SamyGO

smartsmurf wrote:

baco81 wrote: There is a question that is killing me, though...

If FFMPEG is more performant, why did you choose LIBDCA codec instead as the preferred one? I mean... what are the drawbacks of using FFMPEG codec?

OK, answer is simple:
The libdca has got more advanced DTS decoding features (like downmixing options, dynamic range compression, etc.). Despite that it completely lacks (hardware-)acceleration. It's a pure C implementation. So generally speaking, I would expect best audio results by using it.
In contrast the ffmpeg implementation got less features, but more low-level optimizations for ARMV6 and VFP. So it is more performant resulting in less CPU load.
The differences convinced me to put both libraries into the package. The precedence for libdca is also explainable: ffmpeg is used for media parsing, so it cannot be removed and must be there. The only thing which can be removed safely is libdca, since it is only used for decoding, but not for parsing.

nostromo1983 wrote:Well, it's not a 3d movie, it's a standard movie, nothing special. I can upload you a chunk for analysis, how long do you need? 1 minute is ok?

einaris1 wrote: Hi,
all *.mkv files by default created with MKVMERGE > v4.1.0 uses header removal compression. So not only DTS frames are striped, but video frames (h264, VC-1) and audio (AC3, MP3). This MKV file is not playable on samsung tv media player .
Solution is to remux this file without "header striping". On mkvmergeGUI this is done -> "Compression: none" on each stream.

smartsmurf wrote:

nostromo1983 wrote:Well, it's not a 3d movie, it's a standard movie, nothing special. I can upload you a chunk for analysis, how long do you need? 1 minute is ok?

In the meantime I got the file. from analyzing the logfile I could see that the DTS frames are not sent with the header sync "7FFE8001", which indicates a starting DTS frame (see also dca.h in ffmpeg). Instead the frames start with the payload immediately.

What happened? The creator of the MKV uses a so-called "header removal compression", which is descibed in the Matroska specs. Se also here for some details:
http://www.bunkus.org/videotools/mkvtoo ... ompression

Unfortunately the MKV demux routines made by Samsung can not cope with that mode. So they generate wrong frames.

It can be fixed in software. I am working on that...