-V3 and -V4 in 3.97a10

Topic: -V3 and -V4 in 3.97a10 (Read 16437 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.

-V3 and -V4 in 3.97a10

Reply #50 – 2005-07-19 21:34:00

Quote

Quote
OK, i'll also use 3.93.1, this version uses "--preset" and "--alt-prest" instead of "-V" right?

Yes, you have to use either --alt-preset or --preset.

Quote
where can I find 3.93.1 (recommended download source)?
[{POST_SNAPBACK}][/a]

You can find it at rjamorim's reallyrarewares ([a href="http://www.rjamorim.com/rrw/files/lame/lame3.93.1.zip]link[/url])
[a href="index.php?act=findpost&pid=314693"][{POST_SNAPBACK}][/a]

Thanks for the link sTisTi.

As for 3.93.1 not being recommended, there is nothing to be lost by trying something with it, all of us (meybe not ALL) might learn something here.

Also, the sticky about the Lame versions says that 3.93.1's --preset medium is somewhat untested, we just might change that here, or at least figure out how it compares to the other recommended versions and why.

-V3 and -V4 in 3.97a10

Reply #51 – 2005-07-19 21:40:25

Quote

Hmmm... 3.93.1 is not one of the recommended versions. Either 3.96.1 or 3.90.3 (or maybe the latest alpha). I haven't tried it myself, but I think you will find 3.93.1 inferior (in most cases) to the recommended versions, qualitywise.
[a href="index.php?act=findpost&pid=314701"][{POST_SNAPBACK}][/a]

I know, I never said it should be recommended. It's just an option e.g. for portables if you find 3.96.1 --preset medium too big and -V5 too small. In my experience, 3.93.1 medium's bitrate is mostly right between these two and in my (limited) tests it was clearly better than 3.96.1's -V5.

-V3 and -V4 in 3.97a10

Reply #52 – 2005-07-19 21:41:56

sorry, wrong button (can be deleted)

-V3 and -V4 in 3.97a10

Reply #53 – 2005-07-20 01:41:52

Here's the test with 3.93.1 --preset medium:

WAV and 3.93.1 apm:

Code: [Select]

foo_abx v1.2 report
foobar2000 v0.8.3
2005/07/19 16:34:30

File A: file://F:\Samuel's Music tests\Opeth\Blackwater Park\01 - The Leper Affinity.wav
File B: file://F:\Samuel's Music tests\Opeth\01 - The Leper Affinity 3.93.1 apm.mp3

16:35:20 : Test started.
16:36:32 : 01/01  50.0%
16:36:57 : 02/02  25.0%
16:37:15 : 03/03  12.5%
16:37:28 : 04/04  6.3%
16:37:39 : 05/05  3.1%
16:37:52 : 06/06  1.6%
16:38:11 : 07/07  0.8%
16:38:29 : 08/08  0.4%
16:38:47 : 09/09  0.2%
16:39:00 : 10/10  0.1%
16:39:01 : Test finished.

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

With this last test of the medium preset and -V4 i belive that while all of the encodes have noticeable artefacts, trying to figure out wich encoder produces least annoying ones is the challenge.

3.93.1 produces very similar artefacts to 3.97a11, and as evereux and myself noticed, these are easily ABX'ed.

3.96.1 produces better quality than the others most of the time, but sometimes it does throw in some pretty apparent artefacts, artefacts that 3.90.3 and 3.97a11 hide a little better.

If i were to choose between these four above mentioned versions, it would be between 3.90.3, for more modest filesize and 3.96.1 for speed/little quality increase.
The others are just too annoying and noticeable.

Tests with --preset standard -Y on 3.90.3, and -V2 -Y on 3.96.1 and 3.97a11 to follow.

-V3 and -V4 in 3.97a10

Reply #54 – 2005-07-24 07:29:44

This thread got me curious to see if I could switch from -V 2 to -V 3--vbr-new. So after all this time using --preset standard, I finally decided to do my first abx tests to see if I really needed it. I downloaded some of the test samples from ff123's site, and some from HA threads. The "infamous" ones like castanets, fatboy, as well as kraftwerk, the beatles.

