Hello Carl,

yes I'm working in a commercial theatre which has 6 screens, so my tests are in « real » situation ;-)

I have the feeling that upmixing A lowers too much the LFE channel and gives too much room to highest frequencies.

I don't remember using -3db step for C and LFE in my experimentations ;-)




Using sox for distributing ranges of freq. per channel, it looks like :

# gauche - L
sox $dft $ddt/00.wav $ddt/01.wav remix -l sinc 1900-4800
# droite - R
sox $dft $ddt/00.wav $ddt/02.wav remix -r sinc 1900-4800
# centre - C
sox $dft $ddt/00.wav $ddt/03.wav remix 1v0.5,2v0.5 sinc 150-1900
# basses - LFE
sox $dft $ddt/00.wav $ddt/04.wav remix 1v0.5,2v0.5 sinc 20-150
# ambiance gauche - Lsr
sox $dft $ddt/00.wav $ddt/05.wav remix -l sinc 4800-20000
# ambiance droite - Rsr
sox $dft $ddt/00.wav $ddt/06.wav remix -r sinc 4800-20000

where 00.wav is source stereo material, dft="-S -V" and ddt=path to temporary folder

For C and LFE v0.5 is to get L and R at the same level, if v1 then both are added ending in much too loud mix.





Hereafter another way, sounding more like a dolby prologic mix :

# first step, get a mono mix of the stereo source for C
sox $dft -c 2 $ddt/00.wav -c 1 $ddt/mono_mix.wav mixer 0.5,0.5
# get one file for L, one for R
sox $dft -c 2 $ddt/00.wav -c 1 $ddt/00gauche.wav mixer -l
sox $dft -c 2 $ddt/00.wav -c 1 $ddt/00droite.wav mixer -r
# phase inversion on L and R
# substract R from mono mix to get L / substract L from mono mix to get R
# → this is where "human voices" almost disappear from L and R channel unless they were mixed 100% on left or right !
sox $dft -M -c 1 -v -1 $ddt/00droite.wav -c 1 $ddt/mono_mix.wav -c 1 $ddt/gauche.wav
sox $dft -M -c 1 -v -1 $ddt/00gauche.wav -c 1 $ddt/mono_mix.wav -c 1 $ddt/droite.wav
# gauche - L
sox $dft -c 1 $ddt/gauche.wav -c 1 $ddt/01.wav sinc 120-20000
# droite - R
sox $dft -c 1 $ddt/droite.wav -c 1 $ddt/02.wav sinc 120-20000
# centre - C
sox $dft -c 1 $ddt/mono_mix.wav -c 1 $ddt/03.wav sinc 120-12000
# basses - LFE
sox $dft -c 1 $ddt/mono_mix.wav -c 1 $ddt/04.wav sinc 20-120
# ambiance gauche - Lsr
sox $dft -c 1 $ddt/gauche.wav -c 1 $ddt/05.wav delay 0.02 sinc 120-6000 reverb
# ambiance droite - Rsr
sox $dft -c 1 $ddt/droite.wav -c 1 $ddt/06.wav delay 0.02 sinc 120-6000 reverb




I don't pretend these are perfect but quite smooth. I had another "mix" in the upmix script but way too loud for movies, only useful for concert or musical video clips.


Best regards,
G.





Le 07/10/2015 10:31, Carl Hetherington via DCPomatic a écrit :
On Tue, 6 Oct 2015 13:03:03 +0200
Gérald Maruccia via DCPomatic <dcpomatic@carlh.net> wrote:


Hello !

Wow, impressive list of fixes !

* Add a new stereo to 5.1 upmixing algorithm (Upmixer B).

→ the one already in is kind of "metallic" - which is unexpected if it's 
just about distributing frequency ranges into channels ?

Yes, that is a bit strange.  Are you listening in an auditorium?


→ does the new one make use of phase-inversion ?

Yes, I've just added a description of what it does here:
http://dcpomatic.com/manual/html/ch06.html

These upmixers are all somewhat experimental... it's interesting to hear how useful any of them are.

Regards,
Carl




Best regards,

Gérald




Le 06/10/2015 01:10, Carl Hetherington via DCPomatic a écrit :

Hi all,

DCP-o-matic 2.4.0 is now available for download from:

http://dcpomatic.com/download

It has the following changes:

	* Basic support for creation of VF DCPs from imported OV DCPs.
	* Handle non-UTF-8 .srt files better.
	* Add experimental packages for Fedora 22.
	* Fix various problems with subtitle timing when video frame rates change.
	* Fix various preview problems when changing DCP frame rate (#695 and #699).
	* Fix incorrect recognition of a folder of images due to filenames beginning with ._
	* Fix failure to analyse audio of some combined video/audio files.
	* Fix subtitle errors in some cases.
	* Fix errors with .MTS files in some cases (#702).
	* Fix incorrect hint when making a 3D DCP.
	* Fix error on adjusting the audio map in some cases.
	* Fix for analysis of audio when the subject is later in the playlist than some other long piece of content.
	* Fix encoding server shortcut on Windows.
	* Fix failure to display audio analyses in some cases.
	* Fix crash when re-making certificates after deleting existing ones (#690).
	* Fix crash on adding an empty folder as content (#691).
	* Fix non-operational image sequence frame rate dialogue.
	* Speculative fix for hangs at the end of encodes in some cases.
	* Fix subtitle timing when trimmed.
	* Add a new stereo to 5.1 upmixing algorithm (Upmixer B).
	* Allow configuration of default standard (Interop or SMPTE).
	* Speed up checking of existing image data when resuming transcodes.
	* Add a frame counter to transcode progress (#111).
	* Add export button for certificates.
	* Display peak audio value in the audio tab if an analysis has been done.
	* Add option to analyse audio automatically when it is added (#673).
	* Mark SCP as being the protocol for Doremi as well as AAM (#678).

Thanks to Mirjam Bromundt and Thierry Journet.

Best regards,
Carl


_______________________________________________
DCPomatic mailing list
DCPomatic@carlh.net
http://main.carlh.net/cgi-bin/mailman/listinfo/dcpomatic


      

      
_______________________________________________
DCPomatic mailing list
DCPomatic@carlh.net
http://main.carlh.net/cgi-bin/mailman/listinfo/dcpomatic