0
\$\begingroup\$

I'm trying to estimate the case to ambient thermal resistance of a GaN switching device (specifically the GS66508T device from GaN Systems). The datasheet provides the junction to case thermal resistance, but not the case to ambient. Is there any way I can estimate the case to ambient value? The reason why I need that value is because I know the case temperature of the device and I want to calculate the power dissipation and junction temperature. The device package is GaNpx which is fairly obscure - do you know if I can just find a package with a similar structure/shape with a known thermal resistance and use that? The device is top-cooled.

Thanks in advance!

\$\endgroup\$
4
  • \$\begingroup\$ I was going to say something without looking up the package. Then I saw how it's built and what it is. I think you are in boutique-land. Have you contacted the supplier about this question, yet? (I wonder also about the hidden implications behind your wording that suggests only one "junction temperature" to find.) And don't forget that radiation/surface emissivity/environment may also be worth a look-see. (Must be some serious reasons for all the trouble that goes into that package. Makes me curious.) \$\endgroup\$
    – jonk
    Commented Jun 1, 2021 at 4:00
  • \$\begingroup\$ I contacted the manufacturer and hopefully they have some typical values to go off of. That's an interesting point you bring up about the junction temperature. Is it typical to have numerous values for a single "controlled (to the best of my abilities)" condition? \$\endgroup\$
    – ggg123
    Commented Jun 1, 2021 at 5:17
  • \$\begingroup\$ Can you not obtain the power dissipation some other way, for example by electrical measurements? Or you need to work backward from case temp? The junction and case are gong to be about the same temp in a package like that, unless you put a big heatsink on the topside pad. \$\endgroup\$
    – user57037
    Commented Jun 1, 2021 at 7:32
  • \$\begingroup\$ @ggg123 I'm assuming there's more than one junction. Consider the idea of a plate, round, square, or rectangular I don't care, and think about heating; the various ways heat leaves such an "odd" shape. There are edge effects (matter near the "edges") which have quite different surface radiation, conduction, and convection issues from the more central regions, for example. Personally? I don't worry about the average over some entire chunk of matter. I worry about the "hottest spot" -- which is usually not near an edge. If you imagine it is "all of a piece" then that lumped view may be trouble. \$\endgroup\$
    – jonk
    Commented Jun 1, 2021 at 9:35

1 Answer 1

3
\$\begingroup\$

Devices like that are meant to be used with a heatsink.

The junction to ambient thermal resistance will mainly depend upon that heatsink.

You will need to calculate the thermal resistance of the combination. The device manufacturer doesn't have that information to put it into the datasheet.

Low power device datasheets can declare a device to ambient thermal resistance as they are not usually used with a heatsink.

\$\endgroup\$
2
  • \$\begingroup\$ Ah that's what I figured. Currently I'm trying to solve for the dissipation with no heat sink just to see what the limits of the naked device are. When you mention calculating the thermal resistance of the combination, I assume you are referring to the datasheet resistance values contributed by the heat sink and TIM? \$\endgroup\$
    – ggg123
    Commented Jun 1, 2021 at 5:19
  • \$\begingroup\$ @ggg123 - Yes. It will depend a lot on your PCB. One way to dissipate heat effectively in a PCB is to provide a significant area of unbroken copper in the PCB. As a first guess one square inch will provide about 100°C/W. As you can see the device characteristics won't make much difference with that. \$\endgroup\$ Commented Jun 1, 2021 at 14:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.