vorbis 1.0 default channel couplings

[canadian_fbi]

just wondering... does anyone know at what bitrates the channel coupling changes in vorbis 1.0? i know lossless now starts at -q 6, but i was more curious as to where light lossy ends now, for surround encoding implications.

[DSPguru]

coupling :
http://www.xiph.org/archives/vorbis-dev/200207/0002.html

and if you're interested in setting a lowpass filter in BeSweet, you should know that SuperEQU will give you amazing quality.
lowpass, highpass, parametric, whatever.. :D.

[canadian_fbi]

interesting... i tried reading the documentation just now on it to see if i could find any more info on it, but didn't get anything solid that would answer it definitively. i see from the link you gave that darkavenger has the same concern i had about dolby surround info being preserved... and i'm not sure if dolby surround 2 has greater demands on preserving phase information either. i don't have a ds2 receiver to test anything on though.

[canadian_fbi]

well i've gotten off the channel coupling obsession :) and i'm testing out the superequ plugin now... it seems like a pretty powerful tool though kind of confusing. so as i'm using it, you have to download the winamp plugin, create a .eq file with the settings you want, then use the -shibatch switch and -equ to load it? (doing this manually since it's not in the oggmachine/besweet gui). i'm also using -q 1 in the plugin, since i don't see why one wouldn't use the highest quality conversion. if i'm doing this all wrong, let me know :)

edit: while decoding the test files to do some quick frequency analysis on them, i noticed that superequ changes the number of samples in the raw wave file...
OggDec v1.33 (libVorbis I 20020711) Compiled on: Jul 11 2002
Made with BeSweet. http://DSPguru.doom9.org
LWING_GAIN=01.00
Encoder Version: 0
Serial Number: 41

Bitstream is 2 channel, 44100Hz
Scale = 1.0000
Decoded length: 2590710 samples = 0:58 mins.
Encoded by: Xiph.Org libVorbis I 20020711
Decoding: c:\music\audiograbbed\10 khz test (no lowpass).ogg

100% decoded.

OggDec v1.33 (libVorbis I 20020711) Compiled on: Jul 11 2002
Made with BeSweet. http://DSPguru.doom9.org
LWING_GAIN=01.00
Encoder Version: 0
Serial Number: 41

Bitstream is 2 channel, 44100Hz
Scale = 1.0000
Decoded length: 2589636 samples = 0:58 mins.
Encoded by: Xiph.Org libVorbis I 20020711
Decoding: c:\music\audiograbbed\10 khz test.ogg
this has me somewhat nervous about what other changes are going on with the equalizer...

[canadian_fbi]

more interesting lowpass tidbits:

i tried using headac3he's lowpass feature to compare against besweet with the superequ plugin... well, for some reason, headac3he's file was 107 kbps vs. 113 for besweet. (possibly because of the newer libvorbis.dll used for besweet?) headac3he was also the correct number of samples. finally, here are some gifs of the frequency analysis though, and you can see what i was talking about with headac3he's lowpass not being a very solid cutoff...

headac3he: http://www.disconnect.net/temp/hac3_lowpass.gif
besweet: http://www.disconnect.net/temp/bs_lowpass.gif

at the moment i like besweet's lowpass better, though i am concerned as to how exactly it's modifying the source file to accomplish this.

[DSPguru]

please... :logfile:

[DSPguru]

Originally posted by canadian_fbi
more interesting lowpass tidbits:

i tried using headac3he's lowpass feature to compare against besweet with the superequ plugin... well, for some reason, headac3he's file was 107 kbps vs. 113 for besweet. (possibly because of the newer libvorbis.dll used for besweet?) could be..headac3he was also the correct number of samples. finally, here are some gifs of the frequency analysis though, and you can see what i was talking about with headac3he's lowpass not being a very solid cutoff...isn't SuperEQU amazing :D ?

