Are TVS diodes necessary for full bridge?

[Simranjit]

Hi, recently I built a half bridge Tesla coil with almost zero transients at drain of mosfets. So I went to full bridge using same method for winding GDT and building power circuit to connect mosfets. To test it I first just supplied oscillations to circuit via 555 timer. The gate waveform looked very good. Then I tested it with 15v power applied and there was significant spikes on gates of mosfets. I pushed my luck and connected it my ramped ac circuit. I got some sparks but when I increased the duty cycle to 5% the fuse was blown and all four mosfets were shorted.
Also, I was short on mosfets and just used two IRF840 on positive side and two IRF740 on negative side. Could that be an issue? Or I was supposed to use TVS diodes or snubber RC circuit for it ?
Sorry I do not have pictures for waveforms as I do not have any mosfets left. But I will attach pictures for my circuit.

[davekni]

Also, I was short on mosfets and just used two IRF840 on positive side and two IRF740 on negative side. Could that be an issue?

Presuming genuine parts, specs are close enough, so not likely an issue.  For future when using mismatched pairs, I'd suggest using one part type for one half-bridge and other type for other half bridge (rather than + and - sides).  Then any mismatched switching times won't cause cross-conduction (shoot-through).

Or I was supposed to use TVS diodes or snubber RC circuit for it ?

Gate-source bidirectional TVS diodes are recommended.  But I'm guessing lack of TVS diodes isn't cause of your failure.

What line voltage are you using for the ramp?  What Vbus cap(s) are on the back side of H-bridge board?  What frequency is your SSTC?  What feedback are you using (circuit diagram)?
There are many possible causes for H-bridge failure.  Possibilities that come to mind for this case are:
1)  Higher current of full bridge is more than these FETs can handle for long on-time of a line half-cycle.
2) Higher load on driver board (four gates) is causing problems with driver board.  Hand-wired (breadboard) construction can be problematic, especially for the high peak currents of gate driver chips.  Using sockets for driver chips adds further to parasitic inductance and thus possible issues.

Good luck with repair and debug!

[Simranjit]

Presuming genuine parts, specs are close enough, so not likely an issue.  For future when using mismatched pairs, I'd suggest using one part type for one half-bridge and other type for other half bridge (rather than + and - sides).  Then any mismatched switching times won't cause cross-conduction (shoot-through).

I have attached the picture of how they were connected. Sorry for bad drawing. I'm currently at work. Parts I think are not genuine. I'm noticing a pattern since I got the oscilloscope. Some schmitt triggers and flip flop 74HC74 ICs don't work (maybe 1 out of 10 work properly). That is why I resorted to use esp32 for interrupter. Mosfets I used in half bridge were IXTQ52N30P (listing said original disassembled parts) and I was able to get 7.5inch sparks. So I guess it's a hit or miss. Although driver ICs and GDT cores are from digikey. No more fake parts from now on.

What line voltage are you using for the ramp?  What Vbus cap(s) are on the back side of H-bridge board?  What frequency is your SSTC?  What feedback are you using (circuit diagram)?

1) 120v AC mains so 170pk
2) No Vbus caps because I thought they would sooth out the AC peaks
3) 94kHz resonant frequency
4) Secondary antenna

Other things:
1) On time 1ms - 5ms@ 30Hz
2) I will try to etch my own PCBs. I have actually done it before. Thank you.
3) Also if power is not connected and I just supply oscillations from 555 timer. There is ticking sound (at whatever frequency interrupter is running) from somewhere

[davekni]

Mosfets I used in half bridge were IXTQ52N30P (listing said original disassembled parts) and I was able to get 7.5inch sparks.

That's a higher current part, presuming it was genuine.  Changing to full bridge and to lower current FETs makes for a likely issue.

1) On time 1ms - 5ms@ 30Hz

I'd suggest testing at much lower repetition rate, such as 1Hz instead of 30Hz.  Gives more time for FETs to cool between bursts.

2) I will try to etch my own PCBs. I have actually done it before. Thank you.

Focus on a good ground plane and very short connection from IC (especially driver chip) power pins to bypass capacitors.

3) Also if power is not connected and I just supply oscillations from 555 timer. There is ticking sound (at whatever frequency interrupter is running) from somewhere

That's likely from GDT.  Without coil operating, GDT gets just low-frequency interrupter signal through coupling capacitor from driver to GDT.  Likely no issue.

Schematic shows no capacitor(s) from + to - bus voltage (Vbus+ to Vbus-).  If that is accurate, it is a problem and likely cause of FET failure.

[Simranjit]

Mosfets I used in half bridge were IXTQ52N30P (listing said original disassembled parts) and I was able to get 7.5inch sparks.

That's a higher current part, presuming it was genuine.  Changing to full bridge and to lower current FETs makes for a likely issue.

1) On time 1ms - 5ms@ 30Hz

I'd suggest testing at much lower repetition rate, such as 1Hz instead of 30Hz.  Gives more time for FETs to cool between bursts.

2) I will try to etch my own PCBs. I have actually done it before. Thank you.

Focus on a good ground plane and very short connection from IC (especially driver chip) power pins to bypass capacitors.

3) Also if power is not connected and I just supply oscillations from 555 timer. There is ticking sound (at whatever frequency interrupter is running) from somewhere

That's likely from GDT.  Without coil operating, GDT gets just low-frequency interrupter signal through coupling capacitor from driver to GDT.  Likely no issue.

