Different results between Foster and Cauer thermal networks

Hello,

while simulating a three-phase inverter I noticed a misalignment between the temperatures obtained using an explicit Foster network and an explicit Cauer network. Could you explain to me why this happens?

Power circuit:

If I define the thermal network in MOSFET thermal description, the result in terms of junction temperature is the same whether I define the network as a Foster or a Cauer model (obtained by simply converting Foster into Cauer).

I find almost the same result defining the explicit thermal Network using this circuit:

where I’m probing conduction and switching losses of the same switch and applying average over the switching period.

However, if I replace the thermal chain with RC parallels to build the Foster model, the result is completely different.

I would really understand why there’s this big difference considering that network parameters are derived from the same Zth…

Any suggestion?

Thank you

image
Hello
The “capacitors” are grounded in thermal chain.
They are parallel to the “resistors” in your version. Could this be the reason?
By the way, please tell me how you calculate R and C in thermal chain. Thank you.

Hello,

I performed fitting of Zth to find Foster thermal model parameters (R and C). I used them to define the explicit thermal network with separete Rth and Cth.
Then I used the function “convert to Cauer” in thermal description to obtain parameters to be used in thermal chain block.

Thanks

When connected to additional RC elements one would expect the Foster and Cauer to deviate, but when connected to a fixed temperature the results should align.

Since the average values look around the same, but with different transient peaks, I suspect there may be an error in your implementation. Note that the Foster Coefficients aren’t R & C, but R and Tau. Are you converting from Tau to C when entering the capacitance values in your Foster circuit?

See the attached example.

foster_vs_cauer.plecs (29.6 KB)

If they still deviate, could you post your model and step-by-step instructions to recreate your results?

Hello Bryan,

Yes I converted tau in C using C = tau/R for each RC group. It’s very strange to me

Ok, thanks for the update. Without your model I cannot provide more specific guidance.