It’s a retro mod! Frankly, I am surprised it wasn’t designed like this in the first place. Your design is so much better!
@UsagiElectric3 жыл бұрын
Thank you so much! You know, I'm not great at CMOS logic, but I'm sure the designers could have implemented this without running into too extreme of an increase in transistor count and definitely could have kept it all within the same 16-pin DIP package. It would have been a much more universal chip! Hmm, now I want to build a transistorized version of the new UE14500 design...
@CuriousMarc3 жыл бұрын
@@UsagiElectric Uh oh, thinking about using that newfangled solid state technology? It's all witchcraft and company you know.
@UsagiElectric3 жыл бұрын
@@CuriousMarc I can feel the dark side calling to me, enticing me with the speed and power attainable through the ways of silicon.
@gcewing3 жыл бұрын
I don't think the chip was intended to be a general-purpose CPU, but rather to replace the kinds of industrial control systems that used to be built from relays. There's little to no need for arithmetic in those applications, but there is need for efficiently implementing lots of logical operations.
@dans82873 жыл бұрын
You may want to consider building in separate filament power to any additional expansion boards. That way you can power the filaments directly from a filament transformer. The way this project is growing who knows how much power you will eventually need for the filaments. Right now you're almost at 130 watts of filament alone. Sorry I don't mean to come across as a know it all it's your project please do what you want with it. I just wanted to throw that idea out there just in case you hadn't thought of it already. Thanks for doing these videos I always look forward to the next episode. This is such a cool project!
@UsagiElectric3 жыл бұрын
Thank you! And no worries at all! I love hearing good ideas as I'm literally just making this up as I go! Unfortunately, the way I've kind of designed the entire thing doesn't lend itself well to using a separate power supply for the filaments. The filaments all draw off the primary B+ power rail, which is just 24V. So I run four filaments in series from B+ to GND. It's definitely getting a little ridiculous in terms of power draw, but fortunately, finding a beefy 24V supply isn't all the difficult. The supply I'm currently using can supply up to 15A, which is plenty for what I have now. If it gets a little tight on power availability in the future, I can always run two of those 15A supplies in parallel. The concern then becomes if the PCB traces can carry enough current for the entire system. That's one of the reasons I'm trying to keep as many power distribution rails spanning left to right across the system as I can. Each 1mm trace can handle about 3A of current, and each large 2mm trace on the end can handle around 6A of current. If the PCB traces do start to get a little melty, I can always solder some solid core wire along the trace to help carry more current, but my hope is that distribution will be sufficient enough that I won't have to do that. The upside of doing it this is that I can keep the PCBs and the filament power design really simple and compact. So, essentially, I traded simplicity in the PCB design for complexity in the power supply, but I think that was an alright trade. As for the power draw, I think I'm actually closer to about 115W because I'm slightly undervolting the filaments, but you're absolutely right in that that's still a whole lot! However, at 24V, that's only 4.8A of power draw, and the supply can reliably give 15A, so I'm only at about 1/3 the capacity of my supply (the reality is a bit different as I currently have 60 tubes on the 24V supply and 8 tubes on the -12V supply, so the 24V supply is only drawing around 4.2A, which is pretty close to what my salvaged ammeter is showing). But, who knows how things will react as I get bigger and bigger! So, I'm just going to keep plugging away at it and hopefully it all keeps going smooth!
@advarkmerrygoround14252 жыл бұрын
Nomenclature should really be UEV14500, V for valve. Such an excellent insight into logic and electronics. You encompass first principles in such an amazing and informative way. Keep up the good work Sir. :-))
@iygor3 жыл бұрын
Beautiful...
@UsagiElectric3 жыл бұрын
Thank you!
@antonnym2142 жыл бұрын
I agree with the additional Add/Subtract, and Force One instructions. Will you have a Not instruction also? My name for your processor is the Usagi Rabbit. I see that's redundant, but you be the judge.
@rickhole2 жыл бұрын
can use XNOR
@clifffiftytwo3 жыл бұрын
Redesigning the logic unit: I have no idea of the history of the 14500 and what the designers intended or accepted as limitations. Coming up in the power industry, everything was relay logic from the 50s and 60s and I wonder if that influenced the thinking of the early electronic designers - was the 14500 intended as a replacement for that concept. Electrical protective relays were quite sophisticated but analog in function - capacitors and inductors and the magic you could get that way, but again, someone hands you digital logic capability and your first reaction is to emulate what you know. What an interesting evolution you've made -
@UsagiElectric3 жыл бұрын
You actually totally nailed the original intended use for the MC14500! Motorola dubbed it an Industrial Control Unit with the intention of having it replace relay logic designs from the 50s and 60s. That's why having an arithmetic unit wasn't a priority in the design. In the handbook they even state that if the system requires arithmetic, it's best to add in a separate unit for that. Fortunately, since we're building something from the ground up and we're not beholden to tight budgets or orders form on high, we have a bit of freedom to modify the design to better suit our needs!
