LME49830 burning

[Domid31]

Hello,

I am working on a Mos amp using LME49830 and IRFP240/9240
I have already burned many mos, I do not understand why.

Sometimes it burns when powering up, sometimes I can run long time at full power, output is clean, no oscillation.
I am using National application circuit, with my own PCB (there is no mistake on it). Gate resistors are 100 ohms, I did not mount the 12kohm resistor between the 2 gates (Rq in National application), I just add 2 zener diodes between gate and S for mos protection.

When the amp burns, the gate resistors are getting black, the mos and the lme49830 are dead.
I am using a +-50V unregulated power supply, large capacitors (33 000uF).

Any Idea ? Can it be caused by the zener protection diodes ?
Can it be caused by power supply rise (full voltage can be applied on mos before lme49830 because of PCB design) ?
I am a bit lost ...

[Dxvideo]

Hi Domid,

Does your output bias current level adjustable like in the datasheet? And did you measure the bias?
Did you hear any "shaking" noise while your MOSses dieing?

[Domid31]

Hello,

I have setup to bias current to 70 mA in hot condition.
I am monitoring it: run the amp, wait some minutes, cut input signal, measure and adjust bias.
the bias temp compensation is not used for the moment.

Can this destruction be caused by zener diodes in between gate and source ?
Can it be because I have no resistor between the 2 gates ?
I do not know. I'll replace the burned parts, and test it again.

[AndrewT]

Hi,
you must fit a gate stopper resistor to each and every gate lead.
This resistor does it's job better if it is place right on the gate lead with the shortest possible lead length.

[Dxvideo]

Dear Domid,

That MOSFETs should have at least ~20v GS voltage limit. So I think, without any input signal its not possible to reach gate voltage limit to kill them.
But; If youre not using a gate resistor for these MOSFETs then you have a big situation! Because, its strongly possible that they will start to oscillate after a short time! I think this is the reason you lost a lot of MOSFETs...
Vbe multiplier solution for bias stability is not a must for this amplifier but its a recommendation. Because as I see in my LM4702+Lateral MOSFET amplifier, they have very good bias stability against temperature..
However I think 70mA is a bit more for IRFPs... 40mA will be sufficient for best distortion results..
Let me know the results after gate resistors added..

[AndrewT]

for best distortion and sound quality results the higher the better for bias current. just keep on increasing it all the way up to ClassA bias.

Borbely recommends a minimum of 500mA of bias for the whole output stage and a minimum of 150mA for each output FET.

[Dxvideo]

Originally posted by AndrewT
for best distortion and sound quality results the higher the better for bias current. just keep on increasing it all the way up to ClassA bias.

Borbely recommends a minimum of 500mA of bias for the whole output stage and a minimum of 150mA for each output FET.

Wrong!
Look at the AN-1645... They have already tested 4 diffrerent MOSFET types and determined different bias levels for each.. And for IRFPs the best bias value is 25mA, not 150!
Thats what I meant in my last posting..

[AndrewT]

I prefer to accept Borbely rather than National's application note.

I trust him more than the design team.

[Dxvideo]

Ok... Thats your choice. But Nationals application note is dedicated for LM4702-LME498XXX series..
So I prefer (and already used) that bias levels and gate stopper values for this series amplifiers only.

[Dxvideo]

Originally posted by Domid31
Gate resistors are 100 ohms, I did not mount the 12kohm resistor between the 2 gates (Rq in National application), I just add 2 zener diodes between gate and S for mos protection.

Ooops... I have missed that you already have gate stoppers! So the oscillation theory is dead.. As mentioned, its too hard to kill your mosfets in this application by the reaching of gate voltage limit.. I would remove the zeners and test the amplifier with inputs muted.