>
> thanks rafael
>
> > You use numlines[] to calculate your perceptual entropy, yet
> > you are setting it here:
>
Could this explain our problems with detecting pre-echo events?
the pe formula in LAME would detect real obvious ones, but miss the more
subtle ones that the FhG encoder would pick up. That's why we
had to put in a few alternatives. But the real problem might
have just been a bug in the PE formula!
By the way, there is a bug that was introduced into the
3.05pre* codes. It will do weird things to frames that have
a lot of energy in a few coefficients (like perfect sine waves).
it's fixed, and will be on the web site this weekend.
Mark
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
