User:Mjb/Digital audio

Here are some random notes about digital audio...just some things I want to remember.

Is it better to resample to 48 KHz instead of 44.1?
Converting 96 KHz to 48 is dividing by 2, whereas converting 96 to 44.1 is dividing by ~2.17687. Yes, it is mathematically simpler to divide by 2, but the process involves more than simple decimation, and the risk of audible noise really only comes from those other aspects of the conversion process.

See SRC Comparisons to see (visually) just how much variation there can be in implementations.

Downsampling involves first applying a lowpass filter to remove frequencies higher than the Nyquist (half the target sample rate). The steeper the slope at the high-frequency end of this filter, the more difficult it is to implement. I know that 44.1 requires a steeper slope than 48 to avoid adversely affecting the audible range (20 KHz and below), but does it really matter? I'd like to be confident that the audible range is as good and unfiltered as it can possibly be, so I tend to convert to 48 "just in case". But maybe that's not necessary?