[ntp:questions] almost decided which new PPS GPS to buy

David Lord snews at lordynet.org
Mon Mar 5 16:43:40 UTC 2012

Ron Frazier (NTP) wrote:
> On 3/4/2012 5:51 PM, David Lord wrote:
>> Ron Frazier (NTP) wrote:
>>> Hi all,
>>> Some of you know from my prior threads that I've been debating what 
>>> GPS to buy to upgrade from my USB based BU-353.  I think I've just 
>>> about nailed it down, based on a number of conflicting factors.
>>> I'm getting + / - 6 ms accuracy now with the BU-353.  My goals for 
>>> the new device are: 1 ms if using USB-serial interface, and 100 us if 
>>> using a real serial interface.  One problem with the BU-353, is that 
>>> it's average offset from the internet servers seems to wander around 
>>> up to 60 ms or so.  When that happens, NTPD sometimes marks my GPS as 
>>> a falseticker.  ESR, the guy mentioned in another thread who wants to 
>>> connect GPS to 100 routers, called this "wobble", and said it's 
>>> related to the SIRF chipset.  I haven't done any other research to 
>>> confirm that.
>> There is sawtooth error when you get to the nanosecond level
>> and some timing GPS can output a correction factor for this.
>> I suspect you mean the uncertainty in timing of the NMEA
>> outputs.
>> I have added a couple of extra parameters to my ntp.conf
>> /etc/ntp.conf:
>> #######
>> # mindist allows sources with large offsets
>> tos mindist 0.3
>> # mode 18 for the Sure
>> server mode 18 prefer
>> # time2 to give offset around zero
>> fudge time2 0.410 refid GPSb
>> server
>> fudge flag2 0 flag3 1 refid PPSb
>> #######
>> With both ntp-4.2.6p3 and p5 I found that PPS had an "x"
>> flag rather than "o" when I used minpoll and maxpoll
>> although PPS was conditioning the clock and PPS offset
>> was around 0 usec.
>> David
> Hi David L.,
> Thanks for the info.
> I'm already using fudge time 2 to correct the offset in the NMEA data 
> I'm getting on my USB GPS.
> I looked up the data for mindist here 
> http://doc.ntp.org/4.2.6/miscopt.html .  I couldn't find a similar page 
> for NTP 4.2.7.  I'm assuming they're very similar.  I'm starting to 
> realize there is potentially a LOT to know to configure NTP beyond the 
> simple basic case.  The page said this:
> quote on ->
> mindist /mindistance/
>    Specify the minimum distance used by the selection and anticlockhop
>    algorithm. Larger values increase the tolerance for outliers;
>    smaller values increase the selectivity. The default is .001 s. In
>    some cases, such as reference clocks with high jitter and a PPS
>    signal, it is useful to increase the value to insure the
>    intersection interval is always nonempty.
> < - quote off
> I'm still not totally sure what it does, but it says it increases 
> tolerance for outliers and specifically mentions reference clocks and 
> pps, so I might try the default value you suggested
> I couldn't find the data for flag2 and flag3, and I was wondering if 
> they're different for different drivers.

The value that I used is not the default. I ran my GPS for
a while with "noselect" and chose a value for mindist that
spanned the range or offsets from the NMEA. This is not
only different for different NMEA sentence selections but
also differs between GPS units and different ntp versions.

For documentation, I check with the source distribution
from ntp.org. Default for NetBSD doesn't or didn't support
my needs so I use downloaded sources. There is a html
directory that contains the documentation. Note ntpd is
still work in progress so you need to refer to
documentation for the exact version you are running.


More information about the questions mailing list