IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
gt3b2 better than aotuvb5 @ q6?!?
dcote
post Dec 5 2007, 17:10
Post #1





Group: Members
Posts: 8
Joined: 30-November 07
Member No.: 49146



hi guys!

first off: i am new to this forum so if i am asking stupid questions, please be gentle with me! ;-)

i have been an avid user of the gt3b2 vorbis encoder for a long time now. i use it at q6 and find it gives outstanding results. as a matter of fact, i find it extremely difficult to ABX a difference between it and the original wav file! :-)
(whoever wrote gt3b2 -> great job!)

but knowing it is very old i checked here (THE authority in music encoding ;-) for any new developments.

sure enough: aotuvb5 is HA's recommended vorbis encoder. no prob - i happily downloaded and tested it (also @ q6) vs. gt3b2.

then my surprise: there was a clear difference between the two, which i could clearly hear in an ABX test on some more "tricky" music pieces:

WinABX v0.42 test report
11/30/2007 11:40:24

A file: C:\Temp\Ripped Files\Iaora Tahiti.02.Mouse on Mars.Kompod_aotuvb5.wav
B file: C:\Temp\Ripped Files\Iaora Tahiti.02.Mouse on Mars.Kompod_GT3B2.wav

Start position 00:00.0, end position 04:08.4
11:47:36 1/1 p=50.0%
11:48:18 2/2 p=25.0%
11:48:45 3/3 p=12.5%
11:49:27 3/4 p=31.2%
11:50:06 4/5 p=18.8%
11:51:07 4/6 p=34.4%
11:52:04 5/7 p=22.7%
11:53:11 6/8 p=14.5%
11:54:14 7/9 p=9.0%


WinABX v0.42 test report
11/30/2007 11:55:06

A file: C:\Temp\Ripped Files\Iaora Tahiti.06.Mouse on Mars.Kanu_aotuvb5.wav
B file: C:\Temp\Ripped Files\Iaora Tahiti.06.Mouse on Mars.Kanu_GT3B2.wav

11:55:29 1/1 p=50.0%
11:55:50 2/2 p=25.0%
11:56:09 2/3 p=50.0%
11:56:44 2/4 p=68.8%
11:57:18 3/5 p=50.0%
11:57:45 4/6 p=34.4%
11:58:44 4/7 p=50.0%
11:59:22 5/8 p=36.3%
12:00:05 5/9 p=50.0%
12:00:27 6/10 p=37.7%

WinABX v0.42 test report
11/30/2007 12:00:56

A file: C:\Temp\Ripped Files\Unison.03.Angels & Agony.Unison_aotuvb5.wav
B file: C:\Temp\Ripped Files\Unison.03.Angels & Agony.Unison_GT3B2.wav

12:01:15 1/1 p=50.0%
12:01:38 1/2 p=75.0%
12:02:28 2/3 p=50.0%
12:02:44 3/4 p=31.2%
12:03:52 4/5 p=18.8%
12:04:10 5/6 p=10.9%
12:04:40 6/7 p=6.2%
12:05:40 7/8 p=3.5%
12:06:01 8/9 p=2.0%
12:06:12 9/10 p=1.1%


WinABX v0.42 test report
11/30/2007 12:06:38

A file: C:\Temp\Ripped Files\Zephyr- 48_24_aotuvb5.wav
B file: C:\Temp\Ripped Files\Zephyr- 48_24_GT3B2.wav

12:07:09 1/1 p=50.0%
12:07:53 2/2 p=25.0%
12:08:19 3/3 p=12.5%
12:08:34 4/4 p=6.2%
12:09:06 4/5 p=18.8%
12:09:29 5/6 p=10.9%
12:10:22 6/7 p=6.2%
12:10:55 6/8 p=14.5%
12:13:16 7/9 p=9.0%
12:13:56 8/10 p=5.5%


so far so good. BUT: to me gt3b2 sounds BETTER than aotuvb5?!?!?

in my ears:

GT3B2:
more "open", better staging/positioning (like in a concert hall) and delightful lightness.
but a slight tendency to "muddy" sound and seems to make some passages louder (dynamic loss?).

