In a few museums around the world, there lies the special experience of seeing some of the earliest computers. These room-filling monsters have multiple racks of vacuum tubes that are kept working by the dedication and care of their volunteer maintainers. A visit to the primordial vacuum tube computer, Colossus at Bletchley Park, UK, led [Mike] on the path towards designing an entirely new one. He thinks it’s the first to see the light of day in over five decades. ENA, the Electron tube New Automatic Computer, is the result.
It uses 550 Soviet 6N3P double triodes, and its 8-bit Von Neumann architecture is constructed from the tubes wired up as 5-input NOR gates. ROM is a diode matrix, and RAM comes courtesy of reed relays. The whole thing is assembled as eleven PCBs on a wall-mounted frame, with a console that holds the piece de resistance, a display made from an array of LEDs. A Pong game is in development, meanwhile the machine makes an impressive room heater.
If you’d like to see some more vacuum tube computational goodness, we saw Colossus at the National Museum of Computing, back in 1996.
47 thoughts on “The First New Vacuum Tube Computer Design For Well Over Half A Century”
It gives me a warm feeling just thinking about it.
You’ve definitely got a glow about you
Maybe I’m biased, but that grid of tubes on a PCB warms my cathode.
I’d like to try something this, but I already have too much on my plate.
Talking about a warm feeling, I wounder how much heat all those vacuum tubes make.
About 1200 Watts, according to another comment. But that’s just filament power. I don’t know what plate current and voltage he’s using, so no basis other than wild guesses about the B+ power. It is probably less than the filament power, but still significant.
“Who tried to divide by zero??”
looks emp proof to me, though i’m not sure about the diode matrix.
“What’s it do?” means.. “I’m bored”
“Why did you build it?” means.. “I’m really bored”
“Should it spark like that?” means.. “I’m really scared”
“The Ena.Computer is designed using 550 double triode 6N3P electron-tubes, all configured as identical 5 input NOR gates. Registers and counters are built from these single NOR gates”
Did I miss a schematic of these modules at least. All I see on the web page is a single block diagram.
At the end of the technical page he says he hopes to “..add a decent schematic drawings page to this site.”
More exercises in futility. We I suppose being retired is boring enough certain people would find this productive. I’m wondering what it would be rated at speed wise. I also have flash backs of old Scifi movies and Star Trek hearing all those computers bleeping and squacking and humming.
More productive than watching reruns of MASH or Gilligans Island.
Surely there’s a way to 3D print a core memory. How cool would that be?
That would be DAMN cool.
I wonder why Mike called it “ENA” and not “ETNA”….. not tempting fate maybe.
So I’d like to see the clock speed at around 50-100MHz, it should be reasonably nifty at that speed.
Would need much faster RAM, would be surprised if relays could keep up even at 5kHz.
But here’s a wacky idea – could you build dynamic RAM with discrete components? It’d be hilarious to see an array of 100nF ceramics that each hold 1 bit. Addressing the bits seems easy, straightforward logic. The tricky part, I think, would be getting a tube to do what the per-bit FET does, i.e. make a switchable, relatively low impedance connection from cap to the data line, one that can pass current either direction. But hey, crack that and it’d probably get you into at least the MHz range.
Parallel tubes connected cathode to plate, common grid, to replace single FET.
100 nF would be too much capacitance, it would slow down charge time.
Yeah I realized after hitting post, smaller caps equal faster operation, and that 100nF would be kind of stonkin large. Gotta keep that RC constant low.
Also got wondering, would discrete caps be less leaky than what you get in integrated circuit DRAM cells? If so, refresh rate might actually end up lower.
My guess is that it would be difficult to achieve the degree of insulation outside of the capacitors, that can be had with hermetically-sealed DRAM chips.
Well… it runs just about as hot as the volcano, so…
Using reed relays for program memory could allow for fast operation, as long as there is an electronic form of data memory. Just the same, scale that back from MHz to kHz and you’re in the right ballpark.
Cryptocurrency mining ? No, don’t try that…
Noooooo I said !!
That’s a fantastic build. I like his control panel too.
Does anyone else remember the very underrated (IMHO) movie “Colossus: The Forbin Project”?
That was a favorite of mine
If I remember correctly, at the time “Colossus: The Forbin Project” was made, the existence of the Bletchley Park computer by that name was still classified, so the name was a complete coincidence.
It might not have been a coincidence. D.F. Jones, who wrote the novel the movie was based on, was a British naval commander in World War II and it’s quite possible he knew about Bletchey Park.
While building everything from NOR gates is certainly possible, it’s kind of inefficient. It would make more sense to also design a flip-flop from vacuum tubes, which would almost certainly use fewer tubes than a NOR gate implementation of a flip-flop.
Apollo Computer was built with NOR gate only
Hard to gauge size but one of the “pencil” tubes would have been interesting.
I don’t like they have flexible leads – means replacing one is not just “pull out and plug in”.
And putting the word “audio” in front or behind anything to do with a vacuum tube adds at least one zero to the price!
A fascinating project! It’s a great illustrator of what it takes to make a computer with vacuum tubes.
There actually were vacuum tube static and dynamic RAM memories. For example, the Williams tube (invented around 1947) was a dynamic RAM that stored 1024 bits. The Whirlwind vacuum tube computer project (early 1950’s) developed a static vacuum tube RAM that stored about 512 bits; but core memory was invented about that time and replaced the vacuum tube memory.
Calculating Pi could cause a nationwide blackout… 😁
One thing though…. why does it keep flashing on its display… “There is another…”
Hmmm, 6N3P needs 350mA (+/-35mA) @ 6.3v for the heaters, so for 550 tubes that is about 190A!
Or a little over 1200 Watts, if you’re at all concerned about operating cost.
I don’t know which is more impressive, building it or getting all the tubes!
Buy them from Russia.
Still that they’re making tubes at all, Shockley would be shocked!
Well, some years ago there has been the MUNIAC: https://de.m.wikipedia.org/wiki/Datei:GfhR-Muniac.jpg
And in august 2021 there will be another new/old one at the “Heinz Nixdorf MuseumsForum” in Paderborn (Germany). I saw that one working two weeks ago.
You’re not from around here, are you?
And yet you ask ‘why?’.
I spent too much time at SAGE radar sites with BOURROUGH T-2’s to get excited about this computer.
Yeah, not too many people who had to work with vacuum tubes that are nostalgic about them.
The original Eniac was made up of over 17 thousand vacuum tubes! That should have been preserved in a museum!
Please be kind and respectful to help make the comments section excellent. (Comment Policy)