"Gary E. Miller" <[email protected]> writes: > Yo Greg! > > On Tue, 25 Nov 2025 17:54:26 -0500 > Greg Troxel <[email protected]> wrote: > >> "Gary E. Miller" <[email protected]> writes: >> >> > Much better results than the one hour plot, but even weirder. See >> > attached. CEP(50) of 0.0853 m. But one excursion in Latitude of >> > 1.2 m, and another in Longitude of 1.1 m. Clearly not Bayesian. >> >> As always, there is a difference between "results are consistent" and >> "results are correct". But without first getting consistency correct >> is hard to talk about. > > I prefer to say the restuls are the results. They are what they are. > The hard foundation on which theories are built.
My point is that you are saying CEP as if the mean or some such is the true answer. A tight cluster of positions that is 3m from the true position is going to have a good CEP by your analysis, but it's still 3m off. >> My take from watching that is that there are significant, perhaps >> dominant, periods when the device is in RTK FIX. > > I almost never see RTKFIX. I think it is dropping into simple DGPS > (SBAS) mode for a short period. huh. In my book, RTK without FIX is deficient. >> And, periods when FIX is lost, and it's RTK FLOAT (or maybe maybe even >> just autonomous/DGPS). I find that there is a slow wander to FLOAT. > > I hace not seen fix lost. I'll see if I can write a program to extartc > $ of time in each assist mode. That would be most enlightening. >> It would be interesting to process the data by looking at only points >> in FIX, and to plot the non-FIX points separately. > > So yeah, when I'm guiding my car with the GPS, I'll just freeze when > assist mode changes? I don't understand why you said that. The first question is to understand what's going on, and then there's a later question about how useful it is. If it turns out that plots segregated by mode show that different modes have different error statistics, that's really useful to know.
