Vacuum Tube Computer P.30 - Connecting 2-bytes of Tube RAM to the Processor

[@DasIllu]

Tubes as memory wasn't the only thing employed in that era. I remember seeing a documentary showcasing acoustic memory. It was long tubes filled with mercury (IIRC) with a driver and a pickup each end. The signal was constantly refreshed by amplifying the output and conditionally feeding it back to the input. At a high enough frequency you could hold multiple streams of bits in one tube. Until someone slams the door shut and the entire data becomes corrupted :D It is one of the whackiest methods I've ever come across.

[@DenKHK]

Ah! That would be the mercury delay line memory. Came about in the 40s IIRC and yes it was esoteric (and whacky, by today's standards) along with the williams tube. And both susceptible to audio or visual sources of corruption - the delay line, as you pointed out, from sound, and the williams tube from even camera flash bulbs (when reporters took photos of the device). Must have been really fun times as a computer operator then... "Members of the press, we are proud to announce that our AwesomVAC has calculated the answer to the Ultimate Question! It is.... (camera flash) D'OH!"

[@onfire4Jesus2]

After seeing videos like this, now I see why early computers took up entire rooms.

[@lurkersmith810]

...And why they needed full time technicians (or engineers) to keep them running!

[@kuzadupa185]

They had big badunkadunks. And were big enough to have actual bugs inside ehm!

[@W1RMD]

A thing of beauty in a joy forever! Thanks for sharing and keeping tubes alive. Because of you, I've built a 6au6 pre amp and I can't believe how good it sounds! What a great multipurpose tube!

[@drivers99]

I was just showing my daughter this channel and the tube computer (and the “Hellorld” skit on the Centurion) and a project I made with a few KB of memory (Ben Eater’s 6502 project). We talked about TB, GB, MB, and KB and how each was 1024x smaller. So it’s funny seeing this talking about 1024x smaller yet again.

[@headwerkn]

That is by far the coolest Christmas tree I’ve ever seen ;-) Freaking awesome work!

[@davidward9935]

Really like the process you show. I started in electronics in the hybrid period of the 1970s. Perhaps when it is done, you can demonstrate lacing the wiring harness properly as it would have been in the tube computer era.

[@tomking6006]

I've long wondered what would be possible with vacuum tube technology and modern knowledge - and you're showing us! This is a truly fascinating project. P.S. I really like the curved traces!

[@KeritechElectronics]

That's definitely a lot of cool hacking! Nice to see the progress.

[@MrRlnansel]

I'm not quite sure what you have in mind for your magnetic R/W memory system, but I can foresee big troubles approaching it using any sort of reel-to-reel system (including cassettes). The main issue is the controller needed to allow "random" access to any given bit in an inherently serial medium is decidedly nontrivial. This was historically overcome in one of two broad ways. The first (and easiest from a hardware viewpoint) is to use some sort of recirculating memory where each bit in the system periodically becomes available for R/W purposes. This was the way delay line memory and magnetic drum storage worked (and magnetic disks too in the first Titan ICBM digital computer). The key to making practical systems always involved having multiple "tracks" or channels with dedicated analog & digital write, read, and regeneration electronics for each such track. One track was often dedicated to the system clock with all other bits in other tracks explicitly conditioned to it. The tracks would typically be of just two different bit lengths, so-called "minor loops" and "major loops". Minor loops were short (from say 16 to 64 logical bits long) and were most often used for implementing an accumulator or any other desired CPU registers and usually corresponded to one machine instruction cycle. Major loops were for program code and data and were commonly an integer multiple of the logical data length (plus overhead, discussed below) chosen for the minor loop. To keep the word addressing hardware as simple as possible these machines often also had a dedicated address track containing a prerecorded sequence of incrementing digital address words. This meant that most serial machines had to incorporate at least TWO address fields in every instruction, one being the instruction's own -- self -- address and the other being the address of the next logical instruction to be executed. Conditional jump and branch instructions would require a third address for the location to jump or branch to. Now, the clock track mentioned above could theoretically have been used to advance a hardware program counter, obviating the need for the address track and for the self address field overhead in every instruction word, but since everything in the machine has to be handled in bit-serial fashion anyway the added complexity of a hardware program counter wouldn't really buy much more performance. True, each program word would be shorter without the overhead of the self address, and thus each major loop could either have more instruction words crammed into it, or the major loop could be the same number of words but have a shorter period. The logical program memory available to these machines tended to be in the low thousands of kilowords at most, so the addresses likewise would be, say, 12 bits. Most of these machines had fairly long data word lengths to simplify arithmetic operations, 30 bits being common, so 12 bits just weren't that big of an overhead once you figure in the need for a next instruction field. A three-address word might consist of 30 bits for the machine word size, 36 bits for the three 12-bit address fields. Data was usually processed LSB first to simplify the arithmetic hardware, but very often an extra bit duplicating the sign stored in the MSB was included because knowing the sign of a word before the first bit was processed made things easier. The second way to deal with bit-serial storage was to throw an entire separate digital controller nearly as capable as the main computer at it. This was the IBM way.

