[ §1 · regimes ]

SSTC, CW or pulsed: the criterion that decides the topload

The argument goes round forever because both sides are arguing about the circuit. It is not a property of the circuit, it is a question of time scale.

CommonGnd4 August 202612 min read

The argument goes round forever. A slayer exciter is continuous; no it is not, there are gaps; a fast interrupter is basically CW; no it is not. Both sides are arguing about the circuit, and it is not a property of the circuit. It is also not settled by counting the gaps, which is the measurement everybody reaches for and the one that gives the wrong answer.

What the question is really asking

The label is normally reached for in order to settle something else: whether the topload earns its keep. That is a fair thing to want, and the label does not give it. What a topload costs puts that decision on how far the arc drags the resonance, which is a property of the machine rather than of the supply feeding it, and the two answers come apart on the same builder's bench. Richie Burnett runs his 4 MHz resonator with no toroid at all, having not found large ones to help a CW design (hfsstc); his own 350 kHz machine, fed straight off half wave mains, is topped with a 6 by 1.5 inch toroid and averages 14 inches of spark (sstate). Both are his, both get called CW, and only one of them has a topload. So this page does not claim to decide the topload on its own, and an earlier version of it did.

What the regime does decide is two things that are real: what the envelope makes of the discharge, and what the machine has to survive thermally. Those are the two sections that follow.

The criterion

Does the discharge reach its steady state within the time the coil is on? If it does, the physics is CW: a continuous balance of power, impedance and voltage. If it does not, the physics is pulsed, and the shape of the envelope is still writing itself onto the discharge.

That is the right question and this page cannot close it in the form it is written, because the time a corona takes to settle is not a quantity this corpus holds. Nothing here measures it, nothing linked from here states it, no source reached for this page gives it, and an earlier version of this page asserted a fraction of a millisecond with nothing at all behind it. Treat that figure, and any other offered for it, as unsupported until somebody films a discharge against a known envelope.

What can be done is to read the answer off the discharge instead, because the envelope leaves three different marks on it and they are easy to tell apart from two metres away.

Four supplies, one coil

Burnett ran a single 250 kHz half bridge off four different supplies and photographed the result each time, which is as close to a controlled experiment as this question has (sstate). The coil, the primary and the resonator did not change between them.

Half wave off the mains. The RF envelope is rounded bursts lasting 10 ms with 10 ms gaps. Sparks about 6 inches, very straight and sword-like, with the branching conspicuously absent. A muffled 50 Hz buzz. About 160 W.
Full wave. About 300 W, nearly double, and the sparks became fatter and bushier with "no increase in length". The note went up an octave, to 100 Hz.
Smoothed DC, which is true CW at about 350 V. About 420 W. The discharge became a bushy cone off the point, like a jet of burning gas; all the buzz went and left a pure hiss; a lot of ozone arrived quickly, and the thin wire at the base of the secondary overheated and blistered its varnish.
A phase angle controller firing at the peak, so the bus goes from zero to around 350 V in microseconds. About 180 W. Sparks about 6 inches again, but branched like a spark gap coil, spidery, and dancing about.

Two things fall out of that list, and they pull in opposite directions.

The length did not move. Six inches on the half wave, the same six inches with a phase controller, and no gain at all when the average power nearly doubled on full wave. That is what an arc grows as fast as charge arrives predicts, because a step of leader is the topload voltage divided by the breakdown field of air, and the peak bus is much the same in every one of these modes. Average power is not what buys length.

Everything else moved. The same coil at the same peak voltage made a sword, a bush and a spider, and the only thing that changed was the shape of the envelope. That is the observable, and it is the one a bench can actually use.

It also bounds the missing quantity from the useful side. If the corona settled in a fraction of a millisecond, then nine tenths of a 10 ms hump would be the steady state, and the half wave discharge would look like the DC run. It does not; it looks like the sword and the DC run looks like the bush. So on that coil the settling time is not small beside 10 ms [derived, from his four modes above]. How much of 10 ms it actually is remains unmeasured, and a lower bound is not a number.

The bigger machine says something the small one does not, and the confound is worth stating rather than hiding. Burnett's later 350 kHz H bridge, half wave off the mains with no smoothing capacitors at all, averages 14 inches with occasional hits at 18, at about 1200 W RMS input. Run from smoothed DC at a measured 1500 W it gives a 9 inch flame instead, hot enough to melt the end of the breakout point into a ball that went on glowing after the power was off. More power, shorter discharge. But that machine sweeps its drive frequency from the supply voltage in order to follow the arc as it detunes, and on a flat supply that sweep does nothing, so the CW figure is a coil running detuned as well as a coil running continuous. Two changes and one number. It is evidence that the envelope matters, not a measurement of by how much.

