countersinking a TO247 mosfet

If it's a circular table, rotate it until all four legs contact at the same time.

Nah; what you want is to scale heat losses down as much as device sizes have been scaled down. The microfabricators have convinced you that the high power stuff SHOULD be small. Critters smaller than mice are generally cold-blooded because they can't maintain thermal gradients, because high power metabolisms don't work when small.

That obviously doesn't work all the time.

If it did I'd have to eat someone else's food.

Given that I'm designing a dummy load board, I don't want to save power. A 200 watt module will sell better than a 20 watt module.

Yes. I use it a lot.

One trick is to drill screw holes, but don't thread them. Simply screw the steel screw into the somewhat undersize hole in the acetal. Then the screw is not going to back out.

Joe Gwinn

If the legs are the same length, but the floor is warped, it does.

I presume anyone who can budge a table can shift a chair.

Most floors are pretty flat. The tables are usually the problem.

An afterthought: Solid rivets expand when set, and it's not clear that the semiconductor package can handle the resulting radial stress.

Joe Gwinn

I plan to blind tap four holes into the copper CPU cooler. There's not much thickness until the tap would hit fins. I don't know how a rivet would anchor into the cooler, and replacing a fet would be nasty.

Consider packaging the fets in a hockeypuck package, with controlled clamping force that doesn't have to be limited to an M3 screw...

Like here...

Yeah. Can't say that I am able to visualize the arrangement. So I was pointing out possibly relevant options, fasteners with 100-degree cone heads.

Is it the hope that this screw into the copper heat sink be non-metallic?

It might be simpler to use a steel screw into copper, insulated from the fet package.

Or press-fit a delrin plug into a roughened hole in the copper, and run the metal screw into an undersize blind hole in the delrin plug.

Joe Gwinn

lørdag den 24. juni 2023 kl. 00.12.47 UTC+2 skrev Joe Gwinn:

the hole in a TO-247 is already isolated

threaded insert, perhaps? Avoids using a nut. Put the screw in from under the PCB.

Card cage:

My concern is that the transistor mounting screws could short to the solder side of the adjacent board in the card cage, and I'm tight on spacing. If I don't c'sink the fets and use a plastic screw, it becomes a spacer, not a short! Zero clerance to the next board is no problem.

I've ordered some plastic screws to see how they feel. Apparently PEEK would be the best material, but they are like $5 each, so I'd prefer something else.

Insulation is not a problem with TO247's.

What if you countersink the washer?

If instead of a washer you use a plate with recesses machined into it for the TO-247s, the un-machined parts would stiffen it. Also more than one screw would vastly decrease the required stiffness.

You could glue or (otherwise secure) a piece of thin FR4 over the whole thing to provide durable insulation from the next card.

If you have sufficient clearance on the other side of the card, then you could put some washers between the heatsink and your board, which would give more clearance on the side with the mosfet pusher.

There's a neoprene washer under the metal one, so it's still kinda high. And I'd still have metal about to touch the next board.

But that idea suggests something: Use a giant plastic washer for the pusher, something that flexes a bit and avoids the 4-leg-table issue, c'sink that, and use a plastic screw.

Interesting idea.

Shorting is only a problem where the 'next board' is too close? Can you put a few strategically placed holes in that 'next board' so that a sticking-up screwhead avoids contact by going into a void?

Delrin might work for that big washer. In the space between the FET packages there is added depth available, enough that an ordinary flat-head steel screw would fit, and could be screwed directly into the copper heat sink, perhaps with a heli-coil (pure copper is pretty soft, so an insert may be necessary).

Joe Gwinn

The CPU cooler in the pic is tapped for the single big screw. My machinist guy says we can tap the copper cooler, and he did that one. We could also tap four smaller holes.

I do need an AlN insulator under the fets, probably one big thin custom one with four holes. I figure that will add a trivial thermal resistance, like 0.04 K/W per fet.

There will be a lot of possible neighboring boards, at least 10 ultimately, some very dense, and it would be a nuisance to lay all of them out to avoid screw heads on this one.

Plastic screws sound cool, if they work mechanically. I probably don't need a lot of hold-down force.

Yes. How much thread depth is available? If there is enough copper, then helicoils may not be needed.

With the four smaller holes, it would be one per countersunk FET package, so no metallic protrusion? Might be a bit to close for comfort.

One big, thin AlN sheet may be a bit too easy to break in production. Four FET-size little sheets may be easier.

Joe Gwinn