Just for the sake of completeness, shifting samples and time
stretching are two different things. I don't believe that jitter
correction (no matter if it is just a low frequency drift or a high
frequency jitter) can be successfully corrected using time stretching.
If the clock error can be estimated you could interpolate the samples
back onto a regular grid, which is exactly what varispeed does. The
varispeed pitch control would in this case be controlled by the output
of your FIR filter that does the averaging.
2008/1/14, Herbie Robinson <email@hidden>:
> Correct on both counts; however, I believe the existing algorithms
> are correcting for non-locked sample frequency clocks which are high
> frequency and low jitter. The output from parsing LTC is low
> frequency and high jitter. It is possible in a post processing
> environment: You could low pass the frame clocks with a massive FIR
> low pass filter to get rid of jitter and compensate for the delay by
> shifting samples. You still have the problem of picking the corner
> frequency that responds to speed variations fast enough, but doesn't
> let jitter through. Given that the audio quality is compromised by
> gunky heads, too, it would seem to make more sense to redo the
> transfer.
I agree completely.
--th
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Coreaudio-api mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/coreaudio-api/email@hidden
This email sent to email@hidden
