heat sinking onto diecast boxes

[Simon]

I'm working on a project that needs to do some power switching so need to heat sink a mosfet and what better way than onto the aluminium case the thing is in. But what are the surfaces of die cast boxes like ? are they generally smooth enough for heat sinking ? Perhaps I should consider a plastic box with a heat sink sticking out of it ?

[Rerouter]

it depends on the manufacturer, from the ones i have experienced over a short distance they are relatively smooth but can curve and be warped slightly over the length of the side, just remember if not insulated can be a way to blow up a mosfet with ESD,

[Psi]

Use a thermal pad, it will both insulate and provide some padding to make up for any non-smoothness in the metal.

[T3sl4co1l]

Don't expect to dissipate more than, oh, five watts in a reasonably sized (say 4" / 10cm) box, or more than 20W in a given component (in any size).  Sil-pads are pretty awful, and you'll need a fairly pliable kind (= less conductive?) if the surface is uneven as mentioned.

If you have access to a mill, you can of course get the surface cut as flat as you like.

Tim

[con-f-use]

If you're not worried about isolation, you can polish/grind the surface. For maximal conduction you can use silver thermal compound.

[Simon]

The aim is to save on costs so if I have to start getting them milled it will be cheaper to replace the lid with some punched sheet metal.

[HAL-42b]

Sil-pads seem like the way to go. Also mounting the mosfet to a L shaped bracket (bent sheet aluminium) and mounting that to the case may work also.

[Simon]

well I'm also looking for minimum space and parts, I have easily have a box lid/base punched with flanges for fixing it down and folded ends so that there is something in the air to dump heat, it should not get too hot but with nothing it will build up fast.

[qno]

How many watts do you need to dissipate?

[Simon]

I'm not sure as of yet, I'm going to use an IRFZ48V to drive a 30A fan with 500 Hz PWM power drive, I'll have to experiment but I'd like to at least get the mosfet "connected" to the outside world to help loose the heat and not overheat the enclosure. I am also considering separating the "power Drive" as i already have the signal generator solution sorted for brush-less fans that don't need power driving.

[fcb]

Diecast (eddystone etc..) boxes tend to have smooth walls (otherwise they would get stuck in the mold tools), so shouldn't be an issue.

The main problem is that they tend to be thin walls, this limits the thermal resistance somewhat.

As for insulation - probably look at bergquist gap pad, I have had a lot of success with this stuff, and it's available in some pretty thick formulations which can help.  The other thing I do now (for high dissipation) is not to bolt power devices down to heatsinks, but to arrange either a spring clip or a bar to press down on the device.

[Simon]

I think the average die cast box will work. It will need isolating anyway as I don't need the drain of the mosfet connected to a negative chassis as that will just run my fan all of the time. Of course I'll need a back e.m.f. diode too that might as well go in the case with the mosfet.

[SeanB]

So a small aluminium block with the 2 devices mounted on one side with thin silicone insulators and on the other side a similat larger pad to hold it to the box surface probably using 2 screws. This allows heat spread to the box, and gives a nice step to allow the devices to be on the edge of a PCB making mounting of the board easier. You mount the devices and do a quick test before mounting the whole board and block in the box. No messy pastes either.

[Simon]

given that I'm switching up to 30A I think it will be wires all around

[SeanB]

Much easier to use a PCB with well stitched traces top and bottom and wide as well leading to a set of screw terminals soldered to the board. Then a plug and socket or a simple cable connection to the driver board will make assembly so much easier and make assembly errors so much less likely. As well it will not stress the power devices with lead flexing and breakage every time the case is opened along with giving strain relief.

[Simon]

try calculating the trace width for 30A....... it is wider than the TO220 package pin pitch

[Christopher]

Check the material used on the box, I came into this the otherweek with an Eddystone box which was a zink alloy. A few tests proved it to be crap at dissipating 25W..


Quick way to check: Grab a ~10R TO220 Resistor (£5 on Farnell) and mount it to the case, it will dissipate all the power you put through it in a similar way to a Fet would I'd put double the power through the resistor you'd be expecting the mosfet to dissipate in steady-state. Attach a Thermocouple or three, datalog the temps and powers, now you have a Thermal Resistance value.

try calculating the trace width for 30A....... it is wider than the TO220 package pin pitch

Use 2Oz copper, 2 layers, loads of thermal vias? If 2 Oz copper is a prob solder wire strands to the board (not good for high-volume)

[Simon]

yes diecast is crap as material so yes I'd need to do thorough testing. Plan is it won't make too much heat with a 12mR mosfet channel but then only testing can verify switching losses and of course there is the back e.m.f. diode that will only be realistic with the real load.

[SeanB]

try calculating the trace width for 30A....... it is wider than the TO220 package pin pitch

That is why I said double sided...... plenty of space then if you use top and bottom and make a wide trace immediately after the device. The small bit near the device will be fine as the lead will heatsink it, and a wide trace top and bottom with a few vias will handle the current, They can be wide away from the device, and you only need 3 power handling wires in any case so can use pretty much all of the board for them. As well you can put the driver supply filtering on there as well, along with the fuses and transient suppression.

[Simon]

yes but i have a negative wire in to the source and a wire out from the drain to the drain, why bother putting than on the PCB and almost doubling the size of the PCB and making more heat. There may be larger versions going up to 70A with 2-3 mosfets in parallel, there will be no getting away from just wiring it up even if stiff copper (busses" are used).

[Neilm]

I've used a diecast box to dissipate about 20W (from memory from about 12 years ago). The biggest issue I had was the people assembling forgetting to install or incorrectly installing the SIL pads resulting in the FETs shorting out and blowing up.

[Simon]

What I'm thinking now is to use a plastic potting box and use small heat sinks like the ones I stock and pot them in with the fins sticking out of the potting, this will be cheaper and faster to assemble

[Zero999]

I'm working on a project that needs to do some power switching so need to heat sink a mosfet and what better way than onto the aluminium case the thing is in. But what are the surfaces of die cast boxes like ? are they generally smooth enough for heat sinking ? Perhaps I should consider a plastic box with a heat sink sticking out of it ?

I did this awhile ago on a bipolar tracking LM317/337 PSU.

I achieved best results by soldering the regulator tabs to pieces of copper bus bar and putting the insulator between the copper and the case. I can't ever remember the case becoming too hot to handle, even when the power dissipation was about 40W.

[Simon]

That is a large box, I'm going as small as i can. Given that I'd have to pot the thing anyway and that metal cases are much more expensive it looks cheaper to get an open potting box and just use cheap small heatsinks leaving the fins to stick out of the potting compound. I think you can get aluminium cases with fins on them that are open one side and you just screw any TO220 packages to the bottom (fined end) and pot the things up leaving a flat bottom but i bet they are not cheap and I've never seen them for sale, but have seen them in automotive stuff.

[scientist]

If you sand it or get it painted matte black it will dissipate a lot more heat by itself (no fan) than if you just leave it as shiny metal.