N-FUELS and SOPRANO: Educational Tools for Simulation, Analysis and Processing of Satellite Navigation Signals

Transcription

1 N-FUELS and SOPRANO: Educational Tools for Simulation, Analysis and Processing of Satellite Navigation Signals Emanuela Falletti, Davide Margaria, Mario Nicola, Gabriella Povero and Micaela Troglia Gamba Navigation Signal Analysis and Simulation (NavSAS) Group Istituto Superiore Mario Boella (ISMB), Torino, Italy Abstract In recent years, research activities in the field of Satellite Navigation have boosted worldwide. At the same time, it has become evident that few educational opportunities in the field were available for students and there was a need to develop dedicated tools for hands-on sessions. To partially answer this need, the NavSAS Group has developed N-FUELS and SOPRANO. N-FUELS, a MATLAB R -based signal simulator, allows students to understand the physical layer of the Global Navigation Satellite Systems (GNSS) signals and to learn how to manipulate them via software. SOPRANO, a collection of ANSI C language routines, implements the whole chain of GNSS signal elaboration in post-processing and enables testing and validation of new GNSS signal processing algorithms and architectures. Both tools are used in post-graduate courses at Politecnico di Torino with a high degree of internationalization, which opens interesting points of discussion concerning the introduction of novel educational tools able to meet the demand and the learning styles of students with different educational backgrounds and cultures. I. INTRODUCTION In the last decade, research activities in the field of Satellite Navigation have boosted worldwide. New Global Navigation Satellite Systems are under design and deployment, such as European Galileo or Chinese BeiDou, while already existing systems, such as US GPS or Russian GLONASS, are undergoing modernization processes. This has led not only to an enlargement of the GNSS scientific community but also to the birth of companies interested in GNSS-based products and services. As a consequence, a number of new jobs are expected to be available in coming years worldwide in the field of GNSS and its applications. On the other hand, few educational opportunities focused on this topic are presently available for students [1]. This was recognized by the European Commission, that started actions to foster and support initiatives in this field [2] so as to educate the future GNSS engineers and researchers. At the same time, there was a need to develop new dedicated tools which professors could use in courses and hands-on sessions. This fact was particularly evident to our research group, who has been involved in educational activities on GNSS for almost a decade. One example is the Specializing Master on Navigation and Related Applications [3], a post-graduate programme offered by Politecnico di Torino since 2004, where we teach some modules on Satellite Navigation Systems and Receivers. The Specializing Master is offered in English and this opens it to the participation of international students: on average over 50% of students are from outside Europe. This makes the Specializing Master an interesting melting pot where different educational backgrounds and cultures merge. The result is an international community of people trained in GNSS who are expected to be the tomorrow technical and scientific work force in this field. II. TWO EDUCATIONAL SOFTWARE TOOLS FOR GNSS-ORIENTED STUDENTS Learning from experience of many years of courses on signal and system simulation, as well as from the practice of designing and testing algorithms for the core signal processing in GNSS receivers, we felt the need of developing simple and flexible software tools to, on the one hand, simulate a wide variety of GNSS signals at their physical layer; on the other hand, to build a completely open software architecture of a GNSS receiver, modular and simple enough to be used as a training ground for the implementation of functional blocks in real-time architectures, such as those presented in [4] [7]. The two software tools we have developed are a MATLAB R -based signal generator (N-FUELS, FUll Educational Library of Signals for Navigation), and an ANSI C complete GNSS receiver (SOPRANO, fully SOftware fully Programmable GNSS Receiver for Algorithm testing and validation). Their basic characteristics are reviewed in the following subsections. What is worth highlighting right now, is that these two tools have allowed us to provide the students with highly reliable software tools, which overcome the initial difficulty of simulating quite complex signal architectures [8] and propagation effects, as well as realistically reflect the signal flow in a receiver and the logical sequence of the receiver s states. A. N-FUELS: FUll Educational Library of Signals for Navigation N-FUELS is a MATLAB R -based signal simulator designed to offer a complete and intuitive software tool which simulates the physical layer of GNSS signals, as they appear at the Analogue-to-Digital Converter (ADC) output of a receiver s front-end, either at Intermediate Frequency (IF) or at baseband [9]. From an already trained user s perspective, it allows neglecting all the issues related to the correct implementation of the signal structure (including spreading code /13/$ IEEE

