Talk:MP3/Archive 1

Archive 1

Archive 2

Archive 3

→

Archive 5

NPOV complaints

Here are several years'-worth of threads debating how the article addresses the topic of MP3 audio quality. — archive ed.

Claims about MP3 quality

This article reads like it was written by an audiophile. Choice excerpts:

• To many other listeners, 128 kbit/s is unacceptably low quality, which is unfortunate since many commonly-available encoders set this as their default bitrate.
• It is important to know that despite of all the flaws, recent multiformat listening tests (http://www.rjamorim.com/test/multiformat128/results.html) once again show that LAME MP3 easily rivals its technological successor AAC. (Vorbis aoTuV is tied with Musepack at first place, Lame MP3 is tied with iTunes AAC at second place, WMA Standard is in third place and Atrac3 gets last place). (bold original)

Most of the stuff under the encoder comparison is also POVish. The Alternatives is similar. Random speculation has worked its way in. Finally, the Online Music Resources is marginal. - Fennec (はさばくのきつね) 03:34, 10 Jul 2004 (UTC)

What part of that "despite of all the flaws" did you find 'audiophile'-ish? You can't disregard the fact that inherent to this format (MP3) there are quality-issues involved. Everybody wants (and needs) to know that, in order to understand what MP3 is about. So, to then state under some Quality-section that MP3 is some kind of ugly sounding bad quality trashy format, seems very unfair to me. Check these if you are not convinced; http://www.heise.de/ct/00/06/092/ (I believe there is a translated version of it somewhere) http://jthz.com/mp3/#MYTH So, that is why I posted the bold part; use the right encoder, with the right config, and you'll have flawless quality MP3 encoding. It's been proven. 195.64.95.116 01:30, 5 Sep 2004 (UTC)

I agree, this does read like it was written by an audiophile. Feel free to edit if you don't like it.

MP3: Now with NPOV

The old article was full of awful propaganda, so I revamped it and moved some things into the LAME article. D. G. 10:44, 4 Sep 2004 (UTC)

I kinda disagree with you on that. As with Winamp, also LAME is very much to blame for the grand success of the format. Let's not forget that this was one of the first and highest quality FREE MP3 encoders out there (next to Blade), whilst others needed payment or licensing before being used. So, when speaking of MP3, one needs to speak of LAME. The one would never be this popular without the other. I would hardly call that propaganda, or POV, it's simple fact. 195.64.95.116 01:21, 5 Sep 2004 (UTC)

I'm not sure what you're proposing is a "proven scientific fact." There's no such thing as a proven scientific fact. Anyway, anything is "up for discussion." You simply can't make an edit and declare that it is unquestionable. I believe you are saying that it's a "scientific fact" that nobody can detect the difference between 256kbps MP3 and an uncompressed source. That is untrue, although almost nobody can tell the difference, some golden ears listeners can on certain samples. If you're going to insert these statements you should back them up. You cannot just say "numerous listening tests."

I deleted the paragraph from the design limitations section, because this is already discussed in the MP3 quality section. Please clarify it there if you want. Rhobite 23:01, Oct 3, 2004 (UTC)

It is proven scientific fact that 1 added to 1 makes 2. If you can't get that, you're not worth my time, and I'm not going to discuss proven facts concerning our hearing or MP3 quality, I have better things to do. If you want to state silly 'claims' on MP3 quality which have no base whatsoever other than gossipy (trying to sound like an expert) speak, I'm going against that. I would refer to rec.audio.pro and being a member of the Audio Engineering Society. The point where experts, the high-end listeners and the likes, will not be able to distinguish the MP3 file from the original currently lies around 180 kbps VBR, and 224 kbps CBR mp3 files. This is where 50% of them will say that the mp3 is the original and/or vice versa, i.e. where they can't tell the difference. (This is researched on using LAME.)195.64.95.116 18:53, 4 Oct 2004 (UTC)

Can you link to a study, other than that German study from 4 years ago? Can you link to any specific posts on rec.audio.pro?

You can all test it yourselves, this is quite easy to do. If you can't understand that, you don't belong in this discussion anyway. It's like discussing existence of gravity, or the magnetic polar fields. They are there, and you can't babble on about it the way you want to. Furthermore, the articles in rec.audio.pro or elsewhere would not be read or understood by you anyway. 195.64.95.116 16:04, 31 Dec 2004 (UTC)

The burden is on you. If you refuse to discuss your edits or link to any studies, you are not "worth my time" either. Please do not curse in your edit summaries, and do not personally attack me.

I do as I see fit thank you very much. You must have been deserving of me personally attacking you. 195.64.95.116 16:04, 31 Dec 2004 (UTC)

MP3 and CDDA are not directly comparable. I will not defer to your self-proclaimed "authority."

Of course they are; they are both the end-medium formats people listen to.195.64.95.116 16:04, 31 Dec 2004 (UTC)

Also, if you insist on reverting, please don't reintroduce your own grammar and spelling errors. Rhobite 19:13, Oct 4, 2004 (UTC)

If anything I've corrected yours.195.64.95.116 15:54, 31 Dec 2004 (UTC)

It may help here to consider a fact as something that is not known to be disputed anywhere today by otherwise reasonable people. Since there is obviously some dispute here about this "fact", our WP:NPOV policy arises to meet the occasion. Perhaps it is time to "characterize the dispute" if necessary, or merely to to back off on the fact with something like, "many people even claim they cannot detect a difference between P and Q." I will watch this page for a while. Be sure you are well familiar with the contents of the WP:NPOV article. And as always, remember wikilove and have a nice day! Tom - Talk 16:35, 6 Oct 2004 (UTC)

any criticisms toward the technology would be personal points of view made by the users of mp3's.

This makes no sense. Quality of audio-reproduction can be measured. If this wasn't the case, something like MP3 wouldn't exist, nor would it sound as good as it does these days.195.64.95.116 15:54, 31 Dec 2004 (UTC)

as they are based on the opinions of persons they should not be included in the article. instead try adding links to reveiws made by some sort of professional organization and let reader form their own opinions instead of trying to guide their opinions with your own through the artical--Larsie 21:40, 19 Oct 2004 (UTC)

Value judgments

Wikipedia articles don't make value judgments or recommendations, such as recommending that non-professionals never have a reason to use lossless compression. In any case, this statement is not true: "Those who will only listen, do not need to use lossless compression, since they won't hear the difference with MP3." You can't make blanket statements like that, some people can indeed tell MP3 - even with a good compressor - from audio that has never been compressed.

They can up to a measured degree. This can be proven and it often has been. Beyond certain high enough bitrates NO HUMAN will be capable of telling the difference.195.64.95.116 16:16, 31 Dec 2004 (UTC)

Etree and archive.org distribute lossless copies of nearly every show - there is obviously a large group of listeners who feels MP3 is inadequate for their uses. Rhobite 23:06, Nov 15, 2004 (UTC)

The fact that MP3 is now often considered a degraded format, surpassed by other formats, has nothing to do with that. It's simply because

1) there are bad mp3 encoders around (lots of them in fact)

2) MP3 tends to need some type of special treatment beforehand to reach optimum quality, other f

Yes, it's true that people under stress, with bad music equipment or in a noisy or unsuitable environment can have trouble differentiating between mp3 and lossless, or don't care, but that doesn't mean that they still can't hear it. Lossy formats should not be recommended with a clear conscience. BKmetic 00:00, 23 July 2006 (UTC)

This may be completely off kilter but to me it would seem simpler to actually look at the 5 difference in the signal given the identical encoder and differing bit rates. Any time someone uses "excellent" "acceptable", "poor" and "good" are all opinions (as oposed to factual). I personally have tried to hear the difference between the 128 bit and the CD on a good systems and I can hear the difference (after carefully selecting the music and carefully listening). In my car I have a great stereo, that would not be the case there since the road noise etc interferes just enough that I am unable to discern the difference. I will not make a value judgement as to what is acceptable, good or excellent since my opinons would only hold true to me at this instance. If one could measue it objectively it would be much more helpful. Look at it as follows Break down the audio into 5 frequency bands. Weight each band as it impacts our hearing (the center 3 bands are more important for human hearing than those below say 400hz and above 6kHz <--- Example) Using a digital oscilloscope, measure the difference in peaks in each band, then try to determine the loss of the resonant frequencies as opposed to the main frequencies since they tend to give music that "depth" the audiophiles like (yes, that is a qualitative opinion, not measurable but I am trying). Doing so would create a simple, objective comparison of formats and you could also use the data to compare encoders and thus determine the quality comparatively (objectively) in lieu of throwing subjective valuations around. [Moto] --71.112.37.172 19:28, 1 August 2006 (UTC)

Your simple view of what it takes to make objective quality measurements on audio signals is uninformed by the years of engineering effort in this direction. It is a hard problem, partly because hearing is so complicated, and quality measurements need to correlate to what you hear. Dicklyon 19:53, 1 August 2006 (UTC)

NPOV in Quality section

The entire MP3 quality section reeks of NPOV and unverified information, especially statements like this "However, listening tests show that with a bit of practice many listeners can reliably distinguish 128 kbit/s MP3s from CD originals [...] reaching the point where they consider the MP3 audio to be of unacceptably low quality." Needs a rewrite for sure (and not by some audiophile with no sources). 70.45.49.36 04:06, 11 July 2006 (UTC)

I find the main text perfectly reasonable, and these complaints hard to follow and inconsistent. MP3 is an audio format. I have no clue if the author considers him/herself to be an "audiophile", but certainly I see no shame in being an audiophile, or any problem in the idea that an article about an audio format should be written by an audiophile. Who better ? someone who is not interested in sound and isnt able to distinguish high quality and low quality reproduction ? I personally DO consider myself to be an audiophile, yet am almost ashamed to say that I sometimes struggle to distinguish 128 bit MP3 from a CD original. So its good. But its not perfect and under ideal conditions I can easily tell, and for thius reason I consider 128 bit unacceptable, and always rip at max quality 320 bit VBR. Which I think fits perfectly with the authors text. Whats the problem guys ? / Dave smith

The problem is no sources. Even an audiophile would be OK as a source, if published and referencable, but an editor's own audio opinions are not encyclopedic. Dicklyon 18:57, 25 July 2006 (UTC)

As above. It doesn't matter how plausible the text is or how well it jives with your subjective experience. The article makes several unsourced claims about relative "quality" (which is not defined in specific terms) of MP3s of various bitrates, and it makes several unsourced claims that people who make claims about audio quality have been proven to be unreliable. It's not unreasonable to demand sources be cited or the section be rewritten or deleted.

