Quote:
Originally Posted by bezusheist
here are 3rd party stats showing what happens with the normalizing process...
the process was:
generate 0 dBFS source file @ 64 bit float (render)> clip symmetrically to -0 dBFS (render)> gain down 50% to -6.02 (render)> item processing "normalize items" (render)
all files were rendered at 24 bit integer (no dither).
i have tried this same test using all 64 bit float files as well, and the results are basically the same. (normalizes to same value, not 0dBFS)
afclip stats
source file
2_pulses_0dBFS_sym_clip_24bit.wav
1 ch, 44100 Hz, 'lpcm' (0x0000000C) 24-bit little-endian signed integer
-- no samples clipped --
SoX stats
source file:
2_pulses_0dBFS_sym_clip_24bit.wav
DC offset 0
Min level -8388607
Max level 8388607
Pk lev dBFS -0.00
attenuated 50% (spl0*0.5)
2_pulses_half_power_sym_clip_24bit.wav
DC offset 0
Min level -4194304
Max level 4194304
Pk lev dBFS -6.02
normalized in Reaper
2_pulses_half_power_sym_clip_24bit_normalized.wav
DC offset 0
Min level -8380219
Max level 8380219
Pk lev dBFS -0.01*
*actual value is more like -0.00869...+/- dBFS
i attached all of the test files just to "properly" document this with some actual "evidence".
edit: huh...i think i actually just figured out what's going on, but i'll wait to hear from a developer or someone first..  |
I can replicate this.
However, I'm not sure that the value of fixing this is worth it unless it's some very obvious fix for the developers.
edit: Information:
Code:
sox 2_pulses_0dBFS_sym_clip_24bit.wav -n stats
Quote:
Samples read: 132300
Length (seconds): 3.000000
Scaled by: 2147483647.0
Maximum amplitude: 0.999000
Minimum amplitude: -0.999000
Midline amplitude: 0.000000
Mean norm: 0.000015
Mean amplitude: 0.000000
RMS amplitude: 0.003884
Maximum delta: 0.999000
Minimum delta: 0.000000
Mean delta: 0.000030
RMS delta: 0.005493
Rough frequency: 9925
Volume adjustment: 1.001
|