Addition On The Strangest Vacuum Tube

[Uniservo] made a video of a tube he’s been trying to acquire for a long time: a Rogers 6047 additron. Never heard of an additron? We hadn’t either. But it was a full binary adder in a single vacuum tube made in Canada around 1950. You can see the video below.

The unique tubes were made for the University of Toronto Electronic Computer (UTEC). A normal tube-based computer would require several tubes to perform an addition, but the additron was a single tube that used beam switching to perform the addition in a single package. [Uniservo] points out how the tube could have revolutionized tube computing, but before it could really appear in real designs, transistors — and later, integrated circuits — would take over.

The tube did see use in 1950 in a tic-tac-toe machine, but that appears to be the only practical use on record. The datasheet is actually online. However, if you really want to see how it worked, read the patent instead.

This tube was a big improvement over relay adders. If you want to compare a more modern take on an adder, here’s one in Verilog on an FPGA.

55 thoughts on “Addition On The Strangest Vacuum Tube

      1. So do I – it means that the smartest people move here either due to the best pay, the best challenge, resources and opportunity, the culture, or a combination of all of those. It was called the “brain drain” after WWII. It continues, but China is closing fast as many remain there once educated or return there after attending university in the US. Note the percentage of asian authors in US or, for that matter, western scientific papers in general.

        1. Or it means that the US has a huge legacy trust fund and just fronts the money. Our higher education is still top-notch but these days, I’d be surprised if the brain drain weren’t shifting into reverse.

          The iPhone says, ‘Designed in California. Made in China.’ Huawei’s similar offerings do not.

    1. Hijacking the top comment to say: let me save everyone ten minutes of their lives. This is a youtube video of a guy, holding a tube, and turning it various directions, while rambling. He never does anything except hold it. Doesn’t connect it to anything. Zilch.

      Youtube is not a blog, people. Don’t make me listen to ten minutes of rambling on, when what you said could be condensed into two paragraphs or maybe a minute if you knew how to edit your videos properly.

    1. A little. It brings something even more interesting to my mind. Tubes and semiconductors show us that more than one way to make an active device exists. Could there be others we haven’t thought of yet?

      1. In the 1950s and 1960s, there were lots of technologies for building logic with cool names like cryotrons, flux-summation-cores, electroluminescent-photoconductor logic, tunnel diodes, transfluxors, ferractors, and parametrons, using everything from superconductors to microwaves. The point is that many technologies beyond tubes and semiconductors were tried, but transistors won.

    1. A thousand times this!

      A lot of us read things in environments where sound would be intrusive or impractical or on machines with no speakers connected.

      Also video is just about the worst medium for technical documentation unless there is some complex mechanism or operation that needs demonstrating.

    2. Simply because video is the easiest and most effective way to get stuff out of people’s hidden collections so others can enjoy it. Most casual viewers do not want a wall of text and still pictures. Video is the way to engage.

      1. Videos are shitty. No real search, they require all your time, they are usually much slower than what I can read. And can’t be skimmed. This is at best.
        At worst – add jerky camera, finger-pointing obscuring the device/something, and other sins of poor planning/editing. But everyone can make a video, thank cellphones and youtube for that.
        It is just illiteracy at its best/worst.
        No, I do not deny that good videos can be engaging.
        Unfortunately good videos are rare, and require actually much more effort than editing a wall of text and snapping few pictures.

  1. Interesting that there’s no mention of speed of operation in the datasheet. I’d be interested in how fast it could add, compared with modern ALUs. The first CPU I programmed was a Motorola 6800 at about 600KHz clock speed in 1983. I’m sure this Additron tube could’ve beaten that 33 years earlier with no silicon (although it’s not 100% fair comparison, since that’s the clock speed of the whole CPU, not just the ALU speed)

    1. Given that it’s (electrostatic) beam deflection, I would say the tube is good for 200MHz to 400MHz with magnetic shielding.

      The real speed bottleneck would be the wiring between tubes. Including poor wiring I would give it 10MHz or less. All coaxial wiring should get it to 100MHz or so.

      1. The data sheet gives a “time factor” of 1uS or 0.5uS depending on the circuit. I don’t know if this factor is the actual speed of addition or just related to it, but it looks like the tube operates somewhere around 1MHz.

        1. That doesn’t surprise me for the point to point tag board wiring of the day.

          An old 100MHz Cathode Ray Oscilloscope (CRO) has electrostatic (beam) deflection and works fine at um 100MHz.

