[mpeg2] interrupted streams

[Christian Wolff <scarabaeus@convergence.de>]

Hi everyone,

I ran into a strange problem. Let me explain my setup, first.

Last thursday i recorded the show of No Forcefield at the
Beta Lounge (www.betalounge.com), with the KFIR card hooked
directly into the video live mix and the main audio. Very
good quality, so far.

To monitor what i'm recording, i put a Netstream2000 in the
same box (150MHz pentium, 64MB, 10gig ide) and started the
recording somehow like this:
  cat /dev/video > betalounge_xxx.mpg &
  playfile s betalounge_xxx.mpg

That worked well for the first 45 minutes or so, and then
the picture froze. I restarted the recording into another
file, and the same thing happened again after a while. I
did this over an over. I also tried a variation, where
i started the recording somwhat like this:
  cat /dev/video > betalounge_xxx.mpg &
  tail -f betalounge_xxx.mpg > video.fifo &
  playfile s video.fifo

This way i could restart the netstream app without stopping
the recording. And i could confirm that the streams, once they
broke, didn't start playing again.

Since the first file is 850 MB, the next one recorded for 800 MB
without problems (i stopped the _03_ file manually) and after that
it went down to less and less time before it stopped, ending at
50 to 100 MB per file, regardless of restarting the computer,
is uspect a thermic problem of the card.

Did anyone else have problems like that after recording an hour
or more?

Here are my file sizes in kilobytes, to give you an idea:
884529 betalounge_2001_02_22_01_352x480_2_75MBit.mpg
450096 betalounge_2001_02_22_03_352x480_2_75MBit.mpg
366473 betalounge_2001_02_22_04_352x480_2_75MBit.mpg
 53088 betalounge_2001_02_22_05_352x480_2_75MBit.mpg
419609 betalounge_2001_02_22_06_352x480_2_75MBit.mpg
 87094 betalounge_2001_02_22_08_352x480_2_75MBit.mpg
 38134 betalounge_2001_02_22_09_352x480_2_75MBit.mpg
 38841 betalounge_2001_02_22_10_352x480_2_75MBit.mpg
131411 betalounge_2001_02_22_11_352x480_2_75MBit.mpg
 53538 betalounge_2001_02_22_12_352x480_2_75MBit.mpg
 35660 betalounge_2001_02_22_13_352x480_2_75MBit.mpg
 25814 betalounge_2001_02_22_14_352x480_2_75MBit.mpg
 55963 betalounge_2001_02_22_15_352x480_2_75MBit.mpg
 58003 betalounge_2001_02_22_16_352x480_2_75MBit.mpg
 60348 betalounge_2001_02_22_17_352x480_2_75MBit.mpg
 26151 betalounge_2001_02_22_18_352x480_2_75MBit.mpg
 25878 betalounge_2001_02_22_19_352x480_2_75MBit.mpg
 72156 betalounge_2001_02_22_20_352x480_2_75MBit.mpg
 20931 betalounge_2001_02_22_21_352x480_2_75MBit.mpg
120006 betalounge_2001_02_22_22_352x480_2_75MBit.mpg
135700 betalounge_2001_02_22_23_352x480_2_75MBit.mpg
146751 betalounge_2001_02_22_24_352x480_2_75MBit.mpg
 18795 betalounge_2001_02_22_25_352x480_2_75MBit.mpg
 11390 betalounge_2001_02_22_26_352x480_2_75MBit.mpg
 17767 betalounge_2001_02_22_27_352x480_2_75MBit.mpg
 13462 betalounge_2001_02_22_28_352x480_2_75MBit.mpg
 18072 betalounge_2001_02_22_29_352x480_2_75MBit.mpg
117115 betalounge_2001_02_22_30_352x480_2_75MBit.mpg
 31934 betalounge_2001_02_22_31_352x480_2_75MBit.mpg
 62099 betalounge_2001_02_22_32_352x480_2_75MBit.mpg
117147 betalounge_2001_02_22_33_352x480_2_75MBit.mpg
 41684 betalounge_2001_02_22_34_352x480_2_75MBit.mpg
   371 betalounge_2001_02_22_35_352x480_2_75MBit.mpg
  7197 betalounge_2001_02_22_36_352x480_2_75MBit.mpg
 38792 betalounge_2001_02_22_37_352x480_2_75MBit.mpg
142929 betalounge_2001_02_22_40_352x480_2_75MBit.mpg
153433 betalounge_2001_02_22_41_352x480_2_75MBit.mpg
265054 betalounge_2001_02_22_42_352x480_2_75MBit.mpg

I cut out one section where it went wrong, if you like to 
look at it (128k):
http://generator.scara.com:/betalounge_2001_02_22_04_352x480_2_75MBit.broken.mpg

The interesting locations are these:
0DCA0 the last un-broken pack header (SCR 3C95E11, Mux rate 7435) 
0DCAE the last un-broken system header (rate bound 7435, P-Std video 80k, P-Std audio 10k)
0DCC0 first PES packet (0808 bytes)
10D10 last un-broke PES packet (06DE bytes)
113D4 here the next start code is expected, but instead it's 35 C2 80 E4
113D8 pack start coded, followed by a header with only 2 of 5 marker bits set and 4 bytes short
113E2 system header start code, with header data that makes no sense. Fragments of the right P-Std
      fields, but even that in the wrong location, header size is 0000 etc.

After that, mostly single byte values in 32 bit words appear (00 00 00 xx)
or a "header" of 00000000 00000000 00000000 000000FF and then some garbled 
data.

-- 
Christian Wolff, CTO, convergence integrated media, Inc.
2565 Third St. #338,  San Francisco, CA 94107-3158, USA 
phone:415-920 9237  fax:920 9239   wolff@convergence.de
http://www.scarabaeus.org/    scarabaeus@scarabaeus.org 
check out: http://www.betalounge.com/  http://smash.tv/