Constructing an 8-bit relay computer from scratch | Hacker News Reader

Constructing an 8-bit relay computer from scratch

relaycomputer.co.uk

174 pointshggh36 comments

In my senior year of high school (1968-69) Mike Sivy (Yes that Michael Sivy) and I tried to build some components of a digital computer out of military surplus relays. We took the bus and the subway down to Canal Street in NYC and bought a box of large 24 volt relays. We designed a counter and we were able to get some parts working. It took 5 relays for each part, which Sivy was able to drop down to 4 by using a diode. I always thought that the diode was cheating and by using it we lost the purity of the concept, but we had a limited number of relays and we wanted to count to 8. One of our problems was hysteresis and at various times the whole thing would start sparking and chattering like crazy. I remember trying to explain what we did to parents that wandered into the physics lab in the school's open house. They were more impressed with the giant "jabob's ladder" we built that eventually set the lab's window moulding on fire.

The architecture looks vaguely similar to a Motorola 68HC12, down to the big-endian encoding. Quite ambitious considering that it has more registers than a 6502.

Note that while most people here seem to associate relays with large clicking things, there are reed relays which are much smaller and faster, and also practically silent:

https://en.wikipedia.org/wiki/Reed_relay

...and people have built computers using them:

http://www.xpl0.org/rr6/

Clicked for the sound, was not disappointed.

Yes. Also the blinkenlights are very satisfactory. And very nice casing.

What is the clock speed at the end of the demo? It sounds like about 8 Hz.

He mentions in an earlier video that the fastest speed the closk can run at is 32Hz.

The clicking sound was so pleasing.

Having just finished Charles Petzold's Code book, and reading all about relays and building up a fully working computer solely using relays, it's amazing to see a real physical working relay computer! Incredible work!

If you like relay computes, you might also like this: https://www.youtube.com/watch?v=_j544ELauus

1958 FACOM 128B Japanese Relay Computer

Sounds really cool when it's running!

I just got into that channel recently. The youtuber was one of the 4 guys that restored the Apollo guidance computer. The entire process is documented on his channel. It was very humbling to watch these guys debug damn near every transistor and logic gate and eventually run the old lunar landing sequence. The also recovered some lost code that was still stored in the core-rope ram from 50 years ago, as well as built gate accurate emulators for the AGC itself and other peripherals. Definitely worth checking out.

See also TIM a computer made from scrap relays - https://www.youtube.com/watch?v=RfXrT619jBw

Very impressive!

At the last Vintage Computer Fest West before covid I saw this on display [1] and was completely taken with the clicky sound and blinkenlights. It does use semiconductors for the memory and IO.

I built one last March and April, and found it to be surprisingly enjoyable and relaxing. Highly recommended for those who love clicking and blinking things.

[1] http://relaysbc.sourceforge.net/

The author of that board might swing by.

I never got why the relays used in old computers were so large. Typically relays are power devices so they are kind of beefy. But for computers of that era I always thought it would be possible to create boards with much higher relay density. We did have quite advanced manufacturing technology at the time. Maybe it really wasn't as much of a concern and transistors and IC came online before it mattered.

Can anyone explain why the contact DC voltage rating of relays is often much lower (5x is no exception) than the AC rating?

I think it is due to arcing. With AC an arc will extinguish itself itself when the voltage goes through 0, but for DC no such thing occurs and an long arc can damage the relay contacts.

Also, an interesting oddity about relays and switches: many have a minimum current rating because some arcing is actually required, it burns off the oxidation layers on their contacts. This is called "wet switching", if you use a wet-switching relay to toggle a control voltage with negligible current flowing on the line, you can drive the relay to death prematurely due to oxidation. The solution is to use a control current instead of a control voltage, but it has higher power consumption (on the bright side, better noise immunity). There are also specialized "dry-switching" relays to solve this problem, usually for test and measurements applications (but using them to switch live circuits causes the opposite problem - arcing will damage them).

Yes I figured that, but the 50/60Hz frequency seems quite low for that to be the case.

