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

Lossless vs 128kbps mp3 vs 320kbps mp3 blind test

Discussion in 'Sound Science' started by chewy4, Jan 15, 2013.
1 2 3 4 5 6 7 8 9
11 12
  1. sonitus mirus
    Why don't you take the Philips Golden Ear challenge?  This could at least provide you with the knowledge that you can identify basic parameters of sound qualifications with some reference.
    How individuals hear is subjective.  Our brains decipher what we hear, and there are a multitude of factors that could influence what we hear and how we describe it to others.   You can always qualify your opinion by stating that you have Tinnitus, but you might want to include the frequency of the tone that inflicts you.  It probably does not matter, but at least this gives the reader a fundamental benchmark.
    krismusic and snellemin like this.
  2. krismusic Contributor
    ^^^ Good call. Thank you.
  3. krismusic Contributor
    Doesn't seem to work on an iPhone.
  4. sonitus mirus
    I can get the challenge tests to work on my Android LG G2 with the Chrome browser.  Perhaps the iOS version of the Chrome browser will work for you on the iPhone?
  5. krismusic Contributor
    Yep. I'm using Chrome. Either I am being dense or it doesn't work on IOS.
  6. krismusic Contributor
    Ah ha. Working. It's official I have tin ears!
  7. DiscoProJoe
    I was using my V-Moda M-100 headphones plugged into my laptop, and played the test files through Winamp.  I was using a slight treble boost on the Winamp EQ with an equivalent precut (as I normally have it set), and that's it.
    Anyway, my guesses were as follows:

    I correctly guessed the FLAC file on each set, and got the 3rd group right. But the acoustic guitar in group 3 made it really hard for me to make out any difference, so I was basically guessing on group 3.
    I felt somewhat confident about my answers on group 1, but got the MP3 bitrates wrong.
    Group 2 I almost got right, but didn't realize it had a trick to it.
    Prior to this test, I thought I'd be able to make out the 128 file somewhat easily, but couldn't always guess it right.
    Anyhow, I am not an audiophile and am proud of my music collection of almost 10,000 songs, nearly all of which are encoded as MP3s at my chosen bitrate of 192 kbps, so this test doesn't ruffle my feathers in the least!  [​IMG] 
  8. Vorpax
    My results:
    first group: A 128kbps                  B 320 kbps 
                     C flac
    second:      A flac
                     B 128 kbps
                     C 320kbps
    third:   A 128kbps
              B 320kbps
               C flac
    I messed up the first group entirely it seems [​IMG] 
    On the second group, the flac was pretty obvious to me. More impact, juicy sound. The 128 and 320 were guesses, but both clearly inferior to the flac, they sounded thinner.
    About the third group, the 128 kbps one was easy. More liquid and opaque than the other two. Getting flac right was a bit difficult on this one, but I don't think it was a guess, the low level detail sounded more rich. Flac overall not too different from 320kbps like in the second group but still.
    I'm using hisoundaudio Crystal, onboard sound card (Realtek ALC892) and windows media player LOL!
    The first group aside, I'm happy that lossless is not placebo effect for me even with no external DAC nor high-end headphones.
  9. katulu
    In order of what I think is increasing quality (128, 320, lossless):

