Use-case analysis of the BOC/CBOC modulations in GIOVE-B E1 Signal

Transcription

1 Use-case analysis of the BOC/CBOC modulations in GIOVE-B E1 Signal Rui Sarnadas, Teresa Ferreira GMV Lisbon, Portugal Sergio Carrasco, Gustavo López-Risueño ESTEC, ESA Noordwijk, The Netherlands Abstract In the past decade, the current and envisaged GNSS systems have seen major developments, both in terms of interoperability and compatibility, as well as accuracy, performance, and availability. In particular, the emergence of Galileo and the evolution of other Global Navigation Satellite Systems (GNSS) have not only brought forth new opportunities and applications, but also new signal structures. The underlying studies on Binary Offset Carrier (BOC) and Multiplexed Binary Offset Carrier (MBOC) modulations, signal properties, tracking methods, and general performances, have theoretically confirmed and supported the benefits of future GNSS signals. As a result of this diversity of signals, the Galileo E1-L1 signal modulation is an implementation of MBOC, the Composite Binary Offset Carrier modulation CBOC(6,1,1/11). Receiver designers will have to decide and trade-off for tracking the Galileo E1-L1 signal either as CBOC(6,1,1/11) or BOC(1,1), taking into account the benefits and costs of each solution. This study presents the results obtained with processing CBOC data in two conditions: first, using a simulator to generate a multipath environment, and second, using real data from GIOVE-B satellite collected at GMV s rooftop. These data sets were processed using the NUSAR platform to produce the observables. Keywords - GIOVE-B, MBOC, CBOC, GNSS I. INTRODUCTION Due to the unavailability of the new GNSS in the past, the majority of the studies of signal impact on receiver design have been mainly theoretical. Nowadays, the available Signals In Space in the Open Service E1-L1 band include the GPS constellation and Galileo s GIOVE-B, which is (already) transmitting the envisaged CBOC modulation [1]. Although the full Galileo constellation is not yet available, it is already possible to assess this new modulation s robustness at signal processing level, and even compare the different BOC(1,1) / CBOC(6,1,1/11) interchanges in the acquisition and tracking loops. By design, the Galileo signals in the E1-L1 band provide the flexibility for receivers to process either BOC(1,1) or CBOC(6,1,1/11) modulations. The objective of this study is to evaluate the use of different receiver design options when processing each of the modulations, i.e. CBOC as CBOC and CBOC as BOC. The trade-offs are evaluated by comparing multipath performance, code delay and carrier phase observables at the output of the tracking loops implemented in the NUSAR platform. First, the expected performances are discussed, and then the NUSAR platform is presented. Latter, the data collection environment is described, and the results are presented. II. EXPECTED PERFORMANCES IN GALILEO E1-L1 BAND A. BOC and MBOC Signals With the upcoming Galileo and GPS modernisation, it is foreseen that the E1-L1 band will support three different modulations: BPSK(1), BOC(1,1) and CBOC(6,1,1/11) for the Open Service (OS). Galileo s MBOC implementation (CBOC) was the last to be proposed, and it is extensively described in [1]. The MBOC Power Spectral Density (PSD) is given by (1), where G BOC(m,n) (f) is the PSD of a sine-phased BOC spreading modulation, denoted BOC(m,n). G 10 1 = (1) BOC (1,1) BOC(6,1) ( f ) G ( f ) + G ( f ) The PSD and auto-correlation functions of the three modulations are presented in Fig. 1 and Fig. 2. Figure 1. Power Spectral Densities for E1-L1 band /10/$ IEEE

2 as defaults in the NUSAR platform using the dot product discriminator, and considering a two-sided bandwidth of 18 MHz. It can be seen that BOC and CBOC modulations perform better for most multipath delays, and that CBOC is expected to outperform BOC in the mid-range delays. The impact of multipath on the carrier phase error is negligible, compared to the code tracking errors, and is not discussed in detail. As shown in [2], the carrier phase error due to specular multipath signals is given by (2) Figure 2. Autocorrelation Function for E1-L1 band The PSD of the BOC(1,1) modulation splits the spectrum in order to reduce interference with GPS, thus maximizing compatibility and interoperability. Furthermore, the BOC and MBOC modulations spread more power into higher frequencies, thus narrowing their auto-correlation functions: in addition to achieving potentially better tracking accuracies, this reduces self-interference and re-enforces compatibility with the legacy GPS C/A code. The BOC modulations, however, also insert secondary lobes (peaks) in their autocorrelation function that create a challenge into the receiver design. Now, signal tracking loops have to make sure that they are tracking the main lobe and not a side peak. As a result of such properties, the (M)BOC modulations are deemed to be