On Sun, 29 Jun 2003 01:07:46 -0500, Lord Hasenpfeffer
wrote:

I spent time today browsing websites and reading FAQs, etc. to find out
more information about the Psycho-Acoustic Model employed by MP3. It
is indeed similar to MiniDisc's ATRAC compression scheme in that it
removes not only "masked frequencies" but also frequencies which
are determined to be "too quiet to be heard" by common human ears.

This proves to me that MP3s encoded from "quieter WAVs" which have
been ripped from CDs mastered with low, average amplitudes will suffer
more at the hands of a lossy audio data compression algorithm than
will MP3s encoded from WAVs which have first been "appropriately
normalized" (i.e. Linux: $normalize -ba -10dBFS ) prior to being encoded.

There is a similar and very lengthy conversation going on at

http://www.hydrogenaudio.org/index.p...=1&t=10637&hl=

There is at least one lame developer (Gabriel) who hangs out at this
forum, so you might be able to catch his attention.

As others have already posted, newer versions of Lame have a variable
ATH, which depends on how close the signal is to full scale.

There are a couple of questions this raises:

1) Is it better to try to bring up soft music near 0dBFS and then
apply mp3gain afterwards? How much quality is audibly gained by doing
this?

2) Should loud music be brought down so that the resulting encoded
bitrate is smaller? How much quality is audibly lost by doing this?

ff123