Standard THD measurements give a clear indication of amplifier linearity for the signals with which it is measured. Curiously, however, very often THD performance does not seem to correlate with listener preferences. There have been attempts to develop measurement protocols that correlate better with subjective performance, but not with complete success. Ian Hegglun picks up the gauntlet, and argues that much of the problems stems from the fact that human hearing has a sensitivity for harmonics of the signal that varies enormously with the order of the harmonic, making a measurement that lumps all harmonics into a single number a recipe for failure. Ian quotes research that shows that audibility of higher harmonics is many 10's of dB higher than that for lower harmonics. His novel weighting filter boosts relative levels of higher harmonics based on these findings with a resulting performance number that much better correlates with listener preferences.
Towards a reconciliation of measurements with listening tests