I've performed an experiment with surprising results. The object of this
experiment was to evaluate lossy compression codecs. I decided to use an
excellent piano recording of a couple of Rachmaninoff preludes as the
material. I ripped these pieces to wav files, and converted the wave files
to each of 5 different codecs:

LAME mp3 at 256kb/s, for a 5:1 compression ratio
LAME mp3 at 128kb/s, for 10:1
WMA9 at 64kb/s, for 21:1
WMA9 at 32kb/s, for 42:1
WMA9 VBR10, for 50:1

The VBR10 variable bit rate file was the smallest one that could be
converted back to wav and subsequently burned to a CDR. This is what I did,
putting all 5 compressed versions and the original wav file on the CDR in
random order. I made 6 copies of this disc and passed them around to
discriminating friends, with instructions to rate the fidelity of each file,
and rank them in order. I didn't tell them what they should be listening
for. All these friends have excellent audio systems, in one case costing
over $40,000.

So far, 2 people have returned their discs with their comments. Neither of
them chose the original wav track as the best; neither chose the WMA VBR
file as the worst, inspite of its 50 to 1 compression ratio!

I'd like to expand this test to music having a wider dynamic and frequency
range. I'd like some suggestions from the group of a classical selection
running about 3 minutes that might be more revealing of codec differences
than piano music. Try to pick something fairly easy to locate, and
preferably a piece generally regarded as a great recording. I'll then
rerun this test with the new material. Hopefully, the degradation of the
highly compressed files will be easier to pick out.

Thanks for your input,

Norm Strong