@mikegLXIVMM3 жыл бұрын
Sometimes I wonder what things would be like if the transistor came out later, like in the 1980's. Would they have developed some kind of tube IC or some kind of large tube array?
@UsagiElectric3 жыл бұрын
Interestingly, there's actually two types of tubes out there that can give us a bit of insight into what could have been. The first are Nuvistor type tubes. These were tiny, metal can vacuum tubes that were built entirely in a vacuum, meaning they didn't need any ports to evacuate them, allowing them to be really tiny. They were introduced in 1959 and started seeing use in the 60s, and were roughly about double the size of a metal can transistor, despite having all the same components of a triode (or tetrode) stuffed inside. There's some great information on them over at r-type: www.r-type.org/articles/art-150.htm And the other tube that's really fascinating is the Additron. It was designed in the 50s, but never made it to production. It's essentially using very uniquely designed, interlocking plates and grids to create a special beam-forming type of tube. The result is a binary full adder stuffed inside a single glass envelope, kind of like an integrated circuit inside one tube! The US patent has some excellent information on it, but I also did a bit of a write up on it over on my Twitter: twitter.com/UsagiElectric/status/1374087658671050754
@AI_Image_Master2 жыл бұрын
@@UsagiElectric Try getting a few billion tubes in a small space like we can do with a transistor. Also remember that there is no way that we get to the moon without the IC. Couldn't be done without the onboard computers. No way that it could be done with tubes, no mater how small we may shrink them. I basically built the same type of computer, though 4bit, but with discrete transistors. Even that example showed the wonderment of what we have done today. A tiny stick containing tera bytes of data in my pocket.
@rickhole2 жыл бұрын
@@AI_Image_Master It seems vacuum tubes would have reached their Moore's Law minimum rather quickly.
@tstahlfsu3 жыл бұрын
Love the progress! Great vid as always!
@UsagiElectric3 жыл бұрын
Thank you very much!
@frankowalker46623 жыл бұрын
This is going to get big. LOL. You should call it the UE14521, 21 being the year it was born. :)
@UsagiElectric3 жыл бұрын
You nailed it, it's going to be huge! It may have been born in 2021, but I'm not sure it'll be complete in 2021, haha.
@Clancydaenlightened Жыл бұрын
11:30 USG14500-TV cuz its a 14500 implementation using thermionic valves
@ArtemKashkanovLive3 жыл бұрын
Nice progress!
@UsagiElectric3 жыл бұрын
Thank you! I'm totally making this up as I go, so I'm sure there's a lot of areas where things could be done a lot better, but all in all I'm quite happy with how it's evolving!
@JohnDlugosz2 жыл бұрын
11:30 It reminds me of "Bit Slice" solutions. And it's not a microprocessor. It's a 1-bit ALU.
@winstonsmith4783 жыл бұрын
Fantastic! Loving this.
@UsagiElectric3 жыл бұрын
Thank you so much!
@treadmillrepair7543 жыл бұрын
Awesome.
@UsagiElectric3 жыл бұрын
Thank you!
@SeekingTheLoveThatGodMeans76483 жыл бұрын
The kit idea is cool. If you're going to do that, though, consider whether you can design them to use modern production audiophile/guitar-amp vacuum tubes (JJ, Svetlana, Reflektor, etc.) to avert chancy supplies of surplus tubes whose quality and quantity are steadily declining. You might be able to get scads of workable 6AU6's now, but that won't be true forever. Most of the cost won't be the tubes in either case.
@UsagiElectric3 жыл бұрын
The kit idea is pretty much only going to be for the 555 timer and maybe a few other small things in the future. And initially the kit will be based around small pentodes, but I do hope to expand the design in the future. Unfortunately, at the low voltages I'm working at (+24V/-12V) dual triodes just aren't happy. The screen grid on pentodes acts as a sort of electron accelerator, which really helps the tube operate as planned despite such low plate voltage. The 555 kit currently is planned to be compatible with the 6136, 6AU6, 6CB6, 6DT6 and 8136 (hopefully). I have the first prototype boards for that showing up soon, so I'm quite excited to start testing! In the future, I also hope to expand to include some 9-pin pentodes like the EF80, 6EJ7, 6EH7, etc. Now, for the tube computer I'm building here, there's no plans at all to make this a kit, it's just way too intense of a project, haha. For this project though, the 6AU6 makes the most sense. Price-wise, I can often find the 6AU6 for around $0.60 to $0.80 a tube. Dual triodes are all over the board, but the cheapest new tube that I've found (the 6N1P) is usually around $3 to $10 a tube. At the scale I'm building here, that's just beyond my budget. Though, I do have some 6N1Ps in my collection, and I have done some testing with them, but since they're triodes, they have a really hard time at the low voltages I'm at. That screen grid makes a huge difference, particularly when pushing the tube into saturation, where the plate potential drops really low. Also, the sharp cutoff characteristics of pentodes lend themselves really well to digital applications like here as opposed to the very linear characteristics of audio triodes. For example, the 6AU6 needs a little less than a 2V difference on the grid to go from full cut-off to full saturation. The 6N1P needs a larger than 5V difference on the grid to do the same. That makes the design much more difficult!
