Amp Tester
I whipped up this amp tester earlier this evening. This device consists of a linear subtractor and voltage amp. The idea is to sample the output signal from the gNFB summing node and subtract it from the input signal. Theoretically, whatever's left is distortion after you null the output by adjusting the input signal level. I ran some tests on the new project.
1.0KHz Sine (1.0mS / 5.0mV)
This gives the best null, due to minimal phase shift at this frequency. Most of what you see on the screen is interferance from the 50KW AM BCB station about 30 miles north of my QTH. Residual distortion is pretty much swamped out by the noise.
166Hz Sine (1.0mS / 50mV)
More phase shift, so that the null isn't so deep. The residual is quite sinusoidal. There are some faint switching glitches about half-way up the curve. This was running the finals a bit on the hot side: 55mA of cathode current, for about 17.5W of PD (PD= 12W according to the 6BQ6GTB spec sheet) so this isn't quite so far into Class AB as the design Q-Point. As expected, running hot improves the sonics.
15.6KHz (1.0uS / 50mV)
This looks like predominately h3 with the harmonic being somewhere between 0 and 180 degrees. Still not too bad, considering that this is reaching for the top of the audio band where OPT troubles start to appear.
Tester Schemo
This is based on the NTE859 quad op-amp IC. These being low noise, low distortion, high slew rate, JFET input op-amps. The 33K resistors used in the subtractor are 1.0%, 0.125W, precision metal film. The rest of the resistors are your plain old garden variety 0.5W metal films. The whole thing runs off two 9.0V batteries wired for +/- supply. Still have the final voltage amp to wire up, the idea being that you can then listen to the residual to see what the distortion actually sounds like. If that doesn't sound too bad, then the distortion might not be too detrimental to the sonics.
A new and improved version would include a phase shifter for deeper nulls.