The subject of CD loudness has been discussed on many occasions, but it still divides people. Some still express concern that their recordings don't sound anywhere near as loud as commercial products, while some hate the fact that commercial releases are over-compressed, and wish that material was dealt with in a slightly more caring way.
The first thing to get clear is that, although the mastering process often uses techniques to increase the apparent loudness of a piece of music, that is not its primary aim — the loudness aspect of the process is often only introduced because of client demand.
In digital recording, peak levels can never pass digital full scale, or 0dBFS, so assuming you have two or more pieces of music that peak at the same maximum level, what makes one sound louder than the other? The answer is that the human ear is more responsive to average signal levels than to short peaks, so the higher you can get the average level of the music, the louder it will seem to be. The two main strategies used to achieve this are compression and limiting. Compression can be useful in mastering, as it often helps the whole piece to sound more homogeneous, but there's always the temptation to push it too far just to make the music seem a lot louder. Normally, a setting with a low threshold and a low ratio is used, to squeeze the overall dynamic range by a few dBs. Providing you're sensible about how much compression you apply, the music may indeed sound better as a result, as well as sounding subjectively louder. Of course, you can't push up the average level indefinitely, because there are those awkward peaks (such as snare drums, for example) that poke out of the top of the waveform.
Mastering engineers use limiters to prevent these peaks from hitting the end stops and causing distortion, but such limiting is often quite gentle, with only two or three dBs being trimmed away from the loudest peaks. However, for every dB you trim off the top, you can raise the average level by the same amount without clipping the output signal. Doing this in moderation is normal practice, but it's not hard to find records with 10dB of limiting on the peaks, and a huge but un-dynamic sound.
But at what cost? At some point, the amount of compression and limiting you add starts to damage the music, adding distortion that makes it sound fatiguing; at that point you're trading quality for loudness. Music is supposed to have dynamics, and a lot of mastering engineers would like to be allowed to retain those dynamics. But the reality is that everyone is competing to have the loudest (and most aggressively compressed) material out there, so the only way to keep up with the latest trends is to, in effect, damage your music.
This state of affairs has resulted in a wave of new tools that increase subjective loudness with fewer audible side-effects. Examples include Sonnox's Inflator and Waves' Maxx Volume, but once everybody has these tools, we're back to the position where the engineer who's happy to push everything hardest sets the loudness targets for everybody else. The only solution that I can think of is to warn every client of the dangers of a louder mix. Then, and only if they are sympathetic, we may be able to bring the dynamics back into commercial music.
Paul White Editor In Chief