RC snubbers

[AstRii]

Hello!
Lately i've been designing a small mains voltage SSTC for my school.
I ended up using a DRSSTC driver (UD 2.7) because i had one laying around
It works well, the driver gets feedback from secondary via CT and it oscillates at about 250kHz.
I'm using 15R resistors to charge the IGBT gates and 5R6 + diode to discharge them. And i'm still getting pretty big voltage spikes. (The coil is tested at only 30VDC so the spikes may be only from C-E capacitance of the IGBTs.)
But still, i want to design an RC snubber to handle these spikes.

When powered from 30VDC:
Without any RC snubbers the voltage on the low side C-E looks like this:

70V spikes!

With 22R + 4.7n:

around 55V spikes.

And with 1R + 22n:

Pretty much no spikes.

Seems like 4.7n is the bare minimum capacity i should use, but with the 22R resistor it heats up a lot (at only 30V, imagine 325V).
But when i switched to 1R + 22n i was expecting the resistor to burn but it was pretty much the same temp. as the 4.7n + 22R snubber.
I read some papers about the power loss on the snubber resistor and they all concluded the same thing: P = 1/2 * C * U^2 * f.
But i don't think this is correct.
Consider Danyk's (DiodeGoneWild) SSTC:

He's using 4.7n and 4.7R at full mains voltage (325V) @100kHz. The formula would predict power loss to be more than 20W, yet it's going fine with one 3W resistor.
Can anyone explain this?

I'm yet to design a suitable RC snubber, i think i cannot go much lower than C=4.7n to reduce the spikes but for that i would need 30kOhms to reduce the power to few Watts (at least according to the P=1/2CU^2f formula).
I have already tried this 4.7n and 30k RC snubber, needless to say.. it didn't do anything

[T3sl4co1l]

You need C >= 2.5 Coss and R = sqrt(Lstray / Coss).

Whatever power that dissipates, is whatever it does.

You can also use a RCD clamp or rate snubber, which might dissipate less power, or be more or less sensitive to load conditions?

Evidently your loop has 116nH in it (a resonant frequency of ~3.3MHz with ~20nF loading, and 1 ohm being less than sqrt(L/C)), which is awfully high.  Consider redoing the layout.

Which also puts sqrt(L/C) ~ 5 ohms, so the 4.7 ohms suggested is likely close to optimal.  The peak voltage will be lower for smaller R, but there will be more ringing.  Coss is a few nF for most IGBTs of useful size; it will decrease at higher voltages.  C around 4.7 or 10nF seems good enough.

Tim

[davekni]

As Tim said, lowering inductance to the H-Bridge is the way to save power.  When you use snubbing with low R and high C, power dissipation is moving into the IGBTs, unless you have sufficient phase-lead and dead-time for resonant H-Bridge output transitions.

[Mads Barnkob]

We have talked about low voltage testing with you before

Look at this old topic about your DRSSTC II: https://highvoltageforum.net/index.php?topic=1122.msg8060#msg8060

Low voltage testing is deceiving! Look at the IGBT datasheet for the capacitance vs Vce graph and you can see that Coes is higher at 30VDC, and lower at 100VDC, Coes is the output capacitance of the IGBT module and this will give you extremely high switching transients when looking at the inverter output voltage. So these switching transients will get lower at higher C-E voltages!

[AstRii]

We have talked about low voltage testing with you before

Yes you're right, more than once
But in this case i don't think it's the C-E capacitance causing the spikes.

My layout looks like this:







I will twist the wires to the primary, that should help me decrease the inductance.

When i'll do some higher voltage tests, i will upload them here:)

Maybe there's not even need for an RC snubber at all, i was only interested in how would someone calculate accurately what's the power dissipation on the snubber resistors, because i don't think it's quadratically proportional to the input voltage as i've seen in some papers about RC snubbers.

[davekni]

If it's VBus inductance as Tim and I are guessing, scope VBus (high-side collectors).  That depends on layout of the ECB.  If it's like your previous project with traces rather than overlapping planes, then VBus inductance will be high.

If that spike does show up on VBus as I'm guessing, it also implies significant IGBT current at the switching point.  The initial positive spike implies too much phase lead, with IGBTs turning off while still conducting high current.

As Mads pointed out, IGBT capacitances change drastically with voltage.  I agree, this doesn't strike me as primarily due to capacitance.  IGBT capacitance can affect the overshoot shape.  In my experience, it doesn't cause the overshoot waveforms you are showing.

Yes, if nothing else changes, then snubber power goes as VBus^2.  That's because energy in the snubber capacitors is 0.5*C*V^2.

[T3sl4co1l]

Show board layout.

Also, that looks to be a #26 powdered iron core in the GDT, don't expect a great frequency range from it.  It should be hi-mu ferrite, for least loading of the drivers and extended LF range.

Tim