Let’s Build a Vacuum Tube 1-Bit Full Adder

[@michalvarsanyi612]

No wayy, this is super cool! It is such a big reminder of what engineers back in the day had to work with. Keep it up!

[@UsagiElectric]

Thank you so much! The more I learn about this stuff, the more respect I have for the engineers back in the day!

[@CuriousMarc]

Nicely done! I love the prototyping technique on a modern proto board! Would have never occurred to me. I'd just would have gone to a metal chassis and started to drill holes for the sockets :-o

[@UsagiElectric]

Thank you so much! I think I can get away with it because I'm running things at such a low voltage. I'm not sure how confident I'd feel in this method at 150V+, haha. By the way, I'm absolutely loving the HP restoration videos, they've been an absolute treat to watch!

[@CuriousMarc]

@@UsagiElectric Indeed I was surprised by the low voltage too. How come you can get away with such low voltages when more typical tube circuits are in the 100V region?

[@UsagiElectric]

@@CuriousMarc There's a few things that play into it. First, I'm stubbornly afraid of high voltages, haha. More importantly though, I'm not using the output to drive anything heavy. With a 10k plate resistor and just 24V, the most I can get is 2.4mA of push out of it. Enough for an LED and not much else. Fortunately, the grid uses practically 0 current, so the output of one inverter can drive the grid input of another inverter with ease. Also, the tube makes a huge difference. The 6DJ8/ECC88/6922/7308 dual triode tubes work surprisingly well at these low voltages, transitioning from cutoff to saturation with a change of just about 2V on the grid. The 6AU6 and 6EJ7/6EH7 pentodes also give nearly the same response. However, the 12A_7/5963/6201 dual triode tubes are wildly different. They give much more linear response over a much wider grid input. It takes about a 6V change on the grid to get them to transition from cutoff to saturation. These are massively popular audio tubes and I can see why, the wider voltage swing on the grid can give excellent range and linearity without approaching saturation or cutoff in audio applications (I think, I'm mostly deaf in one ear, so I've done very little tube audio stuff). Interestingly, IBM often used the 5963 tube in their designs, but never at the (kind of dumb) 24V I'm at. They tended to use a 150V supply and have a logic high of 150V and logic low of 50V (that's just the tip of the iceberg when it comes to IBM supply voltages as you are definitely aware of!). With that much voltage on tap, it was much easier to get the tube to swing from cutoff to saturation. I have tested around 30 different triodes and pentodes at 24V, and really only come across a handful that work well in this type of digital application at this low of voltage. As mentioned above the 6DJ8 dual triode, 6AU6 pentode and 6EJ7 pentode work the absolute best. The 6CB6 pentode works pretty well if triode strapped, and the 6GJ7 triode pentode works pretty well if you triode strap the pentode half. The 6GH8 triode pentode is surprisingly well balanced if you triode strap the pentode half, but they still require a bit more swing on the input than I'd prefer. The 6BE6 heptode/pentagrid converter is an interesting one too. It works pretty well too and makes a neat little NAND gate, but I just don't have that many of them, so they get criminally underused in my stuff. Sorry for rambling on, but I hope that gives a bit of insight into the madness of my 24V mis-adventures, haha.

[@CuriousMarc]

@@UsagiElectric Thanks, very informative.

[@carnright]

This was a lot of fun watching :-) A while back I did a three value Venn diagram (input A, Input B, and carry-in C) which I then used to show the two outputs, sum and carry-out :-D I like how you pointed out that you can press any combination of 1, 2 or 3 buttons, that helps reduce the cognitive bias, of what each input is.

[@UsagiElectric]

Thank you very much! I always have to remind myself to test every combination of button presses to make sure it's working properly. It's so easy to do the mental math in your head and say "1 + 1 = 2, so I need to flip the first two switches", when that 1 + 1 could be on any of the two switches. Fortunately, by testing every combination, it reveals mistakes in the wiring, like that pesky 3.3k ohm resistor, haha.

[@carnright]

@@UsagiElectric exactly! Any combination of switches! 2^3 combos 😁 and you tested the no switches after the warmup 🙂

[@frankowalker4662]

Works like a charm, Looks like a fever dream I once had in the 80's while fixing a point to point 1940's LW/MW/SW Radio-Gram. LOL.

[@UsagiElectric]

Thank you! The breadboard really did get full on this one! I try not to pack them so tight anymore when building these test circuits as I end up getting dizzy trying to keep all the resistors straight, haha.

[@ropersonline]

16:14: It might be an idea to put some heat-shrink tubing on the diode's exposed legs. That'd be more Victorian, ergo more steampunk, and incidentally less short circuit-prone.

[@c1ph3rpunk]

As my dad always said, never do anything half add, anything worth adding is worth over-adding.

