Core Memory 64 bit (picture heavy)

[Art]

Hi Guys

Over the last couple of months I’ve completed my 64 bit core memory project
(mid 50’s - mid 70’s vintage).
If you did not know this was the mainstream method of RAM hardware in computers,
but is also non-volatile memory I plan to use as EEPROM.
If you did not know that, hang your head in shame.

AFAIK I’m the third to do it in modern times.
Two core memory projects are documented on the web, a 1 bit, and a 24 bit,
so mine is so far the largest, and will be put to practical use in a larger project.
I just read about one here in an old thread, but looks like it didn’t go anywhere?

64 bit Core Memory matrix (1mm cores) and sense wire:

Etching glass window cutout and creases for memory module:

Completed Core Memory module:

Current limiting resistor board:

Wiring current driver module:

Completed Current Driver module with copper plated copper graphic:

Core Memory sense pulse receiver from what will be the rear:

Wiring of the lower shift register board:

Completed Control Hardware without the second top cover:

Second (copper plated brass) top cover:

Three Modules (prior to micro controller and PSU) :

LED Test Jig (plugs in place of the memory module) :

Microcontroller and Power Supply:

Completed:

I have made a start on visual control software here:
https://www.youtube.com/watch?v=YirR9WHJN5w

Here’s a demo video that also shows a permanent magnet being used to corrupt data

Serial terminal demo:
https://www.youtube.com/watch?v=i8wXkI-mQBE

Cheers, Art.

 

[Art]

There is a four part video on the same YouTube channel vblogging hardware progress.

Aside from showing off :D I thought it a good idea to join here because soon I plan to
weave a constant byte data array into core rope memory.
I have less information on the history of core rope in terms of circuitry but a good idea how to go about it.
Cheers, Art.

 

[Art]

ok so this was a vintage computer forum right?

Though I might as well make a core rope ROM as well (as used in Apollo lander).
Got it working last night, and uploaded a demo today.
Cheers, Art.
https://www.youtube.com/watch?v=j8QuhUoC9CY

 

[Chuck(G)]

Ever considered building some core logic? Consider the Univac "Solid State" machines (described in great detail on bitsavers). I've always found that one interesting.

So where does one get largish "hard" (remanent) magnetization ferrite toroids nowadays? Everything I've seen thus far have been the garden-variety "soft" type.

 

[Art]

Hi Chuck
I have more than considered core logic, and when this has an enclosure it will probably be next.
I love the idea of non-volatile discreet logic such as decade counters and shift registers.

I got the toroids for the RAM from an eBay seller in Bulgaria. They are the 1mm cores.
The core rope uses soft ferrite, and it might be a task to find the larger square ferrite for the core logic.
If worst comes to wort I will eBay a cheaper broken vintage thing, but that is a last resort.

Ever considered building some core logic? Consider the Univac "Solid State" machines (described in great detail on bitsavers). I've always found that one interesting.

So where does one get largish "hard" (remanent) magnetization ferrite toroids nowadays? Everything I've seen thus far have been the garden-variety "soft" type.

 

[NeXT]

I just read about one here in an old thread, but looks like it didn’t go anywhere?

I'm sorry I missed this thread earlier.
Yes I tried either last year or the year before that to produce a few frames of core that I could figure out a use for in future projects but I fell flat on my face producing the frames themselves because I did not and still do not have a proper way to slice up perfboard into segments that can make the much larger frames for 128, 256 or 512 bits of storage. I still got a can of beads here and the wire. I might go back to it once I get a proper S100 machine working and see if I can map something like 1K of memory space to core. I sat down and did the math on what it would take to just to 4K on an 8-bit system and unless I found a way to much more tightly weave the beads you would have an impossibly large stack of core.

Another idea I had brew was to just make a massive 4'x4' frame and build the plane using 16 gauge wire and the larger 1" toroid ferrite beads but the electronics would have to be rather heavily scaled up for it to work at all.

 

[chalackd]

I had my own adventures weaving some core memory, results of which can be seen in this thread. I had the weaving part down quite well, but don't have a clue where to start for the circuitry. So this was mostly a practice endeavor to see if it would be feasible to weave larger planes at home.

The plane linked above still has no sense wire, because I wasn't sure what configuration would be best at the time. There's also one core missing (At what I would call J-3 in that photo) which broke when I was nearly done, and I haven't had a good reason to restring that line.

Working on that I learned a few tricks, and if I ever had a use for one of these I think I could cut the size in half (actually might keep the size and double the density from 1024 to 2048 bit) with a custom etched PCB and slight changes to my procedure.

 

[Chuck(G)]

Hi Art,

I still have a bunch of 1401 cores as well as a complete plane--those are about 1mm. But core logic is going to take larger cores than that, as you want multiple windings on a single core.

I suppose that if all that's available is soft toroids, one could construct a couple of Parametron circuits, but those lack the interest for me that the old hard-core logic does. I seem to recall that one of Seymour Cray's early military computer designs used core logic.

Anyway, a nice project--well done!

FWIW, on the Classic Computer list, there has been discussion about re-constructing 1620 core for a few surviving systems--it seems that the solder used to assemble the original frames "dissolved" the copper wires and that the original frames were unsalvageable. Given that a basic 1620 with 20,000 digits of 6-bits each would require in excess of 120,000 cores (you need to allow for spare columns), that seems to be a project at least a couple of magnitudes of effort larger. Currently, the CHM 1620 uses regular CMOS SRAM.

 

[Art]

Interesting replies. Well I’ve been into electronics my whole life, and to be honest the core RAM is the most satisfying thing I’ve done.
Not even so much my own work, but I still hardly beleive the principal behind it works at all.
Having said that, if I worked in one of those factories weaving it I would lose the will to live quite quickly!

I’m sure reproduction of a very large array could be done, but what a humbugs task.
The 64 bits for mine was specifically for a project I wanted to store 2x32 bit floats right from the get go.

@chalackd, I can’t seem to load the larger images, but the core pane, I like your style looking at the smaller forum preview,
nice job! but at the same time I’m disappointed.. that is way past the point I would let a project go!
I used classic snake weave because I knew I needed no inhibit line for multiple stacks.

From what I read, a single sense wire was always limited to 1024 cores, and larger panes just used more sense loops and sense amps.

Well I did come up with a method for only needing a sense receiver that amplifies positive going pulses only,
even though you cannot wire a sense loop in such a way that it will only send positive going pulses.
100 points for figuring that one out :D