Not related to the article, but responding to your comments, you should be aware that using VBR results in the lowest possible bitrate being chosen for each frame to maintain (not drop below) a certain level of quality (and in LAME at least, that quality level is verified by trial and error). If you want max quality per frame, you'd be using 320 kbps CBR, so that the encoder is under no pressure to drop the bitrate on frames where it thinks it can get away with it. However, 320 kbps CBR is unreliable in some encoders (even LAME), introducing a kind of ringing artifact in the high end, depending on what other settings you use (highpass filter, psy model) and the content of the audio. —mjb 19:11, 25 July 2006 (UTC)

I'd just like to say that i think an audiophile is the perfect candidate for writing on this sort of topic if it had good sources. Now i know ive come across articles on the internet (not blogs) that have tested mp3 satisfaction on subjects, it just a matter of finding them. ive read the section and it all seems to match up with what i rememeber so i think leaving the [citation needed] will do for now. also, im goin to activly look for said articles whenever i can now. also, i propose that the NPOV tag get removed and replaced with a {{Unreferenced}}. -(chubbstar) ^{— talk | contrib | 15:42, 26 July 2006 (UTC)}

Quote:

• MP3 files encoded with a lower bit rate will generally play back at a lower quality. With too low a bit rate, "compression artifacts" (i.e., sounds that were not present in the original recording) may appear in the reproduction. A good demonstration of compression artifacts is provided by the sound of applause: it is hard to compress because of its randomness and sharp attacks. Therefore compression artifacts are audible as ringing or pre-echo.

Isn't the mp3 compression scheme 'adaptive' in some sense that it only throws out perceptively irrelevant data-subspaces? It seems like ringing/pre-echo artifacts would be associated with other forms of stuff, not medium-quality mp3... is there a source to verify this? Ninjagecko 04:24, 7 August 2006 (UTC)

It does try to make sure the errors due to coding are perceptually small, or masked. But since the filters are essentially time-symmetric, it puts as much error before a transient, where it's likely to be audible, as after, where it's more likely to be masked. That is the scheme is not perfect, and some of its imperfections are these "precursor" sounds. I don't have a reference about who finds these to be audible at what bit rate. Dicklyon 04:41, 7 August 2006 (UTC)

I rewrote the quality section and provided references. I thus removed the NPOV and references tag of the article, as my opinion is that this section is now based and supported by facts.--Gabriel Bouvigne 15:06, 14 September 2006 (UTC)

I must point out that MPEG's own verification tests show that MP3 is substantially worse in encoded quality at a given bit rate than MPEG-2 AAC LC profile. Woodinville (talk) 23:11, 15 January 2008 (UTC)

Remove NPOV and replace with Unreferenced template?

i propose that the NPOV tag get removed and replaced with a {{Unreferenced}}. I would do it myself but ill be there might be disagreement which i think should be discussed. The only bit i think to be potentially not NPOV is the "Layer 1: excellent at 384 kbit/s, Layer 2: excellent at.. etcetc." part. Anyhoo. -(chubbstar) ^{— talk | contrib | 15:57, 26 July 2006 (UTC)}

Comments on encoders

I absolutely agree with this part :

>>Good encoders produce acceptable quality at 128 to 160 Kbit/s and near-
>>transparency at 160 to 192 kbit/s, while low quality encoders may never reach
>>transparency, not even at 320 kbit/s. It is therefore misleading to speak of
>>128 kbit/s or 192 kbit/s quality, except in the context of a particular
>>encoder or of the best available encoders. A 128 kbit/s MP3 produced by a good
>>encoder might sound better than a 192 kbit/s MP3 file produced by a bad
>>encoder. Moreover, even with the same encoder and resulting file size, a
>>constant bitrate MP3 may sound much worse than variable bitrate MP3

—Preceding unsigned comment added by Eclipsed aurora (talk • contribs) 30 July 2006

Would like to add the following comment to MP3 format performance for low bit rate. I use those bit rates to encode speech, and this comment comes from personal experience:

>>  Low MP3 bit rates (64 bits mono and under) while adequate for speech 
>>encoding, actually offer less compression than a straight mono .WAV 
>>file at the same rate.  The difference can be as much as 2 to 1.
>>
>>  On the other hand, .WAV files at such low rates usually are not 
>>supported in hardware-only players (non-computers).  Also, the ability 
>>of using ID tags is lost as well.
>>

24.181.71.243 16:23, 31 October 2006 (UTC) Alex

No Advantage?

I removed:

• the fact that these alternatives do not generally provide a clear advantage over MP3

Ogg Vorbis claims "For a given file size, Vorbis sounds better than MP3." [1]. If you put that sentence back, temper it (e.g. "the advantages are disputed") and provide citations in favor of and in opposition to Vorbis. Superm401 - Talk 09:33, 9 December 2006 (UTC)

I don't think the advantages are seriously disputed, but would not be significant enough to most users until they gave a file size reduction of more than a few tens of percent. – Smyth\^talk 10:49, 9 December 2006 (UTC)

Bitrate/quality

I'm confused. Several articles on MP3/AAC throughout Wikipedia contradict each other. Here it says that modern MP3 encoders can achieve very good quality at 128 kbps but in 1998 it was equivalent to AAC at 96 kbps and MP2 at 192 kbps. The AAC-HE article states that the quality is excellent at 48 kbps but 64 kbps is lower quality than 128 kbps. So... if 48 kbps is "excellent" then 96 is probably perfect. Therefore, 128 kbps in 1998 was perfect quality??? Maybe the author meant 128 kbps in 1998 was equivalent to 96 kbps MP3 today? That would make much more sense. Also, the average reader has no idea what "192 kbps MP2" would sound like, so best to keep things simple and only compare MP3 quality with MP3 quality from now on instead of introducing numerous different formats? —Preceding unsigned comment added by 70.65.192.118 (talk) 00:05, 7 October 2008 (UTC)

This article is simply wrong. The editors have been drawing incorrect conclusions from the sources they are using. 128 kbps is not excellent, even if it is really VBR averaging 128 kbps, rather than true 128 kbps. --JHP (talk) 20:58, 30 November 2008 (UTC)

128 kbps is not transparent!

The article currently states, "The transparency threshold of MP3 can be estimated to be at about 128 kbit/s with good encoders on typical music as evidenced by its strong performance in the above test." The test being referred to is here. However, as I look through the test results, nowhere do I see the 128kbps compressed audio files being compared to uncompressed WAV files. Transparency does not mean that a compressed audio format is indistinguishable from other compressed formats. It means a compressed audio format is indistinguishable from uncompressed audio formats. Where is the control? The test doesn't demonstrate transparency without uncompressed audio to serve as a control. The last test I saw that compared MP3 side-by-side with uncompressed WAVs found that transparency was achieved at 224 kbps. --JHP (talk) 20:15, 29 November 2008 (UTC)

Let me also point out that in the last graph at the bottom of the page, 128 kbps VBR only scores 4.6 out of 5.0. If we were to assume that 5.0 represents transparency (I don't see a detailed explanation of the test procedures anywhere), then the fact that 128 kbps VBR scores below 5.0 would be proof of non-transparency, rather than proof of transparency. --JHP (talk) 21:06, 29 November 2008 (UTC)

Here's an interesting article on audio quality. It's a shame they didn't test with the LAME codec, though. --JHP (talk) 03:27, 4 December 2008 (UTC)

JHP, if 5/5 is transparent (ergo, perfect quality) then logically 4 and above would be excellent. A 3/5 indicates "adequate" quality as the artifacts are slightly annoying and noticeable without a double-blind test. Anything 4 and above would not be noticed without said test, and a 4.6/5 would not be audible unless the listener concentrated hard on the artifacts. You have a warped opinion of "excellent."

Since I'm too lazy to browse the hydrogenaudio listening tests and edit the article, perhaps you should pay a visit. 128 kbps IS good-excellent quality if the very latest LAME 3.98 encoder is used with VBR. You say you can't notice a difference up until 224 kbps. I regret to inform you that people with your exceptional hearing are in the minority. This article discusses what majority of people perceive as good quality. —Preceding unsigned comment added by 70.65.204.30 (talk) 06:20, 6 December 2008 (UTC)

The issue in question is not "good" or "excellent", but "transparent". Furthermore, how do we know that test subjects define 4/5 as excellent and 3/5 as adequate? Are those the values given to the subjects by definition (i.e. "If excellent, score it a 4. If adequate, score it a 3."), or are you making an assumption of how different test subjects will rate things? Unless told to score it differently, I would naturally consider 5/5 as excellent, 4/5 as good, 3/5 as mediocre, 2/5 as poor, and 1/5 as awful. On that scale, I would say that 4/5 is where adequate begins because I don't want MP3s that are not good.

Again, if the test is just comparing one lossy format against another lossy format, then it is completely invalid with regards to measuring transparency. The only valid way to test for transparency is to compare the lossy format to a format that is guaranteed to be lossless (WAV, AIFF, FLAC, etc.). If that has not been done, then the test should not be used as a reference to make claims of transparency in this Wikipedia article.

It also appears that the test did not control for the quality of the audio equipment. Instead, it looks as if subjects may have taken the tests on their home computers which could have easily had tiny laptop speakers. All of these different issues which can make a test invalid are the reason why I want to read the test procedures. I don't see detailed test procedures being explained on the web site that is used as a reference.

Also, 128 kbps is CBR, not VBR. In the LAME lingo, you can say there is such a thing as 128 kbps ABR, but there is no such thing as 128 kbps VBR. However, unless explicitly explained in this article, I think most casual readers will interpret "128 kbps" to mean CBR because that's what it usually means.

Lastly, I did not say that I can notice a difference below 224 kbps. I said the last test I've seen that compared MP3 (CBR) to lossless audio found that 224 kbps was the point at which absolute transparency was achieved, although 192 kbps was judged excellent but not transparent. I do know that a number of years ago when listening to songs using VBR, I could very easily hear when it switched from 160 kbps down to 128 kbps. I don't know what codec I was using at the time. (FWIW, this past weekend I created my own one-person double-blind test—I was both tester and test subject—and could not hear the difference between "lame --preset extreme -q 0" and lossless on my home audio system. It was not an abx test, but simply a randomized a/b test for my own use.)

I will try to find the test procedures on hydrogen audio. I've been reading through it in the past few days, but so far I've only read an article on MP3, an article on LAME, and a discussion forum. --JHP (talk) 01:41, 9 December 2008 (UTC)

Even the Transparency (data compression) article specifies a higher number than 128 as the MP3 transparency threshold—192 is discussed but not established as the lower limit of transparency. I say take out the reference with its inadequate, no-control test and take out any text that it was supposed to support. Binksternet (talk) 05:33, 9 December 2008 (UTC)

Mr. JHP, since I'm ignorant of Wikipedias quoting mechanism I'll just address each of your paragraphs individually:

(paragraph one)

Because the ABXing software they use has sliders which defines 5 as transparent, 4 as "perceptible, but not annoying", 3 as "slightly annoying", 2 as "annoying" and 1 as "very annoying." Anything 4 and above would not be noticed without a comparison to the raw file, and something like 4.5 definetaly means that it wouldn't be perceptible without direct attention paid to the artifacts. However, many participants on Hydrogenaudio are retards who themselves abuse the rating system and cant do shit without hand-holding.

