my opinion on all that is still that any generic "distortion at xx dB is ok" is a lot more misleading than useful/practical. A giant lot more.
I humbly disagree.
Now -85dB may be a debatable number. I can imagine that in really quiet, purpose built, well treated rooms where people can let loose and play at high SPL, harmonics could theoretically be audible.
Lets take -100dB as an example.
If the harmonics are down at -100dB the # of available scenarios in which an e.g.: 3dB strong harmonic @ a playback level of 103dB would matter are already pretty slim to nonexistent.
1) Transducer distortion rises rapidly when playing above 100dB, especially for bookshelf speakers. My 906, for instance have pretty heavy distortions at 96dB already, as measured by our overlord Amir. Not sure I could even push them past 100dB.
2) a 3dB strong harmonic gets lost in the noise floor of a typical room or even a really quiet one. I measure sth. around 42dB(C) at best in my room. I know human ears can listen "into the noise" and still recognize signals, but I doubt they penetrate that deeply. Especially not when point 3 applies at the same time. (You're the psychoacoustic man, perhaps you have more data on that)
3) Masking effects. It's very difficult for our ear to hear a tiny signal, next to an overwhelmingly loud one. If you blast your ear with 120dB of sound, would you be able to hear a 20dB harmonic? I certainly could not. All I could do would be reaching for the power plug while panicking and my neighbor would probably be like (ノಠ益ಠ)ノ彡┻ ━┻ .... but that's beside the point.
At -115dB your primary signal would need to go past the human threshold of pain in order for harmonics to be audible. Hence this being the strict limit.
As the thread stated: lenient limits should be okay most of the time. No guarantees though. I find that to be a fair assessment and a good rule of thumb to go by. Naturally, you have to check with your own, specific listening conditions, in order to figure out whether you might have a case on your hands, that demands better performance. The strict limit is a safe bet no matter the case.
Unless you're not human. Then ask your own scientists.
