[xine-devel] metronom clock recovery (was Design issues of VDR plugin)

[Simon T. <si...@bi...>]

Hi,

In reply to the VDR thread. The object is to measure the error
between SCR speed and computer speed, not to minimise the error
of the time measurement itself. The trick is to note that we are
looking for the bias error and that buffer jitter should have
zero mean. So long as packets are arriving at a constant average
rate you can recover the clock.

measure the time interval between a given pair of SCR values
and the local clock time between these time stamps. You want lots
of these interval measurements so that the jitter is averaged away.
Take just one interval over a long period of time and you will still
never be able to measure clock drift that is smaller than the
two buffer jitter errors in your measurements. Thankfully it is the
bias we want, and thats what we get when we take the average of
these measurements.

The article listed below shows how to lock an XTAL to +-22ppm
with a jitter error of +-4msec.

check out the synchronisation article on

http://www.bretl.com/mpeghtml/STCsync.HTM

main page at

http://www.bretl.com/mpeghtml/MPEGindex.htm


Simon
--