[ntp:questions] Re: 1pps signal,20us width

Tim Shoppa shoppa at trailing-edge.com
Tue Nov 9 14:48:28 UTC 2004

Fei Wu <fxw115 at yahoo.com> wrote in message news:<mailman.8.1099501654.537.questions at lists.ntp.isc.org>...
> > > I remember I have seen the similar question before
>  in
> > > this list, but I can only get around 200 posted
>  emails
> > > by using email newsgroup. Appreciate if anyone can
> > > tell me how to check all the emails.
> > >
> > > I have a time server with 1pps output. The width
>  of
> > > the pulse is around 20us. I know generally people
>  will
> > > use a gadget box to enlarge the pulse to around
>  100us
> > > to fit 1bit as 9600bps com port.
> > >
> > > Is it possible for me to set the bit rate to 56K?
>  So
> > > 20us pulse can be treat as one bit width;
>  therefore,
> > > ntp can lock the edge of pulse?

20 microseconds will generally be long enough to trigger
PPS detection through a RS-232 DCD pin.  Most hardware
interrupt schemes for the DCD pin are edge-sensitive
(although I've seen some set up for polling a level-sensitive
line, you don't tell us what hardware/OS you have.)

1 or 2 microseconds (as the PPS pulse from a Trimble SV6
is) will not reliably trigger a PC-clone's RS-232 hardware.
Even with fast PC-clones I saw 50% of the pulses missed.

The modem control lines are not really tied to the baud
rate in any meaningful way.

That said, 20 microseconds is pretty short for a RS-232
pulse.  If the RS-232 cabling is more than a few feet
then there may be interface problems.  Stretching the
pulse with a one-shot (ala 74LS123) is trivial.  It also
gives you one more blinking light, if you want :-).


More information about the questions mailing list