@Manawyrm3 жыл бұрын
That's awesome, I really like the modifications. Looking forward to seeing some simple algorithm like Fibonacci's sequence or something similar running on the tube machine :)
@UsagiElectric3 жыл бұрын
Thank you! The Fibonacci sequence is actually one of my ultimate goals! It requires a lot of components to be working together precisely to accurately run. The biggest hurdle is going to be memory, but I have a few ideas for how best to go about making some memory that'll work that we'll get into after the main processor is finished.
@rrb65443 жыл бұрын
That green glowing mini tubes are so cute! Never seen before, which type are? Neon?
@UsagiElectric3 жыл бұрын
They are great aren't they! They're actually little Vacuum Fluorescent Displays (VFD). These in particular are Russian made IV-15s, which I believe are copies of the DM-160 that was developed in Japan. Neons and nixies need a lot more voltage than I have available on the board, but these little VFDs work great with just 24V!
@fabian999ification3 жыл бұрын
I really like this project! To help with the space issue (it is indeed getting rather large, but that's cool!), could there be a backplane system where the boards are stacked (stacked like food plates) with a main bus board connecting to each board? This was the 'de facto standard' way of building tube and minicomputers back in the day. With the heat issue, huge fans could be used!
@UsagiElectric3 жыл бұрын
Thank you! Fortunately, I live on a ranch in the middle of nowhere, so space is less of a priority than it normally would be. So, the goal is to try to build it to be as much on display as possible! My current idea is to build a rack like this one (i.postimg.cc/QxzkgnMv/Rack.jpg). Only instead of having flat shelves, I'll have multiple angled shelves and each shelf will contain some part of the machine (processor, memory, program counter, etc.). That way it's as much on display as possible while also still being accessible for troubleshooting!
@fabian999ification3 жыл бұрын
@@UsagiElectric Wow that's an interesting idea! I like the spaced-out rack idea. Good for visuals and debugging/troubleshooting! This computer is really unique in just about all aspects! A question I was thinking of a long time is how is memory or RAM going to work? Vacuum tube memory seems like it will be the biggest unit, even for only a few hundred bytes. Even capacitor memory (similar to DRAM) that uses an array of electrolytic caps is big as well, maybe 1 bit per square centimetre on a board. Sorry for the questions, I've just been really interested!
@UsagiElectric3 жыл бұрын
@@fabian999ification Thank you so much! No worries on the questions! I'm more than happy to see y'all getting engaged and excited about the project! RAM is definitely going to be a hurdle. Vacuum tube memory is a little intense, though I do plan to build some general purpose registers using vacuum tube flip flops. I need to work on trying to get the tube count down as low as possible for those, but I think at a minimum, even with some clever designing, it'll still be at least 4 tubes per bit. So, if I have two 6-bit GPRs, that's 48 tubes, not including any control glue logic! So, for RAM we definitely have to think outside the box. From a design goal perspective, I would like to have 64-bits of RAM accessible with a 6-bit address.That's an absolutely tiny amount of memory, but it should allow the computer to do some pretty interesting things at least. But, 64 bits of memory, if done with tubes would be 256 tubes at minimum! So, the current idea is to build the RAM using VFDs and LDRs as SR Latches. The LDR is used to create a voltage divider for the plate voltage of the VFD, if an external positive voltage is applied to the plate, the light emitted by the VFD creates a positive feedback loop, latching the VFD on. Then, a quick negative pulse into the plate turns it off and the feedback loop is broken. Here's a small proof of concept: imgur.com/gallery/TBpFgNf That's essentially three SR Flip Flops! To recreate that with the 6AU6 would take six tubes, so it's massively more compact (and cheaper) than regular tube memory. Granted, there still needs to be a lot of glue logic developed to actually read and write to the memory, but it's a start. I have done a little designing of the logic to read the value stored in memory here (the lines look super crooked, but that's just an optical illusion): i.postimg.cc/fLCRBfk0/RAM.png The layout of that design right now is really inefficient because I was trying to make it all fit on one screen in Logisim. The proper design will be a grid of 8 x 8, which only needs around 40 tubes to select and read the value stored. But, if I wanted to double up and use an 8-bit address for 256-bits of memory, the number of tubes only jumps up to about 80. So, it scales really well! Now from a space perspective, there's no good way to keep this compact. Even with just 64-bits, it's still going to be very large! But, the VFD/LDR combo sits very close to the board, so I have a lot of options for how to stack them to keep the overall size compact enough to fit on one of the shelves ultimately!
