output FET overcurrent limiting

[zilog]

I am concerned about overstepping the output transistor's current capability in case of low frequency and low speaker impedance or just simply a short circuit. Is there any convenient way of limiting a class-D amp from going over a certain current without causing rise time penalties?

/Daniel

[IVX]

1) The shunt in serial to mosfet (basic technique)
2) by additional winding on the output coil
3) by Drain-Source voltage measurement when mosfet is "ON", i use it in the SMPS so far.

[Pierre]

I use the Vds voltage measurement method (measures Id via the mosfet Rds(on) with very good results.
The disadvantage is that it only senses current in the low side mosfet. I think that's admisible in Class-D, where current due to load doesn't change from one cycle to another, so if overcurrent is detected in one mosfet, the other one is still safe as its current can't be much higher than the low's one.

If this assumption is correct (please, I would like to hear comments on this), all you have to do is to shut-down the driver chip (or its inputs) as fast as possible once overcurrent is detected.

At the moment, my amp triggers the shutdown pin for 2-3 secs, then it restarts. Once I intentionally made a shortcircuit at the load terminals when playing loud music, and then it stopped. I left the shortcircuit, but unfortunately when the amp restarted after the delay, the mosfet blew up. I still don't know why (current is supposed to start building up relatively slowly, right?)

[peranders]

Have you checked the current sensors from www.Linear.com ?

http://www.linear.com/pc/productDeta...09,C1077,P1780

Maybe something

[IVX]

Pierre,
it seems upper mosfet started first and feedback held this state too long. My SMPS have single mosfet over current protect also, and shortcircuit no problem.

[phase_accurate]

I don't know how Pierre's amp failed with the short applied but there is one scenario that I could imagine:

While the driver was disabled there might have been some drift (NFB is also temporarily disabled in this case) causing the amp to come back with the upper mosfet on for a period long enough to cause damage.

Regards

Charles

Edit: One solution might be to keep the driver on but to short the INPUT signal to ground instead.

[Kenshin]

Maybe use resistor sensing on both rails? clumsy but reliable.

[Pierre]

Charles.
Your proposal of shorting the input makes a lot of sense. However, there are still a couple of points that I don't understand:

a) My shutdown time was around 4 seconds, and I don't have precharge of the bootstrap capacitor, so I think that the HS mosfet cannot turn on at first instance, as there shouldn't be Vb supply until oscillation starts and the cap charges, right?

b) If you only short the input to GND, you will still have 50% duty cycle oscillation. In case of a shortcircuit at the output, would it be safe to have it short-circuited and still switching mosfets? (I know that corresponds to 0V output after lowpass filtering, but what happens when re-arming?

Although that solutions seem quite sensible, Charles, why is it supposed to improve the situation? I mean, as soon as the amp is "re-armed" after, say, 4 seconds, it will try to deliver audio to the shortcircuit and it may blow just the same it happened before, am I wrong?

Do you think that low-side sensing only can lead to a short-circuit proof amplifier under any circunstance?

[phase_accurate]

Although that solutions seem quite sensible, Charles, why is it supposed to improve the situation?

I just imagined that it would be good to let the NFB keep full control of the amp all the time.

Some 13 years ago we did a switching amp that was short-circuit proof. We did exactly what I proposed for frequencies above some 100 Hz. At lower frequencies we injected a correction signal into the input thus limiting the output current.

We used a very small shunt resistor between output and ground (plus some additional circuitry of course) for this purpose. This didn't give us any problems because the amp itself was only single channel and it had differential inputs (so the hot end of the shunt could be grounded.

If the low-side-only detection can be safe at all I can't tell without doing more thinking. Problems can arise at low frequencies IMO.

Regards

Charles

[Pierre]

Thanks, Charles.
You are right in that it is better to avoid disrupting the control loop as the turn-on behaviour can be dangerous for safety under short-circuit circunstances.
I have to think a bit more about this, but perhaps shorting the input to the comparator to GND under short-circuit circunstance can be reliable.