Quite a while ago, maybe 3 years or so Bob Ludwig at an AES mastering panel demonstrated how interstitial clips can and do happen and suggested a margin of 1.1db below full scale when normalizing. Sounds about right. It is unlikely that any interstitial overloads could happen given that much head room. Should be enough to let them live while not running themselves out of numbers.
How about simply normalizing to <0.0 and see where they quit clipping?
Please pardon the mispellings and occassional insane word substitution I'm on an iPhone
> On Mar 4, 2018, at 19:33, Steven Smolian <[log in to unmask]> wrote:
> Final result.
> I can keep the sound as I go from 32 to 24 to 16, with a slight less overall
> air around the sound at 16. If I normalize, I get the crackles.
> Son the record which I used for test today, Coates 1927 recording of the
> Prelude to the 3d act of Meistersinger, I began with a max level of -12 and
> ended up max at -6.
> It does sound amazing at 32. You can hear the resonance of the open bass
> strings at the start. And the massed upper strings are subdued- these were
> orchestras whose players had just heard of Auer and not yet of Suzuki.
> Steve Smolian