A Close Look at GPS SVN62 Triple-Frequency Signal Combinations Finds Carrier-Phase Variations on the New L5
By Oliver Montenbruck, André Hauschild (DLR/German Space Operations Center), Peter Steigenberger (Technische Universität München), and Richard B. Langley (University of New Brunswick)
The recently launched Block IIF satellite (SVN62/PRN25) is the first of a new generation of GPS satellites designed to transmit ranging signals for civil users on three frequencies: the C/A-code on L1 at 1575.42 MHz, the L2C-code on L2 at 1227.60 MHz, and the I5/IQ codes on L5 at 1176.45 MHz. Unlike L2, the L5 signal is located inside the protected Aeronautical Radionavigation Services (ARNS) band, which makes it specifically useful for safety critical aviation applications. In combination with the legacy L1 signal, civil aviation users can now perform ionospheric corrections without referring to the L2C signal. Compared to L2C, the new L5 signal offers a much higher chipping rate (the same as the encrypted P-code signal) of 10.23 MHz, which promises a lower ranging noise and better multipath resistance. L5 signals have already been transmitted for some time by the geostationary satellites of the United States’ Wide Area Augmentation System (WAAS) and are now about to become an integral part of the GPS constellation.
Following a short test transmission on June 17, 2010, the L5 signal was continuously activated on the morning of June 28. According to GPS officials, the checkout of the satellite is proceeding nominally and all signals have been found to fully comply with specifications. This will allow the satellite to be set healthy as soon as all commissioning tasks have been completed.
Scientists have long discussed the potential of new signals for multi-frequency, multi-GNSS applications, and expresed a great interest in signal combinations, particularly those of carrier-phase measurements, involving all three frequencies simultaneously. The use of triple-frequency combinations has, for example, been demonstrated to be of great interest for ambiguity resolution in precise carrier-phase-based positioning, for receiver autonomous integrity monitoring, and for ionospheric research (see the articles in Further Reading).
In consideration of the multitude of proposed applications for triple-frequency combinations, we took a close look at the quality of the new GPS L5 carrier-phase signal. For this purpose, we made use of measurements from the COoperative Network for GIOVE Observation (CONGO), jointly established by the German Federal Agency for Cartography and Geodesy (BKG) and the German Aerospace Center (DLR). CONGO is the first network of multi-constellation, multi-frequency GNSS receivers offering worldwide tracking of the SVN62 space vehicle on all frequencies (see Table 1).
As suggested by Andrew Simsky (see Further Reading), the availability of carrier-phase measurements on three frequencies offers a particularly simple way to assess carrier-phase quality and multipath effects. By forming a linear combination
of the L1, L2, and L5 carrier-phase ranges with the additional conditions
a geometry- and ionosphere-free measurement is obtained, which reflects a weighted sum of the carrier-phase multipath and measurement noise on the individual frequencies. Here λ i with i = 1, 2, and 5, denotes the wavelength of the L1, L2, and L5 signals, respectively. Since the above conditions determine the factors α, β, and γ only up to an arbitrary scaling factor, we furthermore impose the normalizing conditions.
The latter condition ensures that the noise of the tri-carrier combination will match that of the individual carrier phases if the measurement noise is equal on all frequencies. As a result, we obtain the coefficients
Introducing the carrier wavelengths of the L1, L2, and L5 signals, the coefficients attain the values (2)
It can be recognized that the tri-carrier combination is dominated by the L2 and L5 signals due to the proximity of their respective frequencies. Noise and multipath errors of L2 and L5 measurements are thus most prominently seen in the resulting combination, whereas any L1 phase errors are strongly attenuated.
A long pass of L1, L2, and L5 code and phase measurements from the new Block IIF satellite was recorded by the O’Higgins station of the CONGO network shortly after the activation of the L5 signal generator on June 28. The SVN62 satellite was tracked for more than 6 hours and achieved a peak elevation angle of more than 75° on this date.
