[MrLazyAnt]

Quote:

Yeah but this is about 320 mp3s and not (most probably) transcoded from lossy to lossy audio in youtube videos..

 
The point is moot, if chewy is right (which a post I just delivered argues against) the bass should not be the first thing to suffer even in youtube. And it still is what I based my, for the most part, correct assessment on in the DBT.
 
The long and the short of it is, trust your ears not your eyes.
 
EDIT: Just to be clear, I argued that bass sounds better in lossless, Chewy said that this shouldn't be true because bass sounds suffer less from lossy compression to which I answered with this:
 
http://www.head-fi.org/t/570621/flac-vs-320-mp3/270#post_9307565

 

[sonitus mirus]

Quote:

 
The long and the short of it is, trust your ears not your eyes.

 
This is exactly what an ABX test is all about.  Now walk the walk.

 

[MrLazyAnt]

Okay, so I've downloaded the ABX component thingymabob (terminology isn't my strong suit). How do I get it to work?

 

[xnor]

First you add a lossless file to a foobar2000 playlist, then right click - convert - ... - set output format to MP3 320 kbps.
 
Then add the mp3 to the playlist, select both files - right click - utilities - abx two tracks.

 

[xnor]

Quote:

The point is moot, if chewy is right (which a post I just delivered argues against) the bass should not be the first thing to suffer even in youtube. And it still is what I based my, for the most part, correct assessment on in the DBT.

Transcoding can do all sorts of bad things, especially if you do not know what the uploader actually uploaded. It could have been a 128 kbps MP3 transcoded to 128 kbps AAC (mp4).
 
You have to eliminate all those variables so that in fact you're testing lossless vs. 320 kbps MP3 and not something else (transcoding artifacts, volume differences, player differences ...).

 

[MrLazyAnt]

True. I shall do the ABX test in the coming days. I shall either feel endlessly righteous and post the results as victorious proof, and brag about it to any who will suffer me, or shamefacedly admit defeat.

 

[chewy4]

Quote:

True. I shall do the ABX test in the coming days. I shall either feel endlessly righteous and post the results as victorious proof, and brag about it to any who will suffer me, or shamefacedly admit defeat.

 
Haha, just remember that only 15/15 gets you bragging rights in most people's books, and no cherry picking results.
Quote:

From "Sound On Sound" Magazine
 
http://www.soundonsound.com/sos/apr12/articles/lost-in-translation.htm
 
"The MP3 format has a reputation for making bass and low-frequency content sound weak: that slammin’ bass line can easily lose its phatness! (See audio examples L and M referenced above.) Low frequencies are harder for DSP algorithms to analyse because their durations are long, and amplitude differences over the short analysis windows used by the encoders may only be slight — so the analysis system doesn’t get an entire cycle of a low frequency per analysis window. In some situations, the encoder will be presented with less than a half cycle of any frequency below 114Hz. The AAC format fares much better in bass resolution, and it is thus much more forgiving to the bass."
 
I hate to admit I get a disproportional amount of pleasure in being right about this.

That's certainly the first and only bit of logic I've seen explaining why bass would be effected by compression. But remember he is talking about 128kbps mp3. Not sure if it's completely true as I can't find anything else supporting it, but it could be very well be legit.
 
EDIT: OK so I could be thinking about this the wrong way but the analysis window for mp3 is 1,152 samples... How would this ever bring about a situation where it would only show half of a 114Hz cycle? If I'm understanding this correctly shouldn't a full cycle of a 20Hz sine wave be 1102 samples in? Hopefully someone who knows a lot about MP3 at a low level can explain this...

 

[sonitus mirus]

Quote:

True. I shall do the ABX test in the coming days. I shall either feel endlessly righteous and post the results as victorious proof, and brag about it to any who will suffer me, or shamefacedly admit defeat.

This is really only for you.  Don't think of it as a contest to win or fail.  Your results have no impact on my own discoveries.  I know that I cannot tell a difference.  I feel enlightened, but neither shamed nor victorious. 

