AC3 5.1 to DPL2 downmix with Besweet.... Help

[Ody1]

I want to downmix a 5.1 AC3 file to a 2 channel file in mp3 format with Dolby Prologic 2. The .1 channel must be in the final mp3. I don't want any kind of dynamic compresion, just a simple normalization.

What options should I use in Besweet? The sound comes from a DVD and the final mp3 will be muxed in the avi. The sample rate should stay in 48000. I'll use cbr. Can I use joint srereo in Lame, or it will ruin the phase shift and the surround channels?

I tried to encode this AC3 once but the mp3 was almost full silence. Only now and then you could hear some sounds, with very low volume, for a few seconds... I don't have the log file. I used:

Azid 1
------
LFE to LR Channels: 0db
Stereo: Surround2

Lame 1
------
Mode: Joint Stereo
Alt Preset: cbr 112

Lame 2
------
Error protection

Lame 2
------
Scale Input (Multiply PCM Data) by: 1

BeSweet
-------
Lame: Wave-MP3

Apparently these settings didn't get the job done. What settings should I use in Besweet?

Also how can I create a 5.1 AC3 file from 6 mono wavs? I want to do that so I can do some tests with the various settings in Besweet. I created a 5.1 wav with Cool edit pro 2.1 but Besweet said that the input file was an unregogized format (or something similar). Is there any other program that easily creates 5.1 AC3 from 6 mono wavs?

[tiki4]

1) besweet -core( -input in.ac3 -output out.wav -logfile BeSweet.log -wavmp3 ) -azid( -c normal -s surround2 -L -3db ) -ota( -hybridgain ) -lame( --alt-preset cbr 160 )

-> above command line does normalization and dynamic range compression which you should do (it's done when you decode the original AC3 in your software or standalone DVD player).
-> encoding is done at 160 kBit/s constant bitrate, but better use variable bitrate with '--alt-preset standard', then you must mux with Nandub or VirtualDubMod.

2) Sonic Foundry SoftEncode, but that isn't sold anymore, so don't ask me where to get it.

Regards,

tiki4

[Ody1]

Thanks, I'll try with the command line you sugest.
I'm just starting to understand Besweet. I have a few questions...


Would there be a problem if I set "LFE to LR channels" at 0db? I like earth shacking bass...

In Azid 1, if I don't select any Output mode (Dual mono, stereo...) what will happen? Do I need the option "Decode (Front/rear)"? If it is non selected, is it the same as selected at 2/0?

Should I use the Gain - Auto gain options in Azid 1 or in Besweet? What's the difference? What's better?

Should I use "Dialog normalization reduction" in Azid 2?

What's the difference for e.g. "Alt preset CBR 128" in Lame 1 from "Use CBR Bitrare 128" in Lame 2?

In Lame 3 the option "Scale input (Multiply PCM data) by 1" is on by default... What does is do? Is it needed?


Well thanks again... I have more questions but these are more than enought for now...!

Oh... one last thing. If there's not another program out there that makes AC3, can u tell me how to make a multichannel 5.1 WAV that besweet will accept?

[tiki4]

Hell,

it's just 9 o'clock here. I think it's O.K to select -L 0db in the downmix options, but -3db is the default used by all profiles and also by GordianKnot. I think you should try for yourself and always watch your logfiles for downmix overflows or things like that. Also I'd listen to the stuff before burning to CD.

Don't bother with the output mode in Azid. BeSweet does that for you. If you need to know more about the stuff, download azid.exe and read the included documents. Don't get this mixed up with 'dynamic range compression' (light, normal, inverse), that's something you should always apply.