2 metrics [10] [13]: for example, power spectral density estimation, group delay estimation associated to a certain frontend filter model, auto correlation and discriminator curves, multipath error envelope and running average curves, spectral separation coefficients, etc... Again, this set of simple tools for the signal analysis is a fast bridge for students towards the undestanding of the basic properties of the GNSS signals, allowing them to circumvent the obstacle of coding, at least at a very early stage. Fig. 1. A screenshot of the N-FUELS graphical user interface. generation) and of a number of common propagation effects, such as propagation delays, Doppler shift effects, thermal noise, multipath and various models of interference. From a students perspective, it is a tool that directly and simply puts in their hands samples of the signals introduced of during the lectures; observing such simulated signals, students can get an immediate understanding and a hands-on experience of several signal properties and propagation effects. The goal of being an intuitive and wide-use tool is pursued without loosing flexibility and generality. Indeed, the GNSS signals supported by N-FUELS are all civil GPS services (including signals under modernization), all open Galileo signals, and the EGNOS signal. The student can control several parameters of the simulation through a very user-friendly Graphical User Interface (GUI), where he/she can act on several settings. Namely, three families of settings are provided: parameters related to the Signal-In-Space (SIS), including the length of the simulation, the number of simulated signals, the signal modulation, the PRN codes number, the code delays, the Doppler model, and the received signal power; propagation channel parameters, including Carrier to Noise density ratio (C/N 0 ), thermal noise spectral density, multipath (a static N-ray channel model is embedded, with selectable N), disturbance models (continuous wave, filtered noise-like interference, pulsed interference, inter/intra-system interference, emulation of a spoofing attack); receiver front-end parameters, including sampling frequency, intermediate frequency, front-end filter model, possibility to compensate the filter group delay and its transient time, number of quantization bits, format of the output binary file (double, float32 or int8 sample representation). A screenshot of the GUI is shown in Fig. 1. In addition, N-FUELS includes a set of simple signal analysis tools for quickly testing and monitoring several signal B. SOPRANO: fully SOftware fully Programmable GNSS Receiver for Algorithm testing and ValidatiOn As the acronym suggests, SOPRANO is an ANSI C-based fully software receiver, which implements the post-processing GNSS signal elaboration. It processes, i.e., acquires, tracks and uses for PVT (Position-Velocity-and-Time) estimation, both the GPS-SPS (Standard Positioning Service) and Galileo OS (Open Service) signals. It accepts as input binary files, containing the ADC output samples, either generated by N- FUELS or recorded from a GNSS front-end, such as for example [14], with up to 8 quantization bits. Thus it can be employed to process either simulated signals in a completely ideal environment or real signals. The current version of the receiver supports only real IF samples, but future releases will also include the baseband elaboration. SOPRANO has been expressly thought as a lab instrument for signal processing, enabling developers (either trained researchers or graduate students) to describe, test and validate GNSS signal processing algorithms and architectures. It takes also advantage of the processing speed of the C language, differently from other famous software receivers [15] [17] which employ MATLAB R as modeling language. Similarly to the GNSS Software Defined Radio (SDR) receiver in [18], it aims at being an open-source and public reference receiver, both with educational and training purposes. SOPRANO implements all stages of the whole processing chain, by means of state-of-the-art algorithms only, making it a non-patented and completely free solution. Thus, as N-FUELS, it meets the needs of both trained and student users. While a student can get a deep understanding of the GNSS processing flow, looking how the learned theory becomes practice, the trained user can easily found and modify the desired functions, thus testing his own solutions. The handiness in use of SOPRANO is one of its main features and its crucial point. First of all, its ANSI C structure does not need special libraries, which aids the fully portability among different compilers and different operative systems; furthermore, the user can openly access the entire source code, making the desired changes wherever he/she wants. The code structure is easy to read: it is organized in a sequence of folders, each one including a list of source and header files, belonging to the same processing stage. For example, the directory entitled acquisition collects all functions related to the signal acquisition, while that called tracking contains those devoted to implement the tracking loops, and so on. All the signals/data/measurements needed to perform signal elaboration are included in structures which are accessed through the use of pointers, avoiding other less intuitive ways of data access, such as First-In-First-Out (FIFO) data structures, sockets, etc. The name itself of all data

