Speaking of energy balances and negative thermal feedbacks, I've got an SMD board in a sealed box, in the sun, with a need to extract the heat out from inside as best one can.
Just plunking the board in a sealed metal can gives a 71oC rise, even for the very modest total dissipation (presently ~7W).
Obviously the thing needs heat-sinking. Regrettably, all the heat is generated at nodes that are electrically "hot," power nodes swinging
dissipate about 3.5W.
I suppose I could insulate the bottom layer and take the heat out there, but I don't like capacitively coupling EMI to the case. I also don't like bringing two nodes that close that, connected, would break the device. I don't want its 'life' depending on a silpad or a mica washer either.
| || || | .--------------. .---------------. .-|--||--||--|-. | | | | | '--''--''--' | --. .---------------------. .----. | |= | | '----' | Q1 |= | | | | |.-------- |= | | .-----. || |= | | =| |= .-------. --' | | =| Q6 |= | | | | | | L1 | =| |= | |R4 | | --. | | =| |= '-------' |= | | '-----' || Q2 |= | | | | |'--------- |= | | | | | |= | | | | |
---' '---------------------' | | | | | | | | C1 C2 C3 | | .nnnn. | | .nnnn. .nnnn.| | .--..--..--. | '---| |-----' '-| |--| |' '-|--||--||--|-' .---| Q3 |---------| Q4 |--| Q5 |-..-|--||--||--|-. | | | | | | | || '--''--''--' | | 'uuuu' 'uuuu' 'uuuu' || | | || | | .----. .----. .----. || | | |----| |----| |----| || | '------| |--| |--| |-----'| | .------| R1 |--| R2 |--| R3 |------' | | |----| |----| |----| | | '----' '----' '----' | | | | GROUND | | |
I think I'm going to leave the hot stuff swinging hot on top, and use the five layers underneath as ground, with a ring of vias around the periphery to help move the heat from layer #2 to layer #6.
This way the heat only has to travel 0.010" in FR-4 (from top to layer #2), before getting a copper-assisted boost to the bottom (layer #6). 0.010" instead of 0.062" cuts the FR-4 path by a factor of five or so. The five Cu pours also create lots of spreading, which spreads the vertical thermal path through a much wider patch of the board. That helps too.
Net result should be ballpark 2-3oC/W top-to-bottom for this 1in^2 patch, which I can then safely tie thermally to the enclosure for another