Frauenhofer is junk like Blade or Xing. It's easily ABXable in CBR mode and even worse in VBR mode. It has still the smearing and preecho artefacts that the buggy reference encoder had that they once wrote for the MPEG consortium. The MPEG and ISO reference encoders and decoders are just a quickly hacked together piece of code to show how to write a working encoder. Nobody is expected to use it in real life, but sadly most early encoders where based on that reference code (Blade, 8Hz, and LAME up to 3.80). LAME started to use its own psycho accustics, since 3.00 because the Frauenhofer Psychoaccustiks contains some serious bugs.
LAME 3.90 ~ 3.93 where good, but then they lost it somehow. I had to send a remix to a befriended Artist over the Net and he had some problems depacking my FLAC version of the remix. So I tried Lame 3.95 and it sounded crap to me, I tried 3.97 -vbr-new -V 0 -q 0 and it was a bit better but still crap. Because I've read some good remarks about the Helix MP3 Encoder I tried that next. With V150 the anoying artefacts from LAME where gone, but Helix had some problems with the heavy spartial effects I was using in that remix (which LAME 3.97 didn't have). I was about to give up, when I decided to give the alpha of LAME 3.98 a last chance (I think it was about the time of A2 when I downloaded the source and compiled it). I hadn't high hopes, but the -V 0 encode was nearly perfect. The grainy noise was gone and also the spartial effects where perfect. The difference was so minimal to the original (I had a hard time to hear a difference at all) so that I decided to send him the LAME 3.98 -V 0 encode.
The Encoder is one thing, the decoder another. The MP3 decoders of Windows Media and Quicktime are not very good. If you want to make full use of all features (Bitrate reservoir, Freeformat, 24 bit resolution etc.) then you have to look for alternatives. The MAD library that most freeware and shareware players use is probably the best decoder.
There are a lot of urban myths about MP3, like:
- Frauenhofer did develope the MP3 codec (which is only partly true. They did develope a big part and wrote the reference encoder for the MPEG consortium, big main parts where developed by AT&T, Philips and Thomson)
- Stereo is better than Joint Stereo (That was only true for MP2, but even in the beginning of MP3 the results with Joint Stereo was better. The encoder has more room to encode this way as the differences between left and right channel are often marginal).
- CBR is better than VBR (it was only true in the beginning, but modern encoders make better use of the bit reservoir while even save some space that is bloated up with null bytes when you encode to CBR 320 kbps)
- MP3 is sonicaly better than MP2 (which is not true as the MP3 Design has serious flaws. But MP3 was the first format which offered acceptable quality at a low bitrate that you could download with an anlogue modem in no big time and that helped it to become this popular.)
Lame 3.98 sounds better than 3.97 with trumpets, distorted guitars and synthesizers.