@crutschow (and @rpschultz as it was mentioned on your thread). I've now fully recovered after a week in hospital with a bacterial lung infection and after four weeks of antibiotics, I have 82 test results that prove I'm OK now.
So as I now have access to my computer which runs SPICE, here is, as promised, the amplifier that SPICE claims manages 3ppm (-111dB) third harmonic distortion. There are two circuits, almost identical except for the VAS stage, all with very similar figures for each harmonic, but SPICE says that TOTAL harmonic distortion is four times as good for one as the other.
And as I said previously, I really don't see how either of these circuit could possible be as good as it thinks they can.

 

Only 1 diode to account for 2 Vbe? I think this is a mistake rather than planned.

 

Only 1 diode to account for 2 Vbe? I think this is a mistake rather than planned.

The circuit was designed like that, mainly to see what the distortion would be in pure class-b with that circuit. The figures on the FFT were rather a surprise.

 

I did the top simulation and it would appear the feedback loop is fast enough to cancel the crossover distortion.
Even shorting out D1 still gives very low distortion.
I agree that the simulated distortion values seem to be too low to be believable.

 

I did the top simulation and it would appear the feedback loop is fast enough to cancel the crossover distortion.
Even shorting out D1 still gives very low distortion.
I agree that the simulated distortion values seem to be too low to be believable.

I couldn't measure anything that low if I tried. I don't have a signal generator that good! I have a friend who has an Audio Precision analyser but I don't that that can manage below 10ppm, and power supply noise will take over before it gets to that.