Hello. I have simulated a synchronous boost converter with thermal model for mosfets AIMDQ75R011M2H (attached). I noticed that I have switching losses for the synchronous mosfet (in model: M2) only when it turns on. The turn-off switching losses are zero. I searched a bit in Plecs forum and followed your advice to mirror the look-up tables for negative current values, but I get the same results ( Switching losses - MOSFET, negative current ). Also, if I add dead-time (1-2% of Tsw), switching losses do not change either.
In forum, I also found that:
“For turn-on loss, PLECS looks at the voltage directly before the switching event and the current directly after. For turn-off loss, PLECS looks at the voltage directly after the switching event and the current directly before.”
Is it possible for me to check which values Plecs takes into account around switching, in order to calculate the losses?
Is there anything else that I should take into account for calculating switching losses?
first of all, I want to point out that the thermal description is not optimally defined for switching losses. In particular, it is important that the switching losses at 0 A and 0 V are explicitly defined as 0 J in order to avoid incorrect interpolation. I have added this condition to the thermal descriptions.
“For turn-on loss, PLECS looks at the voltage directly before the switching event and the current directly after. For turn-off loss, PLECS looks at the voltage directly after the switching event and the current directly before.”
This behavior can best be verified by probing the MOSFET M2 and inspecting its signals. In the added Scope6, you can see all the relevant waveforms required to pinpoint the loss calculation.
In this specific case, the situation is a bit tricky because the thermal description not only includes a lookup table, but also an additional equation. If you change the Computation method from Lookup table and formula to Lookup table, you will notice that no switching losses occur. The reason is that only negative currents are present, and therefore no switching losses are generated. In this operating condition, the body diode starts conducting and the MOSFET channel can turn on and off under ZVS conditions. You can verify this by disconnecting the gate signal, only the diode will then conduct. While conduction losses increase significantly, there are still no switching losses.
If you switch the Computation method back to Lookup table and formula, you will see that turn-off losses become non-zero. This is caused by the additional term +lookup('Eoss', v, T) that is added to the Turn-on loss formula. For Turn-off loss, the same term is subtracted, but the overall formulation ensures that no negative loss values occur.
The explanation for this additional term can be found in the Comment section provided by Infineon:
Since electrical measurements overestimate the Eoff losses,
we correct this behavior by transferring the Eoss-related
energy to the respective turn-on energies.
Qrr is not modelled explicitly – it is implicitly included in the energy values.
For this application, it may therefore make more sense to explicitly model the reverse-recovery losses of the body diode, as explained in this forum post: