This article is from the CD-Recordable FAQ, by Andy McFadden (email@example.com) with numerous contributions by others.
The first thing to know is that there are two kinds of jitter that relate
to audio CDs. The usual meaning of "jitter" refers to a time-base error
when digital samples are converted back to an analog signal; see the jitter
article on http://www.digido.com/ for an explanation. The other form of
"jitter" is used in the context of digital audio extraction from CDs.
This kind of "jitter" causes extracted audio samples to be doubled-up or
skipped entirely. (Some people will correctly point out that the latter
usage is an abuse of the term "jitter", but we seem to be stuck with it.)
"Jitter correction", in both senses of the word, is the process of
compensating for jitter and restoring the audio to its intended form. This
section is concerned with the (incorrect use of) "jitter" in the context of
digital audio extraction.
The problem occurs because the Philips CD specification doesn't require
block-accurate addressing. While the audio data is being fed into a buffer
(a FIFO whose high- and low-water marks control the spindle speed), the
address information for audio blocks is pulled out of the subcode channel
and fed into a different part of the controller. Because the data and
address information are disconnected, the CD player is unable to identify
the exact start of each block. The inaccuracy is small, but if the system
doing the extraction has to stop, write data to disk, and then go back to
where it left off, it won't be able to seek to the exact same position. As
a result, the extraction process will restart a few samples early or late,
resulting in doubled or omitted samples. These glitches often sound like
tiny repeating clicks during playback.
On a CD-ROM, the blocks have a 12-byte sync pattern in the header, as well
as a copy of the block's address. It's possible to identify the start of a
block and get the block's address by watching the data FIFO alone. This is
why it's so much easier to pull single blocks off of a CD-ROM.
With most CD-ROM drives that support digital audio extraction, you can get
jitter-free audio by using a program that extracts the entire track all at
once. The problem with this method is that if the hard drive being written
to can't keep up, some of the samples will be dropped. (This is similar to
a CD-R buffer underrun, but since the output buffer used during DAE is much
smaller than a CD-R's input buffer, the problem is magnified.)
Most newer drives (as well as nearly every model Plextor ever made) are
based on an architecture that enables them to accurately detect the start
of a block.
An approach that has produced good results is to do jitter correction in
software. This involves performing overlapping reads, and then sliding the
data around to find overlaps at the edges. Most DAE programs will perform