IGBT Gate resistors heating up a lot

[breezetix]

So, I decided to switch to a GDT instead of using optocouplers (acnw3190s and such) because of multiple power supplies (at least 8, 4 for +15V or so of a full bridge, other 4 for around -5V to avoid cross conduction). Big issue I have is the gate resistors heating up to like 150c to 300+ celsius, from about 40khz to 100khz. (100khz is not needed I dont think, but I can do at least 50-70kHz, depending if my ferrite cores can handle that much at 400V 20A 5 turns primary) Note this is just for high voltage arcing, so constant switching, no pulses. I also tried higher watts and resistance (connecting them in series-parallel etc, 3W 3 ohms to 6W 3 ohms, 3W 6 ohms, 6W 6 ohms, more resistance vice versa but at 6W is the highest I did yet) I also tried an RC damping circuit but that didn't work either (I think for changing the rise or fall time) I am using 16V as well. (I was also trying 18, 20 and 24V for other testing) Oh yeah not to mention the resistors I used were metal film. Previously I used 5W 3ohm metal film resistors but they got too hot to the point they burned (got up to 350c according to thermal camera) and now they have lower resistance which increases peak current.

So what should I do? I could just use lower frequency and use a little higher resistance and watts, and under 100c may be fine, with some cooling or whatever, but I'd like to at least try and use my 3C80 ferrite cores I bought from Mads, try it at 400V at about 20A which I have still not done yet before, since I already have them and I want to at least "accomplish" it. It is also for a video also, to try to run the bricks at higher frequency, just to gain some experience I guess. Or, I can just use lower frequency or whatever and run the primary at resonance (if it's not too high freq) which can go soft on the bricks, and do soft switching, which I don't know how yet.

[davekni]

Can't offer any specific advice without context, such as what IGBTs you are driving.  If this is from a previous thread, perhaps clearer to continue that thread rather than start a new one.  Either that or include all the basic information here.

Total gate drive power can be calculated fairly easily.  Look at total gate charge for total swing (+16V to -16V).  That charge needs to be provided by 16V GDT input circuit twice per cycle, rising and falling edges.  So, 2 * freq * charge * 16V = power.  That power will be shared between driver chip or FETs and gate resistors.  If resistor is bypassed by diode, then roughly half power in driver and half in resistor.

[Mads Barnkob]

If you have 6W gate resistors burning up, finding a higher wattage rated resistor is not the solution and you problem is not weak resistors. This sounds like something else if giving you problems.

I did make a IGBT gate drive calculator, maybe you find it useful: https://kaizerpowerelectronics.dk/calculators/igbt-gate-drive-calculator/

[davekni]

If you have 6W gate resistors burning up, finding a higher wattage rated resistor is not the solution and you problem is not weak resistors. This sounds like something else if giving you problems.

Probably accurate.  However, that is why I asked for context.  A CM600 brick at 100kHz and +-20Vge requires 23W per gate from driver, 92W total for a full bridge.  If no parallel diodes across gate resistors, then most of that 23W is dissipated in the resistor.  That power is why I had to enable oscillator only during ramps on my low-frequency (100kHz) QCW experiment.  Otherwise I could have left gate drive on between buck converter ramps to bridge power.

[breezetix]

If you have 6W gate resistors burning up, finding a higher wattage rated resistor is not the solution and you problem is not weak resistors. This sounds like something else if giving you problems.

I did make a IGBT gate drive calculator, maybe you find it useful: https://kaizerpowerelectronics.dk/calculators/igbt-gate-drive-calculator/

I resolved the issue by using 20W thick film TO-220 resistors, heatsinked and fan blowing at them. Helped a lot. So yeah it would be around 10W of power, I didn't realize that.