|
From: Eric W. <nor...@yh...> - 2015-12-13 08:06:30
|
R Kimber <ric...@bt...> wrote: > If I use sox to convert a 16 bit 44100 file to a 24 bit file, test.flac <snip> > But if I do sox test.flac -n stats I get (inter alia):- > Bit-depth 16/16 > Num samples 21.6M > > Why is this? I'm no expert in this, either; but I'll try to explain as well as I can. 24-bit audio has more dynamic range than 16-bit audio (144dB vs 96dB, you get 6dB per-bit) and thus a lower noise floor. However, when you convert a 16-bit file to 24-bit, you do not lower the noise floor automatically. All the background noise in the 16-bit file stays there since it was in the original file. You don't lose anything, going from 16 to 24, but you don't gain anything, either. The extra 8 bits is just background noise (inaudible at normal listening levels). Keep in mind even the best audio equipment cannot record or playback the entire 24-bit range of 144dB. The best studio ADC and DAC equipment is probably around 120dB nowadays (plenty enough). SoX and nearly all audio software converts to 32-bit (or even 64-bit) internally. This gives plenty of headroom for effects and eases calculation for CPUs. So if you run some effects which alter the audio, you'll end up with a file that takes advantage of what 24-bit offers. |