Figure 1 shows the resulting multipath combination computed from carrier-phase measurements of L1 C/A-code tracking, semi-codeless L2 P(Y) tracking (rather than L2C), and L5 I/Q tracking. The data have been leveled to a zero mean over the entire pass to remove the impact of the unknown carrier-phase ambiguities. Except at low elevation angles, near rise and set of the satellite where signal strengths are low, the tri-carrier combination shows a very low noise level that is consistent with the expected carrier-phase noise on all three frequencies. However, a pronounced long-term variation with a peak-to-peak amplitude of almost 20 centimeters may be recognized, which certainly comes as a big surprise and cannot be explained by local multipath. Frequency-dependent differences of the effective phase centers of the receiving or transmitting antennas can likewise be excluded, since these would result in a purely elevation-angle-dependent variation.
Looking at the entire set of measurements from all available CONGO stations, we could rapidly recognize that the variation of the tri-carrier combination with time is essentially the same for all stations with a common visibility of the SVN62 space vehicle, irrespective of the employed receiver and antenna. This suggests the presence of time-varying inter-frequency biases in the L1, L2, and L5 carriers transmitted by SVN62.
Thanks to the global distribution of the CONGO stations, the SVN62 space vehicle is always tracked by one or more stations, which enables a continuous monitoring of the L1/L2/L5 carrier-phase consistency. By adjusting the unknown offset of the tri-carrier combination for individual tracking arcs in such a way as to obtain a best match of consecutive and overlapping arcs, the variation can be traced over multiple days as shown in Figure 2. The graph shows a distinct orbital (that is, 12-hour) periodicity with a superimposed twice-per-revolution harmonic. In addition, a pronounced drift can be recognized for up to one day after activation of the L5 signal generator. Both observations suggest a temperature-dependent line bias in one or more carriers as a likely cause of the observed variation in the tri-carrier combination. (A line bias is a circuitry delay common in all observed satellites and is usually absorbed in the estimated clock offset.) However, an independent analysis of SVN62 temperature data from the onboard telemetry will be required to confirm the validity of this assumption. The space vehicle is in a deep eclipse orbit right now and therefore experiences substantial changes in its thermal conditions. However, the extreme points of the carrier-phase variation in Figure 2 are slightly shifted with respect to the local space vehicle noon (at 01:30 and 13:30 UTC) and the eclipse intervals (07:00–08:00 and 19:00–20:00 UTC).
While the tri-carrier combination provides a very sensitive measurement for the analysis of differential delays between the individual carriers, it does not allow us to uniquely attribute the observed variations to one of the three signals. We therefore made use of code measurements (pseudoranges) to further investigate the consistency of specific sets of measurements. Since the observed variation of the tri-carrier combination exhibits an amplitude comparable to the noise level of the code measurements, a suitably chosen code-carrier combination can indeed help to identify which signal or signals are affected by line-bias variations. To this end, we consider a generalized form
of the well-known code-multipath combination, in which we difference the code measurement Pi at frequency i with an ionosphere-corrected combination of carrier-phase ranges Lj and Lk at frequencies j and k. In so doing, we remove geometric contributions along with clock and atmospheric variations, leaving primarily code multipath, receiver noise, and any signal perturbation that is not coherent on the involved frequencies. In the traditional case of dual-frequency tracking, the frequency of one of the involved carrier-phase measurements is necessarily identical to that of the code measurements. With triple-frequency tracking, in contrast, we are free to consider a larger variety of combinations. For the analysis of the SVN62 signals, we have specifically evaluated the L5 code-multipath combination using (a) the L5 and L1 carrier phases
and (b) the L2 and L1 carrier-phase measurements
The results shown in FIGURE 3 reveal a dramatic difference, which clearly hints at the L5 carrier as the main source of the observed carrier-phase variations.
In the first case, a variation close to that of Figure 1 is obtained, albeit with a 5–6 times larger amplitude that reflects the different weighting of the L5 carrier phase in the corresponding measurement combinations. A good consistency, in contrast, is obtained for the L5 code measurements when differenced against the ionosphere-corrected combination of L1 and L2 carrier-phase measurements.
Overall, we may conclude that the L5 carrier of the SVN62 space vehicle exhibits quasi-periodic line-bias variations with an amplitude of about 10 centimeters in relation to the L1 and L2 carriers. The L5 code measurements, in contrast, appear to be consistent with both the code and phase measurements on L1 and L2 at the respective noise levels. Further observations at a later time will be required to see whether the observed amplitude of the L5 phase variation is specific to the current eclipse orbit and whether it will possibly become lower when a higher angle of the Sun with respect to the orbital plane (the so-called beta-angle) is achieved.
What are the possible consequences of the L5 phase-bias variations for users of the new L5 signal? Evidently, new positioning services building on the L5 code measurements (and possible combinations) will not at all be affected! Even in the case of carrier-phase smoothing, the smoothing time scale will be much shorter than the periodicity of the carrier-phase bias variation. The L5 code measurement quality itself is well within the system specification and no concerns exist that would prevent the satellite from soon being declared healthy.
With respect to carrier-phase-based positioning applications, it is important to note that the L5 line bias acts like an additional frequency-specific satellite-clock offset. This has, for example, been confirmed in preliminary tests of SVN62 orbit determination conducted by the Technische Universität München. Orbit solutions using L1 and L5 measurements from the CONGO network differed by typically 15 centimeters (3D root-mean-square error) from reference orbits obtained by the Center for Orbit Determination in Europe analysis center using the IGS L1/L2 receiver network. At the same time, however, the L1/L5-based clock solutions showed a periodic offset from the L1/L2-based values that reflects the same variations as the tri-carrier combination discussed above.
As a common error for all receivers, the L5 line bias fully cancels in differential processing. Care must be taken though, that satellite clock offsets derived from L1/L2 carrier-phase observations cannot be employed for precise point positioning using L1/L5 measurements without explicit consideration of the inter-frequency carrier-phase bias. Likewise, efforts to correct second order ionospheric effects through the use of triple-frequency measurements are likely to suffer from an imperfect knowledge of the L5 bias and its variation with time.
Whereas some of the proposed ideas for triple-frequency processing may be difficult to materialize at present, a better characterization of the SVN62 L5 signal will certainly help to exploit the available benefits of the new signal and to establish refined processing schemes for scientific and other demanding applications. A continued monitoring of the L5 line bias and its variation with time is therefore deemed necessary and should be supported by a large number of suitably equipped tri-band GNSS monitoring stations.
— Oliver Montenbruck, Andre Hauschild (DLR/German Space Operations Center),
Peter Steigenberger (Technische Universität München)
Richard B. Langley (University of New Brunswick)
The authors are grateful to Tom Stansell and Col. David Goldstein from the GPS Wing for early discussions and their independent assessment and interpretation of the SVN62 triple-frequency carrier-phase data.
The CONGO network makes use of Javad Triumph Delta-G2T/G3TH and Leica GRX1200+GNSS GNSS receivers for tracking GPS signals on the L1, L2, and L5 frequencies. The stations are equipped with Trimble Zephyr Geodetic II or Leica AX1203+GNSS and AR25R3 antennas.
“The WAAS L5 Signal: An Assessment of Its Behavior and Potential End Use,” by H. Rho and R.B. Langley in GPS World, Vol. 20, No. 5, May 2009, pp. 42–50.
“Using Multi-Frequency for GPS Positioning and Receiver Autonomous Integrity Monitoring” by Y.-H. Tsai, F.-R. Chang, W.-C. Yang, and C.-L. Ma in Proceedings of the 2004 IEEE International Conference on Control Applications, Taipei, Taiwan, September 2–4, 2004, pp. 205–210.
“Triple Frequency Ambiguity Resolution Using GPS/Galileo” by O. Julien, M.E. Cannon, P. Alves, and G. Lachapelle in European Journal of Navigation, Vol. 2, No. 2, May 2004, pp. 51–57.
“Three’s the Charm — Triple Frequency Combinations in Future GNSS” by A. Simsky in Inside GNSS, Vol. 1, No. 5, July/August 2006, pp. 38–41.
“Total Electron Content Monitoring Using Triple Frequency GNSS Data: A Three-Step Approach” by J. Spits and R. Warnant in Journal of Atmospheric and Solar-Terrestrial Physics, Vo. 70, No. 15, December 2008, pp. 1885–1893, doi:10.1016/j.jastp.2008.03.007.