sample for vorbis aoTuV, ...and for other lossy encoders Options

[guruboolez]

CODE

Sample name: Pierres réfléchies [reflected stones]
Artist name: Pierre Henry
Composition: Pierres réfléchies [34 min 53 sec]
CD: MIX Pierre Henry 04.1
Label: Philips
Reference: 472202-2

CODE

Lossless format: flac 1.10
Size: 800 kb
Length: 7 seconds

CODE

Lossy encoder tested:
• vorbis aoTuV beta 2, at:
-q6 (233 kbps): ABX = 8/8 [range: 0.0 – 7.0]
-q7 (330 kbps): ABX = 8/8 [range: 3.0 – 7.0]
-q8 (365 kbps): ABX = 17/20 [range: 1.0 – 4.0]
[failed at –q9]

• vorbis 1.01+GT3b2, at:
-q6 (308 kbps): ABX = 8/8 [range: 0.0 – 7.0]
-q7 (353 kbps): ABX = 8/8 [range: 0.0 – 7.0]
-q8 (461 kbps): ABX = 13/16 [range: 1.0 – 3.0]
[failed at –q9]

Problem description: micro-attacks are blured, and sharpness is smoothened by an extra-noise.

CODE

• musepack 1.14 with --xlevel, at:
--quality 5 (196 kbps): ABX = 11/12 [full range]
--quality 6 (227 kbps): ABX = 14/16 [full range]
--quality 7 (255 kbps): ABX = 8/8 [full range]
--quality 8 (284 kbps): ABX = 8/8 [1.2 – 3.0]
--quality 9 (312 kbps): ABX = 8/8 [1.2 – 3.0]
[didn’t test above]

Problem description: from –q5 to –q7 [include] micro-attacks are distorted. Reminds me a typical artifact I heard many times with similar signal and mpc encoding [creaking.wav; jump.wav; tosca.wav]: something like “boiling oil”...
At –q8 and –q9, this artifact disapeared, but a lack of sharpness and a slight noise was audible, like vorbis aoTuV/GT3. Better ABX score at higher bitrate are due to a stronger concentration, not really needed with –standard/--extreme presets, obviously worse.

____
Testing conditions:
- Beyerdynamic DT-531 headphone
- Terratec DMX6fire soundcard
- Onkyo R-A5 tuner/amplifier
- foobar2000 0.82: kernel streaming • 32 fixed bit output without dithering
- foo_abx v1.2 • no replaygain
____
Extra-comments:
- I’ve began with aoTuV testing, and I resume the test 24 hours after with GT3 then musepack for comparison.
- bitrate deviation is impressive with vorbis (especially Garf Tuning), but is controlled with musepack. MPC –q9 have the same bitrate on this short sample than vorbis 1.01+GT3b2 –q6.
- ABX log files are included in the .zip archive.

[QuantumKnot]

The bitrate jump from q 6 to q 7 for aoTuV b2 is particularly surprising (233 -> 330 kbps).

[shadowking]

My results:

mpc q5 1.14

foo_abx v1.2 report
foobar2000 v0.8.2
2004/06/01 18:50:40

File A: file://H:\temp\abx tests\HENRY - Pierres Réfléchies#4.shorter.flac
File B: file://H:\temp\abx tests\HENRY - Pierres Réfléchies#4.shorter.mpc

18:50:41 : Test started.
18:51:39 : 01/01 50.0%
18:52:02 : 02/02 25.0%
19:12:50 : 03/03 12.5%
19:13:33 : 04/04 6.3%
19:20:05 : 05/05 3.1%
19:20:25 : 06/06 1.6%
19:21:44 : 07/07 0.8%
19:22:48 : 08/08 0.4%
19:23:39 : 09/09 0.2%
19:23:46 : Test finished.

----------
Total: 9/9 (0.2%)


mpc q5 1.15r -

foo_abx v1.2 report
foobar2000 v0.8.2
2004/06/01 19:25:54

File A: file://H:\temp\abx tests\HENRY - Pierres Réfléchies#4.shorter.flac
File B: file://H:\temp\abx tests\HENRY - Pierres Réfléchies#4.shorter.mpc

19:25:55 : Test started.
19:26:23 : 01/01 50.0%
19:26:41 : 02/02 25.0%
19:26:48 : 03/03 12.5%
19:27:58 : 04/04 6.3%
19:28:23 : 05/05 3.1%
19:28:47 : 06/06 1.6%
19:29:14 : 07/07 0.8%
19:29:55 : 08/08 0.4%
19:29:59 : Test finished.

----------
Total: 8/8 (0.4%)


AoTuv q6 - 1st attempt

foo_abx v1.2 report
foobar2000 v0.8.2
2004/06/01 20:05:03

File A: file://H:\temp\abx tests\HENRY - Pierres Réfléchies#4.shorter.flac
File B: file://H:\temp\abx tests\HENRY - Pierres Réfléchies#4.shorter.ogg

20:05:04 : Test started.
20:07:52 : 01/01 50.0%
20:09:32 : 01/02 75.0%
20:10:40 : 01/03 87.5%
20:12:08 : 01/04 93.8%
20:14:32 : 02/05 81.3%
20:14:43 : 03/06 65.6%
20:16:19 : 03/07 77.3%
20:17:21 : Test finished.

----------
Total: 3/7 (77.3%)


another attempt AoTuv-

foo_abx v1.2 report
foobar2000 v0.8.2
2004/06/01 20:18:29

File A: file://H:\temp\abx tests\HENRY - Pierres Réfléchies#4.shorter.flac
File B: file://H:\temp\abx tests\HENRY - Pierres Réfléchies#4.shorter.ogg

20:18:30 : Test started.
20:18:41 : 00/01 100.0%
21:06:49 : Trial reset.
21:07:14 : 01/01 50.0%
21:07:41 : 02/02 25.0%
00:57:18 : 03/03 12.5%
00:58:05 : 04/04 6.3%
01:02:57 : 05/05 3.1%
01:04:21 : 06/06 1.6%
01:06:01 : Test finished.

----------
Total: 6/7 (6.3%)



mpc has an obvious artifact. vorbis needed more attention, but it is also muffled in a way.