Maximum switching speed estimation

[TMaxElectronics]

So I finally had some time to look into the speeds of the IGBT bricks I have around (SKM300GB126DH2) so I can estimate the maximum switching frequency if I was to build a coil with them. I've been reading mads's page on the topic (https://kaizerpowerelectronics.dk/tesla-coils/drsstc-design-guide/igbts/) but the formulas given seem to give me erroneous values even at very low primary currents (something like 5kHz fmax2 at 50A). So my question is what have I been doing wrong? The values I used are 1:1 identical with those given in the datasheet.

It also seems to me like there might be an error in the last formula (the one for Fmax2). Doesn't it assume CW switching? It takes into account that conduction losses only occour during the on time, but not the switching losses
Add to that the fact that the die's thermal capacitance would allow for higher momentary power dissipation during the OT, than it would during CW operation. Calculating the thermal capacitance from that would probably be too much work, so I'm wondering if it might be a reasonable enough approximation to create an equivalent thermal impedance, that indicates the maximum temperature seen during a pulse in relation to the peak power. Kind of like this:

That would assume that the maximum die temperature is Tavg+Tpeak (average temperature + temperature rise during the pulse), ignoring the temperature drop below Tavg at the end of the offtime. That would be a slight overestimation and thus include a safety margin.

I also think it might be a good approximation of switching losses at lower currents to linearly scale them back with the ration between switching current and the current at which the losses given in the datasheet are specified.

When combining those two modifications to the formulas, I get a much more reasonable approximation of maximum switching frequency. At least they are more sensible when comparing the theoretical results with actual results I get for power dissipation in my large coil with SKM400 IGBTs.

So with the IGBTs I have I could go to 100kHz at 10% dutycycle, assuming 200us ontime, 800A Ip_peak and 50°C heatsink / 90° peak die temp. Does that sound reasonable?

[Uspring]

I think Mads equation for Fmax2 is ok. One has to observe though, that

a) the transistor is on only half of the time during the burst. That concerns Pcond.

b) the transient thermal impedance is used for RthetaJC.

[Mads Barnkob]

It is like being forced to look at your old code, looking at a page like that... having to read it all again... AND understand my own train of thought at the time You can quickly get me in doubt

Could you please provide the extrapolated numbers for your SKM300, that results in that very low Fmax2 estimation. Something must be off.

The article is one long assumption and derating of losses according to those assumptions. Its not a theoretical master piece, my math skills are simply not good enough for that, but its my best bet of a how-to-only-abuse-a-igbt-brick-just-enough-to-live-as-long-as-you-find-your-tesla-coil-interesting

I think Mads equation for Fmax2 is ok. One has to observe though, that

a) the transistor is on only half of the time during the burst. That concerns Pcond.

b) the transient thermal impedance is used for RthetaJC.

I did not even include the difference between half-bridge and full-bridge. So the article for now really is only true for the half-bridge. But just before calculating the Pcond losses, its the halving of the previously 1/4'ing of a factor of faster soft switching that could be factor 2-10. So its all really +/- 3dB to be honest