"Dave, AA6YQ" wrote in message
et...
Getting core memories to work was black magic with the original designers
at
your side; that's something that you got a 1620 stack working on your own.
For the price of a few electrolytics -- or a new switching PS -- and a few
more hours, it definitely sounds worth resuscitating.

It has enough sentimental value to keep, for now, but I guess I don't see
any real reason to resuscitate it. Emulation on a modern PC would run many
times as fast. I have kept all the paper tape software and the "high-speed"
paper tape reader.

As to the core memory, I lucked out.

The 1620 stack had a two-level scheme to drive the 100 x 100 grid of
selection wires. There were two sets of 100 switch cores that actually
drove the wires in the core stack, one set for each axis. Each of those 200
switch cores was threaded with two wires itself - a total of 20 wires for
each set of 100 switch cores. You put a pulse through the two wires
threaded through a particular switch core to make it flip - thus sending a
pulse on a desired selection wire in the main core stack. You could think
of the switch cores as a decimal decoder. There were ten "10s" wires, and
ten "units" wires threaded through the set of 100 switch cores.
Simultaneously pulse the "70" and the "3" wires to the switch cores and the
"73" selection wire to one axis of the main core stack was pulsed. The
other axis of the main core stack had a similar scheme.

Since I was only implementing 4096 locations, I didn't have to drive all the
"10s" and "units" wires to the switch cores. I had to drive 64 of the 100
switch cores on each axis - and I could use any 64 I wanted. I split the
12-bit address into two 6-bit halves and drove the switch cores with decoded
versions from the 6-bit half associated with that axis of the main memory
stack. The 64 switch cores for each axis only required use of 8 of the
"10s" lines and 8 of the "units" wires through the switch cores for that
axis - a total of 16 drivers per axis or 32 total switch core drivers - far
better than the 128 I would have needed to directly drive the main core
stack selection wires.

The 1620 core stack was thus a little unconventional, but was a big win for
me.

73 de bob w3otc