Version 3 (modified by kmaclean, 15 years ago) (diff)


comparison of the sample formats:

See this link

8-bit: only 256 possible values for each audio sample. Only 256 levels of "loudness", you could say. Gives you a maximum dynamic range of 42 dB (decibels) - i.e. the loudest part of your piece can be only 42 dB louder than the absolute softest pin-drop sound. 8-bit is used for low-quality speech.

16-bit: 65,536 values per sample, 90 dB of dynamic range. This is used by audio CDs. Takes up twice as much space as 8-bit.

32-bit floating-point: over a billion possible values per sample, almost infinite dynamic range. Takes up twice as much space as 16-bit, and is a little more than twice as slow to process. Suffers very little degredation of quality when you repeatedly apply effects.

64-bit floating-point: takes up twice as much space as 32-bit float, necessary for some frequency-domain calculations (like the FFT filter).

Audacity typically uses 32-bit internally, so you get amazingly good quality during editing and mixing (better than a lot of other audio programs!), and then exports to 16-bit, which is audio CD quality.

More Audacity Notes: