IRS2092 High Power mosfet heat question

[ungie]

Hi all,

I'm working on a high power amplifier using IRS2092 with totem-pole driven pairs of IRFB4227 and +/-90V rails. So far I have everything working fairly well with low noise and THD and excellent sound quality. My issue is that at idle I feel the mosfets are getting hotter than they should. I have two channels on one board, so there are 8 fets mounted to an aluminum bar of a similar type to that used on the IR reference designs (IRAUDAMP1,7). This sink is getting to approx 55 degrees C after 30mins of the amp sitting idle, load connected and no input signal. The output filter is a 4th order type with 17uH and 11uH coils wound on T106-2 material. The inductors do not get more than slightly warm to the touch. The switching waveform looks ok, but I have not fully opimized snubbers yet. Any pointers on where to start looking would be appreciated.

[ungie]

I forgot to mention that the switching frequency is set at 420kHz.

[Andrew Eckhardt]

You'll have to post data or pictures about the heatsink before anyone can say much. The mouting interface (bar or block) is not a heatsink.

Quiescent power consumption is good info too. If its taking 30 minutes to reach max temp you have plenty of thermal mass in the mounting block, but you're lacking surface area (Fins!) if it's getting all the way to 50C. That much is sure.

[stoc005]

I would only use 1 pair of devices, not 2 pair. Also, play with DT. Increase it. That should help reduce Pd but it might cause THD to get rise.

[nigelwright7557]

I found my 2092 design got hot at 400KHZ so I reduced it.

Have a play with deadtime too.
I use max dead time with my irfb4227's.

[ungie]

Hi Andrew,

The aluminum bar is 4"x1.25"x0.625" with all 8 fets mounted to one side with thermal grease and mica pads.
I understand that the heat spreader bar is not a proper heatsink, but during testing I do feel it is getting too hot in idle. My version of the IRAUDAMP7S using +/-70V and one IRFI4020 per channel uses the same aluminum bar and gets barely warm after idling all day.
AC line current draw (at 115Vac) is 150mA in standby (output stage not switching) and 360mA in idle.

Thanks!

Originally Posted by Andrew Eckhardt

You'll have to post data or pictures about the heatsink before anyone can say much. The mouting interface (bar or block) is not a heatsink.

Quiescent power consumption is good info too. If its taking 30 minutes to reach max temp you have plenty of thermal mass in the mounting block, but you're lacking surface area (Fins!) if it's getting all the way to 50C. That much is sure.

[ungie]

I am planning on fairly high current demands into 2 or 3 ohms loads, so for safety I am running two pair of output devices. I'll try increasing dead time to see if that reduces idle Pd.

Thanks!

Originally Posted by stoc005

I would only use 1 pair of devices, not 2 pair. Also, play with DT. Increase it. That should help reduce Pd but it might cause THD to get rise.

[ungie]

A few updates:

I removed the 10k Gate-Source resistors and added a turn off diode across the gate drive resistors. I also increased deadtime from DT2 to DT3 as shown in the IRS2092 app note. The amp is now idling at approx 43 degrees C instead of the 54 degrees C previously. I did not suffer any significant increase in THD. I decided to borrow a differential probe from work to have a good look at Vds and Vgs on both high and low sides.
Photos are attached for idle condition, load connected but no signal.
Any comments would be appreciated.

Thanks!

[Andrew Eckhardt]

Have you tried running through the switching loss formula for both amps to see if the dissipation checks out as expected? I don't have data on those transistors, but all other things being equal, double the transistors and increased voltage will definitely make thing hotter.

Edit: just noticed the 4227 is a much bigger transistor. I'd say everything with the heat in your case is "normal".

[nigelwright7557]

Sounds like you had a bit of shoot through.
You can adjust the deadtime to help with this on the 2092.