[time-nuts] Correcting jitter on the 1 PPS signal from a GPSreceiver.
Tom Van Baak
tvb at LeapSecond.com
Sat Sep 13 17:15:28 EDT 2014
> I have 2 Trimble Resolution T receivers and I have compared the 1 PPS signal
> between the 2 units. They are spec'ed at 15 nS accuracy. I am seeing about
> 80 nS of jitter between the two. This is with about 6 satellites in view.
That sounds like too much. Are you also comparing to a reference (e.g., Rb or Cs)?
> I was thinking about ways to improve this. Since this is a stationary installation,
> can you use the jitter in the reported location (latitude and longitude) to correct
> for the 1 PPS jitter?
In theory, yes. This is essentially how DGPS (US Coast Guard Differential GPS) and how "Common View" GPS work. But for proper results the calculations must be done on a per-satellite basis rather than at the final navigation (NMEA output) or timing (1PPS output) level.
> The location data is derived using the internal GPS disciplined oscillator so both
> pieces of information should show the same jitter error.
You'd think so. But it turns out much of the "jitter" is due to the antenna and receiver itself. Remember it's trying to detect and track a signal that's so weak you can't see the waveform for the noise. And it has an imperfect internal clock. Consequently, much of the error you see in the nav and timing solutions is not correlated between two (or more) receivers. Just how much I can't say, but it would be nice if you were able to measure this.
> If you compare the reported location with the known fixed location you should be
> able to use that error to correct for the 1 PPS error.
Again, read up on how DGPS and CV-GPS work; also RINEX. It's good stuff.
AFAIK the Resolution-T can output the necessary per-SV raw data that you need to pull this off. The Motorola VP also does. The "T" versions of ublox too. Of course, all the fancy expensive geodetic receivers support this.
More information about the time-nuts