Comparing two Nero AAC settings Options

[muaddib]

Can we expect more results, or should we bring a conclusion now?

[Alexxander]

I'm done, and if not, tell me what more to contribute. My conclusion is rather clear, although only a few number of samples are tested.

[/mnt]

I tried the Closer To God sample:

56 12

CODE

foo_abx 1.3.4 report
foobar2000 v0.9.6.9
2009/09/01 14:02:40

File A: C:\Downloads\56_12_vs_56_13__second_try_\Closer_To_God__Edit_12.mp4
File B: C:\Temp\Closer To God [Edit].flac

14:02:40 : Test started.
14:03:03 : 01/01 50.0%
14:03:07 : 02/02 25.0%
14:03:11 : 03/03 12.5%
14:03:18 : 04/04 6.3%
14:03:27 : 04/05 18.8%
14:03:35 : 05/06 10.9%
14:03:41 : 06/07 6.3%
14:03:48 : 07/08 3.5%
14:03:57 : 08/09 2.0%
14:04:02 : 09/10 1.1%
14:04:09 : 09/11 3.3%
14:04:27 : 09/12 7.3%
14:04:35 : 09/13 13.3%
14:04:39 : Test finished.

----------
Total: 9/13 (13.3%)

Artifact appears at mid 0:06. Please note that i re-tetested the 56 12 encode, since i spotted the artifact more clearly on the 56 13 version.

56 13

CODE

foo_abx 1.3.4 report
foobar2000 v0.9.6.9
2009/09/01 14:00:22

File A: C:\Downloads\56_12_vs_56_13__second_try_\Closer_To_God__Edit_13.mp4
File B: C:\Temp\Closer To God [Edit].flac

14:00:22 : Test started.
14:00:51 : 01/01 50.0%
14:01:00 : 02/02 25.0%
14:01:08 : 03/03 12.5%
14:01:21 : 04/04 6.3%
14:01:28 : 05/05 3.1%
14:01:38 : 06/06 1.6%
14:01:46 : 07/07 0.8%
14:01:53 : 08/08 0.4%
14:02:01 : 09/09 0.2%
14:02:07 : 10/10 0.1%
14:02:14 : 11/11 0.0%
14:02:19 : 12/12 0.0%
14:02:21 : Test finished.

----------
Total: 12/12 (0.0%)

Artifact still appears at mid 0:06, and IMO this sounds worse then the 56 12 encode.

56 12 vs 56 13

CODE

foo_abx 1.3.4 report
foobar2000 v0.9.6.9
2009/09/01 14:05:26

File A: C:\Downloads\56_12_vs_56_13__second_try_\Closer_To_God__Edit_12.mp4
File B: C:\Downloads\56_12_vs_56_13__second_try_\Closer_To_God__Edit_13.mp4

14:05:26 : Test started.
14:05:48 : 01/01 50.0%
14:05:55 : 02/02 25.0%
14:06:01 : 03/03 12.5%
14:06:07 : 04/04 6.3%
14:06:12 : 05/05 3.1%
14:06:18 : 06/06 1.6%
14:06:23 : 07/07 0.8%
14:06:29 : 08/08 0.4%
14:06:37 : 09/09 0.2%
14:06:43 : 10/10 0.1%
14:06:51 : 11/11 0.0%
14:06:57 : 12/12 0.0%
14:07:02 : Test finished.

----------
Total: 12/12 (0.0%)

At mid 0:06 the artifact is still more annoying on file B (56 13).

[/mnt]

I decided to do a quick test with LAME 3.97 to see if this is a problem sample:

LAME 3.97 -V2 --vbr-new

CODE

foo_abx 1.3.4 report
foobar2000 v0.9.6.9
2009/09/01 14:26:17

File A: C:\Temp\Closer To God [Edit].flac
File B: C:\Temp\Closer To God [Edit].mp3

14:26:17 : Test started.
14:26:36 : 01/01 50.0%
14:26:50 : 02/02 25.0%
14:27:03 : 03/03 12.5%
14:27:14 : 04/04 6.3%
14:27:27 : 05/05 3.1%
14:27:43 : 06/06 1.6%
14:27:58 : 07/07 0.8%
14:28:07 : 08/08 0.4%
14:28:18 : 09/09 0.2%
14:28:26 : 10/10 0.1%
14:28:28 : Test finished.

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

Precho artifacts all over. Sounds better then Nero at 0:50, but worse then 0:60.

LAME 3.97 -V0 --vbr-new

CODE

foo_abx 1.3.4 report
foobar2000 v0.9.6.9
2009/09/01 14:29:39

File A: C:\Temp\Closer To God [Edit].flac
File B: C:\Temp\Closer To God [Edit] V0.mp3

14:29:39 : Test started.
14:29:52 : 01/01 50.0%
14:29:59 : 02/02 25.0%
14:30:06 : 03/03 12.5%
14:30:16 : 04/04 6.3%
14:30:23 : 05/05 3.1%
14:30:30 : 06/06 1.6%
14:30:37 : 07/07 0.8%
14:30:45 : 08/08 0.4%
14:30:53 : 09/09 0.2%
14:31:01 : 10/10 0.1%
14:31:10 : 11/11 0.0%
14:31:20 : 12/12 0.0%
14:31:21 : Test finished.

----------
Total: 12/12 (0.0%)

The artifacts sound worse on V0 .

