Heatsinking needed for FET on FR4 PCB which lies flat on an earthed heatsink

Hello,
We have a 4 layer, FR4 power supply PCB which lies on a large earthed heatsink. We want to heatsink the power FET to the heatsink. It dissipates 13W.
We would like to have a cutout in the PCB, which will be underneath most of the D2PAK FET’s body and drain tab. We will put a piece of copper into this cutout, and then cover it with a layer of gap pad, we will then crush the D2PAK FET down into the gap pad, and then solder the top bit of the D2PAK drain tab to the PCB at one side of the cutout, then solder the D2PAK's legs to the PCB at the other side of the cutout.
This should give us fantastic thermal coupling to the heatsink, do you agree?
Are there special FET packages which are more amenable to this kind of connection?
(also will be a ultra thin insulating layer under the gap pad.)

We are even thinking of doing this with a TO247 FET and with a powerful soldering iron, solder its tab to the PCB at one side of the cutout, and the legs (with a little bending etc), to the other side of the cutout.
Though the D2PAK offers greater ease, since its legs and its tab are kind of “level” with each other.

Sounds like a hack to me. You might want to try a package like a TO-220 where you can bolt it directly to the heatsink.

Thanks, problem is this means we need a threaded hole in the heatsink, which makes that more expensive. We already have five threaded holes for the PCB, and its amazing how costly threaded holes are.

Avoid such practices and design like plague. Please think of the poor user and the engineer who may have to service your super fancy gadget.

Things connected at multiple points can create stress if the temp goes up and down on a regular manner. Solder joints may come out loose. Through hole plating can come off.

If you are really desperate, mount the FET on the heat sink and connect the legs with wires to the PCB. I know it looks ugly but you value peace of mind, right?

treez said:

Thanks, problem is this means we need a threaded hole in the heatsink, which makes that more expensive. We already have five threaded holes for the PCB, and its amazing how costly threaded holes are.

Click to expand...

An extra threaded hole sounds x100 simpler and cheaper than what you describe. And a lot more reliable.

You might also consider surface mount packages with isolated top side cooling, like Directfet, Dual Cool, etc.

treez said:

Thanks, problem is this means we need a threaded hole in the heatsink, which makes that more expensive. We already have five threaded holes for the PCB, and its amazing how costly threaded holes are.

Click to expand...

First of all, the process you describe is going to cost a lot more than one crummy tapped hole. Secondly, it sounds like you're expecting the solder to act as a mechanical fastener holding the device to the PCB. Terrible, terrible idea. It's just going to rip the copper off the pcb with the first hint of vibration. Third, it doesn't even have to be a tapped hole, it can be a through-hole and a nut (if you've got backside clearance).

I do not know why threaded holes are so expensive. Anyway, threads on a thin Al heatsink is not really strong and you can always use a nut and a bolt, as suggested above. If the thread gets damaged, you have to replace the board (if the heat sink is mounted fixed on the PCB)

Or consider using the FET mounting hole as the PCB support as well. It has to be done at the design stage of course but there is no reason why a screw can't pass right through the FET, PCB and mount directly to a heatsink so it serves a dual purpose. If you are having heatsinks manufactured to your own design it could incorporated as part of the product housing or have a raised step where appropriate under the FET to maintain clearance for the legs of any through-hole components or connectors.

Brian.

One more thing to think about. If your baseplate is aluminum and you stick a piece of copper on it, I think you're susceptible to galvanic corrosion.

barry said:

One more thing to think about. If your baseplate is aluminum and you stick a piece of copper on it, I think you're susceptible to galvanic corrosion.

Click to expand...

Possible only IFF water gets into the contact region. Galvanic corrosion is ionic in nature and you need an electrolyte for galvanic corrosion. Sometimes moisture condenses and causes some visible corrosion at the joint but I am not sure that it is galvanic in nature.

c_mitra said:

Possible only IFF water gets into the contact region. Galvanic corrosion is ionic in nature and you need an electrolyte for galvanic corrosion. Sometimes moisture condenses and causes some visible corrosion at the joint but I am not sure that it is galvanic in nature.

Click to expand...

That's why I said "susceptible". Of course, water is such a rare substance this would never happen.