[vlc-devel] TS files that has PCR hickups

[Morgan Tørvolt]

> 1. Do you see anything suspicious in the vlc messages window which might
> point to something?

Good question! Yes, there pops up these messages:
mux_ts debug: adjusting rate at -5400000/200000 (19/1072)
mux_ts debug: adjusting rate at -5600000/200000 (11/1105)
mux_ts debug: adjusting rate at -4600000/200000 (123/1152)
mux_ts debug: adjusting rate at -3600000/200000 (63/870)
mux_ts debug: adjusting rate at -3800000/200000 (17/996)

They pop up in the same way as there appears PCR problems in the
stream, and the exact same number of times as well. This seems to be
related to my problems.
All further tests confirms that this message occurs whenever I get a
wrong PCR timing in the TS stream.

> 2. Do you also see a similar thing happen if you read a regular normal (ts?)
> file and rewrite it to a ts-file?

I took the same TS file and used it as input and that seems to have
removed one occurrence of the issue, the resulting file has four
problems:

mux_ts debug: adjusting rate at -5200000/200000 (19/862)
mux_ts debug: adjusting rate at -5600000/200000 (11/1105)
mux_ts debug: adjusting rate at -4600000/200000 (123/1152)
mux_ts debug: adjusting rate at -3600000/200000 (78/871)

This got me thinking, so I ran this command:
vlc dvd:///dev/scd0@1:2
"--sout=#duplicate{dst=std{access=file,mux=ps,dst="test.ps"}}"
--sout-all
Followed by this command:
vlc test.ps "--sout=#duplicate{dst=std{access=file,mux=ts,dst="test3.ts"}}"
--sout-all

This resulted in a perfect ts file when I look at the PCR. The
language information for the audio tracks is gone though, but that is
not a huge issue. Why reading the vob from the DVD is different from
reading the same file from the disk I cannot really say I understand.

> I would recommend using broadcast-level shaping and DTS delay, such as
>  --sout-ts-shaping=1000 --sout-ts-use-key-frames --sout-ts-dts-delay=500

Adding those parameters really helped alot. The PCR now seems to be
constand between I frames. We did get one occurence of the rate
adjustment like earlier, and it seems to be more severe than before.
As a result, the entire following GOP has very long duration per
packet, since it now uses the rest of the time until the following I
frame to compensate.

Output from my test program on this new TS file:
http://paste.videolan.org/3447


Thanks a lot for your help guys. I am mighty impressed that the two
people who answers actually both comes up with good replies.

-Morgan-