Heatsink for QFN

[Emil]

I'v got a 5*5mm QFN package with a thermal pad. I can't (easily) make the board large enough to provide effective cooling, so I'm looking for a heat sink.

I would prefer one that can be soldered to the back of the board, to get the lowest possible thermal resistance from the die to the heat sink.

I know there are solderable heatsinks for some smd packages, but I'v never found one for QFN or similar packages. Any suggestions?

[HalFET]

I've had good experiences with the Parker T-wing heat spreaders, but you kind of need a lot of space above your board to put those so not sure if it's a viable option. A better solution might be to use a thermal interface pad and to use the structure you mount the PCB to as the heatsink.

[wraper]

I would prefer one that can be soldered to the back of the board, to get the lowest possible thermal resistance from the die to the heat sink.

Given PCB thickness and that thermal transfer will mostly happen through vias, heatsink placed on top likely will produce better result due to being closer to the die and package material having lower thermal resistance than FR-4. Also on the bottom you may find hard to make big improvements as available copper layer already dissipates heat transferred to it well enough. Absence of SMT heatsinks  may be a good hint it's not a good idea. If there is some cooling solution on the back, then it's usually thick thermal pad between metal plate (part of chassis/enclosure) and PCB.

[Emil]

Given PCB thickness and that thermal transfer will mostly happen through vias, heatsink placed on top likely will produce better result due to being closer to the die and package material having lower thermal resistance than FR-4.

Do you have any data on the thermal reistance to the top of the package? I have searched but not found anything yet. The app notes and datasheets I have read covers heat transfer to the PCB in detail, but doesn't say much about heat transfer trough the package. The most specific I found is in AN862: "upwards of 70% of the generated heat flows to the
air through the board. The presence of metal planes in the test PCB can increase the value to over 90%".

Also on the bottom you may find hard to make big improvements as available copper layer already dissipates heat transferred to it well enough.

That assumes the PCB is large enough to disspiate the heat, sometimes there is not enough space for that.

[wraper]

The most specific I found is in AN862: "upwards of 70% of the generated heat flows to the
air through the board. The presence of metal planes in the test PCB can increase the value to over 90%".

Of course most of heat flows through the board, IC area is tiny, it cannot dissipate much by itself. Using heatsink on SMT components is not a thing in most cases. However it works reasonably well on PC motherboard MOSFETs, FPGA, ARM and similar CPUs, GDDR RAM, fast ADCs like in oscilloscopes, most of which have plastic package.
The question is, do you even need a heatsink, how much heat needs to be dissipated?

[apis]

Something like this perhaps:

https://www.adafruit.com/product/1493

[Emil]

The question is, do you even need a heatsink, how much heat needs to be dissipated?

I need to dissipate 3.4W.

(125C max - 40C ambient) = 85C. 85C / 3.4W = 25 C/W max.

The best I can get with a 4 layer PCB is about 30C/W (die to ambient), assuming atleast a 50*50mm board. Add some air flow and that could be good enough, but the available board area is about 16*21mm, so I won't get close to that.

Edit: 25C/W - 6C/W junction to pad - 2C/W vias - 1C/W(?) board to heat sink + 4C/W PCB = 20W/C. So in addition to the board I need a heat sink with a thermal resistance of 20C/W or less.

[T3sl4co1l]

Is the package thermally enhanced, i.e. shiny on top so you can get benefit from a heatsink?

Even if not, there's some value in trying, but it's less obvious how much.  (Some TI parts have thermal characterization parameters now, that show the temp difference between an external measuring point and TJ, and thermal resistance to the top, leads, etc.  Very useful for figuring these things out!)

A heat spreader alone may be helpful, in which case this is an opportunity to use copper foil/sheet, or even pyrolytic graphite (almost as good a thermal conductor as diamond -- but only in-plane!).  You still need grease and clamping.

More likely, just a self-adhesive IC heatsink would do.  These are cheap and common, in fin or pin styles, and will handle that power easily enough.  If the heatsink needs to be a bit larger (or put up with some abuse), you may find it worth gluing down as well, or getting one with retention clips so it stays put (in which case you can also put other interface material under it; soft rubber pads up to 6 W/mK are available nowadays ).

Tim

[Emil]

Is the package thermally enhanced, i.e. shiny on top so you can get benefit from a heatsink?

No.

Even if not, there's some value in trying, but it's less obvious how much.

Maybe in addition to the board side heat sink if I can't dissipate enough heat with that alone, but I would like to leave that as a back up solution.

More likely, just a self-adhesive IC heatsink would do.  These are cheap and common, in fin or pin styles, and will handle that power easily enough.

I'm looking for some self-adhesive heat sinks to try, I have about 12*17mm space on the board, but no hard limit on height. It's also ok if it spreadsout a bit a few mm up, when it's clear of other components on the board.

But I'm leaning toward trying an Aavid 573100D00000G first. It's intended for DPAK/TO-252, but I think it will work.

[T3sl4co1l]

If you have the room for larger heatsinks and a squishy thermal pad, you can suck a lot of heat out through pours on the top side as well.

In other words, pour copper on the top and bottom and inner layers, and stitch things together, then put a squishy thermal pad on top, that presses down to the board, around all the components.  Clamp the heatsink on top of that.

These thermal pads are soft enough to clear nearby QFNs and 1206s and such, so you can get real good contact with them!

Downside, they're silicone based so adhesives are useless; you need clamps or clips or screws or whatever to hold the heatsink down.

One nearly ideal assembly method is thermal pads on both sides of the board, clamping it in the middle of two heatsinks or cold plates.  I haven't even had an opportunity to do this yet, but some day.

Tim