PPP with Civilian Signals

With my involvement in NRCan’s online PPP service, I get to see new trends in data submitted to our system. Until recently, dual-frequency GPS receivers could be categorized into two classes: the C1C/C2W and the C1W/C2W receivers. With Broadcom announcing L1/L5 GNSS chips for smartphones last week, dual-frequency positioning using civilian signals will gain great momentum and might trigger changes in analysis centers’ processing strategies. There is however another category of receivers that has attracted our attention lately: the ones tracking only the C1C/C2C signals.

 

Since we are working on a new version of our PPP engine, we regularly benchmark the new software against the “old” one, i.e. the version currently supporting NRCan’s online service. We noticed two completely different ways of handling the data sets containing C1C/C2C signals:

  • The “old” version considered that having single-frequency measurements on half of the satellites and dual-frequency measurements on the other half indicates tracking problems within the second frequency. For this reason, it discarded the second frequency and processed only C1C.
  • The new version identified satellites providing single-frequency measurements as having problems and processed only the satellites with dual-frequency measurements.

On the one hand, we have one software neglecting the dual-frequency capacities of the receiver. On the other hand, we have the other software rejecting satellites and weakening the geometry. The ideal solution of course would be to include all data available in the solution.

 

The figures below show positioning results of a static receiver processed in kinematic mode with the new software, using the three options described above: 1) use only single frequency measurements (SF); 2) use only dual-frequency measurements (DF); 3) use a mix of both (MIXED). The position differences are expressed with respect to the static DF solution, although we should keep in mind that this reference solution most likely did not have time to converge either since code observations were heavily contaminated by multipath (not shown here). The focus will then be on the stability of the solutions.

Fig. 1 Solution using all satellites but only single-frequency observations (SF)


Fig. 2 Solution using only satellites providing dual-frequency observations (DF)


Fig. 3 Solution using all available observations (MIXED)

 

As we can see, even with additional satellites, the SF solution is definitely not as stable as the DF solution since ionospheric effects cannot be properly mitigated (although the DF solution did suffer from a few outliers). The MIXED solution does not show much improvements over the DF solution in terms of stability, which indicates that most of the weight is given to the dual-frequency satellites. Hence, when there is good geometry, the DF solution is probably a safe bet. While the MIXED solution should, in theory, outperform the other two, it is not yet clear if the contribution of the single-frequency satellites would, at times, negatively impact the solution. Based on these findings, the “old” software will most likely be updated to process dual-frequency observations, at the cost of impacting geometry.

 

My piece of advice from a data analyst point of view: users can’t expect to obtain the highest accuracy with missing signals. As long as they are aware of it, then let’s welcome civilian PPP!



Write a comment

Comments: 0