Cooling SMT transistors by gap-padding them to above-board heatsink.

Hello,
We are wanting to use the IRF6674 metal topped SMT FET in our 15W buck converter. The heatsink is above the PCB so we need to put “gap-pad” in between the FET and the heatsink.
However, the Metal Top of this FET is connected to the drain of the FET. Therefore, we will need to cover the metal top of this FET with insulating “sil-pad” sheet before we apply the gap-pad.

Supposing instead we had used a plastic topped FET such as the RJK0651 (datasheet below), then we are thinking we would then not need to use the sil-pad under the gap-pad. However, the RJK0651 , being plastic topped, would need us to then additionally gap-pad the PCB copper that connects to this FET, to the heatsink…..and that copper is at drain node potential in places…..so once again, we would need to cover that PCB copper in sil-pad before gap-padding it to the heatsink.

Can you think of any way round this, such that we could gap-pad the hot transistor to the above-PCB heatsink, without needing to use insulating sil-pad. After all, the sil-pad application will be fiddley and the assemblers may forget to apply it.


IRF6674 Metal topped SMT FET datasheet:-
https://www.infineon.com/dgdl/irf6674pbf.pdf?fileId=5546d462533600a4015355ec9f0d1a66

RJK0651 datasheet (plastic topped SMT FET)
**broken link removed**

Thin Insulating sheet (sil-pad)
https://www.bergquistcompany.com/thermal_materials/sil-pad.htm

Gap-pad (medium for conducting heat to heatsink from semiconductor)
https://www.bergquistcompany.com/thermal_materials/gap-pad.htm

The Gap Pads are just thermally conductive. I don't see any that are both thermally and electrically conductive. No need to use an additional insulating pad.

Thanks, but over time, cant the gap pad get logged with moisture which exudes out of the components (its all in a sealed enclosure), and then start getting electrically conductive?

Check the specs on the gap pad that you what to use for water absorbsion. My first pass guess would be "no", since everything is in a sealed environment. Most components don't exude moisture anyway.

The drain of the Buck converter fet is 28v max, so its not a high voltage, and there is a 5mm gap between the pcb and the heatsink, so maybe it'll be ok...but we just wonder about some impurity getting into the gap-pad and then we have a conductive "dirt trace" going between the fet drain metal and the heatsink?...remote possibility, but maybe their is some dogmatic regulation out there which speaks about this possibility?

I think that you guys are worrying too much. If there is a possiblity of a "dirt trace", then make sure that there is a fuze in the input that will blow in event of a "dirt short".

What kind of impurities are you guys imagining? The gap pad is non-conductive and in a sealed environment. What chemical magic do you foresee that would convert the gap pad to an electrically conductive pad?

Why not use standard PCB thermal pad techniques or SMT heatsinks for a 15W supply?

Why not use standard PCB thermal pad techniques

Click to expand...

Thanks, but that woudl involve thermal via'ing doen to bottom layer colling copper, and we cannot do that, as there are components eg processors on the bottom layer....the (h20 cooled) heatsink is above the PCB , and that is where we must connect to it.....the other side of the heatsink has the main "power component" on it

I wouldn't hold out much hope for the metal lid as a main
heat path, because access to it is only by the ceramic
package sidewalls (a thin cross section, long-way-around
thermal path). The main big bottom pad is where the heat
flux is supposed to head on out (die attached right above it).

I wouldn't hold out much hope for the metal lid as a main
heat path, because access to it is only by the ceramic
package sidewalls (a thin cross section, long-way-around
thermal path). The main big bottom pad is where the heat
flux is supposed to head on out (die attached right above it).

Click to expand...

Thanks, ive actually said that to our boss, but for financial and spatial etc reasons, the heatsink must go above the PCB, and components must go on bottom layer of the PCB.....so we cant have bottom layer thermal copper and thermal via down to it......

The heatsink is above the pcb because the "main load" is there and needs it there........the components on the bottom layer also need to be there for space reasons, -this is the smallest way to make the product, and smallness is key for this product.

I might suggest, then, that the device be mounted to
the PCB upside down and use copper ribbon to wire
from the now-up pads down to the PCB and then
use a minimal fill / pad with the heat sink in compression.

You may be able to find variants of the device with
ribbon leads already attached. The HiRel market is not
all that fond of leadless package bodies let alone
when it's a large body and prone to temperature
cycling and other board flexure. In fact last I'd heard
anything larger than SMD-1 was still verboten in
space systems.

If you need to make up for the lousy effective thetaJC
by an even larger heatsink, that's probably a "lose".

Maybe a quantitative thetaJL ("Lid") number from the
manufacturer would help your argument.

I might suggest, then, that the device be mounted to
the PCB upside down and use copper ribbon to wire
from the now-up pads down to the PCB

Click to expand...

Thanks, but this would mean too much stray L and ringing, which as you know, in a high current sync buck , is a bad thing.