
Working Papers • September/October 2012
Exploiting the Galileo E5 Wideband Signalfor Improved SingleFrequency Precise PositioningSinglefrequency positioning can undoubtedly be improved with the deployment of new GNSS systems and the accompanying availability of new signals. Among various innovations, the Galileo E5 broadband signal deserves special attention. Its unique features, including the AltBOC modulation scheme, will drastically boost code range precision, both in terms of reduced code noise as well as with respect to multipath. Precise singlefrequency positioning will be feasible at centimeter level, benefiting both scientific and nonscientific applications. This article demonstrates the expected performance of E5 for selected land applications and precise orbit determination of low Earth orbiting satellites.
Share via: Slashdot Technorati Twitter Facebook Working Papers explore the technical and scientific themes that underpin GNSS programs and applications. This regular column is coordinated by Prof. Dr.Ing. Günter Hein, head of Europe's Galileo Operations and Evolution. Can precise positioning only be performed by highend dualfrequency GNSS receivers? Many in the GNSS community will answer in the affirmative because of a prevailing impression that accurate results essentially require dualfrequency receivers — preferably using carrier phase measurements for precise positioning applications. This way of thinking is primarily motivated by the fact that the (first order) ionospheric delay can be eliminated by the use of at least two frequencies and carrier phase measurements are less affected by multipath effects than range measurements. In the past many singlefrequency approaches have been proposed to obtain precise results from a lowcost singlefrequency GNSS receiver. However, the main obstacle to achieve precise singlefrequency positioning with the current core GNSS signals (GPS and GLONASS) remains the high level of code range noise, substantially poisoned by multipath errors, which can be a few decimeters up to some meters. Further, the high level of multipath effects on the currently available frequencies (e.g. GPS and GLONASS L1 and L2 bands) makes precise singlefrequency positioning impossible. Performing singlefrequency positioning with Galileo E5 could offer a possibility to realize precise positioning solutions with moderate budget. Note that positioning is not limited to land applications only. A number of satellite missions require fairly accurate orbits and GPS singlefrequency receivers have become an important means of orbit determination, whereas only a few dualfrequency GPS receivers are in orbit (mainly limited to certain types of could offer an interesting future field for Galileo E5 singlefrequency receivers.
Galileo E5: Broadband AltBOC Signal The time series illustrated in Figure 1 (at the top of this article) was computed by a subtractive combination of code range and carrier phase measurements to obtain the coderange noise (including ionospheric variations) and portrays the values of Galileo E1 and E5 collected from the GIOVEB experimental Galileo satellite. As the figure data show, the level of coderange noise for Galileo E5 is extraordinarily low — much lower than that for Galileo E1. The multipath errors for Galileo E5 are reduced by a factor of three to five (see Table 1). These characteristics open the possibility of performing coderange measurements at the centimeter level and enable a better mitigation of multipath effects. The drastically increased range precision due to the very low E5 range noise allows for obtaining more accurate combined codeandcarrier observables. The minimum we are expecting from a modern highquality receiver is the use of “narrow correlator” or equivalent technologies to mitigate multipath effects. The corresponding models to the use of “narrow correlator” technology are the basis to generate simulated (“synthetic”) data for the positioning experiments using a GPS/Galileo software receiver.
Galileo E5 AltBOC Receiver To achieve the requested flexibility and configurability, a software receiver appears to be the right technical solution. To keep costs low, the software part has to be linked to a suited RFfront end hardware receiver part, to support singlefrequency Galileo E5 operations. Figure 2 shows a block diagram of the s/w receiver implementation scheme.
CodePlusCarrier (CPC) Equation 1 (see inset photo, above right) where ρ is the pseudorange measurement, ϕ is the carrier phase measurement in metric units, r is the geometric distance between satellite and receiver antenna, λ is the wavelength of the Galileo E5 carrier, N is the ambiguity parameter, δT is the tropospheric delay, δM_{CPC} is the influence of multipath on the measurement, and ε_{CPC} are the additional code and carrier phase noises. This method was basically recognized but rarely applied to terrestrial position determination. Indeed, the method was proposed in 1996 under the name of GRAPHIC (Group and Phase Ionosphere Correction) for GPS orbit determination because the majority of spacecapable receivers are of singlefrequency type. In contrast to traditional pseudorange positioning, here we must deal with the unknown position and also the unknown ambiguity parameter (as in carrier phase positioning) in the newly built observable. This requires a longer observation window in order to allow sufficient convergence of the ambiguity parameters. In practice, we make use of double differences in order to minimize the satellite and receiver clock errors and to attenuate other sources of errors. For this purpose, access to a global or continental network (e.g., International GNSS Service or EUREF) is sufficient as experience has shown that the use of regional networks (shorter baselines) will not further improve the positioning accuracy. Tropospheric delays can still compromise the positioning accuracy. For this reason, either external sources providing precise corrections (e.g., numerical weather models) should be incorporated or make use of additional tropospheric delay parameters in the estimation process. Multipath errors are sitespecific and particularly strong on the code ranges. Here, the use of E5 AltBOC provides a key advantage, as this broadband signal shows an ultralow multipath behavior compared to all other GNSS signals.
