The jury is still out

Discussion in 'OT Technology' started by TenSteel, Apr 4, 2005.

  1. TenSteel

    TenSteel Ted Cruz suicide hotline OT Supporter

    Joined:
    Nov 11, 2001
    Messages:
    14,082
    Likes Received:
    544
    Location:
    San Diego
    I'm running my own comparisons against mp3 vs. ogg vs. aac. MP3 recorded w/ LAME 3.96, ogg with the latest vorbis tools, and the aac with iTunes. CD tracks were ripped with CDex directly to wav for a sourcefile.

    Ogg quality = 6 (about 192kbps)
    MP3 = 192 kbps w/ HQ algorithm
    AAC = 192 kbps, stereo

    All EQ's set flat, listening with Grado SR60 headphones straight from my soundcard.

    So far I can only tell that mp3 seems to lose a bit of treble information vs. the rest. AAC and Ogg seem to be very similar at this point. Update to come later.
     
  2. veonake

    veonake OnT poster, OT lurker

    Joined:
    Apr 11, 2004
    Messages:
    4,235
    Likes Received:
    0
    Location:
    NYC
    Are you listening to them seperately on your own or with a friend changing the tracks at random so you have no knowledge of what format you are listening to? Also, why did you choose 192kbps as your base? Shouldn't you go with the highest quality so there is no disparity between the formats? For instance, you set Ogg equal to 6, which you say is about 192.
     
  3. TenSteel

    TenSteel Ted Cruz suicide hotline OT Supporter

    Joined:
    Nov 11, 2001
    Messages:
    14,082
    Likes Received:
    544
    Location:
    San Diego
    I say "about 192" because ogg is a variable bitrate codec, and the filesize is roughly equivalent to 192kbps mp3 file at a 6.0 quality setting (varies from 1.0 - 10.0).

    I'm choosing 192 because that is the bitrate (or quality setting) that I would use to encode my cd collections.

    I'll have my friend come in and do some switching later this week.
     
  4. 04

    04 Guest

    I HIGHLY reccomend using a PC based ABX comparator program for determining if you can hear a difference between audio files.

    http://www.kikeg.arrakis.es/winabx/ I use winabx and it works fantastic.

    If the differences between two audio files are rather large, then its a lot easier for humans to tell if the difference exists or not. But when the differences are small, it becomes very difficult to reliably tell if you can tell a difference without doing multiple trial blind testing.

    All you have to do is decompress the files you've encoded back to wave and load them up in the program.
     
  5. TenSteel

    TenSteel Ted Cruz suicide hotline OT Supporter

    Joined:
    Nov 11, 2001
    Messages:
    14,082
    Likes Received:
    544
    Location:
    San Diego
    Alright, so I got the ABX program, it's really cool. I couldn't find any program that would convert an aac back to a wav file, I guess because of the DRM? :dunno:

    My methodology with the program is to take the original source wav file and compare it first to mp3 then to ogg to see if I can pick out the two. I will also compare the mp3 to the ogg file. So far, with Miles Davis' All Blues, I was able to pick out which was the mp3 file 3/4 times, in the ABXY mode.

    I'll post more updates. My next sample tracks will be:
    BT - Mercury and Solace (a relatively "busy" techno track with heavy bass drum and female vocals)
    Joe Satriani - Gnaaah (well-recorded solo distorded guitar)
    Soilwork - Nerve (this will be a tricky one, it's a screaming metal track :noes: )
    Greg Howard - Ojos Verdes (solo Chapman Stick track)

    I might have to throw out AAC from the tests if I can't find a decoder.
     
    Last edited: Apr 5, 2005
  6. 04

    04 Guest

    I would redo the miles davis comparison, as 3/4 is a tough result. With that few trials, its hard to say if there was a difference, or if it was due to random chance :hs: But if you're satisfied, then that's all that matters.

    When I made my comparisons a few years back, I would always do at least 10 trials to make sure I wasnt guessing :o
     
  7. TenSteel

    TenSteel Ted Cruz suicide hotline OT Supporter

    Joined:
    Nov 11, 2001
    Messages:
    14,082
    Likes Received:
    544
    Location:
    San Diego
    Yes, I plan on doing the Miles test again. That initial posting was just a "pre-test" to see what my initial impressions were. I did, however, bring out the Satriani track, comparing the mp3 and ogg files. I could pick out the different files 6/7 times.

    10 sounds like a fair number (at the rate I was going with the Satch track, I probably could have gotten 9/10 :o) ... I'll do all the trials-- wav vs mp3, wav vs ogg, mp3 vs ogg, 10 times and copy the results from the logfile.

    Fuck this is gonna take a while.
     
    Last edited: Apr 6, 2005

Share This Page