|Originally posted by Distroyed
I still don't like the concept of losing quality in lower volume portions of music...
The whole point of VBR, if it's done right, is to use fewer bits WITHOUT losing quality compared to CBR. i.e. it's possible to compress the original stream using fewer bits somehow, so just go ahead and do so.
Of course, since we're in the world of lossy compression here, this is a very tricky thing, and depending on the encoder, VBR may or may not change the quality. Some of the Xing and Fraunhofer codecs were famous for screwing up your sound if you used VBR, but LAME is supposed to be pretty good about it.
One key point is that with at least some compression systems/encoders, if the psymodel doesn't think it needs more than x bits at a time to represent the original waveform, then throwing more bits at it does *not* improve quality, it just wastes file space, because the encoder won't do anything in those extra bits that it wasn't already doing. Notably, the Musepack encoder (for .mpc files, the compression system I use after HydrogenAudio recommended it) currently does this sort of thing beyond some particular bitrate; the very highest "quality" settings do ABSOLUTELY NOTHING DIFFERENT than the ones just below them, except use more filesize (this statement comes from the designer/coder of .mpc). In fact those settings are only even there for parity/compatibility with other codecs and frontends reasons.
The whole point of the previous paragraph is that it is very false to assume that more bits == more quality. It might, in fact, mean that. Or it might not. It depends on the source waveform, the encoding method, and the particular implementation in the encoder.