Reading the cases

DRSSTC, 300 µs at 200 bangs a second (the interrupter is the note). The bang is over long before anything settles. Pulsed, whatever the bang rate.
QCW, a ramp of six to twenty five milliseconds (the buck is the ramp), or 4 to 5 ms if the ramp is the rising quarter of a mains period rather than a converter. Steady state is deliberately never reached; that is what growing a sword means. Pulsed.
An SSTC straight off unfiltered mains. This is the case the argument is actually about, and cycle counting does not settle it. Ten milliseconds is 2500 RF cycles at 250 kHz and 3500 at 350 kHz [derived, 10 ms times the frequency], which sounds like ample room for a steady state, and Burnett's coil in that mode still makes the sword rather than the bush. The envelope is still deciding the discharge at 2500 cycles, which is the whole reason the gap count is the wrong measurement. Near the edges of each hump the bus is below breakout and nothing happens at all; what fraction of the 10 ms that costs is not established here.
True CW, meaning a smoothed bus. Bushy, hot, hissing. CW, and it is the mode that costs the most and shows the least for it.

One circuit difference does follow from the label, and it is not the one people argue about. Asked what a CW SSTC needs, davekni answers that its performance comes down mostly to how much line power you can draw before the breaker goes, so leave the bulk capacitor off and let the bus follow the rectified line: the current then roughly tracks the voltage and the power factor is far better than with the large bank an interrupted coil has to have. He puts a CW SSTC's coupling higher than a DRSSTC's as well, around 0.3 and perhaps 0.4 (HVF 2646). Note which way that cuts: on a ramped machine the bus page counts a half wave supply as the worse arrangement, because residual bus voltage branches the start of a ramp. Different machines, opposite advice, and the label is what tells you which one you are holding.

The numbers

The threshold: whether the envelope is still moving while the discharge is still growing. Not a voltage, not a duty cycle, not a count of gaps.
Corona settling time: not established. The only thing established is the bound above, that it is not small beside 10 ms on Burnett's coil.
A 50 Hz half wave: 10 ms of RF and 10 ms of gap, stated as the envelope on his coil.
RF cycles inside one hump: 2500 at 250 kHz, 3500 at 350 kHz [derived].
Spark length across all four envelopes on the small coil: about 6 inches, roughly 150 mm [derived], on inputs from 160 to 420 W.
A DRSSTC bang: tens to hundreds of microseconds, which is why it is never CW whatever the bang rate.
What a CW bridge is built from: davekni calls eight FGA60N65SMD in a full bridge plenty capable, and says the heatsinks want close spaced fins and a fan rather than size (HVF 2646).

What will get you

Three more that are specific to running continuously rather than in bangs, all of them reported by Burnett on the coils above:

It burns rather than shocks. He picked up small RF burns from touching metal near the running coil at fairly low power, and reports such burns to be very nasty. A pulsed coil gives you gaps to be lucky in; a CW coil does not.
Ozone builds fast. The smoothed DC run produced a lot of it quickly, in a workshop. Ventilate before the run, not after it.
Anything resonant nearby joins in. With the breakout point left off, an unused resonator two feet away lit up with a crown of corona of its own. That is the near field finding a better place to break out, and it will find people too.

What goes wrong

A topload was fitted and the arc got shorter. This is not the CW question. The topload bought stability against detuning that the coil was not losing, and the volts are the whole cost. What a topload costs is the trade.
A coil advertised as CW hums at mains frequency. It is unfiltered. The corona is modulated at 50 Hz on a half wave and at 100 Hz on a full wave, and the buzz only disappears on a smoothed bus. Not a fault, but the sound is a free measurement of which one you have.
Doubling the input power did not lengthen the arc. Expected. Length follows peak voltage; average power buys thickness and heat.
The discharge went from a sword to a bush when the bus was smoothed. Also expected, and it is the whole content of the CW question.
The devices die after seconds of continuous running. They were sized for a duty they are not being given. Whether an SSTC bridge switches softly at all, which is what decides the heatsink, is its own page.
A five minute video of continuous operation. Usually interrupted; fast bangs look continuous to a camera and to an eye. An arc grows as fast as charge arrives puts the boundary near 500 bangs a second, where the channels barely cool between them and it reads as continuous light.
A transformer feeding a half wave saturates. The single diode is pushing DC through the winding.

Where next

What a topload costs, which holds the decision this page is usually invoked to settle, and settles it on detuning rather than on the regime.
The interrupter is the note, on ontime and bangs per second, which is what puts a coil on the pulsed side of the line and what turns it into an instrument.
An arc grows as fast as charge arrives, for why length followed the peak voltage and not the power in all four of the runs above.

The SSTC department is the one where this question comes up, because it is the topology that can genuinely be run either way. The other two are pulsed by construction.

CWinterruptertopload