STOP TELLING PEOPLE YOU CAN'T TELL 192AAC VS LOSSLESS ILL PROVE IT
Feb 13, 2007 at 9:27 PM Post #31 of 463
Quote:

Originally Posted by Febs /img/forum/go_quote.gif
OK, I don't know what you mean by "challange of the same 10 songs," so here is my proposed challenge to you:

We mutually agree on three songs, in different genres. Since I will be preparing clips, these must be songs that I have available in my CD collection. I will then prepare clips of the three songs as follows:
  1. Make 30 second clips of each of those three songs. (For copyright reasons, I do not wish to distribute entire songs.)
  2. Encode the 30 second clips to 192kbps AAC.
  3. Transcode the original .wav clip and the 192kbps AAC clip to FLAC.
  4. Make the three pairs of clips available for download.

You will then:
  1. Download the three pairs of clips.
  2. For each pair of clips, conduct an ABX test in Foobar or an equivalent ABX program, using 16 trials for each test. You will turn OFF the option to display results as you go, so that you do not know until the end of the test how many you got right.
  3. Post the results of your ABX tests in this thread.

If you are able to successfully ABX all three pairs, I will change my signature to say "How Trose49 taught me the truth about lossless and showed me that I am deaf," with a hyperlink to this thread. If you are unable to successfully ABX any of the three pairs, you will change your signature to "How Febs taught me the truth about lossless and showed me that it was all my imagination," with a hyperlink to this thread. If you are able to successfully ABX some, but not all, of the three pairs of clips, then we both will change our signatures to say "How I was wrong about lossless versus lossy," with a link to this thread.

Note: by "successful ABX," I mean that the results of the ABX test indicate that there is less than a 5% chance that you are guessing.

What do you think?



First Off Im game!

Second what do you mean by three pairs of clips?

There would be one 192AAC, One original .wav and One flac.

I am supposed to pic out all three???

I will pick out which is the 192AAC! I never claimed to pick the others but I would try it for fun!

If I pick out the 192 stuff you change your sig if I dont I change Mine. We could still post the other data for fun.
 
Feb 13, 2007 at 9:30 PM Post #32 of 463
Quote:

Originally Posted by trose49 /img/forum/go_quote.gif
First Off Im game!

Second what do you mean by three pairs of clips?

There would be one 192AAC, One original .wav and One flac.

I am supposed to pic out all three???

I will pick out which is the 192AAC! I never claimed to pick the others but I would try it for fun!

If I pick out the 192 stuff you change your sig if I dont I change Mine. We could still post te other data for fun



No, the point would be that I would take a .wav clip, and:

1. Convert it to FLAC.

2. Convert it to 192AAC, and then convert that to FLAC.

Thus, for each song you will be testing, you will be comparing two FLAC files, one of which represents the uncompressed original, and the other of which represents the lossless file. You will not know which file represents the uncompresssed original and which represents the compressed version. There will be three songs, each with two FLAC files to compare, for a total of six clips.

Song 1a.flac
Song 1b.flac
Song 2a.flac
Song 2b.flac
Song 3a.flac
Song 3b.flac

ABX test 1 would compare clips 1a and 1b, ABX test 2 would compare clips 2a and 2b, and ABX test 3 would compare clips 3a and 3b.
 
Feb 13, 2007 at 9:41 PM Post #35 of 463
Quote:

Originally Posted by Benco /img/forum/go_quote.gif
I love lossless, and have been FLAC only for quite awhile now, but its going to be funny when this arrogant dude loses.

edited because you can't say a word that stars with a c and ends in ocky



So will the lossless files sound any different being converted to FLAC?

Other wise I dont see an issue.

How about Eric Clapton Sessions for Robert J CD Milkcow?
John Mayer Waiting on the world to change?
Yanni live from the acropolis track 1
evanescence anywhere but home CD further away?

how are we doing so far?
 
Feb 13, 2007 at 9:41 PM Post #37 of 463
I agree that there should be at least 5 trials. With only 3, there is a decent (higher than 10%) chance that he'll get all three right just by guessing. Statistically speaking, even if he gets all three right, that does not prove anything. If you go to 5, that would be about a 3% chance of getting them all right randomly, which gets into solid ground.

