[christian u]

If anybody want to set up their own test, these excellent recorded files are recommended.
 

  Free Tracks Format Comparison

Here is a zip file containing samples of 2 tracks in 4 different formats.

A: 96/24 WAV
B: 96/24 FLAC
C: 16/44 WAV (CD)
D: 320kbps MP3

All the different formats have the same source file 96/24 WAV (Studio Master).

We used WAVELAB for the conversion.

When you compare the files start with the lowest resolution: D (MP3 320 kbps) and move on up through example C and B ending with A.

Be careful: If you start with A, and move down through B and C ending with D, your mind will remember the ''Blueprint'' of the higher resolution file, making it difficult to hear the difference even when finally listening to the MP3 file. Don't be frustrated if you can't hear a difference at first. Hearing is as individual as taste but hearing is also something which can be acquired, like the taste of good wine.

1. BATIK - The Defeat

A: 96/24 WAV
B: 96/24 FLAC
C: 16/44 WAV (CD)
D: 320kbps MP3

2. Carmen Gomes Inc - A Thousand Shades of Blue

A: 96/24 WAV
B: 96/24 FLAC
C: 16/44 WAV (CD)
D: 320kbps MP3

 

http://www.soundliaison.com/
 

 

[Baxide]

Jeez, I just tried these with my different headphones. With the HD800 and D7000 it was more obvious due to the delivery differences of the lower frequencies. But on my cheaper cans from the HD650 downwards it was more guess work than factual. 
I used Foobar and just kept clicking on the NEXT track symbol whilst I had my eyes closed.

 

[cjl]

Well, based on a quick inspection in Audacity, it does at least appear that they were using the same master, so they aren't blatantly lying. Their analogy is completely wrong though - you could probably take a random person off the street, give them a 30 second talk on what the most obvious differences between two wines are, and then have them pass a blind test with flying colors. Telling two wines apart isn't hard, and last I checked, nobody is disputing that there are differences between a box of cheap wine and a top shelf bottle. The debate with wine comes due to the fact that people often rank their choices differently if they don't know which is which (in essence, being forced to rely purely on taste preferences). With audio, it's quite different - nobody has conclusively proved that an audible difference yet exists between 16/44 and 24/96. As for your test, Baxide, that's not really conclusive at all - you should really resample the 44/16 track to 96/24 (so there's no difference in loading time, no different sample rate sent to your soundcard/dac, and identical switching behavior regardless of which file you're coming from and going to), and then use the foobar abx plugin to compare it with the genuine 96/24 track. If you're claiming to be able to hear the difference between the 320kbps mp3 and the other three files, that's slightly more believable, but still fairly surprising, since 320kbps mp3 is pretty much transparent with the vast majority of samples (and from what I can tell, this didn't sound like a particularly hard-to-encode track). That does depend somewhat on the encoder used though, so it is possible that you can hear a difference. Again, though, the ideal way to compare it would be to convert the mp3 back to wav, and then resample it up to 96kHz and then compare in foobar with abx, for pretty much the same reasons as I gave above.

 

[Baxide]

Converting mp3 to WAV and upsampling it to 96kHz has already been discredited as a genuine form of decision making. The mp3 encoding algorithm is not an attempt to be accurate, but to conserve space without discarding too much info. You can't recover what has been dicarded when you convert mp3 back to WAV. Once it is gone, it is gone. Even upsampling won't restore the missing info.

 

[cjl]

  Converting mp3 to WAV and upsampling it to 96kHz has already been discredited as a genuine form of decision making. The mp3 encoding algorithm is not an attempt to be accurate, but to conserve space without discarding too much info. You can't recover what has been dicarded when you convert mp3 back to WAV. Once it is gone, it is gone. Even upsampling won't restore the missing info.

That's the point. Converting it to wav and upsampling to 96kHz makes it so that the sample rate sent to the dac and the filesize will be the same, so you couldn't pass an abx on the basis of a dac that behaves differently with different sample rates, poor resampling in the OS, or the slightly longer loading time when playing the much larger 24/96 wav file. I didn't suggest converting it to wav and resampling it for normal listening, I suggested it to get the most accurate abx test between the 320 mp3 and the 24/96 wav (so you would be trying to abx the native 24/96 file vs the 24/96 upsampled wav created from the 320 mp3).

 

[Makiah S]

  Converting mp3 to WAV and upsampling it to 96kHz has already been discredited as a genuine form of decision making. The mp3 encoding algorithm is not an attempt to be accurate, but to conserve space without discarding too much info. You can't recover what has been dicarded when you convert mp3 back to WAV. Once it is gone, it is gone. Even upsampling won't restore the missing info.

which is why Red Book was made standard, lossless quality is just that lossless, once you compress to mp3 your losing quality, the desired effect of mp3 is to conserve space. Now there are forms of lossy that aren't... as bad for quality, like ogg 

 

[castleofargh]

  Converting mp3 to WAV and upsampling it to 96kHz has already been discredited as a genuine form of decision making. The mp3 encoding algorithm is not an attempt to be accurate, but to conserve space without discarding too much info. You can't recover what has been dicarded when you convert mp3 back to WAV. Once it is gone, it is gone. Even upsampling won't restore the missing info.

 the purpose isn't to restore anything from the mp3 by going back to wave, but to make sure that any differences will come from the difference in data and not from the way your source is dealing with different extensions.
not so long ago I ended up with flac on my x3 because mp3 sounded weird. when I never noticed anything that mattered while playing mp3 on my sony or cowon. that's because the x3 itself had trouble dealing with mp3s. by going back to a common file extension you avoid that possible bias.
 
 
 
edit: cjl already said that sorry ^_^

 

[RonaldDumsfeld]

It's not really a genuine AB/X test if the format of the samples is known to the testees ahead of time.
 
In this example all the samples should be 24/96 FLAC. It would make no difference to the accuracy of the test and no one would be aware of which sample was which.

 

Be careful: If you start with A, and move down through B and C ending with D, your mind will remember the ''Blueprint'' of the higher resolution file
 
 

 
I believe this is called 'prompting the witness'.

 

[Baxide]

  It's not really a genuine AB/X test if the format of the samples is known to the testees ahead of time.
 
In this example all the samples should be 24/96 FLAC. It would make no difference to the accuracy of the test and no one would be aware of which sample was which.

 
 
I believe this is called 'prompting the witness'.

Exactly. Which is why it is a good idea to mix up the position of each file and shuffling the deck with random play.

 

[ab initio]

 
Exactly. Which is why it is a good idea to mix up the position of each file and shuffling the deck with random play.

foobar's (or any other) ABX testing plugin is the best way to determine whether or not one can actually hear a difference between any of the tracks. Simply shuffling the play list will not permit the short switching times necessary to overcome the short audio-memory that we (humans) have, nor will it prevent non-auditory cues from poisoning a fair and unbiased comparison.
 
Cheers

 

[Makiah S]

  foobar's (or any other) ABX testing plugin is the best way to determine whether or not one can actually hear a difference between any of the tracks. Simply shuffling the play list will not permit the short switching times necessary to overcome the short audio-memory that we (humans) have, nor will it prevent non-auditory cues from poisoning a fair and unbiased comparison.
 
Cheers

or just accept that objectively Flac is better than mp3 and get over it, now wheather you can hear the differance depends on the person, but all these morons who keep trying to fight the Red Book Standard... get old. Objectively they have no ground to stand on, Subjectively it's a pointless argument since well that's the nature of subjectivism ;3

 

[cjl]

  or just accept that objectively Flac is better than mp3 and get over it, now wheather you can hear the differance depends on the person, but all these morons who keep trying to fight the Red Book Standard... get old. Objectively they have no ground to stand on, Subjectively it's a pointless argument since well that's the nature of subjectivism ;3

Ahh, but how is saying "accept objectively that FLAC is better than (well-encoded 320kbps) mp3" any different from saying "accept objectively that 24/96 is better than 16/44"? In both cases, there is technically more information in the latter, but that isn't the point. The real question is whether there's an audible difference.

 

[Baxide]

  foobar's (or any other) ABX testing plugin is the best way to determine whether or not one can actually hear a difference between any of the tracks. Simply shuffling the play list will not permit the short switching times necessary to overcome the short audio-memory that we (humans) have, nor will it prevent non-auditory cues from poisoning a fair and unbiased comparison.

Audio memory?? What is that? It is not something that any of our five senses is capable of. You either hear something or you don't. Please don't come up with some new unscientific principle.
Foobar's ABX testing relies on a set of source code routines written by humans. The accuracy of those routines is questionable at best, and cannot be considered a definite answer to anything. A direct comparison of an unexpurgated audio sequence is far more reliable for anyone with good hearing and ability to pick up pitch, dynamic range, and frequency response. I for one am capable of that. It happens to be something that my livelihood depends on. And no doubt so do many others who also happen to be music lovers. So called audio memory won't cut it in my job for one. I would be fired on the spot.

 

[ab initio]

Audio memory?? What is that? It is not something that any of our five senses is capable of. You either hear something or you don't. Please don't come up with some new unscientific principle.
Foobar's ABX testing relies on a set of source code routines written by humans. The accuracy of those routines is questionable at best, and cannot be considered a definite answer to anything. A direct comparison of an unexpurgated audio sequence is far more reliable for anyone with good hearing and ability to pick up pitch, dynamic range, and frequency response. I for one am capable of that. It happens to be something that my livelihood depends on. And no doubt so do many others who also happen to be music lovers. So called audio memory won't cut it in my job for one. I would be fired on the spot.

Here you go: human auditory memory

Cheers

 

[goodvibes]

If you want to compare sounds, use ABX. If you want to compare music, listen to each for bit and see if you better appreciate the art.