QCW operation using pulse skip?

[profdc9]

I was reading about using a buck converter to ramp up the voltage during a pulse for QCW operation:

http://www.loneoceans.com/labs/qcw15/

The buck converter regulates the voltage on the bridge by either changing the duty cycle or the frequency of the driving signal.  But could the same be done just by skipping pulses?  For example, by using pulse skip operation, but ramping up the overcurrent threshold, at the beginning of the pulse, more pulses would be skipped and more freewheeling would occur, and then as the pulse progressed and the overcurrent threshold increases, fewer pulses are skipped and the current on the bridge is allowed to rise higher and higher.  This way the main bridge can resonantly regulate the power, which while perhaps not increasing quite as smoothly as a buck converter, still might achieve the same effect if the overcurrent is allowed to ramp up over several ms.

I think a programmable driver such as the UD3 with the PSoc5 could easily be programmed to do this.  Does sound like it would work and would be feasible?

Dan

[malte0811]

Steve Conner tried this (or something close to it) in 2011, without success (http://4hv.org/e107_plugins/forum/forum_viewtopic.php?128784.0). The thread quickly gets away from the topic and moves to experimental tuning modes (the zero in the transfer function and even more unusual things). As I understand it the bad performance is probably due to low ramp quality. Another possibility (loneoceans mentions this in his QCW1 page) is that the primary current envelope doesn't need to be linear. He measured a roughly quadratic envelope with his QCW2 (might be 1.5?), though I recall seeing someone else (Steve Ward?) showing measurements with nearly constant primary current.
Something I'd be interested in seeing would be a 'traditional' buck-modulated QCW, but taking the buck feedback from the primary current rather than the bus voltage. There are probably good reasons why this hasn't been done, I don't have any experience with or proper knowledge about buck converters.

[profdc9]

I know the solution of using the resonance of the primary current to smooth out the ripples in current due to pulse skipping would like be an obvious solution.  The experiments of Steve Conner seems to have been conducted on a coil tuned to 220 kHz which may have been too low for good QCW operation, as most examples seem to be above 350 kHz.  I am just curious because it seems that UD3 could be programmed to pulse skip, which the number of skipped pulses decreasing as time goes on so that the current can rise slowly.  I don't really know what I'm talking about because I'm just getting started, but perhaps there's an obvious reason this would not work.  The UD3 allows more control though, and this might have this strategy a more viable option.

Dan

[Steve Ward]

I have tried regulating primary current with ramping its reference with the UD3 and couldn't get any good sword sparks to happen.  Consequently, I've used phase shift modulation instead to modulate drive voltage with constant bus voltage.  Chris Dickey tried sigma-delta pulse-skipping modulation on the UD3 to put it back in "voltage control mode" (rather than trying to regulate primary current) and also could not get the sword sparks.  Of course, I don't rule out that it's possible, but so far i don't think anyone has made the sword sparks this way.

I know the solution of using the resonance of the primary current to smooth out the ripples in current due to pulse skipping would like be an obvious solution.

It does make sense, but i think there is a very disappointing practical limit here because you don't really want to store much energy in the L/Cs because energy storage is expensive.  Basically, if the coil is storing enough energy to smooth out the missing pulses enough to mimic the ripple performance of a buck converter, then it's probably massively overbuilt or very exotic low-loss components... or someone can prove this hand-wavy theory wrong :-).

This is all under the assumption that to "work" means it produces the pretty branchless "sword" sparks.  Of course, if you don't care what shape or form the sparks tend towards, then pulse skip modulation certainly works fine and is quite efficient.  From what i've seen, to get the branchless sparks to form, you need relatively smooth/continuous modulation of the spark power with little ripple.

In control theory, you could wrap the buck modulator control loop around the primary current rather than the bus voltage.  There would be more delay between your buck PWM modulation and the resulting primary current change, so stabilizing such a control would be more challenging, requiring some additional phase lead or feedforward to get enough phase margin.  However, this additional phase delay between control and feedback depends on the Q of the tesla coil as well.  If the Q was high, for a constant input voltage we expect a growing current without bound, this looks like a 90* lag in controls world and is a good hint that there be stability issues to think about.  For real coils, the Q is relatively low, so the primary current tends to reach its final value within just a few cycles which means this phase lag will be somewhat less than 90* when you add in this "damping" from producing lossy plasma.

The non-obvious part is what amplitude the current should take over time.  My measurements seem to support my intuition that branchless sparks want to be fed such that their input power goes up with the square of the time (multiplied by some other coefficient that seems to be frequency related).  Basically, if you want to double the length of a branchless streamer you can get there by doubling the the driving period and should expect 4X the input power by the end of the driving period. 

For voltage modulated coils, the primary current can look funny because of the energy stored just getting the coil up to high enough voltage to get any streamer power at all.  This gives the primary current shape an offset as you must ring up to that many primary amps to get the minimum useful secondary voltage to make plasma start.  Furthermore, changes in tuning will modulate the effective transfer ratio of the tesla coil (essentially the ratio of primary current to secondary current) which of course alters the primary current envelope shape.  If the coil is tuned for optimal tuning at max spark length, this tends to give a mostly flat current amplitude as the coil starts "out of tune" and requires higher starting current but the efficiency steadily improves as the spark grows, so the required primary current grows only a little.  If the coil is less optimally tuned, the streamer capacitance detunes the system and the primary current may grow more than linearly near the end of the spark event because the poor tuning demands higher primary current.

Basically, if the coil didn't have to store energy (including being out of tune) , the primary voltage and primary current would both ramp up linearly over time, I claim.  It helps to think of the ideal case, then add all the other stuff that messes with that on top of it, to get a picture of how the machine works.

[profdc9]

Thank you for the thorough answer, Steve.

I suppose that using a separate buck converter allows the switching frequency and energy storage to be separately configured from the primary tank circuit, so that the size of the current packets delivered through the buck converter can be determined by the size of the buck inductor, as long as a compromise can be achieved between switching losses in the buck converter (because it is operated hard-switched) and the size of buck inductor needed to store energy in the inductor ferrite.  I suppose it could be asking too much for the primary tank circuit to be storing energy between cycles as well as optimally delivering energy to the plasma spark to achieve the quadratically varying power profile needed to grow the spark as a sword.

With UD3, it could be operated in open-loop control, because if the coil and other components are fixed, in theory a fixed ramp could be used on the buck converter frequency and/or the sequence of skipped pulses, unless there is enough pulse-to-pulse randomness to make this technique not viable.  It is fascinating to watch the sword sparks jump around, so I suppose the variation in their length and endpoint indicates that open-loop control may not work.

I suppose using large electrolytics to store hundreds of joules of energy is probably one of the most economical ways of doing so, and therefore using a buck converter to convert that energy quickly to a form usable by the primary tank circuit is a practical choice.  Perhaps the buck converter could be combined with power factor correction so that the coil wouldn't have to store so much energy per pulse.

Dan