dubmunkey wrote:right ive never been good at maths- someone hold my hand....what volume reference are we using? the peak value? the average volume?
greg
All of it. No reference. We're not compressing so we're not just changing peak, we're taking the level of the whole file down, like lowering a fader on a mixer. If you do this in a digital editor (i.e. destructively, not with a fader in Logic), all the audio information is quantized into the lower bits; the upper ones go unused. The math makes sure the sound uses as many (few) values as you would have when using 12 bits (4096 values, as opposed to 65536 when using 16 bits). It's a way of discarding audio information. When you raise the level back up (normalize), the lost information doesn't magically re-appear, so you keep the '12 bit' sound, but at a normal level. If you were to look at the finished file, you would see a 16bit file, but with 'stepped' levels; from 0 to 65536 in steps of 16.
I could draw it for you...