[MEncoder-users] Dolby downmix

[Rolf Ernst]

Rolf Ernst <rolf.ernst <at> silverlightning.org> writes:
>> The reason that I like to downmix is that I generally keep TV episodes
>> in Dolby surround sound format (The balance of space/quality/experience
>> tilts in favor of space) whereas I like to keep movies at high audio
>> quality.
> If I could just understand what you write here...
> (Do you suggest that there may be a situation where you can get "high(er) audio
> quality" by dropping channels? Sorry, I am not a native speaker...)
>
> If ffmpeg -i yourrecording -qscale 5 -ab 256k -ac 2 out.avi doesn't work
> (*except* for the not-disappearing telecine), I believe we are interested in the
> sample. (ftp server in my previous mail and in your answer.)
>
> Carl Eugen
>
> PS: Note that as far as I follow the avfilter discussions on ffmpeg-devel,
> nobody has yet lamented over the missing filmdint filter (and neither on
> mencoder-users when I suggested using FFmpeg). So it is useful to mention the
> need for this filter (and others) clearly!
> (Don't misunderstand me: I know that inverse telecine is important, but I
> believe many of the active developers live in PAL-country where it has nearly no
> relevance.)
Carl,

sorry for being so diffuse in my answers (btw, I am native German and 
English, half my life here and half there) but of course I do not 
believe that Dolby downmix will improve quality. Maybe let me explain 
the DVR best: It records in either AC 5.1 or 2 channel Stereo (which is 
plain dropping ;fe, center and back which is not very good). So I let 
the device record in 5.1. Now I take the 5.1 6 channel signal and want 
to downmix to two channel surround which preserves at least some of the 
surround information but uses half the space or so. So there is a 
degradation from 5.1 to 2 channel Surround but still better than 
discarding 4 channels from 5.1.

As for the interlacing - It has nothing to do with interlaced video. The 
device records 1280x720 progressive at 59.94 frames per second. Since it 
comes from an analog input there are no rff flags or what not in the 
stream. It essentially grabs 60 frames/sec from this composite signal 
going into my flatscreen. The 60 fps is sufficient of a multiplier to 
make film look relatively smooth. So how to turn such a signal back into 
24 frames of film which is really what we want to compress? This has 
been a challenge for me for a long time.

The trick here is to artificially interlace the progressive signal and 
then have the filmdint do the actual pulldown. I can't really take 
credit for this approach but found it trolling the net. PAL users have 
an easier time as they will probably just half the frame rate (50 fps) 
and wind up with a shorter movie and be ok with it. Unfortunately my 
signal is 60fps. (59.94).

I'll find a clip and let you tinker with it and you will see. It is best 
understood by actually experimenting with the material.

-- 
/re

"My friends, watch out for the little fellow with an idea"

Tommy Douglass