silicone grease is messy

Cured silicone rubber has a similar thermal impedance, if memory serves.

There are thermal interface pads made of mineral-loaded urethane or silicon rubber as well.

.

Joe Gwinn

There used to be a product called Bergpads or something like that.

Thermal gap pads have come a really long way since the original fiberglass Silpads. Nowadays they come in both elastomeric (rubbery) and clay-like consistencies, with thermal conductivities up to nearly 10 W/m/K, versus 0.85 for white thermal paste and something horrible like

0.5 for the old-timey Silpads. We have a couple of products that rely heavily on that.

One is a laser / TEC controller, smaller than a credit card, with a two-sided load and a fair number of 2-3 mm tall inductors and capacitors. There isn't a lot of space for thermal pours, so we use small ones and embed one side of the board in a 3.5-mm thick gap pad to get the heat out.

We use the clay-like ones for that, because it more or less eliminates the bending stress on the board, which might lead to reliability problems otherwise. The material is also slightly rubbery, so that plastic flow doesn't entirely eliminate the compressive preload, preventing delamination under temperature cycling. (They're just about perfect for the job, which I never thought I'd say about any thermal interface material.)

The other is a SiPM front end with bias regulation and TEC control. It's even smaller, but the real parlor trick is that it has to work in high vacuum, inside a SEM chamber. (It's used for cathodoluminescence detection--cool gizmo.) Thus it has to have very low outgassing, which this one does. (It's one of the elastomeric kind.)

(JL uses gap pads too.)

Cheers

Phil Hobbs

The problem with gap-pads and the phase-change things is that they don't squash thin, and don't conduct heat very well, so have a lot of net theta. Silicone grease squashes down to micro-inches, basically metal-to-metal for flat surfaces.

Some people don't want any silicone grease in their clean rooms.

I was thinking about something that flows like grease but eventually sets so is not messy. 2-part or heat cure. I was just wondering if anyone has used stuff like this.

I can get six TO-220 mosfets on a copper CPU cooler, with a custom AlN insulator. I want max power dissipation. It's a tricky packaging puzzle. TO-220 0.24 in^2 2 w/m-K 5 mils thick >> 0.4 K/W

times 50 watts is 20 K.

To spread heat through conduction, there's springy clips, waxy pads, rubbery pads, glues, solder, greasy compounds, and fluorinert. Every vendor claims theirs is 'something better'.

Every one of those solutions is 'something better' than the old mica insulator and grease of yesteryear. In mass production, I'd imagine the micropipet and glue option is attractive. Is that wrong?

And with good reason.

I think that there are alumna-loaded urethane rubber materials that set when catalyzed, which might work.

Master Bond and Henkel would be places to start.

Those AiN chicklets are very good. Some people attach them with silver-loaded epoxy, being careful not to make an electrical bridge in the process.

Joe Gwinn

Go figure. ;)

But remains removable, I gather? Because there's really good silver epoxy available, such as Diemat DM6030HK or Epo-Tek H20E. They're used for die-attach of expensive chips that you don't want to braze.

Cheers

Phil Hobbs

I plan to use a single big AlN insulator with six holes, with the six mosfets bolted down. I have a source that will make custom AlN parts like this for a couple dollars each.

Mica adds a lot of thermal resistance. So do pads. I plan to insulate with AlN, which conducts heat about as well as an aluminum alloy. But it would still benefit from some gap filler gunk, and silicone grease is messy.

Thermally conductive epoxy would work but is nasty to rework if a fet dies.

I guess I'll have to contact a bunch of goo suppliers. I was hoping someone here had suggestions.

Loctite 384:

Not the best thermal specs, but definitely removable. That's what they mean when they say "controlled strength permits field service and repairs" I've been having a hard time finding it test days, so let me know if you track down a distributor.

-Jim M.

Repairable Loctite is an interesting idea. That one has a thermal conductivity of 0.75 W/mK, but that's tolerable if it squashes out very thin under pressure.

0.0005" thick, under a TO-220, would be about 0.1 K/W.

Gotta watch out for delamination, though--the shear stress goes as the reciprocal of the bond line thickness.

Cheers

Phil Hobbs

Indium foil works pretty well--it's super ductile, and you can get little handheld rolling mills for a couple of hundred bucks, so you can make the foil any thickness you like, within reason.

Cheers

Phil Hobbs

For our low-outgassing application, we are using Loctite Stycast 2850 Catalyst 24LV .

2.3 w/mK is good. We could use that to stick the big AlN insulator to the CPU cooler, which wouldn't need to be reworked. That leaves six TO-220 mosfets to be thermally bonded to the top of the insulator, which I prefer to be re-workable.

Is 2.3w/mK = 2300w/k?

no, the "m" is meters, w/mK = Watts per meter-Kelvin

w/mK is the unit of specific thermal conductivity of a material. Pure copper is around 400. AlN is close to a typical aluminum alloy, ballpark 170. Still air is 0.03. Silicone grease is maybe 0.7.

w/K is the unit used to measure a given thermal conductivity in some specific geometry, like a transistor on a heat sink. 1 w/K is a good number there.

It is unfortunate that m means meters and milli. K is Kelvins and Kilo. Gg is a giga-gram. Mm is 1000 Km, which could be kilometers or Kelvins.

Dashes help, as in W/m-K, as long as dash doesn't mean subtract.

So the MKS system (MegaKelvins, Siemens) has ambiguities, just not as bad as imperial.