[ §1 · failure ]

The SSTC second mode, and flashovers mid-winding

A failure that looks like too much coupling and is not. It also explains why a two inch secondary works in a DRSSTC and burns in a bare SSTC.

CommonGnd4 August 202613 min read

A secondary that flashes over across the middle of its winding is usually blamed on coupling. Sometimes it is coupling. But there is a mechanism that produces exactly that failure with the coupling untouched, and it explains why a small secondary works in one topology and burns in another.

What it is

A resonator has more than one mode. The lowest is the one everybody tunes, and its voltage peak sits on the topload. Anders Mikkelsen went after a flashover on his own coil, measured the base impedance of the winding with no topload fitted, and found the higher modes where a resonator's modes are expected to be, with the second at roughly twice the fundamental (HVF 3097). Twice, on that coil, bare. That number matters, because the number that gets you into trouble is three and two is not it.

The topload is what closes the gap. Capacitance at the top pulls the fundamental down harder than it pulls the higher modes, and when he added enough of it to stand for the detuning his own sparks produced, the fundamental and the second mode came to sit three to one apart. A square wave drive carries a strong third harmonic, and at that spacing the third harmonic is sitting on the second mode and pouring energy into it. What that does, in his words, is give "a voltage peak in the middle of the secondary coil".

So the chain has a step in front of the one usually quoted:

topload and arc pull the fundamental down
        ->  the second mode ends up near 3x the fundamental
square wave drive
        ->  strong third harmonic
        ->  the third harmonic drives the SECOND mode
        ->  voltage peak in the MIDDLE of the winding
        ->  flashover

He confirmed it rather than inferring it: at the operating point where the flashovers happened, the base current was mostly the third harmonic of the drive. Spark loading makes it worse twice over. It is part of what pulls the fundamental into the three to one spacing in the first place, and it damps the fundamental mode harder than it damps the second, so the arc both sets the trap and then holds the energy in it.

The damage happens where nobody is looking, because the top of the coil is fine.

How far detuned before it bites, on his coil. Bare, his second mode sat at roughly twice the fundamental, and the trouble arrived at three times. If the second mode barely moves, which is his stated premise rather than something he measured, the fundamental has to fall to two thirds of its bare value before the third harmonic lands on it [derived: 2/3, from his two ratios, 2:1 bare and 3:1 loaded]. Capacitance goes as the inverse square of frequency, so the coil carries about 2.25 times the capacitance the bare winding had by the time it gets there [derived: (3/2)^2]. That is the size of detuning this failure waits for, and it is a long way past anything a bench measurement of the bare coil will show you.

What goes wrong, and how to tell it from coupling

A flashover across the middle of the winding. This mechanism. Racing sparks from coupling strike the winding low down instead, which is how the secondary is a resonator sorts the two by height. Where the ceiling on coupling actually sits is a different page's question, and this failure does not move it.
A small secondary that works in a DRSSTC or a QCW and burns in a bare SSTC. ZakW could not get his old potted 2 in secondary to run without flashovers on a single resonant machine, nowhere near full input voltage, while the same coil gives him no real trouble in his ramped dual resonant one (HVF 3425). The observation is one builder's, on one small coil, in both machines. Why small is worse is not established, and there are two guesses on the table further down this page.
The arc squeals. Not a mechanical noise. ZakW hears a squeaking or screeching arc when his primary is not high enough on the secondary, and davekni, offering what he calls a loosely described guess, puts it down to a beat: the second mode sitting near the third harmonic, with the difference between them landing on the primary's own resonance. What makes it more than a guess is that ZakW's scope capture of the squeal shows the third harmonic appearing periodically in the secondary current, which is what pointed davekni at it. Treat the squeal as a signal that the second mode is being fed, not as a diagnosis of which of the two frequencies is at fault.
The bridge locks onto the second mode itself. A different failure from the one above, and worse. davekni's account is that standard SSTCs with long secondaries and short antennas, or with secondary current feedback, sometimes fail by oscillating at the second mode outright, so the bridge is running at that frequency rather than at the third harmonic of a lower one, and the overly high frequency often kills IGBTs. Note which way round it goes: on a coil with a short secondary and a comparatively large topload he expects antenna feedback to avoid the instability, because the antenna hears the topload far more than it hears the middle of the winding.

What you do about it

Anders lists three cures on HVF 3097, and they are worth having in his order because the first is the only one that gets better as the arc grows.

Put the second resonance above three times the fundamental at design time, by giving a secondary of a given inductance a larger toroid. Once the mode sits above three times, further spark loading only pulls the two further apart, so the coil walks away from the trouble instead of into it.
Move the primary against the secondary to couple less into that mode and more into the fundamental.
Resonate the primary. He calls this the easiest and most robust, and the reason is not that it moves any mode: a resonant primary cuts the reflected impedance of the fundamental until the fundamental draws far more power than the higher modes do. He fitted one and performance improved greatly. It is also the whole of why the same small coil behaves in a DRSSTC or a QCW, and it is not free, because a primary at resonance is a DRSSTC with a DRSSTC's currents and a DRSSTC's need for overcurrent protection.

