Hand-wired core for ROM?
rickb at bensene.com
Sat Apr 11 09:00:37 CDT 2015
> That is amazing! The core rope memory systems I've read descriptions of
> have used far fewer addresses per core. For a 2Kx42 ROM with 1K addresses
> per core, that would only need a maximum of 84 cores (and perhaps only 42
> if the distribution of ones and zeros was uniform enough for each bit of the
> word), but they'd have to be fairly large cores. I'd have expected it to be
> more cost effective for manufacture to use a much smaller number of
> addresses per core, which would allow use of smaller cores, ease the
> manufacturing, and probably have better SNR of the output pulses.
> However, I'm sure the engineers at Wang knew far more about it than I do,
> so they undoubtedly chose the most cost-effective design that met their
> reliability objectives.
The Wang 700/600/500-series ROMs weren't really core rope. They were, as Brent pointed out, more a form of pulse transformer memory.
The ferrite element in these ROMs is horseshoe-shaped, and inserted in a plastic form that has two hollow tubes, one for each "leg" of the horseshoe. One tube has a specific number of coils of wire wrapped around it, and serves as the secondary for the "transformer". The other plastic tube has myriad wires wrapped around it, either clockwise or counter-clockwise. When current passes through one of these myriad wires, it creates a pulse in the secondary coil, either positively- or negatively-going, depending on which direction the wire was wound around the primary tube. You can see details about the construction at http://oldcalculatormuseum.com/wang720.html. Scroll down about 7/8ths of the way through the page and you can see the details.
I built a device that reads the content of these ROMs, both to troubleshoot them, as well as to capture the microcode for the machines so as to write an emulator that allows the code to be run on more modern hardware.
It's quite amazing, as inspecting these ROMs carefully, it can be seen that there are cases where wires broke in the weaving process, and were simply dipped in lacqure at the end (to prevent shorts) and would around one of the wire guides to keep it out of the way. There are also splices done, where the tiny wire was joined with a tiny bit of solder, and then the joint covered with lacquer. The patience involved with doing the kind of work was amazing.
Wang had machines that tested the ROMs and would point of flaws, and they would be sent back to a department of patient ladies who would fix any bugs in the ROMs by putting in new wires, cutting mis-wound wires and re-wiring them, and repairing broken wires.
More information about the cctalk