On Fri, Apr 8, 2016 at 12:06 AM, Tom Van Baak <[email protected]> wrote: > Meanwhile it would not surprise me if each GNSS system gives a slightly > different position and a slightly different time than GPS does.
One could easily imagine a system that when signals from both constellations were strong it fit a model matching a secondary system to a primary. Then when the primary was degraded/unavailable used the secondary data with the model-- which would hopefully fix any bias between the systems-- assuming that it's fairly stable. But that would be a fairly specialized mode of operation. If the solutions were pooled with weighing and knowledge of their process errors adding more shouldn't make it worse... but might not add anything: the ideal weight might be zero. I wonder how much of these negative experience with glonass are due to non-calibrated antenna: Glonass does frequency division multiplexing, and so the antenna's angle/frequency dependent phase can harm performance... many GPS antenna greatly attenuate the glonass signal too. For time-nuts operation where the GPS is conditioning one (or more) nice OCXO or atomic clocks in a world where tens of gigaflops of CPU are among the least expensive toys we can buy, I imagine that fairly different signal processing approaches would be ideal: e.g. instead of just computing second by second solutions and putting the error into a PLL, one could collect hours of observation data and produce after the fact correction data that uses prior assumptions about the stability of the local oscillator. Such a processing mode could be more robust against disturbance from multipath and SV failure, as that data could be excluded as hopelessly inconsistent compared to modes that didn't assume a stable local clock and didn't use potentially hours of observation that watched the SV move across the whole sky. It's somewhat annoying that this kind of experimentation first would require something like getting a software GPS implementation working well; when you just want to try changing around secondary processing after all the coorelators and such. _______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
