Monty wrote:
>
> > Does anyone have a GOOD quality Sample Rate conversion algorithm ?
>
> See any elementary discrete time audio processing book and look under 'integer
> ratio rate conversion'. It's mostly an exercise in upsampling, lowpassing and
> downsampling. The computational complexity is proportional to the product of
> the ratio integers.
>
> For example, going from 22050 to 44100, just make every other sample a zero,
> then apply a lowpass (however you like) with a cutoff just below 11.025 kHz.
> Now downsampling is needed in this case (well, the downsample factor is 1).
>
> This is a potentially expensive algorithm (eg, 44.1kHz to 8kHz requires much
> more suckage), but classically the most 'correct'. It will exhibit no blocking
> effects and minimal aliasing (you'll be limited by the shape of your lowpass,
> but that will be true of any approach). It can be implemented in either the
> time or frequency domain (time domain is faster for low order filters, cheating
> with an FFT is faster for high orders).
>
> There are other slightly messier ways to do things with FFTs so long as one
> pays attention to edge effects.
There is a much better way to do SRC using polyphase filtering. It also allows
SRC between ANY two sample rates.
See the papaer by Julius O. Smith:
http://cm.stanford.edu/~jos/resample/index.html
Erik
--
+-------------------------------------------------+
Erik de Castro Lopo [EMAIL PROTECTED]
+-------------------------------------------------+
"If dolphins are so smart, why do they live in igloos?" -Eric Cartman
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
