Will PPP ever replace RTK? Over the past decade, we have seen the convergence of the PPP and RTK techniques towards “PPP-RTK,” where satellite orbit/clock/bias corrections are augmented by local atmospheric corrections to enable instantaneous convergence to cm-level accuracies. But how close are we to instantaneous cm-level PPP-AR, without local augmentation?

In the July 2018 edition of the GPS World Innovation column, Denis Laurichesse (CNES) and myself explained how single-epoch (or “snapshot”) PPP-AR could lead to cm- to dm-level positioning accuracies, thanks mainly to the Galileo E6 signal. We proved this concept practically using a small network of 6 stations in Australia to compute phase biases on all GPS and Galileo signals.

Now that more IGS stations are tracking all constellations and signals, Denis has pushed this concept further by implementing global phase-bias estimation on a routine basis. Using a global network of stations, he derived phase and code biases for all constellations (GPS, GLONASS, Galileo and BeiDou). These biases allow for multi-GNSS and multi-frequency ambiguity resolution, except for BeiDou (widelanes only) and GLONASS (no AR). Ambiguity resolution capabilities on these signals enable instantaneous horizontal accuracies between 1 cm (full ambiguity resolution) and 20 cm (widelane ambiguity resolution). Real-time PPP monitoring using these products is available on the PPP-WIZARD website.

Denis has also implemented a post-processed solution based on GFZ’s MGEX orbit and clock products. The estimated phase and code biases are available on his FTP server since 10 December 2018. New products are being made available in batches every month or so. Daily bias and attitude files are produced, the latter ensuring satellite orientation consistency (more details in a future post). The GFZ orbit and clock files are available on the CDDIS FTP server.

The following two plots, courtesy of Denis, illustrate the type of results that we can obtain from these products. Figure 1 shows epoch by epoch multi-GNSS solutions obtained from fixing only widelane ambiguities: already, a 10-cm horizontal accuracy is achievable *without* precise atmospheric corrections from a local or regional network.

**Figure 1 – Horizontal estimates from multi-GNSS epoch-by-epoch solutions with widelane ambiguity resolution (courtesy of Denis Laurichesse, CNES)**

The magic really happens when performing ambiguity resolution on the full set of ambiguities, as shown in Figure 2. At this point, instantaneous cm-level accuracies become possible!

**Figure 2 – Horizontal estimates from multi-GNSS epoch-by-epoch solutions with full ambiguity resolution (courtesy of Denis Laurichesse, CNES)**

So how close are we to instantaneous cm-level PPP-AR solutions? With these multi-GNSS, multi-frequency GNSS products, you can try it out and answer this question for yourself!

Thanks a lot to Denis Laurichesse who makes these products available to the GNSS community. This is truly AWESOME!

**Reference**

Laurichesse D, Banville S (2018) Instantaneous cm-level multi-frequency precise point positioning. GPS World, July 2018

Write a comment

12#1kfkuang(Sunday, 18 August 2019 22:01)It is really awesome. I am wondering how the satellite attitude be modeled correctly. As far as I know, there is still some problem (the +/- issue) with the model by Kouba(2009, GPSS) or Kuang(2017, GPSS).

#2Simon Banville(Monday, 19 August 2019 04:57)@kfkuang These satellite attitude files are to ensure consistency between the network solution and the users, regardless if the attitude is representative of the true satellite attitude in space.

#3gnsser(Monday, 19 August 2019 22:20)Dear Simon, thank you for these updates reporting exciting results. I have two questions.

The first concerns figure 1. We know that wide-lane ambiguities are weakly correlated with the coordinates parameters as they can be determined from the geometry-free MW observables easily. In this sense, why fixing wide-lane ambiguities are beneficial to positioning? Have you also check the positioning results befoe wide-lane ambiguity fixing? How much is the difference between wide-lane-fixed and -float positioning results?

The second relates to figure 2. How can the instantaneous (single-epoch-based ) full ambiguity fixing possible without atmospheric correction?

#4Simon Banville(Tuesday, 20 August 2019 07:45)@gnsser You comments regarding widelade ambiguities are correct for *dual-frequency* receivers. In this case, we use 3 frequencies for GPS and BeiDou and 4 frequencies for Galileo. When fixing multiple widelanes, you start to see and impact on coordinates.

