glob-top thermals

Depends on how hot the resistor gets!

My thought is that the area exposed to air is increased a bit by the spreading of the epoxy. A universe that is a hemisphere of epoxy is one extreme, a thin conformal coat is the other.

Underfill would be messy for production. If I annoy them they withhold donuts.

I have assigned a scut bunny to do experiments, using a 1206 thinfilm RTD as the combined DUT and thermometer.

The Vishay aluminum nitrides are an option but are hard to actually buy. AVX and some others have AlN, also hard to buy.

High speed design often requires pushing small parts hard. I need to know how hard is safe.

Well, 0.85 W/m/k is the same as ordinary paste. You could squirt that underneath after soldering.

Does his/her business card say "Scut Bunny 1st Class"?

Well, you could make the baby board out of alumina. ;)

Cheers

Phil Hobbs

Even better would be diamond grit. It is a good insulator, very good thermal conductor and is surprisingly cheap. Ideally use a mix of course and fine grades to get the best packing density and blend it with a small amount of low-viscosity resin.

John

FR4 is about 0.3 w/mk. There are PCB laminates that are 3.

But that's too easy.

I got some diamond-filled epoxy once. It wasn't a radically good thermal conductor. The diamond grains are still mostly separated by epoxy.

It seems to me that a glue I have an interesting circuit that unfortunately dumps 0.8 watts into

It seems to me that gluing a heat sink to the resistor would increase the air surface area more than the surface area of glob of epoxy . A quick sea rch on Aliexpress found some heat sinks that are 9 mm square by 12 mm high . But there are likely problems with the glue de laminating.

Dan

ly dumps 0.8 watts into

e air surface area more than the surface area of glob of epoxy . A quick s earch on Aliexpress found some heat sinks that are 9 mm square by 12 mm hi gh. But there are likely problems with the glue de laminating.

I was thinking little aluminium rectangles bent into a |__| shape, but just a bit of very thin flat ali would help, with a stiff insulating glue layer . Doesn't need much conduction to improve things.

Or you could go another way & solder a loop of copper wire over the top thr ough hole. That might even work.

NT

True; a glob, too, might encourage streamline airflow (which is BAD for heatsinking). The 'big parts' nearby could create eddies at their corners, which would help, and a bristle of fins adds conduction offboard to the mix.

a top heatsink would also conduct laterally, moving a bit of heat from the centre hotspot to the ends, resulting in better board heatsinking.

NT

An aluminum nitride slab could be glued to the top side of the parts and suck a lot of heat out. That would make the board into sort of a mysterious module.

That would be cool but a lot of work. Well, maybe not so much work.

What glue would you use? Removeable?

Interesting approach. Look at one end of the spectrum by using a thermally insulating epoxy as a very ineffective heat sink then careening in the other direction to consider using a rather more exotic and expensive material.

Oh well, he'll come back to center eventually. Aluminum is pretty nice stuff really. Thermal pad material is easy to use.

Rick C.

No, real epoxy. The idea of the little GaN modules is that they are throw-away components.

I have a source for cheap custom AlN pieces.

I agree that a bit of ali would make more sense. Use a copper wire loop & you don't need to glue anything on.

I still don't know why you're not leaving more cu on the board. I guess you need ultra low C.

NT

This is picosecond stuff. On a 4-layer board, with L2 ground plane, each square inch of L1 trace will be about 50 pF. And fast parts are tiny and get hot.

abuse an smd connector as heatsink?

Our mechanical guy wants to get one of these syringe-fed dispensing pumps

and chuck it into our Tormach n/c milling machine, and glob-top a whole panel of tiny boards at once. Or several panels.

You could dispense clear epoxy, which has some interesting applications.

Eg.

--Spehro Pefhany

We tried it. We put a 1206 platinum RTD on a 4-layer board, as a combined heater and sensor.

We measured 111 k/w in free air, and almost exactly the same after glob-topping. So the epoxy doesn't help thermally. At 0.8 watts, rise above ambient for a 1206 would be 89C. Globbing two 0805s doesn't look good either.

Looks like I'll have to wedge a gigantic 2512 resistor into the layout, and also add the new resistor to stock.

We'll still glob the boards for other reasons, but it won't help thermally.