Not sure what I've got to add to that; I see 'nice measurements' as a rather necessary but not entirely sufficient condition.
I think IMD is often underplayed; although the root cause is the same as %THD (i.e. a slightly bent transfer function*) it is IMO possibly the more 'obvious' of the two on audition since even small amounts lend that distinct 'mush' where THD is probably rather less directly audible. By which I mean it can be very hard to hear a slight or even large shift in timbre unless you are intimately aquainted with the very sound that was recorded; and even then it can be a matter of taste - 'would you like the Stradivarius or the Guarneri, sir?'. Possibly why valve vs SS provokes such heated debate - the harmonic spectra differ, and even if (some) effects are gross, they may be very consonant to a given listener. Vive la difference.
* That's half the answer to 'why 19+20Khz?' - it's because OL gain is falling fast so distortion products are grosser, and easier to measure. If it measures well that high up, it's likely well-behaved a decade below.
[Simon if chains of opamps add up IMD more than THD, take a look at the PSU / decoupling; might need more and larger caps or even a touch of series R each if using a shared supply. ]
I think IMD is often underplayed; although the root cause is the same as %THD (i.e. a slightly bent transfer function*) it is IMO possibly the more 'obvious' of the two on audition since even small amounts lend that distinct 'mush' where THD is probably rather less directly audible. By which I mean it can be very hard to hear a slight or even large shift in timbre unless you are intimately aquainted with the very sound that was recorded; and even then it can be a matter of taste - 'would you like the Stradivarius or the Guarneri, sir?'. Possibly why valve vs SS provokes such heated debate - the harmonic spectra differ, and even if (some) effects are gross, they may be very consonant to a given listener. Vive la difference.
* That's half the answer to 'why 19+20Khz?' - it's because OL gain is falling fast so distortion products are grosser, and easier to measure. If it measures well that high up, it's likely well-behaved a decade below.
[Simon if chains of opamps add up IMD more than THD, take a look at the PSU / decoupling; might need more and larger caps or even a touch of series R each if using a shared supply. ]
