Charles, if I may;
OK, a little background on coincident-current magnetic core memory.
The original Random Access Memory, its predecessors were shift registers
and delay lines, either mechanical, like the coil in an old reverb unit,
or a glass tube filled with mercury with a piezo transducer like a crystal
earphone at one end to inject sound pulses into the mercury, and a crystal
microphone at the other end, to pick up the pulses.
The drawback of the shift registers and delay lines was that the data was
available only in the same order in which it entered the storage device,
and only after all the previous data come out.
Needless to say, the idea of being able to access any piece of memory in
the same amount of time, regardless of its location in the memory, was
very attractive.
Mag cores used two drive wires, one for the X Axis, and one for the Y. By
changing the direction of current flow in the drive wires, it was possible
to magnetize the ferrite core toroid clockwise or counterclockwise.
| /
| /
| /Sense Line
= /
// | \\/
X-Drive Line // | / \\
------------ (( ---/--- )) -------------
\\/ | //
/ \\|// Ferrite toroid
/ =
/ |
/ |
/ | Y-Drive Line
Suppose we consider a clockwise magnetization to represent a binary 1, and
counterclockwise to represent a zero. If we want to read the bit stored in
a particular core, just set the X and Y drive lines that got through that
core so that they try to make the core magnetized COUNTERCLOCKWISE. One of
two things will happen:
If the core is already magnetized counterclockwise, there will be no change
detected by the sense line, indicating that the core held a ZERO bit.
If, on the other hand, the core was magnetized clockwise, when its magnetic
field reverses it will generate a current in the sense line, indicating a
ONE bit was held there.
At this point, with the data that was read out held in a register in the
core memory controller, the location has to be re-written, so that the data
isn't destroyed permanently.
Some processors took advantage of the destructive-read of core memory, and
implemented instructions that did a Read-Modify-Write cycle. The assembly
language instruction "INC memory-location" is an example of one such.
The original core memory toroids were large, and the planes made from them
correspondingly bulky. As the technology improved, their size was reduced.
Manufacturers of core memory employed large numbers of young women to string
the cores on the wires. (Men had neither the accute eyesight, nor the fine
muscle control needed for this process.)
One unfortunate individual spent almost a decade perfecting a machine to
weave core planes. When finished, it could produce core memory of extremely
high quality, almost any length required, very quickly and inexpensively.
The patent was granted just about the time that Intel introduced the first
dynamic RAM chips, and forever destroyed the core memory industry.
The only use for core memory these days is in VERY niche applications that
need long-term, no-power-applied data storage that doesn't suffer radiation-
induced soft errors. Like in the shuttle. It actually uses what's called a
"plated-wire" magnetic memory, where the toroid is a thin plated ring around
the sense wire, rather than a separate object.
I once worked with a mini-computer, the Hewlett-Packard 2116, that had
16KWords of core memory. It was housed in a 19 inch ECMA rack mount and
generated enough heat to warm a good-sized apartment - One of the first
things I helped do to it when I went to work there was install dual 800CFM
squirrel-cage blowers in the lowest bay in the rack. Even with THAT much
airflow, blowing right up from below, in an air-conditioned lab, the thing
ran hot.
Ahh, for the good old days...
Dave
To unsubscribe from SURVPC send a message to [EMAIL PROTECTED] with
unsubscribe SURVPC in the body of the message.
Also, trim this footer from any quoted replies.