    1 - C, A, B
    2- Cant tell, music sounds like noise...
    3 - A,B,C

    Wonder how I did...
    Looks like I got groups 1 and 2 right, at least. For me, the "tell" on 1 was the snare, on 3, higher note picking at the end.
    I kind of cheated in a sense - I always listen to my music EQd to the headphone I am using - don't know if that counts as a "bizarre" EQ.  Chain was Stream 7 Tablet, Jetaudio, Equalizer APO with custom settings, Normal USB to Chord Hugo, Norne Draug 2 cable, HD800s.
  10. katulu
    After reading some responses, some more thoughts:
    Like most, experience is a key factor in telling differences in files.  If you have read/experimented a little, you may know some of the usual flaws of lossy files - watery cymbals, smeared instruments. But one thing that baffles me is that in these tests, the factor of familiarity is never considered. Let me explain what I mean:
    I like steak, generally cooked medium. But I know when it's undercooked, or overcooked. All the music I know and love is my steak.
    Now, someone serves three plates of Dodo meat, and asks me to say which is overcooked, undercooked, or just right. What???
    I can use the principles of my previous experience to try to tell, but such as I have experienced before may not be present - thus I fail the test.
    It amuses me to think about creating some music which would purposefully sound liquidy and smeared, and which perhaps could somehow "clean up" with the linearization of lower bitrate! Experts would be baffled at realizing the "worse - sounding" file is the most correct one...
    If I don't know exactly how a file is *supposed* to sound, I can't really be expected to know which one is more correct. One is more readily able to tell quality differences when one has a baseline, I believe. I seriously think tests results would be quite a bit different with songs "everyone" knows...
    And this is without going into the details of music which may not be losing that much even when coded into lossy format... can the listener then be blamed? I saw one such test online a while ago, trying to prove no one could figure out where a file was lossy or lossless, and it proves nothing, in my view... I did a brief comparison of Morbid Angel's Covenant album 320 and FLAC, which I know well, and to my surprise the difference was fairly obvious - other more complex extreme metal is even more abhorrent of lossy files. I have likewise been surprised at how good some 192 kbps music has sounded too.
    Oh, I also forgot to mention I have tinnitus and hearing loss of higher frequencies in my right ear... and apparently I can still tell differences in SQ.
    Just some cents... Cheers, everyone.
  11. gixxerwimp
    +1 ... my feeling on these tests exactly. Haven't taken this one yet, but taken a whack at many others. Without being familiar with the music, it's really hit and miss.
  12. Sonic Defender Contributor
    Nope, shouldn't matter, if the differences are audible they are audible. If you need to study and train your brain to detect minute differences then for all practical reasons there is no difference. People don't train their ear, they listen to music. I still suspect that even with music you are very familiar with in multiple trials the results would be only marginally better. Today at our head-fi meet I had seven people do multiple trial tests between 320 and ALAC and nobody did better than 50%, and this included three people that swore they had proven in the past the could hear the differences. Hopefully nobody comes along with the standard claim that the gear must have not been resolving enough, the gear wasn't the issue.
    One participant did his 5 trials with the test gear and then wanted to use his T1 and PHA3 for two more trials. By this point he was quite familiar with the piece and big surprise, 50% accurate again. We will be doing this again to develop a bigger sample size, but with such a low rate of detection (50% is just guessing) the results are extremely unlikely to change. People would need to be detecting the differences at about 90% accuracy to really be in the area of statistical confidence.
    Brooko likes this.
  13. katulu
    It does matter, because that 50% is not whether people *can* detect a difference (yes or no, 50% probability - right?), it's whether they can identify the "better" quality given they say they detect a difference. If a person isn't familiar with the music, there is no baseline to say which is better quality - which makes that 50% number quite explainable. See my previous post for reasoning. Thus a test where the listener does not know the piece as it's supposed to sound is heavily biased in my view.
    And of course, it's BS to use musical pieces where encoding has negligible effect to "prove" bitrate doesn't matter/can't be detected. Not all music is equally affected
  14. Brooko Contributor
    Seeing as how we are in the Sound Science section - and because you are so confident, how about a little test (if you'd be willing to participate, and if we could set it up)?
    • You provide the files - whatever music you like
    • You provide own headphones
    • We arrange a DAC and amp that you're happy with
    • We arrange a third party observer - give him/her instructions to properly dither and resample the file to aac256
    • We use Foobar2000, volume match precisely, and blind abx - observer present to ensure everything above board
    • 2 lots of tests 15 - in each.  Because it's music you know well, and headphones you are familiar with, you should ace it easily.  We publish results
    If all of this could be arranged - would you be game?
  15. Sonic Defender Contributor
    So you know what music is appropriate to use? Can you describe what the fundamental characteristics are and how this has been demonstrated? The music piece we used is a fantastic recording by Holy Cole, One Trick Pony. The song has smokey female vocals, nice drums, acoustic bass and very nice guitars. It is extremely well recorded and dynamic so I rather doubt the piece of music wasn't up to snuff. As well, people were asked what version the preferred not whether or not the could hear a difference. The logic there is that everybody who assumes that a 320 file is sonically different implies it is different in a negative sense; therefore, you would of course by that logic prefer the lossless file.
1 2 3 4 5 6 7 8 9
11 12

Share This Page