Performance Assessment of the SingleFrequency Positioning with E5 The results are analyzed by a set of different statistical methods to determine the achievable accuracies of the singlefrequency positioning approach using Galileo E5. In generating the results we assumed a full Walker 27/3/1 Galileo constellation, which will be available by 2020 according to the current state of development.
CPC Results for Galileo E5 versus GPS L1 The Galileo E5 coordinate components yielded results in the range of a few centimeters. The horizontal components were around 2 centimeters and better than the vertical components, which ranged from 3 to 5 centimeters. One of the reasons for this is the higher dilution of precision of the vertical component (VDOP). The daily data batches for Galileo E5 singlefrequency yielded a total error of around 3–5 centimeters 3D RMS depending on the measurement environment whilst for GPS L1 they yielded up to 20 centimeters 3D RMS. The great advantage of using Galileo E5 signals to perform CPC singlefrequency positioning can readily be seen at a glance in Figure 5. This figure represents the time series of Galileo E5 and GPS L1 and the amplitude of the single shots of Galileo E5 is at the low centimeter level. Another experiment involved showing the ability of Galileo E5 (AltBOC) CPC results to detect position change in a moving structure over an extended period of time. The average motion rate for a structure such as a rock glacier is expected to range around 70 centimeters per year, see Figure 6. The reference period is set to day 206 in the year 2011, and data were continuously generated. After 64 days the displacement rate between days 206 and 270 was calculated. The data were computed for GPS L1 and Galileo E5 using the singlefrequency CPC principle with the results given as point scatter plots in Figure 7. These plots draw a clear picture of the relative detection ability of the two datasets. By analyzing the statistical distribution of the measurements, we see that the GPS L1 measurements have an error ellipse of around 30 centimeters, which is not sufficient to detect any significant change in the position of the station if we consider an expected motion rate of about 11 centimeters for that observation period as given by Figure 6. The single shots of the coordinates for the two periods overlap (blue and green dots). For Galileo E5 measurements, a clear difference between the two periods of observation is perceptible, which indicates that the station has undergone a displacement. The positioning accuracy is still too low to depict a clear behavior of the movement because of a fivecentimeter error ellipse of the Galileo E5 results. Nonetheless, the Galileo E5 singlefrequency results are accurate enough to detect a position change after a relatively short period of time. After 126 days of data logging, we repeated the experiment for the same station. The point scatter plots on Figure 8 depict the results. Now the displacement of the station is much more noticeable with Galileo E5 measurements. Using Galileo E5 CPC singlefrequency positioning, we are able to detect deformation or displacement in the range of a few centimeters to decimeters. CPC positioning using a Galileo E5 singlefrequency receiver has a certain advantage here compared to carrierphase processing. Because of the moderate price of a singlefrequency system, we can use more sensors to determine an exact profile of the deformation. Due to the long convergence time (20–30 minutes), singlefrequency positioning using the CPC principle could be suited for certain types of landslide or glacier monitoring applications, because changes in such structures can only be detected after a sustained period of observation.
CPC Results for Galileo E5 versus GPS L5 The L5 signal is improved compared to GPS L1 and has the same centerfrequency (1176.45 MHz with a 24 MHz bandwidth) as the E5a subcarrier of the Galileo broadband E5 signal. Hence, L5 seems to have a similar characteristic to at least one part of the E5 signal. Using the same procedures as before, we generated and processed GPS L5 and Galileo E5 synthetic data. Comparing the results confirmed that GPS L5 is more robust than L1 regarding the effect of code noise. The RMS values of the CPC results using GPS L1, L5, and Galileo data are shown in Figure 9. The graph presents results for 1, 3, 6, and 24hour data batches. As with the earlier experiments, the GPS L1 positioning results are outside of the range that we could call precise. Even with a daily data batch, one can only reach a decimeter to subdecimeter level of positioning accuracy. Using GPS L5 produces a noticeable improvement, but the results are still at least two times worse than those for Galileo E5. The results show the need for a certain convergence time to obtain precise coordinates due to the presence of ambiguity terms in the CPC observables. For example, one might have to gather measurements for at least one hour to achieve subdecimeter precision for Galileo E5. This relatively long convergence time must be properly addressed by fixing the ambiguity terms to their integer values. These tests assessed the performances of the Galileo E5 CPC algorithm and showed a 3D positioning accuracy of around 5 centimeters in critical environments and 2–3 centimeters under normal conditions by processing daily data batches. In comparison, these results are four to five times better than GPS L1 and two to three times better than GPS L5. Regarding the first obtained results, the singlefrequency positioning concept using the potential of Galileo E5 clearly shows some innovative aspects with a definite potential to be developed. Due to its very low code range noise and the even lower multipath influence on the positioning solution (compared to other GNSS signals like GPS L1 or L5), the Galileo E5 CPC singlefrequency results are able to fulfill the requirement for many GNSS precise positioning applications.
E5 CPC versus Carrier Phase Processing A user equipped with a multifrequency GNSS receiver can estimate the ionospheric group delay and phase advance from the measurements, and essentially eliminate the ionosphere as a source of measurement error. A relative ionospherefree linear combination of carrier phase measurements can properly eliminate the ionospheric error and delivers results at the millimeter level. Using such a combination for different GPS signals (L1+L2, L1+L5) the results are compared with the CPC Galileo E5 results. As we see in Figure 10, the results obtained with the carrier phase combination are much more accurate than the ones with Galileo E5 CPC. The difference between the two results is scaling around a factor of 10. Even though Galileo E5 CPC results cannot be compared to multiple frequency results (Figure 10) in terms of accuracy, the CPC approach can still meet the requirements of various precise positioning applications requiring decimeter and centimeterlevel accuracies. Moreover, not every precise positioning application requires millimeterlevel accuracy with the associated high cost expenditure for a multifrequency receiver. Hence, the CPC approach can fill a niche between highly precise positioning using multifrequency carrier phase processing and conventional singlefrequency positioning.
CPC Ambiguity Resolution Until now the results only used estimated float solutions of the ambiguity terms. This estimation does not solve the problem, because the results are not sufficient to obtain precise positions quickly. Furthermore, use of float solutions means that a certain observation time is needed to converge to reasonable values. This assumes results at the decimeter level after around 30 minutes and centimeter level after 3 hours. So, resolving the unknown ambiguities of the double differenced CPC observable is the key to rapid and very precise GNSS positioning. Therefore it is imperative to fix the ambiguity term to its integer value. Many algorithms can resolve the ambiguity term. The “Leastsquares AMBiguity Decorrelation Adjustment” (LAMBDA) is a very efficient method used to fix the ambiguity and is less intensive in the computation than other methods. Due to the high level of noise in the GPS L1 code measurements, it is difficult to resolve the ambiguity term in the newly built CPC observable for this signal. Using the LAMBDA algorithms, Galileo E5 singlefrequency positioning presents a completely different picture (see Figure 11). After one minute of processing, the coordinates’ standard deviations yield several meters. The deviation decreases very quickly. After 15 minutes of observation, the ambiguity terms are fixed to their integer values, enabling one to get results with an RMS of a few centimeters. This is a very efficient method to achieve a faster convergence of the coordinates to a precise solution.
Precise LEO Orbit Determination with Galileo E5 Table 2 summarizes selected LEO mission types, their typical orbital heights, POD techniques currently used, and the mission POD’s typical accuracy requirements. Many of these satellite missions are designed for scientific purposes, while others are operated on a commercial basis, e.g., optical (highresolution) missions. The decision to use a single or dualfrequency GNSS receiver on these LEOs is mainly driven by the orbit accuracy requirements, which are dependent on the mission goals, but cost is also a factor: can the mission’s budget support the cost of a dualfrequency receiver? And the price of the receiver is not the only cost factor; depending on the type of GNSS receiver chosen, many other factors have to be adapted accordingly (e.g., power, downlink bandwidth). The CPC (codepluscarrier) method as described earlier is suitable for singlefrequency LEO POD as well. Especially for LEOs with an excellent pseudorange quality (e.g., GRACE B; see the article by R. Kroes cited in the Additional Resources section), one may reach decimeter accuracy for the orbits. The Galileo E5 signal therefore offers an interesting alternative to GPS receivers. The accuracy can be improved over GPS L1only receivers, and a Galileo E5 receiver could save costs compared to a dualfrequency GPS receiver while still meeting highaccuracy POD requirements. Looking over the list in Table 2, the first two mission types — gravity and altimetry — are not suitable for a singlefrequency E5 receiver because the accuracy requirements are too high. Neither are radio occultation missions suitable, because a dualfrequency GNSS receiver is essentially needed for the radio occultation technique. Satellite operators of SAR/InSAR and optical missions with panchromatic sensors would, however, be potential users of such a singlefrequency E5 receiver. Typically, such missions are operated by space agencies (e.g., ESA, NASA, CNES, DLR) and commercial satellite operators (e.g., SPOT S.A., GeoEye).
