Hydrogenaudio Forums > Nero AAC problem sample

Hydrogenaudio Forums > Lossy Audio Compression > AAC > AAC - Tech

slks

Jul 19 2008, 04:28

I've found what seems to be a problem sample for Nero AAC, that I was able to ABX without much difficulty at -q 0.5, -q 0.55, and -q 0.6 (I didn't bother going higher).

I am using the latest version, 1.1.34.2, with no options other than the -q.

I've uploaded a clip here:
http://www.yousendit.com/download/Q01GckhaMHd5UkZFQlE9PQ

To describe the artifact, the hiss-like percussion sounds are not as sharp, sort of smudged once it's gone through the encoder.

I hope this will be useful for tuning...

ABX logs:

CODE

foo_abx 1.3.3 report
foobar2000 v0.9.5.3
2008/07/19 04:21:15

File A: D:\AFX - Analogue Bubblebath 3 (1992) [FLAC] (REPHLEX CAT 008 CD)\08 - [cat 00897-Aa1].flac
File B: C:\Documents and Settings\tberg\Desktop\compress\08 - [cat 00897-Aa1].m4a

04:21:15 : Test started.
04:22:19 : 01/01  50.0%
04:22:53 : 02/02  25.0%
04:23:45 : 03/03  12.5%
04:24:13 : 04/04  6.3%
04:24:49 : 05/05  3.1%
04:25:27 : 06/06  1.6%
04:26:08 : 06/07  6.3%
04:28:00 : 07/08  3.5%
04:28:27 : 08/09  2.0%
04:29:00 : 09/10  1.1%
04:29:46 : 10/11  0.6%
04:29:50 : Test finished.

 ---------- 
Total: 10/11 (0.6%)

CODE

foo_abx 1.3.3 report
foobar2000 v0.9.5.3
2008/07/19 05:00:19

File A: D:\AFX - Analogue Bubblebath 3 (1992) [FLAC] (REPHLEX CAT 008 CD)\08 - [cat 00897-Aa1].flac
File B: C:\Documents and Settings\tberg\Desktop\compress\08 - [cat 00897-Aa1] - 0.6.m4a

05:00:19 : Test started.
05:02:25 : 01/01  50.0%
05:03:03 : 02/02  25.0%
05:03:52 : 03/03  12.5%
05:05:50 : 04/04  6.3%
05:06:26 : 05/05  3.1%
05:07:08 : 06/06  1.6%
05:08:37 : 07/07  0.8%
05:08:45 : Test finished.

 ---------- 
Total: 7/7 (0.8%)

/mnt

Jul 19 2008, 05:23

You should try the Feb 12 2007 build, since it can perform better then the latest version.

Anyway I did a test with that version I have suggested at q 55.

CODE

foo_abx 1.3.3 report
foobar2000 v0.9.5.4
2008/07/19 12:11:10

File A: C:\Temp\[cat 00897-Aa1].mp4
File B: C:\Downloads\[cat 00897-Aa1].flac

12:11:10 : Test started.
12:12:10 : 01/01  50.0%
12:12:39 : 02/02  25.0%
12:12:44 : 03/03  12.5%
12:13:03 : 04/04  6.3%
12:13:17 : 05/05  3.1%
12:13:31 : 06/06  1.6%
12:14:08 : 07/07  0.8%
12:14:20 : 08/08  0.4%
12:14:32 : 09/09  0.2%
12:14:53 : 10/10  0.1%
12:14:54 : Test finished.

 ---------- 
Total: 10/10 (0.1%)

A poping noise at around 0:06, which I keep hearing on some troubled tracks such as The Unthinking Majority by Serj Tankian at q 6. This file was encoded on the Feb 07 version of Nero AAC, since I refuse to use the latest version, due it producing really bad metallic hiss on silent parts of tracks.

muaddib

Jul 21 2008, 03:41

Thanx for reporting this one!
Have you maybe checked this one with other encoders? Is it only Nero AAC Encoder specific problem?

slks

Jul 22 2008, 02:23

I just tried it with both Vorbis aotuv b5 at -q 6, and Lame 3.98 at -V 2... both those encoders seem to have problems with it too. Vorbis performed comparably to Nero AAC, but LAME had much heavier and more immediately noticeable artifacts.

kornchild2002

Jul 22 2008, 03:35

I downloaded your sample and encoded it to 192kbps VBR AAC using iTunes 7.7 which I think uses QuickTime 7.5 for all its encoding needs. I compared that with Nero's latest build at -q0.5 (that was the closest setting I could get) and it appears that the problem persists in the iTunes AAC encoder as well. I also used the February build of the Nero AAC encoder and the results were similar. All-in-all, it appears that February Nero, August Nero, and the iTunes AAC encoders all performed about the same.

Here are my short ABX results for the iTunes AAC encoder at 192kbps VBR:

CODE

foo_abx 1.3.3 report
foobar2000 v0.9.5.4
2008/07/22 03:30:36

File A: C:\Users\Paul\Desktop\[cat 00897-Aa1].flac
File B: C:\Users\Paul\Desktop\[cat 00897-Aa1].m4a

03:30:36 : Test started.
03:31:01 : 01/01  50.0%
03:31:17 : 02/02  25.0%
03:31:48 : 03/03  12.5%
03:32:05 : 04/04  6.3%
03:32:21 : 05/05  3.1%

----------
Total: 5/5 (3.1%)

Edit: I know that normal ABX tests are normally a bit longer but I felt that there was no need to go beyond 5 trials. I can if you want me to though.

This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.