For gain I recommend BeSweet's hybridgain (1.5 betas only) (look at above command line). It does the following: For 5.1 input files (AC3) a gain of 10db is applied in azid.dll, after applying 'dialog normalization reduction' (-n1). Then the signal is scaled to 97 % of non-clipping maximum. The postgain is directly written to the MP3 frames. If you go for Ogg Vorbis output, the postgain is written as 'Ogg comment' into the Ogg Vorbis file. For all other output modes you need to do two-pass normalize. Normalization can be done in BeSweet with three different switches:

-azid( -g max ) -> two-pass normalize by azid.dll
-ota( -g max ) -> two-pass normalize by BeSweet
-ota( -G max ) -> postgain to 97 % of max
-ota( -hybridgain ) = -azid( -n1 -g 10db) -ota( -G max )

Pick what you need, but there are reasons why to use hybridgain if possible. That means also you don't need to bother for 'dialog normalization reduction' by that way (-n on for 5.1 input and -n off for 2.0 source). Don't care for the 'scale input feature' of lame.dll , switch it off if you use BeSweet. This is done by BeSweet for you.

Last thing: LAME switches. For best possible quality use the --alt-presets. There are three different forms of them:

true VBR:

--alt-preset standard (very high quality)
--alt-preset extreme (even possibly higher quality)
--alt-preset insane (= --alt-preset cbr 320) (guess what)

ABR switches:

--alt-preset 128, 160... try's to reach the given bitrate

CBR switches:

--alt-preset cbr 128, 160... uses constant bitrate mode.

For highest quality use VBR and mux with NanDub or better VirtualDubMod. Don't even think about using other switches. The --alt-presets where tuned at hydrogenaudio.org and the VBR ones cannot reproduced by other command line switches as they are internal code tweakds. The ABR switches may get reproduced by other command line switches, but why bother, when life can be as simple as that.

Take care and look for more information about BeSweet in this very forum.

tiki4
--

[Ody1]

So that's why I couldn't find hybridgain... I still use the last stable version... Sould I d/l the 1.5 beta? It isn't unstable or something...?

I only use PowerDVD to play DVDs and I always have dynamic range compression off. I'll try first with none 'dynamic range compression' and go to light or normal if the sound is too low.

For hybridgain you said that it first aplies 10db...? Always? What if this cause clips? From what I understand, hybrid gain does a little bit of everything... (except dynamic range compression), right? Why all this trouble when you can do a simpe two pass normalization?

From these switces
-azid( -g max ), -ota( -g max ), -ota( -G max )
which is the best? What'll happen if two of them, or all three of them, are used?
Until I d/l the 1.5 version and use hybridgain, is the -n1 needed? I can't understant what it does.....

For vbr sound and avi I have done a lot of reading. I know why it's a bad idea to put vbr mp3 in avi and how NanDub's hack works. Anyway I'll stay with cbr for now.
So you say it's better to use "Alt preset CBR 128" than "CBR Bitrare 128" in Lame 2...

I'd like to create an AC3 file that lasts for a few seconds, e.g. 6sec, and every sec there's a sound coming from each speaker. First the FL, then the C, ..... and last from the .1.... How to do that with AC3machine?


I feel Besweet less and less strange and complicated... Thanks for the help.

[tiki4]

Hi Ody1,

yes, if you just do AC3 -> MP3 or AC3 -> Vorbis then it's definitely O.K. to download 1.5b16. It has some issues but you won't miss any problem if you always watch your logfiles. Even latest GordianKnot uses 1.5beta. You need not bother with any of the three gain switches if you use -hybridgain. It applies 10db gain after 'Dialog normalization reduction' but just for 5.1 channel input files. If the input is 2.0 then no normalization is done and no 10db gain is applied (just -ota( -G max )). Quality wise hybridgain should offer you best quality of all normalization modes (while also being fast because of postgain). If you are looking for an in depth discussion just search this forum. I don't know all the details. In LAME you always should use --alt-preset. Have a look for yourself at HA (http://www.hydrogenaudio.org)in the MP3 forums at the recommended settings sticky.

There's been a surround test AC3 file you may find by searching the forums again. I think it does what you are looking for. Search for 'surround test' or similar.

Regards,

tiki4