Choosing snubber capacitance

[Caraffa]

Hi, I'm building my first DRSSTC and I've encountered a problem with choosing snubber capacitor, mainly because the waveform looks a bit strange. I don't have much more than 1uF capacitors and any package will take a long time to arrive so I want to be sure before I order anything.

Back to the topic, I took some screenshots with different subbers. I think i should choose something bigger than 3uF but i'm not sure why square-wave has such sharp end. What do you guys think? 

[Caraffa]

I managed to solder another capacitor, so in total 4uF, but now it looks like too much? I'm only guessing. Maybe there is another problem that i didn't consider.

[davekni]

Back to the topic, I took some screenshots with different subbers. I think i should choose something bigger than 3uF but i'm not sure why square-wave has such sharp end. What do you guys think?

It would help to define what you are measuring.  Presumably a DRSSTC given the post category.  Given the waveform shape, perhaps you are scoping H-bridge output, perhaps differential between the two outputs, rather than Vge of an IGBT.
If this is differential across H-bridge output, are you using a differential probe, or a grounded probe and floating Vbus supply to H-bridge, or a floating scope chassis?  Scope captures of individual H-bridge outputs with probe ground to Vbus- are typically more useful.
If all my above guesses are correct, looks like you are getting resonance of snubber capacitance with lead (or bus bar) inductance to bulk caps.  A picture will help show what inductance might be there.  Pictures showing scope probe connections also help, as some artifacts can be due to inductive pickup in the loop between probe tip and probe ground lead.

[Mads Barnkob]

Did you try a few simple calculations to this problems? https://kaizerpowerelectronics.dk/tesla-coils/drsstc-design-guide/snubber-capacitor/

[Caraffa]

It would help to define what you are measuring.

Sorry! I totally forgot. Yes you guessed everything correctly. It's differential between the two outputs of H-bridge.

If this is differential across H-bridge output, are you using a differential probe, or a grounded probe and floating Vbus supply to H-bridge, or a floating scope chassis?

It's grounded probe and floating Vbus.

A picture will help show what inductance might be there.

It's TO-3PN H-bridge so there is a little bit of cables that are maybe sub-optimal. I probably won't leave it as it is now, I wanted to build it mainly for testing and when I find more time I will design something better than a mdf board and a couple of screws. But it is unlikely to come soon.

Did you try a few simple calculations to this problems? https://kaizerpowerelectronics.dk/tesla-coils/drsstc-design-guide/snubber-capacitor/

These calculations are mainly based on stray inductance, and I don't even know how to estimate it because I don't have much experience in this subject.

[Mads Barnkob]

Did you try a few simple calculations to this problems? https://kaizerpowerelectronics.dk/tesla-coils/drsstc-design-guide/snubber-capacitor/

These calculations are mainly based on stray inductance, and I don't even know how to estimate it because I don't have much experience in this subject.

The estimates are in another article, there is a link between them at the start of the Snubber article

https://kaizerpowerelectronics.dk/tesla-coils/drsstc-design-guide/busbar-and-primary-circuit/

[Caraffa]

The estimates are in another article, there is a link between them at the start of the Snubber article

Thanks, according to those formulas and my calculations 1,2-1,5uF should be enough. Do you think there will be much of a difference between 1,5uF and 1/2uF that i tried?

[Mads Barnkob]

The estimates are in another article, there is a link between them at the start of the Snubber article

Thanks, according to those formulas and my calculations 1,2-1,5uF should be enough. Do you think there will be much of a difference between 1,5uF and 1/2uF that i tried?

I also made a snubber capacitor calculator, you can play around with the numbers to get 1.5uF and 0.5uF, by adjusting the allowed transient voltage, that way you can judge if you find that transient level acceptable.

https://kaizerpowerelectronics.dk/calculators/snubber-capacitor-calculator/

[davekni]

Thanks, according to those formulas and my calculations 1,2-1,5uF should be enough. Do you think there will be much of a difference between 1,5uF and 1/2uF that i tried?

Given your earlier scope traces, I'd suggest using at least 2uF.  Standard snubber design guides such as Mads' well presented one are based on an assumption that snubber ring has decayed before the next H-bridge output transition.  That is typically the case in commercial uses of IGBTs, generally at much lower frequency than DRSSTC use.
When snubber ring has not decayed, resonance can make the ring amplitude build each cycle.  That is what shows up in your scope plots especially at 1uF.  There are two or three other threads on this forum where designs have had the bad luck of snubber resonant frequency hitting very close to 2x operating frequency.  That's the worst case, as Vbus ripple current is dominated by 2nd harmonic of operating frequency for full H-Bridges.

[Mads Barnkob]

Thanks, according to those formulas and my calculations 1,2-1,5uF should be enough. Do you think there will be much of a difference between 1,5uF and 1/2uF that i tried?

Given your earlier scope traces, I'd suggest using at least 2uF.  Standard snubber design guides such as Mads' well presented one are based on an assumption that snubber ring has decayed before the next H-bridge output transition.  That is typically the case in commercial uses of IGBTs, generally at much lower frequency than DRSSTC use.
When snubber ring has not decayed, resonance can make the ring amplitude build each cycle.  That is what shows up in your scope plots especially at 1uF.  There are two or three other threads on this forum where designs have had the bad luck of snubber resonant frequency hitting very close to 2x operating frequency.  That's the worst case, as Vbus ripple current is dominated by 2nd harmonic of operating frequency for full H-Bridges.

