cpu cooler off-label use

The cooling surface is attached to the fan fins. The tabs should face the cooling surface - not as illustrated in mock4.

RL

lørdag den 19. marts 2022 kl. 20.09.24 UTC+1 skrev legg:

the tabs are facing the cooling surface, mock4 is a bottom view, shows the transistors mounted through a hole in the PCB

lørdag den 19. marts 2022 kl. 20.04.20 UTC+1 skrev legg:

no

It's not completely ideal; the airflow favors cooling the side opposite the outlet. A ribbed plate with air flowing parallel to the ribs would ensure some heatflow uniformity regardless of lateral conduction in the plate.

For serious cooling, your fan can be one-per-box, and air channeled not by objects clamped to the printed wiring board, but by vanes riveted to the box. The tiny internal fan, even though it may be effective, is not as easy to inspect or replace as a muffin fan mounted to the box/chassis. Moving parts, substantial failure possibilities...

Nice idea, but am curious: Tradeoffs are to move more air over limited surface area or increase the surface area and move less air. Looks like the heat sink is flat, limited area surface. Limited time buffer from fan failure to destruction of the semiconductors - What is doing the over temp control - CPU or dedicated temp monitor/controller? J

Thermal theories are nice, but I believe experiments.

The little thing is an LM35 temp sensor. That gives us heat sink temperature. We'll know the fet voltages and currents, so that gives us power.

We'll run a realtime junction temperature simulation and shut down at some Tj, probably 150C. The sim will run in an FPGA.

I cut that on our classic Bridgeport milling machine. Just ignore the obvious goof.

lørdag den 19. marts 2022 kl. 21.41.31 UTC+1 skrev snipped-for-privacy@highlandsniptechnology.com:

murphys law also apply to the handles moving the table on milling machine ;)

I measured 0.24, which is pretty good in the thermal business.

The hot air will blow right out the back of the box. The main box fans might help a little too.

I might have a lot of topside copper under the heat sink, with many thermal vias to, well, somewhere. That would help a bit too, maybe

There are hunkier coolers around, but I don't have height for them.

lørdag den 19. marts 2022 kl. 21.58.36 UTC+1 skrev snipped-for-privacy@highlandsniptechnology.com:

you could also skip the hole and put the transistors with the tap side up on the top of the pcb, so the whole sandwich is: heatsink, transistors,pcb and on the back of the pcb a plate with the threads for the screws

The fets will have AlN insulators between their drain tabs and the copper heat sink. All electrical connections will be through the pins, sorta surface mounted.

We'll tap a hole in the center of the copper pedistal. A giant fender washer with some compliant stuff will squash the fets down, with the center screw applying the force.

I guess I shouldn't have any vias under that copper thing.

That might be hard to assemble.

We'll make a 3D printed plastic fixture to align the four fets, with their insulators, before we solder them down and then scrunch down the clamp thing.

That's only moderately hard to assemble.

mandag den 21. marts 2022 kl. 20.13.32 UTC+1 skrev John Larkin:

if you mount the transistor on top with the tap facing up, you just put something like this

It's the perfect design to make a crappy fan die faster than usual.

Those are some pretty optimistic numbers for a new fan, with a clean heatsink, in an infinitely cooled room at 25C.

Good luck with that thing either way, even at 35.

Here's the bottom of the cooler. The metal washer scrunches a neoprene washer onto the fets to push them against the copper heat sink.

I ordered a custom AlN insulator to complete the stack.