I don't have the plots for the original float solution, but look at Fig 5 from the GPS World paper: the GPS-only solution is basically a float solution.

Once you have fixed widelane ambiguities, the precision of your solution becomes good enough to be able to fix all ambiguities. Fixing widelane ambiguities also reduces the uncertainty of your slant ionospheric delay parameters; therefore, atmospheric corrections are not essential (they could help though!)

I suggest reading the GPS World paper referenced in the blog post. You might find a lot of useful details on the approach. Cheers.

#5tjliu(Thursday, 31 October 2019 03:58)Dear Simon:

Thank you for taking time to answer my questions. I have read the GPS World paper and blog post, and had some questions about the Snapshot PPP:

1. There are only two (NL and WL) ambiguity parameters to be estimated in the dual-frequency UC (uncombined) PPP, but in the triple-frequency case, there is much more possible combinations. For example, in PPP-WIZARD, we use Galileo E1+E5a+E6 measurements, we can resolve the NL (or N1 of E1 signal) ambiguity and the two WL ambiguities (E1-E5a and E1-E6), and we also can resolve the NL ambiguity and E1- E5a, E5a- E6 WL ambiguity. How to choose the combination of WL ambiguity in the triple-frequency case by using the UC model?

2. The satellite attitude files contains the attitude quaternions of the satellite. But as a PPP user, how to use this file? Can you give me a detailed user's manual about this files?

Yours sincerely

Tianjun Liu

#6Simon Banville(Saturday, 02 November 2019 20:38)Dear Tianjun,

Our GPS World article mentions that we used the BIE method for ambiguity resolution. This approach is a weighted average of integer vectors and it was applied directly to the uncombined ambiguities. We did not form any linear combinations of ambiguities to generate the positioning results in the paper.

Regarding your question on the attitude file, the answer is too long to fit here. Let me write a blog post on it.

#7tjliu(Sunday, 03 November 2019 23:47)Dear Simon:

Thank you for taking time to answer my questions. I appreciate your help.

#8Lihua(Wednesday, 27 May 2020 02:25)Ionosphere Modelling for PPP-RTK:

About this article, three different Networks are defined with different inter-station distance. I wonder whether the slant ionospheric delays from the Network are feed into the ionosphere models in once or not? Since the sat-by-sat model may not be suitable for a so large area, so did the there Networks are divided into few small areas, and then modeled and tested areas by areas?

#9Simon Banville(Friday, 29 May 2020 09:05)@Lihua Would it be possible to give a full reference to the paper? Often, for a PPP-RTK type of solution, slant ionospheric delays from the network solution are provided on a satellite-by-satellite basis at grid points. Using this approach, a wide area can be covered with very precise ionospheric delay corrections.

However, I would agree with you that trying to fit a model (e.g., a polynomial) over a wide area is likely to result in a loss of precision.

#10Lihua(Tuesday, 02 June 2020 06:50)Dear Simon,

Sorry for not mention it's your blog： Ionosphere Modelling for PPP-RTK.

And I read your paper "Network Modelling Considerations for Wide-area Ionospheric Corrections" published at ION+ GNSS 2018: three different models (single layer, dual-layer and sat-by-sat model) are compared. As for the sat-by-sat conical model, if the slant ionospheric delays from the network solution are provided at grid points (as you reply), have you even addressed the effect of the resolution of the grid? Or could you tell me the resultion you used in your paper (published at ION+ GNSS 2018) , since this is not mentioned in your paper?

#11Simon Banville(Tuesday, 02 June 2020 08:40)@Lihua In this paper, no grid was used. As stated in the paper, the interpolation was done on the server side, i.e. simulating a two-way communication (the user sends his position to the server and slant ionospheric delay corrections are sent back for that location). The interpolation was done using Kriging so it is valid regardless of the size of the area.

If we wanted a one-way communication, we could perform Kriging at pre-defined locations and transmit the values to the users. In this case, the grid spacing would be a trade-off between bandwidth requirements and precision.

I hope this helps!

#12Lihua(Wednesday, 03 June 2020 07:39)It make sense! Thanks a lot!