Negative Loss values for component and Loss variable limit exceeded

PLECS
1

The model was running fine with a higher number of FETs and IGBTs. but now the constraints have changed and I have dropped the number to 4 each of the FET and IGBT per hybrid switch. i have also reduced the peak currents I am running but i keep running into model is too stiff errors or matrix is singular errors, along with loss variable limits exceeded on the fets and negative conduction loss warnings on the IGBTs

HyS_SiCFET_SiCTIGBT_V2.plecs (527.6 KB)

2

As we have previously discussed, this is likely due to runaway thermal behavior where your temperatures approach infinity. You must debug the underlying electrical and thermal circuit to ensure the conditions are appropriate for the device.

Some techniques to debug the thermal model:

  1. Ensure all electrical stresses are appropriate and within the device ratings. You cannot debug a thermal model if the electrical conditions will inherently lead to thermal runaway.
  2. You can fix the case temperature. If the temperature is still too high then the device is in appropriate for the conditions simulated in the model. If things look normal then it will also identify if the case-to-ambient thermal chain impedance is too large / inappropriate for the system requiring a refined design of the heatsink / cooling system.
  3. You can try having just one of your phases model thermally to minimize the combined contribution of all devices.
  4. Pick more appropriate devices or place more in parallel.

I’ll note you didn’t include your thermal model for the A4E120N50 device.

3

Hi Bryan,

I had switched to using the MOSFET model available from Microchip instead of creating my own one which had fixed the MOSFET over temperature issue.

I have reduced the number of devices in parallel, but at the same time i am running lower currents that the switches should be more than capable of handling.

The error i am getting is focused on the diode, which is using a thermal model from Infinieon, not one I have created, they were no issues with the diode initially when i had more in parallel, but now it seems to be only causing issues in that.

IFS100B12N3E4_B31_diode.xml (9.8 KB)

TC_IGBT_A4E120N50.xml (1.8 KB)

Thermal files for the IGBT and diodes i am using (the FET is using the Microchip library) are attached. I have also pasted the screenshot for the error i am getting.

I had added a thermal resistor and a constant temperature source to reduce the thermal runaway issues but it still seems to have not worked?

image
image1782Ă—848 111 KB

image
image1780Ă—838 95.5 KB

4

Those are nuisance messages due to the forward voltage drop in the electrical model. You can add another row to the voltage values as shown in the figure below to clamp the losses at 0 J for small forward voltages.

image
image286Ă—172 17.2 KB

I had added a thermal resistor and a constant temperature source to reduce the thermal runaway

You should only add the constant temperature source to fix the temperature. The resistor isn’t needed as it you will have a temperature delta across the resistor which we are trying to eliminate.

TC_IGBT_A4E120N50.xml (1.8 KB)

This thermal model does not have a thermal impedance from junction to case. You should fix that!

I have reduced the number of devices in parallel,

I assume this is a typo and that you increased the number of devices in parallel. Also note you can significantly improve the model construction. In PLECS 5.0 there is a “Number of parallel devices” option for the native MOSFETs and Diodes. This is already also implemented in the Microchip part. This way you can change the number of devices without having to actually modify the circuit - only by setting this parameter. Furthermore, you could use “Model References” to have just one MOSFET + IGBT + Diode subsystem that you would then reference for each switch position.

5

Hi Bryan,

Thanks a lot for your help. before you had replied i had tried a few things and adjust i removed the thermal chain values i was using as just used a single value for R and t at 0.001 for both. that seems to get the model to run now without errors (the warnings still remain).

I am now however getting very high losses:

image
image1908Ă—1012 64.6 KB

where exactly did you mean to add a thermal impedance to the igbt model?

if i dont use a thermal resistor before connecting the heatsink to the constant temperature source, i get the following error:

image
image2557Ă—631 186 KB

i did see that i can combine the switches, just hadnt gotten round to it yet and wanted to model it all for easier debugging.

6

if i dont use a thermal resistor before connecting the heatsink to the constant temperature source, i get the following error:

See the State-source dependency of components section of the PLECS documentation, especially the section pertaining to setting the thermal capacitance of the heatsink to zero.

where exactly did you mean to add a thermal impedance to the igbt model?

The thermal chain is empty:

image
image833Ă—618 43.8 KB

7

thanks bryan!

managed to work on the model based on your feedback and got it to work now!

1 Like