Let’s Build a Vacuum Tube D Flip Flop

[@MatsEngstrom]

Nice. You're slowly building up more and more complex logic blocks. Soon you'll be building a 8-bit counter and have used up your entire stock of tubes and probably a shedload of additional ones. ;-) I'm guessing that a ring-counter would be a logical (pun intended) next step in the evolution of circuity here. But every time I see one of your videos I'm thinking that the grid resistor network might be simplified a bit. I have a hard time seeing what the 4k7 resistor actually is doing there. The current is already limited by the upper and lower resistor in the preceding voltage divider. If the gate have a very high input impedance the 4k7 is totally superfluous. If the gate have a low impedance (I guess not) then you could just jack up the divider resistors a notch. The only reason I could guess for the 4k7 doing something useful is if the gate is presenting as a non-linear load towards the voltage divider and some magic happens.

[@UsagiElectric]

Thank you so much! It’s funny you mention a counter, because that’s something I’m actively trying to suss out at the moment. A ring counter would definitely make sense, but I can’t seem to get one working reliably with tubes because the obvious answer is to use neon bulbs to make a ring counter. Unfortunately, neon bulbs do really do a whole lot at 24V, haha. I do have some plans for other counters to try to build, but like you said, it’s going to take a boat load of tubes! You know, I never really thought about the little 4k7 resistor, it was one of those things that I was using while I was experimenting with a whole bunch of different resistors of all values for the voltage divider going into it. Once I was happy with the 22k/33k combo I use now, I never gave the 4k7 another thought. But, you're absolutely right that the 4k7 could definitely be superfluous. So, I figured I would do some testing! I set up a 6AU6 as an inverting amplifier like always and then ran a potentiometer into the 22k input, changing the voltage from 6V to 20V in 1V increments. I took a measurement on both the grid and plate/output to see if there was any appreciable difference between the two. Here’s the results: i.postimg.cc/Y9s9qbPk/Data.jpg At first, it looks like there’s nothing in it, but where it gets interesting is what the voltage measured directly at the grid is as the input voltage increases. Once the voltage at the grid passes about -0.5V, it starts to flatten out, despite increasing voltage into the divider input. With an input of 10V, the grid voltage measures just about 0V and barely increases from there. That’s not necessarily a revelation, as I’ve seen that in other tests I’ve done in the past. Also, I’m not entirely sure why that’s the case. I haven’t done too much reading into it, but it definitely is curious. However, the main thing between the 4k7 and NoResistor setups is there is an actual deviation from an input of about 10V on up. It seems the tube hits full saturation (output at 4.9V) with an input of 15V as opposed to an input of 17V with the 4k7 resistor. So, the 4k7 is definitely doing something, although it’s probably so minimal it doesn’t really make a meaningful impact on the circuits I’m building. I’ll have to do a bit more reading on this and see if I can find a solid answer for what exactly is going on on the grid!

[@MatsEngstrom]

@@UsagiElectric Yea, neons are out of the question at 24 volts ;-) When I do ring counters I do it with a npn/pnp combo that makes up a pseudo-scr as the active element and on top of that a diode or led up to the supply line that is having short interruptions in it to force a step in the counter. II guess something similar could be made with thyratrons. The 4k7 seems to actually do something a little bit, which is kinda curious since the grid should have a high impedance (considering I read everywhere that a tube amplifiers input impedance is basically only affected bu the value of the grid resistor that follows after the input dc-blocking cap). Maybe the tubes operate in some strange mode at this low plate voltage.and actually is pulling some current into the grid?Could you measure the voltage drop over the 4k7 at some input voltage and see if there's any? Even a 1uA current should give 5mv that should be visible on any dmm. Does your tube-based meter have any real low voltage settings?

[@UsagiElectric]

@@MatsEngstrom Oooh, I hadn’t thought about trying to build a ring counter out of thyratrons, that’s an interesting idea! I’ll have to look into that more and see if anybody has experimented with them as ring counters. The 4k7 is definitely doing something! I setup the 6AU6 as a standard inverting amplifier with the 33k to -12V and the 22k to a potentiometer input swinging between 5V and 21V. I measured the voltage on both sides of the 4k7 resistor with my VTVM. That Conar meter has a 3V scale which lets me read voltages at 0.05V intervals fairly easily. The results are pretty wild! Data: i.postimg.cc/Jns8wbnY/4k7-Data.jpg So, the voltage measured at the grid itself pretty much plateaus at about 0.2V, but the voltage measured on the resistor network side of the 4k7 continues to rise. I would say that the grid definitely looks like it’s pulling in some electrons, and I have a theory as to why. The divergence happens at about 9V input voltage, and the output of the tube at the input level is about 6.5V, which is right about saturation. The plate is absorbing as many electrons as it can. But, as we crank up the input voltage higher and higher, there’s a stronger and stronger potential flowing into the grid. So, it starts to suck up some electrons. Because the grid is absorbing electrons, the potential measured at the grid is essentially ground, but on the other side of the 4k7 resistor we see where the voltage drop is happening, which appears to be about 1.5V to 2V. My guess is that 24V on the plate just isn’t a strong enough of a potential to provide enough of an attraction to keep the electrons from being pulled into the grid. So, it might actually be a good idea to keep that 4k7 resistor in there. I wonder what the effect on grid longevity will be, considering the grid isn’t supposed to absorb any electrons at all (in theory). At any rate, since I’m using the tubes for weird applications, they rarely run longer than an hour or two per day, which means they’ll probably still last for many years. Very interesting information! Thank you so much for asking about it, I’m having a ton of fun setting up and measuring different things!