[kxy]

With HeadAc3He, this is similar problem that trg100 reported a few weeks back.

trg100's report (http://members.aol.com/trg100uk/ds_tests_2.html#frank)

And this thread here: click (http://forum.doom9.org/showthread.php?s=&threadid=27936)

Any explanations?

[canadian_fbi]

you want the logfiles for the two runs i described earlier? (where the number of samples is different)... i believe these are the ones:

BeSweet v1.4RC3 by DSPguru.
--------------------------
Using Shibatch.dll v0.12 by Naoki Shibata & DSPguru (shibatch.sourceforge.net).
Using libVorbis.dll v1.0 ( Jul 11 2002 ) by John33 (www.inf.ufpr.br/~rja00).

Logging start : 07/16/02 , 11:46:44.

BeSweet.exe -core( -input C:\Music\audiograbbed\tensiontest.wav -output C:\Music\audiograbbed\10 khz test.ogg -logfilea BeSweet.log ) -ota( -g 0db ) -ogg( -q 0.400 ) -shibatch( -equ 10000 lowpass.eq -q 1 )

[00:00:00:000] +------- BeSweet -----
[00:00:00:000] | Input : C:\Music\audiograbbed\tensiontest.wav
[00:00:00:000] | Output: C:\Music\audiograbbed\10 khz test.ogg
[00:00:00:000] | Floating-Point Process: Yes
[00:00:00:000] +------ Shibatch -----
[00:00:00:000] | Equ Preset File : 10000 lowpass.eq
[00:00:00:000] | Equ Quality Level : 1.0
[00:00:00:000] | Source Sample-Rate: 44.1KHz
[00:00:00:000] +-------- OGG --------
[00:00:00:000] | VBR Quality : 0.400
[00:00:00:000] +---------------------
[00:00:58:746] Conversion Completed !
[00:00:58:746] Actual Avg. Bitrate : 113kbps
[00:00:09:000] <-- Transcoding Duration
Logging ends : 07/16/02 , 11:46:53.

BeSweet v1.4RC3 by DSPguru.
--------------------------
Using libVorbis.dll v1.0 ( Jul 11 2002 ) by John33 (www.inf.ufpr.br/~rja00).

Logging start : 07/16/02 , 11:48:10.

BeSweet.exe -core( -input C:\Music\audiograbbed\tensiontest.wav -output C:\Music\audiograbbed\10 khz test (no lowpass).ogg -logfilea BeSweet.log ) -ota( -g 0db ) -ogg( -q 0.400 )

[00:00:00:000] +------- BeSweet -----
[00:00:00:000] | Input : C:\Music\audiograbbed\tensiontest.wav
[00:00:00:000] | Output: C:\Music\audiograbbed\10 khz test (no lowpass).ogg
[00:00:00:000] | Floating-Point Process: Yes
[00:00:00:000] | Source Sample-Rate: 44.1KHz
[00:00:00:000] +-------- OGG --------
[00:00:00:000] | VBR Quality : 0.400
[00:00:00:000] +---------------------
[00:00:58:745] Conversion Completed !
[00:00:58:745] Actual Avg. Bitrate : 130kbps
[00:00:08:000] <-- Transcoding Duration
Logging ends : 07/16/02 , 11:48:18.


i was wondering... would it take very long to include in besweet a switch that would enable the lowpass feature in vorbis? i don't know if it would be a better option than superequ, but it would definitely be faster. as long as it's there, might as well include the option... but i have no idea whether this is feasible or even worth the time.

p.s.: when you enable surround2 on a two-channel AC3, it seems to garble the audio. just encountered this myself - don't know if it's already been mentioned. switching it to stereo solves the problem.

[DSPguru]

yea, when using SuperEQU, the very last samples were missing.
just fixed it.

about your suggestion,
i don't see why should i add inferior features to BeSweet after putting quite some time on adding this super-dooper equalizer to my tool.
if you want more speed/less quality, use -q0.2/-q0.4 ...

oh, and thank you for all your inputs !
Dg.

