Re: [Audacity-devel] Just to see if I could...fix clipping (Martyn Shaw)

[Roger D. <rb...@cs...>]

I completely agree that float samples going over 0dB (= 1.0) should not 
be clipped -- it's extra work to degrade the audio. There are many 
operations (filters with resonance, mixing two signals, even a simple 
delay) where it is impossible to design the effect to avoid making peaks 
higher, which is the main motivation for using floats in the first 
place. Well, I guess another strong motivation is avoiding underflow, 
but if the logic is that large float signals should clip just like 
integers, then does that say that very small signals should be rounded 
to zero? I hope not.

-Roger