Re: Re: Summary A

[Steve VanDevender <stevev@efn.org>]

Timothy van der Linden writes:
 > To Steve,
 > 
 > >Core memory is still used in a few applications, although sometimes just
 > >for historical reasons.  It does have the advantage of being
 > >radiation-hard (find me a cosmic ray big enough to wipe out a 1 mm
 > >ferrite core) and non-volatile.
 > 
 > I never understood exactly how it works, I know there are horizontal and
 > vertical lines, with some small ring around every "knot". Do you know how it
 > works and can you explain it in a paragraph?
 > (What is the funcion of these ferrite rings?)

Here's one core cell as a crude ASCII graphic.

          |   /
       -- |  /
       \ \| /
        \ \/
---------\ \--------- horizontal address
         /\ \
        / |\ \
       /  | -- ferrite core (seen edge-on)
      /   |
 sense    vertical address

All three wires pass through the core, which is a ferrite ring; the core
itself holds a single bit encoded as a magnetic state.

The horizontal and vertical address wires form a grid, with a core at
each intersection.  The sense wire is a single wire that threads through
all the cores in a plane.

To write a core cell, current is passed through the horizontal and
vertical address wires that both pass through the desired core.  The
current in the horizontal address wire or vertical address wire alone is
not enough to magnetize a core, so none of the cores that are on that
same wire are affected, but at the intersection the total current
through both wires creates a field strong enough to change the magnetic
field of the core there.  Since the field is effectively circular around
the intersection, current in one direction magnetizes the core in one
sense, while current in the other direction magnetizes in the opposite
sense; one sense represents a 0 bit, while the other represents a 1 bit.

To read a core cell, it is addressed to write a particular state (either
0 or 1).  The core was either in the same state as the one written, and
reinforces the magnetic field, or was opposite the state written, and
gets "flipped", and these produce different induction effects on the
sense wire, which is read to obtain the core's state before the write.
Since reads are destructive, to preserve the core state it must be
written back immediately after reading.

Although it takes more current to read or write a core than it takes to
update a semiconductor memory cell, no power is required to maintain the
state of core memory.  It wasn't until about the mid-70s that
semiconductor memory became all of faster, denser, less power-intensive,
and cheaper than equivalent amounts of core memory, and core was still
used for some time after that.