From: Michael E. DiFebbo <medifebbo_at_rcn.com>
Date: Thu, 02 Feb 2006 09:37:29 -0500

Frej Bjon wrote:
> There's one thing to watch out for, though: joint stereo. Unless you
> really, really need the compression, it's better to use real stereo,
> because joint stereo will "muddy" the stereo field and make it
> narrower. This type of degradation can easily be heard in even the
> most noisy situations with earphones.

This is a common misperception, but at best it is only partially true.
The term "joint stereo" has two different meanings when used with
respect to encoding MP3s.

The joint stereo method that you refer to is "intensity stereo."
Intensity stereo is only used by modern MP3 encoders for very low
bitrate files (lower than 96kbps if I remember correctly). For higher
bitrates, "joint stereo" refers to M/S stereo.

Intensity stereo encoding functions on the principle of sound
localization - by removing the stereo component of sounds that humans
cannot discern the direction of (i.e. the lowest bass frequencies).
Intensity stereo coding does not perfectly reconstruct the original
audio because of the loss of data resulting in the simplification of the
stereo image, and can produce unwanted artifacts with certain types of
source material.

"Mid/Side stereo," on the other hand, encodes stereo information by
using a mid channel , which is the sum of the left and right channels,
and a side channel, which is the difference of the left and right
channels. Unlike intensity stereo coding, is NOT in and of itself lossy.

M/S stereo can be more efficient than left/right stereo (what you refer
to as "true stereo"). Let's say that you have a recording which has a 5
second portion consisting only of a vocal. Vocals tend to be panned
directly in the center of a mix, and therefore, the left and right
channels are the same. With left/right stereo encoding, identical
information would be stored in both the left and the right channel.
With mid/side stereo, on the other hand, most of the information in
those frames would be in the mid channel, and very little of the
information would be in the side channel. Thus, the encoder can use
fewer bits to capture the information about the side channel. Either
method produces results that are mathematically identical when decoded,
but since the M/S method is more efficient in circumstances like the one
in my example, the use of M/S stereo can produce smaller file sizes (or
higher quality at an equivalent file size) compared to using L/R stereo
exclusively. Note that encoders can use either M/S stereo or L/R stereo
on a frame by frame basis, so even with "joint stereo" enabled, the
encoder still has L/R stereo as an option for frames that require it.
Received on 2006-02-02