filling vias

The best Berg material is 5 W/m-k. A square inch of 20 mil stuff is

A 50 square inch PCB would have a net theta to the cold plate of 4 mK/w. That's 4K per kilowatt.

What? You can't use a PCB if your parts have a cold plate on them? What are you talking about?

He likes Berquist. It works ok and conforms to uneven surfaces, like the tops of DPAKs.

The pixie dust seeds the electroless plating, providing nucleation sites that catalyze the electroless plating, which grows between same to make a continuous film.

Typically palladium dust, IIRC.

But I'm no chemist. Bill Sloman could probably recite us chapter and verse.

Cheers, James Arthur

Sorry, I read "Or get rid of the board entirely."

I already suggested a sandwich, top and bottom, with FETs optimized for that, but John's right, that's hard to service or troubleshoot.

Sadly, heat sinking to the top of a D-PAK is next to useless. I measured about ~20K/W(?) on a monster FET recently, IIRC. (Oh, scratch that, it was a thermally-enhanced SO-8, which is a lot thinner. A D-PAK would be worse.)

Cheers, James Arthur

Service? What "servicing" do you do with DPAKs other than replace them? Trouble shooting is as easy as placing test points on the exposed side of the board. I have a board with all the active parts on the inaccessible side. I wasn't able to add test points and it was difficult to bring up. But if I could have added test points it would have been easy. A board filled with DPAKs would be easy to add a few test points and run with.

The sandwich on the other hand, *would* be hard to trouble shoot. Sounds like some of the mil gear I've worked on.

Maybe TO-247s are in order mounted with the heat sink side up.

_________________ Berg + cold plate ,--------,---Bottom of TO-247

==|============== PCB

number depends on the resistance of the heat sink. The package plus PCB

per board, you can more than triple the power per device given the lack of the insulating PCB. That pretty much makes up for the larger size and results in a whole lot easier unit to build. You also get cooling from the component leads to the PCB as well.

I bet his manufacturing people tell him it is hard to fill the vias

The heat comes out of the metal part. The bottom. The part people solder.

Ricky is obnoxious.

I'm still waiting for affordable bulk diamond. Imagine a diamond PCB. It would be beautiful.

Hard to drill the vias, maybe. Lasers might work.

But the heat comes out of the bottom of an epoxy power transistor, not the top. And that spread heat would still need to go somewhere.

We can do heavy. Here's one of our pick-and-place machines.

Den mandag den 5. oktober 2015 kl. 01.14.40 UTC+2 skrev John Larkin:

-Lasse

Den mandag den 5. oktober 2015 kl. 20.18.27 UTC+2 skrev John Larkin:

you can get parts that are reverse side up

-Lasse

Not every design is limited to the heat you can pass through a PCB.

ill, and

for

ed about

e.)

Thermally-enhanced SO-8s are 0.9 K/W, maybe some are better, and Vishay has parts with thermal contacts top and bottom.

That's why I suggested those to John earlier with a cold plate on top AND bottom, because John's just that kind of an extreme guy. :-)

He can get that or better with conventional assembly and copper-filled vias , saving solving the problem of mounting and heat-sinking SMD parts upside-do wn.

Cheers, James Arthur

I am the soul of moderation. If you disagree, I'll rip your liver out.

That's interesting. If I put them on the bottom of the board, the board could push them down into the gap-pad. The thermal footprint would be small, just what's on the package.

ke

sured about

s a

orse.)

has

I happen to know you brutalize mostly semiconductors. If I were a MOSFET I'd be worried...

Cheers, James

Hah, I mounted a power opamp like that, upside down and soldered copper tape to the thermal pad. across the ends and down to the pcb, worked fine, but not very much power.

George H.

Vias about 12 mil and under don't fill very well with solder, especially with lead-free which doesn't spread as well.

I'd go with 15-20 mil vias, preferably in a wave process, for best results. I don't know if you'll have good results with paste reflow or what.

Solder isn't nearly as conductive as copper, but it can make up for it with bulk. Glomming solder onto traces (that have been exposed with some solder mask openings) can easily double the current capacity.

Tim

Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website:

Has anyone filled vias with solder to improve thermal and electrical conduction?

I figure that if a via has an unmasked pad, and it's solder pasted and reflowed, the solder will melt and flow into the via, partially or completely plugging it with solder.

Maybe arrange for 2x the paste that would fill the via volume?

I found this online:

which seems to contain a lot of nonsense.

Seems to me that plugging a via with solder will radically reduce its resistances, and that lots of small vias are better than one big one.

twice the value for the TO-247 where there is nothing else in the way. So how can the DPAK be better even with silver filled vias?

You don't need to mount SMD parts upside down. The TO-247 is a through hole package. Mounting to a heat sink is not exactly a "problem". It is the *same* problem you need to deal with mounting the cold plate on the back of the board. It is also a "problem" that nearly anyone heat sinking parts soldered to a PCB has to deal with. Do you really think this is something that hasn't been done before?

m?

ide

-fill, and

go for

ured about

rse.)

has

ND

CB

vias,

e-down.

with SMD you just stick a reel in the machine and out comes a PCB

with TO-247 you'd have to bend the leads and hand solder

-Lasse