LAME: -V 2 vs. -V 0

VBR V0 maxes out what MP3 can give you. Also V0 is a very tuned compression option, so I use V0 whenever I have to encode something with LAME.
However most of the time I choose FLAC for lossless compression and Ogg Vorbis or AAC for lossy compression. Vorbis and AAC are just superior, because they're based on newer and better coding techniques.

The thing I don't quite understand is why so many people keep encoding with CBR 320. Doesn't make any sense to me when you get the same with fewer bits using VBR V0.

Jam it back in, in the dark.

Originally Posted by Rimo

-V 0 sure has more potential to max out the result, yet 320 CBR would be even better in that case as it's the absolute best possible result for MP3.

No, you're wrong here.
"Best" for lossy encodings is determined by perceived audio quality divided by bitstream size.

Letting the encoding engine work with constant bitrate disables all fine-tuned smart algorithms that are used to allocate bits. You're wasting bits in the stream filled with zero information.

The audio data has certain "flaws" when coming out at the end of the transform coding step. The "quantize" process now allocates needed bits for the results from the transform step. This is done the smart way when VBR is enabled, brute-force when constant bitrate is used.

That's like a factory producing objects of different size but only using only package format to ship the objects and filling the rest with padding material. That's not very efficient.

MP3 is not: The more bits you throw at it, the better it gets. Most people think so, but MP3 is already limited by design. And with frame bitrates produced by V0 encodings you're pushing the technology behind MP3 to it's limits.

I think most people here don't know. But MP3 isn't limited to 320kbit/s CBR. You can e.g. tell LAME to produced freeformat bitstreams. There you can push bitrate up to 640kbit/s.
The problem is that the perceived audio quality won't increase. Again because the certain flaws I was speaking of won't go away just by throwing more and more bits at them.
There are some special test signals that are encoded better when the bitrate is that high, but nothing that appears in regular music.

Problem with freeformat streams is that the MP3 standard doesn't say that hardware devices have to be able to play it. The existance is covered by the standard but you can call a device MP3-capable even if freeformat streams are not supported. A reason why they are so rare.
However the very accurate libMAD decoding engine can playback these streams.

Originally Posted by Rimo

Going with the logic that -V 0 would be a better choice since you get the same audible quality while reducing the filesize, why wouldn't -V 2 be an even better choice if the huge majority of people couldn't hear any differences compared to -V 0?

I don't want to answer this question in that form.
You see, I rip all of my discs in FLAC. In case I want to have something on my portable player I can always re-encode the file to a lossy encoding. As the hardware decoder of portables isn't very accurate, the DAC often is crappy and the standard headphones don't reproduce the sound very well - I can even go below V2, e.g. V4 or even lower. I probably won't notice the degraded audio quality at all.
Encoding quality is just good enough to drive this low-end playback chain.

On the other hand when at home and listening to music through my "good" equipment (AKG k701 dyn. headphones, DIY headphone amp and DIY USB-DAC) I'm not that limited and certain flaws (like ringing artifacts when audience is applauding) are now detectable.

Furthermore I get tired when listening to highly compressed (encoded, not the compression as in loudness war - I also get tired of this one) audio. I can listen much longer when playing from the original disc (or a lossless encoding), also it's more relaxing for me.
There is a lot that's destroyed when doing lossy encodings. Stereo imaging, dynamic range, all sorts of artifacts.

I'm not saying that I can always distinguish between a lossy and a lossless encoding. But there are differences, which are annoying when at perceivable level.

Originally Posted by Rimo

For the moment, I see the 320 vs. -V 0 usage to be quite similar to -V 0 vs. -V 2 and I'm under the impression many people chose an overkill setting just to feel more secure while not actually noticing any differences.

See above. I consider CBR deprecated. It should be removed. The only thing why it's still there: old hardware players that have problems with variable bitrates.

Originally Posted by Rimo

I'd go all the way with Vorbis if it wasn't for the fact that it is still less widely supported than MP3.

The fascinating thing is that it's quite easy to implement it on embedded hardware (there is the integer-only imlementation vorbis-tremor, which should work on most ARM processors).
I think it's more a political thing...

There's nowhere I can't reach.

I should add that there is some tool around (python based I think, search hydrogen audio forums for it) that can transcode VBR to CBR and vice versa.

VBR to CBR is easy, just use the biggest package format ever used in the bitstream (see the factory example above). The other way is a bit more complicated, I think the author has some information about it in the thread. Sourcecode is also open AFAIK.

This thing is sticky, and I don't like it. I don't appreciate it.

Originally Posted by Rimo

I agree with this definition of "best". It is also a fact that MP3 is not the best codec around. Yet, is -V 0 really achieving the "best" result for this codec or wouldn't it be a little overkill like 320 CBR is to a greater extent, actually going over the limits of MP3 without any significant improvement?

If you want to increase efficiency you have to lift the freeformat limitation. Means the standard had to demand freeformat compatibility for all playback devices. That's not going to happen.
This way you're limited to a maximum frame bitrate of 320kbit/s when VBR V0-encoding, IF that amount of bits is really needed to encode the informaton.

You see, VBR V0 is the end of MP3. You can't get more quality without rewriting the standard. At that's not going to happen.

I am a dolphin, do you want me on your body?

Changelog for the interested user:
LAME Changelog

I was speaking idiomatically.

@Moguta: CBR 320 is wasting bits when compared to VBR V0. That's why it's inferior.

What kind of toxic man-thing is happening now?

@Moguta: My definition of "wasting bits" isn't the same as yours. You define wasted bits by "not so much" improvement on the perceived audio.

That's not my "wasted bits" definition, which is much more concrete: unused bits in the MPEG bitstream, unused because the framesize is too big for the bits delivered from the transform coding step.

Additional Spam:

Originally Posted by LiquidAcid

@Moguta: My definition of "wasting bits" isn't the same as yours. You define wasted bits by "not so much" improvement on the perceived audio.

That's not my "wasted bits" definition, which is much more concrete: unused bits in the MPEG bitstream, unused because the framesize is too big for the bits delivered from the transform coding step.

EDIT: Oh yes, also it wasn't me stating that I could hear artifacts in V2 which wasn't there with V0.

FELIPE NO

Nearly every lossy codec uses transform coding, which is totally different from (linear) prediction coding used by lossless codecs like FLAC and WAVPack.

How ya doing, buddy?

For those interested in the (mathematical) details of audio compression:
Wikipedia has a large number of (quite good) articles about different subjects:
Discrete cosine transform - Wikipedia, the free encyclopedia
Modified discrete cosine transform - Wikipedia, the free encyclopedia
Linear prediction - Wikipedia, the free encyclopedia

Simple linear prediction is very easy math, but for transform coding you need a lot of advanced knowledge about analysis/calculus.

Jam it back in, in the dark.

Yep, that's the basic concept. Lossy audio codecs do a "transform" (fourier, fast fourier, discrete cosine, wavelet, etc.) before doing their "real" works, which consists of deciding (smart) which information can be dropped without perceived loss of quality (quantize).
These transforms also have nice properties for derivation when looking at the fourier transformation, which makes them quite interesting. It took me however nearly four semesters (analysis I+II for the basics, complex analysis and analysis III) to fully understand the concept (and be able to proof why it works).
I guess it's a bit easier with prediction coding, but I didn't look into that very deep. However fitting polynoms to arbitrary functions can be tricky as well :-)

You see that comparing the two compression methods is not really possible. Lossy is happening in frequency domain, lossless in time domain.
What would be interesting: to see someone combine prediction and transform coding (again resulting in a lossless bitstream). Anyone aware of such a codec?

There's nowhere I can't reach.