config, was Re: [LinuxPPS] nmea/atom configuration

[gnu not unix]

In message <45B647E5.4070203 at xs4all.nl> you write:
>gnu not unix wrote:

>> I had had just the nmea/linuxpps configured with GPS (and a couple
>> of LAN hosts) and was seeing +/-20 microseconds of flap. 

>> Then, I just added the atom driver to the same /proc/pps/sources
>> (so that both atom and nmea use the same device and edge) and
>> it has smoothed out the plot by a factor of 10 (+/-2 microseconds).

>So you mean one can combine the PPS (atom) driver with a NMEA (including
>PPS) source and gain better performance?

Yes, in this experiment I  had nmea running as the only
refclock (config in ntp.conf), and then added the atom 127.127.22.0 
using ntpdc (dynamic add of refclock). Of course you need both of
the patches to ntpd for this (the nmea and atom patches).

The plot smoothed right out!

>Interesting.

Yes I kept looking at it, I didn't really believe...

>Any caveats for the config?

Another detail, I was using 2.6.19p4-g5f1bb8aa kernel, a git
pull off Rudolpho's code tomb. I updated in order to test the
parallel port code, which worked nicely.

../Steven