BeSweet 'Click' At Track Beginning

[JohnMK]

The pop/click isn't in the original material, and isn't present when transcoded with HeadAC3he. If nobody else has experienced this or has a solution for it, I'll post a log file and more details.

(aside: which stereo mode is recommended for those who normally listen with 2 speakers? stereo, surround, or surround2 -- I know what they do, but I'm curious if listening to surround/surround2 out of two speakers sounds worse than simply encoding to simple stereo)

[JohnMK]

This is just one example:

You can download the resulting file here (about 800k):

http://students.washington.edu/chs4000/snippet.ogg

(I renamed it for your convenience)

btw, I get the same behavior using stable and beta versions. I'll do some more testing of my config tonight when I get home.

eSweet v1.4RC8 by DSPguru.
--------------------------
Using azid.dll v1.8 (b825) by Midas (midas@egon.gyaloglo.hu).
Using Ogg Vorbis v1.0 dlls (http://www.vorbis.com).

Logging start : 09/27/02 , 11:01:14.

C:\Documents and Settings\Chad\Desktop\BeSweetv1.4RC8\BeSweet.exe -core( -input C:\Documents and Settings\Chad\Desktop\In Progress\star trek\301\301snippet AC3 T02 2_0ch 192Kbps DELAY 27ms.ac3 -output C:\Documents and Settings\Chad\Desktop\In Progress\star trek\301\301snippet AC3 T02 2_0ch 192Kbps DELAY 27ms.3.ogg -logfilea BeSweet.log ) -azid( -L -3db -g max -s stereo ) -ota( -d 27 ) -ogg( -q 0.200 ) -profile( The OggMachine v0.5 )

[00:00:00:000] +------- BeSweet -----
[00:00:00:000] | Input : C:\Documents and Settings\Chad\Desktop\In Progress\star trek\301\301snippet AC3 T02 2_0ch 192Kbps DELAY 27ms.ac3
[00:00:00:000] | Output: C:\Documents and Settings\Chad\Desktop\In Progress\star trek\301\301snippet AC3 T02 2_0ch 192Kbps DELAY 27ms.3.ogg
[00:00:00:000] | Floating-Point Process: Yes
[00:00:00:027] +-------- AZID -------
[00:00:00:027] | Input Channels Mode: 2/0, Bitrate: 192kbps
[00:00:00:027] | Output Stereo mode: Stereo
[00:00:00:027] | Total Gain: 5.625dB, Compression: None
[00:00:00:027] | LFE levels: To LR -3.0dB, To LFE 0.0dB
[00:00:00:027] | Center mix level: BSI
[00:00:00:027] | Surround mix level: BSI
[00:00:00:027] | Dialog normalization: No
[00:00:00:027] | Rear channels filtering: No
[00:00:00:027] | Source Sample-Rate: 48.0KHz
[00:00:00:027] +-------- OGG --------
[00:00:00:027] | VBR Quality : 0.200
[00:00:00:027] +---------------------
[00:01:23.392] E7: CRC error
[00:01:23:419] Conversion Completed !
[00:01:23:419] Actual Avg. Bitrate : 84kbps
[00:00:13:000] <-- Transcoding Duration
Logging ends : 09/27/02 , 11:01:27.here: (http://students.washington.edu/chs4000/snippet.ogg) here (http://students.washington.edu/chs4000/snippet.ogg)

[JohnMK]

Winamp2 & 3. I just tried it in mplayer2 and WMP 8 and get exactly the same behavior. There's no split in the commandline because I used DVDDecrypter to rip only one chapter, and then I further reduced the segment in DVD2AVI to the first minute of that chapter.

I just had someone verify this in #vorbis on irc.xiph.org -- there is a pop. Maybe listen with the volume turned up a little bit. :) The person I had listen to it also noticed a pop at the tail end of the file. When comparing to the HeadAC3he output, it's definitely erroneous, as its output contains neither click nor pop at beginning or end.

[DSPguru]

this ?

[JohnMK]

I'm not an audiologist so I don't know what to make of that.

[DSPguru]

what about the timestamp ? is it reasonable ?

[JohnMK]

yes I guess.

[JohnMK]

Using the same AC3 source file, here's the HeadAC3he encode:

http://students.washington.edu/chs4000/snippetHead.ogg

[DSPguru]

can you split the first second of your ac3 file and attach it in here ? i wanna test it with your commandline.

[JohnMK]

I'll try. Any suggestions on how to procede?

Done.

http://students.washington.edu/chs4000/snipples.ac3

[JohnMK]

I'm going to see now if the whole episode resulting .ogg (45 minutes) contains a pop.

[DSPguru]

i now believe it got nothing to do with the delay or ogg encoding. i believe the pop is a left over from the first pass.
how about using one-pass (normalization or 5.6db) ?

