40k Ferrite Memory Core

[Gavin Melville]

Hi All,

This might help date it.  The Mullard core in these photos was purchased in 
1960, and the planes are 48 x 48, 100 layers.  So far I've resisted the 
temptation to cut it up and make some real money out of it. 
http://www.flickr.com/photos/8788341@N05/1799948181/

Regards,
Gavin Melville

--------------------------------------------------
From: "Rick Bensene" <rickb at bensene.com>
Sent: Thursday, October 30, 2008 8:52 AM
To: "General Discussion: On-Topic and Off-Topic Posts" 
<cctalk at classiccmp.org>
Subject: RE: 40k Ferrite Memory Core

> Brent wrote:
>
>>My read on this is it's a (X=40)*(Y=48) matrix, going by the number of
> pins on
>>the X&Y connectors, with 21 planes. I count 52 pins on the Y-side of
> each plane
>>of the actual matrix. This makes sense as the extra 4 pins (52-48)
> would be 2
>>for the sense line and 2 for the inhibit line of each plane. (The sense
> and
>>inhibit connections do not come out to the connectors, there would have
> been
>>additional wires connecting down the stack when installed). In
>>http://www.thegolfballshop.com/oct28/3.jpg, one can see there are 4
> wires
>>through each core: X, Y, sense (diagonal), and inhibit (parallel to X),
> making
>>it a pure 3D, 4-wire construction (the loopbacks for the inhibit lines
> are also
>>visible).
>
> I agree with Brent's analysis.  I too counted 21 planes, which seems a
> bit odd, as typically each plane represented one bit in a word
> structure, which would indicate a 21-bit word...not something that
> matches up with commonly-used word sizes in 50's/60's era computers (12,
> 16, 18, 20, 24, 32, 36, 48, 60 or 64 bits).  Perhaps some of the planes
> have already been removed?  It is possible that whatever machine used
> this stack didn't use the "bit per plane" methodology, and instead used
> a 48-bit word (Y aspect of matrix), which would mean that each plane
> held 40 words, for a total of 40*21 (840) words of 48-bits each.  This
> stack would likely be just one of many used in whatever computer it was
> part of, in order to have sufficient memory to be useful.  It is also
> possible that this stack was used alone in some type of peripheral
> device.  Frequently devices like line printers, rotating memory
> controllers, and other peripherals of the '50's and early '60's used
> smaller arrays of core for buffering purposes, which may account for its
> rather unusual arrangement.
>
>>It does give the impression of being 50's-era construction, probably
> from a
>>tube machine.
>
> I figure late '50's/early '60's.  By the mid-60's, the size of the
> ferrite cores had been dramatically reduced.  While the general
> construction was similar (in terms of 3D core stacks like this one),
> their physical size was significantly smaller by the mid-'60's.  The
> reduction in the physical size of the ferrite cores was somewhat
> mandated by the use of transistorized driver circuitry.  The amount of
> current required to "flip" a larger diameter core was sufficiently high
> that tube-type driver circuit was required in the early days of
> transistors.  As the cores got smaller, so did the current requirements,
> which led to the practicality of transistorized X/Y/Inhibit drive
> circuits.  The cores in this stack are large enough that as Brent
> suggests, the machine may have (at least) had tube-based drive
> circuitry, if not being a completely tube-based machine.
>
> Later, the three-dimensional aspect of core memory was changed, such
> that multiple arrays were woven into the same plane of core, making the
> core array fit on a single circuit board (typically with additional
> board(s) that provided the address decoding, constant current source,
> X&Y drive, sense, and inhibit functionality) in a 2-dimensional
> arrangement.
>
> Whatever this stack is from, it is definitely a work of art that should
> be preserved intact.  Like Brent, I sincerely hope that it doesn't end
> up being reduced to 21 individual planes to "optimize" its sales value
> on whatever venue it is marketed under.
>
> Rick Bensene
> http://oldcalculatormuseum.com (where some of the old calculators have
> core memory, some 3D, and some 2D)
>
>
>