Schematic shows no capacitor(s) from + to - bus voltage (Vbus+ to Vbus-).  If that is accurate, it is a problem and likely cause of FET failure.

Thank you very much for your great input. I really appreciate it. It's always a delight when you give your input. Although one thing I don't understand is that why we need a Vbus cap ? Wouldn't that smooth out AC ramps ? ( I'm sorry if I'm missing something here I'm using AC mains for my ramps )I didn't use any Vbus cap on half bridge either. I do have some 200v mosfets left (probably fake might as well put them to test).

Edit: Now that I think about it I think it was cross conduction most likely. I did notice for one mosfet rise and fall time was 50ns / 50ns . For the other one it was 50ns / 40ns something like that. I should have paid more attention to it. I don't why I chose to ignore maybe because I thought it didn't matter 10ns. Given that I chose to put one type of mosfet on + side and other one on - side. Should have done like you suggested.

[davekni]

Although one thing I don't understand is that why we need a Vbus cap ? Wouldn't that smooth out AC ramps ? ( I'm sorry if I'm missing something here I'm using AC mains for my ramps )I didn't use any Vbus cap on half bridge either.

Large electrolytic bulk caps as used on normal non-ramped SSTCs and DRSSTCs would smooth out ramp.  Smaller ~0.47uF or 1uF film capacitors directly on H-bridge board are necessary.  They handle ripple from TC primary winding switching frequency (mostly at 2nd harmonic of that frequency).  Analog simulation (LTSpice or other) would be best for understanding this issue (and almost every other design detail).  For your half-bridge, the two caps used to split Vbus were apparently enough.  Usually there is at least one capacitor across Vbus besides the pair splitting Vbus for primary connection opposite FETs of half-bridge.  For full-bridge, at least one Vbus cap is needed.  Several small caps in parallel are even better as that reduces total cap lead inductance.  Mount cap(s) directly on foil between Vbus+ and Vbus-.

I did notice for one mosfet rise and fall time was 50ns / 50ns . For the other one it was 50ns / 40ns something like that. I should have paid more attention to it.

Unlikely enough to be a problem.  Missing Vbus cap is almost certainly the cause of fried FETs.  (For learning, you could replace FETs and scope signals at 15Vbus before adding caps.)

[Simranjit]

Large electrolytic bulk caps as used on normal non-ramped SSTCs and DRSSTCs would smooth out ramp.  Smaller ~0.47uF or 1uF film capacitors directly on H-bridge board are necessary.  They handle ripple from TC primary winding switching frequency (mostly at 2nd harmonic of that frequency).  Analog simulation (LTSpice or other) would be best for understanding this issue (and almost every other design detail).  For your half-bridge, the two caps used to split Vbus were apparently enough.  Usually there is at least one capacitor across Vbus besides the pair splitting Vbus for primary connection opposite FETs of half-bridge.  For full-bridge, at least one Vbus cap is needed.  Several small caps in parallel are even better as that reduces total cap lead inductance.  Mount cap(s) directly on foil between Vbus+ and Vbus-.

Unlikely enough to be a problem.  Missing Vbus cap is almost certainly the cause of fried FETs.  (For learning, you could replace FETs and scope signals at 15Vbus before adding caps.)

Hi, here are the results before and after adding the capacitors. Capacitors definitely helped with the spikes on drain of mosfets. But still there is a lot of parasitic inductance. I think I should work on to make a PCB like you suggested before. Picture are of drain of mosfet ( before adding capacitor, 1uF, 3uF ) and gate of mosfet.  14Vbus 1% duty cycle (0.2ms) @ 1.97Hz. pk-pk voltage over 60v. Mosfets are IRF640. Interesting thing is that in zoomed out waveform of drain of Mosfet the transients disappears after some cycles. Maybe I should also try phase shift secondary feedback a bit with capacitor and resistors like mentioned here:
https://circuitsalad.com/2021/01/30/optimised-half-bridge-tesla-coil-driver/

[davekni]

But still there is a lot of parasitic inductance.

Presuming the back side of your H-bridge board is mostly covered in copper too, for the two bridge outputs, that is already good for low inductance.  Biggest remaining inductance is FET leads.  Forward recovery time of internal FET diodes contributes some to Vds spike voltage.  Some designs add an external fast diode in parallel with each FET drain-source.  However, your remaining very-short spikes are likely not problematic.

Interesting thing is that in zoomed out waveform of drain of Mosfet the transients disappears after some cycles.

Looking at your scope plot, looks to me like spikes are gone only after interrupter pulse ends.  Remaining time is TC primary current ringing down.  You could scope the interrupter signal or a gate to see.

Maybe I should also try phase shift secondary feedback a bit with capacitor and resistors like mentioned here:

No guarantee that it would help spikes.  However, there are other advantages to a self-oscillating circuit like you linked to.  I've posted similar self-oscillating alternatives in various threads.  Helps with startup, especially for ramped coils, as they need to start at low bus voltage.
BTW, looks like you may be using antenna feedback.  Self-oscillating is easy to add to antenna feedback, easier than for secondary current feedback.

[Simranjit]

You were right all along. Capacitor on vbus did solve the mosfet burning out issue. Phase shifting didn't help much. Although I'm using CT for secondary feedback now. I did go over 5% duty cycle. Now the the primary is arcing to secondary  .