wjm@lcuxc.UUCP (B. Mitchell) (01/23/85)
Frankly, I'm somewhat suspicious of the KEF "anything over 16KHz is
insignificant" results. My experience is that the high treble frequencies
do contain some information that does affect our perception of the sound,
even though the fundamentals may be at lower frequencies. Keep in mind that
the ear is a non-linear device and that these frequencies, which are usually
harmonics produced by instruments in the 3-10 KHz range, do not behave
necessarily as one would predict if one assumed the ear were a linear
system.
Since most analog devices (microphones, amplifiers, etc.) used to record
a musical performance can be, and usually are, designed to have a bandwidth
in excess of (almost) DC (let's say a few Hz) to 22KHz, anti-aliasing (low
pass) filters are required before digitizing these signals. These filters
can introduce distortion into the signal (by adding noise, phase shifts, etc.)
besides their intended effect. However, they are necessary for the digitization
to properly represent the BAND-LIMITED signal between DC and 22KHz.
Remember that the sampling theorem only applies to BAND-LIMITED signals, and
therefore a 44+ KHz sampling rate is adequate to accurately represent the
22 KHz bandwidth. The problem is NOT with the communications theory, but
with the selection of the sampling rate, which sets the maximum analog signal
bandwidth that can be represented by the digital signal.
I'd also like to thank Phil Karn, rabbit!jj, Dick Grantges (hound!rfg) and
the rest of you guys for pointing this out.
If you want more information on information theory, you can start with
"Principles of Communication Engineering" by Wozencraft and Jacobs, and
move from there into the literature, e.g. IEEE Transactions on Communications,
IEEE Communications Magazine, and the IEEE Transactions on Information Theory.
186,000 miles per second - its not just a good idea, its the LAW!!
Bill Mitchell ({ihnp4!}lcuxc!wjm)