POD Test Case A version of the Bernese GPS Software tailored for LEO POD has been used to generate precise orbits of ENVISAT based on the simulated GPS+Galileo data. The orbits are derived from undifferenced observations, meaning that no additional data from ground tracking stations were needed for 71the processing. Two different orbit types were generated for this test — a reduceddynamic and a kinematic orbit:
Both types of orbits were generated because kinematic positioning is very sensitive to dataquality issues so that the improvement in the pseudorange quality for the Galileo E5 frequency can be made visible very clearly. In the case of the reduceddynamic orbit determination technique, the resulting orbit benefits from the physical force model and the orbits are of the highest quality.
Results Figure 17 shows the orbit differences in the outofplane direction between the kinematic orbit derived with L2 and the reference orbit as well as between the kinematic orbit derived with E5 and the reference orbit. Besides the systematic differences that are of no importance here, one may notice the very small noise structure of the differences for the E5 solution compared to the differences for the L2 solution. The reason is the low noise characteristic of E5 and confirms the excellent performance of this observable for a singlefrequency approach. These initial orbit results derived from synthetic data clearly show the potential that the Galileo E5 frequency offers for LEO POD missions. Accuracies of 5 centimeters for reduceddynamic orbits and 10 centimeters for kinematic orbits are a great success for singlefrequency orbit solutions, which have never been achieved with GPS L1. Many future LEO missions may profit in terms of money, downlink bandwidth, and power from such a singlefrequency Galileo receiver because the need for a more expensive dualfrequency receiver may be avoided completely.
Conclusion The tests that we performed in this research showed that 3D accuracy of a few centimeters (23 centimeters) can be achieved with Galileo E5 CPC singlefrequency positioning. Compared to the results with GPS L1 or L5 (GPS L1, 10–20 centimeters; L5, 3–6 centimeters), one can see the potential of using Galileo E5. A drawback to the method is the long convergence time (20–30 minutes) to get precise coordinates and achieve an accurate solution. This should be properly addressed in future studies by employing a filtering technique. Further tests showed that carrier phase processing is still more accurate (by an order of 10) than the CPC singlefrequency approach using Galileo E5. However, this kind of processing requires multifrequency receivers that are more expensive than singlefrequency receivers. Nonetheless, not all precise GNSS applications require precision at millimeter level or in realtime. So, the singlefrequency positioning approach with Galileo E5 can fill a niche between carrier phase processing (millimeter level) and the usual singlefrequency positioning (decimeter level). In addition, due to the convergence time required to achieve higher accuracy, the E5 approach seems to be suited for monitoring activities in which precise coordinates are only needed to detect changes after a prolonged period of observation. The CPC method described in this article is also suitable for singlefrequency LEO POD. Especially for LEOs with an excellent code pseudo range quality such as GRACE B, one may reach decimeter accuracy for the orbits. This special characteristic of the Galileo E5 signal should make a singlefrequency receiver based on it an interesting alternative to the current dualfrequency GPS receivers.
Acknowledgements
Additional Resources ManufacturersNarrow Correlator is a technology designed by NovAtel, Inc., of Calgary, Alberta, Canada. The simulated GPS and Galileo data used in the experiments and test case described in this article were generated by the application “Nereus” and the position coordinates were processed by the eXpert application, both are contained in the software package SGSS (Scientific GNSS Support Service) which has been implemented during the SX5 project. The softwaredesigned receiver implementation shown in Figure 2 is based on a patentpending design of IFEN GmbH, Poing, Germany. The satellite imagery of the rock glacier is provided by Google Maps, Google, Mountain View, California.Copyright © 2018 Gibbons Media & Research LLC, all rights reserved. 
