Path: Music > Conversion > CBR v VBR and Bitrates

CBR v VBR and Bitrates

So what does CBR and VBR mean, in the first place? CBR (Constant Bit Rate) means that the music is encoded with a constant bit rate, like 128 or 192 kbit per second. The disadvantage of this approach is that some passges are more difficult to compress than others. Complete silence, for example, needs exactly 0 kbit. VBR (Variable Bit Rate) tries to optimise things so that more difficult parts of the music are encoded in more bits while less difficult parts get less. There are two basic approaches to achieve this effect: one is to define a target bitrate (say 128 kbit) and then shuffle around the bits until the encoder has found the best possible distribution of bits between the hard and easy parts. This generally requires two passes over the music: a first preliminary pass, to collect the bits that make up the music, and a second to distribute them fairly. (That is how WMA VBR works.)

MP3 can do something similar: you just choose a target bitrate and the encoder then tries his best to compress the music and stay within your chosen rate. But MP3 can also do a more complicated sort of VBR that doesn't have a target bitrate as such. Instead you choose one of nine quality levels; the music then is encoded accordingly. On the one hand this approach means that the sound will always possess at least the chosen level of quality but on the other hand you have no idea how big the MP3 files will end up. Some music titles (even tracks from the same album) may be encoded with an average bitrate of 104 kbit while others may need 160 or even more. If you encode a large number of tracks, you will find that this value tends to fluctuate around a median value which is characteristic for that level of quality. (In my case, I have tracks with 96 kbit minimum and 180 kbit maximum, while the median value hovers around 120 kbit.)

There may be exceptions but in general I think one of the VBR methods should produce better results than a simple CBR. The target bitrate method is simpler to understand and produces files whose size is known up-front. The quality-level method doesn't and is slightly more complicated but it guarantees a minimum level of quality. I use this latter method.

What to do? If you just want the music digitised with a minimum of fuss or for casual listening, encode with 128 or 192 kbit target VBR, either with MP3 or WMA. Just encode a few test tracks with the two bitrates and listen to them: if you can't hear much of a difference between the original CD tracks and the compressed music files, use a bitrate of 128 kbit. If you do hear a difference, settle for 192 kbit. (Or even redo the test with 256 kbit.)

However, if you plan on converting a whole CD collection and you want to make sure that you get adequate results, you will probably want to test different levels of quality in a series of ABX tests. That'll take a few days but it will pay off: you will probably listen to the files you are going to produce for a long time to come. In case that's your route I can give you a few more hints and tips on the software and the best way of doing this, just drop me a line. There are also many internet discussion boards (newsgroups or web-based fora) around where these things are discussed, sometimes in exhausting detail:-). See, for example, http://hydrogenaudio.com/ .


$updated from: CBR v VBR and Bitrates.htxt Sat 18 Jan 2014 13:14:24 thomasl (By Thomas Lauer)$