I am trying to calculate the power loss in mosfet,which the input current is a train of pulse. Can anyone help me to know what all parameters to be considered.

Thanks in advance.

 

When it is fully on then you have current squared times rdson. When the edge of a pulse is happening, the it goes through a range of resistances, instead of rdson, as the current is ramping from 0 to max or max to 0. You need to integrate one of the three power expressions, over the rise or fall time, i.e. V i or v squared over R or i squared times R. Then add the dissipation during the rise and fall times to the i squared rdson of the fully-on time.

 

When it is fully on then you have current squared times rdson. When the edge of a pulse is happening, the it goes through a range of resistances, instead of rdson, as the current is ramping from 0 to max or max to 0. You need to integrate one of the three power expressions, over the rise or fall time, i.e. V i or v squared over R or i squared times R. Then add the dissipation during the rise and fall times to the i squared rdson of the fully-on time.

To determine the rise and fall time dissipation it's generally easier to use V and I since the equivalent R is constantly changing during that time. Thus a simple linear approximation is to multiply the voltage times the ON current times the rise-time and then take 1/2 of that to get the average power. The switching dissipation is then this value times the ratio of the rise-time to the pulse period. Do the same calculation for the fall-time.

Similarly the ON dissipation is the ON power, as you noted, times the ratio of ON time to the pulse period.

The grand total dissipation is then the sum of the rise-time, the fall-time, and the ON time dissipations.

 

Yes, the integration can be avoided by a numerical estimation of the "average". This is valid because, hopefully, your rise and fall times are a relatively small percentage of the overall cycle. Any error from your estimation is diluted into insignificance.

If the cycle was in fact dominated by the rise and fall curves, you might need a more precise integration. But then you're studying in great detail a poor design that should be improved anyway.