IGBT thermal calculations scaring me

I just went over the concept of dissipation and thermal resistance calculations for MOSFETs and opened a few IGBT datasheets do repeat the exercise for IGBTs.

Now for MOSFETS, at 24V and 35A with 20kHz switching, I got losses of say 3W max for many cheap MOSFETS (RDSon about 2mOhm). So even a cheap heatsink, say 18K/W would be 72C above Ambient.

For the IGBTS I saw, a general value of total switching energy Eon+off is 5mJ. So switching losses at 4kHz are 20W. Add to that about 2Vx5A=10W of conduction losses and you have 30W per IGBT....

Is this right or am I making a mistake (hopefully).

For 30W, even 3K/W of thermal resistance will make the IGBT unusable.

What am I doing wrong?

surely you can find igbt's with less than 3K/w therm res?

Your calcs sound possibly right, it depends what voltage you are switching though.
If its too hot then you need water cooled heatsink with fans blowing over it, or just stick with fets, or even parallel fets

I'm talking 3K/W total, junction to air through a really really huge and good heatsink.

Hi,

3W at 18K/W gives 54K, not 72K.

The rest is correct, as far as I can see.
***
See it from another point of view.
24V, 35A gives 840W. If now the losses are 3W, then the efficiency could be 99.7%.....

Klaus

Klaus the issue is the IGBT...30W. For MOSFETS its 3W.

- - - Updated - - -

OK just saw the 18x3 error.

But are you saying that my calculations for IGBT losses are correct, and even a tiny 3-5A current through IGBTS require monster heatsink.

The switching loss itself ruins it all. Add to it conduction losses and you have a TO3P size heater, putting out the same heat as my soldering iron.

there are probably a subsection of igbts that are better for higher frequency switching.
IGBTs can turn off short cct currents without dieing, fets cant

Its interesting that with mosfets you are dealing pretty much with rds on which is a resistance. Hence power increases at the rate of current squared. Adding more mosfets reduces both total power dissipation, and the dissipation per device.

IGBTs on the other hand produce more of a constant voltage drop. Its not linear, but the trend is that power tends to increase more linearly with increasing current. And adding more devices splits the current but does much less to reduce the overall voltage drop and total power dissipation.

Switching loss is a whole different issue.

At lower voltages mosfets will have far lower conduction losses, and adding more in parallel works especially well at reducing heatsink requirements.

IGBTs are pretty dismal at low voltages, and adding more in parallel hardly helps.

The situation reverses at high voltage. High voltage mosfets have higher rds on and much higher conduction loss. And power dissipation increases at the rate of current squared. You gain by adding devices in parallel but its a struggle.

IGBTs shine at high voltage, power goes up almost at the same rate as current.
This makes really high power a lot easier to achieve, especially with devices in parallel.

Moral of the story, work out your conduction and switching losses as you have done, and repeat with two or more devices in parallel.
Using two devices on a small heat sink may be a better solution than using one device on a smaller heat sink.

Then compare mosfets to IGBTs. There can be dramatic differences between all the various possible combinations, especially when you start looking at using multiple devices, and its all well worth the trouble.

Many moons ago, I helped an acquaintance to design a hybrid system, IGBTs and Mosfets.

If you feel that IGBTs are nowadays slow, you should have tried the first generation IGBTs.

Anyways, the trick is to connect IGBTs and Mosfets in parallel. The Mostet is driven to conduct briefly only during the IGBT's turn on and off periods, but driven fully-off otherwise.

It actually worked very well, although the drive logic was a little elaborate.

Just got up and understood the importance of taking exact values from graphs.

So Etotal at 5A is 2.7mJ and Vce is 1V.
So the initial 30W reduces to 15W.

So can I safely declare that the IGBT under consideration, FGA25N120 dissipates 15W when average Ic is 5A at 4kHz switching frequency?

Now my question is this.

What is the dissipation for each switch in a 3 phase full bridge, giving a 3 phase output using bipolar SPWM? I mean each switch is conducting throughout the cycle but still.

For 1A output line current, what should I take as the current per switch?

It is a fraction of the line current, determined by the conduction angle.