Simply dropping samples doesn't work properly.  Take a 96kHz signal with
a loud sinewave at 46kHz.  If you go to 48kHz sampling rate by throwing
away every other sample, you should ideally end up with silence (the
46kHz signal can't be represented at that sampling rate) but instead
you'll get a very loud 2kHz whistle that wasn't present in the original
at all; it's an artifact of the downsampling.  To get rid of the
artifacts you need to low-pass-filter (ideally a brickwall at 24kHz, in
this case) first.

And linear interpolation to fill in the gaps after upsampling doesn't
work properly either: it creates artifacts.  The correct approach is
sinc interpolation, but in an ideal sense that is infinitely expensive
to compute.

So it's all about tradeoffs:  the ultra-easy computation produces
really bad results sometimes, but an ultra-correct computation is
really really slow.  And different SRC applications take different
approaches, with different degrees of success.


-- 
inguz
------------------------------------------------------------------------
inguz's Profile: http://forums.slimdevices.com/member.php?userid=1139
View this thread: http://forums.slimdevices.com/showthread.php?t=32958

_______________________________________________
audiophiles mailing list
[email protected]
http://lists.slimdevices.com/lists/listinfo/audiophiles

Reply via email to