1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.

    Dismiss Notice

Public Listening Test of codecs MP3, AAC MP4, OGG Vorbis and Opus.

Discussion in 'Sound Science' started by igorc, Jul 29, 2014.
2
Next
 
Last
  1. IgorC
    Codecs AAC(MP4), OGG Vorbis, Opus@ 96 kbps are compared to classic MP3@128 kbps.

    Which is better? AAC? MP3? Opus? OGG Vorbis?

    The following codecs will be tested:
    AAC iTunes 11.2.2
    Opus 1.1
    Ogg Vorbis aoTuV Beta6.03
    MP3 LAME 3.99.5
    AAC FAAC v1.28 (Mid-Low Anchor)
    AAC FAAC v1.28 (Low Anchor)

    The homepage of the test:

        http://listening-test.coresv.net/

    There are 40 samples. From this test, the webpage automatically allocates listeners to odd group and even group, and each listener tests odd samples or even samples.
    The test ends on August 30th. Thank You. 

     
  2. bigshot
    It's interesting that there are so many tests to determine which of the codecs is best at lower bit rates, and there aren't many comparing different codecs to help people determine the point of transparency. That is the comparison that would have real world value for audiophiles. I see a lot of people saying "Lossy sounds like dog droppings." and at low bit rates, I agree with them. But at high bit rates, it's possible to achieve reproduction that is identical to the source with human ears.
     
    The question that needs to be asked is lossy vs. lossless... If someone did a listening test that covered 1) the threshold of transparency with various codecs, 2) the threshold where multi generation transcoding becomes audible, and 3) relative file sizes of the transparency thresholds, audiophiles would have all the info they need to determine whether lossy is a suitable replacement for lossless for the purposes of listening to music and building a library.
     
    I did that for myself and settled on AAC 256 VBR. But I see a lot of people who haven't done the listening tests saying, "Maybe lossy is good enough, but I'm going to rip to lossless just in case." If you've done the test and established the thresholds, there's no "maybe" and no need for "just in case".
     
  3. ferday

    I think the lossy vs. Lossy debate may be flaring up now that streaming is de rigeur, lossless streaming really is a waste of good bandwidth on 3/4 G....I've noticed on commercials now some of the cell providers are offering "unlimited" music streaming with no penalty to bandwidth use (provided you use their streaming service, of course). What if 96k opus is indistinguishable from 320 mp3? There's something to change the streaming landscape forever...

    People like yourself (and myself) that collect massive amounts of albums are on the way out. Heck even in these vaunted audiophile forums a large percentage is streaming.
     
  4. nick_charles Contributor
     
     
    The idea of streaming a song each time you want to listen to it seems rather wasteful, not that we have any real issue with running out of bandwidth at the moment but even so...
     
  5. bigshot
    Well, when you sign up for Spotify, you effectively have the biggest record collection in the world. Eveyone shares one big library instead of each one of them owning their own copy of Dark Side of the Moon and Abby Road, so from that angle, it's not wasteful.
     
  6. IgorC
    Yes,  128 kbps and higher will be more interesting. 
    But high bitrate are hard to test. People inform that it's already hard to listen difference between a high quality MP3-128 kbps and a lossless. 
     
    People can claim that 128 kbps is low and not enough for a good quality  because they can have a bad expierence with a low quality encoders. Different MP3 encoders produce different quality.
     
    MP3 l3enc 128 kbps VBR - 1.56 (very bad quality)
    MP3 LAME 128 kbps VBR - 4.51 (very high quality) 
    http://listening-tests.hydrogenaud.io/sebastian/mp3-128-1/results.htm
    results.png
     
     
     
     
    If codec ranks higher than  4.70 score it's considered transparent (at least very close to it) for a real usage.
    128 kbps was tested previously
    http://en.wikipedia.org/wiki/Codec_listening_test
    http://listening-tests.hydrogenaud.io/sebastian/mf-128-1/results.htm
    results.png
     
     
     
     
    if You think that 96 kbps is very low  You should try to submit some samples from current open test http://listening-test.coresv.net/
    You will be surprise how good is quality for a modern codecs [​IMG]  even at such bitrate
     
  7. bigshot
    The problem with assigning a score to determine whether a codec transparent is that most people assume transparent means "perfect". I think that is what they really want out of a codec. No one says, "I'm going to encode my music so only 7% of it has artifacting errors. We want a codec to encode everything flawlessly.
     
    The way to determine whether a codec is really transparent is to throw lots of very difficult to encode music at it and find the point where artifacts are no longer present. I have one particular killer track I use to try to get codecs to go splat. I'm sure there are others. It's not a matter of testing a whole bunch of people and finding the average among all of them. You're measuring how well the codec encodes music, not how many people can hear artifacting.
     
    I always see people saying things like, "High bitrate lossy may be good enough for you, but my ears are sensitive and I can hear the subtle difference." That is totally wrong thinking. When a codec artifacts, you, me, my mom, the guy down the street watering his lawn... everybody can hear it. It is a very obvious form of distortion. Higher bitrates don't make subtler artifact distortion, they just reduce the frequency and duration of the artifacting. At a certain point, artifacts don't occur at all. That is the point of true transparency.
     
  8. IgorC
      All right, it's possible to test high bitrates. Though we need a good amount of samples for this test to justify move to higher rate.
     
      Have You submit some samples?
     
  9. bigshot
    Happy to... I spent a week testing all kinds of music with various encoders and bit rates when I was figuring out my specs for my media server. I found one CD that had a string tone that was extremely difficult to encode. You can get a copy for $6 shipped from Amazon used. (it's a great CD too!)
     
    http://www.amazon.com/The-Decca-Years-Sammy-Davis/dp/B000002OF2/
     
    Try encoding this at lower and higher bit rates and listen to the gurgling in the massed strings of the orchestra. I found that Frauenhofer 320 *almost* eliminated it. Very very small amount of artifacting left. 320 LAME and 256 AAC rendered it perfectly. Anything much below that sounded more and more like outer space gargling, depending on how low you went. Remember, these are mono recordings, not stereo. So if you encode in mono, the bit rate would be half that of stereo. It's easier to hear the artifacting if you force it to encode in stereo though.
     
    Also, if you have any live albums with applause, try encoding that. Applause is one of the hardest things for encoders to render.
     
  10. IgorC
    I mean,  have You submited some results for this test?
     
  11. bigshot
    I found that transparency in the various codecs occurred at AAC256, LAME MP3 320 and Frauenhofer MP3 (almost) at 320. With simpler music to encode AAC192, LAME MP3 256 and Frauenhofter 320 were transparent. Some music was fine a full stop below that.
     
  12. IgorC
    Since You don't cooperate and don't even answer a simple question.
     
    All right. Nevermind. Have a nice day. 
     
  13. IgorC
    The test is finished. 

    These are the results
    http://www.hydrogenaud.io/forums/index.php?s=&showtopic=106354&view=findpost&p=874674
     
    2qal28i.png
     
  14. bigshot
    What bitrate the threshold of imperceptibility for the various codecs? That is what I would be interested in. I really don't care which codecs and bit rates are sort of imperceptible. If I am going to encode in lossy, I want to know the line at which it becomes audibly  identical to lossless for all music.
     
  15. eahm
    bigshot, you have to test by yourself, you may not even agree to anything the testers submitted. Get some time and peace, ABX and decide your sweet spot bit rate.
     
2
Next
 
Last

Share This Page