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

FLAC vs. 320 Mp3

Discussion in 'Sound Science' started by icedup, Sep 7, 2011.
30 31 32 33 34 35 36 37 38 39
  1. gregorio
    You'll have to forgive our scepticism. Every so often someone posts here that they can absolutely hear this difference but WITHOUT exception, it turns out they can't. Either they do hear some difference due to some serious flaw in the test (for example, they inadvertently compare different masters) or they are perceiving a difference due to some cognitive bias, in which case there is also a serious flaw in their testing. You should be aware that over the course of many years, many tens of thousands of tests have been conducted by many thousands of subjects (ranging from average members of the public to highly trained and experienced professional sound engineers), using pretty much everything from cheap consumer equipment to the very best pro studio equipment. In fact, for nearly a decade, the designers of codecs no longer test at 320kbps because no one can ever hear any difference, instead they test at much lower kbps rates (typically 128kbps and lower).

    1. No, you can't "absolutely say with confidence that you can hear a difference", although you can maybe say with confidence that you "perceive" a difference (almost certainly due to a flaw in your testing). In effect, your statement tells us more about what you are publicly prepared to "absolutely say" and your "confidence" than it does about any potentially audible difference between 320kbps and lossless!

    2. How can you be 70% accurate with discerning differences in "song quality detail" between 16 and 24bit, when there are no differences in "detail"? As bigshot stated, as digital theory stipulates and as objective measurements confirm, detail is the same at 16 or 24bit, the only difference is the level of the noise floor, which in both cases is well below the threshold of audibility. Your statement therefore further confirms that what you're noticing is actually a flawed test/cognitive bias!

    I'm not saying with 100%, absolute confidence that you cannot actually hear a difference. It's possible, though incredibly unlikely, that for some genetic/physiological reason you have significantly different/better hearing abilities than everyone else but with so much reliable evidence that no one can hear the difference, you need some fairly extraordinary and compelling evidence that it's not just a testing error on your part, BEFORE you can say with ANY confidence that you can hear a difference. You would need to do a proper ABX test (which is easy with foobar) and provide the different files you are comparing, which is also very easy. Every single time anyone here (or elsewhere) has adamantly claimed they can hear a difference, when asked for reliable evidence they either: A. Simply run away, thereby admitting that what they "can absolutely say with confidence" was in effect a lie, B. Actually "man up", do an ABX/Double blind test and discover they can't actually hear a difference once cognitive bias is eliminated and/or C. Discover a relatively obvious fault between their original test files (different masters for instance).

  2. Brooko Contributor
    Must be same master. Must be volume matched. Must be multiple tests to statistically be relevant.

    I'd suggest starting here - https://www.head-fi.org/threads/set...-guide-to-ripping-tagging-transcoding.655879/
    Post is old, software is free - all it takes is time. Try it with a true blind test, and then post your actual results. If you can still tell - post your files so other people can have a go.

    Most of the time, with people claiming diffs, they are:
    • Different masters
    • Not volume matched
    • Have artifacting errors during transcode
    • Not as "blind" as they'd like us believe
    • Not statistically relevant - anyone can pass a single test if its repeated until you get it. Pass 14/15 in a true volume matched double blind abx, and I guarantee people will want more info. Especially as I've seen no-one do it.
    The better headphones one is pretty funny. I have a pair of HD800S (detail monsters), and I can't tell aac256 from lossless with a proper valid test. We've had a young guy who successfully abx'd 2 killer tracks @mp3 320 vs redbook a few years ago. Stax system. We later found that one of the tracks had very minor artifacting during the transcode. When we repeated the tests with aac256 he couldn't pass them. Great hearing, great system - couldn't tell the difference.

    To date - I know no-one who has successfully abx'd aac256 vs FLAC in a valid test.
    TheTrace and sonitus mirus like this.
  3. Tsukuyomi
    The converting was indeed done by me, I got a 24bit track and down converted two tracks. one to 16bit and one to 320kbps.
    I didnt change foobars volume, left it default. and im not sure with the abx component for foobar, i'll have to look into that.
  4. Tsukuyomi
    I Dont mind at all the scepticism, in fact im always open for discussion.
    I need to look more into ABX testing as I havent tried this yet.
    I've tried to make multiple sample copies from higher bit files to smaller bitrates like 16bit/320kbp/128kbp.
    I think you may be right about certain artifacts appearing in either how i master the file or perhaps in my equipment or external variables when listening. I do notice that 16bit files are much "louder" or at least sound louder for certain genres of music than listening to the same thing on 24bit files. to me, 24bit sounds quieter ? maybe thats what you mean by sound floor. (i'm still new to this hobby, not listening to music but buying and collecting better equipment than just regular stuff.) i'm always open to learn more how to do testing to test myself and the material I use.

    I'm aware a lot of people claim sonic differences between things, file formats, power cables, audio cables, etc.. and I do like to test thoes things myself for myself to see if i can hear any change at all, thats part of the excitement for this hobby. :)
    but i will definitely learn from this and try to apply it to further testing.
  5. Tsukuyomi
    I use EAC already, its a great tool. and I do get 1 master file and make multiple copys of it for different rates.
    When it comes to volume matching i tend to not play with it, i set it to a comfortable -26db on my ADI-2 with my DT1990 pros.
    I do have EAC set to give best possible file with minimal errors or artifacts, i can't remember the youtube video i copied the settings from but it was very helpful.
    I wish there was a place we could go to with pro equipment and do more testing with multiple people and different ears :) id sign up gladly.
  6. sonitus mirus
    Since you not using an ABX tool like the one discussed in the links provided by @Brooko with Foobar, the chances are that the converted lossy file is not the same volume level as the lossless version by enough dB to be identifiable. The Foobar player's ABX tool has a ReplayGain option that can be applied to the tracks to ensure they are similarly matched by volume. This volume matching does not permanently change the file and can be removed once the testing has been completed.
  7. bigshot
    I have a lossy listening test that includes ten samples... one each of AAC, Frauenhofer MP3 and LAME MP3 at 192, 256 and 320, along with a lossless sample. It's set up properly so all you have to do is listen to each track and rank the sound quality from 1 to 10. It is a good way to find out where your threshold of transparency lies. If you would like to try the test, let me know.
    Last edited: May 14, 2019
    stonesfan129 likes this.
  8. gregorio
    No, that's not what I meant by noise floor. Noise floor is at the opposite, quietest end of the loudness spectrum. It's the almost inaudible hiss/noise that defines the lowest/quietest limit of dynamic range and comprises microphone/mic-pre-amp noise, the ambient noise of the recording venue/studio and various other noise sources (playback amp and speakers/HPs for example).

    At the other, loudest, end of the loudness spectrum, 16bit and 24bit are completely identical. So if you're hearing 16bit as being "much louder" than 24bit, there's really only two possibilities, either: A. There is some additional processing that been applied before/during/after the conversion to 16bit (compression with make-up gain for example), so you're effectively comparing different masters or B. There's some expectation or other cognitive bias occurring that's causing you to perceive/imagine 16bit is louder. Leave the suggested replay-gain switched off and try an ABX test: If you pass the ABX test that eliminates possibility "B" and tells you there must be some processing other than just conversion to 16bit being applied somewhere. BTW, even if you're not doing the conversion yourself, it's not uncommon for record labels to apply more compression/make-up gain to the 16bit version than the 24bit "hi-rez" version.

  9. bigshot
    I know iTunes alters the level slightly when it reencodes. In order to to a true comparison, you need to level match as the last step.
30 31 32 33 34 35 36 37 38 39

Share This Page