On Wed, 12 Aug 2009 09:56:54 +0100, Jim Lesurf
wrote:

In article , David Looser
wrote:
"mick" wrote in message

Now, whether 1nF is enough to make any difference at all I'm not
experienced enough to know. If Jim says: "if you just put a shunt
capacitor across the output that might provoke oscillations if the amp
isn't already stable for other reasons" then that's good enough for
me. :-)

If you are right in your analysis of Eiron's point (and it sems likely)
then clearly he has little experience of this sort of thing. Yes
indeed, a 1nF capacitor connected directly across the output of an
amplifier can easily cause instability if that amplifier is not
unconditionally stable.

Yes. I used to keep a small selection of capacitors with values ranging
from a few tens of pF up to a couple of microF to do quick checks on
amplifier stability, etc, when experimenting. I often found that values of
the order of an nF or so would cause bursts of oscillation with designs
being played with... erm developed. I also had a variety of inductors and
'evil' loads to torture the amp and see if I could annoy it. Usually ending
with a screwdriver to short the output terminals when the amp was playing
to see it if survived. 8-]

Perhaps worth adding that quite often there was a narrow range of
capacitance values that might upset a design. So, say, 2n2F might cause
oscillations, but 1nF or 4n7F didn't. So the problem often isn't that 'any
value above X' causes oscillations. Hence you may well find that an
amplifier that isn't unconditionally stable will work happily into loads
which have a high capacitance, but then burst into song when used with a
smaller load capacitance. So this is rather a matter of chance for users if
the amp isn't unconditionally stable. The possible range of loads being
presented by all the cable-speaker combinations out there is quite wide!

...oh, and just to add to that something I've just recalled. I also at
times used to leave an amplifier design in the fridge overnight and then
test it before it had a chance to warm up. Quite interesting how often a
design would oscillate when cold, but then stop when warmer... or vice
versa.

Slainte,

Jim

I think capacitor resonance is important here. If an amp is
potentially unstable (usually with a capacitive load) at say 30MHz, a
smallish cap may provoke it. A big one, though, probably resonates
below that frequency, so at 30MHz it isn't a cap, but an inductor. The
amp will be quite happy with that.

This sort of brings up another point. It is easy enough to design an
amp that is unconditionally stable - the maths isn't hard. But this
makes some assumptions, one the big ones being that Cdom (usually
about 100pF) has the right phase shift. If the layout isn't great,
though, it is quite easy to end up with perhaps 100nH of parasitic
inductance in series with it. If that happens, Cdom turns into an
inductor at 50MHz, which may well be low enough to start things
fizzing.

So it is not enough to design an amp which is theoretically
unconditionally stable, a parasitic analysis must be made to ensure
that this also applies to the reality of the physical design.

d