
mshilarious
Banned
Well, here's my take on this, with Wavelab 5:
Using 2 different white noise files, 16/44.1 (I had laying around from mic tests I've been doing), I created a 2 track mix, master section down -6, then normalized (about a +2.7dB gain change, then using the same two files, an 8x 2 track mix, master section down -24dB, then normalized (similar gain change). One of the resulting mixes was reversed polarity, then the two mixes were summed. No dither was applied.
Analysis of the resulting mix shows noise in the least significant bit only (peak level of -90dB). It's possible that performance would be better using 24 bit or 32 bit float files to begin with, but I wanted a worst-case test, since I work with 16 bit files more often than I'd like. Maybe later I'll try the 24 bit test.
It's also possible that the difference is due to the two stage gain normalization process, but then I wanted a "real world" test, and as far as I can tell, it isn't possible to set precise gain changes using the Wavelab Master section, whereas normalization is calculated to three significant digits (in decibels).
Here is the FFT of the resulting mix:
Using 2 different white noise files, 16/44.1 (I had laying around from mic tests I've been doing), I created a 2 track mix, master section down -6, then normalized (about a +2.7dB gain change, then using the same two files, an 8x 2 track mix, master section down -24dB, then normalized (similar gain change). One of the resulting mixes was reversed polarity, then the two mixes were summed. No dither was applied.
Analysis of the resulting mix shows noise in the least significant bit only (peak level of -90dB). It's possible that performance would be better using 24 bit or 32 bit float files to begin with, but I wanted a worst-case test, since I work with 16 bit files more often than I'd like. Maybe later I'll try the 24 bit test.
It's also possible that the difference is due to the two stage gain normalization process, but then I wanted a "real world" test, and as far as I can tell, it isn't possible to set precise gain changes using the Wavelab Master section, whereas normalization is calculated to three significant digits (in decibels).
Here is the FFT of the resulting mix: