Continue to Site

Welcome to

Welcome to our site! is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

SMPS PCB lying flat on a thick aluminium heatsink?

Not open for further replies.


Advanced Member level 5
Jun 13, 2021
Reaction score
Trophy points
Activity points
Please comment on this method of heatsinking SMPS FETs on a FR4 PCB which lies flat in a thick aluminium heatsink?...

We have been very very briefly shown a fleeting glimpse (by Non engineers) of a 400W SMPS to be worked on, but we didn’t get much detail as we haven’t signed NDA yet. (don’t even yet know what needs doing, they said it was “flyback” and there had been a problem with the leakage inductance, and that they needed methods to get through EMC, but wouldn’t say more)

Spec is…
Vin = 24-32vdc
Vout = 30vdc
Pout = 400W
F(sw) = 250kHz (I am guessing this freq)
isolated (assume to 500vrms, 500Vdc?)

Please imagine the PCB in an Alu box (as attached). But walls approx 5 times thicker, box dims are approx. 20cm x 10cm x 5cm. It is machined from solid. The PCB lies flat in the bottom of this.

PCB was green so cant be aluminium PCB.
We saw one planar transformer in there, approx 5cm x 3.5cm x 2cm, and what looked like an SMD output inductor approx. 2.5cm x 2.5cm x1 cm. Saw a few SMD Power FETs DPAK size. All components seemed to be SMD.
Component count seemed rather low, though we suspect the bottom of the enclosure is , in places, grinded out, leaving some room for bottom side SMD components aswell.

With such a thick walled aluminium enclosure, they must be using the enclosure as a heatsink. The DPAK FETs must be thermally via’d down to the underlying aluminium base. However, that still leaves the FR4 interposing, which seems a great shame, and a wasteful diminishment of that very thick aluminium chassis for heatsinking…

1.....As such, would you agree, a better method of heatsinking the FETs would be to have TO220 FETs, layed flat, and with a cutout in the PCB, so that there tabs could be screwed down to the metal base directly?....the legs of the TO220’s would simply be snipped quite short, and they would then lay along pads next to the cutout, to which they would be soldered. This should be ok for a 1.6mm thick PCB?

2.....Also, since the chassis will inevitably be connected to circuit ground, this means there will be a significant common mode EMC problem due to the switching node of the power FETs butted up directly to the chassis. What extra EMC measures would you take in order to mitigate this?

3.....Do you also agree, chassis connection to circuit ground must be only in one place. (to avoid ground current loops and thus EMC issues)?


  • box.jpg
    27.1 KB · Views: 97

1. Maybe better thermally but could introduce other issues
2. Look at putting a return path between earth and the Tab to eliminate the capacitance to earth preventing CM noise
3 If the supply is isolated you may need a CM Filter on both primary and secondary which may mean more than one chassis connection.

As their issue was related to leakage spike you best looking at the actual winding construction as this is the source of the issue along with a drain source capacitance of the mosfet. Also any Diode noise as a 400W flyback is pretty hefty and is normally replaced by less noisey topologies with lower switching currents.
Last edited:
1. Maybe better thermally but could introduce other issues
....Thanks, you are right, it amazes me that they dont make FET packages that are conducive to heatsinking to a metal chassis......i mean, when you have a very thick metal chassis like this, its a great shame not to use it for heatsinking......but the only method available is DPAK FET with "thermal vias in pads -plated over"...........and with the interposing FR4......a thermal insulator!

Another way may be to have the FETs on a piece of Metal PCB....and somehow Surface mount connect that to the rest of the PCB by using a cutout in the main FR4 PCB.

Another way, would literally to have a heatsink, and screw it to the chassis base, and then have TO220 FETs connected to this with spring clips or screws.

I confess to not liking having DPAK FETs heatsunk through an FR4 the PCB then needs hard tightening to the chassis....and hard screwing a PCB to a flat surface can make it bend....and then MLCC caps can crack or pop their sodler joints...specially if the chassis surface isnt truly flat, or the PCB itself is slightly bowed.

I once worked at an LED place, and they had SMD LED PCBs hard-screwed to a flat metal heatsink.....and we got constant failure returns with LEDs having gone open cct whilst with the customer...........must have been due to the slight flexion the solder joints experienced when the PCB was tightened to the flat surface....bad news, but the electronics industry just doesnt have answers for such things.

One thing is, we dont want to put tapped holes into the chassis for screwing the fet tabs into...since once a tapped hole gets shredded by clumbsy assemblers, the entire heatsink has to be chucked.......and this chassis looks expensive...machined_from_solid and with quite complex "contours" to it.........the chassis was so thick that i was looking for the channels into which coolant water might it may have been water cooled.
Last edited:

For the right sort of box design you can use gap pad and similar and compression "springs" ( not literal spiral springs ) to compress the board, Tx, fets, diodes and anything else that needs it to the bottom of the case as the lid goes down. This can be an EMC nightmare though for high dv/dt on tabs.
400W is likely too high for a single flyback, 2 x 200W interleaved would have been a far better choice - as would a regen snubber to the output - to take care of leakage - planar Tx's have awful capacitance - so turn on spikes in the main device will cause a lot of di/dt which gets into everything, not to mention the turn off dv.dt and snubber issues - if there is a serious output choke then it may be a fwd converter - but not a ZVS clamped one as that would take care of the leakage no worries and do 400W easily too as the rms currents in the main switch are a lot lower.

- good luck -
Thanks, they were non engineers...your points are very interesting and i actually now suspect they were possibly really meaning the txformer capacitance....will find out soon. Must admit, from what ive read , these Planar transformers dont impress me at all, from the point of view of hard switched converters. They look rugged against vibration, but surely a vibration proof bobbin can be tooled up for a "normal core and winding transformer"? (such as ETD, PQ, etc).

Thing about Planars, is you need multiple interleaving between pri and sec....and the inter-coil capacitance then gets really high as you say.......the turns ratios you can get also appear to be very limited. I am struggling to see their usefulness other than low profile, high power applications....but then some "normal" cores are quite low profile eg, some PQ cores.
Planars also suffer the "edge" effect w.r.t. gapping.

I really dont think i like them. It is said you can get really low leakage with a planar.....i mean, so what?....leakage is handy to an extent as a turn-on snubber. And you can get low enough leakage for eg a flyback when you use a "normal" transformer, by interleave winding.

I really just am not seeing benefit for these planars other than lowest profile, high power. Even then, i reckon theyre limited to low turn counts, [ie high f(sw)] otherwise interwind capacitance goes through the roof. I think these planars are a "fashion" item in many designs. "i want a planar on my CV, so i will jolly well have a planar in my next design"
Last edited:

Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to