IGBT Losses / High Voltage / Forums

Forums

4hv.org :: Forums :: High Voltage

« Previous topic | Next topic »

IGBT Losses

Move Thread

LAN_403

ScotchTapeLord

Mon Feb 22 2010, 04:21AM

Registered Member #1875 Joined: Sun Dec 21 2008, 06:36PM
Location:
Posts: 635

I have some fast IGBTs, and I'd like to run them at up to 200kHz. My issue is that the switching loss for this model is 6mJ. If my basic calculations are correct, that would mean 1200 watts of loss...! I realize that the test was done at 600V and 100 amps, and that decreasing the values should reduce the losses drastically, but I am not sure how much.

I found a post by Dan with a formula for calculating switching loss, but it doesn't account for fall times and it conflicts with a this pdf from fairchild

which says losses decrease as voltage increases... which baffles me. A lot. I realize I'm failing to interpret something. I hope to have these IGBTs running a small DRSSTC. I realize the soft-switching will help a lot. I read somewhere that it reduces losses by about 25% from hard switching?

Are the wattages really that high? I would imagine they would be and that IGBTs are capable of handling the overwattage for such short amounts of time, but could someone clear up my misconceptions? Thank you. The IGBTs are FMG2G100US60 and I am going to try to parallel 2 full bridges of them if the single bridge works well at moderate power.

Wolfram

Mon Feb 22 2010, 07:41AM

Registered Member #33 Joined: Sat Feb 04 2006, 01:31PM
Location: Norway
Posts: 971

Firstly, those IGBTs seem quite fast, so I don't think 200kHz is out of the question. Their peak current rating is a bit on the low side though, but if you already have them, I would definitely try them out.

Remember that the 1200W of switching losses you calculated is a peak figure, and that a DRSSTC operating correctly is usually not on for more than a few percent of the time. The switching losses will only amount to a few tens of watts of average power, in other words.

Anders M.

ScotchTapeLord

Mon Feb 22 2010, 08:39AM

Registered Member #1875 Joined: Sun Dec 21 2008, 06:36PM
Location:
Posts: 635

It just concerns me that at its maximum rating, 1200 watts should be expected from switching alone...
On that note, I'd prefer not to exceed the pulse rating of 200A. I don't think that should be an issue since it is for a small coil. I don't know how effective it'll be, especially with the gates being driven at 12V. I guess I should raise the gate voltage, seeing as threshold is 8.5 and gate overdriving is in vogue. Still, I'd prefer to stay in specs.

Luca

Mon Feb 22 2010, 08:55AM

Registered Member #2481 Joined: Mon Nov 23 2009, 03:07PM
Location: ITALY
Posts: 134

Hi,
SSTCs operate (ideally) at zero current switching, so, during the turn-on/off current should be a few % of the maximum current... As you said, swithc losses are strongly dependent on device current during the switching. Moreover, in DRSSTC duty cycle is usually very low, so, average power is proportionally lowered...
I would not worry about switch losses...

Regards,

Luca

ScotchTapeLord

Mon Feb 22 2010, 09:08AM

Registered Member #1875 Joined: Sun Dec 21 2008, 06:36PM
Location:
Posts: 635

Okay, great! I can't wait to post my results... Let's just hope the sparks come from the topload!

GeordieBoy

Mon Feb 22 2010, 10:40AM

Registered Member #1232 Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881

Most switching losses with IGBTs in hard-switched applications are a result of current-tailing when a large current is interrupted at turn off. Your datasheet "turn-off energy" figure most likely represents a worst case figure for that device in severe service!

As others have stated the sinusoidal current in a DRSSTC naturally falls to zero and commutates smoothly from IGBT to free-wheel diode some time before the IGBT gate drive is turned off. This naturally sweeps out the stored charge at turn-off so there are greatly reduced turn-off loses.

Turn-on losses are usually very low for IGBTs because they can switch on incredibly quickly. Just make sure they don't turn on faster than the free-wheel diodes can recover. Otherwise you will get a lot of ringing and voltage overshoots!

As others have mentioned the DRSSTC is a pulsed power application as well. This means that you can get away with higher losses during the "on time" than you might otherwise be able to tolerate in a CW application. I once tried some moderately fast 2nd generation ST IGBTs driving a 500W CW coil at 200kHz and they quickly overheated and burned out! Pulsed operation with a low duty cycle is quite forgiving of high losses because there is a long time in between bursts for the heat to be conducted away.

-Richie,

ScotchTapeLord

Mon Feb 22 2010, 03:26PM

Registered Member #1875 Joined: Sun Dec 21 2008, 06:36PM
Location:
Posts: 635

GeordieBoy wrote ...

Turn-on losses are usually very low for IGBTs because they can switch on incredibly quickly. Just make sure they don't turn on faster than the free-wheel diodes can recover. Otherwise you will get a lot of ringing and voltage overshoots!

The diode recovery is 90-130, the igbt turn-on delay and the rise time is 50. Does this mean I will have to implement something to slow turn on? That's a shame... will a gate resistor do the trick, perhaps with a schottky to bypass the resistor for turning off?

Dr. Mario

Tue Aug 03 2010, 04:48AM

Registered Member #3059 Joined: Tue Aug 03 2010, 04:09AM
Location: My turf
Posts: 18

Tying a higher-rsistance resistor together to the Emitter and Gate will do fine.
Use 100 kilo-ohms or higher, so you don't upset those gate driver chip. You don't
want those little things get quite moody... They just explode.

Also, important to consider: Peak IvF (Ampere+Voltage = Loss in frequency roll-off):
The higher requency you shove into the gate, the hotter those poor transistor will get, as
it tries to suck in lot of Amps in futile attempt in cancelling out roll-off losses. Check the
manufacturer datasheet REAL carefully. (They lies sometimes, though. Check for Veg loss plots
and other stuff totally related to frequency in that datasheet, answer's all there.)

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob