44/48 kHz, USB Audio and Helix

[Tony Hardie-Bick]

On 13/09/17 00:37, deeplfo wrote:
> The sampling frequency value represents how many audio data points you measure in a given time period, thus defining your audio quality.  It's the bit depth that represents the dynamic range of each of those data points, i.e. the headroom. So, for more detailed (accurate) representation of the audio, you would record at say 24 bits or more.  
> 
> You probably get some what of an improvement in audio quality by bumping the sampling rate to higher than 44.1 K values, but I'll stop here when it comes to discussing the mertis of higher sampling rate and perceived audio qualities :-)

I can jump in briefly... directly audible material can be fully
reconstructed from 48KHz 16bit samples. However, if you then post
process those samples, for example with EQ, the dither that was used to
avoid quantisation distortion will become louder in the boosted part of
the spectrum. So, it's almost like analogue, in that the noise floor is
audible. This is assuming the code was written well. If not, you'll hear
quantisation noise. Which has an aesthetic. But you want a choice on
that aesthetic. So you go for at least 24bits, to keep any such noise or
distortion much quieter.

As regards the sample rate, even 44.1kHz will reconstruct what you can
directly hear, but any effects that are then applied, such as EQ, or
anything subtle, will be compromised (-6dB per octave not necessarily
being -6dB per octave, for example) by the low sample rate. So, several
kinds of effects will benefit from a higher sample rate, and, to avoid
the high computational cost of sample rate conversion in and out of each
effect, you just run the everything at a higher rate. Preferably 96kHz
or more. Basically, everything gets easier to do well, the higher the
sample rate, even if, in theory, there is basic stuff, like simply
mixing channels together, where it makes absolutely no difference at all.

HTH,

t.