> someone please correct me if im massively confused here... the first
> thing you do to the input when encoding is convert it to frequency
> spectra, right? so why not just treat 22khz input exactly the same
> as 44khz, except half all computed frequencies and fft windows? why
> ever convert back to pcm?
In a nutshell, because such a frequency domain operation is a circular
convolution on the block in the time domain. You end up blurring the edges of
the time-domain block into each other.
Monty
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
