Informed rebuttal of VBR disadvantages appreciated
Reply #21 – 2008-10-16 21:29:07
...So, I choose to encode my MP3's at 320kbps CBR so that I give my crowd the best possible chance at hearing the music in the way it was originally recorded. ... ...I DO NOT believe that a VBR MP3 is acceptable, FOR ME, since I consider myself a professional. And since I listen to other DJ's, I like them to use good-quality stuff too ... This sounds a lot more reasonable than when reading the quoted passages. There's nothing wrong using CBR 320 with regard to minimizing the risk of wrong encoder decisions which can happen. But you should be aware that there are many possibilities for an encoder to go wrong:when deciding for using long or short blocks when deciding for l/r or m/s (decision can be avoided by using plain stereo but at the expense of a lower encoding precision as a tendency) when deciding for the amount of audio data required for a frame. In contrary to what you are thinking this process is involved also with CBR. CBR means constant frame data rate. Audio data rate however is variable also in the case of CBR as audio data can expand beyond frame border as well as not cover up an entire frame. It is true however that the decision process is less prone to errors when CBR or ABR is used. There is no reason however to general disbelieve in the VBR process of a good encoder like current Lame in the case of mp3. I'm sure there are a lot more decision making problems for an encoder no matter whether it uses CBR or VBR. As I said it's okay to play it safe to the utmost extent if you like to and don't have to care much about file size. But in a practical sense you shouldn't feel really safer than when using -V0. If you're looking at seriously bad encoded tracks it turns out that it's not VBR which is to blame. Take for instance extremely bad pre-echo sample eig (you'll find it in this forum). The majority of mp3 encoders will produce a very bad result even when using CBR 320. Contrast this to Lame 3.98's behavior when using best VBR quality -V0: the result isn't perfect but a lot better than that of many encoders' CBR 320 results (the Lame 3.98 CBR 320 result is of course as fine as the -V0 result). The fact that perfection can't be achieved with mp3 is the reason why most members here prefer an encoder setting which produces smaller files than when using CBR 320. The quality achieved is identical in a practical sense no matter whether you use -V0, ABR 270 or similar, or CBR 320. Compared to such a setting most members here prefer a lower quality demand like when using -V3 or -V2, simply because there's nothing wrong with the quality except on rare occasion (in which a higher quality setting often brings only a minor improvement). So it's a personal choice which quality setting to use, quality difference is zero in most cases, and it's only about how to handle the rare exceptions to this. Unfortunately you're a bit on a mission and you're wrong with this. You dislike VBR so much that you wrote a lot of fancy stuff about VBR's speed penalty which is really nonsense. With respect to quality you put the blame on VBR for no really existing reason (or do you have samples to back up your opinion?). You're disrespecting the fact that the results of high quality VBR settings are fine, and in those rare cases where they're not you cannot expect to get better results from CBR 320 compared to those of -V0.