[ntp:questions] Calibrate for Refclock - whats an easy proceedure to get averages from clockstat?

unruh unruh at invalid.ca
Sat Dec 10 21:23:59 UTC 2011


On 2011-12-10, Mark C. Stephens <marks at non-stop.com.au> wrote:
> Goodness me, I asked the question in the subject!
>
>
> Just fiddling around getting fudge times right and I came across the procedure to get the fudge 2 time.
>
> I found I was making a couple of mistakes, I had more than one sentence transmitted a second, now only GPGGA is sent. By the way, that's another question, is using a short sentence like GPZDA instead of GPGGA beneficial accuracy wise?

Should not matter. The biggest problem is that many of the nmea drivers
timestamp the end of the sentence, not the beginning. Since sentences in
general have variable lengths, and since each character takes at least
1/2 ms to be transmitted this in itself gives a jitter. But the same
sentence should occur in the same place in the string of sentences (but
of course the jittr from previous sentences would add.)
>
> Secondly, Should I enable PPSAPI (Flag1 1) on NMEA - I have PPS ready to go, but I will get NMEA times right before I tackle that one. The question is, Can I enable NMEA flag1 _and_ have PPS driver as well? Will it break something or just pointless?
>
> Thirdly, What's an easy way to get an average delay out of clockstats file. I want to leave it running for 24 hours and I thought to do it with excel, but that will require a lot of editing of that big file. Is there an easy  way to do what I want in Unix?

Peerstats, not clockstats. Peerstats gives the raw times, while
clockstats has massaged the times. 
Sure, use awk. Or perl.


>
>
> Thanks for your help so far Guys!
>
>
>
> Mark



More information about the questions mailing list