3 structures and functions always refers to their functionality. SOPRANO s flexibility is allowed by the intrinsic nature of its software implementation. In fact, besides working with different IF front-ends, thanks to a simple text configuration file, also several processing setups can be changed via other simple configuration files, for example the number of coherent/non-coherent sums in acquisition, Doppler and code delay resolutions, gain and bandwidth of the frequency and code discriminators, early-to-late spacing of the correlators and so on. This allows the user to quickly create his tailor-made receiver. III. EXAMPLES OF PROPOSED EXERCISES In the frame of the Specializing Master on Navigation [3] introduced in Section I, the GNSS Introduction course introduces the mathematical description of the physical structure of the GNSS signal, including all the current signals and those under modernization. It also discusses the issues of interoperability and coexistence among different GNSSs. Another advanced course on GPS and Galileo Receivers gives students a knowledge about the fundamental architecture of a GNSS receiver. The course combines theoretical principles about the employed algorithms and hints about their implementation in real GNSS receivers. In order to reach this target, the course combines both theoretical lectures and practical assignments. In this section, a couple of assignments, proposed to the students during the mentioned courses, are shown as examples of the employment of N-FUELS and SOPRANO in education. The first exercise is focused on the use of the Spectral Separation Coefficient as a tool to theoretically assess inter-system interference. The second exercise is about the algorithms to acquire the GNSS signal in a typical receiver. A. Exercise 1: Evaluation of Spectral Separation Coefficients The spectral separation coefficient (SSC) [12] is a theoretical method proposed to assess the level of inter-system interference among GNSS signals (and also between a GNSS signal and an interferer). The SSC is a measure of the spectral coupling between two signals, defined as a sort of inner product of the normalized power spectral densities of the two signals, evaluated over a certain front-end bandwidth. The average SSC is defined as [12] κ is = Bfe /2 B fe /2 G i (f)g s (f) df (1) where G i (f) is the unit-power power spectral density of the signal i (taken as the interference) and G s (f) is the unitpower power spectral density of the reference signal s. Once evaluated the SSC κ is, the effective carrier to noise density ratio can be computed, where the interference component is assimilated to an additional noise effect. Having commented such a metric during classes, it is effective for the students to see examples obtained in simulation. With this aim, a specific assignment based on the use of N- FUELS is proposed in the course. The text of the assignment is reported in Table I. TABLE I. TEXT OF THE EXERCISE ABOUT SPECTRAL SEPARATION COEFFICIENTS Exercise 1 Objectives: Understanding the spectral separation coefficient. Description: The spectral separation coefficient (SSC) is a theoretical metric that measures a normalized level of interference between two GNSS signals (or a GNSS signal and a certain interferer). It is based on the measure of the overlapping portions of the power spectral densities of the signals of interest. Here we want to evaluate the SSC between a CBOC modulation and the other modulation formats of the GNSS family (BPSK(1), BPSK(5), BPSK(10), BOC(10,5), BOC c(15,2.5)). Use N-FUELS to generate 100 ms of each of the above signal formats at the intermediate frequency f IF =0MHz, sampled at 50 MHz, received at the power of 0 dbw without noise. Set the Doppler frequency to 0 Hz. The receiving filter is ideally flat on the digitization bandwidth and there is no quantization. Generate the estimated Power Spectral Density (PSD) of each signal (hint: save your simulations in different output folders, one per configuration, so that you can reuse your data sets). Superimpose each estimated PSD to that of the CBOC signal and observe the portion of overlapping spectra. Then set the file of the parameters for the SSC analysis (ParamSSCoefficients.txt) to produce the SCC for each modulation with respect to the CBOC. Questions: 1) Use N-FUELS to obtain a table of the SSC for each modulation format with respect to the CBOC one; sort in descending order of SSC value. 