[JohnMK]

By one pass, do you mean use postgain?

[DSPguru]

either postgain or a pregain of the value you can find in the logfile.

[JohnMK]

There's no pop if I encode the WHOLE episode (45 minutes) with my old settings (using gain).

In general would you not recommend Gain or "Auto Find Max Gain?"

[JohnMK]

Indeed, using post gain 5.6dB -- there's no pop at all, even in the snippet.

[DSPguru]

in general i would recommend the BeSweetGUI presets that uses a hybrid mode of pregain+dialog normalization+postgain.
if you're using OggMachine, just check the PostGain - and you're set :).

[JohnMK]

Thank you. So I guess then that there is a small bug with Gain? Did I help you at all or was this just a wild goose chase? I think you should update your guide to reflect your advice to me, so that for the time being, nobody encounters this annoying bug.

[DSPguru]

yes, there is a bug.
thank you very much for all those inputs!


* all guides recommends using BeSweetGUI's presets.

[JohnMK]

Not to be picky, but I don't think your oggmachine FAQ does, however, mention presets of any sort:

http://www.doom9.org/index.html?/oggmachine.htm

[DSPguru]

well, there aren't any presets in OggMachine, only BeSweetGUI.
in OggMachine, when selecting PostGain - it does the rest.
anyway, thank you for paying my attention to the guide, it needs to be updated :There are two gain methods, both will ensure that the audio will have the optimal volume without any distortion. First of all you should check Auto Find Maximum Gain so you won't have to guess what gain to apply. Then you can select between the regular Gain and PostGain. Gain performs 2 passes, whereas PostGain only needs one decoding pass. However, PostGain is currently only supported by WinAmp's in_vorbis plugin and the Ogg DS filters do currently not support it (though such support is forthcoming).well, it's already supported.. :)

[JohnMK]

DSPGuru I brought this up before with you and we determined that the 'pop' that occurs at the beginning of a 'Gain'-ed AC3 file is caused by an incomplete movie rip, such that occurs when you rip half the chapters for the first CD, and half the chapters for the second CD. Your suggestion at the time was "Don't use Gain" then. I'm hoping you can fix this, as I've been using 'Post-Gain for several movies now and just now realizes this introduces clipping. Have you included a fix for this in your internal alphas, or have you been too busy with real life things? :)

Thank you,

John

[DSPguru]

found it, fixed it, didn't have time to make a release.

now, what clippings are you talking about exactly ?
(don't forget to post relevant logfiles)

[JohnMK]

BeSweet v1.4 by DSPguru.
--------------------------
Using azid.dll v1.8 (b825) by Midas (midas@egon.gyaloglo.hu).
Using Ogg Vorbis v1.0 dlls (http://www.vorbis.com).

Logging start : 10/10/02 , 15:49:53.

BeSweet.exe -core( -input C:\Documents and Settings\Chad\Desktop\In Progress\Amadeus2\2 AC3 T02 2_0ch 192Kbps DELAY -246ms.ac3 -output C:\Documents and Settings\Chad\Desktop\In Progress\Amadeus2\2 AC3 T02 2_0ch 192Kbps DELAY -246ms.ogg -logfilea BeSweet.log ) -azid( -L -3db -g 10db -n1 -c light -s stereo ) -ota( -d -246 -G max ) -ogg( -q 0.400 ) -profile( The OggMachine v0.5 )

[00:00:00:000] +------- BeSweet -----
[00:00:00:000] | Input : C:\Documents and Settings\Chad\Desktop\In Progress\Amadeus2\2 AC3 T02 2_0ch 192Kbps DELAY -246ms.ac3
[00:00:00:000] | Output: C:\Documents and Settings\Chad\Desktop\In Progress\Amadeus2\2 AC3 T02 2_0ch 192Kbps DELAY -246ms.ogg
[00:00:00:000] | Floating-Point Process: Yes
[00:00:00:-246] +-------- AZID -------
[00:00:00:-246] | Input Channels Mode: 2/0, Bitrate: 192kbps
[00:00:00:-246] | Output Stereo mode: Stereo
[00:00:00:-246] | Total Gain: 10.000dB, Compression: Light
[00:00:00:-246] | LFE levels: To LR -3.0dB, To LFE 0.0dB
[00:00:00:-246] | Center mix level: BSI
[00:00:00:-246] | Surround mix level: BSI
[00:00:00:-246] | Dialog normalization: Yes
[00:00:00:-246] | Rear channels filtering: No
[00:00:00:-246] | Source Sample-Rate: 48.0KHz
[00:00:00:-246] +-------- OGG --------
[00:00:00:-246] | VBR Quality : 0.400
[00:00:00:-246] +---------------------
[00:01:00.485] W7: Downmix overflow (1: +0.1dB)
[00:01:00.490] W7: Downmix overflow (1: +0.4dB)
[00:01:00.496] W7: Downmix overflow (1: +0dB)
[00:01:00.517] W7: Downmix overflow (1: +0.5dB)
[00:01:00.581] W7: Downmix overflow (1: +0.4dB)
[00:01:00.650] W7: Downmix overflow (1: +0dB)
[00:01:01.813] W7: Downmix overflow (0: +0.3dB)
[00:01:01.909] W7: Downmix overflow (1: +1.0dB)
[00:01:02.128] W7: Downmix overflow (1: +0.2dB)
[00:01:02.133] W7: Downmix overflow (1: +0dB)
[00:01:02.682] W7: Downmix overflow (0: +0.1dB)
[00:01:02.730] W7: Downmix overflow (0: +0.6dB)