[@UsagiElectric]

The problem here is complexity. Everything you just described adds a level of complexity and a number of tubes that simply won't fit on my wall anymore. For reference, what you see in this video is approximately 270 tubes. If I wanted to add random access program storage, even if I used something like core memory, I would need probably around 1,000 tubes minimum just for program storage. This is because it's fantastically complex, but also this being a 1-bit processor means we need programs with thousands of instructions just to be able to do something that would normally be considered simple. I'll go into a lot more heavy detail once we start the build, but the way we're going to tackle the random access problem is simply to not have it. The MC14500 architecture was originally designed to replace relay ladder logic in industrial applications, which meant the chip had to be able to handle any kind of memory or program storage type that might get thrown at it. This is where IEN and OEN come into play. It's not entirely graceful and it makes programs considerably longer, but you can do conditional skipping using OEN quite easily. Then, you can either construct your program around a loop (ie. looped tape), or you can create multiple copies of routines on the tape. So the tape will always be trundling along at a constant speed and instructions will be 100% sequential, but we can do conditionals relatively easily, which will allow us some rudimentary type of "jumping". Program control will essentially just be an audio signal coming in off the tape. We'll use FSK to determine between three different signals: data bits, a data clock, and a global clock. Each data bit will get shifted into a shift register on each tick of the data clock. Then, at the end of an instruction word, the global clock will hit that will essentially be the same as me hitting the clock button on the remote control. It's going to be quite obtuse to write programs for it, but with a long enough tape, you can do some really, really tricky stuff with the machine. And, since it's just an audio signal, this opens up the possibility for the audio source to be anything, meaning I can load up really, really long "tapes" in MP3 form on my phone and play them back. I can then experiment with clock speed, or even tricky operations to hunt down bugs or bad designs. Most importantly though, it means that instead of needing 1,000 tubes for program control, I only need about 100 tubes.

[@MrRlnansel]

@@UsagiElectric I understand what you are saying regarding complexity. Perhaps a useful perspective to consider, though, is that back in the day bit-serial computers routinely had fewer than 500 vacuum tubes, including magnetic drum R/W circuitry; that was almost entirely because those designs put most of their complexity into the logical machine they implemented. Indeed, an exemplar of such a machine was the LGP-30. From the Wikipedia article about the LGP-30: "The single address instruction set had only 16 commands. Magnetic drum memory held the main memory, and the central processing unit (CPU) processor registers, timing information, and the master bit clock, each on a dedicated track. The number of vacuum tubes was minimized by using solid-state diode logic, a bit-serial architecture and multiple use of each of the 15 flip-flops. ... It was a binary, 31-bit word computer with a 4096-word drum memory. ... The computer contained 113 electronic tubes and 1450 diodes. The tubes were mounted on 34 etched circuit pluggable cards which also contain associated components. The 34 cards were of only 12 different types. Card-extenders were available to permit dynamic testing of all machine functions. 680 of the 1450 diodes were mounted on one pluggable logic board." You have to admit that implementing a 31-bit computer with just 15 vacuum tube flip-flops is pretty amazing. The way the magnetic drum worked together with the rest of the machine was entirely responsible for this.