[@stphinkle]

@@UsagiElectric There are some dekatron counting tubes. You might see if any of these pop up sometime.

[@UsagiElectric]

@@stphinkle Dekatrons are definitely on my list of must buy tubes! Unfortunately, they're a cold cathode tube, which means you need pretty high voltages for them to start operating, similar to neon bulbs. Still, they're absolutely epic pieces of history and I'd love to build something with them someday!

[@tstahlfsu]

Awesome! Love the continued progress on these gates!

[@UsagiElectric]

Thank you so much!

[@waynethompson8416]

Thinking out loud here...I wonder if you could have done the same thing using 6B7 or 6B8GT tubes? They are Duo-Diode Pentode type tubes. The idea is to replace the diodes you used with the diodes in the tube.

[@UsagiElectric]

That's actually a really good idea, and I actually have a few 6AV6s and 6AT6s, which are dual diode triodes. However, the 6AV/T6 as well as the 6B7/6B8 all seem to have the diodes sharing the cathode of the triode or pentode inside the element, which unfortunately wouldn't work with the NOR design here as I need the cathodes of the diodes to connect up to the control grid ultimately. However, there's actually a way better tube for this, and that is the dual pentode, like the 6HS8. They're set up really weird, but they essentially have a shared cathode and a shared control grid #1, with separate plates and separate control grid #2. So, the share control grid can act like an enable, and the two plates and two separate control grids can act like two separate tubes. This means the tube is essentially a dual NAND gate, which means it can be an SR Flip Flop stuffed into a single tube, which is just awesome!

[@waynethompson8416]

@@UsagiElectric My original thought was to suggest a 6H6 but when I looked it up to confirm what it was, I discovered that it is actually just a Duo-Diode only. But it is a cool little squat tube about 3/4" tall. I am eager to watch your progress as you go through the process of completing a full computer using vacuum tubes. What are you going to use for memory?

[@migry]

A lot of the thin “2 row” side breadboards now have that annoying gap in the middle, usually with no indication that the gap exists. Back in the 80’s when breadboards were manufactured in the good old U S of A, that gap didn’t exist. I note that you have added the two,short links for two of the side breadboards, but there is no link for the bottom one, which I think has the negative bias voltage. So is the bottom breadboard different to the two above?

[@migry]

I’m new to your channel and playing catch-up. Seems to me that you have level sensitive latch and not an edge triggered D-type flip-flop. To make a D-type flip-flop requires two latches, one taking clock and the other clock-bar.

[@ropersonline]

Somehow it bothers me that that push-button isn't between C and E but between D and E. PS: I admit not douible-jamming contacts into a hole in the 0 (actually 1) column would have required an additional wire to move things one over.

[@T3sl4]

Cool! Coincidentally around the same time, I got the itch to do a little blinkenlites with tubes. I decided on a shift register. Schematic: www.seventransistorlabs.com/Images/Tube_Shift_Register.pdf Blinking: imgur.com/gallery/OF4jAxh I also happen to have a box full of pentodes (5702, subminis with wire leads -- very handy for breadboarding!), but figured using logic gate synthesized flip-flops wouldn't scale very well for a register (hard to do it in less than the four you've used). Fortunately, I found a way to chain multivibrators together with diodes and a (fairly high amplitude) pulsed clock, so only needed two tubes (one dual) per bit. Also handy, everything is balanced, so the XOR is pretty easy to synthesize, only needing two tubes. (I did opt for the trick that, depending on gate bias resistor, one triode can perform NOR or NAND between two, maybe three inputs.) One downside to this logic scheme is, the grid divider resistors (in the register) need to be somewhat closely matched (1% or so; I only have 5% on hand so had to bin them!). The neon lamps can misbehave as well; I went through several permutations of bulb and resistors until settling on this one. Note the bulbs and resistors are chosen to give equal voltage drops in the respective 'on' states.

