eac3to - audio conversion tool

[madshi]

Originally Posted by b66pak

avisynth script for 5.1 to Stereo NO LFE:
sox remix line for 5.1 to Stereo NO LFE:
avisynth script for 5.1 to Stereo with LFE:
sox remix line for 5.1 to Stereo with LFE:

avisynth and sox seem to use the same matrixes? But is this a standard matrix? Or did just somebody say "this sounds ok" and everyone else is copying the parameters?

Originally Posted by spork985

I redownloaded and tried it and it does the same thing. I hashed all the files with md5 and they match the hashes of the fiels on the computer that does work.

Hmmmm... Maybe your PC doesn't like the libFlac.dll version for whatever reason. You could try replacing the libFlac.dll with an older version. eac3to ships with libFlac 1.2.1. You can get older versions here in the "flac-win" section:

http://sourceforge.net/projects/flac/files/

You need to download the "flac-x.x.x-devel-win.zip" file. That's the only one where the libFlac.dll file is in.

[tebasuna51]

Originally Posted by madshi

avisynth and sox seem to use the same matrixes? But is this a standard matrix? Or did just somebody say "this sounds ok" and everyone else is copying the parameters?

Is the standard matrix:

fl' = FL + sqr(2)/2 x FC + BL
fr' = FR + 0.7071 x FC + BR

now we have the same acustic power for FC when add the part in FL and FR.

If all channels have a peak at same time we can have overflow reaching values until 2.7071, then we need divide the coefficients by 2.7071:
1/2.7071 = 0.3694
0.7071/2.7071 = 0.2612

fl' = 0.3694 x FL + 0.2612 x FC + 0.3694 x BL
fr' = 0.3694 x FR + 0.2612 x FC + 0.3694 x BR

Now the max value is:
0.3694 + 0.2612 + 0.3694 = 1
After the downmix is recommended a Normalize to avoid low volume.

[madshi]

Originally Posted by tebasuna51

Is the standard matrix:

fl' = FL + sqr(2)/2 x FC + BL
fr' = FR + 0.7071 x FC + BR

now we have the same acustic power for FC when add the part in FL and FR.

Ok, thanks, that makes a lot of sense.

Now the big question is: How should I make this available in eac3to? Right now "-down2" invokes Dolby ProLogic II. How should I name the options for Dolby ProLogic II and for this other downmixing matrix? Maybe I should use "-down2" for the normal matrix and "-downDpl2" for Dolby ProLogic II? Any other suggestions?

Originally Posted by tebasuna51

If all channels have a peak at same time we can have overflow reaching values until 2.7071, then we need divide the coefficients by 2.7071:
1/2.7071 = 0.3694
0.7071/2.7071 = 0.2612

fl' = 0.3694 x FL + 0.2612 x FC + 0.3694 x BL
fr' = 0.3694 x FR + 0.2612 x FC + 0.3694 x BR

Now the max value is:
0.3694 + 0.2612 + 0.3694 = 1
After the downmix is recommended a Normalize to avoid low volume.

I'll simply skip the 2.7071 division. If there's overflow/clipping, eac3to will detect that automatically and fix it in a 2nd run, resulting in normalized audio.

[jpsdr]

Hello.
Question about use. If i want to extract the raw video track from an .mts/.m2ts file, to get an untouched raw video data stream, to be able to use it later in authoring SW like Scenarist.
Is the following enough : eac3to file.mts -demux
or, is there others options to add ? (Maybe to extract only video if possible).
If i want the results file(s) to be put on another place than the source file, what the command line should be ?
Thanks.

[tebasuna51]

Originally Posted by madshi

Now the big question is: How should I make this available in eac3to? Right now "-down2" invokes Dolby ProLogic II. How should I name the options for Dolby ProLogic II and for this other downmixing matrix? Maybe I should use "-down2" for the normal matrix and "-downDpl2" for Dolby ProLogic II? Any other suggestions?

To be backward compatible I suggest use a new parameter "-downstereo" for the new mix and let "-down2" with the Dolby ProLogic II behaviour.

I'll simply skip the 2.7071 division. If there's overflow/clipping, eac3to will detect that automatically and fix it in a 2nd run, resulting in normalized audio.

You are right, maybe you can use the same procedure for dpl2 with:

fl' = FL + 0.7071 x FC + 0.866 x BL + 0.5 x BR
fr' = FR + 0.7071 x FC - 0.5 x BL - 0.866 x BR

[Midzuki]

Let's say, I create an MKV which contains DTS audio, encoded from 16-bit .WAVs. When I demux the MKV with eac3to, this always "patches" the demuxed DTS to "24-bits". What is the point of this behavior?

[SomeJoe]

Originally Posted by Midzuki

Let's say, I create an MKV which contains DTS audio, encoded from 16-bit .WAVs. When I demux the MKV with eac3to, this always "patches" the demuxed DTS to "24-bits". What is the point of this behavior?

DTS and Dolby Digital store audio information in the frequency domain. As such, they do not have a "bit depth". When PCM audio (time-domain) is recreated from them during decoding, the audio may in fact decode to values that have data into the 8 additional bits. Those bits may not be accurate with respect to the source, but are accurate with respect to the frequency-domain representation in the compressed audio.

Remember that a 16-bit PCM source is itself an approximation of an analog waveform, and as such is inaccurate because of the 16-bit quantization.

[tebasuna51]

Originally Posted by Midzuki

Let's say, I create an MKV which contains DTS audio, encoded from 16-bit .WAVs. When I demux the MKV with eac3to, this always "patches" the demuxed DTS to "24-bits". What is the point of this behavior?

