Hello all,

I am going through power up tests of my hobby Class D audio amp but have hit something very interesting but equally as frustrating.

For your knowledge, I have powered the board up with +-24V as depicted 'without' the MOSFET bridge and the power rails hold up fine (i.e. no short circuit etc..).

I then mount in the MOSFET bridge with the few select other components as depicted. The capacitance in series with the inductance was put there to signify stray capacitances (for simulation).

After mounting in the MOSFET bridge (with each FET rated to Vbrss = 150V, i.e. reverse voltage) the power rails hold up for approximately 3 seconds and then the MOSFET goes short circuit and since my +-24V supplies have over current protection, they come down with it.

I am very confused at how this simple setup keeps blowing my FETs. The only theories I have until now are:

1. Inductance somehow oscillates. I cannot see how this would cause an issue as the +-24V supplies would clamp any oscillatory voltages above this. I tried with and without the bus capacitors of 1000uF thinking this may have been an issue but same problem. Simulations showed no oscillation during pulse power up (as expected).

2. The gates are currently floating and maybe they couple in a pulse and create a short circuit. The FETs do not get warm though, so I do not believe this is the case. If they were failing due to overload, they should be smoking hot.

3. (most likely) The gate to drain/source go partially short circuit after failure (100 ohms to each). The Vgs of these MOSFETs are 20V, so it might be a scenario of the inductor oscillating while the gates are floating which creates a Vgs of > 20V which causes a complete failure of the device. Subsequently perhaps there is a coupling path for +24V onto the lower gate, which then sees a source of -24V. I guess I just figured if it was floating for now there would not be enough power behind these voltages to cause a gate breakdown, but perhaps someone else can reflect a bit.

Another interesting fact is that if I power up the +24V but leave the -24V out, it holds up fine and vice versa for -24V but no +24V. It is only when I turn both on the device decides to die.

The actual MOSFET bridge I am using is the IRFI4019.

Any thoughts would be greatly appreciated, I'm pretty stuck on this one for no and afraid to proceed at the expense of continually blowing up my FETs.

 

It would seem more likely that both Fets are on simultaneously thus over current before your supplies can fold back.

 

Are the heat sinks not electrically insulated, connecting the MOSFETs together, or to ground?

 

If short circuit were the case of destruction, from my experience that results in a very HOT MOSFET, which I am not experiencing. This leads me to believe that a destructive over voltage fault is occuring which turns the device into a dead short (no Rds) so it stays cool while easily over powering the supply.

The MOSFETs are meant for heatsink mounting and the two FETs are in the same package (dual N channel).

 

It would seem more likely that both Fets are on simultaneously thus over current before your supplies can fold back.

Further to this, the supplies take 3 seconds to go into a dead short (noticed via LEDs) during a 'first test' (i.e. before the MOSFETs have died). Everytime thereafter the supplies go into a dead short instantly, visible by the lack of LEDs ever coming on.

3 seconds for a SMPS to power up is a very long time. From my experiences most supplies get up to regulation within 1mS.

Therefore I would conclude it is not a dead short at power up, rather a more lingering effect, as described by my first description of the power supplies characteristics.

 

Two questions... Why are the Gates floating? You know that's a no no. Also, why no load at the output node?

 

It's odd that you chose this Fet for an audio amp as the data sheet states that it's intended as a high speed switch. It's doubtful that it will have a large linear transfer curve.

 

It's odd that you chose this Fet for an audio amp as the data sheet states that it's intended as a high speed switch. It's doubtful that it will have a large linear transfer curve.

OP said it's for a Class D amp, which I believe is a "digital" amp -- it works by PWMing at a frequency much higher than audio. You filter out the PWM to get your audio signal back.

I am not terribly experienced at MOSFET design, but the very first thing I'd try if this was mine is: tie the gates to something. I'd guess you're getting some sort of oscillation that's either causing the gate to charge up to more than Vgsmax, or to swing higher than Vgsmax. Your three-second delay may be the time it takes for the oscillation amplitude to get high enough to cause G-S breakdown. However, once that happens (if it happens!) I don't know how one would expect the FET to behave.

 

Tie your gates to their respective source terminals with 10k resistors.

Don't leave your gates floating; bad things will happen.

 

Two questions... Why are the Gates floating? You know that's a no no. Also, why no load at the output node?

Yes, I realize this is a no no, but I was doing a power up test without the gate drivers in place and experienced this odd behaviour.

Before the next power up I will be adding zeners to the gates and ensure the gate driver is in the circuit to hold the gates at power up.

Regardless of this, it was some odd behaviour. I think I've realized that MOSFETs are more delicate than I thought and the advice thus far has been, just get them wired up with adequate protection and have another go.

Thanks for your recommendations.

 

OP said it's for a Class D amp

Oops, I missed that!

 

You should install a 10k-20k resistor from the gate to the source terminals on your MOSFETs anyway, just in case the drive circuit fails. They'll help prevent your MOSFETs from getting fried.