AoTuVb5:
somehow "darker", more "closed" sounding, like in a small room, more restrained.
however, instruments and voices seemed to sound more precise/"clean" and there was less apparent loudening of certain passages.

are there some special settings i need to do to aotuvb5 i simply missed? (yes, i read the wiki) or is it my choice of tracks? (most is 16bit/44kHz electronic, but "zephyr" is a 24bit/48kHz acapella choir) is there a possibility that the decoder i use (a squeezebox 3 with optical dig. out) works better with gt3b2 than with aotuvb5?

thanks for your thoughts, guys! :-)
Go to the top of the page
+Quote Post
benski
post Dec 5 2007, 17:56
Post #2


Winamp Developer


Group: Developer
Posts: 670
Joined: 17-July 05
From: Brooklyn, NY
Member No.: 23375



2 things:

Are the bitrate identical? The quality scale will only have meaning for a particular encoder. You should adjust the quality for each encoder until the bitrates come identically.

Can you ABX against the original WAV file? Perhaps what you're hearing is actually flaws in the gt3b2 encoder. ABX'ing two lossy files doesn't make a whole lot of sense.
Go to the top of the page
+Quote Post
john33
post Dec 5 2007, 18:10
Post #3


xcLame and OggDropXPd Developer


Group: Developer
Posts: 3763
Joined: 30-September 01
From: Bracknell, UK
Member No.: 111



While the above may well be true, Garf focussed on the higher quality settings with GT3b2 whereas Aoyumi has focussed mainly on the lower quality settings in the aoTuV versions.


--------------------
John
----------------------------------------------------------------
My compiles and utilities are at http://www.rarewares.org/
Go to the top of the page
+Quote Post
Alex B
post Dec 5 2007, 19:34
Post #4





Group: Members
Posts: 1303
Joined: 14-September 05
From: Helsinki, Finland
Member No.: 24472



As benski said, it would be essential to compare the encoded files with the original and try to find out which one is closer. Possibly dcote is just accustomed with the "gt3b2 sound".

gt3b2 used to be pretty good (I too used it at q7 and so), but I think the recent aoTuV builds (b4 and later) have a slight edge also at higher bitrates. (I'll try to search some references to back up this).


--------------------
http://listening-tests.freetzi.com
Go to the top of the page
+Quote Post
dcote
post Dec 6 2007, 17:50
Post #5





Group: Members
Posts: 8
Joined: 30-November 07
Member No.: 49146



hi guys!

thanks for your thoughts.

you are right - the bitrates are not identical. gt3b2 generally has a ~10% higher nominal and average bitrate.

and gt3b2 has especially high bitrate fluctuations. in some of those tracks, the bitrate soars to ~420 kbit!!! whereas aotuvb5 is more restrained and hardly goes beyond 280 kbit. (would that explain the more "restrained" sound?)
that that alone could account for the fact that it sounds better to me, couldnt it?

but i thought that the main reason vorbis uses a quality setting instead of bitrates is supposed to ensure compararability, since vorbis is natively VBR. at least, that is what xiph.org says...
that is why i purposely compared at the same q=6 instead of trying to find the same bitrate.
but out of curiosity, i discovered that (for those tracks) both have about the same average bitrate at: aotuvb5=q6.7 gt3b2=q6...

and you are right, i did not compare the vorbis versions with the wav originals yet. since i already found the extreme differences between the two, i didnt think it was necessary.

i am starting to think alex b could be right - i am used to the "gt3b2 sound". one of these days, i will try to ABX vs. the original and see which one is closer.

maybe i'll even discover that i LIKE gt3b2's flaws. ;-)
Go to the top of the page
+Quote Post
Jens Rex
post Dec 6 2007, 18:50
Post #6





Group: Super Moderator
Posts: 605
Joined: 18-December 01
Member No.: 680



ABX'ing two lossy files doesn't make any sense. ABX can only be used to prove a difference between files - you can't use it to tell which of two files sound better. Indeed you've proven there is a difference between gt3b2 and autuvb5.

