My DRSSTC does not want to resonate at correct frequency

[AstRii]

Hello,
I'm working on my first ever DRSSTC, i'm using Steve's Ward Universal Driver UD2.7 as shown on this schematics: https://easyeda.com/Little_Arc/Universal-DRSSTC-Driver-2.7
I've managed to get over 10cm (4inch) sparks on just 30V input. I've measured the resonant frequency to be close to 280kHz (tapped at 8th winding of primary)
During further testing the sparks from the coil shortened a lot, because my driver started to resonate on different frequency of aprox. 560kHz
I've tried to tune the coil by tapping the primary at different positions. On 1st winding i get frequency of about 880kHz, at the last one (8th winding) it gets to around 360kHz.
I'm still using the same amount of resonant capacitors, and i don't think this high change of frequency is because of the surrounding environment change. Something else has to be resonating or not functioning correctly. I tried to change C33 and C32 capacitors but pretty much nothing changed. And the strangest thing about this malfunction is that when i change my resonant capacitor's values, the frequency at which it operates almost doesn't change! Does anybody had similar problem? Or do you know what could be wrong?
Thanks!

[Hydron]

Have you got the CT connected the right way around? I've seen odd behaviour with higher resonant frequencies when it's connected out of phase.

[AstRii]

Yellow waveform is my primary current around 6Vpk, with current transformer with 43 windings and 1 ohm resistor as a load, therefore 6V*43windings = 258Apk
Red waveform is the GDT output measured on Gate of one of the transistors. The phase is not totally correctly adjusted but that shouldn't allow any strange behavior.. at least.. i think.
But as i said, i already got nice output out of it, it was resonating with 280kHz as it should.. but just for a while, then it suddenly started to oscillate at 562kHz (as on the scope waveform), i didn't change anything i was just playing around with the sparks when it changed the oscillating frequency. I tried to reverse polarity of feedback transformer, but that changes nothing. The OCD transformer is not yet installed.

[davekni]

What values do you have for L1, C32, and C33?  Also, what IGBTs are being used?  Since the gate drive signal lags the current sense a bit in your scope image, I'm guessing L1 is low or bypassed.  If the IGBTs are also on the larger slower side, it could be oscillating in a phase-delay mode rather than resonant.  Above resonate frequency, the primary LC circuit will have roughly 90 degrees phase lag.  If the other delays add up to 90 degrees (180 degrees total), and the current sense transformer is reversed, then it will oscillate there.  If the current sense transformer is correctly wired, then it would require a total of 360 phase lag to oscillate (270 degrees in the drive circuitry and IGBT delay), which is unlikely.

If current sense transformer reversal and readjusting L1, C32, and C33 isn't sufficient, then please post more scope traces, particularly H-Bridge outputs (drive to primary LC).

[AstRii]

Hello,
After some more debugging, i tried again reverse polarity of feedback CT and suddenly it works again! I know.. not a reliable system, but this time it seems to be working quite well for a long time now. my L1 is bypassed, C33 is 1nF and C32 is 150pF. I'm using FGA60N65SMD IGBTs. It now produces about the same sparks as before (10cm/4inch) from 30V input. Phase from feedback CT is not corrected, but i hope i'll be able to adjust it correctly. Strangely enough the primary current is not just about 55Apk, which is lower than my 258Apk when it was not in resonance.

[davekni]

Thank you for the update, which makes the issue a bit clearer.  The IGBTs are switching plenty fast.  The total delay from current-sense to gate-drive is excessive, around 900ns, which is almost 90 degrees of phase lag.

Studying the UD2.7 schematic, I'd recommend a few part changes:
1) At low currents, the junction capacitance of D1 and D2 will cause significant delay.  They have 170pF each typically (per spec.) for 340pF total, which is fed from 1K impedance, for a 340ns delay.  This delay will decrease at higher currents, since the average resistance of the diode pair decreases due to forward conduction.  I'd recommend changing to a lower-capacitance diode part such as BAT54GW.  (1N4148 or MMSD4148 for the SOD-123 package or other switching diode will work fine with typical values.  If you have any similar switching diodes around, that may be easier than waiting to order BAT54s.  With worst-case diodes and worst-case TL3116 parameters, 1N4148 diodes would allow the common-mode voltage of the comparator to go a bit above its 2.5V worst-case common-mode upper limit.  Typical TL3116s go to 4Vf per spec.)
2) C33 should be a bit lower for your relatively-high frequency operation of 280kHz.  I'd suggest 470pF, or 680pF at most.
I'd suggest scoping current sense and gate-drive again with the above two changes.  If the delay is more manageable, say under 400ns, then the rest can be handled reasonably with L1 phase lead.

If you feel adventurous, I could suggest some mod's to the circuit.  I prefer circuits that oscillate close to the primary LC resonant frequency without any feedback, then use current feedback to override the initial oscillation frequency and phase.  It doesn't look too difficult to modify the TL3116 circuit to accomplish that.

Best of luck debugging!