I've done a very detailed graphic comparison between many audio formats (Mp3 fhg, Lame 3.96.1, Lame 3.97a, Mp3pro, Wma, Mpc, Ogg, Nero Digital Audio and Vqf) at different bitrates.
The url is: http://www.telefonica.net/web2/audioformats

I try to explain how the comparison is made on the web.
I hope it to be useful to you. I accept opinions.

Regards,
Pablo

These test results mean nothing, this has been said a million times:

A. An artificial sound that you generated is not representative of most music.

B. Comparing pictures or graphs tells us NOTHING about how the file will sound.

C. WMA 9 pro is not the same codec as WMA standard and can't be played on any portable players at all.

Please read the forum rules and stop posting garbage.

I don't think you have any idea how "compatible" WMA really is. Neither WMA9 Pro nor Lossless are supported by any portables nor are they supported on the Mac.

Graphical plots are absolutely useless to judge audio quality, for a number of reasons. Your plots are very pretty though, and give some graphical insight as to certain modes of encoder operation (most importantly cutoff frequency). But your plots are autoscaled on the Y axis which makes them very difficult to compare across encoders (since different cutoff frequencies generate different scales).

Finally, this sentence from the conclusion is, frankly, pure naivete. It's clear that sound quality was never really a factor in this test.


There are a lot of other questions raised by the test - estimated bitrate of lossless codecs, no inclusion of QT AAC, MPC is only presented on the 128k page, allegedly with the --insane preset!

What was the original article that attempted a graphical comparison?

Analysing spectrograms is not considered to be a valid way of comparing lossy codecs. The reason is fairly simple - the codecs are designed to remove as much information as possible from the signal while still sounding identical. What the output looks like on a spectrogram is not a useful metric of the quality of a codec.


You seem to be missing the point of lossy audio compression. These codecs are designed to take a music (or spoken word, etc) signal and mangle it as much as possible while ensuring that the output sounds as similar as possible to the input. Taking a synthetic test signal and feeding it to the codecs then "locating the lossess" is not helpful.

http://www.telefonica.net/web2/audioformat...at_128_kbps.htm

MPC was used at --preset insane for 128 kbps test?!

nothing to see here, move along please people.

Welcome to the forums.

Personally, I would also prefer that the author either take down the pages or put a big/huge/gigantic/monstrously large disclaimer on top of the pages saying that the test is purely for fun and the results are meaningless, lest someone stumbling upon and actually citing that page as a reference for lossy codec quality comparison

Especially on the conclusion page.

Even though these pictures don't tell anything about how it sounds, I find the images pretty fascinating.

I've done a very detailed listening test between many video codecs (XVID, motion JPEG, WMV9, MPEG2) at different bitrates.
The url is: http://www.nowhere.com

I try to explain how the comparison is made on the web.
I hope it to be useful to you. I accept opinions.

I excluded MPEG4-AVC because it stands no chance (it sounds like crap).

WARNING: Turn off your subwoofer before listening with the lowest video bitrates !!! The artifacts may harm it

This reminds me of my 2003 development of a groundbreaking New Lossy Audio Codec.

Ah, memories

Ok, I have corrected my conclusion page. I thought wma pro and lossless would be compatible with players because mine plays them perfectly.
About pictures, I recognize I'm a neophyte, but I have to say that as a encoder version is actualized, the plots become more similar to the original one. You can check it in the web with lame and wma (I also have tested ogg 1.6 and the same occurs)


I don't mention quality because I suppose that everybody want the best quality


I encoded the original sound with insane preset and the result was a 127.5 kbps file. You can check it. I only publish that preset because the others didn't give similar bitrates to the ones I use for the comparison.

I even clicked on your site and si faboulous!!

Of course they get closer looking to the original if you increase the bitrate. The problem is, closer looking is not necessarily better sounding.

BTW, Musepack is really efficient at encoding stuff with simple spectrograms, IIRC.

Not really... mppenc was set with --quality 7, which corresponds to 230...240 kbps on average. The screenshot should be moved in the 256 kbps category.

Come to think of it, 127.5 seems a bit high. I seem to remember seeing bitrates much lower than that when using insane or braindead, when I encoded NES game music with mppenc.