[@MrRlnansel]

Another way to think of this is to realise that the CPU you have implemented is at a really, really, REALLY low level of abstraction, much lower than most any CPU, past or present. It's much closer to the logic microcontrollers use except that THOSE machines are used to implement the programmer's notion of the machine using microcode, and the programmer never sees any that. (I have programmed in microcode, back in my misspent youth, and I'd rather drive a needle in my eye than do more of it.) Your CPU already has almost all the hardware it needs to implement a traditional higher-level Von Neumann register machine, and believe it or not to actually do so will very likely save you a lot of time and hardware. I would guess you don't really need anything like as many additional flip-flops as what you already have in mind to build. You've already built 16 flip-flops for two bytes worth. I'd be surprised if you couldn't do everything you need to do to turn this into a full-featured general purpose bit-serial computer with just another 16 flip-flops. The rest of your tubes, surely fewer than 32, would be allocated to R/W circuitry. At the very modest speeds this sort of machine is capable of, digital select and/or MUX/DEMUX circuitry can be made using relays.

[@kaitlyn__L]

@@MrRlnansel that’s really interesting, avoiding having tube memory at all, relying on magnetic memory, and building the compute units and interpretation/decoding circuitry. Basically exploiting the analogue domain of dimension (length) to reduce the need for binary electronics.

[@MrRlnansel]

@@kaitlyn__L If any one thing could be said to be the secret of bit-serial drum computers, it was that high-speed electronic flip-flops were only used where there was really no other choice. A CPU has to have at least SOME fast state memory, but while it takes just two vacuum tubes to implement a single S/R flip-flop (one being a dual triode and the other being a buffer), most flip-flops require rather more than that; four to eight tubes were the usual cost per flip-flop. Consequently the designers of these machines used the fewest number of flip-flops they reasonably could and worked all of them HARD. For instance, anytime a flip-flop was used it meant that both true and inverted values of the bit were available to drive the combinational logic. This meant that diode logic AND and OR gates could create most logic functions required without the need for separate logic inverters on either inputs or outputs as long as the buffers on the flip-flops could drive all the gate inputs connected. CPU state information that didn't require flip-flops was stored in the serial data streams of tracks on the magnetic drum (or ultrasonic delay lines). Each data stream, whether a minor loop or a major loop required tubes to make the pulse shaping write circuit and a few more tubes to make the pulse regenerating read circuit. No matter the length of the data stream, it required just a handful of tubes to handle. It didn't matter whether the data stream was 16 or 60 or 1024 bits long, the same number of tubes were required. Minor loops also had to have two or three flip-flops extra to allow access to more than just two word bits at a time plus carry/borrow from the previous bit operation, but also to allow logical and arithmetic shift left and right functions. (The operation of right bit shift functions can be a bit tricky to visualise at first, but it's all a matter of knowing when to tap into the bit stream, and the extra flip-flops provide those tap points.) A 30-bit accumulator implemented this way would usually have only 27 or 28 of those bits on drum at any given time, with the other two or three flowing through the track's associated flip-flops.

[@marknesselhaus4376]

That is looking real fine. I love the looks of the construction and wish my Light Logic LLTP ( 4 bits) was as cool looking. I really need to do more and post to Hackaday some updates.

[@Jensen567]

So I just want to say, I love the kitties, but I also miss the rabbit. Otherwise fantastic as always.

[@willjackson1955]

Hey Usagi! What about a little relay-controlled initializer board for the system? In that way you don't have to fiddle with the remote control each time the system is powered up and, as a plus side, having lots of see-through relays click clack around is really cool!

[@senilyDeluxe]

Look Mum No Computer could build that out of old telephone equipment. He could even build an automatic bootloader injector.

[@PixelBrushArt]

Nice to see more progress on this! Makes me motivated to look into my DTL Computer again! But who knows 😅

[@rickhole]

Jumper wires for digital signals, you are right, build the harness! Great stuff. I wonder if a matrix of silicon diodes could serve as ROM to automate the cold-start instructions.

[@absalomdraconis]

Diodes would be a ROM variant, but for density there's actually a better option: a PCB in the HP 9100 calculators provided ROM via inductance. Diode ROM should ideally be limited to configuration stuff that might be changed over time.

[@CrimFerret]

As cool as converting a reel to reel for program storage would be, wouldn't a paper tape reader be more inline with the tech this is being built with? The punch and reader probably wouldn't be that hard to build from scratch.

[@Skibidi_Male_X]

I've learned so much from this series. Very cool content .Thank you sir

[@JurassicJenkins]

Love watching these creations. He not only knows what he’s talking about, he Owns it 😊

[@ropersonline]

I might have said this before, but strictly speaking Motorola's chip is called the 14500B or MC14500B, not just 14500. I think the UE14500 name is fine, but I think if nothing else, stating the name with the B for the Motorola 14500B avoids confusion with the 14500 type of battery.

