SSTC primary coil design: why it is a different part

Most of what is written about primaries on a doubly resonant coil does not apply to an SSTC. Its primary is wound differently, coupled much tighter, and sized by a different rule.

What it is and why

A DRSSTC's primary is half of a tuned circuit. An SSTC's primary is not tuned at all. It is a transformer winding, and it is meant to be a good one.

Both machines have a capacitor in the same place. On a DRSSTC it is chosen to resonate with the primary. On an SSTC it is chosen so that it cannot, and that is the whole difference between the two.

Everything below follows from that. Without a tank there is no resonant multiplication of the current, so the coupling has to do the work instead.

What you decide

How tightly to couple, and where to put the winding

Criterion: energy transfer, with no resonance to help.

A DRSSTC's primary sits loose around the base of the secondary, a few turns at k of about 0.15 to 0.2, which is where Steve Ward brought his DRSSTC-3 down to once arcs began racing up the secondary (pupman archive).

An SSTC's sits close around the secondary's form. Burnett's is less than half an inch from the secondary winding, and the photograph further down shows the same on his first coil. With no tank to multiply the current, the geometry has to supply the coupling.

How tight is tight enough is harder to say. Richie Burnett puts a floor under it on his solid state pages: very tight coupling, above 0.35, is what the primary feed method needs for good power transfer, and he names that as the method's significant disadvantage, because it is what makes the insulation hard. His own coils are given at about 0.40 on that page and at about 0.55 on the design page.

Those are his own words and his own hedges, estimated for the first and approximately for the second, and neither page says how either figure was arrived at. Under the rule this site runs on for k they count as his figures, not as measurements. The two descriptions may not be one coil either. The project page says 19 turns and the design page 18, both at 350 kHz, and he never says whether that is one machine at two stages, so their figures are quoted separately and not combined. His 0.55 is also a different 0.55 from the one argued over in the coupling article. That one is dr. kilovolt's, whose primary has a ferrite rod core through it, while Burnett's is air cored on a plastic form.

Move the primary and you have detuned the coil. Burnett's reflected model says why: raising k couples more of the secondary to the primary and leaves less of it free to resonate, so the resonant frequency seen from the primary rises as the coupling is tightened. He is explicit that a solid state coil has to be re-tuned after any adjustment of the coupling. On a machine where the winding is deliberately slid up and down the form to set k, that matters every time. Your driver decides whether it costs you anything: Burnett's free running TL494 oscillator has to be moved by hand, while a driver that takes its frequency off an antenna or a current transformer finds the new one by itself.

Coverage, meaning how much of the secondary's height the winding spans, has no rule behind it at all. The one build on file is Burnett's, link coupled over the bottom third of his resonator, about 33 per cent, and one coil is an example rather than a specification.

No source in reach gives a band for it, including the 20 to 50 per cent that has been suggested. If you need a starting point, take the real build above and expect to move the winding along the form, because the number is found by trial rather than calculated.

How many turns

Burnett's rule balances two currents that both flow in the primary:

Lp should be made only sufficiently large to obtain an acceptable magnetising current, then k should be maximised as far as possible.

That is from his driver design page, where he also sets out the three ways it goes wrong. Too little Lp and the magnetising current becomes unacceptably high: it draws no supply current, so it does no work, but it heats the switches anyway and it puts ripple into the reservoir capacitors. Too little k and the load current is mostly magnetising current, with little of what the driver supplies reaching the secondary. Too much k and the two windings are close enough to flash over.

The lever is steeper than it looks. He notes that halving the primary turns takes Lp to roughly a quarter, and that in an ideal world the impedance the driver sees falls by the same factor, with the current rising to match. It is the right lever because it moves the load current without moving the resonant frequency, which the coupling adjustment does not leave alone.

What the current will be, before you choose devices

The current is not set directly. It falls out of the reactance. Burnett's design page gives all three inputs for his present coil: an 18 turn, 20 µH primary driven ±340 V at 350 kHz (driver design page).

X = 2·pi × 350 kHz × 20 uH = 44.0 ohm
I = 340 V / 44.0 ohm = 7.7 A peak

Both lines are [derived] from his three figures, and the second reads the bridge's square wave as a sine of the same amplitude, leaving out the primary's resistance and the blocking capacitor's reactance.

That reading is the weak point. A reactance is a sine wave quantity, but a bridge makes a square wave, and across a bare inductor a square wave makes a triangular current. Burnett's model says which current that is. He splits the load current into two parts. One is a magnetising current, triangular and ninety degrees behind the voltage, which flows whether or not anything is coupled to the winding at all. On top of it sits the resonant current reflected back from the secondary, sinusoidal and in phase, which is the part doing work. The reactance route and the triangle both describe the first of those.

Three routes, same three inputs:

7.7 A    340 / 44.0                       [derived] bus read as a sine amplitude
9.8 A    (4/pi) × 340 / 44.0              [derived] the square wave's fundamental
12.1 A   340 V × 1.43 us / 20 uH, halved  [derived] triangular peak

The 1.43 µs is the half period at 350 kHz, over which 340 V across 20 µH moves the current 24.3 A from end to end [derived].

An SSTC primary, then, carries around ten amps of magnetising current and about twenty measured, against the hundred amps and up that a DRSSTC's tank carries. That is the impedance page's subject, and the DRSSTC figures are sourced there. The low current is the other reason an SSTC takes MOSFETs rather than IGBTs.

What will get you

Some things do carry over from a DRSSTC primary. It still melts if you run it hard, it still must not sit on a metal frame that becomes a shorted turn, and skin effect still means the middle of a round conductor is not carrying anything. Those parts of what a DRSSTC primary is for are unchanged.

The numbers

• Coupling: Burnett's floor is above 0.35, and his own coils are given at about 0.40 and about 0.55, neither with a method stated. Against about 0.15 to 0.2 on a DRSSTC.
• Gap between the windings: under half an inch on Burnett's.
• Coverage: not a rule. The bottom third on Burnett's, the only build on file.
• Burnett's own coil, for scale: 18 turns, 20 µH, ±340 V at 350 kHz, k about 0.55, with 18 inch sparks common, which is 457 mm [derived, 25.4 mm/in].
• The worked chain, on Burnett's figures: 20 µH at 350 kHz is 44.0 Ω, and 340 V across that is 7.7 A peak. Two other readings of the same square wave give 9.8 and 12.1 A [derived], and the reflected resonant current is on top of all three.
• The only measured primary current here: 22 A peak, Burnett's H bridge coil.

What goes wrong

• Flashovers that burn the primary form. On Burnett's coil, a smooth toroid with no breakout point. Give the field somewhere to break out other than the primary.
• The switches get hot with a short arc to show for it. Magnetising current: too few turns. It draws nothing from the supply and heats the bridge anyway.
• The bridge cannot get power in however hard it is driven. Too many turns, so there is no step-up, or too little coupling, so the load current is magnetising current and little of it reaches the secondary.
• A device chosen on the calculated current that dies at full power. The calculation gives one component of the current. Put a current transformer on the primary before believing any of the three routes above.
• A flashover in the middle of the secondary rather than the top. Not the primary's doing. That is the second mode, and the handle for it is the primary's height, not its turns.

Where next

• Why an SSTC couples tight, where the tight coupling turns out to decide whether the bridge switches softly at all.
• The capacitor that decides what you built.
• The primary is an impedance, the DRSSTC side of the current question and where the hundreds of amps are.
• What a DRSSTC primary is for, the parts that are common to both machines.

---

The SSTC department diagram draws its tank across the middle of the bridge with no capacitor in it, and that missing capacitor is what separates it from the next department.