Lame -V2 not transparent to me
Reply #20 – 2007-12-23 17:25:55
I guess it was me who first suggested to the OP to give 3.98b6 a try. There are some reasons to do sothe extremely bad pre-echo sample 'eig' has improved a lot with 3.98b6. So we can expect this improvement to be relevant for more pre-echo problems. The OP wrote about issues with drums, so 3.98b6 is a candidate for the solution of these problems. Meanwhile diewaldo has found it is the solution. 3.97 has a specific problem called 'sandpaper noise' which affects voices. This problem is gone with 3.98b6. This kind of problem can be heard for instance with the 'Birds' sample. There were certain weaknesses with Lame's psy model not just with 3.97, but also with 3.96 and other versions, and it was especially bad when using VBR (as VBR enhances the goods and bads of the psy model). These problems are gone with 3.98b6. My top sample for these kind of problems is 'trumpet' (before I heard a 3.97 -V2 encoding of trumpet the mp3 world was fine to me, but with this sample I was totally upset. With 3.98 the world is fine again). I'm not sure whether or not the 'sandpaper noise problem' also falls into this category - from mere listening the problem sounds similar. The HA recommendations are just a guideline, and only final versions will make it into it. With 3.98b6 we're certainly so close to final (qualitywise) that in my very personal opinion it would be unwise to prefer 3.97 final just for the academic sake of using a final version.