Hollerith Electronic Computer

The Hollerith Electronic Computer (HEC) was produced by the British Tabulating Machine Company (BTM) and was based on a design by Professor Andrew Booth of Birkbeck College, London. It was Britain’s first mass-produced business computer. The prototype first worked in December 1951.

In 1950 John Womersley, who had previously led the team developing the Automatic Computing Engine (ACE) at the National Physical Laboratory (NPL), joined the Unit record equipment company, BTM. He recognised that there was a need for smaller inexpensive computers and recruited Andrew Booth as a consultant to develop such as machine. Booth had previously worked for the British Rayon Research Association (BRRA) before moving to Birkbeck College in 1945. The BRRA had sponsored him to develop what became the All Purpose Electronic Computer (APEXC). He needed punched card input and output technologies and struck a deal with BTM, whereby they supplied him with these in return for their copying the machine that he was developing, including its magnetic drum memory. In March 1951, BTM's Dr Raymond 'Dickie' Bird with Bill Davis and Dickie Cox were dispatched to Fenny Compton in Warwickshire where Booth lived and where, in a rotting barn, he was developing the prototype of his machine.

Dr Bird and his team built a copy of Booth's machine in the BTM premises at Icknield Way Letchworth, which they called HEC 1. It was 1.5 m high by 3m wide by 0.5m deep and used simple circuits, with approximately 1000 ex-Government thermionic valves (vacuum tubes) mainly 6J6s which were B7G-based double triodes. The memory consisted of a 5.5 inches (140 mm) diameter, 1 inch (25 mm) wide drum rotating at 3000 rpm containing 32 tracks each storing sixteen 32-bit words giving a total of 2 kilobytes. The drum had a special track from which the electronics were clocked. The machine had an accumulator and a multiplier register which were arranged to allow double length multiplication. As the multiplicand was repeatedly added to the product, it got longer while the multiplier got shorter so that the product could fill the accumulator and then continue into the multiplier register. Multiplication took up to 640ms for a 32-bit multiplier, which needed 32 drum accesses.

...
Wikipedia