[Eeyore]

Don Pearce wrote:

Eeyore wrote:
Don Pearce wrote:
Eeyore wrote:
Don Pearce wrote:
Kevin Aylward wrote:

I don't see that it matters much where the error comes from, it can be
handled in the same general way.
It can be handled the same way, but doesn't need to be. Noise must be
handled by filtering, with with a specific inline filter or effective
filtering by means of a control loop. DC will also be handled that way,
but for DC there is always the option of trimming it away with either an
adjustable or an SOT. You can't do that with noise, which makes them
fundamentally different things.
And offset in an audio power amp is commonly corrected by not making it DC
coupled. A filter of sorts I suppose.
I think that since these days 99% of audio amps are configured as op
amps, DC offset is most commonly handled by the feedback loop - the
second kind I listed.

Gross offset from mismatched power devices, yes. I was thinking of that from the
input pair. Of course lateral fets aren't subject to the gross offset problem.

All depends how the input stage is configured. If the actual input
transistor is suitably grounded at its based, and the feedback returns
to the base of its Blumlein partner, then all offsets through the entire
amplifier chain get added together and corrected as a single item,
resulting in the speaker terminal at zero volts, just like the first
transistor base.

Actually this isn't quite true since that base is offset slightly
negative by the base current through the several k resistor.

Which is hopefully matched by the feedback circuit resistance.


Individual internal nodes will still be non-zero, of course, but if we
treat the amplifier as a black box, with only input and output terminals
visible, the net offset is always efectively zero. Or rather whatever
they would have been if uncorrected divided by the open loop DC gain of
the amplifier.

Indeed.

Graham