An unusual magnetic rod ROM memory

This page 2009 Andrew Wylie all rights reserved

This page shows a rare and unusual early computing device: a magnetic rod ROM memory. At least, I think that is what it is. I am indebted to Paul Symons for providing this extra-ordinary piece of technology. He informs me that it probably came from a UK military computer, but he does not know which.

If you know anything about this, or can say definitely what it is, or what it is not, please

magnetic ROM general view The exterior of the module reveals that it comprises two circuit cards bolted together into a single assembly, about 9 inches deep by 8 inches high, with a central thick aluminium plate. Two smaller edge connector cards (each divided into three groups of 'fingers') are fixed along one side and connected to the two main cards by wires, some of which are glued on the surface of thin flexible cards. The assembly therefore slid into two slots of a card cage although it has no handles or tags on the outside edge. The two sides have thin aluminium plates protecting part of them. A double row of diodes, totalling 68 per card, is present along the interior bottom edge of the two main cards, connected to the edge connectors by soldered wires with black insulation (17 on one side, 15 on the other). Closer examination shows that the diodes are type CV7049, a British military germanium switching diode equivalent to Mullard OA10, date coded SL and SM which means November and December 1961. The two cards are connected together along the back of the unit by a thin flexible insulating card carrying 50 wires.

magnetic ROM with plate removed magnetic ROM rods close up Removing the aluminium plate on either side reveals a grid of ferrite rods, which look like pencil 'leads', numbering 29 rows by 68 columns. These are odd numbers, but further investigation shows that not all were used.

drive/read wires Undoing the circlips and removing the boards holding the ferrite rods reveals that each of the two main 'cards' is in fact made from three superposed pieces clamped together. The exterior piece holds the ferrite rods, on the interior side it has wires associated with them which run up the columns of ferrites from the diodes. The wires change side after every pair of rods, thus making about an eighth of a turn around each rod, and are completely looped around the last rod and so return to the diode end, where every eighth one is wired together and connected to a finger of the edge connector (see image above of the reverse of the card). Only 64 columns are wired, the exterior 2 on each end are not. (You must rotate the image through 90 degrees to obtain the same orientation as the others).

the central card The central piece in the card assembly has no wires connected to it. Instead it is a thin flexible printed circuit comprising 25 copper rows, each made up from 64 connected loops. Each loop is either complete or cut at one side.

drive/read wires The third, interior piece of the card assembly also holds a grid of ferrite rods, identical in number to the first card. However, the card has wires along the rows. Only 25 rows are wired, the exterior two on each side are not. However, on one card only 12 rows are connected to the card edge connector (two wires per row), the other 13 rows show that the two wires per row are in fact one wire that loops around the penultimate rod and returns. All 25 rows have both wires connected through the rear interconnect to the other ferrite row card, which has the 13 rows connected to the edge connector on that card, and the other 12 looped back. The whole assembly is therefore an array of 25 by 128 cells rather than two arrays of 25 by 64.

We now come to the question of what this device is. It is almost certainly a magnetic read-only memory (ROM). Some early computers used such devices, often made from so-called "core rope", which used physical winding around ferrite core transformers to provide presence or absence of a signal, and hence a bit. Such devices were usually word-oriented.

serial numbers This device is more sophisticated. I believe that it stored 128 words of 25 bits (probably 24 data plus 1 parity). The diodes provide drive line decoding. Each drive line is coupled to the 25 bit-output lines by transformer action of the ferrite rods. The central circuit card either allows or prevents this happening per bit, depending upon whether or not the copper loop is complete. As the drive lines only make about an eighth-turn around each ferrite, signals would have been weak, and sensitive read amplifiers would have been needed. I do not have any of these, or any knowledge of the device's manufacturer. However, I suspect that it may be a military offshoot of the Ferranti Argus computer, which had a programmable magnetic ROM system. The 20th issue of Computer Resurrection, the bulletin of the UK Computer Conservation Society, describes a military version of this. However, there are major differences from my device: the article states that the military machine used "square coupling loops", and that each card contained 12 by 25 loops i.e., 12 words.

My device bears no identification other than some numbers shown in this image. It is possible that the AP number is an 'Admiralty Pattern' number, which identified UK naval equipment. If you have any knowledge of this device, please