alsa@xxxxxxxxxxxx wrote:
>
> On Mon, 15 Sep 2003, Michel Bardiaux wrote:
>
>
>> In a long run of my capture application, I see, at random moments
>> but on average roughly once per houre, either 2 or zero frames in
>> one 40msec interval (PAL, 25fps). This could be due to fluctuations
>> in my system clock (it is NTP'ed but nothing is perfect), or in the
>> TV board, or just bad signal from the cable company, but I wonder:
>> what is the stability and long-term accuracy of the 50Hz synch
>> clock in a broadcast TV signal? Anyone know where I can find such
>> information?
>
> according to ITU Document 11/30E (1998), fluctuations of line
> frequency can be +/-0.02% and corresponding frame freq. +/-0.0001%.
Which means 216 frames per day.
> but I think, that card is syncing with the source, otherwise, You
> will get garbage instead of images. :)
I thought so. My question was about the fluctuations and long-term drift
of the synchro signals *in the video feed*.
> really, there is a great problem in all those grabbing hw: no cards
> with synched audio. don't ask me why. so, the difference in ADCs
> frequences makes bad game in video. but, it seems, that problem
> touches nobody.
>
In capture, it's a catch-22 situation. *All* sound boards have rather
crappy xtals, thus wrong frequency (on 1 machine I have 2 boards:
32007.5 and 32006.1 Hz. Temperature-sensitive, of course!); thus all
capture software has to cope with non-nominal sound frequencies. But it
is not made easier by V4L and OSS not having any way to timestamp their
samples.
For playback, I can see how it could be a problem. I dont know much
about game software, but does any of them attempt a calibration of the
sound board?
--
Michel Bardiaux
Peaktime Belgium S.A. Bd. du Souverain, 191 B-1160 Bruxelles
Tel : +32 2 790.29.41