What's causing these MOSFET's to overheat?

Thread Starter

JMan89

Joined Oct 10, 2023
7
I have a half bridge Class-D Amplifier that's not working.

As soon as it's turned on and with no input signal, MOSFET's T1 & T2 in the attached .pdf immediately begin to overheat, hot enough to melt the solder on the Drain tab.

This is a logic level FET and the minimum VGS to turn it on is 3v. When I measure VGS on both FET's when the amp is turned on, I get 0v. This should mean that the FET's are off and no current can flow across Drain and Source, correct?

What then could be a possible reason that these FET's immediately begin to overheat? They are obviously letting current flow somehow.

The interesting thing is that I get no voltage on the speaker output with reference to ground from the FET's. I currently have an 8ohm resistive load across the speaker outputs while testing.

Any help is appreciated.

Thank you.
 

Attachments

DickCappels

Joined Aug 21, 2008
10,122
First thoughts: They may have their source and drain swapped, have an internal short, be “dummies.”

if you post the schematic we might spot something or get more ideas.
 

activerfid

Joined May 30, 2020
12
Hi, thanks for your help.

The schematics are attached.
At first glance the schematics appear to be ok.

I would disconnect the load and recheck.
Also scope the gates and power supply to see if there is any noise.
To ensure the gates really are pulled low connect 10k between gate and source on both FETs (for testing only, remove afterwards).

Otherwise (as DickCappels suggests) the FETs are suspect.

Following on, I have found a number of articles that indicate potential issues with the IRS20957 / IRFR24N15D combination

MOSFETs overheating on power on (another designer fitted 22k gate to source resistors and they are questioning whether the IRS20957 can produce sufficient gate current for the IRFR24N15Ds).

Another article indicated that if the PWM input is floating then the IRS20957 self-oscillates at 500kHz (with obvious consequences on the outputs).

Of course, there may be other issues (knockoff parts etc.) but I would start with 10k's or 22k's gate to source and have something on the input (if only from a generator).
 
Last edited:

HasBeen

Joined Jul 28, 2023
19
You didn't mention if you had tested for AC input to the FETs. You can have ZERO volts DC but still have AC volts (oscillation?) present at the inputs, which could cause current flow that won't show up with only DC measurement.
 

Thread Starter

JMan89

Joined Oct 10, 2023
7
At first glance the schematics appear to be ok.

I would disconnect the load and recheck.
Also scope the gates and power supply to see if there is any noise.
To ensure the gates really are pulled low connect 10k between gate and source on both FETs (for testing only, remove afterwards).

Otherwise (as DickCappels suggests) the FETs are suspect.

Following on, I have found a number of articles that indicate potential issues with the IRS20957 / IRFR24N15D combination

MOSFETs overheating on power on (another designer fitted 22k gate to source resistors and they are questioning whether the IRS20957 can produce sufficient gate current for the IRFR24N15Ds).

Another article indicated that if the PWM input is floating then the IRS20957 self-oscillates at 500kHz (with obvious consequences on the outputs).

Of course, there may be other issues (knockoff parts etc.) but I would start with 10k's or 22k's gate to source and have something on the input (if only from a generator).
I have tried the 10k Gate to Source resistors and this does not change the overheating.

I have however found that when I disconnect the resistive load from the speaker outputs, the FET's no longer overheat. I suspect that this is because the DSP or similar does not "turn on" until a load on the speakers is detected. Still, it is interesting.

All Power Supply lines appear clean when scoped. I'm trying to scope the PWM input into the IRS20957 IC's but as this is a "floating" input, I'm not sure with reference to what I am suppose to be measuring on the scope. Same with measuring the Gate inputs on the scope, should the measurement just be directly across Gate and Source?

I am using 2 ungrounded scope probes and then summing them together to get the waveform.
 

activerfid

Joined May 30, 2020
12
I have tried the 10k Gate to Source resistors and this does not change the overheating.

I have however found that when I disconnect the resistive load from the speaker outputs, the FET's no longer overheat. I suspect that this is because the DSP or similar does not "turn on" until a load on the speakers is detected. Still, it is interesting.

All Power Supply lines appear clean when scoped. I'm trying to scope the PWM input into the IRS20957 IC's but as this is a "floating" input, I'm not sure with reference to what I am suppose to be measuring on the scope. Same with measuring the Gate inputs on the scope, should the measurement just be directly across Gate and Source?

I am using 2 ungrounded scope probes and then summing them together to get the waveform.
Check signal on IN and LO, both referenced (scope ground lead) to VSS

Link CSD to VSS to shutdown the IRS20957.

Check for overheating FETs during each test (load connected and ensure scope probes are not connected to mains earth).

The application note recommends a 1N4002 between COM and VSS (cathode to VSS) as the IRS20957 can easily be damaged by low or missing supply voltages.

Other than this also check the start-up sequence as there can be many issues with lockups.
 

Boggart

Joined Jan 31, 2022
82
Seen amps overheating before due to ultrasonic oscillations, forget the cause now, but definitely get a scope on the outputs and see what's going on.
 
Top