more robust in harsh environment conditions such as noise, multipath or interference, when compared to the BPSK legacy. Examples of such considerations and trade-offs are presented in the following sections. B. Multipath Performance A common way to analyze multipath performance is through code multipath error envelopes. In this case, only one replica is assumed, with a Signal to Multipath Ratio (SMR) of 6 db, and delays between 0 and 1 chip. The upper and lower envelopes correspond to the case where the phase of the multipath replica is in-phase and in quadrature with the Line Of Sight (LOS) component respectively. θ e = arcsin 2π ( τ τ ) / T c [ λ] (2) SMR where (τ 1 - τ 0 ) is the relative multipath delay, SMR is the signal to multipath ratio, and Tc is the chip duration. Note that the phase error is given in multiples of λ, the L1 wavelength, and is in the order of millimetres. As an example for a Galileo E1-B CBOC(6,1,1/11) signal with Multipath delay, (τ 1 - τ 0 ) = ns (100 meters); Chip Duration, T C = ns; Signal to Multipath Ratio, SMR = 1.99 (6 db); Wavelength, λ = 0.19 meters; the resulting delay is given by (2), θ e = meters. C. Front-End Bandwidth Although the performance of the pure BOC(1,1) modulation is often tackled, from the receiver perspective, the actual performance to be assessed is the result of processing CBOC(6,1,1/11) signals as 1) matched CBOC(6,1,1/11), or 2) as a BOC(1,1) signal, discarding the higher frequency BOC(6,1) component. For this purpose, one of the main parameters to be considered is the Front-End bandwidth. Fig. 4, taken from [3], illustrates the power distribution of the real scenario. Figure 4. Power Containment for Different Modulations and Bandwidth, B T Figure 3. Code Multipath Error Envelope Fig. 3 shows the code multipath error envelopes of all three modulations for fixed early late spacing of 0.8 for BPSK, 0.5 for BOC(1,1) and 0.18 for CBOC(6,1,1/11) the values chosen It can be seen that the power distribution in the MBOC signals, especially for the higher frequency BOC(6,1) component, is highly dependent on the applicable bandwidth, usually associated with hardware cost, sample frequency (data rate), and added noise. Note that although over 90% of the BOC(1,1) power is allocated to a bandwidth of 8 MHz, the

3 important high-frequency component of the MBOC signal, added by the multiplexed BOC(6,1), has less than 10% of its power contained in those same 8 MHz and, as a consequence, is filtered out of the signal processing loops. III. NUSAR PLATFORM The NUSAR (Navigation User SoftwAre Receiver) platform was developed under a contract with the European Space Agency (ESA). NUSAR processes Open Service GPS and Galileo signals in the E1-L1 band, following an as fast as possible approach, providing code phase, carrier phase, Doppler frequency and C/N0 measurements. The NUSAR platform is comprised of an antenna (L1 band), a wide Front-End (18 MHz) and a laptop, all acquired as COTS components. The laptop then runs the NUSAR software, which implements the signal processing acquisition and tracking of the GPS L1 C/A BPSK(1) signal, and both the Galileo E1-B and E1-C channels as BOC(1,1) and CBOC(6,1,1/11) signals, all transmitted in the E1-L1 band. The NUSAR platform is depicted in Fig. 5. Figure 5. NUSAR Platform The software provides a user-friendly Graphical User Interface (GUI), Fig. 6, with extended configuration capabilities, e.g. modulations, number of channels, correlators spacing, number of non-coherent integrations, loops bandwidth. The outputs, i.e. code delay, carrier phase, Doppler frequency, CN0, bump jumping and loss of lock information, time stamp, satellite ID and prompt (I and Q) correlation results are output to the GUI, and can be stored to a text file. Offline: processing offline IF data, stored in a data repository; Data Acquisition: data is collected from the antenna and Front-End, and stored to file, so it can be processed later in Offline mode. Interaction with the user is further supported by acquisition and tracking loop s high configurability, providing extensibility through APIs, which ease the implementation of external algorithms, and the creation of inspection points. In addition, the software is developed in a flexible and modular way, so as to be able to accommodate new algorithms, and to inspect receiver behaviour at stages that are often veiled to the end user and algorithm developers. The NUSAR is an attractive solution to support studies on the E1-L1 band, since it provides direct access to the raw measurements as output by the tracking loops: the Delay Lock Loop (DLL) and Phase Lock Loop (PLL), for code and carrier tracking, respectively. IV. DATA COLLECTION The simulation plan considered herein aims at assessing and comparing the performance between the processing of CBOC(6,1,1/11) modulation as CBOC(6,1,1/11) and processing it as BOC(1,1). This corresponds to the scenario that Galileo receivers and receiver manufacturers