Cooling inside of small, sealed, plastic enclosure

[mtwieg]

I'm currently working on a project which is basically a 350W boost DC-DC converter and motor driver. Initially I was told the enclosure would be a certain size, which sounded fine, but now I've found that the enclosure is smaller, and is also a polyamide plastic IP65 (meaning it's sealed tight enough to be protected from dust, and from water sprayed at it from a nozzle). So I won't be allowed any airflow between the inside and outside, and the thing is plastic so it has poor thermal conductivity. And from what I can tell the inner compartment is only ~30 cubic inches (and I'm not sure how much of that I'm actually allowed to occupy).

Even if I through my best SMPS design at it, it will probably end up being up to 90% efficient, so I'm looking at >35W of power to get rid of, and I believe it's supposed to operate at 85C ambient. So I need about 1 K/W overall thermal conductivity. Back of the envelope calculations estimate the case's thermal impedance to be ~0.38K/W (purely based on its own material properties and size, nothing based on airflow or convection).

So at this point I'm wondering what, if anything, can be done to keep temperature under control. Heatsinks inside will be useful to reduce hot spots and keep things even, but what about getting the heat out of the enclosure? I don't know if filling it with urethane or some other conductive solid is an option, or how well that could work in practice. I've seen some enclosures where the heatsink pokes out of the enclosure, which would be great, but is it feasible to do this in a way that preserves the enclosure's IP65 capabilities?

 

[penrico]

What about using peltier cells? You can put a peltier cell inside your circuit, and the hot side of peltier exposed to the external environment, it will work as a thermal pump.

 

[dick_freebird]

I would say your only hope (if you can't modify the enclosure)
is to mate a nice slab of aluminum heat spreader to as many
of the internal faces as you can manage, with a thin and
high thermal conductivity adhesive. Get as much area into
conducting heat as you can.

Seems to me that somebody's clowning the project by
dictating a thermally inferior case design - why design
the case before you know what's in it?

 

[mtwieg]

What about using peltier cells? You can put a peltier cell inside your circuit, and the hot side of peltier exposed to the external environment, it will work as a thermal pump.

But this would still require punching a hole in the enclosure so that one side can be on the electronics while the other is outside. Same issue with a heatsink.

I would say your only hope (if you can't modify the enclosure)
is to mate a nice slab of aluminum heat spreader to as many
of the internal faces as you can manage, with a thin and
high thermal conductivity adhesive.

Yeah that might work better than a polymer/urethane filler, except that such a thing would have to be completely custom, and I don't know a thing about custom heatsinks. It could be feasible, but I don't know if the client will like it... at that point you might as well get the enclosure completely made of aluminum.

Seems to me that somebody's clowning the project by
dictating a thermally inferior case design - why design
the case before you know what's in it?

Well, it probably didn't occur to them that it would be an issue, which is why they contract out the engineering. Not sure who was in charge of that particular decision...

 

[dick_freebird]

Could you perhaps be allowed to breach the wall, provided
you maintain an acceptable seal in the end?

If so, then an aluminum plate on the inside and outside,
with a plethora of aluminum fastners acting as thermal
vias, and a sealant on both faces might work out.

But if it's a no-touch-ie kind of gig, I guess that'd be
off the table.

 

[mtwieg]

Right, having the heatsink pass through the enclosure would be great, but the whole sealing issue is completely foreign to me. Using fasteners or rivets as thermal vias between the inside and outside is a nice idea though. Maybe with rubber washers it would still be a good enough seal. No idea how I'd actually estimate the thermal impedance though...

Enclosure aside, has anyone seen heatsinks which are meant to be used in sealed applications? Like maybe with grooves carved in them for gaskets?

 

[mtwieg]

Update: I think I'll be able to use an aluminum enclosure. Now I just have to figure a way to thermally couple the electronics to the enclosure. The devices are several FETs in the DIRECTFET package mounted on the bottom of the PCB. There will likely be ~5mm of space between those FETs and the bottom of the enclosure. In my head, I want to just fill that volume with a flat slab of aluminum, with some thermal pad on both sides for filling and electrical insulation. But the issue now is how to exert pressure on that stackup in a way that doesn't require making any holes in the enclosure wall? The PCB will be mounted to the enclosure at the ends or corners, but if I tighten those I am afraid I might bow the PCB, which would of course be awful. Any suggestions?

Or might it be feasible to drill mounting holes in the enclosure and keep it sealed using rubber grommets?

 

[dick_freebird]

I have had occasion to customize some of my DC-DC
eval boards for in-vacuum testing. What I did was to
make an aluminum shim that fit the opposite side of
the board from the DUT (surface mount, many thermal
vias) slightly oversized, and relieved where there were
through-hole stubs to respect. Some thin Kapton film
made a decent insulator without too much thermal
impedance, between the board and the slug. I used
the board corner holes and equal height standoffs to
the aluminum slug, mounting to a larger aluminum
plate.

You don't really need a lot of force, just positive
interference. If I were not needing to swap samples
I would have used some of that "arctic" silver filled
epoxy between board and Kapton, Kapton and slug,
slug and plate, and make it all a solid stackup.

I might recommend tapping the slug and using
it, plus corner standoffs, to mount the board to the
case wall.