[@UsagiElectric]

Thank you! A shift register is one of those things that I would love to build for a future bit-serial computing project! But, you’re right in that it’s almost not possible to build a D Flip Flop in less than the four tubes (or two dual triodes) that I’ve done here. What’s more, this D Flip Flop isn’t edge triggered, so it won’t work in a shift register at all (it just oscillates like crazy). In order to do that, we need another two tubes (or three total dual triodes). That’s what I get into in the next video on the Toggle Flip Flop. Having said that, your multivibrator design is really brilliant! I’ll be honest, it’s a bit over my head though, haha. I stared at your schematic for quite some time and I’m only starting to understand bits and pieces of it. The interaction between the pulsed clock and the quad diode junction between grids and plates is a wild one that’s really hard to visualize. But, the results speak for themselves, your blinkenlites mad scientist breadboard concoction is amazing! Is there a bigger project this will be used in? I’m curious what kind of applications you have in mind for such a wild circuit!

[@T3sl4]

Right, I once built a T f/f out of transistors, in whatever you'd call it, ECL or CML, something like that (not straight ECL, but a few stacked tiers of current steering). Really beautiful schematic, right pain to build! The key insight for this (tube) circuit was, the grids are really sensitive. Like, I usually toggle a bit's state by touching its grid with a screwdriver -- body charge is enough to pull it off. Or the next easiest thing, pinching the leads of the negative pull-down resistor (thus biasing it off via skin resistance in parallel). I figured, if I can just charge a capacitor from the plates, with little enough current that the plate voltages aren't disturbed during the up swing (which would cause backwards operation), and on the down swing, ferry that charge over to the next grids, pulling down whichever has the lower voltage -- I'll have a shift register. (And you can't just ferry around a lone capacitor anywhere, but you can split it in half and drive the midpoint from a clock -- same idea.) And it happens that two pairs of diodes can do exactly this, with the voltage levels as they are, given a falling edge of more than about 80V. After poking at resistor and capacitor values, matching grid resistors, and adjusting the clock waveform, it started working! Heh, also 1N4148s are very marginal for this, but I checked, and found most of mine avalanche around 120-130V, which is fine for this. So, whatever, it's close enough. :) Along the way, I had some weird behaviors, like 4 or 8-period cycles with weird double transitions (obviously, the register wasn't shifting as such). Possibly the clock needs to fall quickly, or needs to be a low enough impedance that a cascade of transitions doesn't ka-chunk the whole thing. I'm not sure exactly which things were wrong at the time, but just generally cleaning it up and making a nice solid clock seems to be the trick. Also, I started with an externally clocked version (in turn driven by my function generator in one-shot or auto mode), which was giving marginal clock levels, but did seem to be working. The 12AT7 multivibrator however is able to saturate *extremely* well -- the minimum plate voltage is something like 10V -- so it gives a huge signal level, and the waveform is just about perfectly the pulse I wanted. A downside seems to be, the duty cycle is limited, so that I can't go much further on the ratio of timing capacitors; but the values shown do a nice job so that's not a problem. Also, because this is a sort of dynamic logic -- it's able to run up to a few kHz, but somewhere above there, I can see the light intensity and pattern changing, shifting to some degenerate sequence I'm sure. (I don't have enough channels to scope the full register at once.) There's plenty of room left to optimize for speed; obviously(?) with 12AX7s I wasn't going for anything extreme. :)

[@UsagiElectric]

@@T3sl4 Using ECL is an interesting way to build a toggle flip flop! You don’t really see too much on transistors not being driven into saturation or cutoff. It would be interesting to see what kind of toggle speeds you can get out of a discrete transistor ECL flip flop compared to the conventional way of doing it. I know next to nothing about transistors, but it would be fun to see what the limiting factor is! I’ve discovered that tube grids are really sensitive! Even my low voltage stuff here is sensitive enough to toggle just by touching a resistor, like you mentioned. Your circuit is wild and absolutely awesome! I would love to see a full-length video explaining it and showing just how far you can push it (both in terms of functionality and speed). The 12AT7 is an excellent tube, and I’ve actually found that even at 24V it saturates pretty well. I would be interested to see how well the 6DJ8/ECC88/6922 tube does in your multivibrator. I’ve found that I can get that tube to saturate reliably down to about 5V, which is better than nearly any other tube I’ve tested (weirdly, the 6EH7/6EJ7 pentode actually has a nearly identical curve to the 6DJ8). I’m going to have to do a lot more studying of your schematic, there’s a ton to learn in there! Thank you very much for sharing it!

[@nasalimbu3078]

CPU Athena eco life engrgy core biology core GPU ares war core