50Hz basically means the arc is extinguished after at most 20 milliseconds, limiting how much the contacts heat up. For DC on the other hand there's no guarantee the arc will extinguish at all. You can imagine how a sustained arc would be bad for a relay.

Of course a higher frequency would heat the contacts even less, and I imagine you could get a relay rated for even higher voltages given a higher AC frequency.

But arcs only happen when the contacts are almost-but-not-entirely closed, I suppose (ionization could be a factor). This time-period is, I suppose, very short, much shorter than the 10 milliseconds of the AC half-wave.

I suspect the arc could cause the contact to stick for a while, which may increase the arc time.

At high enough voltages you get anywhere from second-long to indefinite arcing. Not to mention that the contacts are cost optimized, so the manufacturer would like to get away with as little travel as they can as to simplify the mechanism.

I don't think that's a universal effect.

https://www.youtube.com/watch?v=-cM8P6-aAf4

As the other poster pointed out arcing is one component. Another is that DC voltage equivalent of an AC voltage is the RMS value of the AC signal. So, V_AC_RMS = V_peak / sqrt(2).

AC specifications are usually in RMS equivalent voltage so 1A 10V AC is same power as 1A 10V DC. This makes the peak difference larger, not smaller.

AC is easier to extinguish (0A current 100 times / s), and I believe the reversible direction makes for better electroplating characteristics.

Could this kind of computer work at 450°C?

Yes, you could. But there are also specialized electronics that work at 450C (or even 500C). NASA Glenn Research Center has developed them. Transistors and such using Silicon Carbide (instead of regular silicon). A simple computer could as easily be made with such transistors and passive elements. It’d be much faster and lighter than a relay computer.

The issue with all such computers is developing significant storage. Enough RAM for scientific data and images is probably too energy intensive (electrical energy is hard to get if you have to dump heat at 450C for your RTG). So you either transmit constantly as described here: https://ntrs.nasa.gov/citations/20150002090

or you store it using maybe a magnetic tape with high curie temperature.

Ozark Integrated Circuits got a contract from NASA to develop a high temperature RISC-V chip.

> Ozark IC will create the first 500-degree Celsius RISC-V multi-chip system in a package as a way to illustrate the techniques, design procedures and the multi-chip package with high-temperature components that go into creating a high-temperature electronic system.

press release https://talkbusiness.net/2020/05/nasa-to-give-750000-to-ozar...

Their banner says, "We Design for Venus" https://archive.is/w0lZV/image

https://www.ozarkic.com/

Holy crap, that’s awesome!

That's a great question if you're asking about using it on Venus. Copper melts at 1084 C, so you're good there for springs, conductors and contacts. You'd need insulation for wires and circuit boards: something like fiberglass weave would be good up to 537C. No solder, wire wrap, steel posts.

Resistors and capacitors for RC circuits should be similar, carbon is good and capacitors are insulator/conductor layers.

I suspect making a temperature controlled oscillator will be very hard. This means power supply AC current might be hard to make.

Anyway rebuilding OP's computer that can run in your kitchen oven would be a nice hobby goal.

https://www.mcmaster.com/high-temperature-insulation/fibergl...

The spring constant/youngs modulus probably changes with temperature. Making a relay work reliably over that kind of range sounds difficult.

> Copper melts at 1084 C, so you're good there for springs

Would you really be able to make a spring out of copper? It is my understanding that copper work hardens very quickly, which is a big no-no if you're making a spring, I guess

For springs you have to use Beryllium copper

https://en.wikipedia.org/wiki/Beryllium_copper

True, bad idea, anyway there's many metals for that including steel, etc.

What a fantastic machine, amazing work!

is there BOM available? at least for the of the shelf electronic parts

Relays were invented in the 1830s. Why were tubes used instead of relays? Do they switch faster?

yes, a lot faster. Mechanical relays will probably be at best <30Hz, tube circuits can do 10000Hz+

Reply on news.ycombinator.com