Hello

I was wondering whether someone could enlighten me about VBR NEW that is enabled as default in the latest LAME. I have read that it is a new algorithm for "doing" vbr. So I presume this means that the a MP3 encoded at --preset extreme will actually be different from a MP3 encoded with --preset fast extreme (-V 0) ?
And what is the difference between these two settings in terms of the sound?

The reason I ask is that my MP3 player (Cowon iAudio A2) does not like vbr new and I get glitches on playback. I am hoping to understand the issue better so that I can try to take the best course of action.

No audible differences, at least none that were ever clearly proven in an objective ABX test by anyone on these boards. In fact --vbr-new is generally recommended over the old vbr algorithm due to its superior speed without affecting the quality level.

Sad enough that there are still MP3 players out there that don't properly support VBR, no matter whether it's only the new algorithm or both ones. I guess you already tried your luck looking for a firmware update, hence in that special case you won't have any choice but to encode using the old VBR mode. The encoding process will be noticeably slower, but quality certainly won't suffer at all.

I agree with Junon. In practice most people will struggle to correctly identify differences among any of the four settings from -V3 to -V0 when listening to music, or for that matter CBR at -b 320. Very young people might identify some missing high frequencies on occasions with -V3. As -V0 (formerly known and still aliased as extreme) provides an extra margin of safety over the just-transparent settings (either -V3 or -V2) so the minor differences between VBR methods are even more unlikely to be enough to get close to breaching transparency.

Regarding your Cowon's problems with VBR, either:

1. use the old VBR mode as suggested by Junon, if that works, or

2. you might try to use MP3Packer (or WinMP3Packer), which can convert VBR files into the smallest possible CBR file that can encode all that data. Despite a few 320 kbps frames in most VBR files, you might find that 224 kbps or 256 kbps CBR is often enough for conversions from -V0 thanks to smart use of the bit-reservoir to borrow unused bits from earlier frames. The CBR file will be larger than the original VBR file, thanks to being padded out to a fixed CBR bitrate but it will produce the exact same audio output and
would presumably work in your Cowon and present only a modest file-size increase. Very simple files (e.g. near-mono) will still benefit from low bitrates in CBR (e.g. 128 kbps), while rare and very hard-to-encode files may become 320 kbps CBR , so an album of 8 to 14 tracks of different CBR rates might come out at an average bitrate that isn't itself a CBR rate yet ought to still remain transparent.

If, for example, you converted -V2 VBR files to CBR with MP3Packer, you might perhaps (by guesswork, not representative bitrates) come up with an album of various CBR files that averages 232 kbps over the album, compared to 197 kbps for the original album of VBR files. But there might be one file that needed CBR 320 to cope with a difficult-to-encode section of successive 320 kbps VBR frames, so that file got 320 kbps when converted to CBR. This might indicate that if you'd encoded in CBR from the start you'd have needed to encode the whole album at 320 kbps to remain transparent on the tricky file.

Thanks for the replies.

I guess there's nought to be done but encode a little slower.

Great tip about MP3packer btw, as this little prog will save me loads of time with the VBR NEW mp3's that I've already got

VBR NEW = The new VBR algorithm. It's alot faster than the old algorithm.

Aim I missing something here? Isn't MP3Packer for removing padding from 320 kbps CBR MP3s; i.e., for converting 320 CBR --> very high-bitrate VBR?

yeah. but it can also reduce the size of VBR'd mp3 files by a few bytes. not a whole lot but it's still something.

and the point for me is by converting vbr new to cbr, the mp3 will play on my portable player

Hmm I spoke too soon.
Repacking did not fix my problem.

So all it seems that I can do is to re-encode my MP3's.

Anyway, what this means is that a VBR NEW encoded MP3 is not the same as the same file encoded using the old, slower encoding method.

From the previous comments, it seems the two encoding methods produce mp3's that are identical acoustically. However, they must be different somehow...

Anyone know what the difference is?