This article is from the CD-Recordable FAQ, by Andy McFadden (email@example.com) with numerous contributions by others.
An audio CD cuts a one-second "egg" of sound into 44100 slices, and
measures the "height" of each slice from 0 to 65535 (16 bits). It does
this independently for the left and right stereo channels, using a format
called Pulse-Code Modulation, or PCM. The technical shorthand, which you
may have seen in a sound editor, is "44.1KHz 16-bit stereo PCM".
Measuring the "height" of each slice is called quantizing. The round-off
error in the measurements is called quantization error. The problems
associated with the error can be reduced by applying "dither" (low-level
The reason for the number 44100 is explained in section (2-35). The choice
of 16 bits is also fairly arbitrary, but extremely convenient on a computer.
There are other problems when digitizing (e.g. aliasing) and when converting
back to analog form (e.g. jitter). See
http://www.tc.umn.edu/~erick205/Papers/paper.html for an introduction.
Newer audio formats, such as Super Audio CD and DVD-Audio, offer different
sampling rates (up to 96000), quantization (up to 24 bits), and numbers of
channels (e.g. 5.1 surround-sound).