Whenever I have calculated (using my calculator) the power dissipation in switched-mode supply MOSFETs, I can’t say I‘ve been impressed by my accuracy.
How good is SPICE?

 

Spice quality = (the models X the operator).

 

Spice quality = (the models X the operator).

Trying models of similar MOSFETs from different manufacturers would give some clues about the former: can’t do much about the latter!

 

For most available MOSFET Spice models, and if you are properly simulating the SMPS, the power simulation should be close enough for engineering design work.
Edit: And likely better than any hand calculation using a calculator

 

For that matter, how accurate are the datasheet values of the power handling capability of 'typical' parts?

 

For that matter, how accurate are the datasheet values of the power handling capability of 'typical' parts?

The quoted power handling is likely somewhat unrealistic, but I think the stated thermal-resistance values for the parts are accurate enough to determine proper heat-sinking (or if it is needed).

 

A further note on Spice power calculations:
Spice calculates the power at each incremental timestep data point (instantaneous volts x amps) when it's doing a Transient Analysis, and these can then be summed together in the plot window to get the total average power during the simulation time.
Thus the power calculation is fairly accurate, since it determines the incremental power at all times, including during the transient device switching of devices.

 

That means that knowing the size of the transients is really important, so the overall accuracy will depend on having accurate values for such things as supply impedance and transformer self-capacitance, or running the simulator with a span of values for them and seeing which make a difference.
Thanks for your insights.

 

so the overall accuracy will depend on having accurate values for such things as supply impedance and transformer self-capacitance, or running the simulator with a span of values for them and seeing which make a difference.

Yes, obviously the accuracy of the simulation is dependent on the accuracy of all the components, including strays.

 

I added 20mΩ for the battery (24 cells at 830μΩ per cell for a 1000Ah battery), a similar amount of ESR for electrolytic capacitors and finally a film capacitor with very little ESR, and none of it made much difference to the overall power dissipation in the transistors. The current spikes became smaller. 1nF for transformer self-capacitance made no difference at all.
Next thing to explore is different half-bridge drivers, but very few manufacturers supply models. Analog Devices being one of the few exceptions.

 

Regarding LTspice's capabilities