On the bench, with the coil already built, the handle is the height of the primary, and it is a search with a symptom at each end:

Too low and the arc squeals and the switching goes strange. Raise it and the switching goes back to normal.
Too high and you get a flashover anyway.

That is ZakW's own procedure, and he reports the same behaviour on his older ramped hardware, so it is not a property of one driver. Anders arrives at the top end of it independently on a later resonator of his own (HVF 3097, later in the thread): he attributed occasional flashovers to an optimistically high primary and cured them by moving it down and spacing the turns out, which took coupling from 0.45 to 0.50, and a new primary later reduced the flashover risk again while taking coupling to about 0.55. He does not say those particular flashovers were the second mode, and that resonator is not the one he diagnosed it on, so read it for what it is: the same builder finding the primary's height to be the flashover knob, on a coil where coupling went up while flashovers went down.

The other handle is the series capacitor, which ZakW calls the DC blocking cap and davekni the coupling cap, the same part either way. It sets the primary's own resonance, and davekni's guess makes that resonance the thing the beat has to match, so moving it moves the trouble. His suggestion is to make the capacitor much larger, which drops the primary resonance low enough that little of it gets through the gate transformer and lowers the primary's resonant Q as well. What that page of the thread also says, and what a one line version of this advice loses, is that a larger capacitor does not remove the mechanism. It relocates it: with the primary resonance lower, the beat can only match it if the second mode sits closer to three times the drive, and once those two are close enough you are in the flashover case instead of the squeal.

What the evidence does not support

That the beat is established. davekni offers it as a guess and says so. What is measured is the third harmonic in ZakW's secondary current during the squeal, and the correlation between the squeal and the primary's height. The rest is a plausible mechanism that nobody has tested.

That thin secondaries are simply worse. Two explanations are on the table for ZakW's small coil and neither was tested. davekni's is that the smaller coil may have had its second mode closer to three times the operating frequency; he writes it with a "perhaps". ZakW's own is geometric and has nothing to do with modes: on a coil that short the primary is simply too close to the middle of the winding, and the bus voltage sets how small you can go. Anders' fix list points at a third reading, since the design cure is the ratio of toroid to secondary inductance rather than the size of the secondary at all. A small secondary is not condemned by any of this.

That there is a number for how close is too close. The corpus's arithmetic below turns davekni's guess into a prediction, and a prediction is what it stays.

The numbers

Where it strikes: the middle of the winding. Racing sparks strike low down, and that height belongs to the coupling pages.
What excites it: the third harmonic of a square wave drive, once the fundamental and the second mode have come to sit three to one apart.
Anders' bare coil: second mode at roughly twice the fundamental, measured at the base with no topload. His coil, his measurement, and not a property of resonators in general.
The detuning it takes on that coil: fundamental down to two thirds of its bare value, which is about 2.25 times the bare winding's capacitance in place [derived: 2/3 and (3/2)^2 from his 2:1 and 3:1].
ZakW's primary resonance at 1.0 µF: about a third of the working frequency, davekni's figure. Put through the criterion on the blocking capacitor page, X_C / X_L = (f_res / f)^2, that is a frequency margin of 3, which cancels a ninth of the primary's reactance, 11.1 per cent [derived: (1/3)^2]. It sits between the two published SSTCs that page records, at 17 per cent and 1.1 per cent, so davekni's figure describes an ordinarily tuned single resonant coil rather than a strange one.
The cross-check on that figure: ten times less capacitance raises a resonance by the square root of ten, so a third of the working frequency at 1.0 µF becomes 1.05 times it at 0.1 µF [derived: 3.162 / 3]. That is exactly what ZakW calls close to resonance, and neither man wrote his statement to agree with the other's.
What the beat predicts, if it is right: the difference between the second mode and three times the drive equals the primary resonance, so on that coil the second mode would sit at 2.67 or 3.33 times the drive [derived: 3 ± 1/3]. Measurable, and unmeasured.
ZakW's coil, for scale: 31 in of arc from a 4 in secondary at 40 A, off a 340 V half wave doubler into a half bridge, his figures. He does not say whether the four inches is the diameter or the winding length.
Not established: whether a smaller secondary is worse, and if so why.

Where next

Winding and measuring a secondary, the rest of what goes wrong with one, sorted by the height of the damage.
DRSSTC coupling: how far above critical, and what raising it buys, the other flashover mechanism and its ceiling.
An SSTC primary is a different part, since the handle here is where that winding sits and how much it covers.
The capacitor that decides what you built, which is where the series capacitor's value is actually decided.

The mechanism, the measurement and the three cures are Anders Mikkelsen's, on HVF 3097. The squeal, the primary height procedure and the small-coil comparison are ZakW's, on HVF 3425, where davekni proposed the beat and the second mode lock. The arithmetic tying their figures to this site's own capacitor criterion is ours and is marked where it appears.

flashoverharmonicssecondary