Head-Fi.org › Forums › Equipment Forums › Sound Science › Blind Test - Take Now!
New Posts  All Forums:Forum Nav:

Blind Test - Take Now! - Page 3

post #31 of 88

B seems to be the lossless, the its in the gaps eh?

post #32 of 88
Quote:
Originally Posted by nick_charles View Post



 There is an easier way, in audacity just load the first few seconds into the spectrum analyzer, that is a dead give-away.

 


The spectrum analysis is a give-away for figuring which file is the 128 kbps MP3.  But I wasn't looking at the waveform to try to figure out which file was which.  I was looking for why there was some distortion at the 30 second mark.  Looks like the distortion may be due to the flat-topping of some of the peaks in that section.  Maybe.

post #33 of 88

I just noticed I could check the files.

 

One has this: blocksize=4096 type=LPC order=8

 

The other: blocksize=1152 type=FIXED order=2

post #34 of 88

The files were intentionally compressed with different FLAC compression settings so that it wouldn't be immediately obvious which one was originally lossy.  Otherwise, if both were compressed with the same settings the smaller file size would be the MP3 lossy version.  Using different random compression settings for each eliminates that potential give-away.

post #35 of 88
Thread Starter 

B is the lossless.

 

I can't tell the difference, but alot of you could.  I am impressed.

 

When I do get my music, I do try and shoot for 320 or lossless, as there really is not any reason to use 128.  Space is cheap.

 

To me it seems that headphone amps have the same effect.  The difference was so small if at all, I was never able to tell.

 

 

 

 

 

 

 

post #36 of 88

yay! I got it right! and that was with 20$ Philips I had lying around.

 

Anyway, cool test! you should definetly another one sometime soon!


Edited by Thecoolguy - 10/21/10 at 9:05pm
post #37 of 88
Quote:
Originally Posted by Ham Sandwich View Post

The files were intentionally compressed with different FLAC compression settings so that it wouldn't be immediately obvious which one was originally lossy.  Otherwise, if both were compressed with the same settings the smaller file size would be the MP3 lossy version.  Using different random compression settings for each eliminates that potential give-away.


I understand order=2 to increase the size, but I don't get why type=FIXED, I thought the approximation method give better accuracy when set to LPC, maybe Crazy*Carl or someone else could explain if type had an influence on the file size.

post #38 of 88
Quote:
Originally Posted by Crazy*Carl View Post

B is the lossless.

 

I can't tell the difference, but alot of you could.  I am impressed.

 

When I do get my music, I do try and shoot for 320 or lossless, as there really is not any reason to use 128.  Space is cheap.

 

To me it seems that headphone amps have the same effect.  The difference was so small if at all, I was never able to tell.

 

 

 

 

 

 

 



Thanks for putting this together. I have a hunch that most of us couldn't have decided so easily if the original mp3 file was 320 (or even 256) kbps.

post #39 of 88
Thread Starter 

I will do another soon with maybe 3 or 4 different file types.  64mp3 up to lossless.

 

Any specific requests for a well recorded track?


Edited by Crazy*Carl - 10/22/10 at 8:35am
post #40 of 88

One of the 3 here: http://www.mastersfromtheirday.com/

post #41 of 88

Quick back of the napkin calculation of odds of getting 14 out of 17 correct responses assuming null hypothesis (that responses on 2 choice test are purely random) says that the odds of this outcome is 0.0052.  This clears the P < 0.05 significance by a mile.  So assuming that all those 'b' responses were truly independent and not swayed by other responses, the conclusion to the hypothesis test of whether people can hear the difference between 128kbps and lossless files is - yes.

post #42 of 88
Thread Starter 
Quote:
Originally Posted by eucariote View Post

Quick back of the napkin calculation of odds of getting 14 out of 17 correct responses assuming null hypothesis (that responses on 2 choice test are purely random) says that the odds of this outcome is 0.0052.  This clears the P < 0.05 significance by a mile.  So assuming that all those 'b' responses were truly independent and not swayed by other responses, the conclusion to the hypothesis test of whether people can hear the difference between 128kbps and lossless files is - yes.



No that is not really the conclusion I get.  I think the conclusion is some people on some rigs can tell a difference.

 

You sound like a robot


Edited by Crazy*Carl - 10/22/10 at 6:15pm
post #43 of 88

 

You're welcome.  Of course the test was not perfectly conducted.  But if someone was curious a p value, I thought I'd furnish it.  This is after all the science forum.
post #44 of 88

Everyone should know statistics smily_headphones1.gif.

 

Nice test Carl. I could easily hear a difference on my rig and thought B was the lossless. I personally use 320kpbs to save space as I cannot regularly hear a difference between that and lossless. Do that test next.


Edited by MomijiTMO - 10/22/10 at 8:19pm
post #45 of 88

Carl,

just listened to the tracks before finding out the answer.  while not a night and day difference, i did prefer file B via my y2/B22/K702.  

 

thanks for putting the effort into this test and hipping me to some new music.

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Sound Science
Head-Fi.org › Forums › Equipment Forums › Sound Science › Blind Test - Take Now!