A very early real time digital computer, one of the first (ran a program a year before Whirlwind did). One built, one sold. "The 152’s plug-in units were based on printed circuits on glass plates, with plated-through holes, deposited resistors, semiconductor diodes and sub-miniature pentodes. This cutting-edge technology unfortunately became a bleeding edge: the 152’s reliability left much to be desired and the MRS5 contract was eventually curtailed by the Admiralty."
- http://www.cs.man.ac.uk/CCS/res/res42.htm#i
- from this book review
I looked into this as a result of a misdirection, but it's interesting anyway. The machine was to control a battleship's anti-aircraft guns, which means responding to radar in millisecond timeframes, and normally means doing trigonometry too. But there was no time for that. So, the ten-inch optical encoding disks which were read by a photocell were not merely calibrated in fractions of a degree, but were calibrated as trig functions of the angles - the encoding disk acts as a physical lookup table directly. As it happened, they patented this great idea but missed a trick so it was widely used without bringing in revenue.
There's a big expensive book all about Elliott, some of which can be read online:
The Elliott 152 Computer
Technical Details of the Elliott 152 and 153
There's a paper about the 152 and other early Elliott machines here:
Systems architectures for the Elliott 152, 153 and Nicholas computers.
Quote:
The first version of the 152 computer was completed in September 1950. It contained a fast multiplier; a single-word accumulator made from a circulating loop of delay elements with facilities for addition, subtraction, shift and collation (AND); and a double-word (40-bit) accumulator. Data transfers were bit-serial. Arithmetic and logic circuits were based on sub-miniature pentodes (vacuum tubes). Storage was provided by 64 words of CRT (electrostatic RAM) memory, using the anticipation-pulse system invented by F C Williams
As with all early Elliott machines, it was bit-serial - static storage was expensive, even for a few bits, and serial storage technology was available and even favoured (mercury lines, nickel wires, cathode ray tubes, disks.)
The machine operated on fixed-point numbers and had a fast multiplier. It had an overflow bit within the data word, which might be a very nice idea compared to a global overflow bit in a status register. Intended to run at 1MHz, it only went up to 333kHz, which became a standard speed. A multiply took 60 microseconds.
A bit more info here:
http://www.chilton-computing.org.uk/acl ... e/p033.htm
(See the bit starting "I should now like to describe some of our work at Borehamwood.")