I’m trying to model the thermal losses for a NPC leg with 4.5 kV IGBT and IGCT from Hitachi.
As you can see in the model attached, I modified the plecs library NPC block to add the possibility to select IGCTs.
I first tested my model with the thermal chain setup in the .xml file and the case to heatsink thermal resistance from datasheets setup in the thermal interface resistance parameter for each semiconductor. It worked well (i.e. I had the same results than with analytical expressions) except that I had to simulate over a long time to get steady-state temperature (because of the RC time constant). The steady-state analysis tool also worked well.
All the semiconductors blocks were put in a common ambient heatsink.
I tried to modify the thermal model in order to also model the heatsink to ambient thermal resistance Rthha. As you can see I tried to add heatsink for each semiconductors and connect them to the thermal resistance Rthha and then to the ambient block.
But the results I have now are not compliant with what I calculate. The juction temperature for IGBTs is too high. Also the Steady-state tool cannot converge to a solution.
Have I done something wrong in my simulation or is this normal that the thermal transient is so long ?
Thanks for your help and feel free to ask if you need any more informations
Have I done something wrong in my simulation or is this normal that the thermal transient is so long ?
Although the thermal time constants in your device models appear to be ~2 seconds, there’s a slower feedback that can occur. The higher junction temperature induces higher losses, which in turn induces higher temperature (that takes a bit of time to increase) which in turn leads to higher losses and so on. That’s the benefit of the steady-state analysis tool!
But the results I have now are not compliant with what I calculate. The juction temperature for IGBTs is too high. Also the Steady-state tool cannot converge to a solution.
Regarding your loss calculations, it could be that your simpler model may not be calculating thermal runaway effects where the temperature keeps on increasing. You can try and operate at a lower current level or smaller Rthha and see if the solution converges. Your component model is also quite complex, so it’s possible you’re doing some double accounting of losses within the model.
I tried to remove the thermal time constant by removing the foster chain in the .xml and putting a simple thermal resistance to simulate a steady state Rthjc. I also put a 1 J/K thermal capacitance in the heatsink enclosing each semiconductors to avoid simulation errors.
When I do that the junction temperature for each semiconductors except clamping diodes skyrocket to 10^15 °C 10^16 °C… Even if I try to reduce the load current (althought my semiconductors should handle with a 1000 A load) I got the same result.
I found something weird : I put a thermometer in my lowest subsystem connected to the ambient block :
As you can see the temperature is unrealistic whereas it should be at 50 °C because it is connected throught the ambient block to the 50 °C constant temperature source outside the subsystems.
I feel I’m not using correctly the ambient block. Am I wrong ?
Hello,
I replaced the ambient block with a thermal port to propagate the ambient temperature inside my subsystem and now it works perfectly. I got the same results as my calculation.
I don’t know why but the ambient block did not propagate the temperature inside the 3rd layer of subsystems.
