IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
Problem sample for MPC 1.14b standard
Gecko
post Dec 5 2002, 13:28
Post #1





Group: Members
Posts: 948
Joined: 15-December 01
From: Germany
Member No.: 662



Hi, I have found a sample which does not sound transparent with the standard setting. "Angelic - Can't Keep Me Silent (Julian D'or mix)" from the "Eye Trance 03" trance sampler. The artifact occurs during the breakdown at ca 2:25 mins into the track. The dominant sharp synthy instrument sounds more noisy than in the original. Other words describing the artifact would be rough or coarse sounding. ABX results 8/8 (pretty easy imho). This sample also makes LAME 3.90.2 APS and OGG -q 6 stumble, allthough in a different fashion. This sound is of course purely synthetic but not uncommon in trance music.

Encoder version: 1.14 beta
Decoder version: 1.93h
Settings: --standard --xlevel
Avg bitrate: 201.3 kbps
A lossless clip can be found here. (14 seconds, 1.5MB)

edit: The artifact is heard during seconds 0.5 - 2.2 and 7.5 - 9.3.

This post has been edited by Gecko: Dec 6 2002, 12:32
Go to the top of the page
+Quote Post
R.A.F.
post Dec 6 2002, 10:23
Post #2





Group: Members
Posts: 281
Joined: 4-August 02
From: Nuremberg/Bavaria
Member No.: 2924



Shit! I have problems, to depack your FLAC. I downloaded Speekīs flac-frontend, the flac-file and the codec. Put the codec into the directory, where the frontend is. And tried to decode the "angelic_silent_short" - but the dos-windows just opens for the part of a second. - What am I doing wrong?


--------------------
My used codecs and settings:
FLAC V1.1.2 -4 / APE V3.99 Update 4 -high / MPC V1.15v --q 5 / LAME V3.97b2 -V2 --vbr-new / OGG aoTuV V4.51 Lancer -q5
Go to the top of the page
+Quote Post
R.A.F.
post Dec 6 2002, 10:51
Post #3





Group: Members
Posts: 281
Joined: 4-August 02
From: Nuremberg/Bavaria
Member No.: 2924



Ok, I managed to depack it. Seems, that the frontend does not run properly on my system.

But now to the artifact. Well, I encoded the sample in all quality-steps (--q5, q6, q7, q8), but couldnīt hear any differences to the original. But, maybe my equipment or my ears are too bad. It would also be helpful, if you would give the time-stamp in the sample (in seconds from beginning), where the artifact appears. "2:25 Mins." isnīt very helpful, when the sample itself is only 14 seconds long .... wink.gif

This post has been edited by R.A.F.: Dec 6 2002, 10:59


--------------------
My used codecs and settings:
FLAC V1.1.2 -4 / APE V3.99 Update 4 -high / MPC V1.15v --q 5 / LAME V3.97b2 -V2 --vbr-new / OGG aoTuV V4.51 Lancer -q5
Go to the top of the page
+Quote Post
SometimesWarrior
post Dec 6 2002, 11:48
Post #4





Group: Members
Posts: 671
Joined: 21-November 01
From: California, US
Member No.: 514



Confirming the artifact... listening to the first three seconds, it sounds like extra fuzz has been added to the synth.

Using mppenc v1.14, mppdec v1.93j (both at defaults):

First ABX: 9/16, simply listening for a good spot on the sample to ABX
Second ABX: 12/16, would have been higher if my finger hadn't slipped on the "next trial" button a few times wink.gif

with mppenc --xtreme, I abx'd 9/16 twice, so I can make no conclusions about that.

This post has been edited by SometimesWarrior: Dec 6 2002, 11:48
Go to the top of the page
+Quote Post
Gecko
post Dec 6 2002, 12:31
Post #5





Group: Members
Posts: 948
Joined: 15-December 01
From: Germany
Member No.: 662



The artifact is heard during seconds 0.5 - 2.2 and 7.5 - 9.3. Sorry for not being more accurate in my first post. The 2:25 mins referr to the original song. Btw, I used flacdrop to encode the wav. I will test --xtreme or higher when I have more time.
Go to the top of the page
+Quote Post
JohnV
post Dec 6 2002, 13:38
Post #6





Group: Developer
Posts: 2797
Joined: 22-September 01
Member No.: 6



Ya, 5/5 in ABA.

edit:
hmm, 5/5 also with xtreme with first try. Probability of guessing 5/5 in ABA is 0.4%

This post has been edited by JohnV: Dec 6 2002, 13:47


--------------------
Juha Laaksonheimo
Go to the top of the page
+Quote Post
Gecko
post Dec 6 2002, 15:19
Post #7





Group: Members
Posts: 948
Joined: 15-December 01
From: Germany
Member No.: 662



Update:
xtreme xlevel -- ABX 10/10 p < 0.1%
insane xlevel -- ABX 21/32 p = 5.5%

I reached 15/19 on the insane sample where p = 1.0%, but then I pretty much lost it. Fatigue would be my excuse. smile.gif Same sample for over a hundred times in 15 minutes. (Yes, I do those abx tests quick and dirty)

edit: just for kicks I also tried the old 1.7.9c encoder. Sounds similar, I would even say worse than the current 1.14 beta (without xlevel, but no clipping anyway). Of course I should abx those as well, but I'm fed up with this sample atm.

