After winning the Catalonia Regional Price of the Galileo Masters 2018 for the MEDEA single-frequency GNSS receiver prototype, Rokubun is currently upgrading it to a MEDEA multi-frequency GNSS navigation computer, which is based on the u-blox ZED-F9P multi-frequency and multi-constellation chipset.

MEDEA consists on a sensor board (with the GNSS chipset, IMU, barometer, extra magnetometers, ...) and the application board, with a small yet powerful processor to embed any program you need (NTRIP caster, RTKLib, your own navigation engine, webserver UI...)

The MEDEA sensor board (SB) can operate independently from the application board. This means that is able to collect all the data (GNSS raw observables, IMU, time events, ...) in the SD card for later processing, specially well suited for PPK applications such as drone photogrammetry. The MEDEA SB has the same functionalities as our previous Argonaut receiver. Also, due to the fact that the Ublox chipset contains an RTK engine, you can use it also as a navigation unit if you provide it with connectivity. Actually, the upcoming application board will have 3G and WiFi connectivity to the sensor board for this purpose.

With the sensor board already very advanced in terms of implementation, we are starting to perform some first tests on the data quality we can get from the Ublox receiver. To do this we have set-up a series of dynamic (automotive) and static tests using a multi-band Tallysman TW7972 antenna.

Signals tracked and tracking sequence

The data samples show that it takes several steps and time to fully track all signals in a cold start (with no almanac data stored in the receiver). We have observed the following sequence:

Usually the GPS satellites in L1 frequency are tracked first. The Galileo and Beidou satellites follow after a small delay of few seconds in a cold start or simultaneously (if the almanacs are stored in the receiver).
In a cold start, the GLONASS satellites were tracked after few seconds, but instantanously in subsecuent takes (with probably the almanac stored in the receiver). Usually the L2 is tracked with some delay after L1.
For GPS, the ublox chipset starts tracking the L2C medium code (C2S) and after some samples (always under 1s) switches to the longer code (C2L).
Finally, the Galileo E5a signal is tracked first by one satellite and then the others follow one by one.

It is worthy to note that no GLONASS carrier phase measurements where apparently tracked by the receiver.

A summary of the signals tracked by the receiver is provided (header of Rinex 3 files generated from MEDEA SB Ublox raw data)

G   12 C1C L1C D1C S1C C2S L2S D2S S2S C2L L2L D2L S2L      SYS / # / OBS TYPES
R    8 C1C L1C D1C S1C C2C L2C D2C S2C                      SYS / # / OBS TYPES
E    8 C1C L1C D1C S1C C7Q L7Q D7Q S7Q                      SYS / # / OBS TYPES
C    4 C2I L2I D2I S2I                                      SYS / # / OBS TYPES

Signal-to-noise (SNR)

The SNR of the tracked signals are included in the following plots for GPS, Galileo and Glonass. The following points are noted:

GPS L1 signals higher SNR than L2C, which is compatible with the fact that the IIR-M satellites (that transmit L2C) have a slighlty lower transmitted signal power in L2 (as per the GPS ICD, IS-GPS-200H, Table 3-Va)
Galileo E5b signals not only have higher SNR (1 to 2dB more) but are also more robust (i.e. have less jitter). This is consistent with the higher transmitted power for E5b frequency (as per the Galileo Signal-In-Space ICD Issue 1.3, Table 11)
Similarly to the GPS case, the Glonass L1 signal is steadily stronger than L2, again consistent with the extra 6dB of the Glonass L1 carrier frequency (as per the Glonass ICD, version 5.1, Section 3.3.1.6)

Navigation solution

The automotive dataset gathered by the MEDEA receiver in this campaign has been processed using Jason, Rokubun's positioning-as-a-service (https://jason.rokubun.cat). Jason was able to go for a PPK strategy (i.e. post-processed RTK) due to a nearby base station in Girona (GIRO) from the Spanish IGN network.

As you can see in the plot below (and attached full track in KML), the resulting track is much smoother (i.e. less solution breaks and jumps due to obstructions) than compared with single frequency receivers. There are some spots where vegetation seems to upset the navigation filter a bit, but this may be solved using an IMU such as the one onboard the MEDEA (results of loosely coupling to come in future posts).

You can download this datapackage, that contains a sample Rinex3.03 file (used to generate the solution with our Positioning-as-a-Service) as well as the resulting KML. Please feel free to play with these files as you see fit.

As expected, the benefits of multi-frequency GNSS over single-frequency are obvious, not only in terms of accuracy but, more importantly, in terms of robustness and convergence time. Being MEDEA one of the first receivers featuring the new Ublox ZED-F9P, multi-frequency GNSS is nowadays an affordable solution. In particular, the MEDEA Sensor Board could bring affordable accuracy for applications that do not require real-time accuracy such as drone photogrammetry or surveying.