Steve Ward wrote ...

I think the control scheme worked to keep the ramp the same until the DC bus cap "ran out" at which point the RF current suddenly collapsed. With the extra capacitance this wouldn't happen, but I think the streamer had reached its final length before even the small capacitor ran out of charge.

Well, it would be useful to know how much real power the bridge delivers throughout your bang. I recently started messing with my QCW system and found a tuning that actually limited the power of the system quite a bit, where the primary current actually went down with increasing drive voltage! I suppose it might be possible, with current regulation, that your pulse skipper starts jumping more pulses and makes the effective voltage lower, hence the sparks dont grow any longer... maybe? The nature of your control makes it somewhat difficult to know what the real power is, since you have to average all the driving cycles vs recycling cycles to know what the actual power factor is.

Lucky for me and my bus modulation scheme, i can assume my power factor is unity (zero current switching, V and I are in phase) and just look at the volts and amps going to the DRSSTC bus. Of course, im currently investigating other techniques to get rid of the bus modulation, namely class DE switching applied to a phase-shifted bridge. Initially this might look a bit lousy since my existing QCW system needs to modulate the voltage from 50V to 350V. More investigation of the behavior of a double resonant transformer shows there are in fact 3 modes that can give zero current switching: the upper and lower poles*, and a zero frequency somewhere in between**. When operating at the zero, the voltage gain is simply due to the transformer ratio of the system, where as when you operate at either pole, the voltage gain is limited by output loading, so the voltage will just keep going as high as the spark lets it. Anyway, operating at the zero gives a more direct control of the toroid voltage and the whole thing acts like a voltage source. I measured a 5 foot straight spark and found that i need a starting voltage of 56kV and a top voltage of 65kV, so only about 15% modulation depth, which suggests to me that class DE switching is probably not a bad candidate. Of course the control may end up quite a bit more sophisticated.

* in my typical primary feedback schemes, the system will jump to one of these pole frequencies because they satisfy the zero phase requirement for ZCS. Depending on the tuning, one pole frequency gets more energy in it than the other and takes over once this dynamic settles out.

** It takes proper control to get the thing to oscillate at the zero frequency because there is not much gain there compared to the pole frequencies. Its worth noting that a system oscillating at pole frequency will actually converge to the zero frequency as the Q of the system drops sufficiently, and the system effectively looks like a voltage source. I believe my original QCW setup transitions from upper pole operation to zero operation, which is marked by a sharp increase in primary current (vs voltage) near the end of my ramp, that is, the power goes up more than linearly. It also causes my sparks to branch. Tests with a lower impedance resonator avoided this issue, likely because the lower impedance resonator maintains operation at the pole frequency.

Id like to write up a paper about all this QCW DRSSTC theory once i get it all worked out on my own. Turns out there are very many ways to tune a system, including some that are actually intentionally out of tune to exploit specific behaviors.