Evaluation of L2C Observations and Limitations

Transcription

1 Evaluation of L2C Observations and Limitations O. al-fanek, S. Skone, G.Lachapelle Department of Geomatics Engineering, Schulich School of Engineering, University of Calgary, Canada; P. Fenton NovAtel Inc., Canada BIOGRAPHY Ossama Al-Fanek is a Ph.D. candidate in Geomatics Engineering at the University of Calgary. He received his B.S. (2004) in Mechanical Engineering at the Jordan University of Science and Technology, Jordan. Susan Skone, Ph.D., is an Associate Professor in Geomatics Engineering, Schulich School of Engineering, at the University of Calgary. She has a background in space physics and conducts research in ionospheric and tropospheric effects on GPS. She has developed software for mitigation of atmospheric effects and is currently chair of the Canadian Navigation Society. Dr. Gérard Lachapelle is a professor and CRC/iCORE Chair in Wireless Location in the above department, where he heads the PLAN group. He has been involved with GPS development and application since For more information visit Pat Fenton is Chief Technology Officer of NovAtel Inc. He is responsible for the new technology developments and activities at NovAtel Inc. He has been involved in the development of 7 generations of GNSS receiver equipment over the last 20 years. ABSTRACT With the recent launch of Block IIR-M satellites and modernization of the GPS, a new L2C signal has been introduced for civilian applications. It is anticipated that ranging measurements on L2C will offer improved observation quality and independent tracking performance, as compared with L2 semicodeless observations. PRN 17 with L2C capabilities was launched in late 2005, with PRN 31 and PRN 12 following in late In order to make use of new L2C observations in conjunction with legacy L2 P(Y) a number of issues must be resolved. Due to differences in code modulation offsets (L2C versus L2 P(Y)) a satellite- and receiver-dependent differential code C2-P2 bias arises that must be quantified and accounted for. Additionally, the tracking noise and multipath characteristics of L2C observations are expected to be similar to those for the C/A code. An investigation of L2C observation quality and robustness is necessary to assess potential capabilities of exploiting these new measurements. NovAtel OEMV3 receivers with L2C tracking capabilities, and equipped with specialized firmware that allows acquisition of both L2C and L2 semicodeless observations for a given satellite using a single receiver, are used. At present observations from as many as three Block IIR-M satellites are available simultaneously in Calgary, allowing inter-satellite comparisons. Zerobaseline tests and inter-receiver comparisons are conducted using live data for the Block IIR-M satellites currently available, in order to assess the quality of L2C versus L1C/A and L2 semicodeless observations. In this paper, investigations of phase and code observation quality for the L2C signals are presented. Zero-baseline tests are conducted with multiple receivers to assess measurement noise, and linear code-phase combinations are used to compute multipath statistics. Tracking performance is assessed for degraded signal strengths to determine performance under challenging scenarios such as weak signal environments. Differential code biases (C2-P2) are determined, and the limitations in combining new L2C observations with legacy L2 P(Y) are assessed for practical implementations. INTRODUCTION Since December 2006 there have been three Block IIR-M satellites in operation transmitting L1C/A, L2P(Y) and L2C signals. The L2C signal has a unique structure, which consists of the L2CM (civilian moderate) code multiplexed with the L2CL (civilian long) code with both codes having a frequency of khz. The resulting L2C signal has a code frequency of MHz, identical to the L1C/A code. It is therefore expected that the magnitude of code multipath errors will be similar for L1C/A and L2C observations. Block IIR-M signal characteristics are listed in Table 1.