Also, transparency CAN be reached (I repeat, transparency, not high quality) at 128 kbps for many songs. I have validated this myself many times by conducting my own double-blind tests.

(paragraph two)

I agree 100%. THe tests at hydrogenaudio DO use the raw file as a reference, and I personally never heard of any serious test having an already-compressed file as the reference.

(paragraph three)

Maybe, but high quality headphones are easily affordable, recommended to be utilized for the HA.org tests. I myself use high quality headphones and my evaluations of quality doesnt deviate much from said tests on HA.

(paragraph four)

1998 was 1998, now is now. I dont know how you drew the conclusion that "VBR" is non-existant. However, CBR is still popular so it's possible that readers may interpret it as that. In such a case, clarify it by adding "if VBR mode is used" or simply "128 kbit/s VBR."

(paragraph five)

Yes, CBR. CBR is an outdated concept so I wouldnt be surprised. All modern encoders and devices support VBR. A test conducted half a decade ago using a technique not from this millenium would not be a good reference to use, and should removed from this article.

Yes, 160 kbps should be distinguishable from 128, as 128 kbps commonly smears cymbals, as well as percussion to a lesser degree. --70.65.204.30 (talk) 07:24, 9 December 2008 (UTC)

Just a reminder that comments about personal test results and personal opinions of what makes for a good bit rate don't hold any water in the matter under discussion. The question is about test procedures that may or may not be up to Wikipedia's standard for a reference, and whether to delete the text that such a reference is supposed to support. As for the difference between a score of 4.0, 4.5 and 5.0—if 4.5 had been intended as the line in the sand where audible differences vanish from perceptibility, then the test designer would have dropped the 5.0 completely and made 4.5 the top score. Ridiculous? Yes. That's why I see 5.0 as establishing transparency, not 4.98 or 4.99 or whatever. Gimme five... Binksternet (talk) 08:20, 9 December 2008 (UTC)

Umm, there is no "fact" about what makes a good bitrate. There's a good reason why MP3 is called a "perceptual" coding algorithm. All quality at every bitrate is perceived a.k.a subjective. Which is why subjective (double-blind ABX) testing is best known method for evaluating the quality, and all are "personal opinions" blended into one. Also, where did I say that 4.5/5 is transparent? I said 4+ is not immediately noticeable without a comparison, hence is high quality or "transparent" for all intents and purposes. A 4.0 would be a perceptible flaw but not distracting in any way, a 4.5 would be much harder to notice. I'd give a 4.9 whenever I would notice a difference when concentrating extremely hard but not being able to explain what the difference is, just that there is one.

About providing a reference, browse HA.org's "Listening tests" forum and look for the recent 128 kbps MP3 test. I'd say it definetaly qualifies for a reference, but the fact that any idiot who can cheat the test is allowed to participate personally bothers me. I'd prefer a more official test where strict rules are imposed upon the rating system, and expunging any illegal score like when one participant would rate an obviously low quality sample 4/5 or vice versa. I've seen a couple, such as the 64 kbps AAC vs. WMA done by Microsoft, but most are outdated and don't provide details of what equipment/software/version/settings they used. —Preceding unsigned comment added by 70.65.204.30 (talk) 10:19, 9 December 2008 (UTC)

Transparency means that the audio is indistinguishable from the original. Therefore, in an abx test, a transparent MP3 would get the same score on average as the original (allowing for statistical error). If it gets a lower score, that is proof of non-transparency. Within the range of statistical error, a transparent MP3 would have a 50% chance of scoring better than the original. If an MP3 has any artifacts audible to the human ear, it is not transparent—by definition. Your own statement that "160 kbps should be distinguishable from 128, as 128 kbps commonly smears cymbals, as well as percussion" means that 128 kbps is NOT transparent. How can you claim in one statement that 128 kbps is transparent and in another statement claim that there is a difference in the audio quality between 128 and 160? The fact that they sound different at all is proof of non-transparency.

Also, I never said that VBR is non-existent. I said 128 kbps is not VBR. 128 kbps is a bit rate. VBR is variable bit rate, which means that it changes. Sometimes during a song it may be 128 kbps, other times during the same song it will be a different bit rate. I suggest use of the phrase "128 kbps average variable bit rate" to describe VBR averaging 128 kbps, because it is not 128 kbps all the way through the song.

Also, CBR is not an outdated concept. I believe most online music stores sell downloadable songs at a constant bit rate. Apple uses 128 kbps AAC, except for iTunes Plus which is 256 kbps AAC. Real used to use 192 kbps AAC, but is now selling 256 kbps MP3s. Amazon.com uses 256 kbps MP3s. (I have not actually purchased songs from these retailers since their move to DRM-free songs, so I have not verified these bit rates by playing songs. I'm just going by what they advertise.) --JHP (talk) 02:22, 10 December 2008 (UTC)

How can you claim in one statement that 128 kbps is transparent and in another statement claim that there is a difference in the audio quality between 128 and 160? The fact that they sound different at all is proof of non-transparency. No, I said SOME songs can be transparent at 128 kbps, but for many of them, the cymbals and percussion may slightly be smeared due to MP3's gayass framesize intervals design.

