I know a little attenuation is always needed in every medium to compensate for people who apply equalisation etc (which boosts the volume of certain frequencies): to avoid clipping, there has to be a bit of head room. But 10db!?
Surely you get lower quality audio that way anyway as the sampling resolution isn't using the full spectrum of the wave (a bit like saying "we're only using integers, now rate these between 1 and 10%" - you get much more detail by offering 1 and 100, the full range).
I bet someone here knows! (hopeful smile
