This article is from the Audio Professional FAQ, by with numerous contributions by Gabe M. Wiener others.
Both MD and DCC use lossy compression algorithms (called ATRAC and
PASC respectively); crudely, this means that the numbers coming out of
the machine are not the same as those that went in. The algorithms use
complex models of the way the ear works to discard the information
that it thinks would not be heard anyway.
For example, if a pin dropped simultaneously with a gunshot, it may be
reasonable to suggest that it isn't worth bothering to record the
sound of the pin! In fact it turns out that around 75 to 80 per cent
of the data for typical music can be discarded with surprisingly
little quality loss.
However, nobody denies that there is a quality loss, particularly
after a few generations of copying. This fact and others make both MD
and DCC useful only as a consumer-delivery format. They have very
little use in the studio as a recording or (heaven forbid!) mastering
format. [Chris]
 
Continue to: