ECEF results do not match expectations

I’m setting up a permanent RTK Base Station, following the process detailed here: How to Build a DIY GNSS Reference Station - SparkFun Learn

I’ve collected 24 hours of RAWX data, converted it to a .obs file, and sent it to the Canadian CSRS-PPP service. The results weren’t as good as those in the example:

vs.

+/- 12 cm still seems pretty good, but should it be better? I noted that my 24-hour file was ~50 MB, not 300 MB as described in the tutorial, so maybe I’m not collecting all the parameters I need?

Size could depend on number of satellites, constellations and observation interval.

The error likely more related to the amount of noise, distance to the closest reference station, and placement/Skyview the antenna has.

ECEF, doesn’t really seem to be illustrating that here. In what datum?

20-30 mins should be sufficient to get cm level results. At 6km I used to be able to get deviation numbers sub-cm when post-processing uBlox M8 L1 GPS data against CORS.

300 MB seems more probable for 10 Hz raw measurements. Many CORS are at 1/5th Hz, and archived at 1/30 Hz

Here is the Cartesian coordinate information:

Interestingly, “GNSS Data” only indicates GPS and GLONASS, while I saw Chinese and European satellites in u-center.

I was wondering about your comment about “distance to closest reference station”–I was under the impression that this process was dependent solely on the satellite data, and not ground stations, but am I mistaken?

Post-Processing is the geometry of your observations vs observations in the same instance of time from a fixed and known location. ie the Reference. The baseline is the vector between your unknown location and that of the reference, and then the carrier cycle distance between the reference location and the same satellite/signal, and then the carrier cycle distance and your location as expressed in the RINEX file you generated. The satellites are moving, and the Earth is rotating, so the solutions are in motion, and the error reflects in how those solutions jitter in time vs them being tight and predictable, ie cycle lock vs cycle slipping, or not cleaning resolving. Think of the local clock stability, and any wobble / wind-loading on the antenna and mounting pole.

If the reference station doesn’t have data for BEIDOU or GALILEO satellites, it’s not going to be able to do the geometry on that data which you collected. It requires a common view of a set of satellites and signals.

Much of the US CORS network only provides GPS, and not GLONASS in recent year. Depends on the equipment installed locally and political stances.

Ok, thanks. I’ll try submitting my data to NOAA, and see if I get better results.

Hope the NOAA submission gives you cleaner results — this was a really informative thread to follow. The constellation point was new to me.

@Jack3, that is correct. CSRS-PPP is a PPP solution. It does not use a baseline or a reference station.

You results say ultra-rapid orbits were used by CSRS (which is GPS + GLONASS only).

Submit again at least 18 hours after the end of the UTC day when you finished collecting. Then CSRS will use Rapid orbits with GPS + GLONASS + Galileo.

The best results will be Final Orbits, 2 weeks later.

But 12cm seems bad for a 24 hour PPP mission even with Ultra-Rapid Orbits. Did you antenna have a clear view of the sky with no trees or buildings nearby?

Your solution looks noisy. Maybe the antenna is not in the best location? Some buildings/trees
blocking part of the sky?

For PPP local reference stations don’t come into play. They are not used in any way.

For static data CSRS-PPP only uses 30 second data. You can send it one second data, but it will just decimate the file.

CSRS-PPP works with GPS, GLONASS, and Galileo data. With the Ultra-rapid orbits it is just GPS and GLONASS, no Galileo as they don’t have the appropriate Galileo product until the Rapid orbits are available roughly a day after the data is collected. You also need to have the appropriate signals for each constellation. It’s all explained on the web site.

If the antenna is in a good receiving area with 24 hours of data just a GPS only solution will be good. The other constellations won’t add much. Also, wait a day or so until the Rapid orbits are available. That provides a better solution especially on shorter duration files. Waiting 2 to 3 weeks for the final orbits is best, but the accuracy jump isn’t as big as going from the Ultra-rapid to Rapid orbits.

The estimated error in the PPP solution is not necessarily the error in the returned coordinates. Estimating errors is very tricky.

I agree that probably your raw data isn’t good enough. DIg into the rest of the PPP report, which should show cycle slips etc. You did not mention which antenna you are using, and if it has an NGS/IGS cal file, if you selected that, and which receiver.

Uploading to NGS OPUS will also be interesting, but if you are using an F9P it is unlikely to work; the F9P doesn’t sample precisely on the second, for no apparent reason. If you are using a Mosaic X-5 it should go much better.

I also suggest that you gather 4h of data on multiple days (5 to start) and then graph those results and error circles. Until you’ve done that, you are assuming a lot.

The point about ultra-rapid vs precise orbits is valid, but I don’t think it’s your biggest problem.

I tried again, with the antenna located in a more open space, and the results were much better:

Here is the original antenna location, mounted to the barn:

And here is the antenna located in the field:

I only got about 8 hours of data, perhaps the laptop didn’t like spending overnight in that red bin.

I see I’ll need to sort out a better permanent location for the antenna, but I think I can proceed with the next steps for setting up the base station.

Thank you everyone for your help!

@Jack3 - Thank you very much for the photos! This really shows, what I would consider, a medium amount of tree coverage on one horizon will do to a survey. I would not have expected that much interference.