@m.l.5284 Жыл бұрын
@@fabian999ification Oh wow. Discrete transistor CPUs are not retro enought. Now we need tubes. What's next? Relais?
@colinstu3 жыл бұрын
What's the worry about inrush current? Overloading powersupply or damaging the processor itself? Also what can be done to deal with the inrush current?
@UsagiElectric3 жыл бұрын
Eventually my power supply is going to think that I've got a short circuit and refuse to power up. The tube processor itself is just about bulletproof and pretty tough to actually damage. I've been tossing around a couple of ideas, and really, I just need to get a little power into the filaments for their resistance to change pretty dramatically. So, I've been thinking about building a time delay relay circuit that runs the primary supply through some high wattage dropper resistors before clicking over and latching on fully. That should get some heat into the filaments so that when ti kicks over the inrush current is minimal.
@jorgeferreira67272 жыл бұрын
You feel the lack of the arithmetic operations because you are thinking computer and Motorola was thinking industrial controller, like in PLC. Your tube boards are quite nice, but the cutest thing around are the bunnies. :)
@user-folk19873 жыл бұрын
Good
@UsagiElectric3 жыл бұрын
Thank you!
@bronka423 жыл бұрын
Would you consider using a ptc to relieve the inrush current?...by the way, you're insane ... keep up the good work!!
@UsagiElectric3 жыл бұрын
Haha, thanks! I actually think that since the current is going to be quite high once finished, it might be best to just build a time delay relay switchover. When first turning the power on, initally power will flow through some big power resistors, dropping the voltage the heaters see to something quite low. Once the heaters get a little warmth in them and their resistance climbs, the relay will switch over to full power. I like this cause it's a super simple approach that should be relatively easy to build (plus I get to hear an audible click clack when the relays switch over which is just cool).
@bronka423 жыл бұрын
@@UsagiElectric I like the concept ... are you going to use neon lamps for the blinkenlights?
@UsagiElectric3 жыл бұрын
@@bronka42 Thanks! Unfortunately, I'm running the entire system at incredibly low voltage - just +24V and -12V. So, I don't have enough voltage on tap to power some neons. Instead, I'm going to use more little VFDs like you can see at about 17.15 in the video. These require very little current and work great at 24V!
@rickhole2 жыл бұрын
@@UsagiElectric I thought that also. You could put the inrush limiter resistor in the transformer primary where the current would be less hence the resistor value higher and easier to source.
@RWBHere9 ай бұрын
Wow! You're already at 80 tubes, if you include the four VFD's. It will keep you warmer in Winter. I see that your logic levels are beginning to degrade a little. It will be interesting to see which method you choose for restoring them, if they become problematic.
@janikarkkainen39043 жыл бұрын
Had to recheck the title of the PCB timelapse music... "Why does this sound like FF7 music"
@UsagiElectric3 жыл бұрын
The Shinra Shuffle! Definitely check out OCRemix, they've got a ton of fantastic game remixes! (And speaking of FF7, I've actually avoided the Remake so far, so that way I can buy it all at once and binge the entire thing once it's fully released. I can't wait!)
@advarkmerrygoround14252 жыл бұрын
Question? As I always watch KZbin not signed in and never watch adverts as I have Goshtery and Addblocker on. Are the real audience for this content actually counte````d? I was a traffic analyst for BBC and Unique users was the only way of counting correct traffic. But as this sight would appear to people like me who dislike adverts, I'm sure the registered views on this sight are way higher than are registered. My apologies, I'll subscribe
@rickhole2 жыл бұрын
Yes. UE14500, but it also needs a nickname, such as BunnyOne. By the way, I am getting BORED for the lack of smoke! Not.
@mylestechnological70313 жыл бұрын
I think I may build one too in about 2 years I really don't have a job and I'm too young to get a job
@UsagiElectric3 жыл бұрын
You're never too young to start building crazy projects! Logisim is an awesome program that's totally free and can really help with visualizing ideas.
@mylestechnological70313 жыл бұрын
@@UsagiElectric that may help if only I had a computer I do everything from my phone
@UsagiElectric3 жыл бұрын
@@mylestechnological7031 Hmm, Logisim and many other excellent design programs are available on Linux and have very low system requirements. It could be a lot of fun to build up a Linux machine out of scrap components that people don't want any more (or even grab the Raspberry Pi 400). Then again, I'm a bit biased as I'm terrible at using a phone for anything!