Inconsistent Efficiency Results Using Switch Loss Calculator in SiC MOSFET AC/DC Converter

Hello everyone,

I am relatively new to PLECS and have been working on designing an AC/DC converter using SiC MOSFETs. To estimate the switching and conduction losses, I am utilizing the Switch Loss Calculator tool.

However, I’ve encountered an issue with the efficiency calculations that I’m hoping someone can help me with. When I calculate the power at the input and output sides separately, the efficiency behavior appears as expected: efficiency is lower at low power, increases with power, and then eventually decreases again at higher power levels.

On the other hand, when I calculate efficiency using the results from the Switch Loss Calculator, the behavior is quite different and unexpected. The efficiency starts off high at low power and then decreases as the power increases.

Has anyone else experienced this issue, or can anyone suggest what might be causing this discrepancy? Any insights or suggestions would be greatly appreciated!

Hello, please see this separate thread (https://forum.plexim.com/4117/calculate-efficiency-converter-including-thermal-resistive?show=4117#q4117) and let me know if it answers your question.

In PLECS, the thermal domain is parallel to the electrical domain- it doesn’t directly affect the electrical waveforms and ultimately the efficiency. Therefore, you should use 1-(Ploss(thermal)/Pin(electrical) for a more accurate calculation, rather than Pout(electrical)/Pin(electrical), where Pout(electrical) only assumes optional user-specified electrical domain based losses from resistive elements such as the on-resistance parameter of the MOSFET block.

I assume this fixed-value on resistance can lead to a much different loss at low vs. high currents and this is likely why you see the higher efficiency at low power vs. at high power/high current.

" I assume this fixed-value on resistance can lead to a much different loss at low vs. high currents and this is likely why you see the higher efficiency at low power vs. at high power/high current."

How can solve this issue. Resistance value is fixed.

Generally, using the electrical resistance value is to calculate conduction loss when you aren’t using the thermal modeling domain with temperature-dependent IV curves to define the losses. So if you are going to use the thermal model for the device, I would typically ignore the electrical value, or if you want to keep it you should then manually tune it to a realistic value based on the steady-state operating point. But to be clear you also can’t change the fact that I^2*R losses will increase as the current through the device increases…

The issue is that with fixed resistance values, the system shows high efficiency at low power levels. At the same power When we manually reduce the RDS(on)​, the losses decrease, leading to higher system efficiency.

I can share my simualtion file on your email to varify it.

We are using a reference design from Wolfspeed, and according to page 39, the efficiency is low at 2 kW, then increases to a peak before decreasing again. I am incorporating the same Wolfspeed SiC MOSFET used in the reference design to make the results more realistic.

Wolfspeed_PRD-02282_CRD-22AD12N_22kW_Bi-Directional_Active_Front_End_User_Guide (3)_compressed_compressed.pdf (1.31 MB)

Hi Muhammed, what’s important though is how you are calculating efficiency. To incorporate switching loss and temperature dependency, you should use the formula I provided above. We would argue that adjusting the electrical value for Rdson has little significant effect on the overall system-level performance and negligible influence on the waveforms. Only if you are worried about current sharing, e.g. when adding an external anti-parallel Schottky to suppress the body diode losses during dead time, do the electrical values influence the circuit behavior to much effect.

Can you confirm if you are reporting on your efficiency findings using our approach or by calculating Pout/Pin (all electrical just by probing the source/load)?