Core storage consists of tiny doughnuts, in mathematical terms a toroid,
made of ferro-magnetic material, that can be magnetised one way or the other.
Magnetised one way means the value "1" is stored.
Magnetised the other way means the value "0" is stored.
To magnetise a core an electrical current is passed through a wire threaded through the core.
The current in the wire generates a magnetic field around the wire.
If the magnetic field is strong enough the core will be magnetised the same way as the magnetic field.
An electrical current in the opposite direction can magnetise the core in the opposite direction.
The figure shows the state of a core.
The horizontal axis is the electrical current through the core, at "A" it is zero.
If the core is magnetised at "L" it stores the value (One), at "Q" it stores a (Zero).
Say it is storing a (Zero) so at "Q".
If a small current is passed through the wire through the core the state moves from "Q" towards "R",
nothing much happens to the state of the core.
In fact if the current is removed it will relax back to "Q".
However if the current is increased beyond "R" it changes it magnetised state to "J".
If the current is further increased no further change takes place.
When the current is removed the core relaxes to "L". It now stores a (One).
Beginning at "L" if a negative current is passed through the wire,
nothing much happens as the current reaches "M" but after that all of a sudden it flips to "N",
and when the current is stopped it relaxes to "Q" and so stores a (Zero) again.
The shape of the figure is called a "Hysteresis Loop".
If two wires thread each core it is possible to select one core from many.
If 1/2 the current to magnetise a core is on an X wire and 1/2 on the Y wire
then just one core will be selected where the X and Y coincide.
The other cores on the X and Y wires will be just 1/2 selected so not changed.
The above figure is a 4 x 4 = 16 core plane just for diagrammatic purposes.
Actual core planes in 1965 were 128 x 128 = 16,384 cores.
Core planes were stacked one above the other.
The X and Y wires threaded every plane so energising them
with 1/2 current selects a core in each plane.
Every core in a plane is threaded by a third wire,
the "Sense/Inhibit" wire.
The way core storage works is as follows:
To read what is stored a negative 1/2 current is applied to a pair of X and Y wires.
This will cause a zero to be stored in the selected cores.
Where there was a zero before no change will take place.
But where a one was stored the magnetic state of the core will change
and this magnetic change will induce a current in the sense wire for that plane.
This is a destructive read, all the cores selected are changed to zero.
So a read operation is automatically followed by a write operation to restore the data.
A write operation consists of energising the X and Y wires with a positive 1/2 current
that will write a one in every core selected.
However where a zero is to be stored a negative 1/2 current is applied to
the inhibit wire so preventing the core from being changed.
There is a lot more about how core storage works but it is to do with secondary effects.
When a voltage is applied to a wire that causes a current through it
it creates a magnetic field around it that causes induced currents in neighbouring wires or cores
the resultant "noise" has to be ignored in favour of the desired signal.
Such considerations are not in the above but are crucial to the operation of core storage
and had to be dealt with by the manufacturers.
In this case it was the International Business Machines Corp.
from one of whose publications the above diagrams and figures have been reproduced.
|Mus.Cat. NEWUC:2003.10||Mnfctr: IBM||Date: 1968||Form No: Z22-6608-2||Pages: 69|
|Comp: Manual||Height: 279 mm||Width: 216 mm||Thickness: 40 mm||Weight: 796 g|