What is the best heat transfer method

[kinarfi]

I have lost several TO3 devices due to heat because they lost solid contact with the heat sink, I used the grey silicone/fiberglass insulator pad with the nylon screw isolator and the nylon melted and let the tab loose tight contact with the heat sink. I've started using mica for the insulator and put a piece of shrink tube around the screw thread and using a small piece of fiberglass circuit board to insulate the screw head to mount my FETS to the heat sink and use grease on both sides of the mica.
What's the best heat conductor? is there a better way to securely mount a TO3 to a heat sink?
Thanks,
Kinarfi

 

[kubeek]

You could use a spring loaded mechanism to keep the transistors in place, or just run it a little cooler so that stuff around it doesn´t melt

 

[Pommie]

Use a proper insulation setup and if that doesn't work get a bigger heatsink. By proper I mean the kits they sell. https://www.altronics.com.au/index.asp?area=item&id=H7100 for example.

Mike.

 

[tomizett]

I believe that mica is about the best for thermal conductivity - it should be easy enough to find thermal resistance figures from manufacturers datasheets. Metal screws with plastic top-hats (as used for metal tab devices like TO220) might allow you to tighten down harder than nylon screws and be easier than the heatshrink+PCB solution.
Of course, make sure your heatsink is completely flat under the device too.

It does seem that you might be running near the edge though, so a bigger heatsink is probably in order too.

 

[ronsimpson]

Are you using split washers that help keep the bolts tight when the temperature changes?

Can you insulate the heat sink and not the transistor? (no grey silicone/fiberglass insulator pad)

Get a bigger heat sink! Get a bigger heat sink!

 

[kinarfi]

#1 Thank you all, & my apologies, it's the TO220 that I'm using, loosing, it was late when I posted #1. The lost FET got so hot that the surface of the tab no longer looked smooth and uniform and was freely movable.
#2 a larger heat sink may be in order but would not have saved my FET, IMHO.
#3 Direct contact with heat sink not feasible as there are 4 FETs on this heat sink that form an H Bridge.
I think the problem is that the FET gets hot, the nylon softens, the contact pressure lessens, poorer heat transfer, FET gets hotter, nylon softens more and on and on. If the nylon didn't melt, the FET temp would rise, but the temp would level out within the operational range of the FET, thus my switch from nicely formed nylon insulators to the tedious use of shrink tube and fiberglass board.
Thank for the confirmation of my thoughts about the mica insulator.

 

[kubeek]

If it´s TO220 then definitely try somthing like these https://www.tarapath.com.au/productRange_image/240/3120_3121.jpg

 

[ronsimpson]

Kinarfi, How many watts are you dumping into the part? The insulator can not do 100 watts. You need to find why so many watts. If you are milting insulators then you are at/above the milting point of silicon.

 

[mab2]

one good thing about FETs is that you can parallel two or more fairly easily - which would make heat transfer less problematic.

 

[4pyros]

Can you use two heat sinks?

 

[kinarfi]

Can you use two heat sinks? no my problem is not heat sink, it's heat transfer to heat sink, I'm using IRF3205 55V 110A 8.0mΩ FETS and running them at about 30 and maybe up to 60 at motor stall.

 

[kubeek]

That is 30W at most, they shouldn´t be running that hot.

 

[tomizett]

I've suffered from FETs dieing too, when I though they where not running that hot but - it turns out - not making good contact with the sink.
If it's TO220, a clip that presses on the plastic body is generally better than mounting through the hole as it puts the pressure where you want it and can't work loose.

 

[chemelec]

Can you use two heat sinks? no my problem is not heat sink, it's heat transfer to heat sink, I'm using IRF3205 55V 110A 8.0mΩ FETS and running them at about 30 and maybe up to 60 at motor stall.

It Sounds more like your Mosfets are being Distroyed by Transient SPIKES from the Motor.
Not Just Overheating!

 

[MrAl]

Hi,

If this is a PWM application my first guess would be the transistors are not being switched on and off properly (gate drive). Switching losses add up fast with fast PWM.
If it is just forward and reverse it could still be that though depending on gate drive levels but then it would be more like the voltage levels themselves rather than the turn on turn off switching speeds.

 

[kinarfi]

I should not be having problem with transients, there is a full wave bridge across the motor.
Yes, it is PWM

 

[chemelec]

Still can be Transients.
Can you Post the Schematic?

 

[richardb]

This sounds a familiar thread.
how many degC/W is the heatsink? its gonna be vast

that part is package limited to 75A MAX with perfect cooling.

 

[ChrisP58]

What is your PWM fequency? What are you driving the gates with?

Power losses in mosfets can be much higher than expected if the gate drivers aren't able to drive the gates on and off very quickly. Even a few microseconds can hurt performance if the switching speed is high

 

[4pyros]

If its alright with you we would like to check for overheating possibilities.
It could be the fets are not being driven fully or to small a heat sink.
Would you mind posting the schematic and a pic of board and heatsink.