> On Jul 20, 2021, at 10:04 PM, John Foust via cctalk <[email protected]>
> wrote:
>
> At 12:13 PM 7/20/2021, pspan via cctalk wrote:
>> I worked at a company called DMA located in Amery Wisconsin during the 80's
>> and 90's that did do core mat repair.
>
> And why were they in Amery, WI - a very small town?
>
> - John
Perhaps because someone there started the business, did a good job, and since
the field wasn't big enough to support two companies he ended up owning the one
and only core repair company in the US.
Re an earlier comment about 3-wire vs. 4-wire topologies, if you want to see
even wilder stuff take a look at the design of the CDC 6000 core modules. It's
documented in one of the training manuals on Bitsavers. It uses a FIVE wire
design, with two inhibit wires rather than one. And the inhibit wire doesn't
run through the whole plane as is usual; there are four horizontal and four
vertical inhibit wires.
The description doesn't give a reason, but my assumption for why this was done
is that the total inductance of each of the driven wires (X, Y, InhX and InhY)
in the 12-bit stack is roughly the same. That means a single driver design
works for all those wires. The other thing that's interesting is that the
drivers don't switch current on and off; instead, they switch current from an
idling inductor to the wire. Again, it doesn't say why; I assume it is for
speed (short pulse duration). Certainly the performance of those memories -- 1
microsecond read/restore cycle and under 500 ns access time -- is astoundingly
fast for 1964.
paul
