According to this thread:
http://www.hydrogenaudio.org/forums/index....=16&t=7790&st=0
it seems that a "regression" was introduced during february in 3.94 related to abr/cbr.
I have uploaded 4 compiles from 4 different february dates on http://gabriel.mp3-tech.org/lame
I'd like you to tell me on which ones do you hear the problem.
(I am posting this in the news section in order to stay on front page a few days)
Full Version: Lame 3.94 - need your help in identifying problem
Sample : APPLAUD.WAV
Preset : --preset 160
a12, xx16 & xx19 are incredibly bad.
Preset : --preset 160
CODE
ABC/HR Version 0.9b, 30 August 2002
Testname:
1L = C:\Temp\lame\001_applaud00_394a12.wav
2L = C:\Temp\lame\001_applaud00_xx15.wav
3L = C:\Temp\lame\001_applaud00_3902.wav
4L = C:\Temp\lame\001_applaud00_xx19.wav
5L = C:\Temp\lame\001_applaud00_xx16.wav
6L = C:\Temp\lame\001_applaud00_xx08.wav
---------------------------------------
General Comments:
---------------------------------------
1L File: C:\Temp\lame\001_applaud00_394a12.wav
1L Rating: 1.5
1L Comment:
---------------------------------------
2L File: C:\Temp\lame\001_applaud00_xx15.wav
2L Rating: 4.0
2L Comment:
---------------------------------------
3L File: C:\Temp\lame\001_applaud00_3902.wav
3L Rating: 4.0
3L Comment:
---------------------------------------
4L File: C:\Temp\lame\001_applaud00_xx19.wav
4L Rating: 1.5
4L Comment:
---------------------------------------
5L File: C:\Temp\lame\001_applaud00_xx16.wav
5L Rating: 1.5
5L Comment:
---------------------------------------
6L File: C:\Temp\lame\001_applaud00_xx08.wav
6L Rating: 4.0
6L Comment:
---------------------------------------
ABX Results:
Testname:
1L = C:\Temp\lame\001_applaud00_394a12.wav
2L = C:\Temp\lame\001_applaud00_xx15.wav
3L = C:\Temp\lame\001_applaud00_3902.wav
4L = C:\Temp\lame\001_applaud00_xx19.wav
5L = C:\Temp\lame\001_applaud00_xx16.wav
6L = C:\Temp\lame\001_applaud00_xx08.wav
---------------------------------------
General Comments:
---------------------------------------
1L File: C:\Temp\lame\001_applaud00_394a12.wav
1L Rating: 1.5
1L Comment:
---------------------------------------
2L File: C:\Temp\lame\001_applaud00_xx15.wav
2L Rating: 4.0
2L Comment:
---------------------------------------
3L File: C:\Temp\lame\001_applaud00_3902.wav
3L Rating: 4.0
3L Comment:
---------------------------------------
4L File: C:\Temp\lame\001_applaud00_xx19.wav
4L Rating: 1.5
4L Comment:
---------------------------------------
5L File: C:\Temp\lame\001_applaud00_xx16.wav
5L Rating: 1.5
5L Comment:
---------------------------------------
6L File: C:\Temp\lame\001_applaud00_xx08.wav
6L Rating: 4.0
6L Comment:
---------------------------------------
ABX Results:
a12, xx16 & xx19 are incredibly bad.
I have similar results.
15 - "ok". 16 - something was messed up.
15 - "ok". 16 - something was messed up.
Thank you.
After listening to both applaud and a sample of my preference, and throwing an encode with 3.92 into the mix, in order of worst to best:
x19
x8
3.92
x15
x16
x19
x8
3.92
x15
x16
I have uploaded a modified 3.94 which should fix the problem.
Could you confirm/infirm?
Could you confirm/infirm?
Gabriel, i've done a quick test and found that the problem is still present with the executable contained in lametest.zip. Even with spectral view applaud.wav has "spikes" similar to 3.94a12 ones.
I have uploaded a new test version. There are now test and test2.
lametest2 has the same problem.
Well, that's really annoying...
Could anyone else confirm?
Something else that I missed until now:
Could you try a regular 3.94 (that sould exhibit the problem) and add -X1,1 when testing abr?
Could anyone else confirm?
Something else that I missed until now:
Could you try a regular 3.94 (that sould exhibit the problem) and add -X1,1 when testing abr?
QUOTE(Gabriel @ Apr 19 2003 - 02:14 PM)
Could you try a regular 3.94 (that sould exhibit the problem) and add -X1,1 when testing abr?
Listening applaud with regular 3.94a12 -X1,1 the problem seem to disappear !
See also these captures:
3.90.2 --alt-preset 128 --scale 1
test1 --preset 128 --scale 1
test2 --preset 128 --scale 1
3.94a12 --preset 128 -X1,1 --scale 1
Well, we are progressing...
What about -X3,1 ?
What about -X3,1 ?
QUOTE(Gabriel @ Apr 19 2003 - 02:55 PM)
What about -X3,1 ?
-X3,1 is bad, same problem.
Ok, thank you very much.
Problem is found, source code will be updated during this weekend.
Btw, I am annoyed because theorically that should not be the case.
Problem is found, source code will be updated during this weekend.
Btw, I am annoyed because theorically that should not be the case.
Test on the TEST1 and TEST2 fixes. While encoding, not listening, noticed a possible change in joint stereo. The first LAMETEST used more LR than MS than LAMETEST2. Both are more like x19 and x16. x8 and x15 are both more like 3.92 for JS.
On with the test. Compared with the 3.92 and x16 encodes. Test1
was the worst, and it was a tie between 3.92, x16, test2. Though, on my test sample, which is from music I like (kick drum, cymbals, woman's voice), test2 was the worst, most noticeable in its softening of the kick. 3.92 and x16 tied.
Just for the record, I don't bark when dog whistles are blown. I hear up to about 18.5, but after 17, the clarity of the sounds fade.
On with the test. Compared with the 3.92 and x16 encodes. Test1
was the worst, and it was a tie between 3.92, x16, test2. Though, on my test sample, which is from music I like (kick drum, cymbals, woman's voice), test2 was the worst, most noticeable in its softening of the kick. 3.92 and x16 tied.
Just for the record, I don't bark when dog whistles are blown. I hear up to about 18.5, but after 17, the clarity of the sounds fade.
QUOTE(Gabriel @ Apr 19 2003 - 02:05 PM)
Ok, thank you very much.
Problem is found, source code will be updated during this weekend.
Btw, I am annoyed because theorically that should not be the case.
Problem is found, source code will be updated during this weekend.
Btw, I am annoyed because theorically that should not be the case.
Code has been updated, therefore, new win32 binaries are available at 'Other Stuff' Mirror 1.
Please wait before rushing...new alpha will follow very soon
QUOTE(Gabriel @ Apr 20 2003 - 04:00 PM)
Please wait before rushing...new alpha will follow very soon
OK, sorry!! False alarm, apparently.
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.