In short, I could successfully abx all of them with -V 3 --vbr-new, and also also had success with -V 2 on 2 of 3 samples. But there was a difference between the successful tests from -V 2 to -V 3: the -V 3 tests were very quick and very confident, and the -V 2 tests took a bit longer and a few trials were pure guesses on my part, even though the overall test turned successful. So I'm sticking with -V 2.

The tests were done using some nondescript earbuds into an SBLive Value card using kx project drivers.

I didn't have the time or energy tonight to abx these samples with -V 2 --vbr-new, but maybe another time.

ABX logs for -V 3 --vbr-new:

Code: [Select]

foo_abx 1.2 report
foobar2000 v0.9 beta 5
2005/07/23 21:49:33

File A: C:\My Documents\test files\castanets.flac
File B: C:\My Documents\test files\v3new - castanets.mp3

21:49:34 : Test started.
21:50:14 : 01/01  50.0%
21:50:22 : 02/02  25.0%
21:50:30 : 03/03  12.5%
21:50:39 : 04/04  6.3%
21:51:00 : 05/05  3.1%
21:51:06 : 06/06  1.6%
21:51:19 : 07/07  0.8%
21:51:32 : 08/08  0.4%
21:51:38 : 09/09  0.2%
21:51:46 : 10/10  0.1%
21:51:53 : 11/11  0.0%
21:51:57 : 12/12  0.0%
21:52:01 : 13/13  0.0%
21:52:06 : 14/14  0.0%
21:52:20 : Test finished.

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

Code: [Select]

foo_abx 1.2 report
foobar2000 v0.9 beta 5
2005/07/23 21:59:15

File A: C:\My Documents\test files\kraftwerk.flac
File B: C:\My Documents\test files\v3new - kraftwerk.mp3

21:59:19 : Test started.
22:02:21 : 01/01  50.0%
22:02:27 : 02/02  25.0%
22:02:33 : 03/03  12.5%
22:02:41 : 04/04  6.3%
22:02:45 : 05/05  3.1%
22:03:02 : 06/06  1.6%
22:03:19 : 07/07  0.8%
22:03:28 : 08/08  0.4%
22:03:33 : 09/09  0.2%
22:03:44 : 10/10  0.1%
22:03:53 : 11/11  0.0%
22:04:00 : 12/12  0.0%
22:04:05 : 13/13  0.0%
22:04:07 : Test finished.

 ---------- 
Total: 13/13 (0.0%)

Code: [Select]

foo_abx 1.2 report
foobar2000 v0.9 beta 5
2005/07/23 21:53:39

File A: C:\My Documents\test files\fatboy_30sec.flac
File B: C:\My Documents\test files\v3new - fatboy_30sec.mp3

21:53:43 : Test started.
21:56:37 : 01/01  50.0%
21:56:57 : 02/02  25.0%
21:57:17 : 03/03  12.5%
21:57:36 : 04/04  6.3%
21:57:42 : 05/05  3.1%
21:57:49 : 06/06  1.6%
21:58:00 : 07/07  0.8%
21:58:08 : 08/08  0.4%
21:58:14 : 09/09  0.2%
21:58:20 : 10/10  0.1%
21:58:34 : 11/11  0.0%
21:58:45 : 12/12  0.0%
21:58:54 : 13/13  0.0%
21:58:57 : Test finished.

 ---------- 
Total: 13/13 (0.0%)

ABX tests at -V 2

Code: [Select]

foo_abx 1.2 report
foobar2000 v0.9 beta 5
2005/07/23 22:08:38

File A: C:\My Documents\test files\castanets.flac
File B: C:\My Documents\test files\v2 - castanets.mp3

22:08:39 : Test started.
22:10:23 : 01/01  50.0%
22:10:28 : 01/02  75.0%
22:10:38 : 02/03  50.0%
22:10:47 : 03/04  31.3%
22:10:56 : 04/05  18.8%
22:11:02 : 05/06  10.9%
22:11:08 : 06/07  6.3%
22:11:13 : 07/08  3.5%
22:11:23 : 08/09  2.0%
22:11:43 : 08/10  5.5%
22:11:55 : 09/11  3.3%
22:12:08 : 10/12  1.9%
22:12:18 : 11/13  1.1%
22:12:33 : 12/14  0.6%
22:12:52 : 13/15  0.4%
22:13:01 : Test finished.