[@SeekingTheLoveThatGodMeans7648]

After looking at your design, I did some Boolean logic to see if there is a way to do this that requires fewer tube sections. I think there is. Let A, B, Ci (carry in) be the inputs Let S (sum) and Co (carry out) be the outputs + is OR ! is NOT juxtaposition is AND Co = (AB + ACi + BCi) any two or more being high means there is a carry out S = ABCi + (!(AB + ACi + BCi))(A + B + C) an odd number being high means the sum is high But notice this is the same as S = ABCi + (!(Co))(A + B + C) Diode logic will get you Co = (AB + ACi + BCi), and a tube inverter will get you !Co. Diode logic will get you ABCi. Diode logic will get you A + B + C. Diode logic will get you (!Co) (A + B + C) Diode logic will get you S = ABCi + ((!Co)(A + B + C)) Buffering is wise for Co and S, so that would be possibly a total of three triode sections... that's it. PS thought... a cathode follower won't restore logic level lost with diodes, however, so that would be five triode sections, leaving one triode section unused if using dual triodes. So it won't save you much in the way of tubeage, and will need scads of diodes.

[@UsagiElectric]

Straight up, that's some brilliant work! I'm blown away, that's some seriously impressive stuff. I actually went ahead and built up the circuit you described in Logisim and it works a treat: i.postimg.cc/Zqz9gMW8/DDL-FA.png And honestly, with the right values, you could probably get away with just a single tube for the one lone inverter. Everything else could be DDL. You'd have to be super precise, and more B+ voltage would give more flexibility, but I think it could be done. Of course, getting it to interface with other things would be a nightmare, but on its own, a 1-tube full adder is all kinds of impressive! I absolutely need to try to build that out to see if I can get it to work!

[@daniellevy5248]

Give it a shot. I'd be curious how far it can be pushed. There's a limit on how much you can chain diode logic till you compromise your logic levels (a full adder is typically used in byte or larger arrays) but with careful design you don't have to restore logic levels at every point. Those old tube computer engineers had to think through things that we don't have to any more with zillions of transistors to throw at problems.

[@UsagiElectric]

​@@daniellevy5248 As soon as I can free up some breadboard space (and steal away a bit of time), I definitely hope to give it a go! The SUM outputs would definitely need to be buffered at a minimum and restored and then buffered at worst, but the big question is, could the carry out be ganged into the carry in of another adder without causing issue? There's only one way to know for sure and that is to build it!

[@SeekingTheLoveThatGodMeans7648]

By all means fiddle around more with the logic and electrical engineering math. As another observation: in a larger architecture it can also make sense for some signals to be active low. Every trick to cut the number of tubes for a functionality buys a hobby project like this more mileage.

[@SeekingTheLoveThatGodMeans7648]

Personally I haven't been quite as afraid of higher voltages, but I have drilled well in how to respect them. I wouldn't run high voltage in a board that I need to directly touch while operating, even if it means lever handled switches rather than push buttons. I like those cool 300v modules but in my day I would have gone more modest by using a half wave voltage doubler with 120v line power. That was a common drill in transformerless TV designs in the USA.

[@duality4y]

this channel is so underrated :)

[@UsagiElectric]

Thank you so much!

[@tstahlfsu]

Great video! If you used pentodes, could you have done away with the diodes? Just curious if you could make it with only tubes, resistors, and capacitors (if necessary).

[@UsagiElectric]

Thank you very much! You could actually build the entire thing with tubes, resistors and caps (although I don't use any, there are a few places that I probably should be). The design itself wouldn't even have to change, just instead of using silicon diodes, you could use the 6AL5 dual diode tube. I've actually built a few NOR and NAND gates using both a 6DJ8 dual triode as well as a 6AU6 pentode couple with the 6AL5 dual diode, and they work really well! The main reason I didn't do it here is it would take 6x of those little 6AL5 tubes, making it quite a bit larger! Although, the tube diodes may actually work a bit better because there is zero electron leakage in the reverse direction. Alternatively, you could also use a tube like the 6JU8, which is a quad-diode in a single 9-pin tube. In which case, you'd just need three of those. That would be very cool indeed, but getting all the jumpers in the right spots would be a mental exercise for sure, haha.

[@adrianobueno6984]

Sorry my stupid question: Why an inverter after the NAND gate? Wouldn't be better to use only an AND gate?

[@topilinkala1594]

One triode makes a NAND-gate. So to make an AND-gate from triode you need an inverter. You can't make an AND-gate out of one triode?

[@erikr007]

Nice series of videos, but please say "K" instead of "thousand" -- e.g. "a 33K resistor" instead of "a 33 thousand ohm resistor" or even just "a 33K".

[@UsagiElectric]

Thanks for checking the videos out! I'll be honest, I never even noticed that I was doing that, but I'm fairly certain I get better about this in later videos.