IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
Improving vorbis q6 quality?
Djnx
post Jun 3 2007, 03:07
Post #1





Group: Members
Posts: 2
Joined: 15-May 07
Member No.: 43480



Hi.

I just made some ABX tests, I started from q2 up to q6. From q2 to q5 I got 5/5 in every test. And these are the results of the ABX test in wav vs q6:

CODE
foo_abx 1.3.1 report
foobar2000 v0.9.4.3
2007/06/02 20:10:30

File A: C:\Documents and Settings\lalilulelo\My Documents\My Music\Marilyn Manson\Mechanical Animals\12 Fundamentally Loathsome.wav
File B: C:\Documents and Settings\lalilulelo\My Documents\My Music\Marilyn Manson\Mechanical Animals\12 Fundamentally Loathsome.ogg

20:10:30 : Test started.
20:12:40 : 01/01  50.0%
20:13:30 : 02/02  25.0%
20:14:52 : 03/03  12.5%
20:16:02 : 03/04  31.3%
20:16:13 : Test finished.

----------
Total: 3/4 (31.3%)


I have an Audiostorm soundcard and a pair of E2 earphones. I used aotuv beta 5 with impulse_noisetone set to -15. I didn't finished the last test because whatever the result was I knew I needed to do further tweaking on the vorbis encoder options, unfortunately I'm not that techsavvy and I don't know what to do to make vorbis q6 sound a little better. Of course at this level we are talking about very very subtle differences between the loseless and the lossy samples. Soon I will buy a Delta 1010LT soundcard and the Shures SE530 earphones and I'm affraid that could make the subtle differences... less subtle, and I feel that using -q7 would be only brute force.

So I would appreaciate any advice on what to do.

Thanks in advance.
Go to the top of the page
+Quote Post
Cygnus X1
post Jun 3 2007, 03:43
Post #2





Group: Members (Donating)
Posts: 676
Joined: 5-June 02
From: New York
Member No.: 2224



Getting 5 trials correct in an ABX test is not enough for statistical significance. The generally accepted minimum for significance is .05 or a 5% chance that your results are a false positive (i.e., Type I error). Please re-do your test to include at least 10 trials; 15-20 is optimal.

This post has been edited by Cygnus X1: Jun 3 2007, 03:44
Go to the top of the page
+Quote Post
shadowking
post Jun 3 2007, 03:43
Post #3





Group: Members
Posts: 1494
Joined: 31-January 04
Member No.: 11664



Abx 3/4 ??

Sorry this abx result is not valid. You need 3% or less. If its very difficult to hear , try 2 x 4 trials then add the results. You need to score at least 7/8 or better yet 8/8.

This post has been edited by shadowking: Jun 3 2007, 03:44
Go to the top of the page
+Quote Post
Djnx
post Jun 3 2007, 03:59
Post #4





Group: Members
Posts: 2
Joined: 15-May 07
Member No.: 43480



QUOTE (shadowking @ Jun 2 2007, 21:43) *
Abx 3/4 ??

Sorry this abx result is not valid. You need 3% or less. If its very difficult to hear , try 2 x 4 trials then add the results. You need to score at least 7/8 or better yet 8/8.


As I said in my first post, I didn't finished the last ABX test, in the other tests I got 3.1% (q2 to q5).

I just wanted some advice on what to tweak on the encoder settings before finishing the q6 test.
Go to the top of the page
+Quote Post
Firon
post Jun 20 2007, 20:36
Post #5





Group: Members
Posts: 830
Joined: 3-November 05
Member No.: 25526



How many trials did you do for the other tests?

And try doing it WITHOUT altering the impulse profile options.
Go to the top of the page
+Quote Post

Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 22nd May 2013 - 19:42