will be facing in the near future: in fact, Galileo satellites will be transmitting the signal as a CBOC(6,1,1/11) modulation, and receiver designers will be able to process the signal by either/both using CBOC or/and BOC modulated local replica signals. The final design of the receiver is then a trade-off between available power (e.g. sampling frequency, data size and rate, hardware bandwidth) and performance (e.g. code delay errors, carrier phase errors, Doppler frequency estimation and C/N0 at the receiver). Two main sets of data 1 were used to support the analysis: Data set 1: simulation data with multipath; Data set 2: real data from GIOVE B; A. Data set 1: Simulated Data In order to look at the effect of multipath in the receiver performance in terms of code tracking error, simulated data was collected using the GSVF Galileo Constellation RF Signal Simulator with a linear multipath model for a static user. The NUSAR platform was then used in Data Acquisition mode to record IF samples for 70 seconds of data, at a sample frequency of. The GSVF simulated four satellites with the multipath characteristics shown in Table I. TABLE I. SIMULATED DATA PRN Ephemeris Multipath Delay [m] Figure 6. NUSAR Graphical Interface and Channel Configuration dialog The NUSAR platform supports three modes of operation: Online: processing real-time data received directly from the Front-End and antenna components; Same as Same as Note that for additional comparison results and statistical relevance, other GIOVE-B real data and GSVF simulator data are used and processed under the same conditions.

4 All multipath rays have a relative power of 6dB with respect to their direct Line Of Sight (LOS) ray. Satellite Vehicles (SV s) with PRN s 11 and 12 contain the same ephemeris as SV s 8 and 9, respectively 11 is a clone of 8, and 12 is a clone of 9. This way, SV s 11 and 12 are used as references, to assess and compare the code delay errors associated to multipath. B. Data set 2: Real Data To complement the results obtained with the simulated data, the NUSAR platform has been used to collect data from GMV s rooftop with a static receiver for about 60 seconds, using the same method as above. The objective is to record data from GIOVE-B using different Front-Ends and sample frequencies, and then process the data using NUSAR in offline mode, with several different configurations. Note that although no information is available regarding the real multipath conditions for this data set, different assessments can be made on processing real live GIOVE-B signals, including parallel processing with currently available GPS signals. The data was collected using different Front-End bandwidths: at 18 MHz, so as to capture most of the CBOC(6,1,1/11) modulation spectrum, and at 8 MHz, capturing only BOC(1,1) modulation spectrum, and filtering out the CBOC(6,1,1/11) high-frequency spread spectrum components, BOC(6,1). V. RESULTS A. Data set 1: Simulated Data with Multipath In order to analyze the behaviour of the simulated data with multipath, in terms of code ranging error, the data sets (for both 40m and 110m of multipath delay) were run in NUSAR with different configurations and sample frequencies, f s, as summarized in Table II. The different configurations aim at assessing the impact of 1) processing the CBOC signals as matched CBOC(6,1,1/11) and as BOC(1,1), and 2) reducing the available signal bandwidth in the processing chain, by reducing the input sample frequency. The NUSAR software was then configured to process the full sampled data, at a sampling frequency of (no decimation) and with decimation factors of 2 for CBOC(6,1,1/11) and 5 for BOC(1,1), keeping the resulting bandwidth compliant with Nyquist criteria. Furthermore, different settings where used to enable acquisition and tracking of both the LOS signal and the corresponding delayed ray under the same conditions. Figure 7. Code Delay Errors for Multipath Data As Fig.7 shows, the results from processing the multipath data paired with the LOS rays confirms that the (absolute) values are within the expected envelopes for each multipath delay. One immediate result is that processing the incoming signal as BOC(1,1) (blue circles and crosses) results in a higher code multipath error that processing it as a matched CBOC(6,1,1/11) (red triangles and diamonds). This is due to the fact that the presence of the high-frequency BOC(6,1) component in the CBOC(6,1,1/11) modulated signal increases multipath robustness, when compared to the BOC(1,1) processing. It can also be seen that the results using any decimation on the sample frequency f s = (represented as circles for BOC and diamonds for CBOC) increase the resulting error in code multipath, in comparison to the full bandwidth signals (represented as crosses for BOC and triangles for CBOC), and a detail of the results is presented in Fig. 8. This is because the reduced sample frequency limits the available signal bandwidth that