Calculating Hard-Switching losses

[Phoenix]

Hello 

I would like to calculate the maximum frequency i can drive an IGBT-Brick within its thermal limitation. I am using SKM200GB128D Bricks.

I am calculating the losses for 180A Hardswitching and a frequency of 20kHz. I need to find out the turn on/off energy first using the datasheet.


And here is my question: When reading the energy of the chart, do i need to use the 180A of current or the change of current (dI/dt)? The change of current would only be around 22.6A/µs high for 180A at 20kHz, which would result in much lower turn on/off energies.

Greetings,
Phoenix

[Mads Barnkob]

It is your peak current, that is how I calculated it for soft-switching here: http://kaizerpowerelectronics.dk/tesla-coils/drsstc-design-guide/igbts/

I guess you can use my example all the way through, just ignore the reduction factors for resonant switching that I use

[Phoenix]

Thank you for your reply 

I have already followed your guide for the calculations but i am stuck at that point. In this application i am only switching  an inductive load, which results in a triangular waveform. The IGBT's always switch at the peaks of this waveform (180A). There is no resonance circuit, so the IGBT's always switch at full current. Because of that, i think i have to read the chart at 180A of current. But if i read the chart at the operating current, i can't keep up with the manufacturers claim of 220A at 20kHz with a junction temperature of 150°C and a case temperature of 80°C. Here are my calculations:


As you see i am only able to switch at a maximum frequency of 8.462kHz but according to the datasheet the bricks should be able to switch 20kHz at 220A.

Is there something wrong with my calculations?

Edit: I just tried something new and i think i have found the problem: A square wave has a RMS-Factor of 1, but a triangular wave only has a factor of 0,577. The maximum current rating in the datasheet is for DC, so there are no switching losses included. If i put in the numbers for 127A (220*0,577) into the equations, the result is 23,462 kHz and it should work for my inverter. Could this result be correct?

Greetings,
Phoenix

[Mads Barnkob]

I was just about to mention that, just like I had to reduce for soft-switching short pulses, as hard-switching is seen as a square wave, a triangle waveform would also have to use a reduction factor, but you already figured that out

[Phoenix]

I have now used the reduction factor for my 180A which gives me about 104A RMS. I used the RMS value for the CE Voltage and in the calculation for the conduction losses. But i have not used it for the turn on/off energy, because the IGBT still has to switch the 180A peak.

According to my calculations, if the junction has a temperature of 125°C and the Case has a temperature of 50°C, i can run my Inverter up to 20,6kHz. I also converted the equation to tell me the temperature difference between junction and case, at 20kHz the difference would be 73°C. I could now use a circuit which cuts of the supply voltage to the inverter using a relay, when the case temperature reachs 57°C, so the junction would never get hotter than 130°C 

Greetings,
Phoenix

[rikkitikkitavi]

Just a thought, wouldnt average value instead of RMS be more accurate?

The loss is the integral of the current*voltage across the switching device during the switching event.

[Phoenix]

I only used the RMS Value for the calculation of the conduction losses during the ON-State.

For the losses during the switching event, i used the full 180A, because the IGBT's switch at the current maximum.

[Hydron]

Conduction losses for an IGBT should probably be done with the average (mean) current rather than RMS - they behave more like a diode than a resistor in forward conduction. RMS is appropriate for MOSFETs though, and it probably won't make a lot of difference anyway.