DPak heat sinks

I need to do a calculation, Google search or whatever, but I'd rather watch the NE Patriots playoff game, so I'm going to toss my problem over to you guys. Perhaps it can only be solved with a bench measurement. Has one of you been there and done that?

I have three DPak MOSFETs located on 2.5 square inches of copper on both sides with heat-transfer vias. Additional heat-sinking is required, and I'm evaluating a 35x60mm flatback heatsink with fan to remove heat and increase the maximum allowed dissipation.

I've located a decent heat-spreader layer, by t-Global. I can easily place it on the DPaks with the heatsink on top. We know the heatsink will have a poor thermal path to the DPak tabs. Alternately the heat transfer layer and heatsink can go on the other side of the board, arguable with better heat transfer from that side's copper. Which scheme will work better? Can we add performance numbers to the evaluation?

If I use Aavid 0.5-inch-high 7109 DPak PCB-mount SMT heatsinks on top, with a fan, we might be able to handle more than 5 watts total for the three parts. I'd like to do much better than that. The MOSFET channel-to-tab Rth is 6-deg-C/watt. My co-conspirator says I should stand up the part and treat it like a TO-220. Ugh! Infineon offered an upside-down TO252, with metal exposed to the heatsink. Don't wanna try bending leads and playing that game.

Any advice or comments will be appreciated.

--
 Thanks, 
    - Win
Reply to
Winfield Hill
Loading thread data ...

You can surface mount the dpaks onto topside copper pours, thermal vias to the bottom side to big copper pours there, have a thin layer of a gap-pad type stuff, and then a real heat sink. That should be good for several watts per part. Figure the gap-pad could be 6 w/m-k. Use 2 oz copper and a thin board if you can.

Very roughly, 6 k/w for your fet, 6 for the vias, 1 for the gap-pad, maybe another 2 for spreading resistance, total around 15 K/w from the junction to the heat sink surface.

More extreme would be to route holes in the PCB and use a ceramic insulator (alumina is mediocre, AlN is great) and heat sink on the other side. That would have no vias or PCB copper in the thermal path.

20 watts per transistor would be easy.

I'd use surface-mounted TO220s and an insulator with a hole, and bolt things together. Sort of like this:

formatting link

That's actually a TO247 with a TO220 sized AlN insulator underneath. I had a bunch of AlN TO220 size insulators made for this.

formatting link

Those guys will make anything you want.

I guess you could glue a DPAK to an insulator without a bolt. Or clamp mount a DPAK to an insulator and then to the heat sink, again through a routed hole in the PCB.

--
John Larkin   Highland Technology, Inc   trk 

jlarkin att highlandtechnology dott com 
 Click to see the full signature
Reply to
John Larkin

Why DPAK over, say, PDSO-8 (5x6mm) or DFN (3x3mm)?

The latter can dissipate just as much -- you're limited to the thermal capacity of the board in either case -- and is smaller and thinner (some are even characterised for thermal conductivity through the plastic top).

A goopy thermal pad clamped on top will soak up heat from the component and surrounding pours nicely, and a 2oz top layer will grant much needed lateral conductivity. Even better with inner layers.

If you need to dissipate more than 5 or 10 watts, it's really not a good idea, any way you cut it. Better to use a THT in that case.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Design 
 Click to see the full signature
Reply to
Tim Williams

The smaller the footprint, the harder it will be to spread out the heat of a surface-mount part.

--
John Larkin   Highland Technology, Inc   trk 

jlarkin att highlandtechnology dott com 
 Click to see the full signature
Reply to
John Larkin

There's no choice about this. This is my super-fast high-voltage amplifier, requiring MOSFETs with very low capacitance. For this case, it dictates a DPak.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

I've run into that lately, the tradeoff between heat sinking and capacitance.

--
John Larkin   Highland Technology, Inc   trk 

jlarkin att highlandtechnology dott com 
 Click to see the full signature
Reply to
John Larkin

On Sunday, January 20, 2019 at 4:26:56 PM UTC-8, Winfield Hill wrote: ...

Well, a case can be made that the die and backplate are already coupled well, and you want to continue that thermal coupling into a heatsink. So, mount the DPAK over a hole, not just stitched vias; drop in a copper slug after soldering, to fill the sandwich, and top-bread is a holddown plate, bottom-bread is a heatsink with a simple flat.

It ain't pretty, but blanking a handfull of copper slugs and keeping the capacitance nearly unchanged with a mechanically strong hoddown plate are not big problems. It's a loose-hardware issue, still, unless you can locate it all near a board edge and finagle a clip-on 'wrap' instead of multipart sandwich.

I've always kinda admired, too, the threaded-stud-attached packages for big thyristors, and the occasional CPU. Never liked the stitching of PC boards, because the good cryogenic insulating spacers we used to use, are... fiberglass reinforced plastic (heavy on the glass, light on the epoxy).

Reply to
whit3rd

So, the SiC parts, innit?

It annoys me that they don't put them in small packages. SMA diodes are almost unheard of. Evidently the dies are minuscule too, a DPAK is so much wasted space (which translates to more pF for you).

Your power requirement doesn't help the capacitance much, either. :-/

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Design 
 Click to see the full signature
Reply to
Tim Williams

The complication is that he may need electrical isolation. There will be a direct tradeoff between thermal resistance and capacitance to the heat sink, parameterized by the properties of the insulating material.

formatting link

One cheat is to cool with mass transfer, flow.

--
John Larkin   Highland Technology, Inc   trk 

jlarkin att highlandtechnology dott com 
 Click to see the full signature
Reply to
John Larkin

What about Wakefield-Vette's ulTIMiFlux PhaseChange Thermal Material? They specify 0.107 C/W at 100psi for 1-in^2, amazingly low. Low cost: $7.50 per 5x5" piece, qty 10. Could stack multiple 3-mil layers to reduce capacitance.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

That works out to about 1 w/m-k, which is mediocre if capacitance matters. And if you believe their numbers. That stuff is poorly specified. I'd test it for both theta and capacitance.

I don't like phase-change material; it's yukky.

I use some gap-pad stuff from 3G Shielding. It has the general consistency of bubble gum and it's 6 w/m-k.

I got tired of doing the math so wrote a little program.

formatting link

--
John Larkin   Highland Technology, Inc   trk 

jlarkin att highlandtechnology dott com 
 Click to see the full signature
Reply to
John Larkin

Yes, that's a very useful little program! Thanks!

--
 Thanks, 
    - Win
Reply to
Winfield Hill

Tauno Voipio wrote an HTML version. It needs javascript.

formatting link

You can't run it directly because Dropbox doesn't want to host web pages. So copy it to a local file and then a browser can run it.

--
John Larkin   Highland Technology, Inc   trk 

jlarkin att highlandtechnology dott com 
 Click to see the full signature
Reply to
John Larkin

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.