Vacuum Tube Computer P.13 - Building a NAND Gate to Control the Input Enable Register

[@pnpplpnppl]

Your enthusiasm is so infectious. I often watch your videos to get motivated to work on a project called nand2tetris, so this was a highly appropriate entry.

[@UsagiElectric]

Nand2Tetris is an amazing resource and serious project to undertake! Looking forward to your progress on that project!

[@pnpplpnppl]

@@UsagiElectric Aw thanks for the encouragement! :3 I'm on chapter 7/12, translating VM commands to assembly. Learning Rust very slowly. It's been a great way to jog my memory after 10 years neglecting CS.

[@jorgeferreira6727]

Hi. Have you considered/tested the option to build an AND gate by just skipping the inverting amplifier and applying the output of the diodes circuit directly to the grid of the cathode follower? Possibly with a slightly smaller pull up resistor on the diodes.

[@ooze9808]

ummm.... i just want to say, this is officially the coolest video series on youtube

[@UsagiElectric]

Thank you so much! I'm quite happy with how this thing is coming together, though it's getting properly large, haha.

[@HoffmanTactical]

Super cool! I need to go back through your vids on this.

[@remyluisant2264]

Thank you for your optimism.

[@UsagiElectric]

Thank you for checking the video out!

[@powerofthemind21]

"Quiet" interesting series!

[@SeekingTheLoveThatGodMeans7648]

blinkenlights and glowentubes... what's not to like

[@UsagiElectric]

The more blinkenlights and the more glownetubes I can get into this project, the cooler it's going to look!

[@possible-realities]

Nice video! Thinking more about NAND gates, I'm starting to think that using NAND gates as the default kind of gate might actually give better properties than using NOR as the default, if done correctly. My basic reasoning is that the inverting outputs that you get with tubes should be much better at sinking current than sourcing it, and the NAND gate needs inputs that are good at sinking current but don't need to be good at sourcing it (unlike the NOR gate, which needs the opposite). That should allow for a lot more fan-out before you need to buffer the outputs. (The output must be better at sinking current because it can overpower the plate resistor, which is what sources the output's current. ) If you don't need your outputs to be able to source much/any current, you can have a much bigger plate resistor, reducing power consumption. But seeing as the heaters are the main power draw, reducing the number of tubes through less buffering should be the major gain. I also think that to reduce buffering needs, you should try to design your gates to load the inputs as little as possible (but at the same time be careful with using too big resistors, because they will likely make the gate slow - like the 1 Mohm resistor in this video will probably take some time to charge whatever capacitance is there). With the NAND gate design in this episode, each input feeds into a voltage divider before the diode, which will always load the input. I think you should be able to design it so that the voltage divide is after the diodes instead. You'd put a pull-up to plus 24 V on the other side of the diode, and a voltage divider between the diode and -12 V to get the grid voltage. Then, the inputs are only loaded when they are pulling the grid low.

[@frankowalker4662]

Brilliant. I realy love the look of the PCB's you make. I think you might need a custom PSU for it soon, Maybe with a valve rectifier that warms up gradualy ? :)

[@UsagiElectric]

Thank you! The curvy traces are sometimes a little difficult to design around, but they just look so cool! I've got a few ideas I'm working on for the soft-start, but I think the easiest way is going to be using a couple of relays. Though, there may be some Thyratrons involved to ensure the proper time delay!

[@frankowalker4662]

@@UsagiElectric Good idea with the relays, You'll end up with a clicky start-up sound. :)

[@CrashPilot1000]

That rabbit is solid.

[@UsagiElectric]

She's so spoiled, but by far the most important member of the Usagi Electric team! (And by team, it's just the two of us and she always gets top priority, haha)

[@CrashPilot1000]

​@@UsagiElectric Two of you are enough!!! BTW I have been developing Quadrocopter software for an open source project (so I had to learn C first to make that piece of sh** fly better) - I find your work great, but I am not an electronics guy but hey, you can hit me up on discord and we have a chat --- Well German is my 1st language, so.... just if you are interested.

[@CrashPilot1000]

BTW: I had 3 cats and one dog, they all ran by and decided to stay - so never bought or adopted a pet - and a rabbit never stayed as far as I know.... but I had ducklings...

[@leyasep5919]

Hi ! Have you ever considered using a pair of mutually-blocking tubes to implement XOR ?

[@UsagiElectric]

Hello! If I'm thinking of the right think that you're thinking about, I have! I actually have a short 1-minute video on building that: kzbin.info/www/bejne/d3_VhIKinKyGhbc

[@leyasep5919]

@@UsagiElectric well indeed that's that thing :-D Well done !!!

[@PaulasTechStuff]

very nice work!

[@UsagiElectric]

Thank you very much!

