to: navigation, search

The Bell Telephone Laboratories relay computing devices were a series of digital computing devices, using relays for both logic and memory, all architected by George Stibitz at Bell Laboratories in the late 1930's and early 1940's, and built with the design assistance of Samuel B. Williams, C. E. Boman and E. G. Andrews, and used on a variety of computing problems. They grew in capabilities, complexity, and size from the earliest one (which seems to have been a sui generis invention by Stibitz), through the rest of the series.

They were binary internally, suitable for the relay technology of which they were built, but could do I/O in decimal, for the convenience of their users. They included:

Model I - Also called the 'Complex Computer', it was a calculator capable of performing arithmetical operations on complex numbers; these are much used in analog electrical engineering (a major field for engineers at Bell Labs), in which addition is relatively easy to do by hand, but multiplication is not. To ease the conversion of large numbers it used binary-coded decimal internally, in 'plus-three' form - i.e. 0 was represented as '0011', and 9 as '1100'. It handled operands of up to 8 decimal digits, with 2 extra digits internally to minimize round-off errors.

It used 450 relays, and was completed in October 1939, and placed in use on 8th January, 1940. It was remotely demonstrated at a meeting of the American Mathematical Society at Dartmouth College in Hanover, New Hampshire on 11th September, 1940; the machine remained in New York, but was connected to an I/O station at the meeting over telephone lines. (A predecessor of the 'Mother of all Demos' of the Augmentation Research Center!)

Model II - Also called the 'Relay Interpolator', it was a specialized machine built to solve fire control problems; it was completed in July, 1943. It (and all following Models) used bi-quinary internal representation, adopted for its error-checking capabilities. It was programmable, with the program stored on a paper tape loop; it also used paper tape for I/O. It was slightly smaller than the Model I; 440 relays.

Model III - Also called the 'Ballistic Computer', it also was used for fire control problems; it was completed in June, 1944. It was a more general-purpose machine than the Mark III; it included a number of novel features, such as the use of some registers as index registers. It was considerably larger than the Model II; 1,400 relays. It (and all following Models) could operate in un-attended mode.

Model IV - Very similar to the Model III, it could also perform trigonometric functions. Completed in March, 1945, it was about the same size; 1,425 relays.

Model V - Two built (one delivered in December, 1946; one in August, 1947); architecturally, it was a generation past the Models I-IV. It was designed to have up to 6 "computers" (what today would be called ALUs); as soon as one was idle, it automatically started the next "problem". It only had 2, as built; it used one paper tape reader for input data, up to five for instructions (which allowed considerable easy use of subroutines), and six for table data. It had support for floating point (the first Bell relay machine to do so), using exponential notation. It was considerably larger again; 9,000 relays.

Model VI - Similar to the Model V, but somewhat simpler; it had another novelty: each of its three paper tape readers could hold either data or instructions. It was considerably smaller; 4,600 relays; its designers had apparently decided that the increased powers of the Model V were not worth the extra expense, or attendant complexity. (Perhaps the very first case of Second System Syndrome?)

The Models II-V were produced for the armed forces (all remained in use until around 1960); only the I and VI were used by Bell. The bi-quinary machines were very reliable (in the sense of never producing erroneous results), even though they used relays, which were inherently un-reliable; the continuous self-checking possible with the bi-quinary representation was responsible.

See also

Further reading

  • George Stibitz, Computer, in Brian Randell (editor), The Origins of Digital Computers: Selected Papers, Springer-Verlag, Berlin, Heidelberg, New York, 1973, 1982 (3rd edition)
  • George Stibitz, Early Computers, in Nicholas Metropolis, Jack Howlett, Gian-Carlo Rota (editors), A History of Computing in the Twentieth Century, Academic Press, New York, 1980
  • E. G. Andrews, Telephone Switching and the Early Bell Laboratories Computers, BSTJ Vol. 42(2), 341, 1963 - the author was present at Bell Labs from the 1920's on
  • O. Cesareo, The Relay Interpolator, in Brian Randell (editor), The Origins of Digital Computers: Selected Papers
  • Chapter 4: Number, Please - Computers at Bell Labs, in Paul E. Ceruzzi, Reckoners: The Prehistory of The Digital Computer, From Relays to the Stored Program Concept, 1935-1945, Greenwood, Westport, 1983
  • Paul E. Ceruzzi, Chapter Six: Relay Calculators, in William Aspray (editor), Computing Before Computers, Iowa State University Press, Ames, 1990