Second paragraph: Let's not get nitpicky here. Text like "128 kbps VBR" is publicly recognized. If the reader wants more detail, he can click on the wikified 'VBR' and learn more about it.

And just because big companies are stupid enough to use outdated technologies doesnt make them less outdated. They aim for compatibility, not quality. Otherwise we would all be using MP4s now instead of MP3s as they are way higher quality. They simply use CBR because not all devices support VBR.--70.65.129.18 (talk) 05:46, 10 December 2008 (UTC)

I said SOME songs can be transparent at 128 kbps, but for many of them, the cymbals and percussion may slightly be smeared... —Are we in agreement then, that 128 kbps is not usually transparent? Can I remove the claim from the article?

Text like "128 kbps VBR" is publicly recognized. —I'll accept that as long as "VBR" is linked to the variable bitrate article.

Otherwise we would all be using MP4s now instead of MP3s —I won't get into a debate over what is meant by "outdated". As far as MP4 (AAC) is concerned, the iPod supports it, but most other devices don't. iTunes Plus downloads are in the MP4 format (.m4a extension), but they're still CBR. I was disappointed to see that Real switched from AAC to MP3.

By the way, Wikipedia does have a special block quote template but I think you can also use the HTML <blockquote> element. --JHP (talk) 09:44, 10 December 2008 (UTC)

I removed the 128 kbps transparency claim. I replaced it with a 192 kbps VBR transparency claim after reading through the hydrogenaudio.org forums. They don't constitute a reliable source, so I didn't use a reference. I'll let other Wikipedians decide whether the bit rate transparency claim should be removed altogether. --JHP (talk) 22:37, 10 December 2008 (UTC)

Dude, what you removed was not a transparency claim. It read "128 kbps was somewhere between annoying and acceptable with older encoders, but modern encoders can provide adequate quality at such a bitrate." "Adequate" and "transparent" are two different things. Change that shit back.

Btw, do you have MSN? Would you be willing to ABX a couple samples and prove you can discern a 128 kbps output from the original?--70.65.209.57 (talk) 06:25, 11 December 2008 (UTC)

I removed that paragraph because I felt the claims made in the paragraph were dubious, but I have put it back for now. However, I have removed the paragraph with the transparency claim altogether. If someone has a reliable source for a transparency claim, they can put it back.

No, I don't have MSN. I didn't realize anyone used MSN anymore. Asking me to do an ABX test is irrelevant because it would violate Wikipedia's rules against original research. However, the ABX test cited in the article proves that 128 kbps average VBR is not transparent. Even this newer test proves non-transparency. If it was transparent, 5.0 would be within the range of the sampling error. --JHP (talk) 09:18, 11 December 2008 (UTC)

I yanked the word 'heavily' and the sentence "While quality around 128 kbit/s was somewhere between annoying and acceptable with older encoders, modern MP3 encoders can provide adequate quality at those bit rates." The words 'annoying' and 'acceptable' and 'adequate' were not proven by the reference and without those words, the sentence was worthless. 'Heavily dependent' was not proven by the reference, either, though the simpler 'dependent' was. Binksternet (talk) 12:34, 11 December 2008 (UTC)

Thank you. Those are the reasons I had deleted the paragraph to begin with. I have removed the "dubious" tag because I don't think the wording is dubious in its current form. --JHP (talk) 00:34, 12 December 2008 (UTC)

I just bought my first MP3s from Amazon.com last night. I need to correct my comment about them using CBR. Although Amazon may list the bit rate for a particular song as 256 kbps, they actually use VBR most of the time. It looks like they are using the equivalent of LAME -V0, which is the highest VBR possible. According to the hydrogenaudio.org forums, that is exactly what they are using most of the time. However, this is all original research. For a reliable source, Amazon lists their MP3 specs here. --JHP (talk) 04:35, 12 December 2008 (UTC)

I doubt the Hydrogen Audio wiki counts as a reliable source, but here it claims that LAME encoding at ~170 kbps VBR and above "will normally produce transparent encoding". I have no idea what empirical tests that claim is based on, but it is still considerably higher than 128 kbps VBR. --JHP (talk) 04:53, 12 December 2008 (UTC)

Hmm... guys, don't you think the formality of references and "original research" is a little ridiculous for proving whether a bitrate of a perceptual codec is annoying or adequate? The proof is in the pudding, and any normal person can easily evaluate the quality himself. Can't we just agree that 128 kbps is OBVIOUSLY not annoying, especially with modern encoders?

I realize a sentence must be supported with a reference, but I personally don't trust tests conducted by big companies as they dont list details and for all we know they could be using an encoder not from this millennium. This is why I like to validate such claims myself with my own unbiased tests. —Preceding unsigned comment added by 70.65.201.108 (talk) 08:26, 12 December 2008 (UTC)

References are not required on Wikipedia, they are just preferred. However, questionable claims are likely to be removed by another editor if they are not backed up by a reliable source. On the other hand, Wikipedia's rule forbidding original research is absolute. Again, the ABX tests and the Hydrogen Audio LAME wiki page both contradict the claim that 128 kbps VBR is transparent. --JHP (talk) 05:30, 13 December 2008 (UTC)