I'm not totally sure though, it was a long time ago, and I did it just for fun.

Using Nero VBR at ::extreme:: in fast mode -> 123 kbps, and graphical representation is near perfection.

Funniest thing, I could ABX without difficulty mpc --insane on this sample: the beginning suffers from strange coarseness.

Last thing: the author recommands WMApro for quality and compatibility with most portable device, and seems to ignore that WMApro can't be played by standard WMA decoder.

From the test conclusions:

I don't know what kind of players you have in mind. The iPod recognises AAC in a MP4 container. Is that not the reason you included Nero in the test?
Also no known hardware player recognises WMA Pro.


Once again Nero is compatible with iPod.


The "spectrums" sure are pretty.
As of factors that affect sound quality, here is a start: the looks of the sound spectrum for an artificial waveform have REALLY VERY, VERY LITTLE to do with them.

At the end of the day it is better to trust your own ears and compare how the formats perform on the music you normally listen to. If you use the ABX methodology and get more people to do the same comparison, you can do a statistical analysis to obtain a scientifically valid comparison of formats.

Hydrogenaudio.org is the right place to look for an objective comparison of formats.


Lossless may not be necessary for your needs, but it is very necessary for number of people. You don't mention that the sound information missing after lossy compression is irremediately lost. If you decide to switch format later you would need to retrieve the original sound to take advantage of the new format.

I'm sure that this will be the guys first & only thread at HA, congratulations.

Gambit do you get a hardon when you see threads like this? you seem to love it

Ok, now I comment this on the web.


Artificial waveform is a waveform. I know very much music based on artificial sounds and synthesizers. I think encoders developers cannot only think about guitar waveforms.


I forgot to say that the plot of nero in 256 kbps category is certainly from the original file compressed with cbr 256 kbps option, but the result was a 195 kbps file (more or less, I can't check it now). Maybe that's because both audio channels are identical. With Nero occured the same at all bitrates. For normal music this doesn't happen.
I understand here there are many people in love with mpc and ogg, but i'm not surprised about Nero is the best audio encoder since Nero Digital is the best one for video (Doom9 test here)

This has to be a troll... I think it's hilarious. The inclusion of MPC --insane in the 128k category is classic. It's graph is crappy looking, but I wonder if it will sound better than 128k mp3... Hmmm - tough one there!

On a more serious note, I think the graphs are interesting even if they don't tell you anything about quality. I think you can see ringing in the WMA samples, for instance, and you can see what SBR looks like in the MP3pro graph in the 64k section.

Actually I think the above quoted sentence is what makes your test meaningless...Basically you state that your signals are completely different from real music. At least I think so

For his contribution here, the OP has so far been ridiculed, laughed at, told to stay away and close his page, and of being a stereotypically disruptive and abusive character. The more this treatment continues, the more it will attract trolls to the forums and it will be deserved. I suppose that will be non stop laughs then.


-even if is right, even if - there is no one around with the necessary skills to inform an analysis these visualations besides -"these mean absolutely nothing (to me) dont post this rubbish etc etc"

Thats the interesting stuff, Ive come to expect more of from you.

Please. There is a TOS here for a reason.


Graphs like this are meaningless for measuring quality. That point must be emphasized lest some poor neophyte stumble upon this thread and be misled otherwise.


I'm not here to impress anyone, ChiGung. I come here because HA is known for having scientific and objective discussions, something sorely lacking in the general audiophile community. I'm not particularly sorry if you are upset by my lack of patience for people who think wraping quackery in technical jargon makes it somehow scientific. Should we accept baseless claims about green pens and magic quantum stickers? Of course not. To do so would be to abandon the core principle of this community. These graphs, as a measure of relative audio quality, are no more valuable than any of the snake-oil they peddle in Stereophile magazine.

Many people come here for help, or to learn, or to ask questions, and I am willing to be one of those that helps where I can, just as others have helped me when I've had questions. There's nothing wrong with being a newbie, but we are a community with rules. Those who are unwilling or unable to work within those rules are free to look elsewhere, newbie, expert or otherwise.

"Ignorance more frequently begets confidence than does knowledge."
-- Charles Darwin