Dynamic for F9P isn't static

[Jef239]

Add CFG-NAVSPG-DYNMODEL,2 to U-Blox_ZED-F9P_rtkbase.cfg

By default dynamic is PORTABLE. It's wrong for base station.

[Stefal]

In my opinion, this is unuseful because the navigation engine isn't used in a base station.

[Jef239]

Ooooops! I didn't expect that!

I see the following options for checking.

1. If you have a splitter and two receivers, you can record measurements on the zero base and calculate the second differences in different dynamic modes. That is, you can actually see the code noise and phase of the receiver in different modes.
2. If you can connect to Onocoy, you can see the values of the "Code RMS" and "Phase RMS" scales for a day in different modes.
3. If you trust me, I can take second difference graphs for you. True, on a ZED-X20P, but these are details.
4. You can read my lecture and just believe it.
5. You can double-check my words with any developer of GNSS receivers, or with Tomoji Takasu (the author of RTKLIB) or with the competent developers of Centipeda.

[Jef239]

How does the phase meter work? It is clear that we do not count each period at a frequency of 1574.75 MHz. This is impossible. The phase meter in coherent mode calculates the difference between the ideal signal and the real one.

When at rest, this difference is tenths and hundredths of a hertz. That is, you can increase the gain (we say "clamp the phase tracking ring") so that the difference occupies the entire amplifier scale. Fractional noise will have a minimal effect. As an example, the entire scale will be +/- 0.25 hertz, and fractional noise at 8 bits of ADC is 0.002 hertz, that is 0.4 mm.

But what happens when moving? For example, when a pedestrian moves? In 1 second, a pedestrian moves 1 meter. That is, 5 wavelengths (if the movement is directly towards the satellite), and the phase changes by 5 hertz. This will cause the phase meter scale limit to be exceeded, i.e. SLIP. SLIP is the skipping of a certain number of phase meter revolutions (hertz) due to too much gain.

Therefore, in Pedestrian mode, we must loosen the tracking ring, i.e. have a lower gain. For example, the entire scale is +/- 5 hertz, fractional noise at 8 ADC bits is 0.04 hertz, i.e. 8 mm.

Approximately the same picture in the correlator (measures pseudorange) and in the dopplerometer (measures frequency).

And it is the maximum clamping of the tracking rings that distinguishes different speakers. All other numbers are secondary, they are 90% a consequence of different gain factors.

The remaining 10% are coefficients in the navigation calculation, showing what to put into motion according to the coordinates, and what to the unevenness of the clock generator. Plus different coefficients of the Kalman filter at the output of the solution.

This is an old scheme. In modern receivers - it is more complicated. The movement can be noticed by a rapid change in the value of the phase difference and "unclench the tracking ring", that is, reduce the gain. Therefore, instead of speeds, the dynamics were first limited by accelerations, and then jerks (acceleration derivatives) came to replace accelerations.

Well, if you put it very simply - then so.

[Jef239]

A little about myself. Experience in RTK development - 20 years, PPP - 10 years. I have my own, proprietary RTK/PPP code. Very roughly - my own RTKLIB. The last 3 years - I have been working at GNSS.STORE. Author of ELT_RTKBase, which is an extension of your rtkbase. https://gnss.store/95-cors-fanless-stations

A small tutorial on GNSS was written, here it is
https://gnss.store/blog/category/gnss-basics.html
https://gnss.store/blog/category/about-gnss-receivers.html

What is interesting to you personally - here are recommendations for setting up Mosaic-X5 and Unicore UM980 in base station mode https://gnss.store/blog/post/optimizing-onocoy-quality-via-receiver-settings.html

[Stefal]

If you can share a method to analyze and compare the signal, I could test it on 2 UM980. I can't do it with 2 F9P because I have only one available.

[Jef239]

About the zero baseline double difference test - https://www.comnavtech.com/about/blogs/388.html

I can't provide the source code - it's proprietary. I might be able to provide Windows binaries, but I'll share them with you personally, not post them on GitHub.

It's useless to look at the UM980; when MODE BASE is selected, it always sets the dynamic range to STATIC. You can create an undocumented MODE MOVINGBASE, but I haven't tested whether it allows you to change the dynamic range.

The F9P always uses "moved base" mode. It switches to true baseline mode when setting fixed coordinates and static dynamic range.

We (GNSS.STORE) have quite a few receivers and 4- and 8-channel splitters. So, I'll take some pictures for you on the X20P. And if that's not enough, I'll take some pictures on the F9P.

For now, these are old pictures from the F9P; I don't remember which dynamic range was used.

These are the second differences in the code with a common leading satellite.

Second рhase differences with a common leading satellite. It's clear that there are intersystem shifts.

Second phase differences for a common master satellite for GPS and GALILEO. Everything is fine here.

Second phase differences for GLONASS. Inter-character shifts are visible, that is, shifts between satellites operating on different frequencies.

Second phase differences for Beidou. A half-cycle shift relative to GPS is visible.

Second phase differences with a separate leading suppository for each system.

All of this was taken five years ago or more. So, everything could be much better with the new firmware.

The code scales are in meters, and the phase scales are in cycles.

Well, now you know how RTK receivers are tested.

I'll take fresh measurements; it shouldn't be too difficult for me.

[Jef239]

Tests on ZED-X20P HPG 2.00 (1e963fM)

STATIC. Code noise is approximately 25 centimeters.

PORTABLE. The noise level of the code is approximately 75 centimeters. THREE times greater.

STATIC. Phase noise is approximately 0.01 cycles, or about 2 millimeters.

PORTABLE. Well, there's a lot going on here. Systematic trends, shifts. The firmware is clearly raw.

Is this enough for you, or would you like to see it on F9P?

Just in case, here's the raw data.

portable3_portable9_Code.csv
portable3_portable9_Phase.csv
static3_static9_Code.csv
static3_static9_Phase.csv