Hmmm... sampler maybe, but I doubt ADC was ever done vacuum state. Simply no need for it, it's faster to plot on an oscilloscope or X-Y plotter and make a photograph of it. They didn't have the widely available data processing, let alone DSP, we have today.
When did Tektronix release their first sampling plugin? I'm guessing it was already transistor...
As for inventing one from scratch, tube S&H would probably be best made with a diode gate (as far as switching signals, Tek's used this since the beginning, when they created the "dual trace" oscilloscope). Triodes don't make as good switches as JFETs, and certainly aren't bidirectional. You still have to contend with leakage, which is astronomical (low uA) compared to SS (<pA easily). That will fundamentally limit the time a signal remains valid for a given ratio of sampling vs. hold time (i.e., longer sampling time = more time to charge a bigger capacitor = more time the voltage remains within tolerance).
As for making an ADC, you could build a comparator + SAR with not too many tubes. Tubes are fine at storing logic states, if not particularly fast. The comparator will be plenty noisy and drifty, when you get to the lower bits (~10mV), so even with compensation, you'll have a limit there. Fortunately you can amplify the signal to accommodate a more tube style voltage (100V instead of 5V?), but that still only gains you a few bits accuracy. One bit sigma delta would be easy, and not too bad on accuracy due to the integrator.
And of course the complement, a tube DAC, would be kind of silly. You're better off stacking calibrated xxx meg resistors from the logic lines to the output, using voltage division to brute force the conversion. With 200V+ to burn, this isn't a big deal. You could still switch currents with diode gates, but you need to generate accurate currents, too.
Tim
When did Tektronix release their first sampling plugin? I'm guessing it was already transistor...
As for inventing one from scratch, tube S&H would probably be best made with a diode gate (as far as switching signals, Tek's used this since the beginning, when they created the "dual trace" oscilloscope). Triodes don't make as good switches as JFETs, and certainly aren't bidirectional. You still have to contend with leakage, which is astronomical (low uA) compared to SS (<pA easily). That will fundamentally limit the time a signal remains valid for a given ratio of sampling vs. hold time (i.e., longer sampling time = more time to charge a bigger capacitor = more time the voltage remains within tolerance).
As for making an ADC, you could build a comparator + SAR with not too many tubes. Tubes are fine at storing logic states, if not particularly fast. The comparator will be plenty noisy and drifty, when you get to the lower bits (~10mV), so even with compensation, you'll have a limit there. Fortunately you can amplify the signal to accommodate a more tube style voltage (100V instead of 5V?), but that still only gains you a few bits accuracy. One bit sigma delta would be easy, and not too bad on accuracy due to the integrator.
And of course the complement, a tube DAC, would be kind of silly. You're better off stacking calibrated xxx meg resistors from the logic lines to the output, using voltage division to brute force the conversion. With 200V+ to burn, this isn't a big deal. You could still switch currents with diode gates, but you need to generate accurate currents, too.
Tim
Sample & Hold
Thanks for reply. I know all that, but it's just for my own culture, and if I get schematics they will be the base for an experiment I want to realize with an electrometer tube CK5886 for a switched integrator. I've tested with special IC's and homemade circuits using JFet and MosFet but problem is speed (always staying in an amateur environment where it's not easy to get high range components).
Thanks for reply. I know all that, but it's just for my own culture, and if I get schematics they will be the base for an experiment I want to realize with an electrometer tube CK5886 for a switched integrator. I've tested with special IC's and homemade circuits using JFet and MosFet but problem is speed (always staying in an amateur environment where it's not easy to get high range components).
5886, "electrometer pentode". Well I'll be, 10^-15 -- that's 1fA nominal grid current! Haha, and at really low voltages too, like 4.5Vg2, cool. Microamps plate current, of course.
It's my understanding that FET op-amps that good have only recently hit the market. But they have much higher transconductance (or in the case of op-amps, voltage gain). In fact, the insulation quality of even a regular 2N7000 is pretty impressive, just don't get fingerprints on it.
You will certainly not get any speed improvement using tubes.
Tim
It's my understanding that FET op-amps that good have only recently hit the market. But they have much higher transconductance (or in the case of op-amps, voltage gain). In fact, the insulation quality of even a regular 2N7000 is pretty impressive, just don't get fingerprints on it.
You will certainly not get any speed improvement using tubes.
Tim
Amazingly, it was, and it was audio. This month's AES journal has a fascinating article about the SIGSALY system that allowed Churchill and Roosevelt to talk securely during WWII. The digital encryption techniques were startlingly modern. Unsurprisingly, the terminal equipment was rather big (forty bays of equipment).
Mike's Electric Stuff has a bit about Vacuum state ADC:
Glass Analogue to Digital converter
It is well worth exploring the site - lots of interestnig and fun stuff to do with, erm, electricity. I want a Mercury Arc Rectifier for my MIG welder...
Thanks for the pointer to the AES article - I'll have to try and find that one.
Glass Analogue to Digital converter
It is well worth exploring the site - lots of interestnig and fun stuff to do with, erm, electricity. I want a Mercury Arc Rectifier for my MIG welder...
Thanks for the pointer to the AES article - I'll have to try and find that one.
I went to the EIPBN ("3-beams") conference some 5-6 years ago. Those guys deal with just about everything on a miniature scale that has photon/ion/electron beams in it. Someone was presenting a MEMS-based ADC based on exactly that principle. Pretty neat stuff.