2 The L2C signals are transmitted with signal strength 1.5 db lower than L1C/A (Table 2). The L2CM and L2CL codes have longer repetition periods, however, compared to the L1C/A code, which will limit cross-correlation interference between satellites. This is expected to allow mitigation of receiver code noise for L2C observations versus L1C/A. For commercial receivers L2P(Y)sc observations are derived using semicodeless techniques in which noise is amplified. This results in larger code noise and multipath errors, and degraded phase tracking capabilities - particularly at the lower elevation angles. For applications in which reliable dual frequency phase measurements are required, errors in L2P(Y) tracking can limit ambiguity resolution. This is particularly an issue in weak signal strength environments where there can be many phase cycle slips for L2P(Y) observations. The availability of more robust L2 phase measurements from L2C signals will be a benefit for dual frequency applications. A number of biases exist between GPS observables. L1- L2 interfrequency biases are computed by JPL for each GPS satellite and are provided in the broadcast navigation message as Tgd values (Wilson et al., 1999). These values are referred to the P1 and P2 observations, and must be applied by single frequency users. Additional C1- P1 differential code biases have been quantified by the Centre for Orbit Determination Europe (CODE) and updates are provided regularly for all GPS satellites (Hugentobler et al., 2005). These values can be as large as 60 cm. Such biases must be accounted for when using receivers reporting C1 and P2 dual-frequency observations. With the addition of an L2C code, a new differential code bias is introduced which must also be quantified: the C2- P2 bias. This bias must be accounted for when mixing observations from legacy receivers (e.g. C1/P2 or P1/P2 observations) with new C1/C2 combinations from nextgeneration receivers. This is important for ambiguity resolution and wide area GPS network applications, among others. Previous studies of L2C signal quality have been conducted by Simsky et al. (2006) for a Septentrio PolarRx2C receiver and by Sükeová et al. (2007) for Trimble R7 and Trimble NetR5 receivers. In these studies L2C code noise and multipath statistics were found to be similar to L1C/A values. In this paper, L2C code noise and multipath errors are quantified for NovAtel OEMV3 receivers. Limitations in L2C tracking performance are also assessed for degraded signal strengths, and the C2-P2 differential code biases are determined for available Block IIR-M satellites. Results are compared to L1C/A and L2P(Y)sc observations. Table 1: Block IIR-M Signal Characteristics L1C/A L2CM L2CL Carrier Frequency MHz MHz MHz Code Type Gold Maximal Length Maximal Length Code length Repetition 1 ms 20 ms 1500 ms Code frequency MHz khz khz Data rate 50 bps 25 bps N/A Table 2: Received Minimum Signal Strength (IS-GPS-200D, 2004) SV Blocks ChannelSignal P(Y) C/A or L2C II/IIA/IIR L dbw dbw L dbw N/A IIR-M/IIFL dbw dbw L dbw dbw DATA COLLECTION Block IIR-M satellite observations are available at elevation angles as high as 82 degrees for the University of Calgary location. Figure 1 shows the elevation angles for PRNs 12, 17 and 31 observed from University of Calgary over 24 hours. For periods of several hours, two to three Block IIR-M satellites are observed simultaneously at Calgary. For the tests conducted in this paper, two dual frequency NovAtel OEMV3 receivers (herein referred to as Receiver 1 and Receiver 2) were used to collect L2C data at University of Calgary. Equipped with a specialized firmware version 3.100S19, these receivers are capable of acquiring L1 C/A, L2C and L2 semicodeless (herein referred to as L2P(Y)sc) simultaneously for a given satellite. The fourteen-channel OEMV3 receiver has seven channels which can be assigned to combined L1C/A and L2C dual frequency tracking, with the remaining seven channels used for combined L1C/A and L2P(Y)sc tracking. Having this capability, data quality and tracking performance for L2C versus L1C/A code and L2P(Y)sc can be assessed directly for a given satellite.

3 Figure 1. Elevation angles for Block IIR-M satellites observed from Calgary. The L2C data were collected for an extended 16-day period, 25 July to 9 August 2007, using two NovAtel OEMV3 receivers connected to one NovAtel 600 GPS antenna via an antenna splitter (zero-baseline configuration). The antenna was located on the rooftop of the University of Calgary CCIT building. Data were collected for a mask angle of five degrees and data rate of 1 Hz, and carrier smoothing was disabled to collect raw code measurements. In several sections of this paper, one representative day (25 July 2007) is analysed in detail. The entire 16-day data set is used to determine the magnitude and long term stability of the differential code biases (C2-P2) for Block IIR-M satellites. SIGNAL STRENGTH According to IS-GPS-200D (2004), the transmitted signal power of L2C should be 1.5 db lower than L1C/A and 1.5 db higher than L2P(Y). Figures 2 and 3 show carrier-tonoise densities for two Block IIR-M satellites for PRN 12 and 31 data collected using Receivers 1 and 2 respectively. The values of C/N 0 for L1C/A vary from to db-hz for the lowest versus highest elevation angles, respectively. For L2C, the C/N 0 varies from to db-hz, while the C/N 0 varies from to db-hz for L2P(Y)sc. Figure 2. Carrier-to-noise density for PRN 12 data collected with Receiver 1. Figure 3. Carrier-to-noise density for PRN 31 data collected with Receiver 2. Figures 4, 5 and 6 show the C/N 0 values as a function of elevation angle, for observations binned in five-degree increments. The values of C/N 0 for PRN 12 are shown in Figure 4 for Receiver 1 and Figure 5 for Receiver 2. Figure 6 shows the C/N 0 values for PRN 17 data collected using Receiver 1. There is consistency of C/N 0 values between different receivers and different satellites. L2P(Y)sc carrier-to-noise densities are lower than those for L2C. The L2C C/N 0 values are approximately 5 db- Hz lower than L1C/A values at high elevation angles. For elevation angles of 5-10 degrees L1C/A and L2C C/N 0 values converge. This trend has been noted by others (Simsky et al., 2006) and is likely due to different antenna gain patterns for L1 versus L2.

