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 mp3 ABX results. (Among other ABX's)

Discussion in 'Sound Science' started by deadlylover, Sep 30, 2011.
2 3 4 5 6 7 8
Next
 
Last
  1. deadlylover
    Thought you guys might find these results slightly interesting.
     
    I got pretty mad bro in another thread where people claimed it was impossible for a human to distinguish between high bitrate mp3 and lossless, so I grabbed life by the balls and tried it for myself.
     
    For these tests, LAME 3.98r was used, both tracks replaygained and tested using foobar ABX.
    Stax Omega 2's used for all tests. I have a feeling these bad boys brought up all the differences into my weapons range.
    No musical training, but I'm young and dumb, so that might count for something. [​IMG]
     
    Results:
     
    Lossless v 256k cbr, some kind of Japanese rock(?)
    No audible compression artifacts, the differences were in the focus and dynamics of the poorly recorded vocals.
    You can see me resetting quite often, I was checking different parts of the track and looking for a good spot to focus on.
     
    foo_abx 1.3.4 report
    foobar2000 v1.1.8
    2011/09/29 20:30:41

    File A: C:\Users\Alice\Music\祈り歌.flac
    File B: C:\Users\Alice\Desktop\05 祈り歌.mp3

    20:30:41 : Test started.
    20:31:07 : 01/01  50.0%
    20:31:31 : 02/02  25.0%
    20:31:42 : 03/03  12.5%
    20:31:54 : 03/04  31.3%
    20:32:12 : 04/05  18.8%
    20:32:51 : 05/06  10.9%
    20:33:05 : 06/07  6.3%
    20:33:21 : 07/08  3.5%
    20:33:36 : 08/09  2.0%
    20:33:59 : 08/10  5.5%
    20:34:21 : 09/11  3.3%
    20:34:40 : 09/12  7.3%
    20:34:42 : Trial reset.
    20:34:53 : 00/01  100.0%
    20:34:54 : Trial reset.
    20:35:03 : 00/01  100.0%
    20:35:04 : Trial reset.
    20:35:15 : 01/01  50.0%
    20:35:43 : 01/02  75.0%
    20:35:44 : Trial reset.
    20:36:05 : 01/01  50.0%
    20:36:12 : 02/02  25.0%
    20:36:28 : 03/03  12.5%
    20:36:56 : 03/04  31.3%
    20:37:04 : Trial reset.
    20:37:20 : 01/01  50.0%
    20:37:30 : 02/02  25.0%
    20:37:35 : 02/03  50.0%
    20:37:43 : 03/04  31.3%
    20:37:45 : Trial reset.
    20:37:59 : 01/01  50.0%
    20:38:18 : 01/02  75.0%
    20:38:21 : Trial reset.
    20:38:30 : 01/01  50.0%
    20:38:40 : 02/02  25.0%
    20:38:49 : 03/03  12.5%
    20:38:58 : 04/04  6.3%
    20:39:10 : 05/05  3.1%
    20:39:28 : 06/06  1.6%
    20:39:36 : 07/07  0.8%
    20:39:54 : 08/08  0.4%
    20:41:08 : 09/09  0.2%
    20:41:21 : 10/10  0.1%
    20:41:25 : Test finished.

     ----------
    Total: 27/36 (0.2%)
     
    320kbps, no idea what this genre is, it's kind of futuristic doof doof with female vocals.
    There is a loss of detail in the hard panned low level sounds, with a LOT going on up center, so I guess the psycho compression thought I wouldn't notice. This one was pretty easy, despite not hearing any obvious compression artifact giveaways.
     
    foo_abx 1.3.4 report
    foobar2000 v1.1.8
    2011/09/30 03:29:53

    File A: F:\Music\wave edebn\WAVE - 隻眼のエデン -A.D.2109-  Eden , the One-eyed.cue
    File B: C:\Users\Alice\Desktop\05 ola grande.mp3

    03:29:53 : Test started.
    03:31:57 : 00/01  100.0%
    03:32:00 : Trial reset.
    03:33:53 : 01/01  50.0%
    03:34:23 : 02/02  25.0%
    03:35:05 : 03/03  12.5%
    03:36:22 : 04/04  6.3%
    03:38:57 : 05/05  3.1%
    03:39:56 : 06/06  1.6%
    03:40:35 : 07/07  0.8%
    03:45:30 : 08/08  0.4%
    03:46:19 : 09/09  0.2%
    03:47:09 : 10/10  0.1%
    03:47:12 : Test finished.

     ----------
    Total: 10/11 (0.6%)
     
    320kbps, more Japanese rock.
    The loss of some treble energy in her vocals caused a lack of focus, especially after any high bits.
    You can see me screw up big time at the start, when I thought I was imagining obvious compression artifacts. There are none =P.
     
    foo_abx 1.3.4 report
    foobar2000 v1.1.8
    2011/09/30 09:22:24

    File A: D:\Music\sc\01 Kimi no Shiranai Monogatari.flac
    File B: C:\Users\Alice\Desktop\01 Kimi no Shiranai Monogatari.mp3

    09:22:24 : Test started.
    09:22:39 : 01/01  50.0%
    09:23:37 : 01/02  75.0%
    09:24:32 : 01/03  87.5%
    09:24:55 : 01/04  93.8%
    09:25:09 : 02/05  81.3%
    09:25:26 : 03/06  65.6%
    09:25:37 : 04/07  50.0%
    09:25:53 : 05/08  36.3%
    09:26:05 : 06/09  25.4%
    09:26:15 : 07/10  17.2%
    09:26:30 : 08/11  11.3%
    09:26:46 : 08/12  19.4%
    09:26:56 : 09/13  13.3%
    09:27:06 : 10/14  9.0%
    09:27:20 : 11/15  5.9%
    09:27:30 : 12/16  3.8%
    09:27:40 : 13/17  2.5%
    09:27:52 : 14/18  1.5%
    09:28:06 : 15/19  1.0%
    09:28:20 : 16/20  0.6%
    09:28:44 : 17/21  0.4%
    09:29:10 : 18/22  0.2%
    09:29:42 : 19/23  0.1%
    09:30:08 : 20/24  0.1%
    09:30:11 : Test finished.

     ----------
    Total: 20/24 (0.1%)
     
    320kbps, yep, more j-rock.
    This one was extremely difficult because it was on nodding terms with 'well recorded'.
    No compression artifact giveaways here, I had to fight to the death by locking onto the vocals. Same thing with the other tracks, it seems that the loss of some treble information gives female vocals less attack and focus.
    You can see me struggle at the start because I was trying to use the drums to check for obvious compression artifacts in the cymbals, and as you can see, no luck there =P.
     
    foo_abx 1.3.4 report
    foobar2000 v1.1.8
    2011/09/30 17:21:32

    File A: C:\Users\Alice\Music\Gazelle\Cover of Parallel Filter\Flac and secure\magnet (prkrock edition).flac
    File B: C:\Users\Alice\Desktop\05 magnet (prkrock edition).mp3

