LMG3522R030-Q1: Datasheet missing Eon/Eoff; errors in PLECS thermal model

Checking in again. The fixed PLECS model would be especially helpful as it should provide an estimate for switching losses so we can properly size a heatsink.

Hello Steven,

I will look into this and get back to you about the Updated PLECS model.

Regards,

Adithya Ankam

Hello Steven,

We are in the process of updating our PLECS models right now. Previously, we had two libraries for high voltage and medium voltage devices, which caused some issues. Now we are working to merge the libraries and create one for all TI GaN parts, which is ongoing.

I can share with you our LMG3522R030 PLECS model which has Eon/Eoff data, you can use this for helping with heat sink sizing.
LMG352xR030_PLECS_V2.0.zip

For the Foster model, I will have to follow up with you, we are not aware of any issues with this so I will need to look and check with our team.

Thanks,
Zach Soviero

Hi Zach,

Thank you! I tried using model v2, but I am unable to run any simulation as the mask has initialization errors looking for parameters that existed in the v1 model but no longer exist in v2 (Vf, R_coss, Id_max to name a few).

I tried manually filling in those values, but it was then unable to find 'device' (looks like another thing that was renamed). Is it possible you have a new version of the file 'TI_GaN_Version_A.plecs' that is compatible with the v2 model for LMG532xR030?

RE the thermal model: I see the Foster model was updated in the LMG532xR030 v2 model file. However, if I sum all Rth components it still does not match the datasheet value of 0.28 C/W as expected. See another post discussing this issue here:

https://e2e.ti.com/support/power-management-group/power-management/f/power-management-forum/1281860/lmg3522r030-thermal-model-plecs

Hi Zach,

Have you had a chance to take a look at this?

Hi Steven,

I am involved in out PLECS modeling, and caught up with the previous thread you mentioned here.
First about the Foster modeling:
When adding a single thermal resistance value in our datasheet, the process with our packaging team takes all the data related to each device, and works through a number of thermal simulations to produce a single number representative of the thermal resistance.
Our PLECS models use the foster model to simulate thermal resistance and help simulate transient thermal behavior in addition to steady state. The method for generating and using the Foster model within PLECS is still valid and will help accurately model thermal losses. As our PLECS models improve, we will continue to work on our Foster models and ensure they are as accurate as possible

Second about the PLECS initialization errors:
As mentioned we are still in the process of updating our PLECS models, which is why you see these errors. We are working to fix the initialization file with the PLECS software team and will update shortly.

Thanks,
Zach S

Hi Zach, thanks for clarifying. I'll wait for your updated PLECS model (hopefully soon, as we are in the process of development right now).

RE the thermal model, my understanding is if the thermal resistances in the Foster chain don't add up to the datasheet Rjc value, this indicates an error and will cause the PLECS thermal model to converge to a different temperature than the datasheet Rjc suggests. So even though the packaging team is modeling many different thermal paths, the sum of those paths must equal the datasheet Rjc. Otherwise one or the other Rjc value is incorrect.

Note this is just about steady-state conditions. The thermal capacitances affect the dynamics, but right now I'm only concerned about the steady-state which is determined entirely by the Rth values in the Foster chain.

I forgot to mention that based on the PLECS model, the thermal resistance is ~3x higher than the datasheet. If true, this part is not-usable for our application. So this is why I'm trying to get to the bottom of this point. Any feedback is greatly appreciated. Thanks!

Hi Steven,

Thanks for the additional information, let me confirm the foster model first since thermal design is very critical. I will get more details and share with you this week

Thanks,
Zach S

Hi Zach, any update on this?

Hi Zach, pinging again.

Hi Steven,

After a lot of digging and communication with our packaging team, I have a full answer for you.
First, why the PLECS foster model was wrong: the foster model loaded into the PLECS model you are using is for the wrong device which is why it is so different than the datasheet. There are a number of reasons why this happened but the result is that the foster model is not correct.

Previously, our foster models have not been optimized for PLECS, and after working with the PLECS team we are updating all our foster models to be only 3-5 resistances in the network since additional values become redundant.

Below is a corrected version of the foster model for LMG3522R030, is it still being worked on by packaging team, but you can see that it is matching the datasheet properly:

you can use this table to help run simulations.

Thanks,
Zach S

Hi Zach,

Appreciate the reply. To use the updated Foster model you provided, I really need a working mask (icon). It seems someone changed much of the init code and the variable names, and they no longer work with the underlying model (rev B) when using the provided mask (rev A). So if you can provide that I'd be happy to close out this thread. The file that is broken is 'TI_GaN_Version_A.plecs'.