OTA question

[Haible_Juergen#Tel2743]

>> feeding a capacitor/resistor to ground and buffer it with an FET
>> source follower.  This will keep the circuit free of nasty opamp-style
>> feedback.
>
> Nasty feedback? Interesting. Can you elaborate a little?

Don has posted about this before, and he has some good points
about it, IMO. Pumping signals thru dozens of nonlinear transistors,
all in one heavy feedback loop, has its disadvantages. (Especially
when the forward path is rather slow, due to the artificially reduced
bandwidth, which is necessary for stable unity gain applications.)
I tend to go for discrete circuits (and *local* feedback) more and
more. For audio paths, that is. But sometimes I want my modules
to be usable for both, audio signals and VC's.
I built the mixers for my JH-3 Modular with discrete BJT's: The
inputs are buffered by NPN emitter followers, then there is the
pot for mixing between two sources, then the pot is buffered
by a PNP to compensate for the voltage drop of the input buffer.
Well, it doesn't compensate to 100%, so there is an offset trimmer
as well, but at least the temperature drift is 1st-order compensated.
This is quite usable for DC-coupled mixing applications, though
there is a slight variation in offset voltage dependent on the pot's
position. Guess using FETs instead of BJT would have been better
for this - but then again the GS-voltage of FETs has a mouch wider
spread than the BE of BJT's, so *selection* would have to be
done ...
... it is *not* that easy to avoid this nasty overall feedback !

JH.