The coil runs and the bridge gets warm, but nothing comes off the breakout. Or something does, and it is a third of what it should be.
Work through the checks below in order. The cheap ones come first, and they are also the ones that are usually wrong.
Three checks before anything else
They cost nothing, so do them first.
- The feedback phase is inverted, 180 degrees out. A current transformer wound the wrong way round drives the bridge against the current, and almost no energy gets transferred. Turn the ring over to test it. Where the ring sits is in two rings, not one.
- The secondary's ground is bad. It is one side of the resonator's circuit as well as a safety connection, see what RF ground is.
- The feedback signal is weak, usually because the current transformer has the wrong core or the wrong material. Measure it in volts per amp and compare it with the band working boards sit in.
None of these is a tuning problem, but all three get misdiagnosed as one.
One observation looks like a pass and is not. A clean, full amplitude sinusoid on the secondary, with the gate frequency equal to it, tells you the tracking loop is working and nothing more. A resonator rings and a loop follows it at power levels far below anything that will break out, so a locked loop is perfectly happy with a machine that delivers almost nothing to the arc. It rules out the first item above and nothing else.
Then the ones that look like a small coil
Much of what follows comes from one first-time builder's thread on the High Voltage Forum, First DRSSTC, No Sparks on Breakout. People are named where their advice is used.
On that thread Uspring tells the builder to lengthen the breakout rod somewhat, to at least three times its present dimension, which will reduce the breakout voltage and simplify tuning somewhat.
That gets a coil started. It is not the right length once the coil runs. davekni reports the opposite at the other end of the same problem. His longest arcs came off a short breakout on the upper side of the topload, a thin cone with the wire extending slightly past the point, the cone holding a high field at the tip while keeping corona down. He says what governs it is sharpness and distance from the topload, and length is not on his list. Both are coil specific and both are set by the voltage you want arcs to start at. Lengthen the rod to get going, then shorten it to grow the arc.
Keep metal away from the coil. A chassis or a case near the secondary kills the Q and loses energy to induction heating. The minimum clearance is half the secondary's diameter, so on the 110 to 160 mm secondary of an ordinary coil that is 55 to 80 mm derived from [this site's own size classes]. A clearance of 100 to 200 mm is for a 200 to 400 mm secondary, the large and very large end.
Test in the dark. Tesla coil arcs are not very visible in normal daylight. davekni's own DRSSTC running 4 to 5 kW is, by his account, faint in evening light just before sunset. In daylight you will undersell your own machine and go looking for a fault.
The primary current flattens too early
If the current stops rising partway through the bang, energy is being lost before it reaches the arc. Uspring's three places to look are below, in his order of likelihood. He thinks the first is the most likely, and the second is the one with explosive consequences.
- The primary leads are too thin.
- Excessive collector-emitter drop.
- The ESR of the bus capacitors.
The wire gauge for those leads comes from someone else. davekni gives it on the same thread. A few 16 AWG wires in parallel work great, he says, even slightly better than the equivalent copper area in a single conductor, since it is just a bit of home-made litz wire.
The bang is too short
A DRSSTC does not behave like a spark gap coil here. Cut the bang off early and most of the primary's energy goes back into the bus capacitors instead of into the arc. Uspring reads that builder's scope shots and puts the first at a very short burst of about 50 µs and the later ones at about 120 µs, which he calls still on the short side but it might be ok. Those figures are for one coil at one power level, and we have no general figure for a mid-sized coil. Longer ontime gives a longer spark, especially at modest peak current.
Tuning, in the order that works
Measure the secondary's resonance first, with the topload and the breakout point in place. There are three ways. The first two are davekni's, in his own words, from a thread about a driver killing IGBTs. One is a signal generator wired through a resistor to a scope input, say 1k to start, with the frequency swept until the voltage drops the lowest. Once it is roughly found, raise the resistor to 10k, which will make a sharper amplitude dip. Another is a ring-down. Put a capacitor of about 1 nF across a scope channel input, touch the secondary's ground lead to a somewhat high voltage source, say 170 V, and then to that input. The ring-down is at the secondary's frequency. The third is an antenna near the topload, looking for the field peak.
