[time-nuts] GPS message jitter (preliminary results)
kb8tq at n1k.org
Mon Jul 18 16:19:56 EDT 2016
Very nice data !!
Hmmm….. software averaging against the MCU (or FPGA) clock source …
NTP seems to get some pretty good results down into the single digit ms (or lower)
doing that based on some equally jittery inputs.
Looking at the data by eyeball (maybe not the best way): A 10 to 30 sample finite
impulse average (boxcar) would do a lot for the result. Yes, you could get much fancier.
> On Jul 18, 2016, at 3:50 PM, Mark Sims <holrum at hotmail.com> wrote:
> That plot was from a 12 hour run. I've done 48+ hour runs and did not see anything strange. I'm not currently measuring the offset of the message from the 1PPS, just the stability/jitter in the timings of the last byte of the timing message. If the timing message offset time from 1PPS jumped, it would show up as a spike in the message jitter (and if the offset jump was not particularly large, would be hidden in the normal noise of the end-of-timing-message time stamps.
> Adding the message jitter measurements to Lady Heather was a quick and dirty 5 line code hack to investigate whether one needed a 1PPS signal to usefully drive a clock display. So far it looks like you could easily get +/- 30 msec stability from a cheap GPS without the 1PPS (and much better with some software averaging)... and that is on a non-real-time, multi-tasking operating system going through who knows how many layers of serial port message buffering, etc. Using a dedicated micro to do the deed should be quite a bit better. You should, of course, measure and compensate for your chosen receiver's message offset time to the 1PPS to get an accurate time display.
> How long did you run your tests?
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
More information about the time-nuts