[@paulawillaminachandler-ren3725]

You really should consider using a 48 to 60 vdc supply. The 24V supply might be fowling up your results. Looking good so far though, I can't wait for the teletype peripheral part of this video series, ha ha ha. Nice workspace btw.

[@nonoyorbusness]

Tommy Flowers would be proud of you!

[@avientapiedras2961]

They did marvelous things with two bytes.

[@laurdy]

I wonder if Neon bulbs would make good memory? Connect 60v DC through a resistor to one electrode and the other through a resistor to ground, then connect a coupling capacitor to each electrode. supplying a 60v +ve pulse through the top capacitor will turn the bulb on and a pulse through the bottom capacitor will turn the bulb off.

[@UsagiElectric]

Neons are something I actually considered, because they can indeed make a good memory cell, in multiple different ways (I love Neons). But, for this entire build, I'm restricting my voltage levels to +24V and -12V only. Which is incredibly low for tubes, but also it means that there's no danger on it. So anyone can get handsy with it without and fear or danger of shocking of themselves. But, that also means I don't have enough potential to use Neons.

[@lindoran]

Very excellent! I bet you are dealing with some parasitic signals with regards to the signals changing when you manually jumper the grid / row select lines.

[@user-fb9os7hy2y]

You join us as a nice man reinvents wheel.. We're just approaching the singularity (The point where the ratio of spokes built to kittens in basket exceeds parity) 😀

[@frankowalker4662]

Every time I see this computer, it always brings a big smile to my face. I love it.

[@Ice_Karma]

The end kitties are *_gorgeous!!_* 😻😻

[@BadAndUgly]

Will you be building a discrete transistor logic computer later with LED's showing each gate working, it would be a sight to behold!

[@rauldempaire5330]

It sure looks like a "Flash Gordon" movie!.....Very interesting project!...

[@skfalpink123]

Wow! That is a LOT of valves!

[@МатвейКоротков-х8щ]

Нигуя зебе! Радиоламповый компьютер на 2 килобайта. (Wow vacuum tube computer.) this is amazing.

[@josecarlosalvarezalonso3020]

That is a lot of job sir!. Good to whatch real content creators!

[@landspide]

In Australia we say Tea Ack. :)

[@stefanfennek]

So cool.... Maybe it's time to create a punch hole card reader to automate some tasks?

[@Clancydaenlightened]

1:17 they also used transistors to do logic functions that where easier implement than with a tube, like ram

[@Clancydaenlightened]

and output buffers

[@Clancydaenlightened]

look at the 1946 eniac specs and get real tired of buying diodes and tubes, and 3 phase power, make u think which one has more processing power...

[@garyives1218]

.....but still will never be as cool as a Telecaster with Blues Jr. amp :)

[@lohphat]

TEAC. Did you just say “teak”? Son, I am disappoint.

[@Kikay0n]

We always said Teac as Tee - ack

[@ME-ty8rv]

Thank you so much for explaining all this ... amazing....

[@svenvandevelde1]

Your video is used to teach young teenagers what a computer IS.

[@MLampner]

Love the machine and you have very cute cats!

[@gavincurtis]

Now you need to add battery backup for the RTC feature.

[@absalomdraconis]

Better is a magnetic core memory: Wayne Holder has a page about one, and even found some of the cores on ebay.

[@adambarber100]

You should build an arduino test harness to automate board QA.

[@TechBench]

Does the horizontal mounting not create an amplified heating problem as the heat rises through the various rows of tubes? i.e. does the top row not face a river of upward flowing heat?

[@the-retro-center-espanol]

wow, i love this!!, i have a some termoionic tubes on audio amplifiers.

[@brendanvogele2531]

Not to be a peanut gallery comment, but I do think core memory would have been better. ferrite cores and magnetic wire are pretty easy to find still with the cores available on eBay. I know one has to create refresh circuitry, But I think overall it scales better. just my two peanuts.

[@absalomdraconis]

Tape scales better than core. At any rate, maybe he'll use core as programmable configuration storage, which honestly I think it's more appropriate for than ordinary registers.

[@wktodd]

How about a simple cone shaped aluminium reflector slipped over the vfds to make them easier to see.