Yes, we've agreed that 128 kbps is usually not transparent. However, if I recall the paragraph correctly, it read something like: "The transparency threshold can be estimated at 128 kbps." Which doesnt imply "128 kbps is transparent." Moreover, claiming that one stating "128 kbps provides adequate quality" constitutes original research is like saying that someone who claims that all elephants have a trunk is also doing such. Here's another reference from a site that's something like Wikipedia, 'cuz it allows any person to evaluate random samples in a double-blind fashion. This is the closest thing to a reliable reference, as they list details and everything, but again, I don't like the samples they use. IMO they dont have a variety of them.--70.65.229.62 (talk) 07:41, 13 December 2008 (UTC)

Saying "The transparency threshold can be estimated at 128 kbps" implies that 128 kbps is transparent. If that's not what it means, then I don't know what the hell it means. Also, I didn't say that "128 kbps provides adequate quality" is original research. I said that your opinion constitutes original research, as does mine. --JHP (talk) 03:37, 16 December 2008 (UTC)

It means exactly what it says: that transparency MAY be estimated at 128 kbps because some audio samples can be transparent at that bitrate, but not all. The original paragraph was: The transparency threshold of MP3 can be estimated to be at about 128 kbit/s with good encoders on typical music as evidenced by its strong performance in the above test. However, some particularly difficult material, or music encoded for the use of people with more sensitive hearing, can require 192 kbit/s or higher.

It clarified itself later by saying that higher bitrates are required for transparency, that 128 kbps is just the "sweet spot" because it can sometimes be transparent, and good/excellent when it cant, and on top of all that: a significantly lower bitrate than ones required for guaranteed transparency.

What part of any of this is misleading or unfair?--70.65.229.62 (talk) 23:05, 16 December 2008 (UTC)

First of all, the references don't even back up that claim. Second, the wording leads readers to think that 128 kbps frequently achieves transparency, when you seem to be arguing here that it will sometimes on rare occasions achieve transparency. Since 128 kbps usually does not achieve transparency, it is no "sweet spot". You really are stretching the truth and presenting weak arguments to make the case that Wikipedia should say to encode at the bit rate that you apparently choose to encode at. I'm done arguing this with you. You haven't made your case. Quite the opposite, the listening tests demonstrate that 128 kbps is not transparent in most cases. --JHP (talk) 02:39, 17 December 2008 (UTC)

I wouldn't say rarely, it does more often than that, though I can't say how much 'cuz I've only done 3 total tests myself where it was transparent once, and I dont know any official test that can confirm this (why would they care?)

A "sweet spot" is a bitrate that's a balance between quality and filesize, and in MP3s case: 128 since it's 50% shorter than 192 for just a minor loss in quality.

Wikipedia should say to encode at the bit rate that you apparently choose to encode at. Actually, I don't use MP3 at all, as it is far outdated IMO, and I think everyone else should also trash it and move on. I got no special interest to promote any particular bitrate of MP3 or MP3 itself, in any way whatsoever. Just like you, I wanna improve the article and make sure newbies are properly informed about the various aspects of MP3. Everything I do on Wikipedia I do with good faith. I'm not perfect but I contribute what I can when I have time.

But, back on topic: why do you think the majority of people encode their music at such bitrate? Because they were brainwashed to perceive such low quality as high? No! Because it sounded perfectly listenable to them and they wouldnt notice a difference without a direct comparison, much like the quality differences between 44 kHz and 22 kHz. But mainly it was to relieve HD space. That was back then. Nowadays, people encode at 128 kbps because the technology improved and gives even better quality than before... and they still save the same space.--70.65.229.62 (talk) 14:21, 17 December 2008 (UTC)

Forgive me for not assuming good faith. Like you, I think MP3 is outdated. I personally favor AAC, but the hardware support on non-Apple devices is lacking. Hopefully that will change as iTunes moves to DRM-free AAC files.

While encoding at 128 kbps was popular a decade ago, I don't believe it is what most people use anymore. Among online music stores, Apple is alone in using 128 kbps, but they don't even use MP3. For years most competing online music stores have been encoding at at least 160 kbps (but didn't use MP3 either). The new norm—looking at iTunes Plus, Rhapsody MP3 Store, and Amazon MP3—appears to be 256 kbps. As for the sweet spot, based on this graph, I'd say the sweet spot for LAME encoding is either -V0 (~240 kbps) or -V3 (~170 kbps). File size declines very rapidly between lossless and 320 kbps with little loss of audio quality, then declines less rapidly from 320 to V0, again with little loss of audio quality. Below V0, file size starts declining at a much slower rate. As for audio quality, it declines very slowly between 320 and V3, then begins to drop off rapidly at bitrates below V3. 128 kbps VBR would be LAME encoding with V5. --JHP (talk) 08:49, 18 December 2008 (UTC)

I didnt say you werent, I was simply describing my position (that I'm not promoting MP3.) The encoding preference of a large entity does not denote the preferences of the public, and isn't a good reference in general. If say Amazon encodes at 256 kbps, what does that prove? That 224 kbps is not transparent? The reason they encode at such high bitrates is to preserve more technical quality -- to make it as close to CD-quality as possible, since bandwidth really isn't an issue like it was a decade ago. 128 kbps MP3s are slowly being replaced by 160/192 on filesharing networks, but a quick search with Limewire clearly shows 128 kbps songs in the majority (and 112/96 for longer songs are also fairly common.) I could take a couple screenshots if this passes off as a reference.--70.65.229.62 (talk) 21:38, 18 December 2008 (UTC)

<--outdent... References are required when there are differences of opinion between editors. I think 128 kbps sounds like ass, making for a strong disagreement between at least two editors on the subject of 128 being "adequate" or not. Adequate for what? 128 is certainly adequate for identifying the song's lyric or artist or beats-per-second. It's adequate for mild background music under noisy conditions. It's adequate for subwoofer bandpass sine wave test tones. It's not adequate in a critical listening situation for a picky listener who knows what kind of artifacts to listen for. This is why we need to find reliable sources in order to back up the text. Binksternet (talk) 15:03, 13 December 2008 (UTC)

128 is adequate for all intents and purposes except professional, where the waveform must be perfect to the bit (they also use a 176 KHZ sampling rate in those studios for extra precision.)

