Hi everyone. I've made an observation with the versions of the Fraunhofer ACM codec
that I installed in about January some time & Lame 3.62. Yes, they're old versions, &
maybe a lot has changed since then, but... this is what I did-
1- Encoded a sample with moderately low stereo separation with both Fraunhofer &
Lame, to 22050Hz - 56kBit/sec - Joint Stereo.
2- Decoded with Fraunhofer.
3- Mixed to Mid/Side as PCM (manually) & compared the voiceprints.
Lame's output appeared to just have the Side channel at a lower bitrate, with both
channels with a lowpass filter of 7756Hz (I think). Fraunhofer looked to be using a
much lower LPF for the Side (around 3.5 kHz?) than it was for the Mid (which was close
to full bandwidth).
Is Fraunhofer's trick like the "Side channel starving" some of you were talking
about earlier? Well, whatever it is, I think it would be a great asset to Lame if the
user could set the filter of Mid & Side independently of each other, & independently
of encoding normal L/R frames at least until some good default settings are found
(that's if it hasn't already been implemented). We might be able to save a lot of bits
that way. All Lame should have to do is discard the high frequencies on the Side
channel before the FFT data goes through the masking algorithms.
I haven't tried filtering Mid/Side separately before encoding yet... I might try
that later & see if it makes a difference with Lame.
I know these filters are cheating, but MP3 in itself is cheating, right?
Shawn
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
