Hi gang. I'm working on the design of an amplifier to be used in a
tube mixing console. (Please, let's ignore for the moment the question
of WHY someone would want to design a tube mixing console in 2003!).
Anyway, I cooked up a usable circuit in short order: two triodes in
cascade, with negative feedback from the plate of triode 2 to the
cathode of triode 1. I'm using both sections of a 12AU7. The open-loop
gain of the circuit is 123, and with feedback it's 76. So, there's
4.18dB of feedback, a very modest amount. I didn't like the idea of
using NFB at all at the beginning, but this application really demands
predictable gain that isn't unduly affected by tube aging or
manufacturing spread.

Thing is, I've come across a reference to a study done by P. J.
Baxandall in the '70s, in which he showed that using less than 15dB or
so of feedback can actually INCREASE the generation of higher-order
distortion products. My distortion analyzer is out of order at the
moment, so I can't look at the distortion products of my breadboarded
circuit. I guess I'm just wondering if Baxandall's assertion (which
apparently arose from experiments with a FET amplifier) also applies
to tubes. Your comments are very much appreciated.

Why not just use, say, a 12AX7 (with an open-loop gain of 1600 in this
particular circuit)? Well, I tried it, but it clips with a much lower
input voltage (0.14VRMS versus 0.67) and has a much lower maximum
output swing than the 12AU7 (27 VMRS versus 45).