Re: [LAU] Requirements for huge multitrack editing

From: Fons Adriaensen <fons@email-addr-hidden>
Date: Tue Nov 13 2012 - 12:48:57 EET

On Mon, Nov 12, 2012 at 02:39:17PM -0800, Len Ovens wrote:
 
> To change things to 48k you would have to do that outside of ardour. I
> would suggest against doing that except as a final step after mixdown as
> any time differences from track to track in the resample would be
> magnified by such a long track. Even in shorter cases stereo imaging could
> be affected.

You must be using a *really* crap resampling app if you get that
sort of effects.

If you import the tracks in Ardour with resampling, or if you
use sndfile-resample, the work is done by Erik's libsamplerate.
This uses (double precision) floating point to maintain the
filter phase, and that will indeed lead to accumulating roundoff
errors. But in practice these are *absolutely insignificant*.

I actually did the test, checking the timing after 10 hours
of resampling from 44.1 to 48 kHz. Within the precision of
the test (around 1 millionth of a sample, 25 picoseconds)
there was *zero* error.

And if you use zresample the problem doesn't exist by design,
as it uses only rational arithmetic (i.e. all integers) to do
the same calculations.

Ciao,

-- 
FA
A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user@email-addr-hidden
http://lists.linuxaudio.org/listinfo/linux-audio-user
Received on Tue Nov 13 16:15:01 2012

This archive was generated by hypermail 2.1.8 : Tue Nov 13 2012 - 16:15:02 EET