 ---------- 
Total: 13/15 (0.4%)

Code: [Select]

foo_abx 1.2 report
foobar2000 v0.9 beta 5
2005/07/23 22:13:25

File A: C:\My Documents\test files\kraftwerk.flac
File B: C:\My Documents\test files\v2 - kraftwerk.mp3

22:13:28 : Test started.
22:15:21 : 01/01  50.0%
22:15:30 : 02/02  25.0%
22:15:48 : 03/03  12.5%
22:15:55 : 04/04  6.3%
22:16:09 : 05/05  3.1%
22:16:19 : 06/06  1.6%
22:16:26 : 07/07  0.8%
22:16:35 : 08/08  0.4%
22:16:43 : 09/09  0.2%
22:16:51 : 10/10  0.1%
22:16:58 : 11/11  0.0%
22:17:04 : 12/12  0.0%
22:17:13 : 13/13  0.0%
22:17:14 : Test finished.

 ---------- 
Total: 13/13 (0.0%)

Code: [Select]

foo_abx 1.2 report
foobar2000 v0.9 beta 5
2005/07/23 22:17:56

File A: C:\My Documents\test files\taxman.flac
File B: C:\My Documents\test files\v2 - taxman.mp3

22:17:57 : Test started.
22:20:35 : 01/01  50.0%
22:20:58 : 02/02  25.0%
22:21:13 : 03/03  12.5%
22:22:11 : 04/04  6.3%
22:22:23 : 05/05  3.1%
22:22:41 : 06/06  1.6%
22:22:52 : 07/07  0.8%
22:23:00 : 07/08  3.5%
22:23:10 : 08/09  2.0%
22:23:21 : 08/10  5.5%
22:23:35 : 08/11  11.3%
22:23:58 : 08/12  19.4%
22:24:18 : 09/13  13.3%
22:24:36 : 09/14  21.2%
22:24:57 : 10/15  15.1%
22:25:40 : 10/16  22.7%
22:25:42 : Test finished.

 ---------- 
Total: 10/16 (22.7%)

About this

Quote

If you use -V2 -Y you will for sure also see a dramatic bitrate drop, the result will probably be not much larger than -V3.
[a href="index.php?act=findpost&pid=310539"][{POST_SNAPBACK}][/a]

I just wanted to mention that I could not successfully encode with the -Y switch with 3.97. I didn't see this switch in the documentation, and foobar just did nothing when I had the switch in the parameters.

-V3 and -V4 in 3.97a10

Reply #55 – 2005-07-24 17:31:01

For the above tests, you encoded everything with 3.97 and vbr-new?, maybe you should try -V3 on 3.96.1, with and without --vbr-new, and -V3 without vbr-new on 3.97.

I did some tests on some of my own samples (including the opeth sample) with -V2 -Y (on 3.90.3, 3.96.1, 3.97a11) and failed, mostly getting ~6/10, and i did guess a lot, but seeing -V2 -Y encodes with that much HF cut makes me paranoid, even if i can't hear a difference.

-V3 and -V4 in 3.97a10

Reply #56 – 2005-07-24 18:22:43

I was testing -V 3 --vbr-new and -V 2 (vbr-old).

If I get more time I may test -V 3 (vbr-old), but looking at the bitrate distribution of some recent rips, it looks like the difference between -vbr-new and old at these V levels is marginal, and I suspect the abx results would be similar to my previous tests. In other words, I suspect the audible difference between these -V levels is greater than the difference between -vbr-new and old within the same -V level.

Mostly I'm interested in testing 3.97, so I don't think I'll run any tests on the older encoders (sorry).

I think my next tests will be speed tests between encoding -V 2 and -V 2 --vbr-new, followed by abx tests for -V 2 --vbr-new.

Also I figured out my problem with -Y was capitalization. So I'm gonna try -V 2 -Y also. I hope I fail the tests actually because I'd like the bitrate savings.

-V3 and -V4 in 3.97a10

Reply #57 – 2005-07-24 18:27:33

No problem. 3.97 needs to be tested.

Notice