The usual total harmonic and intermodulation
distortion figures do not reveal the abrupt output stage distortions accurately
because of the averaging factor involved in such measurements. A spike of
crossover distortion may reach 2 per cent, but if it occurs only over 5 per cent
of the waveform, it averages out to a respectable 0.1 per cent distortion
figure.
What he doesn't say which is important is that the crossover distortion is
fixed (it is not amplified as you crank up the volume, at least that is my
understanding) - therefore Class A becomes critically important at low power
output => this is what makes designs with sliding bias or several stages that
include a Class A stage attractive. (it allows distortion to remain low at low
output levels or quiet passages but then still gives you the ooomph needed
for loud passages....the crossover distortion is less critical during a loud
passage as it is a relatively small amount on a good Class AB design
compared to a big signal output)
This also makes an active speaker topology attractive. Since the distortion is
high order then some of the worst stuff is going to come out the tweeter. In a
biamplified active speaker design you simply feed clean Class A for the feeble
power requirements of the tweeter whilst you can power the mid/woofer with
a beefy Class AB design without much worries over high order harmonics or
intermodulation distortion being all that audible