Poll: Round 2, which file is the MP3, which is the CD?

[glen]

<< I would suspect adding such wupport would not be trivial >>


I agree, but mp3 is good enough, so space is not an issue anymore.
I think we will soon see more and better multi channel solutions.

 

[ff123]

Here's that article, translated to English:

http://fets3.freetranslation.com:5081/?Sequence=core&amp;Language=German%2FEnglish&amp;Url=www.heise.de%2Fct%2F00%2F06%2F092%2Fdefault.shtml

They tested 300 audiophiles, with both B&amp;W Nautilus 803 speakers and Sennheiser Orpheus Electrostatic Reference headphones, and not one of those people was able to consistently pick out the mp3 from the source material. That's a pretty bold test, with some of the best equipment in the world, why would an audio engineer be able to do it any better?

Actually, they tested only 14 people, and only two of whom were professionals. And actually one guy (a professional involved in mastering/production) did do significantly (two standard deviations) better than the average of the 14 people. And actually, some music was significantly easier to pick out than others. Moreover, many people were surprised to learn that mp3 sounded so good -- i.e., they were not trained to hear artifacts.

In other words, put demanding music (in an encoder sense) together with a truly discriminating (and trained) listening panel, and you will likely get different results.

I find the results of true formal listening tests much more valid. For example, read the &quot;Results of the AAC subjective Tests,&quot; at the MPEG site:

http://www.cselt.it/mpeg/quality_tests.htm

Unfortunately, there is no rating of mp3 at 256 kbit/s, but reading the paper gives you an idea of the pains they went through to properly evaluate codec quality.

ff123

 

[Eug]

This is much harder than the original thread. I would guess sample 2 is the original and sample 1 is the MP3, but I'm on shaky ground. But again, I'm competing against computer fan noise on my secondary (low-end) system since my main system is in pieces, and I can't find my damn headphones (just moved).

Anyways, I say this because it seems that some of the highs are a bit harsh on sample 1 on this sh!tty system.

But I would be happy with either for a MP3 file, at least on this system.

Just a note, I'm still convinced that with the absolute best settings with crapola Xing VBR, most (but probably not all) of you would be perfectly happy with it. I know I am, even on my main system, although every once in a while I wonder about a song's quality (highs). But as convenience format Xing VBR with the best settings is pretty good.

I want answers dammit!

EDIT:

Just saw the spectrum analysis. WTF? Why the humungous dropoff at the high end??? Anyways, it's nice to know my ears are still acceptable, and that my support for Xing VBR is not only because my ears are too old now (age 31).

 

[Eug]

Hmmm... Come to think of it, most of my VBR Xing testing was with songs having predominantly vocals and not a huge amount of cymbals, highhats, etc. With those songs I had a helluva time hearing any differences.

But like I said, on some songs I have wondered if I had heard a difference on the highs, but I couldn't reliably nail it every time, although this was just listening to the song, and not side by side testing. I will pick new songs and retest once my main system is up and running again.

What freeware software is available for doing the spectrum analysis?

I can encode a few with Xing VBR now if anybody is interested in testing them. I personally can't do a reliable test until I get my other system up. (Apex DVD/MP3 player, RCA DVD player, Kenwood receiver, Paradigm Monitor 7 speakers. Both the RCA and Apex are hooked up via digital, and I can start songs simultaneously and flip back and forth between them.)

 

[NaughtyusMaximus]

I think sample 2 is real, and sample 1 is the mp3. The highs sounded much more realistic, and less harsh than sample 1.

 

[rlism]

hrm... i think 2 is real too, 1 is mp3.

why? highs were higher? dunno, so close.

 

[Boonesmi]

they both sound good to me

 

[ff123]

<< WTF? Why the humungous dropoff at the high end??? >>



A spectrum analysis averaged over the entire sample might imply a big cut in high frequencies above 16 kHz, when in fact the true story (what one actually hears) is better represented by a spectral view (frequency vs. time). For example, here is a spectral view of a portion of sample123.mp3.

http://fastforward.iwarp.com/export/sample123.html

Frequencies less than 16 kHz are generally kept more or less intact (i.e., the music is lowpass filtered), but when needed, the encoder (in this case, Fraunhofer VBR) will go all the way up to 20 kHz. The spectral view also shows that when there are soft portions below the threshold of hearing (even below 16 kHz), the encoder throws those frequencies away.

ff123

 

[Quickfingerz]

What if you think the Mp3 version sounds better than the original wave?

You can tell with two files you need a base file to compare it to.

 

[DannyLove]

<< I want answers dammit! >>


yes me too, or i will be forced to not host, jk

 

[DannyLove]

stats on how many people downloaded sample1 and sample2: maybe the poll is biased!!!!!

FILE NAME........................HITS.........Data in KB

/glen/sample1.wav................220..........708,524

/glen/sample2.wav................121..........546,907

people are downloading sample1 and not sample2..........makes you think huh?

 

[rlism]

actually, it's pretty simple why the first sample gets more hits... it's listed first. someone'll click on it, realize that it's going to go directly to the audio player or they realize that it's too big and just cancel it.

if see how much data is actually downloaded per hit you'll see that more people completed download #2, or at least downloaded more of it. i don't think that anyone who downloaded just one of thie files is gonna vote anyway. =P

 

[glen]

Sample 1 is the MP3.