LAME 3.97 320kbps CBR

CODE

foo_abx 1.3.4 report
foobar2000 v0.9.6.9
2009/09/01 14:36:27

File A: C:\Temp\Closer To God [Edit].flac
File B: C:\Temp\Closer To God [Edit] b320.mp3

14:36:27 : Test started.
14:36:40 : 01/01 50.0%
14:36:48 : 02/02 25.0%
14:36:53 : 03/03 12.5%
14:37:03 : 04/04 6.3%
14:37:11 : 05/05 3.1%
14:37:27 : 06/06 1.6%
14:37:35 : 07/07 0.8%
14:37:43 : 08/08 0.4%
14:37:54 : 09/09 0.2%
14:38:13 : 10/10 0.1%
14:38:25 : 11/11 0.0%
14:38:26 : Test finished.

----------
Total: 11/11 (0.0%)

Still ABXable at 320kbps, due to precho artifacts.

V2 vs V0

CODE

foo_abx 1.3.4 report
foobar2000 v0.9.6.9
2009/09/01 14:32:01

File A: C:\Temp\Closer To God [Edit] V0.mp3
File B: C:\Temp\Closer To God [Edit] V2.mp3

14:32:01 : Test started.
14:32:25 : 01/01 50.0%
14:32:32 : 02/02 25.0%
14:32:45 : 02/03 50.0%
14:33:03 : 03/04 31.3%
14:33:13 : 04/05 18.8%
14:33:19 : 05/06 10.9%
14:33:32 : 06/07 6.3%
14:33:44 : 07/08 3.5%
14:33:58 : 08/09 2.0%
14:34:09 : 09/10 1.1%
14:34:18 : 10/11 0.6%
14:34:26 : 11/12 0.3%
14:34:34 : 12/13 0.2%
14:34:50 : 13/14 0.1%
14:34:58 : 14/15 0.0%
14:34:59 : Test finished.

----------
Total: 14/15 (0.0%)

The V2 encode artifacts sound softer, much better then V0 IMO.

[IgorC]

Can we expect more results, or should we bring a conclusion now?

Do you need more results from the same persons ( from me?) or just people who haven't participated yet.
If you do then there is one problematic sample for 1.3.3 encoder at -q0.5 (maybe -q0.55/0.6 also. I will try.) http://ff123.net/samples/applaud00.flac

Also I will try to finish to test Pet Shop sample.

This post has been edited by IgorC: Sep 1 2009, 17:12

[IgorC]

Another easy to ABX sample Autechre from this test http://www.hydrogenaudio.org/forums/index....c=70442&hl=

Nero 1.3.3 -q 0.60. Sound collapse on transition around 3.8-4.3 seconds

CODE

foo_abx 1.3.4 report
foobar2000 v0.9.6.9
2009/09/01 13:25:29

File A: C:\0\01- DCT Killer Samples (Lossless)\01- Artefact+Context\02- Autechre (Artefact+Context) Lossless.flac
File B: C:\0\01- DCT Killer Samples (Lossless)\01- Artefact+Context\q060 02- Autechre (Artefact+Context) Lossless.mp4

13:25:29 : Test started.
13:25:48 : 01/01  50.0%
13:25:51 : 02/02  25.0%
13:25:55 : 03/03  12.5%
13:25:59 : 04/04  6.3%
13:26:02 : 05/05  3.1%
13:26:04 : Test finished.

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

This post has been edited by IgorC: Sep 1 2009, 17:36

[muaddib]

Here is the summary:

• Observations from vpa can’t be accepted because they are not backed up by an ABX test.
• rpp3po, sauvage78 and /mnt reported that they can’t hear a difference on emese, they could ABX both against the original.
• Alexxander and /mnt reported that they can’t hear a difference on bibilolo, they could ABX both against the original.
• rpp3po reported that he can’t hear a difference on bibilolo, he didn’t ABX against the original.
• C.R.Helmrich reported that he can’t hear differences, he didn’t ABX against the original.
• IgorC prefers version 12 for emese, he could ABX both against the original and against each other. He reported subtle differences between 12 and 13 that are hard to ABX (16/20 overall in 4 ABX sessions).
• Alexxander prefers version 12 for emese, he could ABX both against the original and against each other. He scored 13/15 between 12 and 13 in an ABX session.
• Alexxander prefers version 12 for pet_shop_boys, he could ABX 13 against the original and not 12 against the origianl. He scored 15/20 between 13 and the original in an ABX session.
• IgorCprefers version 12 for pet_shop_boys, he could ABX 13 against the original and not 12 against the origianl. He scored 15/20 between 13 and the original in an ABX session.
• /mnt prefers version 12 for closer_to_god, he could ABX 13 against the original and not 12 against the origianl. He could ABX 12 vs 13.

In spectrum plots, it can be noticed that version 13 has more of higher frequencies preserved. This is due to the differences in settings of encoder, where version 13 preserves also very quiet high frequencies. Differences at low frequencies can’t be spotted in the spectrum plots. As a consequence version 13 has a bit higher bitrate.
It seems, however, that people prefer version 12 a bit.
First conclusion based on the results would be that there is no point in comparing spectrum plots. Listening is the only way to test encoder quality.
Second conclusion would be, that differences between version 12 and 13 are very subtle. As a consequence for the next version of Nero AAC Encoder, settings in the middle, but closer to version 12, will be used. This is only for very high bitrates (starting from 220 kbps) in order to preserve a bit more of high frequencies than in previous version from 2008. At lower bitrates, settings which are more strict than version 12, will be used to provide best possible quality. Only at very high bitrates (starting from 280 kbps), all high frequencies will be reintroduced as it was in version from 2007.

Big thanks to all that participated!

[IgorC]

I was searching some information on the forum and saw this topic.
It was very hard to hear difference in past.
But now with a new pair of headphones and other hardware it's very easy.

This post has been edited by IgorC: Apr 19 2011, 09:50