nierika wrote:"Variable BitRate encoding is designed for size & quality optimalization. Where there is silence in the music, it is less "demanding" in terms of its encodability, it makes sense to drop the bit rate, simply because there's not much there to encode, and the wasted space is overkill."
I would hope that people (especially on this board) would desire just as much sound quality in the "less demanding" parts of a track.
Also, considering how cheap 500GB+ HDs and DVDR discs are these days, I don't know why anyone would bother with any lossy compression when recording/archiving audio.
It can be kind of a hard thing to visualize but think about watching movies on digital cable or Divx / Xvid. Everything might look superb until you see something like fast running water on Digital Cable you can see artifacts of macro blocks. This is because the bitrate is not fast enough to accommodate the change in data. Most digital cable signals are probably processed on the fly or the media is mass processed for the format.
Professional DVD authors master DVD but adjusting bitrate based on the section of video being processed. That is why you don't see much artifacting in big budget movies from Hollywood, even in action movies.
Audio compression isn't much different except that its harder to quantify or less obvious to point out. You can freeze a frame of video and examine the compression. Its not so easy to freeze a moment of sound because without time sound doesn't really happen.
This isn't an direct representation of how the encoding of the bits and bytes work but just a method of simplifying how I conceptualize the workings of VBR vs CBR. Each block could represent a whole song or just a small section of a sound or music.
Each character represents a piece of data.
This top line is an abstract way of representing the complexity of a song or sound that is to be encoded.
* character represents null or nonexistent or unnecessary data
Code: Select all
Left 320 CBR, Right VBR
1123415318313136 1123415318313136 <- abstract way of representing the complexity of a song or sound.
**************** *********1****** @500 kbps
**************** *********1******
**************** *********1*****1
0000001001000001 ******1**1*****1 @320 kbps
0000101001000001 ****1*1**1**1**1
0001101101101011 ***11*11*11*1*11
0011101101101011 **111*11*11*1*11
1111111111111111 1111111111111111
0000000001111111 0000000001111111 Reference Positions
1234567890123456 1234567890123456 Read in a vertical manner, These are just for our reference.
With the CBR we are guaranteed to reproduce high quality audio that is guaranteed to be piped though at a constant rate of 320kbps. This is going to produce acceptably good quality sound and the data will be processed at a constant rate. The trade off is that we have to process the same amount of data regardless of the content of the audio.
With VBR the data is only processed if it is needed to reproduce the sound accurately. Other data is disregarded.
CBR gives good sound but you have a lot of unnecessary overhead when processing simple sounds or silence.
Looking to the complexity aspect...
Go to reference point 6 (RP6)
Complexity = 1
CBR = 00001
VBR = ****1
The sound is not complex and is represented accuately in both CBR and VBR.
CBR must process 4 additional units of data (0's) so that it meets the 320kbps bitrate.
VBR disregards the extra data and saves the space.
Go to reference point 10 (RP10)
Complexity = 8
CBR = 11111
VBR = 11111111
The sound is more complex here.
CBR 320 bks bit rate is insuffient to accurately reproduce RP10 to the same standards of RP6.
VBR increases the bitrate to 500kbps accommodate this higher level of complexity.
The amount of data for RP 10 would actually be larger then 320 CBR thus creating a better represention of the original sound.
If you add up all the units of data for CBR and VBR (disregarding the - data that VBR disregards)
CBR 80 Units
VBR 47 Units
VBR ends up producing (mostly) smaller files and better represents the original audio.
Note that at RP10 the complexity of audio of rated at 8.
CBR 320 can only represent 5 Units of data while CBR increases its data size to 8 to accommodate the complexity.
This would mean that 320bks CBR is sacrificing 3 units of complexity in its representation of the data which results in poorer representation of the original audio.
This is why highest quality VBR files will show bit rates that are well over 320 kbps. I have seen bitrates above 600kbps in VBR files. Just taking a quick glance here but seems most of my 16 bit Wave files have bitrates of 1411kbps for a point of reference.
Granted this is just an abstraction of what is going on but I think it makes sense. I have been reading about this for awhile and I think I am being won over by the VBR proponents.
To be honest its very hard to really tell the difference with a lot of tracks but I am starting to hear differences the more I listen. Could all be in my head though