Dragan Noveski wrote:
If using IEEE754 floating-point, your data should have been normalized before it sent to disc - so you lose nothing (although if you're using a flush-to-zero mode for FP arithmetic you may introduce crossover distortion).
cheers R