another thermal thing

formatting link

I'd like to dissipate 40 watts from a D2PAK.

This little experiment has no vias, and the copper thickness is 1 oz. All the thermal conduction is through the epoxy-glass. Theta is measured from the fet to the aluminum plate. The topside copper is hot near the mosfet but drops off rapidly, so thicker copper would really help spread the heat around... 6 oz maybe?

With thick copper and a zillion vias, and maybe a thinner board, I suspect we can get this below 1 K/w.

This D2PAK fet is rated for 100 amps and 300 watts, both ludicrous numbers.

Reply to
John Larkin
Loading thread data ...

Tacking on some copper foil / sheet would simulate thicker copper.

The better the vertical path, the more important the horizontal will be.

Cheers, James Arthur

Reply to
dagmargoodboat

Must be IR brand... Direct attach to a diamond substrate; thermal conductivity i understand is better than silver and it is a rather decent electrical insulator.

Reply to
Robert Baer

I think this has been discussed in this newsgroup before.

Some manufacturers seem to specify the current handling of the actual chip performance, while the performance is much worse than what can be achieved in any package they offer. Look for "Package limited" or similar phrases in the data sheet.

Power transistors are rated at Tc = 25 C which is realistic only, if you dump the device in flowing water :-).

Reply to
upsidedown

I understand that the copper is your model for the final PCB top side? In that case, and if you hadn't planned on packing a lot of components near the MosFET, instead of using a PCB with a thick copper layer, I'd consider soldering a sufficiently thick copper plate on top of the PCB ('s vias) on which you solder the MosFET. You could also emulate the more traditional heat sinks by soldering copper wires, radially emanating from the MosFET to help spread the heat horizontally, if there is space for that.

If there's only little space around the MosFET, you could consider improvising a heatsink by putting 2 or 3 copper foils between PCB and MosFET, cut the parts of the foil that sticks out from under into small strips, and fold them up, like 5 mm high?

Paint the surface around the MosFET black with radiator paint might also help a bit.

joe

Reply to
Joe Hey

Where does the heat go once it leaves the pcb? (Is the aluminum plate to simulate the back panel?) George H.

Reply to
George Herold

100A might make sense, depending. They have to be calculating 300W from R?jc to an infinite sink or some such.

I used a pair of these IRF (now Infineon) FETs to handle ~1kA surges a few years ago, best I could find at the time.

formatting link

095845

0.4K/W and 340W-rated at Tc=25C jives with Tj(max)=175C.

Rugged. They handled the surges with ease, battle-tested.

Global warming. (John's naughty.)

Cheers, James Arthur

Reply to
dagmargoodboat

The baseplate will probably be a water-cooled cold plate, which I'm assuming will be fairly cool and isothermal. I'd rather pick-and-place a lot of (20 or 25?) small cheap surface-mount fets, rather than bolt down and hand solder a half dozen TO247s.

Reply to
John Larkin

I'll try to beg some scraps of heavy-copper FR4 from our PCB suppliers. I've never worked above 2 oz before, so that would be interesting.

Yeah, the vias will be the next limit.

Reply to
John Larkin

Den tirsdag den 20. oktober 2015 kl. 05.53.02 UTC+2 skrev John Larkin:

try some of this:

formatting link
?

maybe have some copper sheet lasercut/etched to fit on top, it could probably be reflowed

-Lasse

Reply to
Lasse Langwadt Christensen

How about a heat pipe as used as CPU cooler in laptops, since you have that cold plate ?

Reply to
upsidedown

Putting the MOSFET on a copper plate would significantly increase board fabrication difficulty.

Making a form-fitting copper plate with "U" notches that nestle around the FETs may help. You'd need to find someone who could punch them out and keep them flat.

I think John's going to try vias next, which seems like a good thing to try.

--

Tim Wescott 
Wescott Design Services 
http://www.wescottdesign.com
Reply to
Tim Wescott

Wouldn't this be part of a much larger board, with 6 oz on the whole thing? If so then wouldn't it be cheaper to get a piece of copper sheet and tack it on?

Reply to
Tom Del Rosso

It's like the way they test the mileage and emmisions of cars. Maybe Volkswagon has the answer.

Reply to
Tom Del Rosso

Do you have experience of the aluminium PCB pricing versus standard PCB pricing? (I know it's a generalistic question...)

Cheers

Klaus

Reply to
Klaus Kragelund

no, I've never had the need for an aluminium pcb.

looking at that page, about the same price as 4 layer FR4

formatting link

-Lasse

Reply to
Lasse Langwadt Christensen

We've occasionally had custom shielding cans made, the expensive part is the tools to do the bending. The sheets are etched just like PCBs

-Lasse

Reply to
Lasse Langwadt Christensen

Injectorall Electronics has 0.01" 1 oz 2-sided FR4 in retail. They clam they can build 0.005" inch 4 oz 2-sided FR4.

I use the 0.01" 1oz PCB glued to a heatsink for projects with little hot SMDs. The catch, and it's a big one, is that it tends to wrinkle up like drying paper when the solder cools. There's some trial and error in finding ways to keep each board flat. With any luck, that's figured out by the time I have a final circuit ready to glue down.

BTW - Is that silicone glue? It's not a very good heat conductor and it's prone to never curing in thin sheets. At best you can clamp it with great force and hope that atmospheric pressure holds it together after the edge cures.

--
I will not see posts from astraweb, theremailer, dizum, or google 
because they host Usenet flooders.
Reply to
Kevin McMurtrie

I've emailed our usual board supplier to see what they can do, and begged for some thick-copper FR4 scraps to play with.

Now I want 12 oz copper with aluminum nitride dielectric. Seriously, one can buy small slabs of copperclad AlN, but I don't know if anybody makes PCBs that way.

It's ordinary epoxy, pressed down during cure to be thin. It's thermally minor compared to the 0.062 FR4. We might use thin compliant gap-pads in production.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com
Reply to
John Larkin

You can probably keep the die below its maximum temperature rating if that ludicrous number suggests that the thermal resistance junction to case is really nice and small (if 300 watts spec, then it most likely is), but then the question is... How hot are you willing for the part to run in the real world day to day ? The hotter it is for loing periods, the shorter its life is bound to be as you know.

I don't suppose you can parallel 2+ devices and reduce that dissipation to a much lower number ? 40 watts out of a TO220 or D^2 pak is a lot to ask for a production unit. Maybe this isn't for production in which case, you might be OK.

Dividing the current into two parts should reduce each parts' dissipation to 10 watts disregarding switching losses. Don't know if this is all conduction or switching loss related ?

boB K7IQ

Reply to
boB

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.