Josef Wolf wrote: >On Sat, Oct 29, 2005 at 10:37:12AM +0400, Manu Abraham wrote: > >Thanks for your reply! > > > >>>- Most of the time the bitrate on the video-pid is quite low (about >>>2000..3500 TS packets per second, which gives something below 7MBps). >>>But sometimes (every 10..20 minutes) the bitrate raises dramatically[2] >>>for a couple of seconds (sometimes just one second, sometimes up to 20 >>>seconds) and comes down again to the 2000..3500 range. I really don't >>>understand what is going on here. How come this dramatic raise without >>>increase in the motion or in picture-details? Is this raise really >>>caused by the video stream? >>> >>> >>Since the TS is constituted of more than the Video ES, the bandwidth >>occupied might very well depend on other factors. >> >> > >Sorry, I should have stated more clearly that this bitrate-raise is only >on the video-pid (pid 101). > > > >>I remember Johannes once saying that the EIT rate on the German Das >>Erste is quite high. >> >> > >But isn't the EIT on a completely different PID? Thus the EIT should >have no impact on this raise. > > > Sorry, I thought you were looking at the transport stream. >>>I would like to try this recoding (TS->PS). Extracting PES from the TS >>>is not really a probem. But unfortunately, I could not find any >>>description how PS streams are to be constructed. Any pointers? >>> >>> >>You can probably, check out replex from the Metzler bros, >>http://www.metzlerbros.org/dvb >> >> > >I'll take a look at it. But since I need to do it on the fly (since I want >live-streams) I'll probably need to roll my own. > > > A TS2PS can take a little bit of CPU power, another option was dvbstream, it had a ts2ps conversion, i don't know whether it is still there. Once Nico had plans to remove it. If it is removed, you can check for some older versions. >I found a description in iso13818-1 section 2.5, but this is somewhat hard >to read =:) Are any more comprehensive descriptions avalable? > > > I think compared to other documents (eventhough not directly related, EN50221 and friends), 13818-1 is far better. Manu
