On Mon, 23 Nov 2015 11:44:00 -0500
Paul Davis <paul(a)linuxaudiosystems.com> wrote:
On Mon, Nov 23, 2015 at 11:33 AM,
jonetsu(a)teksavvy.com
<jonetsu(a)teksavvy.com
> wrote:
> This "up-sample" as far as I understood
from the course, does not
> exist in reality as far as the quality of audio is concerned. A
> demonstration can be easily made using Audacity and zooming up to
> see the actual sampling points of both files, the original and the
> "up-sampled": the file that has been resampled from lower sampling
> rate will not add anything at all but more sampling points on
> straight segments of the audio. It cannot add quality. It cannot
> create curves.
Your understanding is incorrect. Please watch this:
https://xiph.org/video/vid2.shtml
If you have time (40mins to an hour), I'd also
suggest you watch
this):
https://www.youtube.com/watch?v=BYTlN6wjcvQ
The meat of this one starts at 16:48
https://youtu.be/BYTlN6wjcvQ?t=16m48s but the preamble is worth
watching for other reasons too.
Thanks for the links I will watch them later on this week.
There's one thing clear though. If you have a sine wave with say 16
sampling points, much of the curves will be lost. If this is saved to
a file and the file is resampled with 128 sampling points, it will not
add the curves. The original is gone, the resampling does not know about
the original source, it cannot and it will not make it up.
So if the sine wave ended up sounding like a robot in the file,
a linked collection of edges, direct resampling at a higher rate will
not add a more natural expression. This is what I described above.
There might be special processing that extrapolates, creates
points between existing sampling points thus adding slopes and curves,
although w/o knowing the source only fiction can be created,
your audio interface runs at one sample rate. all
tracks not at the
same SR would have to be resampled. Not a good idea.
OK, that's the limiting factor then. It is not possible to have various
sampling rates inside a session for hardware interface reasons.