Ferrite Core Memory

Ferrite Core Memory

Ferrite core memory was the best way to make computer memory from the early 1950's to the mid 1970's. It is a fascinating technology that feels completely alien today, and it is amazingly difficult and complicated. Better explained in videos. I have several of them on the subject: an introductory but comprehensive explanation, probably the place to start, having fun with Soviet era core memory and demonstrating writing and reading a few bits from it, looking at a spectacular 64k stack from an IBM FAA mainframe computer. And let's not forget the videos about the Apollo Guidance Computer core erasable memory and core rope memory, both of which gave us considerable trouble.

Core Memory Explained

Hopefully you will be able to understand how this all works after you watch this introduction to core memory, which we explain with animations and then demonstrate using an IBM 360 core plane.

Playing with Soviet Core Memory

In this video I mount two spectacular old Soviet era core memory planes in a plexiglass display, hook them up and attempt to make them work. And yes, they still do work!

The complete schematic of the experiment is available further below.

IBM 9020 Core Memory Module from an FAA Mainframe

In this video, we explore a massive memory module from a large IBM mainframe. All 64k of it.

Apollo Guidance Computer Core Memory Module B12

Our Apollo Guidance Computer 2k core memory module has a fault. We throw just everything we have at it, from X-Rays to time domain reflectometry, to find what the problem is and if it is repairable.

Apollo Guidance Computer Core Rope Bringup

If you thought core memory was weird, wait until you see core rope memory, which was used as the read only memory in the Apollo Guidance Computer. It contained the actual flight software, which was literally woven into a wire and core tapestry.

Diagram of the Soviet Core Memory Experiment

Here is the actual schematic of the experiment I demonstrate in my Soviet Core Memory video. It contains all the elements necessary to read and write a bit from the memory.

The Saratov-2 is in the rack on the right.

The top drawer has logic cards that look to be a copy of the DEC "flip-chip" modules. The core stacks are in the drawer underneath, 4 can be seen, for a total of 4 kW.

Here is a core stack. You can clearly recognize the core plane with its four quadrants, and even the little dangling spare cores in the extra line. The stack seems to have 14 bit planes, which does not jive very well with a 12-bit word of a PDP-8, even if you add parity (you'd expect 13 bits). Maybe they expanded the bit width to 14.

Comparing to mine below, it appears to be the same: