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

ABX Test of 320kbps VS FLAC - Results

Discussion in 'Sound Science' started by berlioz, Jun 27, 2009.
1 2 3 4
  1. nick_charles Contributor

    Originally Posted by RonaldDumsfeld /img/forum/go_quote.gif
    I cheered up a bit when I found out that no one else can either. Not in a properly conducted test. So I believe. If anyone knows to the contrary I'd be pleased to hear.

    Generally I cannot tell MP3 VBR 0 from lossless. However I was able to DBT them on one specific track to a P < 0.05 level (14/17 or something like that) , on the track in question the original file was heavily saturated , peaking frequently at 0db and the endoding appeared to add a small extra amount of distortion, the telling them apart was really hard and I had to focus on one small (5 second ) segment and go back and forwards between the two of them many many times. Nevertheless I managed it. In casual listening I would have been unable to tell them apart only with extreme concentration and mind-numbing repetition and only with one tricky sample...
  2. Young Spade
    ^Wow alright man thanks for sharing [​IMG]
    I really should AB some files one day. All of the current stuff I get now is ripped to FLAC and used so I'm probably just gonna wait until I'm forced to convert and see what happens [​IMG]
  3. Chef
    I know you probably just mistyped, but please know the difference between AB and ABX testing. The former is not useful in this context.

    I think you'll find if you're ABX testing a random song for the first time, you'll probably even have trouble comparing 128 vs flac. You will be surprised how hard you need to listen, and how much experience it takes knowing where to listen. There are some rare files where it is very obvious, but on the whole lossy encodings are very good and you will find yourself confused.
  4. Young Spade
    ^Thanks and yea I meant ABX [​IMG] It was getting a little too late for me [​IMG]
  5. Horken
    That describes me pretty well! I just tried my first ABX test using foobar2000, my desktop PC (Gigabyte motherboard with good audio), and my Etymotic HF5's. I took some source flac files and converted them to MP3 with various Lame presets. Then I ran them through the foo_abx.
    At first, I tried using quality V2 with Radiohead's "15 Step" from In Rainbows. It's a very noisy track with a lot of range. It came out to an average of 187 kbps.  I didn't even bother making any guesses because I couldn't find anything by which to differentiate A from B, and I tried very hard in a number of passages. So, I tried V5 instead (avg 139 kbps). Still nothing. At this point I really wanted to hear a difference so I made a V8 version (avg 98 kbps) and tried again. OK, this time it was obvious right at the start and I got 100% correct after 5 guesses and stopped because it was so obvious.
    Then I tried "Satellite Mind" by Metric. It's dense, but not noisy like the last track. I figured it would be harder to tell the difference, and it was, but I had the exact same results: I couldn't tell the difference until V8, but then it was 100%. However, it wasn't nearly as obvious. I had to listen to a couple passages and could only really hear a difference in the cymbals.
    I thought it was going to be easier to tell the difference. Boy was I wrong! Lame is a really good encoder, and I have 40 year old ears. Despite having very accurate headphones, which are super crisp and clear, it was a struggle to find any difference until V8!
  6. Eternal Schism
    I've used the same method on Foobar myself. I've done it with Reflections by Tool (the early drumming parts), Helms Deep from the Lord of the Rings Trilogy Soundtrack, Money by Pink Floyd and Bust a Move (quick synth arpeggios) by Infected Mushroom. I couldn't reliably discern the differences between MP3 320 or Flac files using a HD800 powered by a NFB 28 (S/E output).
1 2 3 4

Share This Page