SparkFun Forums

[wholder]

Call this more of a nostalgia project, but I've designed and documented a simple circuit that use an Arduino Nano to magnetically flip a ferrite core between 0 and 1 state and read back the result. It's just a one bit memory, but the design is easy to scale up to larger arrays. With a little experimentation you might be able to adapt it to read a surplus core plane, such as those sold on eBay. For more on my project see:

http://sites.google.com/site/wayneholde ... ore-memory

I suppose this it the kind of project that only someone who worked on computers back in the day when a "core dump" actually meant what it says, but perhaps it will be educational for the later generation, too.

Wayne
Back to top

[salvador]

WOW! this brings back allot of memories. For many years I worked on a CDC 6600, water cooled, core memory computer and I can appreciate your project. Nice!

[leon_heller]

When I worked at English-Electric-LEO-Marconi Computers at Kidsgrove, in the early 60s, there was a large room with about 50 ladies assembling core memories. They often used to sing hymns while they were working. Core memories were used on all our mainframes at the time, machines like the KDF9 and KDF7.

[bill190]

I saw the title of this topic and did a double take!

Magnetic core memory? In 2010?

My goodness! I don't remember the size of the mini computer magnetic core memory boards I worked on, but the number 128 rings a bell. I think it was 128 bits x 128 bits.

If you are using that, be sure to also get a removable platter disk drive which was the size of a dishwasher. I think these were about 72 megabytes.
http://www.techtree.com/India/Reviews/E ... 8-632.html

And don't forget the paper tape boot loader!
http://en.wikipedia.org/wiki/Punched_tape

[wholder]

When I worked at English-Electric-LEO-Marconi Computers at Kidsgrove, in the early 60s, there was a large room with about 50 ladies assembling core memories. They often used to sing hymns while they were working. Core memories were used on all our mainframes at the time, machines like the KDF9 and KDF7.

If you can recall any details about how these women went about wiring up the arrays of cores, I'd appreciate hearing about it. I've managed to do the X/Y wiring for an 8x8 array, but threading the sense wire through all 64 cores is a real pain. While I've managed to track down quite a few articles on the electronic details of core memory, the manufacturing process seems to be lost to time, which is a real shame.

Wayne

[leon_heller]

I never went in there to see in detail how they assembled them. I vaguely remember that they might have used something like a sewing needle to thread the cores, but I'm not sure.

When I was sent to the LEO factory in West London one of the things I did was test the microcode units for the prototype LEO III machine, which were made at Kidsgrove. They used ferrite pulse transformers to achieve the logic functions. Rather than return them to Kidsgrove if they were assembled wrong, which I was supposed to do, I rewired them myself, which was quite easy. Their construction was completely different from the core memories, we called them 'banjos' because they looked vaguely like one and were about the same size.

[wholder]

On a tangential note, in the course of learning about how magnetic core memory works, I stumbled across an article on how later model Seeburg jukeboxes (the ones made between 1955 and 1978) used magnetic cores to record the selections a user made. Apparently, pressing the 'B' and '7' buttons, for example, would send a current through a particular core and magnetize it. Then, as the player head scanned across the records from side to side, it would read out the state of each core and stop when it found one that was written. Seeburg called this the "tormat" module and there is more on how ti worked it here:

http://home.pacbell.net/fmillera/tormat.htm

Al those years popping spare change into jukeboxes and I never realized it used magnetic core memory...

Wayne

[Conrthomas]

Wow, talk about cool! I'm definitely going to follow this thread.

[redwire]

I have info on the Apollo spacecraft guidance computers, designs started in 1961 at MIT. It solved the navigation equations to maintain knowledge of the spacecraft's position and direct control of the rocket engines.
The core memory RAM was only 1008 words, 11.7usec access time. Op-amps needed to pick up the sense line pulse.
For program ROM, "rope memory" was used where you threaded the core to make a "1" and skipped it to make a "0". 2nd generation computer Apollo Guidance Computer AGC4 1963 first to use IC's and 16-bit wide instructions of 12,288 words ROM. Rows of women with microscopes weaving doing the assembly. God forbid you need to make a change to your code.
87lbs and 85 watts power, 4 refrigerator sized cabinets. I think it was Apollo 5 or 6 that used the AGC4? Check Wiki as I think schematics are on-line somewhere.
http://en.wikipedia.org/wiki/Apollo_Guidance_Computer
Reference (excellent book):
Hall, Eldon C. (1996), Journey to the Moon: The History of the Apollo Guidance Computer, Reston, Virginia, USA: AIAA, ISBN 156347185X

*** core RAM with an 11.7usec access time 0.0855 MHz might be a bit s l o w .

[thebecwar]

Also impressive is the fact that even the cheapest AVR is several magnitudes more powerful then the Apollo guidance system.

[wholder]

The technique used to create the Apollo rope memory was also used commercially. One example is the Nixdorf "threaded ROM" which can be seen here:

http://www.technikum29.de/en/devices/threaded-rom.shtm

and the more advanced, user configurable "stick memory" which can be seen here:

http://www.technikum29.de/en/devices/ma ... emory.shtm

One thing I've learned from studying older tech is that the engineers who came up with these innovations were very creative lot, as they had to often invent not just the end results, but often also the materials and components needed to build them. For example, I'm still trying to figure out exactly what sort of metallic mix is needed to create cores that are suitable for use in a memory plane. Before I started to research this, I just assumed that all ferrite cores were just about the same. But, it turns out there are "hard" and "soft" ferrites and many different chemical formulas in each of these two types. It must have taken a lot of perseverance to work out all the details needed to build the first core memory systems.

Wayne