Using Multiple Heatsink on one Transistor?

[skipjackrc4]

I am designing a constant current sink (yeah, really creative project, I know) that uses 2-3 MOSFETs in TO-264 packages.  Each MOSFET will have its own CPU heatsink with heatpipes, with a thermal resistance of roughly 0.11.  I can safely get 200W from each, but I would like to squeeze a bit more out of the system.  I am pretty well set on using air cooling vs. water or oil.

So, my question is: would sandwiching the MOSFET between 2 different heatsinks be effective?  It seems like the thermal resistances would then effectively be in parallel, but I'm no thermal expert.  What do you guys think?   

[c4757p]

Use more MOSFETs. Hundreds of watts is not a project where you can skimp like that. The poor guy shouldn't have to handle all that power himself, he'll fail in no time at all no matter how fast you get the heat out. Real, commercial loads use a lot more than three TO-264s for hundreds of watts.

[c4757p]

And hundreds of watts is nowhere near a need for water or oil cooling, except in the rare case where you're forced to dissipate all that heat on one die (fast-as-hell PC, that's about it). The fact that you even considered this makes me recommend putting a little more effort into learning how others handle this much power before you ruin some expensive MOSFETs and start a fire.

[skipjackrc4]

OK, thanks for the reply. 

At 100 degC, the MOSFET is capable of dissipating 1000W, according to the datasheet.  I'm using this MOSFET partly because of it's junction-case thermal resistance of 0.05.  That is extremely low.

FDL100N50F: http://www.mouser.com/ds/1/149/FDL100N50F-42314.pdf

If I can keep the heat out, how would 200W dissipation be a problem?  I'm not trying to argue with you, but rather, I want to know what I am missing here.  A 200W target is far below the derated power dissipation limits.  The reason for using fewer MOSFET's is decreased control circuitry, which reduces PCB size and cost. 

[M. András]

you also need to consider the case to sink resistance. and take a look at the ixys linear fets those have nice specs for linear operation pass element/current sink, they are expensive but less likely to fail then those switching fets

[c4757p]

Holy crap, that is one hell of a MOSFET. Definitely designed for switching, though, I wouldn't use it in linear mode, especially with that much power.

200W is a lot of dissipation. If you can keep it within its temperature range, it won't be that much, but good luck with that. They quote 0.1 °C/W case to heat sink "typical", but I'm not too sure how typical that is. Be careful.

Fewer MOSFETs isn't going to reduce your control circuitry that much if you're doing it right. An op amp per MOSFET is nothing when the FETs themselves are already this expensive.

[skipjackrc4]

Yep, that MOSFET is a beast.  I was originally going to do a large bank of 10 or transistors, and then I found that one.  With some rough calculations, it wound up being cheaper to use these big MOSFETs  instead of smaller ones.  Not a big difference, though.

I didn't trust the 0.1 typical case-sink resistance either, and was assuming about 0.25 or so. 

Its SOA includes DC operation, and I would be well within that, but it still made me a bit nervous. 

To my original question, though, mostly just for curiosity: do multiple heatsinks act as parallel thermal resistances, or is it more complicated than that?

[SeanB]

As there is only one thermal interface connected to the die ( the case one is pretty poor) you need the lowest resistance on that side. Adding a second one will not do much more, the best is to use the recommended mounting method and hardware along with the right fasteners and torque method and go from that.  Forced air cooling over the case will lower it by a small amount, but keeping the case bottom at as low a temp as possible is the thing you need to do.

[mikeselectricstuff]

Much easier & cheaper to spread out over several devices either on one big heatsink or several small ones.

[Jay_Diddy_B]

Hi,

That is good advice from AcHmed99   .

I posted some comments on using MOSFETs in the linear applications in this thread:

https://www.eevblog.com/forum/projects/did-i-forget-anything-constant-current-load/msg170915/#msg170915

If you use multiple MOSFETs it is better to use one op-amp per MOSFET to ensure the current is shared equally.

If you watch Dave's Teardown video of the BK8500 electronic load you can see the quad op-amps that are used for the sharing at 14:30.

http://www.eevblog.com/2012/05/23/eevblog-281-bk-precision-8500-electronic-load-teardown/

The same technique is used by HP in the 6060B Load. You can search for the Service manual for this load on www.agilent.com

Jay_Diddy_B (not my real name, I am in disguise    ->  O0 )

[c4757p]

I second the safety glasses suggestion. I have had more than one transistor decide to jettison a piece of its package. It's usually spectacular. That video someone recently posted of an LED blowing up is not atypical at all.

[c4757p]

My favorite was a TO-92 that shorted a 70V rail because I stupidly set the PCB down on some solder dust. A worthy rival to the Independence Day fireworks.

[skipjackrc4]

I totally agree with the safety glasses.  As part of a previous job, I was involved in purposely blowing up large numbers of IGBT's for failure testing.  Very fun to watch, especially when the conformal coating on the board caught on fire.

I was on the fence about trying those transistors anyway, so you guys have convinced me.  I won't even bother doing experiments with them since they cost so much.  Excellent point about the 0.05 C/W possibly being a typo.  That is absurdly low.

Thanks guys!