I'm trying to be pragmatic about designing a PCB and circuit that will have to dissipate heat. It's pretty simple: a surface-mount power transistor charges up a 4F 5v supercapacitor with a roughly constant current of 2A. It will not do this repeatedly - so it only needs to dissipate about 20 joules over 10 seconds, once, from room temperature.

I like the idea of killing two birds with one stone: keep calculations simple by building in safety margins and taking worst-case simplifications. But calculations or thermal dissipation over time, especially combined with electrical dynamics is not simple, mainly since the PCB's copper tracks and their layout have such a significant effect, but also since the transistor is not ideal.

So I imagine the only two practical ways of designing this are one or both of: with specialist CAD software, or testing of a physical prototype, simply measuring the temperature parts of the PCB and its components reach.

Realistically, what would be the simplest, effective way of doing the electrical and thermal calculations/assessments necessary?

  • \$\begingroup\$ Why do you expect to have a full 1 V across the channel (assuming it's a FET)? \$\endgroup\$
    – The Photon
    Commented Nov 8, 2016 at 3:16
  • 1
    \$\begingroup\$ A link to a datasheet of the transistor you want to use (or one you're considering) would help give us something concrete to talk about (like an SOA graph). \$\endgroup\$
    – The Photon
    Commented Nov 8, 2016 at 3:18
  • \$\begingroup\$ @ThePhoton - good point - I will work on coming up with specifics. So it sounds like you would advocate logically looking for dominant factors and manually working it out from first principles, rather than CAD or just try it and see? \$\endgroup\$
    – Jodes
    Commented Nov 8, 2016 at 3:36

1 Answer 1


20 Watt-seconds/10seconds * (Rjc+Rcs+Rsa) = 2[W] * Rth(total)['C/W] = temp rise

I^2*RdsOn is one of your choices in addition to choice of mechanical thermal resistances to apply the thermal version Ohm's Law for convection based on datasheets and after you define the max Tj rise for reliability concerns.

  • \$\begingroup\$ But this wouldn't take into account the thermal mass of the PCB and components, or the increasing effect of convection as temperature rises. from what I can tell, the thermal mass could play the biggest role, and the changing temperature gradient would need to be calculated? It just seems to spiral in complexity! \$\endgroup\$
    – Jodes
    Commented Nov 8, 2016 at 3:34
  • \$\begingroup\$ Of course but then you have not defined anything yet, have you \$\endgroup\$ Commented Nov 8, 2016 at 3:37
  • \$\begingroup\$ True - and I apologise - but I was hoping for a more abstract thought on approaching this kind of problem \$\endgroup\$
    – Jodes
    Commented Nov 8, 2016 at 3:38
  • \$\begingroup\$ Start with steady state then determine how much thermal time constant U need if you blow the budget for T rise.. \$\endgroup\$ Commented Nov 8, 2016 at 3:45

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.