It’s always interesting to see what will come out of a hacker meet-up. At the Observe, Hack, Make festival earlier this month [Bertho] was talking to a guy named [Erik] about how flip-dot displays work. [Erik] mentioned that the control theory is the same as core memory. So when [Bertho] got back to his home workshop he started playing around with it to see if a flip dot display can be made to behave exactly like core memory.
We’re really glad a successor to core memory was found since it’s pretty slow. But the concept still makes for some fun exploration (here’s the obligatory Arduino implementation of core memory). It uses magnetic rings with two conductors running through them that pass at right angles to each other. Sound familiar? This is exactly how flip-dot displays work.
There are, of course, some differences. The biggest one being that the displays don’t have the sense wire present in core memory. That was an easy enough thing for [Bertho] to get around. He added the grey sense wire by threading it through the inside of the hardware. The other hurdle he had to overcome was to alter the controller firmware to match the destructive tendency of core memory (reading the state also resets it).
So far he’s just set this up as a proof of concept, reading the sense wire while repetitively reading and writing to the “memory”. But it’s engaging to see what was captured on the scope. We asked him about his future plans, specifically what he would use to automatically read from the sense wire. His response is found after the jump.
Mike Szczys wrote:
Just out of curiosity, if you are adding your own sense wires to the display, what kind of circuit do you think you’ll use to read from them?
Bertho responded:
I’m not going for a sense-wire. It is too difficult to amplify in a stable fashion (I only got 20mV with a lot of noise). Moving the wire will also change the flux-coupling and makes it even harder.
Real core-memory has a well-defined coupling and has an inductance that is orders of magnitude lower than flipdots. Therefore, the sense-wire will see well-defined pulses, whereas the flipdot version sees rather slow flux-buildup. The stored energy in the flipdot coils make rather nasty spikes where real core-memory has negligible energy stored.
It is much easier to do current-sensing on the H-bridge setup, which has a distinct shape based on the “seen” inductance. Simple amplification of the current-sense voltage combined with a comparator gives current-rise-time and that is indicative for what was stored.
The flipdots need to have 350mA for 450us to flip, so that is a fixed timing frame (see my other flipdot posts on my website for description how that works). The current-shape is then easily transformed into a memory-content analysis.
Hahahaha, YES! Someone finally did it! I’ve got a slab of this stuff waiting for round tuits. I’m never gonna get a round tuit.
Further ideas:
This is a display whose contents can be read back, but it’s also a memory device whose contents can be viewed. Think about it from both angles.
With a handheld magnet, you can manually remagnetize the cores, and manually “write bits” into the display. Can the core-memory circuitry read these bits out? Can the core-memory circuitry detect the presence of a handheld magnet over a particular pixel, perhaps by noticing that it’s a stuck bit whose contents don’t verify after a flip? That would make it an input device.
Again assuming the presence of a user with a handheld magnet, can the user feel vibrations in their hand if the core-memory circuitry repeatedly toggles a particular pixel? That would make it a haptic feedback device.
Also, you can play music on them…
It is quite hard to change the state with a permanent magnet. The square hysteresis loop magnet is hidden inside the casing. Sweeping a permanent magnet over the dots will simply move dots around and they will flip back (unless you demagnetize the cores too much). Not much of a practical input device though.
You can possibly feel the disc flip with a very low-power magnet above the disc (probably better to use a piece of iron). However, we are talking about very very little force. The disc flips in 20..30ms (small) or 40..60ms (large), whereas the core can flip magnetic direction 800 times per second.
Hm, noise is possible, already done that, but playing music, /that/ I will have to try too :-)
You’re right, sweeping the magnet over the dots doesn’t fully remagnetize the cores. On my specimen I’m able to keep the dots flipped after removing the magnet, but it stands to reason that only the top portion of the structure is affected.
I do still wonder whether you’d be able to detect the presence of a handheld magnet by seeing a change in the current required to flip a nearby core. It should take more energy to flip one way than the other, right? Whereas with no nearby magnet, it should be approximately symmetrical.
And who said anything about practicality? ;) This is fascinating stuff! I’m glad someone’s not just doing it, but documenting it.
(I’ve also found that our maglock door magnet can play music…)
Why not read the memory optically? Noy I the true spirit of core memory but would be faster..
Actually, I was planning to use some mirror version flipdots to make a (laser) lightshow.
So… DLP
Flipdots aren’t quite the same as core memory in construction – core memory has independent row and column wires, while a flipdot display has rows and columns joined by the inductors for each dot.
True, the flipdots have an integrated coil. However, you /could/ split the coil (it already is partitioned) and then have real X/Y drivers. It would require some serious hacking with uncertain outcome; the coil already is the weak link (very thing wire).
A serious problem you would encounter is the induction of the coils. It is about 3.4mH when the magnet needs to flip direction. Adding multiple coils in series (X and Y) would make it immensely slow and proportionally more difficult to detect the current state, let alone coping with the coil’s DC resistance at 18.2Ohm. Core-memory has virtually no inductance nor DC resistance and having 32 cores in series is no problem. Having 32 flipdots in series would be very hard to get going.
Some flipdots have 3 tap coils, but some only have two and rely on having bipolar drivers. I think the author’s current sense idea is on the money, though.
huh. ive always wanted an electric othello board, just never got around to it.
Does anyone know of US sources for surplus flip dot displays?
Old displays are very hard to get your hands on. However, flipdots are still being produced.
I designed the 7×7 modules and controller from scratch with new flipdots and components (which I used to test the core-memory). The manufacturer and I are currently in discussion to bring the modules to the DIY market. You can find some more information on my website too.
Would help if you mentioned what your website is!
As the original author of the featured article here at HaD, you can quite easily follow the link in this article (vagrearg.org) and look at the left-side index of articles ;-)
In Canada, I’ve seen ’em at Active surplus in Toronto. Never run across any in the US, but they must be around. Hmmph.
I want to find one big enough to display a minimal QR code. Mine’s just a few rows too small.
Gotta love core memory, I have a chunk of it framed on my wall that I point at when someone complains about the speed of their computer.
I also would love to find some surplus ones of these. I feel like they have to be around, in a warehouse or garage somewhere. If anyone knows where to get some of these I would love to know. The best I have seen is 400 EUR for a 14×28 display. That is way to pricey for me to play with.