[ntp:questions] Re: NEMA driver on Fedora Core 3 or 4
davids at webmaster.com
Mon Oct 17 12:03:21 UTC 2005
"Hal Murray" <hmurray at suespammers.org> wrote in message
news:-ZydneiIDqU_687eRVn-hA at megapath.net...
> Anybody got one running happily?
> I've got a GPSClock 200. The serial port is off by 500 or 600 ms.
> That's running FC 4. I've got a similar setup at work running FC 3.
> It has the same quirk.
That's not a quirk. The GPSClock 200 specifies a precise relationship
between the NMEA output and the actual second boundary. Specifically, the
NMEA output is the time that it will be at the next PPS output. You *cannot*
use the NMEA output without the PPS output.
> Both systems are recent Dell PCs: P4/Xeon with hyperthreading.
> I tried the non-SMP kernel. That didn't change anything.
> I'm pretty sure the GPS hardware is OK since I get the right
> answer from the other half of the splitter on the serial cable
> which is plugged into an old system.
> I do get sensible results on the new system with a HP Z3801A.
> I'm talking about the serial port, not the PPS signal. (So
> this isn't a PPS inversion mixup.)
You can't use the serial port without the PPS signal. The NMEA output
alone has no specified time accuracy.
I designed the clock you are using, so I happen to know what I'm talking
There is a GPSClock 200 driver for NTP. However, you can get away
without one by fudging the NMEA driver by .5 seconds and using the standard
atom driver. The NMEA driver will get you to within .1 seconds, which will
allow the PPS code to lock on.
If you are using Linux and don't want to use a PPS kernel, you can use
my user-space PPS code that works with NTP.
More information about the questions