This circuit uses high voltages, which are dangerous. Don't attempt to build it if you aren't experienced. Use of the information on this page is entirely at your own risk. I will not take any responsibilities for any resulting injury or damage.
After Ken Stone presented a current controlled thyratron VCO, I decided to try thyratrons myself. At this stage I have successfully breadboarded this circuit with the exception of the output stage. The voltage swing at the 1M pot is comparatively low (a couple of volts), thus so far I decoupled it with an opamp voltage follower. This might already be an option for those who want to interface such a circuit to solid state based gear.
I won't go into too much detail about the expo convertor which is pretty straight forward. If you're interested in how these things work, you can find a tutorial at my start page. For the friends of linear VCOs I might say that one can simply replace this by a linear sink going to the inverting input of the current mirror stage in which the expo current is going. The only thing exotic about this mirror is the sink element, which is a triode. Still this part of the circuit works just as if it were a FET. The thyratron is conducting when it sees a positive voltage with respect to its cathode. Initially the 2.2nF and 68nF caps are discharged and the cathode is at 0 volts. The grid is held at about 50V and thus the thyratron fires. It charges the caps rapidly up to nearly 100 volts, the larger fraction is over the 2.2nF, and a smaller one over the 68nF. The grid cathode voltage difference is now negative, and the thyratron goes out of conduction. Now the current sink tube slowly discharges the caps until the voltage across both is less than 50 volts, where upon the thyratron can fire again. The cycle now repeats.
The most notable circuit feature is the capacitive divider, which translates the 50V voltage swing at the 2.2nF cap into a more manageable amplitude, which does not overdrive the output stage. There are two compensation circuits, the one labeled bias trim is to back up for the grid current, which is subtracted from the exponentiator current in the current sink tube. This is caused by electrons accidentially hitting the grid structure, the more negative the grid the smaller the current gets. (Also see the tube expo convertor for circuit that tries to make use of this.) The compensation tries to compensate for the average over the cycle. The network comprised of the 47k trimpot and the 1N4007 diode is a Franco compensation network. The diode bypasses the trimmer on the large inrush of current when the thyratron is fired. The value of the compensation is fairly high. That is due to the deionisation time of the thyratron. It takes a couple milliseconds until the gas in the tube goes out of conduction and the normal ramp down can take place. This deionisation is clearly visible as a small flat portion at the start of the waveform. However the steep portion of the sawtooth is very fast, infact so much that the TL081 opamp I had used as a follower wasn't fast enough to properly follow it, it started to overshoot. Monitoring straight at the cap showed no ringing. Its quite interesting to note that the capacitive divider also does an impedance transformation much like a transformer, so the voltage at the lower cap appears with a low impedance, which could be loaded with a resistor without causing detuning effects.
The dual triode Tubes I used can be replaced by types which source sufficient current at low plate voltages. About 5 milliamperes at 50V plate voltage are desireable for the current sink tube, preferable if there are still -1 to -2V of Vg at this point, to get very little grid conduction. (See the tube expo convertor... There still is a tiny grid current at low negative grid voltages.) A 6N1P, ECC88 or ECC82/12AU7 might be suitable. Paralleling of sections yields more current. This current is defining the maximum current of the current sink, and determines how far up you get in frequency. The russian miniature dual triode 6N16B seems to be designed for low voltage operation, and is well suited for this application.
At present I only made a temporary bench setup, leaving up some of the dimensioning to a later stage where a permanent version of this oscillator will be built. (Well actually three, for my tube synthesizer.) I haven't done extensive measurements on the circuit, but the impression of the performance was pretty good. The tuning curve could be made very good. I could track solid state VCOs pretty well over at least 8 octaves. I have to delay a test of the thermal behaviour until the final circuit is built. Anyway, I don't expect stability wonders. We'll see if thermal insulation will do the trick. There is some problem at lowest (couple Hz) frequencies, the thyratron fires errantly, possibly because of hum. But for normal audio frequencies all is well.