4 Figure 4. Carrier-to-noise density as a function of elevation angle for PRN 12 data collected with Receiver 1. Figure 5. Carrier-to-noise density as a function of elevation angle for PRN 12 data collected with Receiver 2. the resulting measurement will include code noise and multipath, phase noise and multipath (which is at the centimeter level), and twice the ionospheric delay. By single-differencing such code-minus-carrier observations between receivers for a given satellite, the ionospheric delay (which is a propagation error common for the two receivers) is eliminated. The remainder consists of the difference between code noise measurements for the two receivers and a negligible carrier phase noise plus multipath contribution. From these observations, a statistic describing code noise can be derived and analysed as a function of C/N 0 or elevation angle for a given signal/satellite. Figure 7 shows the code noise for L1C/A, L2C and L2P(Y)sc versus elevation angle for representative day 25 July These statistics were derived using singledifferenced code-minus-carrier observations for all three Block IIR-M satellites, with results binned as a function of elevation angle in five-degree increments. For elevation angles in the range five to 15 degrees, the L2P(Y) code noise is m, the L1C/A code noise is cm, and the L2C code noise is cm. At higher elevation angles the code noise values converge to approximately 10 cm for all observations. Figure 8 shows the code noise as a function of C/N 0, where the code noise values have been binned in 4 db-hz increments. The overall lower code noise values for L2C versus L1C/A can be attributed to two properties: 1) minimal L2C cross-correlation interference due to the low number of Block IIR-M satellites available and the longer L2CM and L2CL codes, and 2) a modified L2P(Y) tracking loop (with L1 aiding) has been implemented for L2C tracking in the NovAtel OEMV3 receiver, resulting in a narrower L2C tracking loop bandwidth. Figure 6. Carrier-to-noise density as a function of elevation angle for PRN 17 data collected with Receiver 1. CODE NOISE ANALYSIS To estimate the code noise, a zero-baseline configuration was used. In this case the two receivers are connected to the same antenna, and measurements of a given code observable for a common satellite are expected to differ between receivers as a function of code noise and receiver clock error. By forming the linear code-minus-carrier observation for a given signal (L1C/A, L2C or L2P(Y)sc) Figure 7. Code noise as a function of elevation angle for all three Block IIR-M satellites combined.

5 Figure 8. Code noise as a function of carrier-to-noise density for all three Block IIR-M satellites combined. MULTIPATH ANALYSIS Code multipath can be computed by forming a linear combination of dual frequency pseudorange and carrier phase observations as follows: Figure 9. Antenna set-up on the roof of the CCIT Building, University of Calgary. Figure 10 shows the estimated multipath time series for PRN 12 using Receiver 1 (top) and PRN 17 using Receiver 2 (bottom). It can be observed that multipath errors for L2C and L1C/A code are of similar magnitudes, while errors are much higher for L2P(Y)sc. MP1 P Φ Φ (1) MP2 P Φ Φ where MP1 and MP2 are the multipath observables for L1 and L2, respectively. The observations P i and Φ i are the pseudorange and carrier phase measurements for L1 (i=1) or L2 (i=2) frequencies. These MP observables also contain code noise and carrier phase noise and multipath. Since carrier phase noise and multipath errors are relatively small (centimeter level), this contribution can be neglected. Magnitudes of code noise were estimated in the previous section. Times series of these multipath observables for a given satellite will contain a bias due to L1 and L2 integer ambiguities. To derive absolute estimates of multipath errors, the mean value of a given satellite MP time series is computed and removed. The multipath test environment is shown in Figure 9. The antenna was located in an open-sky environment with medium-level multipath created by reflecting surfaces and building rooftop structures. Results are shown in this section for a representative day 25 July Figure 10. Code multipath estimates for L1C/A, L2C and L2P(Y)sc, for PRN 12 using Receiver 1 (top) and for PRN 17 using Receiver 2 (bottom).

