This article is from the CD-Recordable FAQ, by Andy McFadden (email@example.com) with numerous contributions by others.
A common problem when creating an audio CD compiled from many different
sources is that the sound is at different volume levels. This can be
slight or, after you've cranked up the volume to hear the first track, very
much the opposite of slight.
There are actually two issues that determine how loud the music sounds.
The first is the signal amplitude. Put simply, if you open a WAV file,
this is how close to maximum the squiggly line gets. You can adjust the WAV
file so that the highest amplitude is at maximum with the "normalize peak"
function of a sound editor. Some programs, such as Roxio's Spin Doctor,
may even do this for you automatically.
The second major issue is the dynamic range compression. This differs
from data rate compression in that it doesn't make the WAV file smaller.
Instead, it can make the quiet parts louder and the loud parts quieter.
A CD-DA has a dynamic range of about 96dB. If a symphony is recorded with
a range of more than 110dB, it has to be compressed to fit on a CD-DA.
In practice, you don't want whispers to be inaudible and shouts to be
deafening, so the audio is often squeezed into an even narrower range.
Radio stations often compress their broadcasts "up" so that music can be
heard more clearly by listeners in cars or work environments.
(According to Ken Pohlmann's _Principles of Digital Audio_, 4th edition,
page 35, ideal 16-bit quantization of a sinusoidal waveform is 6.02n+1.76
decibels, or 98.08dB. Using "dithering" techniques, it's possible to
extend the effective resolution well beyond this, because of the way
the ear perceives sound. There is an *excellent* introductory article
at http://www.digido.com/ditheressay.html. Compression is more often
employed on pop music recordings, where louder is better, than something
like classical music, where accurate reproduction is desirable.)
To make a CD that sounds like it has equal volume across all tracks, you
need to have the average sound level uniform across all tracks and have the
peak volume be about the same on all tracks. One program that does
essentially this is Audiograbber v1.40 and later, available as shareware
from http://www.audiograbber.com-us.net/. (As of v1.41, you went into
"Normalize Settings" and hit the "Advanced" button.) The tool is a little
clumsy for serious audio mastering, but should do fine for preparing a
"mix" CD that you'll be listening to in your car.
Another tool is "WAV file leveler", at http://www.plompy.co.uk/software/.
Some programs approximate compression by letting you normalize against
average RMS power. In this case, you are using a value that more closely
matches the apparent loudness of the recording.
If you aren't dissuaded yet, http://www.digido.com/compression.html has
an excellent article on compression, intended primarily for the budding
recording artist but a good general reference nonetheless.
has an excellent article entitled "Over the Limit" about the Louder is
Better phenomenon in professional recording. The author examines the
progress of the trend by analyzing clipping and power levels in five
different Rush CDs recorded from 1984 to 2002.
Sidebar: "dB" is the abbreviation for "decibel", a signal strength ratio
measured on a logarithmic scale. In a WAV editor like Cool Edit, which
can show the sound level in dB, the signal level doubles every time you
add 6dB, and the "loudness" doubles every 10dB. This is different from
signal power levels, which double every 3dB (what you see in a WAV editor
is analogous to voltage, not power). Detailed information is available
from the Acoustics FAQ at http://www.campanellaacoustics.com/faq.htm.
See also http://www.ews64.com/mcdecibels.html and
http://www.modrec.com/about/excerpt.php. There is a comparison table at
http://www.gcaudio.com/Archives/volatgeloudness.htm that breaks things