[canadian_fbi]

Originally posted by DSPguru
yea, when using SuperEQU, the very last samples were missing.
just fixed it.
ok, thanks :)

about your suggestion,
i don't see why should i add inferior features to BeSweet after putting quite some time on adding this super-dooper equalizer to my tool.
if you want more speed/less quality, use -q0.2/-q0.4 ...

well, the thing about the vorbis lowpass is that it theoretically should increase the speed of a lowpass without really reducing the quality, and without modifying the original signal. and as i understood it, the feature was already there, it just needed to be invoked from a besweet switch. whether or not this would actually be better than superequ would be doubtful, but it would just be another option to do it inside the codec. but if this would take too long, you're under no pressure from me to do it since the functionality's already there in the form of the equalizer and in the form of headac3he supporting it. i admit when i suggested it i didn't know that the equalizer could do all that; it's pretty handy. :) and it did shave about 1.5% of the size off the ogg vorbis file, and i shouldn't be able to notice it.

[DSPguru]

you're invited to post some fft graphs (+logfiles!) when using -q0.2 / -q0.4

[canadian_fbi]

are you referring to the quality setting on vorbis or on superequ? the tests i did today were -q 0.4 for vorbis and -q 1.0 for superequ. i'm not really interested in going below 0.4 for vorbis because of potential channel coupling problems, low quality, and the fact that it starts to impose a lowpass on its own.

[DSPguru]

you wanted to use Vorbis' inner, low quality, lp filter. so i'm suggesting that you lower the quality of sequ with lowering the -q switch in -shibatch().

[canadian_fbi]

ok, i'll try that. i'll have to test with frequency graphs and bitrates primarily though, as i don't really trust my ears to be able to detect difference in lowpasses (unless they're pretty severe).

[canadian_fbi]

ugh... i hate to pester you yet again, but when muxing the final ogm with the lowpassed audio, it's out of sync. i muxed one with the non-lowpassed version and it's better. maybe this has something to do with the missing samples you mentioned earlier - could i try the fixed dll and see if that solves the problem? i cut an example from both to show the sync loss if you want to see. here's the output of oggdec when decoding the two files:

OggDec v1.33 (libVorbis I 20020711) Compiled on: Jul 11 2002
Made with BeSweet. http://DSPguru.doom9.org
LWING_GAIN=01.00
Encoder Version: 0
Serial Number: 41

Bitstream is 2 channel, 48000Hz
Scale = 1.0000
Decoded length: 264237984 samples = 91:44 mins.
Encoded by: Xiph.Org libVorbis I 20020713
Decoding: clerks audio q 4.00.ogg

100% decoded.

OggDec v1.33 (libVorbis I 20020711) Compiled on: Jul 11 2002
Made with BeSweet. http://DSPguru.doom9.org
LWING_GAIN=01.00
Encoder Version: 0
Serial Number: 41

Bitstream is 2 channel, 48000Hz
Scale = 1.0000
Decoded length: 264225024 samples = 91:44 mins.
Encoded by: Xiph.Org libVorbis I 20020713
Decoding: clerks audio q 4.00 lowpass 17.5.ogg

100% decoded.


it looks like there's .27 seconds missing from the lowpassed file... that would be about the amount that it's out of sync.
and here's the logfiles, if necessary:

BeSweet v1.4RC3 by DSPguru.
--------------------------
Using azid.dll v1.8 (b825) by Midas (midas@egon.gyaloglo.hu).
Using Shibatch.dll v0.12 by Naoki Shibata & DSPguru (shibatch.sourceforge.net).
Using libVorbis.dll v1.0 ( Jul 13 2002 ) by John33 (www.inf.ufpr.br/~rja00).

Logging start : 07/16/02 , 13:16:21.