[@gammaleader96]

As always, nice progress. One question I had for a while now: Is there a specific reasen why you use solid state diodes instead of tubes? I like the way you layout your boards, it gives them a kind of retro style with the organic curves. Very fitting for such a project. Greetings, Michael

[@UsagiElectric]

Thank you so much! There's a couple reasons I'm using silicon diodes instead of tubes, but the primary one is cost. I could very easily build everything I've built up to now using the 6AL5 dual diode tube, but that would end up doubling the amount of tubes being used, which would require more than my budget is happy with, haha. It also would require pretty large rethinking of the power supply since the filament power draw would also be doubled. But, it's not unheard of for vacuum tube computers to use silicon diodes. The IBM 604, 650, 700 series machines regularly used silicon diodes in conjunction with vacuum tubes, much the same way I'm doing here! Granted, they often used Germanium or Selenium diodes, and I'm using Schottky diodes, which generally use more exotic metals. But, the design could very easily run vintage Germanium diodes if I could get my hands on about a thousand of them! Thanks for the complement on the boards! I really like the look of the curvy traces, though sometimes I have to use a proper 45-degree angle in some places. Still, I think it really fits the theme of the project though!

[@rickhole]

@@UsagiElectric Semiconductor diodes were available long before the transistor was invented and were used in tube circuits, so you are good to use the diodes and still have a VT computer

[@skeezixcodejedi]

VFDs are _always_ a win. I'm amazed you designed all those little boards so you coudl test them independantly, and lego them all together into that pattern (and others, tubes are pricey I guess, so you want these pieces reusable?) Which logic software is that for testing? (you're going fast, making videos, makinmg designs, making boards.. my goodness! And a giant computer to play with!) You're home etchign boards.. using a home mini CNC with a dremel or something? or are those etched using chemical submersion?

[@UsagiElectric]

Thank you! And I agree, VFDs are always an excellent choice, haha. The primary limitation I have is that all my designs must fit within 150mm x 100mm as that's how big my single sided PCBs are that I start with are (with some exceptions - I've got a few long skinny boards that I cut out of a larger stock). That plays well into the idea of building individual modules on a single board. That also helps with trouble shooting - if the IEN register is having an issue, I can just replace that board instead of having to reproduce everything. The logic software I'm using for testing is called Logisim ( www.cburch.com/logisim/ ) and it has been an absolute life saver! It's easily one of my most used programs, and for being free is surprisingly capable! I'm actually using a full size Bridgeport EZ-Trak 3-Axis CNC milling machine to cut my PCBs, which is way overkill, but it's something I had ready access to. I design the board in DesignSparkPCB, then use FlatCAM to convert the gerber files to G Code, then I just run out to the garage and set the board up for a cut! I actually show the process in a bit more detail here ( kzbin.info/www/bejne/hYPWdYqlfNl4qbs < Should be timestamped to the right spot).

[@markpeter9919]

thank you for this exciting video :-) what software are you using for electronic simulations ?

[@AlejandroLopez-qd3xm]

Very nice addition to the proyect. When completed its gonna look like those model train setups! Will need a pretty big table, just for the processor ! What kind of implementation are thinking of for the USAGI 1405? Is it going to be also tube based?

[@UsagiElectric]

Thank you! Since I primarily cut all my boards out of 150mm x 100mm PCB stock, it ends up with this really weird lego like construction, but I kind of like the ability to swap and change parts if needed without having to rebuild everything else. Plus, it feels cool to connect pieces together like a puzzle, haha. The ultimate goal is to have the full computer split into three separate modules: microprocessor, program control, and memory. And, the hope is to keep it all based entirely on vacuum tubes (with some silicon diodes to help keep the size of the project somewhat manageable). In the end, I'm thinking it'll be around 600 tubes, but that could definitely balloon up some. So far though, we're only at 80 tubes used, so we've still got a long way to go!

[@AlejandroLopez-qd3xm]

@@UsagiElectric That's thinking ahead!. I'll keep tuned!. Its a very interesting way of getting into how a microprocessor works, by splitting it into its very minimun components. In these days of múltiple core, zillion transistor chips, processors looks hard to understand. This type of exercise, is very educational. Vacuum tubes just make it cooler! (warmer?)

[@aislanorberto9440]

What name about this simulation software?

[@Clancydaenlightened]

3:20 so would you called that Diode-Valve-Logic since DTL is already taken by transistors? kinda a combo of DTL and RTL with diode-resistor logic, imagine building this cpu out of diode resistor logic gates, you'd need a negative voltage rail to implement any inverter based logic, and some zener diodes, so one might be able to nand, not, nor wit dat shit

[@Clancydaenlightened]

it would eat kw of power and be slower than frozen shit during middle of winter in antarctica too