Invert your thermal model to good effect

Electronic thermal simulations are most commonly formed by specifying a power dissipation within a package and using numerical solution techniques to derive what the resulting temperatures are within and around the package, PCB etc. If the package temperature is too high you’ve got issues that you’ll need to resolve via various thermal management approaches. You can turn the model definition on its head by specifying a package temperature as an input are getting a solution for power dissipation. From the solvers point of view it’s just two sides of an equation. From your point of view it provides useful information.

In an IC package heat is dissipated on the silicon die. This leads to the die reaching a certain temperature, the value of which depends on how easily heat can exit the die, through the package, into the PCB, the air, a heatsink, through an enclosure etc. This temperature at the heat source is called the junction temperature. It’s this value that is related to a specified maximum junction temperature and thus thermal compliance can be quantified.

Instead of a package power dissipation value (which is difficult to obtain accurately) you can set in your electronics thermal tool of choice (that will likely or of course be FloTHERM, FloTHERM.PCB or FloEFD) the silicon to be at the maximum junction temperature that the package is rated at. This value should be much easier to obtain. It should be, you were going to use it to judge thermal compliance anyway! Perform another thermal simulation and you can observe what power dissipation comes out of this fixed temperature silicon.

Such a power dissipation is the maximum power dissipation that could be dissipated by the package whilst the junction temperature does not go above its maximum value. Useful information in and of itself.

Some things to be aware of though… that predicted maximum power dissipation will be valid only for that package in that layout for that PCB in that environment. Any change to layout, environment etc. and you’d have to do a recalc.

Try it, turn your simulation on its head, see what comes out.

21st July Nottingham