Efficiency calculation: test conditions and final considerations

Question

Hello everyone!
My question is related to this thread.
I didn't quite understand whether the last comment (by Kris) actually confirms the assumptions made about calculating efficiency with a thermal model.

Is it correct to include Psw at the denominator as the user suggests?

I noticed this problem also in my simulations and you don't notice it until you calculate low loads efficiency but in general, it should slightly change the efficiency also at higher load in some conditions.

For example, I attached a modified version of the DAB demo model. I decreased by a factor of 100 all the MOSFET and diodes resistances (and Co_esr) to not consider them in the efficiency calculation (Pohm=0) and to get the same power at the input and output.. Let's say I only want to consider switching and conduction losses from the thermal model to calculate the efficiency in two ways:

- eff=1-Psw/Pin
- eff_real=1-Psw/(Pin+Pws)

If the output resistance is set to 1kW at a nominal voltage (Vout=380V), the two efficiencies are almost the same also for lower power.
Let's say now that the DAB is in a voltage mismatch condition between primary and secondary DAB sides (Vout=200V) so we have an increase of the reactive power and the RMS current.
If the resistor is set to 1kW, efficiency does not change too much (more or less 0.2%) and in this condition, we have more or less Psw=40W (only MOSFET losses without diode losses but If I included diode losses, Psw would get worse.).

Now, if I keep the voltage mismatch condition (Vout=200V) but I reduce the output resistor power to 20W, here we have eff<0 and eff_r=32%. I know that 20W is ridiculous compared to the nominal power of the converter but this test is only to stress out this behavior and try to understand the "real" efficiency formula. In general, the efficiency difference increase with the voltage mismatch increasing.

Are my considerations correct? Am I being too fussy? Am I overlooking something?
Actually, since the Psw are not supplied by Pin because it's the electrical model, they should always be included at the denominator, right?

Thank you for your help!

Bye

Answer 1

Nikilito,

The assumption that the semiconductor device losses are much smaller than the processed power is fundamental to the lookup table based approximation PLECS uses for thermal modeling.  That is, Psemi << Pin. This is the goal with most power conversion systems and so the simplified formula of 1-Psemi/Pin in the documentation is a very good approximation of the efficiency in most cases.  The formula is simple and easy to understand.

Thermal models should only be applied under the conditions outlined above.  Variations on the efficiency formula can provide additional insight, but may not be accurate as the different formulations impact the estimated efficiency only when the above conditions are not met.

Regards,

Bryan

Comments

Thank you very much for your reply!
This is exactly what I expected...

Bye!