[zotjen]

Quote:

Talking about true VBR and mp3 is a mistake. The VBR mode in mp3 has got nothing common with truly VBR algorythms implemented in mpc, ogg or aac.

So based on this, encoding using VBR for mp3 is not as efficient which results in what? Not as good sound quality? Bigger files?

 

[ILikeMusic]

Quote:

Originally Posted by zotjen /img/forum/go_quote.gif So based on this, encoding using VBR for mp3 is not as efficient which results in what? Not as good sound quality? Bigger files?

Well it could be either but the simple answer is to accept a slightly larger file size and then no quality difference. But really you don't even need to worry about it all all. Beyond the low-level details VBR is more efficient in terms of file size for a given quality level and that's what's important. Some formats are slightly more efficient than others but the difference is negligible at higher bitrates and there is no reason to avoid MP3 based on this factor, and MP3 has one big compensating advantage that no other format can offer, almost universal support. If the filesize is 5% larger than something else (and even that is an 'if') then who cares...

 

[LnxPrgr3]

Quote:

Originally Posted by zotjen /img/forum/go_quote.gif So based on this, encoding using VBR for mp3 is not as efficient which results in what? Not as good sound quality? Bigger files?

VBR actually uses the same bit reservoir CBR uses, so the situation is not as bad as it looks. If we needed 237kbps, we can either write 192kbps if we can borrow enough from the bit reservoir to get the size we really need, or write a 256kbps frame and add to the bit reservoir for the next frame. Either way gets us there.

But without that ability, the worst is that you waste some bits to get up to 320 when you really only needed 300kbps for that part. Wasteful, but not the end of the world, and certainly still an improvement over CBR, which can still vary in bitrate over time, but must average at whatever you set over a very short time frame.

 

[synaesthetic]

I can barely tell the difference between 192kbps and CD. If it's 192kbps or better, it's all good.

 

[majkel]

Quote:

Originally Posted by monolith /img/forum/go_quote.gif You should go have some scientific tests conducted. If you can truly hear everything you say, I imagine you'll be quite a curiosity to scientists. They aren't used to superhumans.

ABX isn't enough? I use foobar2000. Believe me, there is quite a lot of music that the 320kb/s bitrate makes difference. There is just lots of music and equipment especially, that does not allow you to hear the difference. Sorry for being the worse case. 90% of population does not distinguish the 128kb/s bitrate from the original. Let's say, we are here the 10% on this forum and several others audio related. Then you extract another 10% of those 10% who might hear the difference. Maybe it's me? All I can say is that staying with the -V0 forever is limiting yourself from hearing many new, interesting, endorsing things from your albums when the upgrade time comes. Do what you want but memory is cheap nowadays, there is no reason for avoiding lossless. I use ogg Vorbis Q10 encoded by Lancer SSE3 instead due to my player's limitations. If you ask me if I hear the difference - the foobar2000 ABX test says I don't. And I'm sure I don't hear the difference. Same case with the AAC Q0.95

 

[monolith]

Quote:

Originally Posted by majkel /img/forum/go_quote.gif ABX isn't enough? I use foobar2000. Believe me, there is quite a lot of music that the 320kb/s bitrate makes difference. There is just lots of music and equipment especially, that does not allow you to hear the difference. Sorry for being the worse case. 90% of population does not distinguish the 128kb/s bitrate from the original. Let's say, we are here the 10% on this forum and several others audio related. Then you extract another 10% of those 10% who might hear the difference. Maybe it's me? All I can say is that staying with the -V0 forever is limiting yourself from hearing many new, interesting, endorsing things from your albums when the upgrade time comes. Do what you want but memory is cheap nowadays, there is no reason for avoiding lossless. I use ogg Vorbis Q10 encoded by Lancer SSE3 instead due to my player's limitations. If you ask me if I hear the difference - the foobar2000 ABX test says I don't. And I'm sure I don't hear the difference. Same case with the AAC Q0.95

I do mostly use lossless. There are some albums for which there is a definite and obvious (i'm talking a child or hard-of-hearing person could hear it, not some audiophile magical subtle difference) difference between lossless and V0, but I remain very skeptical that you can hear a difference between V0 and 320.

 

[Stereodude]

Quote:

Originally Posted by monolith /img/forum/go_quote.gif I remain very skeptical that you can hear a difference between V0 and 320.

He might be able to. From all the reading I've done on hydrogen audio the LAME developers have put pretty much all their time into optimizing the VBR portion of LAME, not the CBR side. I wouldn't be surprised if the VBR 0 version was actually closer to the original than the CBR 320 version.