This is because standard DTS isn't lossless. The internal precission of the samples, in frequency domain, is equivalent (not equal) to 24 bits in time domain (no matter what is the source bitdepth).

Then decode to 24 bits is always better than decode to 16 bits, BTW if you want only 16 bits you can add -down16.

edit: I don't see the SomeJoe post with, more or less, the same thing.

[b66pak]

@madshi in my opinion we should have:

-down2stereo
-down2dpl
-down2dpl2
-lfe

the present "-down2" will be equal to "-down2dpl2" for backward compatibility...

also for everybody to be happy i think you should also give full access to the mixer with something like:

-down2custom 1v0.3694,3v0.2612,5v0.3694 2v0.3694,3v0.2612,6v0.3694

this will also solve the present eac3to inability to downmix anything but 5.1...

for example a 3ch (FL,FR,FC) to stereo will luke like:

-down2custom 1v0.5858,3v0.4142 2v0.5858,3v0.4142

_

[tebasuna51]

Originally Posted by b66pak

...
also for everybody to be happy i think you should also give full access to the mixer with something like:

-down2custom 1v0.3694,3v0.2612,5v0.3694 2v0.3694,3v0.2612,6v0.3694

...

That's could be interesting.

A 'Quadro' audio (2/2.0 : FL,FR,BL,BR) can be downmixed to Dpl2 with:

-down2custom 1v1.0000,3v0.8660,4v0.5000 2v1.000,3v-0.5000,4v-0.8660

(the sample is for remember accept the '-' sign)

[spork985]

Originally Posted by madshi

Hmmmm... Maybe your PC doesn't like the libFlac.dll version for whatever reason. You could try replacing the libFlac.dll with an older version. eac3to ships with libFlac 1.2.1. You can get older versions here in the "flac-win" section:

http://sourceforge.net/projects/flac/files/

You need to download the "flac-x.x.x-devel-win.zip" file. That's the only one where the libFlac.dll file is in.

FLAC 1.1.4 worked, flac 1.2.0 would not. Thank you very much, I have been going crazy about this for a few days now!

[Midzuki]

tebasuna51 wrote:

BTW if you want only 16 bits you can add -down16.

Doesn't work. MediaInfo says the demuxed DTS still "is" 24-bits. Fortunately, there are workarounds — namely, TSMuxer and AVI-Mux GUI.

SomeJoe wrote:

Remember that a 16-bit PCM source is itself an approximation of an analog waveform, and as such is inaccurate because of the 16-bit quantization.

Even a 64-bit @ 768kHz PCM file will be "just an approximation". Remember, there is no such thing as a "perfect" Analog-To-Digital conversion.

Anyway, thanks for the accurate answers.
Actually, I was expecting to receive an answer from madshi himself, but it seems he really is "much busier" than I thought.

[Snowknight26]

It doesn't matter what MediaInfo says about lossy tracks regarding bitness. It's not 16-bit and it's not 24-bit.

[raquete]

Originally Posted by Snowknight26

It doesn't matter what MediaInfo says about lossy tracks regarding bitness. It's not 16-bit and it's not 24-bit.

all right, can be but...
i was sending pm to Midzuki telling that i have few dvds where olaying DTS from video_ts folder my receiver/decoder inform in the display: 24b..this info stay there until the dvd end!
others dvds don't inform and i don't know how to encode in 24b.
i need lessons and explanations about this informations. how and what program can be used to do that?!?
Porcupine Tree is one of my DVDs that give 24b info, i have others!

@ Midzuki
i could not resisit to post here too but i'm waiting yours infos too.

[Midzuki]

Originally Posted by Snowknight26

It doesn't matter what MediaInfo says about lossy tracks regarding bitness. It's not 16-bit and it's not 24-bit.

I already knew that, since DTS "Digital Surround" is a lossy format. BUT, a binary file-comparison between a 16-bit DTS and a "patched-to-24-bits" DTS shows they are different beasts (on a "frame-by-frame" level, I think).

[Midzuki]

4 raquete: lossy "24-bit DTS" requires 24-bit sources
@ 96kHz or higher.

Also, take a look at your PM inbox.

[raquete]

Originally Posted by Midzuki

4 raquete: lossy "24-bit DTS" requires 24-bit sources
@ 96kHz or higher.

Also, take a look at your PM inbox.

great, i got the lesson Mid.

thanks so much!

[tebasuna51]

One more time, AC3/DTS standard (lossy) don't have bitdepth. The parameter quality for this files is the bitrate, forget the bitdepth.

Only DTS have a value in the header than say the bitdepth of the source wav files used to create the dts. And Surcode put always 24 bits also when source wavs is 16 bits, then is a useless info. I say to madshi don't put this info (sometimes wrong) with dts files because confuse the users, like you can see.

AC3/DTS files must be decoded to, at least, 24 bit to obtain the best aproach to the source, no matter the original bitdepth.

[raquete]

ok tebasuna,
i was curious based on some informations from my decoder like i posted here:
http://forum.doom9.org/showthread.ph...84#post1386384

to tell the true i was not only confused, i was really lost with that infos.

thanks so much!

[Midzuki]

Originally Posted by tebasuna51

Only DTS have a value in the header than say the bitdepth of the source wav files used to create the dts. And Surcode put always 24 bits also when source wavs is 16 bits, then is a useless info.

AC3/DTS files must be decoded to, at least, 24 bit to obtain the best aproach to the source, no matter the original bitdepth.

So, let's just hope there don't exist too many
" 'dumb' DTS decoders ".