Basic MOSFET Thermal Design: Parameter Calculations from Datasheet

[KyleGranger]

TL;DR Summary

I would like to calculate parameters, such as power dissipation, for a given MOSFET based on the manufacturers datasheet.

I would like to be able to determine the current through a device for a given junction temperature. I am looking at a datasheet and notice that it gives the device power dissipation with different case temperatures. Since the maximum junction temperature is 175 C, I believe that means that lower case temperature with higher power means it would require a better cooling system compared to higher case temperature and lower power dissipation.

First I wanted to calculate the power dissipated by the device for a given case temperature, assuming the junction temperature is 175 C. To do this, I thought I would use the equation q=ΔT/Rth where q represents the power in watts, ΔT represents the temperature in degrees C, and Rth represents the thermal resistance C/W.

I checked this logic using the datasheet values for P=441 W @ Tc=25 C and Rth=0.25 C/W. I get (175-25)/0.26 = 577 W. The datasheet value is 441 W, so this method doesn't check out. What am I missing here? Or is there a better way to do this?

What I hope to do next is determine the number of parallel devices for a given cooling system at a given current. This part will be done using thermal simulations, but I need to know a power to apply as a heat source and have a way to check my basic simulations to ensure I'm getting close to the expected case temperature.

A link to the datasheet is below.

https://unitedsic.com/datasheets/DS_UF3C065030K3S.pdf

 

[DaveE]

Sorry, I don't have time to go through and check your work. However, you are on the right track with your thermal equations (Rth, etc.). The device manufacturers specify this in different ways, so some translation is required. They also don't know how you will cool the case, so they usually just specify a case temperature for their characterization and leave the thermal modelling for your situation up to you. The key things I want to know are maximum junction temperature and R_Θjc, then combined with the (hopefully standard) package, that is all the manufacturer can tell you, the rest is your design. There are a bunch of places on the web that you can read about this. I put some below, which is just the first couple that I found that looked reasonable.

https://www.eeeguide.com/heat-sink-in-transistor/
https://info.boydcorp.com/hubfs/Thermal/Air-Cooling/Boyd-How-to-Select-a-Heat-Sink.pdf

BTW, if they say the junction can get to 175C, don't try that. You'll want some design margin. My designs were nearly always for pretty hi-reliability, I would limit it to 140-150C or so. A cheap audio amp design would probably use 160-170C.

 

[KyleGranger]

Sorry, I don't have time to go through and check your work. However, you are on the right track with your thermal equations (Rth, etc.). The device manufacturers specify this in different ways, so some translation is required. They also don't know how you will cool the case, so they usually just specify a case temperature for their characterization and leave the thermal modelling for your situation up to you. The key things I want to know are maximum junction temperature and R_Θjc, then combined with the (hopefully standard) package, that is all the manufacturer can tell you, the rest is your design. There are a bunch of places on the web that you can read about this. I put some below, which is just the first couple that I found that looked reasonable.

https://www.eeeguide.com/heat-sink-in-transistor/
https://info.boydcorp.com/hubfs/Thermal/Air-Cooling/Boyd-How-to-Select-a-Heat-Sink.pdf

BTW, if they say the junction can get to 175C, don't try that. You'll want some design margin. My designs were nearly always for pretty hi-reliability, I would limit it to 140-150C or so. A cheap audio amp design would probably use 160-170C.

Thanks for the references. I read them and that's pretty much my understanding of it. I need to design an enclosure and will likely have forced air cooling so would like to also do simulations so I have a custom solution.

The equation I listed above is equivalent to the one in one of the references you provided. The power is pretty far off though; 577 W vs 441 W from the datasheet. This is probably the simplest part of this but I still don't see what I'm missing.

 

[Tom.G]

Summary:: I would like to calculate parameters, such as power dissipation, for a given MOSFET based on the manufacturers datasheet.

I checked this logic using the datasheet values for P=441 W @ Tc=25 C and Rth=0.25 C/W. I get (175-25)/0.26 = 577 W. The datasheet value is 441 W, so this method doesn't check out. What am I missing here? Or is there a better way to do this?

That's because you used the Typical thermal resistance of T_J-C from Page 2 of the datasheet. The 441W is calculated using the Maximum thermal resistance.

Also, note that even if the MOSFET is immersed in oil, the case temperature will be above the ambient temperature.

Cheers,
Tom

 

[KyleGranger]

That's because you used the Typical thermal resistance of T_J-C from Page 2 of the datasheet. The 441W is calculated using the Maximum thermal resistance.

Also, note that even if the MOSFET is immersed in oil, the case temperature will be above the ambient temperature.

Cheers,
Tom

Thanks, I'm getting the right number now! I plan to use thermal simulations to get a better idea of the cooling performance with different airflow rates and heatsinks. Since I'm building my own model, I wanted something to use as a reference so I can validate that it's at least somewhere in the range of where I'd expect it to be with just convection.

I understand the case won't be the same temperature as ambient. I'm not really expecting much cooling from the case though if I add a fan though. I just wanted it to have a baseline for my model comparison as a sanity check.

 

[DaveE]

Power capabilities listed on data sheets are usually way too optimistic, in my experience. T_jmax is pretty standard (150C or 175C), so I would just always look for R_Θjcas a figure of merit; It is almost always directly a function of die size. In any case power semiconductors almost always need significant heatsinks of some sort.

 

[Baluncore]

@KyleGranger Welcome to PF.

In any case power semiconductors almost always need significant heatsinks of some sort.

The cost of the thermal management will often exceed the cost of the semiconductor.

 

[DaveE]

@KyleGranger Welcome to PF.

The cost of the thermal management will often exceed the cost of the semiconductor.

Yes, especially if you account properly for assembly labor costs.

 

[Tom.G]

I plan to use thermal simulations to get a better idea of the cooling performance with different airflow rates and heatsinks. Since I'm building my own model, I wanted something to use as a reference so I can validate that it's at least somewhere in the range of where I'd expect it to be with just convection.

For comparisons and sanity checks, try the catalogs of the various heatsink manufacturers. They have explanations, data tables, along with curves for their stock heatsinks.

Here is a link to the first one that came to mind:
https://www.mouser.com/catalog/supplier/library/pdf/Aavidselectionguide.pdf

Cheers,
Tom