To his other claim. ABX'ing a mp3 to the original doesn't really mean squat unless you properly level match them. By default they are not level matched. MP3's will play back distorted unless you MP3gain them to prevent clipping. After you MP3gain them they are not as loud as the original. So, unless you compensate for the volume differences and inherent clipping on decode the ABX test is flawed.

 

[ILikeMusic]

Quote:

Originally Posted by monolith /img/forum/go_quote.gif There are some albums for which there is a definite and obvious (i'm talking a child or hard-of-hearing person could hear it, not some audiophile magical subtle difference) difference between lossless and V0,

Ah, excellent. Please let us know which albums those are so we can all experience the definite and obvious difference between lossless and V0 that a child or hard-of-hearing person can hear.

 

[Stereodude]

Quote:

Originally Posted by cowpat /img/forum/go_quote.gif does anyone else think there's just something a bit wrong about the idea of VBR? Having an unthinking computer decide whether something "deserves" to have a higher bitrate, and so constantly flicking between many different levels of quality? Does anyone take a stance against it, or has anyone noticed any sound quirks that would set it apart from a constant bitrate of a similiar quality? It seems to me that some realism could be sacrificed by the constant changes, and a lot of my music ends up at around 200kbps - does this mean that the difference between that v0 file and a 320kbp mp3 is negligible, or is the computer just taking a "that'll do" attitude. all a bit jumbled i know, but i just got my laptop back from the shop with all my files lost (sadface) so i have to re-upload all my CDs again (super sadface). I just wanted to see what you audio bods thought on the topic, and if this is just a psychological barrier on my part also, sorry if this is the wrong board

The general idea is that a VBR file will sound better than a CBR file of the same size. The largest portion of bits are allocated to the most difficult portion of the music to compress. This allows for the best usage of the bits and better sound for the same file size. The software uses the same model of human hearing to determine where the bits go to as it does when figuring how what is the least audible portion of the music to discard. If you accept that the software understands how to do the second half accepting that is knows how to do the first half shouldn't be a stretch.

The other nice thing about VBR is that it's tied to quality not filesize. You can do some ABX testing with Foobar to determine what setting is good enough. Maybe the default V2 is transparent to your ears with your headphones. Maybe it's V3, V1, or V0. You can decide based on what sounds best to your ears. Once you know this you know you will get the same relative quality every time from the compressor. With one one album the average bitrate might only come out to 180kbit/sec, but on another the average might come out to 225kbit/sec. In a CBR situation, unless you really crank the settings up some albums will be transparent, while other aren't.

Basically VBR lets you get more consistent transparency level while letting you fit more music on your portable player.

 

[LnxPrgr3]

Quote:

Originally Posted by ILikeMusic /img/forum/go_quote.gif Ah, excellent. Please let us know which albums those are so we can all experience the definite and obvious difference between lossless and V0 that a child or hard-of-hearing person can hear.

I can think of one, but not with the current LAME.

Album: Pink - I'm Not Dead
Track: 14 - I Have Seen The Rain
Encoder: LAME 3.96.1
Settings: -V0, with and without --vbr-new (--vbr-new is worse)

The difference is not exactly subtle. It clicks, loudly: check out the waveform!

Pink noise, -90dBFS, also freaked out this encoder, producing a 297kbps mp3 with the same artifact. 3.97 produced a 32kbps file with... -90dBFS pink noise

The point? LAME isn't perfect -- it does produce loud artifacts sometimes, even at settings that are generally good.

 

[ILikeMusic]

Quote:

The point? LAME isn't perfect --

Wasn't meaning to imply that it was, only that claims to be able to easily and casually resolve -V0 from 320 or lossless (meaning other than rare cases of outright encoder failure) should be taken with a great deal of skepticism.

Do you have any links where I might download that sample? I'm not sure really what to make of the waveform plot... would really like to hear it.

And would like to see a list of monolith's examples as well.

 

[majkel]

Very plain example - Papa Roach "Infest". Only the 320kb/s bitrate is able to fully reproduce the guitar's timbre.
Quote:

The general idea is that a VBR file will sound better than a CBR file of the same size.

ROTFL!

Less bits = higher quality.