This post has been edited by Gecko: Dec 6 2002, 15:25
Go to the top of the page
+Quote Post
Diocletian
post Dec 6 2002, 15:44
Post #8





Group: Members
Posts: 45
Joined: 11-October 02
Member No.: 3517



QUOTE (Gecko @ Dec 6 2002 - 07:49 PM)
Update:
xtreme xlevel -- ABX 10/10 p < 0.1%
insane xlevel -- ABX 21/32 p = 5.5%

I reached 15/19 on the insane sample where p = 1.0%, but then I pretty much lost it. Fatigue would be my excuse. smile.gif Same sample for over a hundred times in 15 minutes. (Yes, I do those abx tests quick and dirty)

edit: just for kicks I also tried the old 1.7.9c encoder. Sounds similar, I would even say worse than the current 1.14 beta (without xlevel, but no clipping anyway). Of course I should abx those as well, but I'm fed up with this sample atm.


Do you also tried the 1.93 encoder? Is this encoder better or worse than 1.14?

This post has been edited by Diocletian: Dec 6 2002, 15:45


--------------------
Diocletian

Time Travel Agency
Book a journey to the Diocletian Palace. Not today!
Go to the top of the page
+Quote Post
Seed
post Dec 6 2002, 17:07
Post #9


Musepack Project Coordinator


Group: Developer
Posts: 161
Joined: 24-June 02
Member No.: 2385



1.11b, 1.14, and 1.93j all produce ABXable samples
at --insane (10/10). Transparency is achieved halfway between --quality 7 and --quality 8 with all encoded
samples. I only tested seconds 3.7-4.2 and didn't listen
to the whole sample.


--------------------
And if Warhol's a genius, what am I? A speck of lint on the ***** of an alien
Go to the top of the page
+Quote Post
-Nepomuk-
post Dec 6 2002, 18:47
Post #10





Group: Members
Posts: 47
Joined: 20-November 02
From: Germany
Member No.: 3824



1 of 1, p = 0.500
2 of 2, p = 0.250
3 of 3, p = 0.125
4 of 4, p = 0.063
4 of 5, p = 0.188
5 of 6, p = 0.109
6 of 7, p = 0.063
7 of 8, p = 0.035
7 of 9, p = 0.090
8 of 10, p = 0.055

mpcenc 1.14 / standard

But i can't explain what the difference is. it sound less accurate.
Go to the top of the page
+Quote Post
Q!
post Dec 7 2002, 14:06
Post #11





Group: Members
Posts: 98
Joined: 26-August 02
From: PL
Member No.: 3199



QUOTE (Gecko @ Dec 6 2002 - 01:31 PM)
The artifact is heard during seconds 0.5 - 2.2

Aaaah, that helped a lot. 11/12 (1.14, stnrd)
Btw, is anyone reporting these samples to Frank or are we testing them just for fun? smile.gif


--------------------
http://qisgod.host.sk
Go to the top of the page
+Quote Post
JohnV
post Dec 7 2002, 14:35
Post #12





Group: Developer
Posts: 2797
Joined: 22-September 01
Member No.: 6



QUOTE (Q! @ Dec 7 2002 - 03:06 PM)
QUOTE (Gecko @ Dec 6 2002 - 01:31 PM)
The artifact is heard during seconds 0.5 - 2.2

Aaaah, that helped a lot. 11/12 (1.14, stnrd)
Btw, is anyone reporting these samples to Frank or are we testing them just for fun? smile.gif

I've emailed Frank. smile.gif


--------------------
Juha Laaksonheimo
Go to the top of the page
+Quote Post
Seed
post Dec 8 2002, 04:34
Post #13


Musepack Project Coordinator


Group: Developer
Posts: 161
Joined: 24-June 02
Member No.: 2385



I spent a little more time, this time I picked seconds 0.4 - 0.8 and tested with ABC/HR at --braindead. 1.11b, 1.14, and 1.93j are all ABXable (10/10). At --quality 8.5 I could not distinguish between the original and the sample encoded by 1.11b at all. With 1.14 and 1.93j I scored 7/10... This is my limit.

I was curious to see how both SV7 encoders handle the part I tested. mppenc 1.11 goes as high as 666.1 kbps for the short segment slightly after 0.5 seconds. mppenc 1.14 allocates 655.4 kb (both at --quality 8.5). What I can hear in all samples at --quality 8 and below is a change of amplitude in the right channel. Right when the synth hit plays, there's a change in modulation and a very slight change in amplitude. At --quality 8.5 it's extremely hard to notice, especially with 1.11b.

Headphones: Sennheiser HD495. Card: Diamond Monster MX400.

I hope Frank can learn something from this report.


--------------------
And if Warhol's a genius, what am I? A speck of lint on the ***** of an alien
Go to the top of the page
+Quote Post

Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 26th December 2014 - 11:14