Do I still need to heat sink my MOSFET given the following conditions?

[kanmaedexandzelbladex]

I would just like to ask about the need for heat sinking. Well, I have a boost converter and I'm planning to use a MOSFET as the switch which has 170W as its maximum power dissipation in its datasheet. It's Rds,on is 6.5 milli ohms max and let's say worst-case stress is when current flows through it at 10A with 80% duty cycle. Well, conduction losses are really small. The switching loss, I compute it as (1/2) Coss Vds^2 fs. Switching frequency is at 100kHz, and the output voltage of the boost is 45V so Vds = 45V. Coss is obtained from the datasheet and it's about 650pF. So the losses are really small and I don't think I even need to heat sink it since it is at TO-220 package which can dissipate heat to air with a maximum of 26W rate. My loss based on my computation is only less than 1W. I'm not planning to package my circuit, so it's at open air.

The problem is when I simulate my circuit in SIMetrix and I plot the instantaneous power vds(t)ids(t) in my MOSFET I get really low watts in general but I get huge power spikes at switching transistions. Like during transistions where it goes on-off or off-on, there is a 250W spike there. The spike is huge and goes down horizontally and becomes as low as 1W only after 0.2us. This I think is due to the switching loss. Do I need to heat sink my MOSFET? and umm, the 170W rating, is my MOSFET going to get broken? Should I get another MOSFET?

 

[FvM]

Your simulation can tell you an exact power dissipation number by averaging switching losses, so you don't need to guess.

The other question is if your simulation is giving a realistic view of switching losses, particularly if rise/fall-times are modelled correctly.

 

[chuckey]

So you are dissipating 250 W for 0.2us and 1W for 5us , so in 5.2us you have 250 X 0.2 X 10 ^-6 + 1 X 5 X 10 ^-6 watt seconds or 50 + 5 watts/micro secs , mean power is 55/5.2 = ~10 W. Looks like a small heat sink to me, but you say < 26W is OK. (due to huge switching loss ~ 10 X "on" loss).
170 watts is OK, how about the max current for this device, as the transistor discharges the parasitic capacitance,its Vcc decays exponentially from Vcc to 0, and the current exponentially falls to 10A from VCC (volts across the Cap) / .0065, which looks huge!! (45/6.5 KA). Never done FETs at college (weren't invented in 1964!), so I guess there is some internal feedback mechanism that stops the current rising so high), but the point remains, how high is the peak current?
Frank

 

[kanmaedexandzelbladex]

Well, my MOSFET gets approximately 10A during turn-ON in which it has Vds approximately 0.065V and when it is OFF it gets 0A with voltage across it approximately 45V. I see, thanks guys. I was just wondering about the 170W rating of the MOSFET. It only says 170W at 25 degrees celsius but I am not sure what that 170W is, whether it is an instantaneous power rating or is it an average power rating. From your explanations, it seems like it's the average power rating which indeed if calculated is small. So I guess indeed a small heatsink will do. Thanks!

I have a question, suppose I would like to reduce the switching losses so that the efficiency will be high and that the heatsink will be smaller, how can I do it without replacing my MOSFET? Is putting a capacitance in series with the MOSFET gate going to help, like to provide base-current peaking for faster turn-on? And, when choosing a FET, should I look at the Coss in the datasheet or should I also look at the turn-on times and turn-off times, when picking out a more switching efficient FET?

I'm trying to get much ideas before simulating again because my simulations take long. Approximately 45 minutes.

 

[FvM]

I presume the 170 W is the continuous (average) power specification, but a rather theoretical number requiring an infinite heat sink. For pulse power, you'll look into the SOA and dynamical thermal resistance specifications of the datasheet.

Switching can be speed up by using stronger gate drivers rather than a series capacitors. But it's only reasonable if your circuit layout can handle the speed. In many cases, people are slowing down switching speed with series gate resistors to keep ringing and inductive overvoltages within acceptable limits. You won't see these effects in your simulation unless you model all circuit parasitics.