6 The code multipath errors statistics for PRN 12 are plotted in Figure 11. The L2C multipath errors are lower (by a factor of two) compared to L2P(Y)sc multipath errors. L2C and L1C/A multipath errors are similar at low elevation angles. For elevation angles above 35 degrees, the L2C multipath errors are similar to those for L2P(Y)sc and are higher than the L1C/A code multipath errors. Such similarities between the L2C and L2P(Y)sc statistics may arise from the nature of L2C tracking loops where modified L2P(Y) tracking methods have been implemented for L2C. At the highest elevation angles, the three code observables have similar errors (which consist primarily of code noise). Table 1 lists the overall multipath statistics for PRNs 12, 17 and 31 for both receivers. It is observed that the average code multipath errors for L2C and L1C/A are almost equal, and these values are significantly lower than for L2P(Y)sc. Figure 11. Code multipath estimates for L1C/A, L2C and L2P(Y)sc as a function of elevation angle. Table 3. Multipath Statistics for L1C/A, L2C and L2P(Y)sc. PRN 12 PRN 17 PRN 31 RMS (m) RX1 RX2 RX1 RX2 RX1 RX2 C/A L2C L2P(Y)sc PHASE TRACKING ANALYSIS Receiver tracking performance can be degraded during periods of weak signal strength. Such occurrences can be due to ionospheric scintillation effects or presence of RF interference. With the opportunity to derive L2 phase observations directly from L2C signals, it is expected that the availability and quality of dual-frequency carrier phase observations will be improved. The phase observations derived using L2P(Y)sc tracking methods are more susceptible to cycle slips and outages during reacquisition periods. To investigate performance of the NovAtel OEMV3 receivers under degraded signal strengths, a controlled test was conducted using an in-line signal attenuator. The attenuator was placed between antenna and receiver and set to desired levels forcing degraded signal strengths of live L1 and L2 signals. Using one antenna, two receivers were connected in a zero-baseline configuration one with 6 db and the other with 12 db attenuation. Data were collected for 24 hours. The exact effect on the C/N 0 is not known due to the fact that the antennas have high gain internal low noise amplifiers of greater than 20 db designed to overcome the cable loss between the antenna and the receiver. It is not known how much the attenuators overcome the internal gain and residual effect on the C/N 0 but relative differences should at least result. Results are shown in Figure 12 for PRN 31, for the case of 12 db attenuation. Results are similar for other satellites. Values of carrier-to-noise density are shown for L1C/A, L2C and L2P(Y)sc tracking. Values at epochs where a cycle slip occurred are plotted in red. Percentages of corrupt observations are also shown. These percentages are computed as the number of cycle slips or missing observations versus number of expected observations. The percentage of corrupt observations is clearly higher, as expected, for the L2P(Y)sc phase observations with percent corrupt observations prior to full loss of tracking capabilities. Percentages for the L1C/A and L2C phase observations are almost identical. For most L1C/A cycle slips occurrences, there were simultaneous cycle slips for L2C. This is attributed to the L1 aiding implemented in the L2C tracking loops for the NovAtel OEMV3 receiver. In this case it is not possible to assess the independent tracking performance of L2C. In this test L1C/A and L2C tracking ended for C/N 0 thresholds of 27.1 db-hz and 26.9 db-hz, respectively. The satellite elevation angle at this epoch was approximately 20 degrees. The L2P(Y)sc channel lost tracking capabilities 8 seconds before the L1C/A and L2C channels. Average reacquisition periods (after cycle slip occurrences) were 1-3 s for L1C/A and L2C, and 3-8 s for L2P(Y)sc.

7 % Corrupt Phase C/No (db-hz) C/No (db-hz) C/No (db-hz) L1C/A C/No Cycle slip L2C L2P(Y)sc L2P L1CA L2C Hour (UT). Figure 12. C/N 0 values for L1C/A, L2C and L2P(Y)sc, and percentage of corrupt observations (bottom). DIFFERENTIAL CODE BIASES (C2 P2) Using NovAtel OEMV3 receivers equipped with specialized firmware, measurements of C2-minus-P2 differential code biases (DCB) can be computed. In this manner the C2-P2 DCB can be estimated directly. This bias includes satellite-dependent and receiver-dependent contributions. The 16-day data set described in the Data Collection section was used to investigate the magnitude and stability of DCB values. A mask angle of 20 degrees was applied to the data to reduce the influence of noise at low elevation angles. Since there are only three available Block IIR-M satellites currently, results are only derived for PRNs 12, 17 and 31. The combined receiver and satellite DCB was computed as the average of all C2- minus-p2 values for a given satellite/receiver combination over the 16-day period. Figure 13 shows the DCB (C2-P2) for PRNs 12, 17 and 31 using the two OEMV3 receivers. The average DCB values over 16 days are listed in Table 4. It is observed that these values differ between satellites but are consistent between the two receivers within millimetres. Since these values represent combined satellite-dependent and receiver-dependent contributions (from receiver/transmitter hardware), a Spirent hardware simulator was used to estimate the receiver DCB contribution alone. The receiver C2-P2 bias was estimated to be 5 cm with ±5 cm uncertainty introduced by the simulator (Spirent, personal communications). Absolute satellite DCB values are therefore in the range cm lfor the three Block IIR-M satellites. These values are consistent with the C1-P1 satellite DCB values determined by CODE (where maximum values are approximately 60 cm). Table 4. Average Differential Code Biases (C2-P2) for PRNs 12, 17 and 31 and Two NovAtel OEMV3 Receivers PRN 12 PRN 17 PRN 31 RX1 RX2 RX1 RX2 RX1 RX2 DCB (cm) Stabilities of the DCB values were also evaluated for the 16-day period. Figure 14 shows the daily averaged DCB values for each satellite/receiver combination. Variations of the DCB values are consistent between the two receivers. For a given satellite, there are variations of ±2 cm with respect to the 16-day mean value. The DCB values therefore appear to be stable at a level of several centimetres over a few weeks.

