Stereo sound can be thought of as independent left-right channels or as mid-side channels. Mid-side, or the representation of stereo sound as the sum and difference of the two channels, is similar to YUV (YIQ, YCC) color coding in that it is a good way to decorrelate audio signals. FM radio uses mid-side stereo, encoding the stereo signal in a 38 KHz amplitude modulated subcarrier. (NTSC and PAL use an AM subcarrier for color information.)
MP3 joint stereo encoding also uses mid-side, as many songs' vocals are almost entirely in the mid channel, and the ear doesn't hear as much distortion in side as it does in mid, letting encoders such as LAME play psychoacoustics and allocate fewer bits in the side channel. Ogg Vorbis does not use mid-side, as U.S. Patent 5,481,614 restricts the use of mid-side in audio compression.
There is another use for mid-side: vocal elimination. Often, you can remove the vocals from a rock song through this procedure:
- Rip the CD, or otherwise obtain a stereo linear PCM wave file.
- Convert the file to mid-side using your wave editor's channel mixer.
- Low-pass filter the mid channel at about 150 Hz. (You might want to adjust this figure for the style of music you listen to; a 6th-order Butterworth filter at 150 Hz worked for "Black Hole Sun" by Soundgarden.) This removes everything from the mid channel except the bass.
- Convert back to left-right.