    17:21:32 : Test started.
    17:22:03 : 01/01  50.0%
    17:23:13 : 02/02  25.0%
    17:24:34 : 02/03  50.0%
    17:24:51 : 03/04  31.3%
    17:26:14 : 03/05  50.0%
    17:27:11 : 03/06  65.6%
    17:29:08 : 03/07  77.3%
    17:29:51 : 04/08  63.7%
    17:30:06 : 05/09  50.0%
    17:30:22 : 06/10  37.7%
    17:31:11 : 07/11  27.4%
    17:32:35 : 08/12  19.4%
    17:33:34 : 09/13  13.3%
    17:33:55 : 10/14  9.0%
    17:35:42 : 11/15  5.9%
    17:36:08 : 12/16  3.8%
    17:36:41 : 13/17  2.5%
    17:36:59 : 14/18  1.5%
    17:37:24 : 15/19  1.0%
    17:37:54 : 15/20  2.1%
    17:38:14 : 16/21  1.3%
    17:38:51 : 17/22  0.8%
    17:39:10 : 17/23  1.7%
    17:39:22 : 17/24  3.2%
    17:39:41 : 18/25  2.2%
    17:40:05 : 19/26  1.4%
    17:40:15 : 19/27  2.6%
    17:40:50 : 20/28  1.8%
    17:41:06 : 21/29  1.2%
    17:41:28 : 22/30  0.8%
    17:41:54 : 23/31  0.5%
    17:42:46 : 24/32  0.4%
    17:43:19 : 25/33  0.2%
    17:44:29 : 26/34  0.1%
    17:45:26 : 26/35  0.3%
    17:45:51 : 27/36  0.2%
    17:48:23 : 27/37  0.4%
    17:48:36 : 28/38  0.3%
    17:49:13 : 29/39  0.2%
    17:49:22 : 29/40  0.3%
    17:49:45 : 30/41  0.2%
    17:50:44 : 30/42  0.4%
    17:50:55 : 31/43  0.3%
    17:51:24 : 32/44  0.2%
    17:52:11 : 33/45  0.1%
    17:52:51 : 34/46  0.1%
    17:53:45 : 35/47  0.1%
    17:54:11 : 36/48  0.0%
    17:54:17 : Test finished.

     ----------
    Total: 36/48 (0.0%)
     
     
    Well, I'm all ears for some input and suggestions. I can try a few v0's vs lossless if anyone wants, but I have a feeling it might be worse than 320kbps.
     
    I'm not too sure how to interpret these results, there were no obvious compression artifacts like pre echo or anything like that, so I don't know if they're all killer tracks or not. It seems electronic music might be easier to tell with than rock music. I'm going to try some easy listening music to see if I'm still successful on music that isn't so 'busy'.
     
    Perhaps the guys who claim to hear differences between mp3 and lossless aren't so crazy after all, but I'd like to see some more ABX results from the people who say that they are able to hear the differences.
     
    I'm surprised I could pull it off, hey, maybe one day I'll be able to hear the differences between cables. [​IMG]
     
  2. Satellite_6
    I'm pretty sure almost anyone could do this with some concentration and the right mindset. It doesn't take magic ears or anything, it's just challenging - and the difference is there.
     
    Now I want to try to do this again. 
     
  3. ChipnDalebowl
    Also realize certain genres are more susceptible to being easily ABXed than others. I don't know what "Japanese Rock" means but if it has lots of cymbals and other high frequency sound effects or processing, I'm not surprised it can be done. Lossy audio traditionally has a very hard time with that stuff. Go to a different genre like soft rock or easy listening classical, and its much easier for the lossy encoder to do its job...and much harder to tell a difference--unless your goal is to listen hard enough to find it. Most people don't care to find the difference for the sake of doing so and just want to enjoy the music. :)
     
  4. deadlylover
    Quote:
     
    Yep, since I've already established that it's possible to ABX 320kbps, I'm setting out to find any genre's/tracks that I have that are impossible to ABX.
     
    All samples are 320kbps cbr.
     
    foo_abx 1.3.4 report
    foobar2000 v1.1.8
    2011/10/01 01:16:51

    File A: F:\Music\amenoumi\08 環.flac
    File B: C:\Users\Alice\Desktop\08 環.mp3

    01:16:51 : Test started.
    01:17:06 : 01/01  50.0%
    01:17:14 : 02/02  25.0%
    01:17:27 : 03/03  12.5%
    01:17:40 : 04/04  6.3%
    01:18:01 : 05/05  3.1%
    01:18:11 : 06/06  1.6%
    01:18:19 : 07/07  0.8%
    01:18:27 : 08/08  0.4%
    01:18:42 : 09/09  0.2%
    01:19:09 : 10/10  0.1%
    01:19:12 : Test finished.

     ----------
    Total: 10/10 (0.1%)
     
    Acoustic guitar lost a fair amount of impact from the loss of the higher frequency harmonics?
    Same deal with the vocals as well.
     
    foo_abx 1.3.4 report
    foobar2000 v1.1.8
    2011/09/30 21:14:25

    File A: F:\Music\Pizzicato Strings - Equinox -Limited Time-\Pizzicato Strings - Equinox -Limited Time-.cue
    File B: C:\Users\Alice\Desktop\02 ミニチュア模型.mp3

    21:14:25 : Test started.
    21:14:32 : 01/01  50.0%
    21:14:37 : 02/02  25.0%
    21:14:42 : 03/03  12.5%
    21:14:47 : 04/04  6.3%
    21:14:53 : 05/05  3.1%
    21:15:00 : 06/06  1.6%
    21:15:05 : 07/07  0.8%
    21:15:10 : 08/08  0.4%
    21:15:15 : 09/09  0.2%
    21:15:21 : 10/10  0.1%
    21:15:27 : 11/11  0.0%
    21:15:34 : 12/12  0.0%
    21:15:36 : Test finished.

     ----------
    Total: 12/12 (0.0%)
     
    Vocals were recorded with an insane amount of high frequency energy, so that one was a piece of cake.
     
    foo_abx 1.3.4 report
    foobar2000 v1.1.8
    2011/09/30 21:17:42

    File A: F:\Music\amenoumi\05 マリンスノー.flac
    File B: C:\Users\Alice\Desktop\05 マリンスノー.mp3

    21:17:42 : Test started.
    21:18:09 : 01/01  50.0%
    21:18:45 : 02/02  25.0%
    21:19:03 : 02/03  50.0%
    21:19:36 : 03/04  31.3%
    21:19:45 : 04/05  18.8%
    21:20:08 : 05/06  10.9%
    21:20:26 : 05/07  22.7%
    21:21:26 : 06/08  14.5%
    21:21:43 : 07/09  9.0%
    21:22:20 : 07/10  17.2%
    21:22:43 : 08/11  11.3%
    21:23:01 : 09/12  7.3%
    21:23:39 : 10/13  4.6%
    21:24:11 : 11/14  2.9%
    21:24:31 : 12/15  1.8%
    21:24:42 : 13/16  1.1%
    21:25:19 : 14/17  0.6%
    21:25:44 : 15/18  0.4%
    21:26:37 : 15/19  1.0%
    21:26:47 : 15/20  2.1%
    21:26:56 : 15/21  3.9%
    21:27:38 : 16/22  2.6%
    21:27:47 : 17/23  1.7%
    21:28:01 : 17/24  3.2%
    21:28:11 : 18/25  2.2%
    21:28:31 : 18/26  3.8%
    21:29:06 : 19/27  2.6%
    21:30:29 : 19/28  4.4%
    21:30:44 : 20/29  3.1%
    21:30:51 : 20/30  4.9%
    21:31:05 : 21/31  3.5%
    21:32:00 : 22/32  2.5%
    21:32:11 : 23/33  1.8%
    21:32:49 : 23/34  2.9%
    21:33:36 : 24/35  2.0%
    21:34:00 : 24/36  3.3%
    21:34:15 : 25/37  2.4%
    21:35:01 : 26/38  1.7%
    21:35:12 : 27/39  1.2%
    21:36:11 : 28/40  0.8%
    21:37:59 : 29/41  0.6%
    21:38:37 : 30/42  0.4%
    21:39:09 : 31/43  0.3%
    21:39:31 : 32/44  0.2%
    21:40:22 : 32/45  0.3%
    21:40:42 : 33/46  0.2%
    21:41:26 : 34/47  0.2%
    21:41:58 : 34/48  0.3%
    21:42:38 : 34/49  0.5%
    21:43:05 : 35/50  0.3%
    21:46:04 : 36/51  0.2%
    21:46:39 : 37/52  0.2%
    21:47:15 : 38/53  0.1%
    21:48:51 : 39/54  0.1%
    21:49:19 : 39/55  0.1%
    21:50:23 : 40/56  0.1%
    21:52:08 : 40/57  0.2%
    21:52:14 : 41/58  0.1%
    21:52:24 : 41/59  0.2%
    21:53:48 : 42/60  0.1%
    21:54:02 : Test finished.

     ----------
    Total: 42/60 (0.1%)
     
    This one was difficult, I don't even know if this result is a pass or not.
    Fatigue started kick in the last 10 minutes, I'll have to revisit this track later, but I'll call this a semi-positive result =P.
     
     
     
    I'm starting to suspect that ANYTHING with female vocals in it will have a high probability for it to be distinguished. I need to test a few more easy listening tracks, but I don't think the result will change as I'm not hunting for the obvious compression artifacts like pre echo or wishy washy cymbals, I'm really only listening for impact/dynamics and focus/presence in the vocals.
     
    I personally don't give a toss about the differences in normal listening, as most of my library is in mp3 anyway. The differences are too small to bother with (really, they are), but I just want to dispel the myth that it's "impossible" to distinguish between high bitrate mp3 and lossless. That, and confirm that this forum isn't full of crazy people =P.
     
  5. Willakan
    Nice! Whilst I would imagine that the STAX rig probably helps, this seems to show you have very nice hearing.
    I don't believe anyone said on that thread that it is always impossible (likely impossible for most people, yes) but as we can see there are those with the transducers and the ears who can genuinely distinguish differences which the majority of us can't (will test myself once I have put my new rig together).
     
     
    NB: The first person to somehow twist the latter part of that sentence into an argument about hearing differences in cables should curl up and die.
     
  6. RexAeterna

    cables are night and day man, night and day differences. just haters here.
     
  7. mikeaj
    Thanks for going way out of your way to do all these tests.

    For me it is difficult or impossible to ABX -V0 (or 320kbps) with lossless on most tracks, but I've had a lack of very high frequency hearing (say like a half octave or a bit more) since I was a kid. Whatever it is I'm hearing, it's not the lowpass filtering that's discernibly different, that's for sure.


    If you ever get motivated again for more ABX, could I suggest trying a different lossy encoder rather than different music? e.g. whatever the state-of-the-art Vorbis (or AAC, etc.) is.

    Best Vorbis encoder should be 2.87 (which uses aoTuV 6.03):


    Try something like -q8 to -q10. There was at least one time when LAME 3.98r -V0 was not transparent but around Vorbis -q8 was (to me and somebody else), for only about 10-30 kbps higher average bitrate.
     
  8. deadlylover
    Quote:
    [​IMG]
     
    I was mainly referring to the 'hardcore high bitrate crusade', whom say it's impossible and what not.
     
    I think the Stax is helping me big time, as there is less impact/muddiness from being a stat, letting me focus a little bit better.
     
    Quote:
     
    No probs, thanks for the input. ^^
     
    I tried the aoTuV beta 6.03, it was the latest on the aoTuV website, results are below.
     
    foo_abx 1.3.4 report
    foobar2000 v1.1.8
    2011/10/01 04:27:57

    File A: D:\Music\sc\01 Kimi no Shiranai Monogatari.flac
    File B: C:\Users\Alice\Desktop\01 Kimi no Shiranai Monogatari.ogg

    04:27:57 : Test started.
    04:30:02 : 01/01  50.0%
    04:30:21 : 01/02  75.0%
    04:30:58 : 01/03  87.5%
    04:31:27 : 02/04  68.8%
    04:32:00 : 03/05  50.0%
    04:32:09 : 04/06  34.4%
    04:32:16 : 05/07  22.7%
    04:32:25 : 06/08  14.5%
    04:32:30 : 07/09  9.0%
    04:32:35 : 08/10  5.5%
    04:32:43 : 09/11  3.3%
    04:32:56 : 10/12  1.9%
    04:33:02 : 11/13  1.1%
    04:33:11 : 12/14  0.6%
    04:33:25 : 13/15  0.4%
    04:33:33 : 13/16  1.1%
    04:33:44 : 14/17  0.6%
    04:33:53 : 15/18  0.4%
    04:34:04 : 16/19  0.2%
    04:34:14 : 17/20  0.1%
    04:34:17 : Test finished.

     ----------
    Total: 17/20 (0.1%)
     
    I think it is better than 320kbps lame, I felt that there was less treble information lost, but it was still lacking slightly in impact/attack. For that particular track, on the mp3 version I used the off focus vocals to distinguish between them (was missing a fair bit of treble), but that trick didn't work for the ogg file (evidence in the start of the test =P).
     
    Sometimes I see the bitrate spike to ~380-400, which probably explains why there is less treble lost in the busy bits. But I do feel that there is more of a impact/dynamics difference vs the mp3, which is why I was able to blaze through the test after I discovered that.

    The setting -q10 starts at like ~500kbps, that doesn't seem worth it considering the FLAC is only 1056kbps. So I think I'll stick to -q9 setting for further tests which should put it at about ~320kbps, which is similar to the mp3.
     
  9. mikeaj
    Yeah, I agree at some level like -q10 the bitrate is so high, may as well go with some lossless. Interesting thoughts on the difference between the codecs. I think mp3 is just more likely than more modern formats to produce noticeable encoding artifacts like pre-echo, sizzles, and so on, at the lower bitrates, and not just some slight perceived loss in quality or information.

    IMHO 1056 kbps FLAC (max compression level 8?)** is a high bitrate, meaning that the tracks are very busy. I mean, 16-bit stereo 44.1 kHz is 1411.2 kbps. A symphony recording may be like 400-650 kbps FLAC. I do have some electronic and synthesized music that is more like 1050 kbps when in FLAC though.

    ** FYI those who don't know, FLAC is lossless regardless, just has to do with encoding speed, so a "default" compression level of 4 or lower just encodes faster and produces a slightly larger file size

    As a complete guess, I think the dynamic range compression, maybe some of the processing, lots of percussive or noiselike sounds, and so on makes those tracks harder to compress well? I guess if the entropy is higher in some sense, that will make it tough for the lossy encoders as well.

    Out of curiosity, what was the bitrate on the easy listening, vocals + symphony, and other music you listened to? I'm going to guess it was way below 1056 kbps.
     
  10. Head Injury
    Quote:

    I think this is true. After all, the easiest sound to compress is silence. Dynamically compressed tracks like those don't have much silence.
     
    Awesome results, DeadlyLover. Couple questions:
    1. What volume are you listening at?
    2. How clear are the differences, overall? As in, would you notice if someone replaced all your FLACs with MP3s overnight?
    3. Who said it was "impossible" to distinguish between them?
     
  11. deadlylover
    Quote:
     
    They're roughly ~550 - 850, and everywhere in between for most of my easy listening lossless tracks. I'm not good at classifying genres, but they're the easiest stuff I have to listen to =P.
     
    The ~1056 (lvl 8) file is brick-walled to hell, so that probably explains the high bitrate as there is barely a moment of silence/quiet bits.
     
    Quote:
     
    I'm not listening very loudly most of the time, I'm guessing 65-70db or so, I really need to get my hands on a db meter. For me, the louder I listen it actually gets harder for me to discern differences (I'm not hunting for obvious artifacts), as when I switch A / B the switch appears more sudden/amplified(?) and it seems like there is more attack. I think it has something to do with that something reflex that makes things softer when you're listening too loudly.

    The differences were small, I really don't think I would notice if someone swapped all my tracks. That being said though, it kind of depends on the track. I'm not going to miss/detect any lost impact/dynamics but I think I might notice any out of focus vocals if there are any, as I really just tunnel vision the vocals when listening to music. I rarely 'just' listen to music, I'm always doing something else as well, so realistically, I won't be able to notice if somebody replaced my files. But I think it's very possible for somebody else to notice it if they tend to focus while listening to their well known tracks, the differences are clear enough for that.
     
    You can answer the third question yourself, there's always a couple of people who say something like that in any random mp3 vs lossless thread.
     
    Some more results, the track is some kind of electronica with vocals, it's not very 'busy' in the parts I used for the comparison. (yeah im terrible at genre's)
     
    foo_abx 1.3.4 report
    foobar2000 v1.1.8
    2011/10/01 15:51:57

    File A: C:\Users\Alice\Desktop\01 Hysteresis.mp3
    File B: C:\Users\Alice\Music\Electrocutica - Hysteresis\Hysteresis.flac

    15:51:57 : Test started.
    15:52:26 : 01/01  50.0%
    15:52:31 : 02/02  25.0%
    15:52:54 : 03/03  12.5%
    15:53:03 : 03/04  31.3%
    15:53:21 : 04/05  18.8%
    15:53:37 : 05/06  10.9%
    15:53:47 : 06/07  6.3%
    15:54:07 : 07/08  3.5%
    15:54:19 : 08/09  2.0%
    15:54:28 : 09/10  1.1%
    15:54:36 : 10/11  0.6%
    15:54:57 : 11/12  0.3%
    15:55:10 : 12/13  0.2%
    15:55:29 : 13/14  0.1%
    15:55:31 : Test finished.

     ----------
    Total: 13/14 (0.1%)
     
    foo_abx 1.3.4 report
    foobar2000 v1.1.8
    2011/10/01 15:55:52

    File A: C:\Users\Alice\Desktop\01 Hysteresis.ogg
    File B: C:\Users\Alice\Music\Electrocutica - Hysteresis\Hysteresis.flac

    15:55:52 : Test started.
    15:56:44 : 01/01  50.0%
    15:56:52 : 02/02  25.0%
    15:56:59 : 03/03  12.5%
    15:57:06 : 04/04  6.3%
    15:57:16 : 05/05  3.1%
    15:57:34 : 06/06  1.6%
    15:57:44 : 07/07  0.8%
    15:58:04 : 08/08  0.4%
    15:58:12 : 09/09  0.2%
    15:58:25 : 10/10  0.1%
    15:58:30 : Test finished.

     ----------
    Total: 10/10 (0.1%)
     
    Interestingly, I was unable to ABX mp3 vs aoTuV b6.03 for that particular track. I'll have to try it again with something else, but I'm not too surprised that I couldn't do it, considering that it seems much more unlikely to be able to discern differences between lossy formats. I think the only way that it might be possible is if only one of the files are transparent. If neither are transparent, then I suppose I'll need golden ears and/or an SR-009 and/or proficiency in The Force to pick out the differences in psycho acoustic compression.
     
    I'll have a go later with a very sibilant track, to see if the CBR mp3 removes some of the higher frequencies more so than aotuv's VBR.
     
  12. wnmnkh Contributor
    Interesting and good results.
     
     
    Remind all of you folks, even with best encoder (LAME) mp3 is not without fault. There are still certain genre or tracks that it will fail (see Hydrogenaudio for more examples). It is just that most of time they are virtually transparent enough after V5, but it does not mean all of them are good enough. :p
     
  13. Willakan
    @Deadlylover:
     
    Whilst I realise it might be inconvenient, I would be very grateful if you could repeat the test with one of the "harder" MP3 tracks using some cheaper headphones - just interested to see how big a role the supremely revealing Stax played in it.
     
  14. Brooko Contributor
    Nice deadlylover!
     
    I remember last time you and I discussed this, you called yourself 'tin-ears'.  Apparently not the case!  If you ever get the chance, I'd love to see if it's similar results with 256aac.  It's supposed to be as good if not better than 320 mp3 - have a feeling it won't make a difference to you though.
     
    I'll have to try again some time with my DT880 and HD600 - but I'm pretty sure it will be same as last time (I am a true tin-ears).  The good thing is that this has its advantages (get a lot of music on a Touch 32Gb at 256aac).
     
    Nice job anyway [​IMG]
     
  15. deadlylover
    Quote:
     
    That is a fantastic idea!, I will do some further testing later when nobody is home. (I want to use my speakers as well, and kidnap an AD900 off family)
    We may or may not be able to confirm if gear matters at all, which would lead to an end of the "if I use xxx is mp3 good enough". On the other hand, we might also be able to conclude that no matter how good your gear is, mp3 vs lossless does not matter one bit, so no need for folks to break their balls over re-ripping their library when they go into the high end. Well, they can still break their balls, but at least we can lol at them.
     
    Here are some results that I've tried just then:
    First up are the O2's, then my SR-202, then my computer speakers. I reset the test when I did the switches.
     
    foo_abx 1.3.4 report
    foobar2000 v1.1.8
    2011/10/01 18:22:06

    File A: C:\Users\Alice\Music\Gazelle\Cover of Parallel Filter\Flac and secure\magnet (prkrock edition).flac
    File B: C:\Users\Alice\Desktop\05 magnet (prkrock edition) .mp3

    18:22:06 : Test started.
    18:22:31 : 01/01  50.0%
    18:27:10 : 02/02  25.0%
    18:27:24 : 03/03  12.5%
    18:28:20 : 04/04  6.3%
    18:28:45 : 05/05  3.1%
    18:29:13 : 06/06  1.6%
    18:29:24 : 07/07  0.8%
    18:29:30 : 07/08  3.5%
    18:29:57 : 08/09  2.0%
    18:30:14 : 09/10  1.1%
    18:30:27 : 10/11  0.6%
    18:30:52 : 11/12  0.3%
    18:31:05 : 12/13  0.2%
    18:31:22 : 13/14  0.1%
    18:31:30 : Trial reset.
    18:32:08 : 01/01  50.0%
    18:32:15 : 02/02  25.0%
    18:32:23 : 03/03  12.5%
    18:32:40 : 04/04  6.3%
    18:32:57 : 05/05  3.1%
    18:33:14 : 05/06  10.9%
    18:33:30 : 06/07  6.3%
    18:33:44 : 07/08  3.5%
    18:33:51 : 08/09  2.0%
    18:34:05 : 09/10  1.1%
    18:34:38 : 10/11  0.6%
    18:34:53 : 11/12  0.3%
    18:35:17 : 12/13  0.2%
    18:35:31 : 13/14  0.1%
    18:35:34 : Trial reset.
    18:36:29 : 01/01  50.0%
    18:37:18 : 01/02  75.0%
    18:37:25 : 02/03  50.0%
    18:37:29 : 03/04  31.3%
    18:37:49 : 04/05  18.8%
    18:38:10 : 05/06  10.9%
    18:38:31 : 06/07  6.3%
    18:38:37 : 07/08  3.5%
    18:38:42 : 08/09  2.0%
    18:38:47 : 08/10  5.5%
    18:38:55 : 09/11  3.3%
    18:39:08 : 10/12  1.9%
    18:39:25 : 11/13  1.1%
    18:39:31 : Trial reset.
    18:42:32 : Test finished.

     ----------
    Total: 37/41 (0.0%)
     
    It would appear that gear does not matter one bit, but that may be because I know what to listen for. I'm going to do some further testing starting the other way around on a new track, to see if gear really does bring things into weapons range or not.
    Subjectively, the speakers were only a little bit harder than with the stats, but that may have been because I didn't want to turn the volume too loudly as my music is terribly embarrassing. I'll chime in with some more results in a day or two, whenever the house is free.
     
    Quote:

    Heheh what can I say, people change. [​IMG]
    Maybe I just needed some more confidence.
     
    I will try out 256aac and see if it is comparable or better than 320mp3. I have a feeling it still wont be transparent, but I'll ABX it against 320 to see if there are any differences. And again, if I had an ipod full of mp3's or whatever, I still wouldn't flush the whole thing out with lossless, the differences are WAY too small to bother with, but then again, I'm not a picky person.
     
2 3 4 5 6 7 8
Next
 
Last

Share This Page