I recently performed a small 96k listening test as practice for the upcoming 128k test. I was going to do an exhaustive 28 sample test but life got in the way and to date I have only got through four.

I tested the following formats:

LAME 3.98a2 -V7 --vbr-new
aoTuV Vorbis b4.51 -q2
MPC 1.15v --quality 3
Nero AAC Internet Profile
WMA Std q50
Shine mp3 as Anchor



I was only able to hear a difference between Nero and Vorbis once, and, trust me I tried very hard to tell the top two competitors (which I assumed were Nero and Vorbis) apart in every test; the difference in final notation must be within the margin of error.

Uploads Thread