4
\$\begingroup\$

I'm designing a board with some MOSFETs that dissipate a fair amount of power. The best device (low Rds and low Qg) available in this scenario would run at 30W max. This is available in D2PAK and TO-220.

What is the preferred option for cooling D2PAK at those power levels? Some of the options:

  • SMD stamped heatsinks on the drain pad, straddling the device (like these goofy things). Aavid 7109D resistance from mounting surface to air is 2.75C/W at 400fpm and 2C/W at 800fpm, this seems marginal in this case (since the connection between heatsink and device is thin copper conducting sideways).

  • thermal vias under device and a big heatsink epoxied to the board on the other side (one per device, not isolated)

  • thermal vias under device and a big aluminum plate heatsink on the back of the whole board (isolated with a pad)

  • aluminum core pcb or extreme thickness copper pcb. I'm guessing this makes the boards significantly more expensive, and prototyping more difficult. A metal core board would still need to connect to a heatsink somewhere, or the board would have to be extra big to give it enough heatsinking ability.

  • multiple devices to reduce dissipation per device. this almost seems cheaper than a complex cooling solution.

  • give up on using SMD and use TO-220 where <1C/W is fairly easy :) some concern about manufacturability in this case.

\$\endgroup\$
9
  • 3
    \$\begingroup\$ liquid nitrogen \$\endgroup\$
    – PlasmaHH
    Commented Sep 26, 2015 at 21:43
  • 1
    \$\begingroup\$ "The best device (low Rds and low Qg) available in this scenario" - what is the scenario? Unless you are operating the FET in its linear region I suspect your power loss is much higher than it should be. \$\endgroup\$
    – Bruce Abbott
    Commented Sep 26, 2015 at 22:04
  • 1
    \$\begingroup\$ I've only had to deal with up to ~7W with OPA551's in DDPAK; my solution was to use [two] parallel 551's mainly for geometric reasons (kept board profile low) and also because good a DDPAK heatskink (solid ones like Ohmite makes, not stamped) cost nearly as much as another IC of the kind I used. \$\endgroup\$
    – got trolled too much this week
    Commented Sep 26, 2015 at 23:00
  • 1
    \$\begingroup\$ I've prototyped the thing on 1.27mm-pads, square-pad protoboard (std. 35um copper though) to actually measure the thermals at various surface areas before doing any CAD work; I used straight, 0.7mm-thick jumpers in a grid pattern to make the heatsink copper areas on the protoboard. I did not just solder the pads to each other thinking that solder has much worse thermals, but a jumper between each row of pads. It turned out than my prototype had somewhat better thermals than the real PCB (probably due to the surface not being flat), but it's not currently manufacturable like that. \$\endgroup\$
    – got trolled too much this week
    Commented Sep 26, 2015 at 23:06
  • \$\begingroup\$ This is what I used to prototype/solder each DDPAK on (note, no holes). They also have (bigger) prototype boards with plated-through holes (e.g. this), but the asking price for those is rather steep (you can get an actual custom PCB done at those prices). \$\endgroup\$
    – got trolled too much this week
    Commented Sep 26, 2015 at 23:26

1 Answer 1

Reset to default
2
\$\begingroup\$

I suggest that if none of the 'standard' approaches work that you should look into one of these power peg thermal connectors.

enter image description here
(from here, also here and here)

They're by far the best solution I've seen for thermal management of SMD parts, but I've never had the need for them so haven't prototyped with them.

(No affiliation, just remember seeing them a while back)

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.