If you want to test the average loudness of your files you can use libebur128. I found amazing differences between same album re-editions. I'm not an audiophile, I just have a High-Enough-Fi systems.<p><pre><code> # cd /tmp
# wget http://www-public.tu- bs.de:8080/\~y0035293/libebur128-0.1.11-Source.tar.gz
# tar -zxvf libebur128-0.1.11-Source.tar.gz
# cd libebur128-0.1.11-Source
# cmake .
# make -j 10
# gcc minimal_example.c -I../include -L../ -lebur128 -lsndfile -o r128-test
# r128-test "t/Yann Tiersen/1999 - Black Session/03 - Life on Mars (feat. Neil Hannon).flac"
global loudness: -17.2 LUFS</code></pre>
The loudness war is actually pretty much over. Lots of the newest dance music from popular artists (deadmau5, Justice, etc.) have lower RMS than mid-2000s rock, which is the genre and time when lookahead loudness maximizers were seeing the most abuse.<p>(The article is full of technical inaccuracies. Compressors, depending on how they are set, do not reduce the attack transients of a sound like a snare drum, they <i>exaggerate</i> it. Also, it's from 2009.)
Related and also somewhat dated: Apparently youngsters are starting to prefer the artifacts of MP3 sound in general:<p><a href="http://radar.oreilly.com/2009/03/the-sizzling-sound-of-music.html" rel="nofollow">http://radar.oreilly.com/2009/03/the-sizzling-sound-of-music...</a>
The "Digital Compression" section feels misleading. Maybe that's my misunderstanding, though.<p>As I understand it, when we compress audio file, so from a 100s MiB of raw wave data it'll become a several-MiB file, we perform a (discrete) Fourier transform, get rid of less significant frequencies (thus compressing the data), and record the coefficients. Please, correct me if I'm wrong.<p>So I don't get the "rough edges you end up with in the digital recording" part. It is applicable to pre-compressed data, but that's another topic.