Magnetic Core Memory Reborn

16 bits of magnetic core memory (click to enlarge)

Magneic Core Memory Reborn

I’ve been waiting a long time to write this post so it is with great satisfaction that I finally sit down to do so.

I recently finished working on a fairly long project which I carried out jointly over the past year or so. My friend Ben had been reading about magnetic core memory and had discovered that it was possible to buy old, surplus magnetic cores on eBay. This gave him the idea that it might be fun to try to build a modern core memory module. He mentioned this idea to me and explained a bit about the technology. It wasn’t long before we had agreed to take on the challenge. As far as we knew, it was going to be the first time somebody had made a core memory module for decades!

In fact this post is really just an advertisment for the real report which is hosted here. Please visit if you are interested.

It was an enormously satisfying project for a number of reasons. We learned about the physics of magnetic materials (in much greater than we needed as it happened), we learned exactly how this interesting technology works, we learned how to design PCBs, how to have them fabricated and how to reflow solder them at home and we acquired huge respect for the original inventors of the technology whose task was so much more difficult than ours. Also, importantly and despite some unexpected challenges, we were ultimately successful. Taming the analog world and building a reliable digital memory module was indeed a rewarding experience.

The original 8-bit prototype (click to enlarge)

There were quite a few stages to the project and we hope to write up the details of our project in detail in the near future. However we decided we would first write up a clear, clean report of how we think one should go about building a core memory module, rather than exactly how we did it.

Lastly, we couldn’t resist the temptation to publish on the 60th anniversary of the patent which cast core memory into its current form. Core memory had several inventors as is evident from the US patent history. The first patent, 2992414, filed on May 29, 1947 by F.W. Viehe seems to be the earliest mention of the idea. Just over two years later on Oct 21, 1949, A. Wang (founder of Wang Laboratories) filed patent 2708722 in which the idea was apparently reintroduced. However it was on May 11, 1951 that J.W. Forrester filed patent 2736880 in which the final key idea: coincident current addressing was introduced. Coincident current addressing was the idea which made large core memory arrays a reality. Indeed, IBM paid MIT $13m for this patent in 1964 according to the computer history musem.

In any case, here again is the link to our report on this the 60th anniversary of Forrester’s patent application.

The final 32-bit version (click to enlarge)

Please do get in touch with both me and with Ben if you are interested in our work.