BeSweet.exe -core( -input C:\Video\DVD ripping\clerks AC3 T01 2_0ch 192Kbps DELAY -66ms.ac3 -output C:\Video\DVD ripping\clerks audio q 4.00 lowpass 17.5.ogg -logfilea BeSweet.log ) -azid( -g max -c normal -s stereo ) -ota( -d -66 ) -ogg( -q 0.400 ) -shibatch( -equ 17500 lowpass.eq -q 1 )

[00:00:00:000] +------- BeSweet -----
[00:00:00:000] | Input : C:\Video\DVD ripping\clerks AC3 T01 2_0ch 192Kbps DELAY -66ms.ac3
[00:00:00:000] | Output: C:\Video\DVD ripping\clerks audio q 4.00 lowpass 17.5.ogg
[00:00:00:000] | Floating-Point Process: Yes
[00:00:00:-66] +-------- AZID -------
[00:00:00:-66] | Output Stereo mode: Stereo
[00:00:00:-66] | Total Gain: 3.7dB, Compression: Normal
[00:00:00:-66] | LFE levels: To LR -INF, To LFE 0.0dB
[00:00:00:-66] | Center mix level: BSI
[00:00:00:-66] | Surround mix level: BSI
[00:00:00:-66] | Dialog normalization: No
[00:00:00:-66] | Rear channels filtering: No
[00:00:00:-66] +------ Shibatch -----
[00:00:00:-66] | Equ Preset File : 17500 lowpass.eq
[00:00:00:-66] | Equ Quality Level : 1.0
[00:00:00:-66] | Source Sample-Rate: 48.0KHz
[00:00:00:-66] +-------- OGG --------
[00:00:00:-66] | VBR Quality : 0.400
[00:00:00:-66] +---------------------
[01:31:44:638] Conversion Completed !
[01:31:44:638] Actual Avg. Bitrate : 110kbps
[00:18:52:000] <-- Transcoding Duration
Logging ends : 07/16/02 , 13:35:13.

BeSweet v1.4b14 by DSPguru.
--------------------------
Using azid.dll v1.8 (b825) by Midas (midas@egon.gyaloglo.hu).
Using libVorbis.dll v1.0 ( Jul 13 2002 ) by John33 (www.inf.ufpr.br/~rja00).

Logging start : 07/13/02 , 20:51:10.

BeSweet.exe -core( -input C:\Video\DVD ripping\clerks AC3 T01 2_0ch 192Kbps DELAY -66ms.ac3 -output C:\Video\DVD ripping\clerks audio q 4.00.ogg -logfilea BeSweet.log ) -azid( -g max -c normal -s surround2 ) -ota( -d -66 ) -ogg( -q 0.400 ) -profile( The OggMachine v0.5 )

[00:00:00:000] +------- BeSweet -----
[00:00:00:000] | Input : C:\Video\DVD ripping\clerks AC3 T01 2_0ch 192Kbps DELAY -66ms.ac3
[00:00:00:000] | Output: C:\Video\DVD ripping\clerks audio q 4.00.ogg
[00:00:00:000] | Floating-Point Process: Yes
[00:00:00:-66] +-------- AZID -------
[00:00:00:-66] | Output Stereo mode: Dolby surround compatible
[00:00:00:-66] | Total Gain: 3.7dB, Compression: Normal
[00:00:00:-66] | LFE levels: To LR -INF, To LFE 0.0dB
[00:00:00:-66] | Center mix level: BSI
[00:00:00:-66] | Surround mix level: BSI
[00:00:00:-66] | Dialog normalization: No
[00:00:00:-66] | Rear channels filtering: No
[00:00:00:-66] | Source Sample-Rate: 48.0KHz
[00:00:00:-66] +-------- OGG --------
[00:00:00:-66] | VBR Quality : 0.400
[00:00:00:-66] +---------------------
[01:31:44:958] Conversion Completed !
[01:31:44:958] Actual Avg. Bitrate : 112kbps
[00:16:38:000] <-- Transcoding Duration
Logging ends : 07/13/02 , 21:07:48.