I have been post-processing some Emlid Reach 24 hour stationary files collected at 1Hz, after trying a few different processing parameters I ended up with a few files with large errors (10 metres) on a dataset that I would consider benign in terms of raw data quality (see screenshot).
Can some one please take a look at the RINEX files (base and rover) located here: https://drive.google.com/file/d/1LxZcnaeVoOfm-vFD0tJrOSjZiEQfXdem/view?usp=sharing
And let me know why you think the solution keeps jumping off fixed ambiguities towards float even though there is no physical change in the antenna? Note that I have tried Tomji Takasu 2.4.3 build 29 and the results are even worse than with Emlid b28 build but I obtained slightly better results with RTKlibexplorer b29b demo5 build (the initial convergence displayed fairly large errors kept for quite a bit of time but after fixing the results was better)…
This is why post processing is so cool.
If you look at the clock, around 00.00 something happend and it throws the calculation of.
Its easy to see but if you process data before and after 00.00 separately and merge them later, you get nice continuously fix
I’m feeling lazy today, shall I understand that your first screenshot was gathered out of RTKlib Emlid b28 forward only PPK (sorry for the laziness)?
When you say “look at the clock” I’m fairly sure you are not looking at the RINEX file as no clock information is stored there other than the pseudoranges (I did look in to the ubx file and surprise surprise it is the night from saturday to sunday so it is the transition from GPS week 1980 to 1981 and the clock reset flag is green)
You nailed the problem however I’m a bit concerned about this, how often the NEO-M8T resets the clock, only once a week? is there any plans from Emlid to iron the behaviour of Tomoji’s implementation?