~Tom
Actually, hollow state ADCs did exist. In 1955, Epsco, Inc. introduced its Datrac B-611 hollow state ADC which offered 11 bit resolution at a 44KHz conversion rate, which is still pretty impressive. The whole unit weighed in at some 150 lbs, at a price of $8500. The performance is still impressive, but definitely pricey. ADCs were used in the space program, and for the first radar signature analysis techniques.
Hollow state ADCs were based on a VT derived from electrostatic deflection CRTs. In the place of a screen, you had two plates: one solid, and another one with perforations in front of it. The pattern of perforations determined how the analog signal was digitally encoded. Horizontal scanning determined the conversion rate, while vertical deflection would put the electron beam at the proper place to digitally convert the corresponding voltage level.
By 1959, discreet BJTs were being used to implement ADCs with the corresponding reduction in size, weight, and cost.
I don't get it - the most straightforward ADC to implement is (brute force type) flash converter. No sample-and-hold circuit, just vast number of active elements thrown against the problem. The number of comparators needed goes up exponentially with desired resolution (number of bits per sample). Low resolutions aren't exactly cheap, but not impossible either (computers running tens of thousands of tubes existed and operated long time ago). If OP really wants a tube version of ADC, a truckload of russian dual triodes would be an excellent starting point.
S&H ? If desired resolution is too high, it mighjt be needed anyway. What is S&H anyway ? It it a tiny sampling capacitor, followed by a buffer capable of supplying/sinking large current, feeding much larger storing capacitor, so voltage reading can be stored long enough to perform two (or more) consecutive operations on it without losing more than half of a bit worth of information in due process. Again, nothing that couldn't be implemented with tubes.
Obviously the starting point would be to set the limitings factors, that is the desired resolution, sampling rate and size of conversion stage (number of active elements). Once constraints are set - and sane - a converter could be designed. There are faaaaaar too many solutions possible until these constraints are set.
S&H ? If desired resolution is too high, it mighjt be needed anyway. What is S&H anyway ? It it a tiny sampling capacitor, followed by a buffer capable of supplying/sinking large current, feeding much larger storing capacitor, so voltage reading can be stored long enough to perform two (or more) consecutive operations on it without losing more than half of a bit worth of information in due process. Again, nothing that couldn't be implemented with tubes.
Obviously the starting point would be to set the limitings factors, that is the desired resolution, sampling rate and size of conversion stage (number of active elements). Once constraints are set - and sane - a converter could be designed. There are faaaaaar too many solutions possible until these constraints are set.
Ahhhh yes, the ADC tubes! I forgot about those.
Point the beam at the left hand side (S&H holds for a scan, which can go as fast as you'd like), keep vertical constant, scan horizontal at your "clock rate". If you got a fancy model, I suppose it could generate start and stop bits automatically, so you could literally buzz this thing into a modern RS-232 port and read data live with about four tubes total.
God, did they ever cheat, back in the day.
Tim
Point the beam at the left hand side (S&H holds for a scan, which can go as fast as you'd like), keep vertical constant, scan horizontal at your "clock rate". If you got a fancy model, I suppose it could generate start and stop bits automatically, so you could literally buzz this thing into a modern RS-232 port and read data live with about four tubes total.
God, did they ever cheat, back in the day.
Tim
*Diassapears into workshop carrying large quantities of glass, steel and scarce metals*
*Emerges a month later with a laptop interfaced to an analog signal, only using glassware*
*Cackles*
God, did they ever cheat, back in the day.
Oh, yes! "How can we use less tubes?" "Hey, how about we make some of the most complex glassware known to man?" "Yeah, that'd do it!"
The thing is, that given how much power multiple valves use for the heaters, it does make some sense to use fewer, more complex valves rather than a boatload of simpe ones.
That, and the fact that I can't image that valve manufacturing scales anywhere near as well as transistors, meaning that large quantities of simple valve would be extremely expensive.
I like it - it's a nicer way of doing it than brute force with millions of transistors, IMHO. Maybe less accurate, and efficient, but more elegant.
The ONLY thing transistors have going for them is printability. Well, they still take the cake in terms of fundamental properties like current density, transconductance and Q factor (i.e., transconductance vs. capacitance, ultimately limiting bandwidth). But the main limitations are heater current and physical size.
It would be interesting what might happen if transistors hadn't been invented. Would tubes be manufactured photolithographically? Could they print, say, a very fine tungsten wire between electrodes, deposit a layer of rare earth for emission, then print insulators, grids, plates and metallization on top of that? The whole thing still has to be sealed in vacuum, but that's not a terrible downside; semiconductors are sealed in glass today (glass passivation), only a small envelope would be required to contain the extra vacuum. Manufacture would probably be similar to nuvistors, but more automated and less tedious.
Of course, the ultimate devaluation of our beloved tubes would be encapsulation in the glassophile's natural enemy, the black plastic TO-247 package.
As bad as TO-247s look, I must say I would love to have a 300B an inch across that can dissipate 300W.
Tim
It would be interesting what might happen if transistors hadn't been invented. Would tubes be manufactured photolithographically? Could they print, say, a very fine tungsten wire between electrodes, deposit a layer of rare earth for emission, then print insulators, grids, plates and metallization on top of that? The whole thing still has to be sealed in vacuum, but that's not a terrible downside; semiconductors are sealed in glass today (glass passivation), only a small envelope would be required to contain the extra vacuum. Manufacture would probably be similar to nuvistors, but more automated and less tedious.
Of course, the ultimate devaluation of our beloved tubes would be encapsulation in the glassophile's natural enemy, the black plastic TO-247 package.
As bad as TO-247s look, I must say I would love to have a 300B an inch across that can dissipate 300W.
Tim
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.