Can you decrease the quality by using more bits? Think. 320kb/s LAME sounds better than -V0, sounds better than -V0 -q 0 as well. What I observed in the mp3's SQ is: 320kb/s FhG > 320kb/s LAME > 256kb/s FhG > 256kb/s LAME > VBR -V0 LAME. If you don't hear the difference, be happy with your VBR. Here are some clues how to recognize the following codecs and their flaws:
- 320kb/s FhG - very slight rolloff in the highs
- 320kb/s LAME (preset insane) - too sweet midrange, slightly instable soundstage
- 256kb/s LAME - slightly incomplete midrange, instable imaging
- 256kb/s FhG - slight treble rolloff, delicate midrange decoloration
- LAME -V0 - cold midrange, slightly blurry sounds
- wma 320kb/s CBR - slightly hardened bass, decolorated midrange
- wma VBR 98 - slightly U-shaped sound, nasal midrange coloration (applies to most wma actually)

As a general rule - the soundstage in LAME mp3's sucks in all modes besides VBR and -m s switch usage. Using 320kb/s in ABR mode saves 0-20kb/s per track causing audible treble rolloff, so makes no sense. Direct comparing 320kb/s LAME vs, FhG shows slight rolloff in the FhG encoder but there is more textural information in the FhG mp3's and more correct spatial imaging. LAME loses some details in comparison to FhG, and comparing both 320kb/s to the original, the FhG is closer. So, use WMP10/11, or invest in the dBpowerAmp Reference, it's not that expensive.

 

[Agnostic]

Quote:

does anyone else think there's just something a bit wrong about the idea of VBR? Having an unthinking computer decide whether something "deserves" to have a higher bitrate, and so constantly flicking between many different levels of quality?

Who says the computer is deciding? The lame encoder uses algorithms that have been developed and tweaked by humans for about 20 years now. The computer is just applying the algorithms.

 

[Febs]

Quote:

Originally Posted by majkel /img/forum/go_quote.gif ROTFL! Less bits = higher quality. Can you decrease the quality by using more bits? Think.

Do you contend that using more bits always equals an increase in sound quality?

 

[krmathis]

Quote:

Originally Posted by Stereodude /img/forum/go_quote.gif The general idea is that a VBR file will sound better than a CBR file of the same size.

Yeah, thats most probably correct.
Since a VBR file will have higher bitrate when needed (complex audio data), and lower bitrate when not needed (silence or non-complex audio data). Compared to CBR where the bitrate is constant all through the file.

Two examples below.

lame -V2:
Code:

[left]lame -V2 04.\ One.wav LAME 3.98 64bits (http://www.mp3dev.org/) Using polyphase lowpass filter, transition band: 18671 Hz - 19205 Hz Encoding 04. One.wav to 04. One.wav.mp3 Encoding as 44.1 kHz j-stereo MPEG-1 Layer III VBR(q=2) Frame | CPU time/estim | REAL time/estim | play/CPU | ETA 17131/17131 (100%)| 0:20/ 0:20| 0:21/ 0:21| 22.222x| 0:00 32 [ 107] %* 40 [ 0] 48 [ 0] 56 [ 0] 64 [ 1] * 80 [ 2] % 96 [ 0] 112 [ 3] * 128 [ 23] * 160 [ 1441] %%%%%*************** 192 [ 5041] %%%%%%%%%%%%%%%%%%%%%%%%******************************************* 224 [ 4509] %%%%%%%%%%%%%%%%%%%***************************************** 256 [ 3582] %%%%%%%%%%%%%%********************************** 320 [ 2422] %%%%%%%************************** ------------------------------------------------------------------------------- kbps LR MS % long switch short % 228.1 30.0 70.0 89.3 5.9 4.8 Writing LAME Tag...done ReplayGain: -7.8dB[/left]

lame --preset cbr 224:
Code:

[left]lame --preset cbr 224 04.\ One.wav LAME 3.98 64bits (http://www.mp3dev.org/) Using polyphase lowpass filter, transition band: 19383 Hz - 19916 Hz Encoding 04. One.wav to 04. One.wav.mp3 Encoding as 44.1 kHz j-stereo MPEG-1 Layer III (6.3x) 224 kbps qval=3 Frame | CPU time/estim | REAL time/estim | play/CPU | ETA 17131/17131 (100%)| 0:23/ 0:23| 0:24/ 0:24| 19.198x| 0:00 ------------------------------------------------------------------------------- kbps LR MS % long switch short % 224.0 45.3 54.7 92.8 4.3 2.9 Writing LAME Tag...done ReplayGain: -7.6dB[/left]

Notice how the "V2" compressed file have quite a bit of data compressed with 256kbps and 320kbps. While the "CBR 224" file are 224kbps through the entire file.
They are almost identical in bitrate 228.1 vs. 224kbps (and hence file size. 12.2 vs. 12MB), but I expect the first one to be more transparent.

Arrest me if I am wrong.