The idea with the new "auto" replay gain mode is that each playlist is as loud as possible while being the same volume *inside* the playlist. The volume *between* playlists will be very different. (a playlist is any collection of tracks in Playing Now)
Here's a summary of the process:FOR EACH FILE:
1) figure out how much we could safely turn a file up using the peak level
2) use (1) to see how much we could safely add to the replay gain level
Auto preamp will be the minimum (2) from all the files.
Example with 2 filesA: peak level: 50%, replay gain: -10 dB
B: peak level: 75%, replay gain: -4 dB
A(1) = 50% -> +6dB could be added safely
A(2) = +6dB - -10dB = 16 dB could be added to RG
B(1) = 75% -> +3dB could be added safely
B(2) = +3dB - -4dB = 7 dB could be added to RG
Auto preamp = min(A(2), B(2)) = +7 dB
During playback:
Actual gain (A) = -10 dB + 7 dB = -3 dB
Actual gain (B) = -4 dB + 7 dB = +3 dB
From Dragyn:replay gain (auto, playlist based) isn't adjusting to 0dB. replay gain analyzes to the 1/100th (-11.85dB). dsp studio adjusts to the 1/10th, shows -11.9dB , adjusts +14.0dB, changes to +2.2dB instead of 0dB.
It won't, because it factors in the "peak level" and realizes that it could safely turn it up even more. And internally we use lots of precision for replay gain, but round when displaying it.
Think of it as normalizing but using reply gain values. If I play each file one at a time in playing now, it will do exactly what I would like.
If you play just one file in "auto" mode, it just ends up "peak level" normalizing the file, and doesn't even use Replay Gain. Replay Gain only kicks in if there are multiple files -- it makes sure they're all the same volume.
Hope that makes some sense...