[@allobject]

Anazing... looking to get your dsiplay gadget of one bit to a single tube - 6BN11?... and using it for fun in my Espruino world.

[@FrankMakesStuff]

Noob question - is the sideways orientation of the tubes alright? Can vacuum tubes be oriented in any direction, or should they be upright?

[@absalomdraconis]

They can usually be oriented any direction, but they _can_ be manufactured in such a way that they sag and tangle- given the need to ship, such tubes would have been rejected from designs fairly early after it's introduction to, leaving them rarer than other tubes.

[@edgeeffect]

The pair in my Korg Electribe MX lay on their side.

[@ReviewUSA-ri5dv]

To think, this thing would have to be 1000 times bigger just to get 2 KILObytes of RAM.

[@loginregional]

Teak? TEAK? *TEAK?* ahem. Tea Ack.

[@UsagiElectric]

Whoops! You grow up pronouncing something wrong and it sure is hard to rewire your brain after!

[@loginregional]

@@UsagiElectric Not to worry. Dad used to call it bassss, not BASE. Love your stuff, especially those mnemonic memory circuits made using BEAR SKINS and STONE KNIVES.

[@soniclab-cnc]

you need a little cnc for making your backing boards.... interpolate everything!

[@edgeeffect]

I don't think "little" is quite the word.... "you need a stonking great huge CNC for making your backing boards" seems more appropriate.

[@soniclab-cnc]

@@edgeeffect lol... I have stonking great huge machines, as you say, in my shop... My smallest machine has a cutting area of 24"X49" and could still cut those parts. I am thinking more like the Shapoko or Xcarve. My 60"x120" AvidCNC machines are massive compared to those so my perspective my be different.

[@jacobjones1622]

Wait, did you start your memory on a soft start circuit you didn't even test?

[@FinnBojorgensen]

Lovely project ! Totally useless for any real purpose except for the satisfaction and the learning but lovely nevertheless. Brings me back to my teens in the late sixties when I learned programming on a germanium transistor based computer. That was leading edge technology back then 🙂. I still have some paper tapes from back then.

[@magnusmalmborn8665]

Can you build some core memory too?

[@mohinderkaur6671]

A margining control on your power supply may help you weed out marginals flip flops and logic gates. reduced heater voltage reduces tube transconductance.

[@ziotdev5250]

Wooooo GREAAAATTT

[@mariomarcelosiad3112]

amazing !!!!!

[@ManVersusWilderness]

Is it just me, or whenever he says what a needle reads on his equipment for volts, etc... it's wildly different from what I am seeing in video. At 6:57 in this video he says the psu ammeter jumps to 11A, but I am seeing 16A. I know it shouldn't but it's been bugging me throughout this series. Love this series though.

[@TheEmbeddedHobbyist]

Just goes to show why the invented magnetic core storage, had 1k of core storage once but for the life of me I can't remember what happened to it :-(

[@Astinsan]

Awesome..

[@larryroyovitz7829]

Maybe regional, but I've only heard TEAC pronounced "tee-ack"?

[@davidjohnston4240]

Will it run Crysis?

[@leninalopez2912]

Awesome! Can I Telnet to it? (?)

[@danaeckel5523]

Telnet unsecure, need SSH!

[@Simcore999]

👍

[@williamsquires3070]

I want to know how many KILOWATTS of power this thing is going to consume (including the cooling, lest you burn down your lab there)?! Just the one section pictured in the thumbnail image has more tubes than many 1950’s era tube TVs, for Heaven’s sake, and those often ran 350-400 Watts or so. Four of those are going to consume as much as a large microwave oven! 🤔

[@edgeeffect]

He's running them at a very low voltage.

[@eDoc2020]

I remember around 400 watts, which was probably for just the CPU. The entire unit probably _will_ be as much as a large microwave oven. Btw it's only color TVs that use that much power. BW TVs usually are under 200.

[@rolfjohansen5376]

do this guy got high electricity bills ?

[@marvintpandroid2213]

No need for a room heater.

[@Unfinished80]

Nice! I love this computer. This thing is going to be a space heater when you're done! Thanks for the great content

[@mohinderkaur6671]

maybe great to dry clothes on it

[@Ahmedhkad]

This is good in winter times

[@larryroyovitz7829]

@@Ahmedhkad Needed more in the north rather than where he is!