You say 128 sounds like shit, care to prove it? I'll upload a link containing 10-20 seconds samples and you ABX them (see JHP's talk page for details.) And THEN tell me if 128 still sounds like ass.

I'm not trying to get nitpicky or act like a smartass, but I believe your opinion is heavily biased by the low appearance of the number "128" and you suffer from some placebo effect (mini-delusion.) But even if you can discern a difference (you should anyway) and you insist a barely-audible washed-out cymbal artifact bothers you so much, then your opinion is in the far minority. This article, again, discusses what majority of people accept as high quality with great confidence. Besides, why do you think majority of MP3s on filesharing networks are encoded at such a bitrate if the quality sucks as much as you make it out to be?

It's not adequate in a critical listening situation for a picky listener who knows what kind of artifacts to listen for. Well, if you directly concentrate on the barely-audible artifacts and ignore the actual music then I wouldnt wonder. And... if you listen to the music like a normal person then you wont notice artifacts, unless of course the artifacts are so prominent that they're distracting, and that is NOT the case with 128 kbps, but something below like 96, 80, 64 etc. As I said, conduct some actual evaluations yourself and then get back to us.--70.65.239.174 (talk) 12:41, 14 December 2008 (UTC)

You presume quite a lot about the way I listen to music and about the kinds of data compression evaluations I might have taken part in and you end up being wrong on all counts. You also put the number "176 kHz" into the category of "professional" without specifying who, where or for what purpose. Where did this information come from? Hey, I'm a professional audio guy and I don't work in a studio. What number do I use? Every audio engineer has his or her own opinion about what makes for 'transparent' or 'adequate' compression, and rarely do the professionals that I've come across use the same bit rates that satisfy the masses. At any rate (pun intended), this is a talk page for improving the article. My opinion doesn't count, your opinion doesn't count nor would any original research undertaken by me or you. Any discussion that doesn't involve tightening the references isn't a discussion working toward improving the article--what makes the article good is a fair number of on-target references to well-designed tests undertaken by acknowledged experts in the field. Binksternet (talk) 04:53, 15 December 2008 (UTC)

I don't remember, I read somewhere that studios use 4x the sampling rate of 44.1 KHz for accurate engineering and reproduction of the waveform. That wasn't the point. My point was that if you merely want to listen to the music, 128 provides excellent quality. If you plan to engineer that song later, remix it or whatever then you shouldnt be using a lossy codec at all. the way I listen to music and about the kinds of data compression evaluations I might have taken part in so you have ABXed an 128 output with the original?

FYI, I have provided a reference, two of them. Hydrogenaudio's latest 128 kbps listening test and soundexpert.com's 128 kbps page. Both rate around 4.3/5 which logically indicates excellent quality. Now restore that paragraph that says this bitrate provides adequate quality.--70.65.229.62 (talk) 08:29, 15 December 2008 (UTC)

Neither the Soundexpert.com reference nor the Wiki Hydrogenaudio reference characterize 128 kbps stereo mp3 files as sounding adequate or acceptable. I don't think a restoration of those words is supported. Binksternet (talk) 13:43, 15 December 2008 (UTC)

I get it now, anything (even lightly) below transparency is automatically garbage to you. Unfortunately, the public doesn't follow your logic. If the artifacts are not distracting, then the song is perfectly listenable to their ears, thus adequate. And those who cant afford any loss use FLAC.--70.65.229.62 (talk) 14:07, 15 December 2008 (UTC)

I'm tempted to blatently advertise FLAC here. This is the most hilarious discussion ever. Accepted bitrates! On wikipedia! I never seen anything like it, really. Elm-39 (talk) 14:36, 15 December 2008 (UTC)

What is up with the poor performance of LAME 3.98 at 224 kbps in this test? It seems to be doing significantly worse than LAME 3.97b1 with --vbr-new, even though the two algorithms should be essentially the same. (If either were better, it should be the newer one.) --JHP (talk) 02:54, 17 December 2008 (UTC)

All of them (except the last) are rated above 5 which means they're all transparent, hence all sound the same. Though I wonder how the hell their rating system works if it can overlap the maximum. And look at MP2, since when was it ever better (even technical quality) than MP3? Someone should read their PDF and find out how ratings exceed the maximum.--70.65.229.62 (talk) 14:21, 17 December 2008 (UTC)

Since my last post, I read their description of the tests. For audio tests that are near transparency and would normally require golden ears with premium equipment, they use software to audibly magnify any defects. It's how they overcome the problem of not controlling for hearing ability and audio equipment quality that is a flaw in any internet-based audio tests. It seems pretty slick actually. --JHP (talk) 07:22, 20 December 2008 (UTC)

Magnify artifacts, how? You can't without using a lower bitrate. On the other hand, they could probably subtract the 224 kbps output from the original, amplify the resulting difference and mix to original, but clearly this is pointless, especially if MP2 was rated better than MP3.--70.65.229.62 (talk) 22:46, 21 December 2008 (UTC)