How are switching and conduction losses calculated by the model if the waveforms are ideal?

Question

Hello everyone,

How does PLECS thermal look-up models calculate losses? As the MOSFETs/IGBTs are switched with ideal waveforms, the effect of the output capacitance on the switching is not seen. Also the current waveforms are quite ideal in these cases.

 

My guess is that the thermal model basically checks the current and voltage at switching instant and calculates the losses based on the loop-up table. In that cases, how accurate are the losses calculated? Should they be strictly used for estimating purposes?

Thank you.

Answer 1

Hello,

This is a great question. Basically, the losses can be very accurate _if_ you have accurate data, especially when provided from datasheets or the manufacturers directly (see some available device thermal loss models here: https://www.plexim.com/download/thermal_models). The effects of output capacitance can be built into the data, but essentially, PLECS linearly interpolates/extrapolates from the two nearest datapoints for voltage, current, and junction temperature, with the ability to add other dependencies such as gate resistance. Of course, you can also use the tool to provide a rough estimate if you do not have highly granular data, or data for the very specific operating point you may need to study. But generally the goal when the manufacturers benchmark their thermal loss models for PLECS is to be within a few percentage points of accuracy, which is anyway the best you can expect from a model, idealized or physical in nature. I hope this helps!

Kris

Comments

Hello Kris,

Thank you for your answer. I have accurate data from the manufacturers. So based on the current/votlage values during switching, I have the right losses.

My question now is that if my waveforms are ideal, without dead-time and output capacitance variation, the shape of current waveform and its value at the switching instant will not be the same as the intra/extrapolated value. In my simulation of the DC-DC converter, the value of magnetizing current should determine if the MOSFETs hard-switch or not. But if there are no non-linear output capacitors across my MOSFET; my waveforms would be ideal and the switching losses from the thermal model would be wrong. Am I missing something?

Pramod

---

Pramod,

You can certainly still include dead-time so that's not an issue. As for the non-linear output caps, you can also add these manually if you think they will have any significant effect on the V/I values passed to the lookup tables for loss calculation. Have you tried this yet?

Kris

---

Hello Kris,

Yes, I did try it. As the charging/discharging current flows through the capacitor, the waveforms become non-ideal, but in this case the current flowing through the MOSFET itself doesn't account for this capacitor current. When the look-up table is made for these devices, the device current is taken into account which is now not the case in simulation. 

In short:
Measured current = I_MOSFET + I_DIODE + I_CAPACITOR
Simulated current = I_MOSFET + I_DIODE

Pramod

 

---

Hi Pramod,

I understand what you are saying. Can you post your model files or a minimally working example so I can see if I can customize this logic for you?

Thanks,

Kris

---

Hello Kris,

I'm sorry for the delay. I had to take care of some confidentiality issues. I just couldn't find any Infineon PLECS models available online. Most of them were IGBTs. The one I have is confidential and not be shared.

I have attached the basic simulation model here. If you have a 400V or 1200V MOSFET thermal model, please add it to this model. Thank you.

Best regards,
Pramod

---

Please remind me then, what are you hoping to see regarding the losses. You have effectively forced ZVS I believe and therefore should not expect to see switching losses. Are you just hoping to reduce the switching losses slightly then by including the caps and therefore reducing the current through the switch at the time of turn-on/turn-off?

---

Hello Kris,

Here I have only added the capacitors to mimic the output capacitors of the MOSFETs.
In reality, these capacitors would be responsible for the soft/hard switching of the switches.

In the model I attached before, the capacitor currents are not seen by the MOSFETs thermal model, so it shows zero switching losses. But if you add the MOSFET and the capacitor currents that would be the real MOSFET current which should be used to calculate SW losses.

PLECS model without capacitors -> Ideal waveforms, current and voltage at switching used to calculate the switching losses by the thermal model

PLECS model with capacitors -> Real waveforms with limited voltage (and current) slope. Switching current flows through the caps, the model doesn't see it and reports zero losses.

***Please do not forget to add a valid thermal model for the switches. Though they have ZVS, they still have turn-off losses.

---

Hello, let me ask a colleague about this and get back to you.

---

Hello again, I am going to send you further feedback in an email.

---

In general, PLECS is not a proper tool to simulate the switching transient. One can only add body capacitance to qualitatively verify if a ZVS is achieved (whether the capacitor is fully discharged), but then the switching loss amount cannot be determined accurately by the existing thermal model any more. Or in other words, the power loss calculated by the thermal model without a capacitor connected in the circuit is an over estimation of the true losses.