 
Now my audio equipment quests are all about transparency, where before I was victimized by misinformation and incorrect assumptions.  Again, though, this applies only to me with any certainty, as I do not hear things with any other ears.

 

[MrLazyAnt]

Quote:

This is really only for you.  Don't think of it as a contest to win or fail.  Your results have no impact on my own discoveries.  I know that I cannot tell a difference.  I feel enlightened, but neither shamed nor victorious. 

 
Now my audio equipment quests are all about transparency, where before I was victimized by misinformation and incorrect assumptions.  Again, though, this applies only to me with any certainty, as I do not hear things with any other ears.

I was being (or at least trying to be) humorous in the whole victory/defeat thing. I'll still post the results when I get round to doing the testing. For kicks and giggles.

 

[MrLazyAnt]

And the results are in! I must admit it was much harder than I thought it would be, and in the lesser production I was pretty much guessing. But I am happy that my ears function well in the low-mid registers, and despite my sound-system being a far cry from hi-fi (ha! A rhyme! ='P ) I still got it more or less right.
 
 

 

[sonitus mirus]

Did you make sure that the files were volume matched, or did you simply convert the file and test?  What encoder were you using?  Why not test one file 15 times for statistical credibility?  Any way to upload 30 seconds of your files so that others might try a test for themselves?

 

[MrLazyAnt]

Quote:

Did you make sure that the files were volume matched, or did you simply convert the file and test?  What encoder were you using?  Why not test one file 15 times for statistical credibility?  Any way to upload 30 seconds of your files so that others might try a test for themselves?

I just converted and compared. I can't make attatchments, so I can't upload the files (sorry about that), I used the LAME converter on highest quality, and I figured 15 tests in total would be as good as 15 tests on the same file, the statistics remain the same I scored 10/10 on proffesionally produced peices, which I find solid enough, and 3/5 (non-conclusive I know) on the home-mastered track. Either way, I have had my views on the quality of 320kbps altered. I think more highly of it now than before, and the only reasons I shall keep using FLAC files is so that i know it is lossless (I'm silly that way) and that when I eventually do bring my set-up closer to something that can be called hi-fi I will be more likely to profit from the codec.
 
Either way, I've drawn my conclusions. Feel free to disagree, at this point, it matters little to me.

 

[sonitus mirus]

I just wanted to make sure the tests I performed were as accurate as possible. I was striving for the highest degree of objectivity that I could achieve. You just seem to want to convince yourself that your preconceived notions are true. These results are like so many others we see in this forum, in that they are not particularly convincing, with potentially flawed data being used.

Why wouldn't you want to be sure? This is where we differ.

There are instructions showing how to volume match files for use in an ABX test with Foobar in this forum. Before I felt comfortable that I was easily able to distinguish between two files that most people could not, I'd want someone to examine the files or show me how to make sure there was not some glitch in the conversion process.

 

[MrLazyAnt]

Quote:

I just wanted to make sure the tests I performed were as accurate as possible. I was striving for the highest degree of objectivity that I could achieve. You just seem to want to convince yourself that your preconceived notions are true. These results are like so many others we see in this forum, in that they are not particularly convincing, with potentially flawed data being used.

Why wouldn't you want to be sure? This is where we differ.

There are instructions showing how to volume match files for use in an ABX test with Foobar in this forum. Before I felt comfortable that I was easily able to distinguish between two files that most people could not, I'd want someone to examine the files or show me how to make sure there was not some glitch in the conversion process.

I can't easily distinguish the difference. I've no qualms admitting it. And like I said. The results were solid enough for me. I'm not a die-hardist (not that I see anything wrong with that, it just isn't me), and I'm rest happy in the knowledge that my Portaplayer is now usable for a more extensive library with very little compromise, at least to me. Which for me was the main point of this exercise.

 

[nick_charles]

Quote:

I figured 15 tests in total would be as good as 15 tests on the same file

 
Statistically no, guessing 5/5 is hard but not beyond random guessing. A single track 12/15 is statistically more powerful. This does not mean that you cannot distinguish between the files but it is insufficient proof for any one example.