is seen by the processing loops: the higher frequencies are cut off, resulting in an increased error in the tracking loops. Note that the values extracted are mean values of the code multipath error (i.e. differences between the LOS code delay and the reflected ray code delay), and are of course subject to dynamic behaviours and errors, as a result of the tracking loops. TABLE II. Signals (PRN pairs) NUSAR CONFIGURATIONS FOR MULTIPATH ANALYSIS Transmitted as... Processed as... f s [MHz] 8 11 (40m) CBOC(6,1,1/11) 9 12 (110m) CBOC(6,1,1/11) CBOC(6,1,1/11) BOC(1,1) CBOC(6,1,1/11) BOC(1,1) MHz MHz MHz MHz The results gathered with the configurations from Table II are depicted (as absolute values) in Fig. 7, overlapped with the code multipath error envelopes for clarity. Figure 8. Absolute Code Multipath Delays, with and without decimation, for BOC(1,1) and CBOC(6,1,1/11) processing

5 These multipath envelopes and results were determined using real conditions, in terms of hardware and software. Parameters such as Front-End and input signal bandwidth, sample frequency, acquisition and tracking loop settings (e.g. early-late spacing), were selected to reflect current and envisaged settings used for GNSS signal processing, taking into account the trade-offs in processing GPS BPSK or Galileo BOC/CBOC data: Front-End bandwidth = 18 MHz (double sided); Intermediate Frequency = MHz; Sample Frequency =, with decimation by 1, 2 and 5 (, MHz, and MHz); Code Discriminator = Dot Product; Early-Late Spacing = 0.8 chips for BPSK(1), 0.5 for BOC(1,1), and 0.18 for CBOC(6,1,1/11); Figure 9. Carrier 1-sigma Errors for different Bw and f s Integration Time = 20 ms for GPS L1 C/A, 4 ms for Galileo E1-B, and 20 ms for Galileo E1-C. NUSAR result rate = 1 code period. B. Data set 2: Real GIOVE-B Data In order to further compare the different BOC(1,1) and CBOC(6,1,1/11) interchanges in the acquisition and tracking loops of a GNSS signal processing chain, real GIOVE-B data was acquired and processed at GMV s premises. These data sets allow for studies to be performed under real world conditions, processing the real CBOC(6,1,1/11) signal transmitted with several configurations in the NUSAR platform. To analyse the influence of Front-End bandwidth and sample frequency used, the different configurations depicted in Fig. 9 and Fig. 10 represent the carrier and code 1-sigma errors for multiple input signal bandwidths. The dashed line divides the plots in two sections: the upper left section represents a benefit of tracking CBOC as CBOC, as opposed to tracking with BOC. Of course, the lower right section represents the opposite, when tracking CBOC with CBOC yields lower performance than when tracking with BOC. As shown, the resulting limitations from the input signal bandwidth and sample frequency can be interpreted as: For the same conditions, tracking the CBOC(6,1,1/11) signal with CBOC always yields lower code tracking errors than tracking with BOC(1,1); As the sample frequency decreases, the amount of data to correlate, and so the equivalent signal bandwidth, decrease. This results in increasing the magnitude of the code tracking errors, as shown in Fig. 10; For the carrier tracking errors in Fig. 9, the modulation used has a negligible effect on the result (the points are all very close to the BOC vs CBOC line); The carrier error impact is more relevant but always in the order of millimeter level 1-sigma errors with the compromise between Front-End bandwidth and sample frequency. This can be explained by the noise levels and filtering (possibly aliasing) properties of the input signals in the tracking loops; Figure 10. Code 1-sigma Errors for different Bw and f s Reducing the bandwidth, to a point where the higher frequencies of the PSD distribution of the input modulation are clipped out of the processing loops (i.e. Bw = 8 MHz), results in a decrease in C/N0, thus increasing the code range errors; The different combinations of Front-End bandwidth and sample frequency produce mixed results in terms of C/N0: while the matched CBOC(6,1,1/11) processing yields a C/N0 around 0.5 to 1 db-hz higher than the corresponding BOC(1,1) processing, the reduction of available signal power and sample frequency results in a decrease of C/N0. VI. CONCLUSIONS With the choice of the MBOC modulation for the Galileo and modernized GPS future signals, an increasing number of studies and trade-off analysis have been presented and discussed, in order to analyse the implications of the new modulations and implementations in receiver design and signal processing. One of the major topics for the discussion is the trade-off in processing such MBOC signals as BOC(1,1), discarding the high-frequency spectrum transmitted, and is motivated by the impact of the wider MBOC power spectrum on both hardware and software design, implementation, and manufacturing. One example of the analysis over the consequences of MBOC approximation by BOC can be found in [5].

