[sdiy] Simple discrete Unity-Gain Follower ?

[jhaible]

> I agree that feedback is great for DC though.

I'm using discrete circuits where ever I can, so you're preaching to the
converted. But nevertheless:

If the the amp is fast enough, audio is "almost DC".
So with modern opamps getting faster and faster, what would you consider
fast enough for that dreaded servo loop thing?

Of course in 8 of 10 cases these is no need for opamps in audio circuits.
Always consider a single transistor emitter follower instead of an opamp
unity gain buffer, just to name one example.
But - to stay with this example - if you need really low output impedance,
it's hard to do it better than with a full blown opamp structure. (The
npn-pnp
darlingtons which also give low output impedance have their own
problems, and they have very tight feedback, too.)

Sometimes circuits show a quite unexpected behaviour - problems where you
never expected them. Recently I've built a line driver for the output stage
of my PS-3200 clone. It's a discrete class A opamp with opto electronic
gain control, driving a transformer:
http://www.oldcrows.net/~jhaible/polykorg/jh_polykorg_amp_midi.pdf
I started with the (single supply) Urei 1176 compressor's output stage
and tweaked the circuit until it fit into the new application.
I added one transistor to make an opamp structure where there was an
ordinary amplifier with feedback to the emitter of the (single) input
transistor. Will the extra transistor in the signal path (in common base
connection!) degrade the performance? I don't think so; but I get rid
of the output electrolytic capacitor, which I consider an improovement.
Of course this circuit has the same tight feedback loop as the original,
which may or may not be a problem. But the voltage gain stage is
tamed a little by R58, and a carefully chosen compensation (by Urei;
I just copied it) with a 1k resistor in series to the "integrator" cap
C30 might have a positive effect on the performance, too.

But here comes the unexpected thing: In order to get control over the
quiescent current (setting Class A), I tried to replace the 3 diodes
with the often-used "variable zener" transistor circuit.
Then I did some extensive Spice simulations. Mainly AC runs to check
stability under various capacitive loads (1pF ... 1uF). But also
Transient runs to look for unwanted effects when the circuit is
heavily overdriven. And I found that the variable zener biasing
produced a very short current spike in one of the output transistors
when the amp was overdriven. I didn't even see an effect on the
output voltage, but even then such a current spike made me
very suspicious. So I went beack to a diode circuit.
I don't know *why* the behaviour is different, but simulation
says there *is* a problem, and I would never have found it
without simulation. And even though the effect would have
been too small to be visible on the output voltage, someone's
golden ears might have noticed it.
(Thinking of it again, it might just be the higher capacitance
of 3 high current diodes vs. a small signal transistor ...)

My conclusion is: There are many effects that can happen in a complex
amplifier, especially when you leave the more or less linear
region, but the choice of open loop gain and feedback is only
one detail among others.

JH.