Tune the primary slightly below the secondary. The arc adds capacitance, so the secondary slides down as the spark grows, and the detuning should increase with it. Adding primary turns lowers the primary's frequency, and the useful range is wider than it looks. On that same thread Marek Novotny tells the builder not to worry about detuning the primary well below the secondary, and to try it all the way from that coil's 170 kHz down to 100. That is a range to search. No one on the thread reports having swept it.
What you are tuning to is the lower of the two poles. Bert Hickman puts it at about 90 to 91 per cent of the uncoupled secondary for k between 0.18 and 0.20, adding that the folk figure of ten per cent low is indeed in the ballpark. The relation used elsewhere on this site, f0/sqrt(1+k), gives 92.1 and 91.3 per cent at those two couplings [derived]. The two agree to about a point and a half, and the difference is a matter of convention. A flat ten per cent, though, is k = 0.235 [derived], not 0.18 to 0.20, so the folk rule and Hickman's band are not the same statement.
JavaTC gives a good starting estimate. Nearby metal and walls throw it off, so try taps by hand from there.
Uspring's method is to tune at low power for the maximum ratio of secondary current to primary current. That ratio should be independent of the input power, so you can find it at a voltage that breaks nothing. He adds a limit in the same post. The maximum ratio is not exactly the best for high power with long arcs, because the arc detunes the secondary by adding capacitance to it, and from there it is better to bring the primary's resonance down a bit. Find the maximum first, then detune off it.
Coupling for an ordinary DRSSTC is around 0.15 to 0.2, which is where Steve Ward brought DRSSTC-3 down to in the note above. Higher coupling transfers more and is less sensitive to detuning, until racing sparks arrive, and those are what made him reduce it. Lowering the secondary raises it. There is one recorded step, in JavaTC's figures rather than in a measurement. Uspring reads that builder's JavaTC file at a coupling of 0.134 and suggests lowering the secondary a bit with respect to the primary, and the builder lowers the secondary and topload assembly by 2 cm and reports that this increases coupling to 0.172.
Where the current transformer goes
And the things that kill the bridge while you are looking
Two faults show up as a bridge that dies now and then, rather than as no spark.
The first is the bootstrap trap. If the driver stops switching in the high state, the bootstrap capacitor never recharges and the high side desaturates. Shock heating kills it quietly, and then the low side goes in shoot-through with a bang. That sequence is TMaxElectronics' own post-mortem. His driver simply stopped switching instead of pulling both legs of the bridge low. A large bootstrap charge resistor and the drop across the bottom transistor together took the bootstrap voltage below 13.5 V. The top IGBT shock heated and shorted quietly, and the bottom one then had, in his words, shoot through from hell. To avoid it, keep the low side on during pauses, or use a floating gate supply, or add a fast or Schottky diode across the driver chip's internal bootstrap charging diode. The diode is davekni's suggestion, for having some hope of charging the capacitor in one half cycle at the start of each burst. The resistor to go with it is TMaxElectronics' own 1 ohm, which he reckons gives five RC time constants in one low switching pulse even after the diode recovers.
The second is bus resonance. The film capacitors and the layout inductance resonate. An H bridge is essentially a reverse bridge rectifier, so the ripple appears at twice the primary frequency, the way a 50 Hz line supply bridge ripples at 100 Hz. If that lands on the resonance, the bus swings hard. All of this, arithmetic included, is davekni's on the same thread. With 4 by 3.9 µF and a 5.83 µs period he comes up with 55 nH for the resonance, which he thinks reasonable for the standoffs, the bulk capacitor busbars and the capacitors' own internal inductance. He gives three fixes. Lower the inductance with copper foil bent onto the existing busbars. Detune by doubling the film capacitance, which puts the resonance at the square root of two times the primary frequency instead of at twice it. Or, where a high Q resonance has to be avoided, add series resistors to some of the film capacitors.
The figures and cases here come from published threads and other people's benches. The DRSSTC diagram puts the current transformers on the return lead, at the low potential end, for the reason in the warning above.