What is the design methodology of a Heatsink for a high-power diode?

[Ramazan Duzgun]

TL;DR Summary

Either natural or forced air cooled convection can be thought, I have a 400 A (Vforward=1.55 V) power diode and I need to design a heatsink for that.

In reference of Ned Mohan's book, I have designed a heatsink, firstly natural convection taken into account, but laboratory test results and theoretical calculations are not matching.
The test results:
For 40 A (means 62 W), 30 Celsius ambient temp, it raised up 51 Celsius. So, Rs-a = (51-30)/62 = 0.338 Celsius/Watt.
For 60A (means 93 W), it raised up 64 Celsius. Then, Rs-a = (64-30)/93 = 0.365 Celsius/Watt.
In theoretical calculations, for instance, for 40A, it supposed be Rs-a = 1.43 Celsius/Watt. This is not feasible.
I have also tried the other author's calculation methods, like B. W. Williams, but it was also not feasible.
However, there must be exact solution for heatsink calculation. What is the design methodology of a heatsink ?
Thanks!

 

[jrmichler]

What is the design methodology of a heatsink ?

The design methodology is something like this:
1) Determine the thermal contact resistance from the diode to the heat sink. This must be measured experimentally.
2) Calculate the thermal resistance of the heat sink. This requires FEA.
3) Calculate the thermal resistance from convection. Include the effects of nearby obstructions. The tool for this is CFD.
4) Combine the three steps above. You have now analyzed one heat sink, with one contact resistance, one heat sink material, in one installation.
5) If not good enough, change something and try again.

There is no such thing as an exact solution, but there are approximate solutions. The exact performance will always be affected by installation factors that affect air flow past the heat sink. In the end, you still need to run a test.

 

[Ramazan Duzgun]

Thank you for your reply, but first of all I have no design data. I only have a diode and its poor datasheet. How can I specify the heatsink dimensions ?

 

[Dullard]

I agree with jrmichler's answer (as a specific response to your specific question). I'd add:
Convection mode is necessarily going to be harder to model/predict than forced air cooling. Modelling is a fantastic tool, but I suspect that you'll invest more time empirically sorting out your model than just empirically sorting out your cooling. Uisng forced air will reduce the difference between 'predicted' and 'actual.'

 

[Windadct]

First off how did you measure the Ts? For our company it define as a point 2mm "inside" the heatsink, directly UNDER the heat source, so you need to make a HS with an embedded sensor in that location.

 

[essenmein]

Another source of data, at least starting point and some idea how good various heatsink are in convection, some suppliers will give you convection cooling data for their heatsinks eg:
https://media.digikey.com/pdf/Data%20Sheets/Nuventix%20PDFs/63730_Datasheet.pdf

If you are looking to do the analysis on a computer, eg Autodesk CFD is not bad lower cost option, then as long as you build your model as complete as reasonable (ie model the diode case, the thermal interface material, the heatsink etc, then it will do the first three steps jmirchler mentioned in one hit, calculates heat transfer via conduction, then will compute the convection cooling as well, this is a coupled simulation since heat transfer via convected air depends on flow and flow depends on heat transfer. What ever software you use you must be careful to use the right advection scheme for convection otherwise results can be quite wrong.