For any ABX test to be usable in this scenario, you must compare each Vorbis file with the original WAV.
Go to the top of the page
+Quote Post
Seiitsu
post Dec 6 2007, 18:57
Post #7





Group: Members
Posts: 82
Joined: 26-January 07
Member No.: 40026



QUOTE (JensRex @ Dec 6 2007, 18:50) *
ABX'ing two lossy files doesn't make any sense. ABX can only be used to prove a difference between files - you can't use it to tell which of two files sound better. Indeed you've proven there is a difference between gt3b2 and autuvb5.

For any ABX test to be usable in this scenario, you must compare each Vorbis file with the original WAV.

Indeed, all this test has proven is that the files don't sound the same. Comparing both files with the wav should prove which is closer to the original, and therefor better.
Go to the top of the page
+Quote Post
[JAZ]
post Dec 6 2007, 22:05
Post #8





Group: Members
Posts: 1797
Joined: 24-June 02
From: Catalunya(Spain)
Member No.: 2383



QUOTE (JensRex @ Dec 6 2007, 19:50) *
ABX'ing two lossy files doesn't make any sense. [...] Indeed you've proven there is a difference between gt3b2 and autuvb5.

So... it makes sense, or it does not?

QUOTE (JensRex @ Dec 6 2007, 19:50) *
For any ABX test to be usable in this scenario, you must compare each Vorbis file with the original WAV.


That's an ABCHR test, not an ABX test.


Not wanting to be too rude, but better have things clear. If he didn't find any difference, the ABC-HR test wouldn't be needed for him to justify an election.
Go to the top of the page
+Quote Post
dcote
post Dec 7 2007, 10:17
Post #9





Group: Members
Posts: 8
Joined: 30-November 07
Member No.: 49146



i will run a test vs. the orginal wav files soon, just to make sure. and, as you guys rightly pointed out, that would show me which comes closer to the original. and, of course, any differences introduced by vorbis.

wierd thing is, i tried an ABX vs. the original many months (years?) ago to determine wich q-level would be good enough for me and if i would use gt3b1/2 or xiph reference.
i did this unti my "guessing probability" consistently went p>50%.
turns out gt3b2 (At the time actually b1) sounded "transparent" to me @ q6, whereas xiph reference only @ q8. stll allowing for gt3b1/2's higher bitrates, the compression was better, so i went with it.
aotuvb at the time (it was said) was primarily optimized for lower bitrates, which was why i didnt even bother with it.

@JAZ: thanks for clarifying. maybe i need to as well: i actually performed an ABXY test. becasue of that, there is indeed a chance that i was only picking the encoder whose sound i "knew" better from experience. ;-)

bearing that in mind - if i am looking for the encoder that *subjectively* sounds best to me (including any and all artifacts and problems), i would still ABX the two *lossy* files, no?

i am starting to think that there could be a difference between *objective* and *subjective* quality here. quite similar to the way a tube amp sounds better to my ears than any solidstate amp, even though it's tech specs are almost always inferior.
to me, "musicality" is at least as important as "objectivity" and "precision"... ;-)
Go to the top of the page
+Quote Post
Aoyumi
post Dec 14 2007, 12:45
Post #10





Group: Members
Posts: 236
Joined: 14-January 04
From: Kanto, Japan
Member No.: 11215



QUOTE (Alex B @ Dec 6 2007, 03:34) *
As benski said, it would be essential to compare the encoded files with the original and try to find out which one is closer. Possibly dcote is just accustomed with the "gt3b2 sound".

gt3b2 used to be pretty good (I too used it at q7 and so), but I think the recent aoTuV builds (b4 and later) have a slight edge also at higher bitrates. (I'll try to search some references to back up this).

This?
MPC vs VORBIS vs MP3 vs AAC at 180 kbps, 2nd checkup with classical music (2005)
http://www.hydrogenaudio.org/forums/index....showtopic=36465


And other...
listening test at 160 kbps, pre-echo with aoTuV, GT3... (2004)
http://www.hydrogenaudio.org/forums/index....showtopic=22495

MPC vs OGG VORBIS vs MP3 at 175 kbps, listening test on non-killer samples (2004)
http://www.hydrogenaudio.org/forums/index....showtopic=23355
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: 20th December 2014 - 00:01