[ntp:questions] Re: Question on loopstats and PPS

David L. Mills mills at udel.edu
Fri Dec 16 01:15:25 UTC 2005


Mauro,

The idea is to run the thing at the fastest poll rate you can, then 
calculate the Allan deviation involving every sample. Then remove every 
other sample and do it again. Then from this series remove every other 
sample and do it again. Continue until no less than ten samples remain 
and bake for thirty minutes in a hot oven. Serve on log-log plots. Takes 
about a dozen lines in Matlab.

Dave

Mauro Fiacco wrote:

> Hal,
> 
> The PPS signal is derived from an _unsynchronised_ rubidium source. It
> provides a 1sec pulse signal which is interfaced to the serial port of
> the server (/dev/cuaa0 on freeBSD) (no time infos are available)
> 
> I do not use refclocks, I let the kernel to discipline the oscillator
> (The only thing on the conf file is "enable pps")
> 
> I have also slightly modified the ntp_loopfilter.c file such that the
> PLL is not used, only the FLL correction (I modified the status passed
> to the kernel).
> 
> One could use the  127.127.22.0 as preferred clock (not much choise
> anyway!)
> 
> The system works well; however, it is difficult to quantify how good it
> is. The frequency error on loopstats refers to the PC oscillator.
> 
> Thanks again.
> 
> Regards,
> 
> Mauro
> 
> 
> 
> On Thu, Dec 15, 2005 at 04:14:42AM -0600, Hal Murray wrote:
> 
>>Date: Thu, 15 Dec 2005 04:14:42 -0600
>>From: Hal Murray <hmurray at suespammers.org>
>>To: questions at lists.ntp.isc.org
>>Subject: [ntp:questions] Re: Question on loopstats and PPS
> 
> 
>>>I am tryng to evaluate how good is my NTP server. It runs freeBSD and
>>>uses a rubidium PPS signal to discipline the frequency (no time
>>>correction).
> 
> 
>>How are you connecting it up and/or what's in your conf file?
>>Which refclock are you using?
> 
> 
>>I didn't know NTP could do what I think you are describing.
> 
> 
>>The PPS driver (refclock_atom) expects the PPS signal to
>>arrive on the second boundary.  It needs some other source
>>of time (a server marked prefer) to know which second the
>>PPS signal matchs.




More information about the questions mailing list