"Jim Lesurf" wrote in message


As an experiment, I made up a 'test CD' a while ago to
try on some friends and colleagues. This consists of a
set of tracks of various types of music where the
original peaks well below 0dB, and versions I
deliberately clipped. Apart from the clipped sections the
two versions of each example are sample-by-sample the
same.

It has been interesting to see how hard/easy people have
found identifying the clipped version to be. :-)

Depending on the music, it can be hard or easy.

This seems to agree with something I discovered 20+ years
ago. When I designed the Armstrong 730/732 amps I fitted
a clipping indicator. It turned out to be quite difficult
to hear the clipping in many cases - although admitted
this is at levels well over 200Wpc so I am not sure what
the speakers (or ears!) were doing in some cases at these
levels in a normal UK domestic situation. 8-]

Case in point is the clipping indicator on QSC power amps. It starts
visibly illuminating at clipping that amounts to something like 0.02% THD.
Again depending it can be found to be flashing quite a bit, and yet the
sound may not be all that bad-sounding.

Do remember that lighting the top bit light does not
necessarily imply clipping - it is just another value,
and if the signal isn't trying to go beyond that, it
hasn't clipped.

Pop recordings use heavy compression, and when this is
done in the digital domain it is quite possible to have
sufficient control to peak to the same value every time.
There is no reason not to normalize the result up to max
level.

Many converters, espcially the cheap ones, won't convert cleanly right up to
FS. Keeping peaks 0.5-1 dB below FS will help ensure clean performance with
cheaper equipment.

Alas, my recent experience confirms that a number of CDs
have successions of samples well within 0.05dB or so of
the peak values allowed on CD-A. Level compression seems
much more common, but flat-top clipping seems far from
rare.

Agreed.

As you say, this seems utterly insane when many
rock/pop CDs squash the sound into a range of about 10dB
- on a medium that should be able to offer a range over a
million times greater!

There are two sorts of logical reasons to clip music. One is that it
constitutes an EFX. It's pretty well known that distorted music often sounds
louder than music that is cleanly reproduced. The other is the fact that
music with limited dynamic range can be more suitable when listening to
music is not the most important thing that the listener is doing.