I am modeling the thermal behaviour of the FS820R08A6P2B HybridPACK Drive Module. Right now I am considering the heatup of the IGBT. I worked from some datasheet values and measurements to develop a model that handles steady-state pretty well, but it does a poor job with transient cases.
I read several application notes, and learned that the resistances and time constants in the transient thermal impedance curve in the datasheet represent the thermal resistances and capacitances in a Foster Network:
I transformed this Foster Network into a Cauer Network, and compared the measurements: it seems like the third node of the Cauer Network is pretty close to the NTC temperature, which I reason should be similar to the base plate temperature.
I have read dozens of application notes, and the closest thing I could find was this:
"The individual RC elements can be assigned to the individual layers of the module (chip, chip solder, substrate,
substrate solder, and base plate)." (AN2015-10: Transient thermal measurements and thermal equivalent circuit models)
But from this perspective, the model doesn't match very well with the measurements.
Could someone please confirm or clarify this? Am I making a mistake somewhere in my reasoning or modeling?
Thank you very much
Solved! Go to Solution.
Thanks for posting on the Infineon community page.
Your understanding was correct and the application note also says the same.
Regarding your question: But from this perspective, the model doesn't match very well with the measurements.
I would like to know what is the basis of the comparison and how you are measuring?
Hi Akhil Kumar,
Thank you very much for your reply 🙂
I will share my results, explain the model behind them, and then I hope my question will be more clear.
I model the IGBT losses based on the suggestions in this application note:
Application Note, V 1. 1 , January 2009 -- IGBT Power Losses Calculation Using the Data-Sheet Parameters
The switching and conduction losses are calculated and passed to the Cauer model. I look at each state variable here, and compare it to the temperatures measured from the module's NTC(s). Here are the results with the measured temperature, and the temperature of each state, called Cauer Nodes 1 to 4 (all temperatures in Kelvin):
Nodes 1 and 2, I guess the junction and "chip solder", of course don't have strong enough thermal coupling with the NTC. But nodes 3 or 4 are getting close to the NTC:
There will be some error in the loss calculations, and for now I am ignoring the diode loss/temperature contributions. But right now, to me it seems like node 3 most closely reflects the transient behavior of the IGBT when it gets a current impulse. There is some steady state error that starts to accumulate, but I thought maybe this is because the "normal operating" thermal impedance model should be used, not the transient thermal impedance. But if node 3 corresponds with the substrate solder and not the baseplate, then should that also mean that it is not a good reflection of the NTC temperature?
I am sure I am making many mistakes, this is my first time doing a project like this, so my apologies in advance. And thanks again very much.
NTCs can give a steady-state temperature for the device and not for the transient state. NTCs are placed near the hot spot temperature of the module and it can be considered as a temperature close to the junction temperature of the module.
Can you please let me know the reason for doing the above analysis?
Thanks for your reply and for that information. Am I right in saying that the NTC reading during a current impulse is not easily predicted? We can say that the temperatures at different nodes of the module can be modeled well with this Cauer model, but the NTC is not so easy to predict?
I am modeling the temperature reading of the NTC.
Thank you very much