[arnyk]

I didn't say they judged based on them; more that cats have curiosity.

 
I think that curiosity is great. Given that my formal education in audio and engineering ended over 40 years ago, just everything relevant that I know today is the result of curiosity.

 

[nick_charles]

  I found an interesting article with samples of what is removed by MP3 compression of a song.
 
http://theghostinthemp3.com/theghostinthemp3.html
 
 
What are your thoughts?

 
Putting on my reviewers hat and referring to the published paper rather than the article itself I have to say it is a pretty bad paper on a number of levels, but it is not for a hard core scientific conference but a general purpose composition/tech conference and is a position paper i.e there is no real hypothesis being tested, no real research question and it should be judged  as such i.e does it raise interesting questions.
 
However it does have several weaknesses first it is highly opinionated and where it could use actual research it does not do so. For instance in the opening paragraph it totally fails to reference the long research on masking necessary for MP3 development presenting MP3 listening tests as based solely on what some old European guys wanted to listen to, this is highly misleading
 
It contains a number of material inaccuracies, it berates the iPod yet we know that the iPod even the gen 1 iPod has GOOD measured performance. It presents as fact opinions about MP3 listeners and is very light on the references. The author also misunderstands his references,  he attributes "MP3's create audible artifacts" to Pras et al. - in fact in Pras et al the authors concede that there were no significant differences in the perceptions of 256/320 Kbps MP3 and CD nevertheless they end with the general statement that

Mp3 compression does introduce audible artifacts.

Which is highly misleading as clearly the artifacts were not audible for some samples but in the abstract they alter this to "potentially audible artifacts" 
 
 
Also gems of overstatement such as 
 
Quote:

Many listeners today listen exclusively to MP3 files, even in settings where the gains from a higher fidelity format would be clearly perceptible

I listen to high bitrate MP3 in a quiet home setting with a competent outboard DAC and decent headphone amp and good headphones and I have DBT'd high bitrate MP3 and WAV numerous times and almost never found a clearly perceptible gain and many other members here have mirrored my experience
 
and so on.....
 
 
But it is a student paper and you have to cut them some slack as the PhD process is an apprenticeship, ****%$%%$ !! when I think about some of my early papers that got published,  lucky just isn't in it 

 

[castleofargh]

the all idea of mp3 was to use masking to remove what most likely wouldn't be heard anyway. removing the masking part obviously will make the otherwise masked parts audible. not sure what we gain from doing that? 

 

[Currawong]

That's what I was thinking at the start. For me, thinking back to that test of LAME 128 vs. lossless that was posted, I remember picking the difference at a particular point with a particular instrument (though I don't remember what instrument). It would be interesting to get the "ghost" of that mp3 and listen at the part where I picked the difference. It might be interesting, for curiosity's sake to note the loudness levels of both the original and the ghost at that point too.
 
Quote:

  Do you not agree that any reasonable discussion of the science must start with the true facts of the situation?

 
Yes, but not mixed with ad hominems, weasel words and other propaganda techniques that have no place alongside them. 

 

[arnyk]

   
Yes, but not mixed with ad hominems, weasel words and other propaganda techniques that have no place alongside them. 

 
 
This article while not exactly error free seems to cover similar territory 
 
http://faculty.georgetown.edu/irvinem/theory/Sterne-mp3-2006.pdf
 
and do a better job of avoiding those things that are so obnoxious in 
 
http://theghostinthemp3.com/theghostinthemp3.html
 
If you are fascinated with difference testing this software does it about the best:
 
http://www.libinst.com/Audio%20DiffMaker.htm

 

[Terja]

Just finished reading the synopsis on the link Currawong posted but haven't read the full conference paper just yet. I think this is a very worthy study if approached from a neutral perspective. There tends to be a lot of defensive posturing around this subject that I find quite unusual for such a basic concept. Yes MP3s and other lossy codecs are lossy. The question of whether what is "lost" or discarded or removed can be recomposed and viewed (heard) as an artifact of the compression methodology is fascinating. One should not assume such an artifact would be the same as the music, but it would be something, and of course the question is - "What Is It?" I for one would be interested in finding out, without any preconception that it would be music - it would simply be something - but what?

 

[nick_charles]

  Just finished reading the synopsis on the link Currawong posted but haven't read the full conference paper just yet. I think this is a very worthy study if approached from a neutral perspective. There tends to be a lot of defensive posturing around this subject that I find quite unusual for such a basic concept. Yes MP3s and other lossy codecs are lossy. The question of whether what is "lost" or discarded or removed can be recomposed and viewed (heard) as an artifact of the compression methodology is fascinating. One should not assume such an artifact would be the same as the music, but it would be something, and of course the question is - "What Is It?" I for one would be interested in finding out, without any preconception that it would be music - it would simply be something - but what?

 
 
AudioDiffmaker, if you can get it work properly will extract a difference file so you can take a wav make an mp3 from it then create a new wav from the mp3 run AudioDiffmaker and the result will be "the ghost" you are after, note , I've used many different versions of AD and to date never got it to work properly i.e it can create a difference file with content when the two files compared are the same...but in principle ......it...... could.......... work

 

[Terja]

So if I had a wave file of a track and a compressed version of the same track, AudioDiffmaker would be able to compare the two and extract the difference into an mp3 file, is that right?

 

[Terja]

Oh .. it sounds like AudioDiffmaker requires for both comparison files to be the same format, thus the need to convert the Mp3 back into wave.

 

[nick_charles]

  So if I had a wave file of a track and a compressed version of the same track, AudioDiffmaker would be able to compare the two and extract the difference into an mp3 file, is that right?

 
 
That is how it is supposed to work, I have had issues with false positives, but it is definitely worth a go ! - the files both need to be in the same format (iirc) 

 

[Terja]

It would have been a better program if it was able to compare native file formats. Reencoding an Mp3 back into wave could introduce some other artifacts into the equation, essentially skewing the results, but it's definitely worth checking out.

 

[RRod]

You can also do a simple workflow via lame and SoX:
 
#encode and decode with lame
lame -b 320k original.wav - | lame --decode --mp3input - compressed.wav
#make the difference file with SoX
sox -m original.wav -v -1 compressed.wav difference.wav

 

[RRod]

  It would have been a better program if it was able to compare native file formats. Reencoding an Mp3 back into wave could introduce some other artifacts into the equation, essentially skewing the results, but it's definitely worth checking out.

 
You can't natively compare mp3 to wav; completely different formats. The mp3 has to be decoded back to PCM format before it goes to the DAC anyway.

 

[Terja]

In Foobar ..? It's been a while since I played around with encoding-decoding strings.

 

[RRod]

  In Foobar ..? It's been a while since I played around with encoding-decoding strings.

 
Foobar would have to decode the mp3 before doing the comparison as well. It would be like my giving you a text file and a zipped version of the text file, and asking you to compare them without unzipping the latter.