[@dakinlowe36]

While I honestly can't follow half of what you're saying, it doesn't matter as I'm just mesmerised and in awe. Your passion is clearly displayed and infectious. Keep up the great content, as soon as I finish work on Monday's I immediately check for the latest video, the project doesn't matter as they're all fantastic. When you first spun up the hawk drive I just had to tell someone how cool and ridiculously noisy it is and mentioned it to my mother only to be told she used to work in a mainframe environment in the 70's and remembered drives like the hawk and magnetic tape reels. A whole section of her life I'd have never found out about if it wasn't for you. Thanks.

[@ostsan8598]

Glad to see the UsagiVAC project is still moving along.

[@edgeeffect]

That's quite a good name... it's got a good cadence to it AND the most important thing with vacuum tube computers is that their name ends in "...AC"

[@youfrancis]

That's really awesome. Tube and electromechanical technology has always fascinated me. I genuinely feel that: when you first learn electronics, that should never be left out or forgotten. Because the fundamental concepts originated there.

[@douro20]

I would try delay line memory. I know it would complicate the design a bit but it would take up much less room than a board full of tubes and it would allow for a good deal more storage capacity.

[@senilyDeluxe]

A Tape Echo! It's not like this computer will ever run faster than a few kHz anyway. And some people will probably chuckle at seeing ancient recording studio equipment used for memory.

[@alexquant1335]

Amazing, as usual! :) I'm intrigued at your intention to use a reel-to-reel tape machine as program storage - will the tape have to move back and forth over the head for random access? It would be fun if the Centurion could be utilised to create code for the VTC, or used as a terminal and storage to the VTC?

[@UsagiElectric]

Thank you! I'll go into a lot more heavy detail once we start the build, but the problem with program storage is that it's essentially memory, and memory is so, so hard. But, the MC14500 architecture has a way around that. Since the chip was originally designed to replace relay ladder logic in industrial applications, the chip had to be able to handle any kind of memory or program storage type that might get thrown at it. This is where IEN and OEN come into play. It's not entirely graceful and it makes programs considerably longer, but you can do conditional skipping using OEN quite easily. Then, you can either construct your program around a loop (ie. looped tape), or you can create multiple copies of routines on the tape. So the tape will always be trundling along at a constant speed and instructions will be 100% sequential, but we can do conditionals relatively easily, which will allow us some rudimentary type of "jumping". It requires a bit of a shift in thinking, and it's not going to be particularly powerful, but at the end of the day, it is a 1-bit computer, so our expectations weren't super high, haha. We are working on a UE14500 emulator for the Centurion, and since the program control will essentially just be an audio signal (we'll use FSK to determine sequential data bits, a data clock into a shift register, and a global clock that executes the instruction from the shift register), we can pretty easily build an extremely simple audio card for the Centurion and write a custom CPL program that dumps a specific audio pattern based on a latched in value and the Centurion could be running long programs through the tube computer. Then it gets even more fun because after program control, we'll start on input and output and the goal there is to bit bang 5-bit baudot or 7-bit ASCII, which could make it back into the Centurion as an input!

[@parttimelaowai1771]

Someone please calculate how many tubes it would take to simulate a modern Intel CPU, and how many TW's of power it would consume.

[@Phil-D83]

Several km across and power of a small city

[@edgeeffect]

Pick an expletive, append "...load" to it. It's that many!

[@Raz82000]

Same size of eiffel tower if you are going to build it vertically.

[@johnpassaniti4417]

Love this series... and the kitties.

[@darkwinter6028]

Ya know, one of these days you’re gonna have to build an interface card so you can cross-connect this with the Centurion…

[@mwwhited]

I recently built myself a byte of memory using DTL(diode transistor logic) and ran into an issue with stability. I haven’t fixed it yet but it mostly seems to be caused by manually toggling address lines. I have slightly better stability when they are electronically addressed. I think I need to add some zeners or comparators to act like a Schmitt trigger... you probably have something similar going on.

[@OZtwo]

Yes! It is back! Not saying I didn't love all your other videos but still. :)

[@BrainboxccGames]

Very cool and always so interesting! Keep up the good work, i can't imagine how much space even 1kb of ram would take up, and how much heat it would give off!

[@cairsahrstjoseph996]

I can't imagine how you are going to get a program from the tape drive to this computer. But it is interesting to see how a computer works right down to the circuit!

