Up for Auction: Memory from the First Computer in Space

[Brent Hilpert]

On 2015-Oct-29, at 10:22 AM, Jon Elson wrote:
> On 10/28/2015 11:48 PM, Brent Hilpert wrote:
>> Very interesting to hear of another scheme, but it's not clear whether it applies to the Gemini auction memory. The BiAX scheme shows cores with the holes (apertures as they're called in the business) perpendicular to each other. In contrast, the Gemini auction cores have two apertures with the same orientation (a figure 8). 

> Yes, but I'm pretty sure the concepts are related.  The remanent flux in the non-volatile side of the core affects the flux hysteresis in the volatile side, so when you flip the flux polarity on the volatile side, you can see some effect caused by the non-volatile side.

At  a general level, as you suggest, they could be said to be related, however what description is out there does seem to indicate them being distinct in the detail.

Refs on this page:
	http://www.ibiblio.org/apollo/Gemini.html
refers to the Gemini memory as being of the MARS type  "Multi-Aperture Readout Sensing".

MARS is described here if one wishes to delve into the magnetics:
	http://www.computer.org/csdl/proceedings/afips/1961/5058/00/50580443.pdf

A cursory description of BiAX
	http://encyclopedia2.thefreedictionary.com/Biax
does describe different functional principles as suggested in one sentence here:
	http://adsabs.harvard.edu/abs/1961P%26SS....7..184W

Photos of the Gemini auction memory:
	http://historical.ha.com/itm/explorers/gemini-3-flown-random-access-non-destructive-readout-4096-bit-memory-plane-from-the-gemini-spacecraft-computer/a/6146-52008.s
See the third photo for detail.

. . and a photo of BiAX cores:
	http://www.computerhistory.org/revolution/memory-storage/8/253/984

While researching this I ran across a couple of other patents for multi-aperture techniques.
A fair effort seems to have been put into developing a non-desctructive readout for core memory, to little effect in the marketplace. Looking at the photos one can see why, it looks like the methods doubled or better the complexity of construction, so the 'standard' core memory techniques remained the most cost-effective.

Amusingly, there was also a magnetic core device called a "transfluxor". Take that, flux capacitor.

I like this one however, another technique for NDRO, using standard core construction, from the 50s: 
	http://www-isl.stanford.edu/~widrow/papers/j1954aradio.pdf
The magnetised core acts as a non-linear mixing element for two RF signals sent down the matrix wires.
The magnetic polarisation of the selected core affects the mixing in such a manner that the phase of the difference/beat frequency can be observed to determine the polarisation and hence the stored-bit contents.