kent williams a écrit :
Sean Deason was kind enough to send me a CD of the Model 500
'Starlight' mixes, which I'd already bought in MP3 form from Boomkat.
After the discussion yesterday about how MP3s are noticeable on a good
sound system, this seemed like a golden opportunity to challenge the
golden ears amongst you to a blindfold test of MP3 sonic quality.
http://www.cornwarning.com/m500-starlight-blindfold test.wav
How this was made:
1. I ripped the uncompressed digital audio from the CD.
2. I cut a short representative sample out of the track.
3. I saved it to an uncompressed WAV file, with no processing whatsoever
4. I saved it as an 320KBS fixed-bit-rate MP3 file. I don't know
which MP3 encoder Sony uses this days, but I'd guess they've licensed
the current Frauenhofer CODEC.
5. I loaded both the WAV and the MP3 samples back into Sound Forge,
and put them together in one file.
Here's the challenge -- can you tell which sample comes first?
There's a very tiny 'click' that will tell you where one sample ends
and the other begins.
Listen to it however you want, but if you load it in Ableton Live, set
all level faders to 0dB and make sure that warping is turned off for
this loop. According to the appendix of the Live Manual, with those
settings, the sample passes through without change.
If you listen in Winamp or ITunes, make sure that you turn off EQ and
effects, and set the volume fader to 100%.
If you don't think this is a fair track to try this on, feel free to
send me a sample from another record that you think would more clearly
show the difference in sonic quality between MP3 and Wav.
Link is not working.
I'll be trying this using high end headphones, however in my experience
WAV and 320K are indistinguishable (but I always buy WAVs).
And a test it is a bit meaningless, as if any guess it right there's 1/2
chance it's due to luck.
To do that kind of test proper you need to provide both separate WAVs
(without us knowing which is which), use an ABX program (such as WinABX)
and do at least 10 runs and post the results.
I did such a test once and failed it miserably. There's a few golden
ears who can pass it successfully: for mos of the population 30K is
considered transparent.
