Retro tech – magnetic logic

Jeri Ellsworth is facinated by retro computing technology. In her latest endeavor she’s researched the principles behind magnetic core logic, a technology used to store data in the pre-transistor days.

We’d have appreciated a close up of the device under test as well as the oscilloscope screen, but an interesting presentation nevertheless.

For more information on all magnetic logic computing, Jeri suggests a visit to SRI International.

Join the Conversation


  1. Still trying to figure out why Jeri is hung up on old tech. Yes it was pretty cool because it’s what got me into engineering. There are so many new / cool technologies today. Would love to see some forward thinking tutorials.

    1. Still can’t beat magamps for robust power electronics in some cases. Now we have nanocrystaline magnetic materials and superconductors to make them with though. Just look at the inductive superconduting fault current limiters.

      1. Another application is the FRAM devices, though they don’t work on exactly the same level, they still have issues of desctructive write and magnetic state. Keeping “old tech” in mind can be important.

        On a mostly unrelated note MEMS research is also looking at microfabricated relays and vacuum electronics as lower-loss replacements for FET and BJT devices.

  2. Yes, the new batch of people coming into electronics often are unaware and disinterested in the old ways, but it’s often important to know when you need something a bit different – Not everything can be sorted with the latest micro! It’s a bit like people who only learn digital electronics and have no idea about Analogue, AC theory, RF theory, Transmission line theory, etc, they can come awfully unstuck later on.

    I actually have a real magnetic core memory board along with a bunch of other old 70’s computer boards. The cores themselves are very small and the array is about 4 inches (~100mm) across from what I recall.

    1. Just look at all the “pocket scopes”. The digital / MCU / FPGA side is fine, but the AFE on all of them is awful. Input impedance and bandwidth change depending on attenuator settings and they only reach ~10% of “claimed” bandwidth (because they don’t seem to know how to measure it).

  3. I think Jeri explores the old tech stuff because she’s a tinkerer at heart. It’s true that there is plenty of new tech stuff that is interesting, but, from my perspective, there are millions of people using that stuff right now. A lot of the old stuff has basically abandoned because it’s seen as no longer useful (or so dated that it’s uninteresting). I’m of the opinion that some of the abandoned technologies may actually have some unexplored potential, as absurd as that may sound. When Jeri spends time toying with the old stuff, she’s staying in touch with her tech roots. I don’t think it’s any different than someone working on old cars.

    (responding to JerseyTechGuy)

  4. A lot of the old tech becomes new tech once the rules (materials) change. We may find that magnetic circuits are the way to go as nanotechnology evolves. Jeri might be way out in front of this.

  5. Jeri,
    I am going to attempt to paste in some notes I printed on the back of a core memory plane for display. It has a few smashed cores, but I would be happy to donate to your personal blast from the past! I’ll send a pic if you shoot me an email addy.

    Iron Core Magnetic Computer Memory

    This iron core memory plane probably dates from about 1964, where it may have lived in a CDC 6600 computer. Since that computer had a 60 bit word size, there would have been 60 of these planes in a parallel wired vertical stack, 512 wires connecting to every plane. Each of the four arrays is 64×64, or 4096 bits (1/core). Therefore, this plane stored 16,384 bits. A stack would have been a 16K word page in the memory. Here’s how it worked.

    This plane has 128 X and 128 Y drive lines, each connected to an H bridge of discrete components which can drive current pulses in either direction through each drive line. The current is carefully set to create a magnetic field too small to affect the magnetism of a core. However, at the spot where an energized X and Y line cross, the vector sum is (square root of 2) times stronger and can flip the magnetization of a core. If the state of a core flips, a current is induced in the sense line which runs through every core on a diagonal to the X/Y lines. These lines come off of the planes through the colored wires at the edges. A flip would be a one and no flip of the intersected core would be called a zero. The core had to be pulsed back to its original state after every read/sense operation which resulted in it being flipped, called destructive readout.

    This technology was the fastest RAM available at the time. The 6600 memory had a 1 microsecond cycle time. The small black rectangle glued in the center of the plane is a circa 2008 DRAM chip having a 50 nanosecond access time, twenty times faster than the cores. But there is a metric which has changed more than that.

    Best guess is that the 16K bits of memory on this plane cost CDC about $12,000 (in 2008 dollars). At that rate, the 1Gbit chip at the center ($2) would cost you $78,432,000! How many products can you think of which have increased in power by a factor of twenty, while decreasing in price by a factor of 40,000,000? (price/performance ratio change of 800,000,000) Not too many, I would guess!

  6. Wow. The comment is awaiting moderation, so I assume it’s effectively an email to you. ‘Sorry that your blog code stripped the returns defining paragraphs. I’ll use ellipsis as substitution. . . . . .

    Yeah, I’m an old geezer dating back to that era. I built a gamma ray spectrometer with a Nuclear Data pulse height analyzer with core memory. Also, our Scientific Data Systems 925 computer had cores too. Funny, when the campus power was off for a half hour, the computer would not work for another half hour. ‘Turns out that computer operation was inhibited unless the core memory oven was up to its set temperature! So do remember that setting your Schmoo Figures is a temperature dependent operation. Also, core memory in computers lived long into the space age as the cosmic ray bursts experienced outside of earth’s magnetic field shield were known to massively corrupt flip-flop memories, but left iron cores unaffected. That is why they flew to the moon! ……..Have fun with your Techno-Retro!

    ‘Ever wonder who did the extremely fussy work of building these things

  7. A continuation because post was somehow clipped……The fussy work was done exclusively by little old ladies! Manufacturers sought out women who did lacework and needlepoint embroidery. They were sent core planes and wire, and got paid big money when their fussy little sewing project tested out. Funny, eh?

Leave a comment

Your email address will not be published. Required fields are marked *

Notify me of followup comments via e-mail. You can also subscribe without commenting.