SG3525 destroyed MOSFETs

[SMPS3525]

Hello all,

I designed a regulated push pull SMPS using PWM Controller SG3525 and 2x MOSFET IRFB41N15D connected via 10R.
Input voltage is +36VDC, output voltage is +32V/1A.
Loads are audioamplifiers and some relais logic used in an
intercom system.

The SMPS works several days and suddenly one or both
MOSFETs are destroyed. Who has any idea, what could happened?

I used SG3525, IRFB41N15D, the same transformer and other same parts in several application in SMPS up to 200W with no problems. I am absolutely sure, that there is no problem with overvoltage, undervoltage, transients, overheating.

Charging current is limited by softstart function of SG3525.

I use a common shunt for both MOSFETs to detect short and
overload. The shutdown pin of SG3525 is connected to the shunt via a 470R resistor, so pulse by pulse current limiting is sure.
IRFB41N15D is overdesigned for this low output power, so
i assume, that all parameters are within limits.

Measurements at gates and drains show clean signals.

Is it possible, that under special conditions the FETs are destroyed by shoot through? How could this occure?
Has anyone experience with layout problems? I use a 4 layer board with power planes and think that the gates of MOSFETs are connected short.

[richie00boy]

What voltage are you running the SG off? Did your FETs get warm at all under any conditions?

[SMPS3525]

The SG3525 runs at +12V, stabilized by usual fix voltage regulator. The FETs are mounted on a sufficient heatsink.
There is no significant heat up. Also in current limit mode there is
no thermal problem.

[N-Channel]

SMPS3525 (I like that name);

What is your minimum load? I seem to recall that without a minimum load on such a supply, the pulses to the MOSFETs becone very narrow, causing spikes on the transformer primary, possibly in excess of the MOSFETs' 150V(dss) rating. Is the amplifier connected all the time, or did the you run the supply no-load?

Steve

[sivan_and]

Did you incorporate snubber for the MOSFETS...
Did the inductive kick back from the relays was the reason..just a thought..

[SMPS3525]

The minimum load is about 200mA, the quiescent current of the audio amplifiers. Maximum current is only 1.5A.

First i though about problems with transients.
So i added suppressor diodes P6KE100A from Drain to Source
of both MOSFETs. At the output there is a 1.5KE33A.
There are Zenerdiodes 13V from gate to ground.
There is bypass capacitor placed very close to the FETs.
The lines to the Transformer are very short.

I tested surge pulses of 4kV (5/50us) to the input and to the
output. I could not destroy the MOSFETs.

But after several days or weeks one or both FETs are short.

thanks for your replies

[subwo1]

Possibly, your MOSFET gates are getting spikes that you don't see, which can explain why it could take a while for the oxide layer to be destroyed. The zener diodes could be a cause.

[sivan_and]

subwo1: wrote

The zener diodes could be a cause.

How could it be..? the zeners clamp the gates to 13Volts..]
Look for any dry joints between pwm o/p of IC to mosfet gates.
when did the Mosfets blew ? at switchon or during Music peaks?

[subwo1]

How could it be..? the zeners clamp the gates to 13Volts..]

Supposedly, they can ring at very high frequencies. ??

[jahonen]

Flux imbalance problem? Push-pull is very prone to flux imbalance, so to prevent this p-p should always be designed with current mode control.

Regards,
Janne