Effect of electrical parameters on conduction/switching losses calculated from thermal model

Question

I was calculating losses for a certain resonant converter converging to a nominal value of switching frequency of operation. I used thermal models available from the manufacturers (Wolfspeed, Dynex) and filled all the thermal parameters applicable to MOSFET+body diode (in primary side) and a secondary side diode (i.e. Rgoff, Rgon for MOSFETs, Tamb for all the three).

I ran two scenarios: one without any electrical parameters filled- Ron for MOSFET/ Vf and Ron for body and secondary diodes vs one with all these parameters filled. I got two different values of switching loss (though a small increase owing to the increase in the switching frequency of the converter because of the apparent shift of the closed loop frequency) and conduction loss (a substantial difference of around 7%). I was given to understand that both the electrical and thermal domain run effectively in parallel (Interpretation of conduction losses - PLECS User Forum (plexim.com)) and those electrical parameters are optional to fill (as per the instructional video for thermal modeling in PLECS website). Which approach is better to calculate losses? And why is there a discrepancy between the two if these run in parallel without any feedback of Rds,on/Vf/Rdiode,on (at least in the original model) affecting the other domain?

Answer 1

You are correct in that the thermal model does not influence the electrical behavior of the circuit.  However, the electrical circuit can impact the thermal behavior, in particular when you have a separate thermal description for the MOSFET and Diode.  By changing the electrical parameters you are influencing which devices are conducting and if there is any current sharing between the two. There also may be small impacts from the actual electrical operating condition of the circuit.

You can verify this by looking at the current through the MOSFET and Diode in your simulation. I suspect there will be a significant difference for the cases you are evaluating. In practice, when the gate is closed nearly all the current will flow through the channel instead of the body diode in the third quadrant. The current distribution will alter the losses as the body diode has higher losses.

The recommended approach that many vendors are moving towards is "integrated" models which provide the correct electrical parameters as part of the thermal description.  This reduces the likelihood of user error.  

In the interim the use should determine the electrical parameters from the datasheet. I'll review the video on the website and see if it can be improved.  If you could point me towards the video and approximate time that would be helpful.

Comments

Thanks for the answer Bryan.
I had a few follow up queries about this:
"By changing the electrical parameters you are influencing which devices are conducting and if there is any current sharing between the two."- The only electrical parameters I changed were: Rds,on (1.6 mOhm) of MOSFET and Vf,body diode= 4.5V, Rdiode,on= 2.5mOhm which should ideally lead to a small voltage drop and hence an increased demand from the input. I don't understand why would they influence which device conducts as this would depend on when a MOSFET is supplied gate voltage and what the instantaneous direction of current is.

"You can verify this by looking at the current through the MOSFET and Diode in your simulation. I suspect there will be a significant difference for the cases you are evaluating. In practice, when the gate is closed nearly all the current will flow through the channel instead of the body diode in the third quadrant."- I also checked the current waveforms with/without these parameters and there is, in fact, not a noticeable difference because as soon as the MOSFET is turned on, all the current from the diode across it commutates to the MOSFET (mainly because of a very high forward voltage drop of body diode) as it should for ZVS.

"The current distribution will alter the losses as the body diode has higher losses."- There is no current distribution happening (before MOSFET turned on and during deadtime- body diode conducting and as soon as MOSFET is turned on, diode current is dropping to zero immediately.

Also, regarding the video: Webinar: Thermal Modeling in PLECS | Plexim (somewhere around 14:34, he mentions that these electrical parameters are optional and then he calculates losses keeping these parameter as 0. There would have been a deviation had he changed these parameters to some non-zero values.)

---

Suyash, at this point it might be helpful for you to post a model so we can have a deeper discussion in the context of your application.  Could you share a representative example including the *.plecs file and thermal descriptions?

Thanks for pointing me towards the relevant portion of the video!