This sounds like something I just add to the guide

[davekni]

This sounds like something I just add to the guide

That would be a great addition to your well written tutorial.  I expect you'll write a clearer and more general explanation than what I ramble on about.

[Mads Barnkob]

This sounds like something I just add to the guide

That would be a great addition to your well written tutorial.  I expect you'll write a clearer and more general explanation than what I ramble on about.

I can see my wording is confusing, I meant to say NEED to add to the guide, not JUST add to the guide

Something like with SGTCs NSTs to take into account for LTR (larger than resonant, to avoid destroying transformer) in the calculation or calculator.

[Caraffa]

So... for now i chose 3uF snubber (i couldn't fit more on the board). It worked good enough for few minutes but then half of my bridge died. It was quiet death so I assume it was caused by overvoltage (IGBTS were cold). I will say right away that there is a chance that it was caused by interference on the gates coming from the coil, but I decided to buy some capacitors for more testing and I think i found another potential issue.

While (i assume) primary discharges itself it's causing initial voltage rise and then relatively high voltage spikes. Adding snubber capacitance helps but it feels more like workaround and not actual solution. Is there something that I can do?

[davekni]

So... for now i chose 3uF snubber (i couldn't fit more on the board). It worked good enough for few minutes but then half of my bridge died. It was quiet death so I assume it was caused by overvoltage (IGBTS were cold).

Which pins of IGBTs are shorted or open?  I think over-voltage is likely to cause a CE short, usually resulting in a bang.

While (i assume) primary discharges itself it's causing initial voltage rise and then relatively high voltage spikes. Adding snubber capacitance helps but it feels more like workaround and not actual solution.

First, I presume final scope capture is Vce, not Vge.
You are seeing the sum of two effects.  The short spike is caused by IGBT package pin inductance and inductance of snubber caps and wiring from snubber caps to IGBTs.  Adding more snubber capacitance has minimal affect on this short spike.  Reduces snubber lead inductance with more caps in parallel, but physically larger caps often have more inductance per cap.  The larger ring is caused by the same resonance of snubber capacitance with wire inductance from snubber to bulk caps.  Increasing snubber capacitance reduces amplitude of this larger ring, roughly proportional to 1/sqrt(capacitance).
I doubt either is enough to be the cause of your IGBT failure.  Unless your bus voltage is close to IGBT Vce rating.  I'm guessing you are using 600 or 650V IGBTs on ~340Vdc bus voltage.  That should be enough margin.  Larger slow ring is proportional to primary current, not to bus voltage.  Short spike is usually less than proportional to current, as IGBT switching slows down at higher current.

[Caraffa]

Which pins of IGBTs are shorted or open?  I think over-voltage is likely to cause a CE short, usually resulting in a bang.

CE are shorted, but there was no bang and IGBTs looked fine, no cracks or anything. If i recall correctly it was running at roughly 80-100V AC.

First, I presume final scope capture is Vce, not Vge.

Yes, sorry my bad.

I doubt either is enough to be the cause of your IGBT failure.  Unless your bus voltage is close to IGBT Vce rating.  I'm guessing you are using 600 or 650V IGBTs on ~340Vdc bus voltage.  That should be enough margin.  Larger slow ring is proportional to primary current, not to bus voltage.  Short spike is usually less than proportional to current, as IGBT switching slows down at higher current.

Yes you are right, I'm using FGA40N65SMD on ~320V DC bus. If these spikes were not the cause of failure then I don't know what could have been.

[davekni]

CE are shorted, but there was no bang and IGBTs looked fine, no cracks or anything. If i recall correctly it was running at roughly 80-100V AC.

Interesting.  I presume both IGBTs of one side (one half-bridge) are shorted.  Is that accurate?  Is either gate shorted?  What peak primary current are you running (what is OCD set to)?

If these spikes were not the cause of failure then I don't know what could have been.

There are many possible causes for IGBT failure, such as insufficient dead time or excess peak current.  Looking back at your build pictures, a few specific possibilities come to mind:
1) Looks like there are TVS diodes across IGBT CE.  Sometimes those fry (shorted) first, which then fries IGBTs.  I'd check for both shorted or open TVS diodes.  (Open because sometimes devices fail shorted initially, then short current burns bond wires open.)
2) GDT output connections to IGBTs are best connected to gate and emitter leads close to IGBT body.  Connecting GDT emitter leads to the circuit board instead adds emitter lead inductance switching spikes to internal gate voltage.
3) Mounting devices by their screw hole can lead to poor thermal contact.  Often compresses pad under screw, which angles IGBT slightly up at lead end, with little pressure to pad under IGBT die.  I'd suggest something like this example to apply pressure to IGBT packages directly above IGBT die area:
https://highvoltageforum.net/index.php?topic=2191.msg16147#msg16147
Even though you say IGBTs were "cold", unless measuring temperature during operation, they can heat up and cool down quickly.

Hope your debug and repair goes well!