countersinking a TO247 mosfet

mandag den 19. juni 2023 kl. 20.09.37 UTC+2 skrev John Larkin:

like this?

mandag den 19. juni 2023 kl. 22.37.26 UTC+2 skrev John Larkin:

I'd be worried about the wedge effect of the countersink

maybe xray or decap a transistor to see how much room there is for a counter bore, I think I've seem some manufacturers say that a washer must be used

Yes, I'll probably use K199. It measures about 0.25 K/W. It's miles better than machining a heat sink and adding a separate fan. It's copper, too, a great heat spreader.

I plan to remove the spring thingies and bolt the cooler hard to the PCB, with the fets on the bottom side. We tried squashing the fets metal-side-up, between the cooler and the board, and I didn't like that. Fet contact force would be iffy and nothing would be inspectable. The thermal grease situation would be ikky.

With fets bolted on the bottom (with a big hole in the board) each fet is accessable, including for temperature measurement and possible rework.

I'll break one open, but I don't think the hole is near the silicon.

That epoxy is HARD. It was tough to countersink.

Thermal expansion on the plastic is going to be what, 5-10x that of the metal screw (SS is best here) so each heat/cool cycle is going to mash the countersunk screw into the plastic a little more, then loosen it on cooling. It all depends on the flow characteristics of the plastic well below the yield point. Bolt a complete assembly together and cycle the fet power from zero to max with the fan on, holding at hot and cold for as long as you think appropriate, and rinse and repeat as many cycles as you can. Check the torque (either loosening or tightening) on one of the screws every 100 cycles, every 1000 cycles on the next screw, every

tirsdag den 20. juni 2023 kl. 18.15.01 UTC+2 skrev John Larkin:

you have room on the top side? if you space out the cooler and mount the TO-247s through a hole with the legs surface mount they only stick up about 2.5mm

The countersink for M3 is 90 degrees, that for #4-40 is 82 degrees; with the right countersunk washers, head is flush with the washer top, so total height is the washer thickness.

It does NOT need any modification of the MOSFET, it's just a washer with a flush-to-surface screw.

I would get a piece of metal and use that as a clamp over the top of the FET package. The metal can be countersunk in regions either side of the FET but not over the FET package. If you want to get fancy you could start with a thick plate of metal, maybe stainless that is 1mm thicker than the transistor package, and have a recess machined in it for the FET package to fit into, and drill the screw holes either side of the recess andcountersink from the other side. You may say that this adds manufacturing cost, but you can easily outsource an order of metal clamps to be machined by the lowest bidder, whereas machining the FETs will not be a job many companies will want to do for you, and when they mess it up it will introduce delayed failures. Also when a countersunk FET blows, you can't just order a new FET and fit it, and the technician will have to not over-do the torque, whereas with a metal clamp you can order a standard FET and the torque will be non-critical.

The classic solution is a small belleville washer.

.

Or a wave wave-spring washer.

.

Joe Gwinn

My preferred idea was

Machining, even tricky stuff, is no problem for us. Nor is specifying procedures for manufacturing to follow, like fastener torques.

We could have two stock numbers and corresponding bins, purchased fets and machined fets, with a drawing to define the machining. All that is standard operating procedures.

Thanks for the suggestions.

The stuff a TO-247 is made of is rock-hard. It was really tough to countersink with a hand drill. It's sure not a thermoplastic.

A flathead would need some repairable locktite or even an exotic countersink-type lockwasher, which would add height.

It's worse with beer.

Well, there is PEEK, but it's like $5 per screw.

.

With plastic, beware creep.

Joe Gwinn

Amazon bought out Small Parts Inc some years back, and still sells things like Delrin screws, way cheaper than elsewhere.

Dunno if it’s all NOS or if they’re still making them.

Cheers

Phil Hobbs

Keep both elbows on the bar.

If we bolt the fets with plastic screws, they change from being liabilities (shorting to the adjacent board) to benefits (insulating standoffs!)

onsdag den 21. juni 2023 kl. 16.44.24 UTC+2 skrev John Larkin:

if you have space on top you could bend the legs on the fet and push them further through the board

.

Acetal is very slippery, so it may be difficult to prevent such a screw from working loose.

But acetal does not creep under steady load.

Joe Gwinn

It also doesn’t get brittle when it dries out.