[ntp:hackers] Fwd: Video Frame Rate Ref Clock

Mark Martinec Mark.Martinec at ijs.si
Fri Dec 14 13:23:24 UTC 2012


> > I expect that it would be a SMPTE timecode.  I found this mentioned on
> > Wikipedia.  Maybe Bart will describe the video standard being used and
> > we can check what it supports.
> > 
> > I would be worried that the time code might have the time from when the
> > video was created rather than the current time.
> 
> It generally does. And there's a few different time code fields and
> standards to worry about as well (time within the picture and time filmed
> at least). There's another field for the current time, but I don't know
> how extensively it is deployed outside of terrestrial analog video
> broadcasts.

Don't mix SMPTE and other timecode that is used internally in video
and film production, with what reaches the receivers. There is
no longer any timecode there. The only useful data there is
is the horizontal line rate, or the frame rate, which according
to Bart they have it stabilized with GPS. Possibly a colour
burst frequency (if stabilized too) can be used for the same
purpose.

Timestamps will need to come from some other source, perhaps from
another NTP server.

So, the most straightforward approach would be to derive a
pulse-per-second signal from the video signal, as Greg Dowd suggests,
and use that as a PPS source to stabilize the ntpd frequency.

Don't know how complicated it would be to adapt NTP to accept
some other low frequency pulses (not exactly 1 Hz) to stabilize
the clock. For example the NTSC frame rate of 29.97 Hz is
somewhat awkward to convert to 1 Hz.

  Mark


More information about the hackers mailing list