Need help in a technical "320cbr vs. v0 MP3"-discussion

[rhymesgalore]

No, this is no thread about what actually sounds better (well, still kind of), but more about the inner workings of cbr and vbr encoding.

I'm debating with a user on another forum, if 320cbr is the best setting for a MP3. I - and the wiki on hydrogenaudio - think that it is, since it is the maximum sample rate that one is able to achive with MP3. He says, that v0 is better than 320cbr.

His argumentation about the working of mp3 goes like this: If you work with 320cbr you use 160kbps to save one channel, and the other 160kbps to save the difference between the other channel. But it will never be more than 160kbps per channel. So he says with VBR this restriction doesn't exist. So if the channels are pretty much the same it uses more than 160kbps for one channel, and the rest to encode the difference between the two.

In my understanding he's doing some mixup between vbr and joint stereo...

My question now is, how does the encoding in cbr or vbr really work? As far as i understand 320cbr would give you 160 per channel on every sample. Is vbr really able to give one channel more than 160kbps? And wouldn't joint stereo affect the two methods in the same way?

[chinesekiwi]

the audible difference between 320 kbps mp3 and VBR -v0 mp3 are none except that VBR uses less room.

While 320 kbps mp3 gives 160 kbps each channel all the time, VBR only gives the bitrate needed for that channel at that time.
Audible different = zero. Both ar eone and the same but VBR's better as it uses less filespace.

I'd, if you can, encode in Ogg Vorbis tbh as it's a much much better true VBR algorithim and sounds better at lower bitrates. What I mean is Vorbis -q6 (@~192 kbps) sound sbetter than 320kbps mp3.

[vegaman]

Originally Posted by chinesekiwi

the audible difference between 320 kbps mp3 and VBR -v0 mp3 are none except that VBR uses less room.

While 320 kbps mp3 gives 160 kbps each channel all the time, VBR only gives the bitrate needed for that channel at that time.
Audible different = zero. Both ar eone and the same but VBR's better as it uses less filespace.

I'd, if you can, encode in Ogg Vorbis tbh as it's a much much better true VBR algorithim and sounds better at lower bitrates. What I mean is Vorbis -q6 (@~192 kbps) sound sbetter than 320kbps mp3.

You forget to mention that this is all in your experience and opinion.

I do prefer v0, for filespace savings and I personally haven't ABX'd it from a 320cbr version yet (though I personally use FLAC, for archival purposes). But I would not try to suggest that it is of exactly the same quality as 320kbps cbr. v0 averages only around 225kbps (can't remember the exact figure), i.e. it cannot store the same amount of detail.

[chinesekiwi]

Originally Posted by vegaman

You forget to mention that this is all in your experience and opinion.

I do prefer v0, for filespace savings and I personally haven't ABX'd it from a 320cbr version yet (though I personally use FLAC, for archival purposes). But I would not try to suggest that it is of exactly the same quality as 320kbps cbr. v0 averages only around 225kbps (can't remember the exact figure), i.e. it cannot store the same amount of detail.

See a couple of threads down 'LAME vs. Vorbis @ 210-250 kbps VBR' for opinions aobut Vorbis and how it's better than LAME mp3.
Anyway, it's ~245 kbps and is exactly the same as 320. I've encoded -v0 mp3 with 95% of time, the bitrate is 250 kbps+ (Public Enemy's 'Welcome To The Terrordome').

The '~' rating doesn't mean too much with VBR mp3 while in Vorbis, it's more exact.

[rhymesgalore]

Thanks for the answers so far. But this is not strictly about the audible difference, or what format to use. This is really more about the technical differences when encoding in 320cbr or v0.

Sadly i haven't found any documentation on that...

[Publius]

Don't waste your time asking here. Ask on HA.

[rhymesgalore]

Originally Posted by Publius

Don't waste your time asking here. Ask on HA.

My first idea as well . But i wasn't registered there before, and now i have to wait 5 days before i can post, since I registerd with a free email account.

So i thought in the meanwhile i could as well try it here...

[psyllium]

With the new versions of LAME, a 320kbps file will actually use as much of that 320kbps as it can. There will be silent and non-complicated parts where it obviously won't need the whole 320kbps, which is where VBR comes in to play. Another thing that LAME will do is make full use of the bit reservoir.

Conceptually, it works like this:

A B C D*

Where D is *really* complex, it will use frames A B and C before it to store that 'extra' information for part D. This is how you can get one frame of an MP3 using up MORE than 320kbps. If you are encoding at CBR 320, it will have more space available for those extra complicated bits. The VBR encodings will probably take advantage of the bit reservoir as well, but, as there is less room to move, you might not get as much quality.

But yeah, ask at HA

[E.B.M.Head]

Well you should keep in mind that LAME normaly uses "Joint-Stereo", unless you force to use a fixed stereo profile (Left/Right Coding, Mid/Side Coding or Intensity Stereo) - but that would be in most cases not very smart. "Joint-Stereo" decides automaticaly which encoding profile to use for a frame. In most cases it will use Mid/Side coding as left and right channel are often very identical in most music. This way you have the main "mono" channel and stereo difference in the second channel. As the stereo differences usually don't need much encoding space, you have more left for the main channel: e.g. 64 kbps for stereo differences and 256 kbps for the main channel. If the music is very different in left and right channel, then LAME uses the left/right encoding and in this case you are limited to 160 kbps left and 160 kbps right.
Intensity stereo is only used at very low bitrate CBR encoding. This downmixes the the channels to one mono channel and the stereo differences for the midrange and high frequency range. The encoding is not phase correct.
Intesity Stereo is in MP1 and Mp2 also called "Joint Stereo" and therefore lots of Audio Noobs think that the LAME "Joint Stereo" is inferior to Left/Right Stereo. In fact it's just a bad name choosen and "Joint Stereo" is the best you could do to your music if you encode it to MP3.
If you think that 320 kbps ain't enough, there is also the Free Format switch in LAMe that gives you up to 640 kbps in CBR mode. But beware: Lots of Players refuse to play anything above 320 kbps (MAD.lib based players are one exception) and you could also reach bitrates with efficent lossless encoders around 600 kbps like Monkey Audio (APE), OptimFrog, True Audio (TTA), TAK or WavPack. For WavPack, Monkey Audio and TTA there are even hardware players like the Cowon D2 or Cowon A3. FLAC and Apple Lossless usually compress worse (around 800 kbps), but are less demanding on the hardware.

[LnxPrgr3]

Originally Posted by rhymesgalore

His argumentation about the working of mp3 goes like this: If you work with 320cbr you use 160kbps to save one channel, and the other 160kbps to save the difference between the other channel. But it will never be more than 160kbps per channel.

With CBR, the restriction doesn't exist either. The encoder is free to, say, allocate all 320kbps to the left channel (suppose the right channel is silent), or use mid/side coding and allocate everything to the mid channel (if the sound is completely monaural).

Well, unless you use dual-channel mode, but then you'd be insane