[@bbqgiraffe3766]

Ah so you're the one that's been causing the power outages in our godforsaken state

[@UsagiElectric]

Yup, but for a brief moment, I'll have the warmest room in the state!

[@bryede]

Quick note: TEAC is pronounced TEE-ack.

[@UsagiElectric]

You're absolutely right! And being a student of Japanese, I absolutely should have known better, but old habits of mispronunciation sure are hard to break.

[@NATHANJK3]

Bro awesome work, amazing...im folling you and im not a expert, just a watcher but your work is so inspiring. I live the computers since 1991 : IBM machine. BEfore ATARi and video games. Now days it is just crazy and changing so fast. This channles is so important.

[@AppaTalks]

Great video! Great content! Love your channel Usagi Electric.

[@shinedom]

Finally some real progress with the vt computer. Very interesting and learnjng

[@brianhginc.2140]

@5:45 I've done this before. The correct way to do precision drilling is to NOT drill through the paper. Get a point tip or tiny Phillips tip screw driver or large nail and hammer. Hammer through the paper a point/dent into the wood for every drill hole. Remove the paper. Now you will have dents for every hole which your drill bit will sink into making each drilled hole perfectly dead center every time. Another solution is to do this through the PCB for the best accuracy. Also, if you want to stick with the paper print reference, some of the really-really old Epson wide carriage color paint-jets which can do 11x17 in color can actually do 22inx17in (56cmx43cm) in B&W. Your precision will vastly increase if you can find one on ebay dirt cheap.

[@senilyDeluxe]

I tried the denting method on my second alarm clock. It looked worse than the first. I totally suck at fine motor control.

[@Funkybulb]

I cant imagine how large a tube version of a 486 cpu and memory

[@jaredwright5917]

I would recommend looking at acoustic delay line memory, specifically using nickel and steel wire. They were used in some early electronic calculators as their main memory. It's been a while since I did any research on the topic, but they seemed pretty simple to make.

[@absalomdraconis]

I really want to recommend that he reproduce a drum memory based on old wire-recorder technology, but I'm only seeing non-magnetic 316L, so your version is probably better.

[@absalomdraconis]

Never mind, McMaster-Carr.

[@jaredwright5917]

Another option would be to try using tape from an old VHS or audio cassette wrapped around a cylinder. Audio cassette tape heads are also cheap and readily available online.

[@kennethrhyne9581]

i still felt like a new comer before i hear your intro. love the channel

[@someusername1]

Fantastic work! This is incredibly impressive and hugely educational.

[@goatman7362]

Maybe you can use that really beat up terminal that came with the centurion to use this computer when it’s done!

[@hicknopunk]

Bytes!? Not "words" of memory 😅

[@your_utube]

Retro - Retro or Retro Squared is what this computer is all about. Cannot wait to see how it will perform when the tape storage comes in.

[@ElianWailmer0323]

Great video. I really like this computer and it's awesome to have an update on it; keep up the great work ᐠ₍⁽˚⑅̆˚⁾₎ᐟ

[@wdavem]

Very impressive, so much fun to watch! I built part of a relay computer out of 120v elevator relays 20 years ago. I got nowhere near as far as you have with yours. But I did actually have working relay-diode logic with 40 bits of ram and a weird counter/shifter register unit that could count in base 10 and convert to binary, turn itself into a looping shift register and output or input serial data, base 10 (didn't quite get to the keyboard, I want to get back to it someday). That was hard but fun!!

[@blitzroehre1807]

The colossus computer developers would have been mighty proud of you....👍

[@LeonirZimmermann]

Why do you use 24 volts in anodes, instead of 90 volts for example? (forgive me if you already explained the reason and I didn't understan)

[@H2Oredfirefox]

I know it's a bet Old fashioned about interesting to be able to play an old game of pong on that of course you would have to create an output display it doesn't have to be a display you would normally expect I wondered what you could use is it old fashioned display from that era

[@AJB2K3]

Random question but is it possible to make an LED flashing circuit using 6AU6 Tubes?

[@StevenIngram]

I have to say that ALU you added was a very interesting twist. I mean, if you ever think about miniaturization via FPGA, you could hypothetically build an 8-bit computer that could operate on 8-bits in parallel. No idea what such a thing would be good for. Just an errant thought that occurred to me watching this series. :D