What's everybody's thing with VBR here?

Originally Posted by Iceboy

I used to use APX all the time, but now I tend to use API. I just don't see how any frame of an MP3 can suffice with anything less than 320k and still make a perfect sound. And who decides what bitrate is needed during VBR encoding? Is it programmed or all automated? Surely you need 320k for every frame to ensure perfect sound?

I did my own encoding tests using the same source Wav and encoding to APX and API. The API yeilded a slightly higher frequency response.

I also have two huge HDDs so space isnt a problem for me.

What made you think that the 'perfect' value is 320 kbps? Technically speaking, MP3s can never be perfect since data is removed from the source regardless of the compression level used. WAV files taken from CDs are something like 1400 kbps.

I think when you say 'perfect', what you actually mean is 'transparent'--meaning no audible difference between the input source and output MP3. The point of VBR is realizing that since the format can't be perfect due to it being lossy, a set of algorithms to determine the complexity of each frame and allocating a number of bits to accomodate it reduces the filesize while still resulting in an output file that is transparent to most everyone.

So, to answer your questions: first, it is both programmed and automated since the algorithms were programmed and they automatically determine the bitrate for each frame. VBR is an attempt at compromise between reducing filesize (which, of course, is the whole point of the MP3 format) while sacrificing only audio data that does not affect the transparency of the output file.

I would be interested to learn the procedure of your encoding tests, since it has been shown via double blind tests on high-end equipment that the difference between high-end LAME encodes is pretty much undetectable. When you say 'a slightly higher frequency response', that sounds to me like you used some software application to determine the frequency range of the file. I'm sure you know this already, but the human ear is not suited to detect frequencies at the high and low extremes while software can.

Have you tried doing your own double blind test with the two files? In case you didn't know, a double blind test in this case is when you listen to two files and eliminate any placebo effect by not knowing which is which. Foobar2000 has this functionality built in, so you might want to try that. All you do is pick the two tracks in your playlist, right click them and go to Utils, then 'ABX Two Tracks' and follow the instructions. Basically it gives you the option to play A, B, X, and Y. A and B are the same as X and Y, and you have to match them up based on how they sound.

Finally, space isn't the only issue of course--there's also bandwidth since we're transferring these files to one another. And even though I have a terabyte of storage space between my 6 hard drives, space is always an issue.

Jam it back in, in the dark.