Galileo Intra-Frequency Biases

NRCan has been operating a zero-baseline GNSS calibration site on its premises for the past few months. The goal of this investigation is to better quantify the interoperability of different receiver/antenna types. Preliminary results obtained from this exercise were presented at the IGS workshop 2017 a couple of weeks ago (MacLeod et al. 2017), and I thought that I would comment further on some of the Galileo results obtained.

 

The setup available at our calibration site currently consists of an AOA antenna feeding a 16-way signal splitter. Data from 5 geodetic receivers is being collected at the moment: Javad, NovAtel, Septentrio, Topcon and Trimble. I was a little bit baffled when I first looked at the data signals provided by each receiver, as shown in the table below:

 

 

As you can see, the codes tracked for each frequency are quite different from receiver to receiver which, I thought, is a significant limitation to Galileo data processing. Even in terms of RTK, mixing receiver types should require the availability of satellite intra-frequency differential code biases (DCBs), for example C5X-C5Q, etc. Otherwise, these unmodeled biases would propagate into the initial position estimates and, depending on their magnitude, could prevent successful ambiguity resolution.

 

Data collected on a zero baseline (i.e. receivers sharing an antenna) is a great way of assessing the magnitude of these biases. Differencing measurements between two receivers cancels all error sources but the difference in receiver clock offsets. This quantity can then be estimated in a least-squares adjustment and the mean of the code residuals is considered to be representative of the between-receiver code biases. Our investigation replicated previous studies showing that decimeter-level biases can exist between receivers even for GPS signals on the same modulation, such as Trimble C1C – Septentrio C1C.

 

Surprisingly, the picture was quite different for Galileo signals. Even when mixing modulations, there were no (significant) biases detected, even though I was expecting to see intra-frequency satellite DCBs. The following figure shows the estimated biases between the Javad ‘X’ codes and the Septentrio ‘C/Q’ codes on different frequencies (similar results were obtained with the Trimble receiver). The error bars show the repeatability over 7 days.

Fig 1 Galileo intra-frequency code biases between Javad and Septentrio receivers

 

Although this finding was new to me, an analysis of the interoperability of Galileo signals has already been reported by Sleewaegen (2012). He mentioned that, in the satellite, E5a and E5b D/A converters are at IF and not baseband, which should considerably reduce I&Q mismatches and lead to smaller biases than for the GPS L5 signals for instance. From the plots presented in his presentations, he showed cm-level biases between the pilot and data signals but it is not possible to attribute the biases to either the satellite or the receiver. In Fig 1 above, any receiver-dependent bias that is common to all satellites would be absorbed into the clock parameter.

 

As a subsequent test, I computed a code-based positioning solution based on the C8X and C8Q signals for epochs with at least 5 Galileo satellites available:

Fig 2 Kinematic position estimates for a zero-length baseline using the Galileo C8 signals with mixed modulations (C8Q - C8X)

 

Even though more work is needed to confirm these results, the unbiased position estimates obtained support the hypothesis of insignificant intra-frequency biases at the satellite. This great feature, along with the low-noise of the C8 signals allowing for cm-level code-based positioning on short baselines (see figure above), will make Galileo an invaluable asset for RTK.

 

I would like to thank Ken MacLeod of NRCan for all of his efforts in bringing this calibration site to life.

 

Reference

MacLeod K, Banville S, Ghoddousi-Fard R, Collins P (2017) Analysis of GNSS receiver biases and noise using zero baseline techniques. IGS Workshop 2017, Paris, France.

 

Sleewaegen J-M (2012) New GNSS signals: how to deal with the plethora of observables? IGS Bias Workshop 2012, Bern, Switzerland.



Write a comment

Comments: 2
  • #1

    Javier Tegedor (Tuesday, 01 August 2017 00:52)

    Hello Simon,

    very interesting article, thanks for sharing the results.

    This is consistent with our experience when trying to do double-difference ambiguity-fixing in network solutions using the Melbourne-Wübbena approach. It works with Galileo (to the same level as GPS) even when mixing different Galileo signals from different receiver brands. The obvious explanation is that there are no significant biases between different signals/modulations, which is confirmed by your zero-baseline study.

    Perhaps surprising, but certainly a good feature of Galileo for high-accuracy applications!

    Kind Regards,
    Javier

  • #2

    Simon Banville (Wednesday, 02 August 2017 20:40)

    @Javier Thanks a lot for sharing your findings. Indeed, this is great news for Galileo processing!