8 Figure 13. Differential code biases for PRN 12 (top), PRN 17 (middle) and PRN 31 (bottom) for two OEMV3 receivers (Receiver 1 left and Receiver 2 right).

9 The differential code biases C2-P2 were estimated for Block IIR-M satellites. Values in the range cm were derived, which are similar to the known C1-P1 biases. These values were observed to be stable at the level of ±2 cm over a two-week period. ACKNOWLEDGEMENTS The authors acknowledge financial support from the GEOmatics for Informed DEcisions (GEOIDE) Network Centre of Excellence. REFERENCES Figure 14. Daily average differential code biases for PRNs 12, 17 and 31 for two receivers, 25 July - 9 August CONCLUSIONS L2C signal strength, code noise and multipath errors were evaluated for NovAtel OEMV3 receivers. L2C C/N 0 values were observed to be 5 db-hz lower than L1C/A values for high elevations. At elevation angles of 5-10 degrees L2C C/N 0 values approached L1C/A values. Similarities at the lower elevations are attributed to different L1 versus L2 antenna gains. Code noise values were found to be consistently lower for L2C versus L1C/A, by a factor of two, for the lower elevation angles. For example, the L2C code noise estimate was 25 cm for elevation angles of degrees compared with 50 cm for L1C/A. It was expected that L2C and L1C/A code noise characteristics would be similar. The reduced code noise values for L2C are attributed to two effects: limited cross-correlation interference, and the NovAtel L1-aided tracking loop implementation for L2C in the OEMV3 receiver. Multipath errors were assessed for a medium-level multipath environment. Magnitudes of multipath errors were similar overall for L2C and L1C/A, although larger values were generally observed for L2C at the higher elevation angles. Hugentobler, U., M. Meindl, G. Beutler, H. Bock, R. Dach, A. Jäggi, C. Urschl, L. Mervart, M. Rothacher, S. Schaer, E. Brockmann, D. Ineichen, A. Wiget, U. Wild, G. Weber, H. Habrichand and C. Boucher (2005). CODE IGS Analysis Center Technical Report 2003/2004. available at papers /codar_0304.pdf IS-GPS-200D (2004). NAVSTAR GPS Space Segment/Navigation User Interfaces. GPS Joint Program Office. Simsky, A., J.M. Sleeewaegen, P. Nemry, and Septentrio NV, Belgium. (2006). Early performance results for new Galileo and GPS signals-in-space. Proceedings of ENC 2006, Manchester, UK, 7-10 May. Sükeová, L., M. C. Santos, R. B. Langley, R. F. Leandro, O. Nnani and F. Nievinski (2007). GPS L2C Signal Quality Analysis. Proceedings of the ION 63 rd Annual Meeting, April 23-25, Cambridge, MA. Wilson, B. D., C. H. Yinger, W. A. Feess, and C. Shank (1999). New and improved: The broadcast interfrequency biases. GPS World, September, Vol. 10, No. 9, pp L2C phase tracking performance was also assessed for weak signal scenarios. Live signals were attenuated during data collection by placing an in-line attenuator between antenna and receiver. In the case of 12 db attenuation, L2C and L1C/A tracking capabilities were observed to be almost identical in terms of cycle slip occurrences and simultaneous loss of L2C and L1C/A tracking capabilities. While L2C phase observations were of better quality (less cycle slips and outages) compared to L2P(Y)sc tracking, independent L2C tracking performance cannot be assessed (due to L1-aiding of L2C tracking in the NovAtel receiver).