Experiment: Which mosfet thermal interface material is best?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
It could be very mixed results depending on how flat each surface is. If not very flat the contact area could be very small. If both surfaces are 100% flat and the surfaces are mirror polished the contact area will be large. If surfaces are 100% perfect they may never come apart again if a "cold welding" is taking place and this will probably be the best thermal contact of all. It will probably require that both surfaces are same material…..e.g. aluminium.
 
First: Eric: excellent testing, thank you for your efforts.
Second: Patrick, nice to see you linking your stuff.
Third:ignore that silly infineon writeup, it's useless.
I would recommend a graphene layer to a copper spreader, followed by graphene to an aluminum oxide insulator and grease below that. Make the copper spreader at least .1 inch thick.
Oh, and compress the package using a clip that goes over the mosfet. I didn't catch the package being used, but TO's are notorious for screw torque issues. Too high, package warps. Too low, inadequate pressure.
Oh, almost forgot. Don't worry about the heatsink temp so much, the largest thermal drop will be just below the semi baseplate where the thermal flux density is greatest. Concentrate there.
Also, the top center of a good plastic TO package will correlate quit well to die temp, no need to go crazy with accuracy concerns. (Yah, I just know someone is going to gig me for that comment .
Jn
 
Last edited:
As Infineon put it: there is no correlation at all between the chip temperature achieved in the test and the datasheet value given for thermal conductivity p.24 middle row
https://www.infineon.com/dgdl/Infin...n.pdf?fileId=db3a304343e0037c0143f875b52e3343
As an engineer, I am embarrassed by that article.
The errors, sigh...I don't know where to start.
Actually, I do. Start with Patrick's link, very good.
I don't know if my stuff is linkable, there's never a grandchild around to help with this techno-stuff when you need one..:D

Jn
 
I just visited Linear Audio, I note that my 45 degree thermal model spreadsheet is available to download. Unfortunately, as I look at it without my article from volume 9 in front of me as a guide, I find it very difficult to figure out.

But even without changing die sizes, the sheet does show the effective areas where the heat transport occurs. If you unlock it and play with die sizes, spreader thicknesses, and material properties you can see how the various layers impact the overall thermal resistance.

jn
 
Jason: I'm sure that lapping the surfaces would further improve the heat flow from the transistor to the sink. I've lapped engine valves, but never thought about doing this with transistors.

Patrick: Thanks for posting a link to your work. Somehow I missed that the first time around - great stuff. Thank you for writing this up and sharing it!

Preamp: Mounting the mosfet directly to the sink with no other material being used would be random at best and not transferable to any other situation. Some of the sinks I've had were well machined (VERY flat and polished), others have had deep grooves left over from the extrusion process. Yet others sinks have had obviously curved backs. It depends on too many variables...

Finally, since I completed my MoFo amp last week, I no longer have an easy test bed for this comparison.
 
I will read through later so apologies if you have covered this-
The torque applied to the mounting bolt has a very significant effect.
If you haven't tested for this it would be a worthwhile addition to your results.
If you don't have access to a torque wrench I could share my results when I get around to doing it.
 
Member
Joined 2011
Paid Member
Motorola application note AN1040 includes LOTS of data about the effect of mounting bolt torque. Or perhaps I should say the non-effect. Here is one of their numerous figures.

_
 

Attachments

  • AN1040.png
    AN1040.png
    88.4 KB · Views: 250
Motorola application note AN1040 includes LOTS of data about the effect of mounting bolt torque. Or perhaps I should say the non-effect. Here is one of their numerous figures.

_
I am talking about single screw mounting.

Those graphs as well as the TO-220 are not junction to heatsink values, but are case to sink. As such, they can be misleading.
On page 11 directly under fig 11, they state that the TO-220 cases will lift as a result of screw torque.

When the case begins to lift, the heat path is no longer direct to sink, but has to travel horizontally to the high pressure area.

If you are going to believe a write up like that, look carefully to see that they are actually measuring thermal performance by monitoring both the heatsink temperature and the junction temperature. Again, case temperature will be misleading.

I've performed these thermal resistance tests using two techniques..drive the die to the exact temperature using an active control circuit and see how much power is required for a specific heatsink temp, or drive an exact power and monitor a die temp dependent entity.

Driving the die to a programmed temp was really fun, as you can actually de-mount the device and watch the power drive plummet in response.

Jn
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.