6 As GIOVE-B satellite is now in orbit, transmitting both data and pilot channels E1-B and E1-C as CBOC(6,1,1/11) modulated signals, an opportunity rises to acquire and process real live data, as if it were a fully operational Galileo satellite. This study focuses on analysing different implications of processing GIOVE-B s CBOC(6,1,1/11) signals as CBOC(6,1,1/11) and BOC(1,1) and shows several trade-offs and results over real data. The analysis was conducted on the NUSAR platform, developed by GMV, as it allows for several different configurations and settings to be used over the same acquired data. Simulated data is also used for multipath performance analysis, as well as other result comparison. It is shown that the CBOC(6,1,1/11) modulations, when tracked with a matched replica, perform better in terms of multipath mitigation and code tracking errors, even with smaller bandwidths lower than the theoretical value of 40 MHz for MBOC [3]. Furthermore, the NUSAR platform was used to configure the signal processing loops, e.g. in terms of sample frequency and early-late spacing, and run the same data in different scenarios. The presented results also focus on assessing the impact of processing CBOC(6,1,1/11) data as BOC(1,1), and show that CBOC outperforms BOC in all cases, at the expense of larger hardware bandwidths, sample frequency, and complexity. In conclusion, as soon as the MBOC modulations are widely available for users and manufacturers, the performance enhancement will be available to those who desire to use it, but still ensuring maximum interoperability and minimum performance degradation over the (less demanding) BOC(1,1) solutions. Of course, the choice will ultimately reflect the application envisaged by the manufacturers and receiver designers. As the discussion in [6] clearly shows, there is no need for a definitive decision to be made, and the mere fact that the CBOC(6,1,1/11) modulation was designed in such a way that maximizes compatibility and range of choices ensures maximum flexibility. ACKNOWLEDGMENT The authors would like to acknowledge José Maria López- Almansa at GMV, for his collaboration and support on data collection and analysis. REFERENCES [1] Hein, G.W. Avila-Rodriguez, J.-A. Wallner, S. Pratt, A.R Owen, J. Issler, J. Betz, J.W. Hegarty, C.J. Lenahan, S. Rushanan, J.J. Kraay, A.L. Stansell, T.A., MBOC: The New Optimized Spreading Modulation Recommended for GALILEO L1 OS and GPS L1C. in Position, Location, And Navigation Symposium, 2006 IEEE/ION, April 25-27, 2006, on pages: [2] Van Nee R, A Spread Spectrum Code and Carrier Synchronization Errors Caused by Multipath and Interference (1993) IEEE Trans. on Aerospace and Electronic Systems, Vol. 29, No. 4, pp , October 1993 [3] Lohan, E. S. & Renfors, M. 2007, Correlation properties of Multiplexed, Binary Offset Carrier (MBOC) modulation, Proceedings of the European, Wireless 2007, EW 2007, Paris, France, April 1-4, p [4] Andrés, D.S., Risueño, G.L., Effect of Receiver Parameters on MBOC Tracking Performances, NAVITEC 2008, S20, Noordwijk, The Netherlands, 2008 [5] Macchi, F., Petovello, M.G., Lachapelle, G., Consequences of MBOC approximation by BOC modulation, ION GNSS 2008, Session F2, Savannah, GA, September 2008 [6] Stansell, T., Gibbons, G., et al., BOC or MBOC? The Common GPS/Galileo Civil Signal Design: A Manufacturers Dialog, Part 1 and 2, InsideGNSS, July/August and September, 2006