[linux-dvb] Re: Interesting DVB-T issues

[Nico <nsabbi@libero.it>]

Edward Wildgoose wrote:

> (well, I thought so...)
>
> First, I haven't run scan for perhaps 6-12 months, but for various reasons
> ran it again recently, and just happened to notice that quite a few of the
> channel tuning details have changed..!  This is for the crystal palace
> transmitter in the UK and quite a few of the channels now have different fec
> values for both hp and lp params.
>  
> Now is this because a previous version of scan was not returning proper
> values or is this something which I should have been looking out for?
>
> The practical upshot is that after correcting these I now get instant,
> perfect audio-video sync on BBC ONE and BBC TWO using "mplayer dvb://" and
> all the other channels exhibit symptoms of just starting out of sync and
> gradually audio/video synchronise over a few seconds.  Previously I had a
> lot of issues with ITV 2 and having the updated tuning params has helped
> enormously.
>  
do you mean that the newer values are correct and the older ones aren't 
or the opposite?
In its current  shape mplayer hardcodes LP, hierarchy and inversion to 
XX_DEFAULT;
in the future I will remove this mess, and respect the content of 
channels.conf

> However, that audio-video sync brings me to my next observed peculiarity.
> Basically, BBC ONE and TWO have instant and perfect audio/video sync in
> mplayer, but the other channels appear not to have video pts and so players
eh? no video pts? are you sure? run mplayer with -v and grep ADDED, it 
will tell you the pts

> like mplayer seem to gradually bring everything back into sync, presumably
> based on the audio timestamps.  This corresponds with BBC ONE and TWO being
> the only two with the PCR pid being the same as the video pid, and the other
> channels have an independent PCR pid.
>  
there are 2 problems currently:
1) in DVB transmissions often there are huge preload times in the audio 
stream
(such as audio_pts ~= video_pts - 5);  it's an old story, and mplayer 
resyncs after few seconds;
2) the pcr pid is totally ignored, so sync is made respect to audio (if 
I remember correctly);
in mplayer-dev-eng of september there's a message by Arpi explaining 
this behaviour

I may fetch the pcr pid, and pass the clock to mplayer, but I don't know 
if it's worth doing it in g1
and what/how much implications this change would have

Nico





--
Info:
To unsubscribe send a mail to ecartis@linuxtv.org with "unsubscribe linux-dvb" as subject.