Those 'downmix overflow' messages continued for perhaps hundreds of pages.

[DSPguru]

yea, i've noticed that 2/0 streams are usually louder than 3/2 with ds downmix.

the particualr solution for those streams would be the use of -g8db instead of -g10db in azid's section.

i might offer an automatic solution for this, so when 2/0 streams are transcoded, a softer gain value would be used.

[JohnMK]

I don't see a way to affect the -g8db change you recommend. There's no place for that in OggMachine. At this point, I'm damned if I do and damned if I don't.

I think I'll weather the small little click. It's rather minor, and only affects my 2 CD encodes. Will using Gain absolutely certainly prevent any clipping whatsoever, with the way I use OggMachine?

[DSPguru]

no, you shouldn't be satisfied with it!
in this case, it would be better to live with a sigle tick at the beginning (two-pass), rather than with lots of ticks in a one-pass with 2/0 stream & 10db gain.

anyway,
i will be offering a new switch, probably -ota( -postgain ) that will use 10db for 3/2 and 7db for 2/0.

[JohnMK]

Does the single click phenomenon throw off the sync of the track by any milliseconds at all?

[JohnMK]

Can I be sure that if I encode 2.0 channel tracks, I'm not getting clipping for the meantime, as long as I'm using Gain, and not PostGain?

[DSPguru]

you'll never get overflows when using a two-pass, except this 'click' on the start you mentioned.
no drift in synchronization will occur.

[JohnMK]

Thank you. You're very generous with your help. I appreciate it. :cool:

[JohnMK]

Frankly all this gaining stuff worries me. I still am not certain that my movies don't come out clipping. If DSPGuru could assure me that his tools produce absolutely positively 100% clipping-free results, I'd appreciate it. :D

[DSPguru]

Originally posted by JohnMK
Frankly all this gaining stuff worries me. I still am not certain that my movies don't come out clipping. If DSPGuru could assure me that his tools produce absolutely positively 100% clipping-free results, I'd appreciate it. :D for 3/2 streams, the default profile is perfect, no clips are expected (and you can always check the logfile).
for 2/0 streams, i suggest to go for only postgain or only pregain (without dialog norm), and not use the famous hybrid preset.
anyway, i plan to offer a new switch -ota( -hybrid ) that automaticly activates the gain (pregain,postgain,dialog norm) according to input type (2/0 or 3/2).

also, as you know in (the unpublished) v1.5b1, there's a fix for the sporadic 'tick' in the very beginning of the file.

[JohnMK]

Presets schmpresheps. I don't use them. I use OggMachine. :D I think a lot of people probably do. The improved behavior should be default in OggMachine, IMO. The ugly details of how to avoid clipping, etc., shouldn't be exposed to the end user. Making a faulty output file that's clipped like mad and ticking at the beginning absolutely shouldn't be an option in the BeSweet + OggMachine combo, so please do more than implement an obscure switch that simpletons like me won't activate. Just make it default behavior. :)

Thanks.

[JohnMK]

By the way, have all my movies I've made using Gain + DNR (Dialog N. Reduction) been made junk, according to your latest advice? If I have clipped output, would the log file let me know, or would it slip by unnoticed, just waiting for my bad ears to hear it?

DNR is default behavior for Gain in OggMachine. Is this good or bad? I still don't understand at all what DNR is. Could you try one more time, except use extreme layman's terms? :D

Sorry to be so repetitive, but I never interpret what people say literally. Things just never seem clear to me, I read too much into the possible interpretations of 'may' 'shall' 'should' 'can.' Etc.

[DSPguru]

- overflows, if exists, can be found in logfile.
- dialog normalization is used to decrease the total output level, in order to supply "same" loudness for different tracks.
- you should remember that after all, i am a busy man, and i can't release a new version on a daily basis. i said i have a transperent solution for all the gaining issues and different ac3 streams, but i didn't find the time to publish it. please keep your doubts to after i release the new version.
- a new OggMachine would be inroduced as well.

[DSPguru]

v1.5b1 is out. time for your testings and feedbacks.