          The wiring is probably most of the difference though there will be a extra delay for the carry out bit.

    2. Remember this is early too, iterative development could have made it a lot better. Bearing in mind that tubes got into the tens of Ghz before they got the first transistors over half a ghz, then potentially could have been seeing 50-100GHZ clocked computers….. BUT they would likely have been very simple with few instructions…. also they’d have probably remained at least desktop size.

    3. The datasheet says 1 microsecond for an unbuffered direct connection, 0.5 microseconds when buffered with a cathode follower. That means no better than 2 microseconds for a 4 bit adder. A processor based on these would be lucky to hit 100 kHz.
      The problem is that the tube was low current, meaning that load resistors had to be high ( 63 kilohms was mentioned ) and the resultant time constant with just 15 pf is 1 microsecond. The tube would have had to be redesigned to get better performance.

  2. In the 1920-ies there was a Loewe (I don’t have German keyboard layout installed) tube that had several triodes in a single glass envelope. What was unique about it was the fact that coupling capacitors and resistors were also inside the envelope. It was used as an amplifier.

      1. Sylvania tried that with the CFF-102 Circuitron, way too late in the game. Of course, in the late 1930s there were some tubes that were almost there – the direct couple dual triodes like 6N6G, for example.

    1. I gave copper sulphide based memristors a shot a couple years ago, but both designs I actually constructed were not resilient. The first attempted to use an insulator with a pin prick hole in it, mounted over a copper pad that had been reacted with sulfur. I used a silver based paint pen to fill the hole in an attempt to make a point contact, as well as to make the connection to the mating pad. Unfortunately, any attempt to use applicable liquid conductors always resulted in absorption into the rather fragile sulphide layer.

      http://richfiles.solarbotics.net/robots/Memristor_PCB.png
      http://richfiles.solarbotics.net/robots/MemristorSulfidePads.jpg

      My second attempt had two PCBs with matched copper pads on either board, sandwiched together with aluminum balls between them, and some aligned mounting holes. I pressed pem-nuts into the lower board and used model paint to paint a mask to cover all copper not meant to be exposed to the sulfur reaction process. I learned from the first attempt, that that gets messy. What I learned from this method was that there MIGHT have been a chance to get one or two memristors working, but the entire mechanism was too unstable. Once you tightened the screws enough to get contact with more than the first one or two, you were basically pressing the aluminum balls down with enough force to break through the fragile sulphide layer. It might have worked if the aluminum balls were mounted on some manner of spring, and the compression force limited, but this design was unfortunately, a dead end. Maybe if the top PCB was made of thinner material, or had

      http://richfiles.solarbotics.net/robots/MemristorProto2_4Fit.jpg
      http://richfiles.solarbotics.net/robots/MemristorProto2_7Sulfur.jpg
      http://richfiles.solarbotics.net/robots/MemristorProto2_8Clean.jpg
      http://richfiles.solarbotics.net/robots/MemristorProto2_11FrontBack.jpg

      In the end, neither design ended up being viable, however I had one remaining idea. Unfortunately, I’ve never had a chance to try making it. The diagram ought to be rather self explanatory. By using 0.6 mm diameter aluminum balls, and 0.5 mm thick PCB material for the spacer layer and the comb layer (a thick, rigid PCB is used for the base layer), the combs ought to act as very light springs to gently apply contact force between the aluminum ball and the sulphide surface. While I came up with the idea in 2014, I have yet to learn any PCB layout software, and have been caught up in other projects (a custom mechanical keyboard, and a control+instrument panel to be mounted into my desk for the video game Kerbal Space Program.

      Personally, I think that last one has potential.

      http://richfiles.solarbotics.net/robots/MemristorProto3.jpg

    2. Well, they did have spinning drum capacitive dynamic memory in the late 1930s. Look up the Atanasoff–Berry computer (ABC). It was a literal drum of capacitors that had their charges read and refreshed on every revolution. The photo memory thing is interesting, though I’m not sure photodetectors of the era were fast enough to read the data quickly enough.

      Of course, there were later devices, like the RCA Selectron RAM tube, used in the JOHNNIAC computer, up until 1955, when they replaced the tube RAM with core memory. I once saw a former JOHNNIAC sourced Selectron tube go for over five grand on ebay. Man, I wish I could have snagged it, but it was about a hundred times more than I could budget at the time.

      Another interesting memory that they had used a Williams tube to store data. A Williams tube used a CRT to store data as “dots” on the screen, readable by the electrostatic charge on each dot.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.