[ntp:hackers] MFS clock driver?

Folkert van Heusden folkert at vanheusden.com
Mon Feb 6 10:39:44 UTC 2006


> >After that, I can read() a new character every second from /dev/ttyS0.
> A much better way would be to attach the signal to DCD so you could
> get precise timestamps with the PPS-API (RFC2783).

True, but I'm trying to get it to work without any modifications. The
windows program (enclosed with the device) can work as is, so should the
unix driver :-)

> I have written a DCF77 receiver which uses what I call "blame-allocation"
> to interpret the timegram and that works great when the DCF77 signal
> is disturbed by noise/fades.
> Instead of trying to find out which pulse is the first pulse in
> the minute, I work backwards and eleminate all the ones that can
> not be.
> For instance in the DCF77 timeformat, the tens-hour field cannot take
> a value of '3' and the tens-minutes field cannot have a value of
> '6' or '7' and therefore a potential minute startbit that would
> result in such illegal values obviously isn't valid.
> With the DCF77 signal this means that I can get minute synchronization
> with as much as 30-40 missing bits per minute and therefore I can
> use quite sharp discriminators on the pulse width to eliminate
> noise pulses.

Clever code!


Folkert van Heusden

-- 
Temperature outside:    7.312500
----------------------------------------------------------------------
Phone: +31-6-41278122, PGP-key: 1F28D8AE, www.vanheusden.com


More information about the hackers mailing list