If you make it 10, with each of you getting to choose 5 songs, I would consider putting up a pretty good prize if trose puts up his UE10s.
 
Feb 13, 2007 at 9:45 PM Post #38 of 463
Quote:

Originally Posted by seacard /img/forum/go_quote.gif
I agree that there should be at least 5 trials. With only 3, there is a decent (higher than 10%) chance that he'll get all three right just by guessing. Statistically speaking, even if he gets all three right, that does not prove anything. If you go to 5, that would be about a 3% chance of getting them all right randomly, which gets into solid ground.

If you make it 10, with each of you getting to choose 5 songs, I would consider putting up a pretty good prize in trose puts up his UE10s.



5 is good also,

I am trying to think of more songs Actually Yanni CD isnt recorded well to begin with.

How about SRV Sky is crying CD empty arms?

Sorry I listen mostly to guitar players, blues etc.

we could do 6 song I pick three you pick three?

That would be fair!
 
Feb 13, 2007 at 9:47 PM Post #40 of 463
Quote:

Originally Posted by seacard /img/forum/go_quote.gif
I agree that there should be at least 5 trials. With only 3, there is a decent (higher than 10%) chance that he'll get all three right just by guessing. Statistically speaking, even if he gets all three right, that does not prove anything. If you go to 5, that would be about a 3% chance of getting them all right randomly, which gets into solid ground.

If you make it 10, with each of you getting to choose 5 songs, I would consider putting up a pretty good prize if trose puts up his UE10s.



Weren't the rules to do 16 trials per pair?
 
Feb 13, 2007 at 9:51 PM Post #41 of 463
...

Maybe I shouldn't mention this before people start but the source has a lot to do with how transparent the encodes will be. There are some types of sounds that mp3 will choke on no matter the bitrate, and I don't know how good AAC will do. For almost all modern recorded music... you won't be able to tell at v0. If you plan on being able to tell with your PC as a source, make sure there's no resampling going on and your hearing above 18khz is still intact (v0 cuts off at 19.5khz I believe.)

The only way to win this is to purposefully pick bad/noisy masters and find the artifacts generated from the noise.
 
Feb 13, 2007 at 9:54 PM Post #43 of 463
Quote:

Originally Posted by DarkJC /img/forum/go_quote.gif
Weren't the rules to do 16 trials per pair?


Exactly. Three separate tests, each test consisting of 16 separate trials. If I am doing my math right, the odds of getting all 16 trials right purely by guessing are 1 in 65,536. The odds of getting all 48 trials right purely by guessing are 1 in 281.5 trillion.

Trose49, we haven't connected yet on any specific songs, but now that I have an idea of your preference, let me look over my collection and make some suggestions.
 
Feb 13, 2007 at 9:57 PM Post #44 of 463
Quote:

Originally Posted by Thelonious Monk /img/forum/go_quote.gif
i will start a "cool headfi users" list in my sig with febs as #1 if he wins lol


And what if I win?

There was a good point I can use my computer as a source my sound card sucks If you send me the files I can put them on my Ipod so I can listen to them the way I am used to IPOD>CRYO>HORNET>UE-10's

This is the only config I know I can hear a difference so Im not changing sources/amps or anything.

I know I can pick em out im telling ya!
 
Feb 13, 2007 at 9:58 PM Post #45 of 463
Quote:

Originally Posted by Febs /img/forum/go_quote.gif
Exactly. Three separate tests, each test consisting of 16 separate trials. If I am doing my math right, the odds of getting all 16 trials right purely by guessing are 1 in 65,536. The odds of getting all 48 trials right purely by guessing are 1 in 281.5 trillion.

Trose49, we haven't connected yet on any specific songs, but now that I have an idea of your preference, let me look over my collection and make some suggestions.



I could also send you some lossless stuff to use or part of some songs etc..

What about the source issue? My sound card in my PC is crap!
 

Users who are viewing this thread

Back
Top