2) Observe that the amount of overlapping portions of the spectra, as seen from the PSD figures by simple inspection, is compliant with the numerical results put in the table. TABLE II. SPECTRAL SEPARATION COEFFICIENTS BETWEEN THE CBOC MODULATION DEFINED FOR GALILEO E1 AND OTHER POSSIBLE COEXISTENT MODULATION FORMATS i BPSK(1) BPSK(5) BOC c(15,2.5) BPSK(10) BOC(10,5) κ i,cboc (db/hz) ) Comments about the assignment: The students are expected to comment the result of this assignment during their oral exam. The simulation through a numerical computation of the formula (1) enables the students to fill a table with SSC values for different modulations, with respect to the CBOC one. The comparison between these results and the figures, where the corresponding PSDs are actually superimposed, should ease the understanding of the SSC concept. For example, the table of the SSC values is reported in Table II, while the estimated power spectral densities for the two extreme cases of maximum and minimum SSC s are shown in Fig. 2. B. Exercise 2: SW Implementation and Test of an Acquisition Algorithm for a Galileo E1b/c Signal In a GNSS receiver, the signal transmitted by one satellite is received using local replicas of the carrier and the spreading code employed by the transmitting satellite. The main task of the receiver is to obtain the alignment between these local replicas and the received signal: the receiver must produce a

4 Power Spectral Density [db/hz] BOC(10,5) BPSK(1) Signal Spectrum CBOC CBOC BPSK(1) BOC(10,5) Frequency [MHz] Fig. 2. Power spectral densities of the modulations analyzed in Exercise 1. The minimum separation of the spectra, as measured by the SCC, is obtained between CBOC and BPSK(1), as their main lobes partially overlap. On the contrary, the maximum spectral separation is obtained between CBOC and BOC(10,5), whose main lobes are well separated in frequency. local replica of the spreading code with the same code rate and delay as the incoming signal, and a local replica of the carrier with the same frequency and phase of the incoming carrier. Such synchronization is the key point that allows to correctly estimate the distance between the satellite and the receiver and, consequently, to evaluate the position. The acquisition stage gives the receiver a first rough synchronization with the incoming signal, whereas the following stage, named tracking, refines and updates the alignment [10] [19]. 1) The assignment: After a lecture about the theory of the GNSS signal acquisition, students are required to implement main acquisition algorithms that can be found in literature [15] both in MATLAB R and in C language. The text of the assignment is reported in the Tables III and IV. 2) Comments about the assignment: In Exercise 2a (Table III), students implement acquisition algorithms using MATLAB R. This approach allows the students to focus on the employed algorithms and almost ignore implementation details. In this way, they can acquire a deeper knowledge of the subject if compared to what can be obtained with a pure theoretical lecture. Under these conditions, the use of N-FUELS is a key point. Using signals generated by N- FUELS, students have a perfect control over the unknowns of the acquisition (carrier frequency, code rate and delay): this facilitates the debug of the implementation. Moreover, the use of N-FUELS enables an easy generation of signals at different carrier to noise ratios, making possible the part of the exercise when students verify the effects of noise on acquisition results. Fig. 3 shows a graphic obtained solving Exercise 2a. The MATLAB R implementation of the acquisition algorithms hides some unnecessary details, but it is very far from an implementation able to elaborate samples of a signal from an antenna or from a hardware signal generator. In order to teach students the architecture of a software receiver able to elaborate real signals, Exercise 2b requires the use of SOPRANO (see Table IV). As stated in Section II-B, SOPRANO does not ex- TABLE III. TEXT OF THE FIRST EXERCISE ABOUT ACQUISITION Exercise 2a Objectives: Implementation of the Parallel Acquisition Schemes in Time and Frequency domain. Description: Following the MATLAB R given example (SerialSearch.m) for the Serial Search Acquisition scheme, write two functions that implement the Parallel in Time and Parallel in Frequency acquisition schemes. Use N-FUELS to generate the code and signal sequences necessary to test the behaviour of the SerialSearch function: generate a few ms of a GPS L1 signal (f IF = MHz, fixed Doppler shift at e.g. 1.5 khz, without noise, without front-end filter, without quantization). Then generate 1 ms of the local code (same PRN as before, same sampling frequency, code delay = 0 s, Doppler = 0 Hz, without noise, without front-end filter, without quantization) Questions: 1) Read from the corresponding data file the code and signal sequences and pass them to the SerialSearch function. Visualize the obtained search space and determine the maximum; determine and verify the corresponding estimated code delay and Doppler frequency shift (acquisition parameters). Plot the 1-D search functions along the delay axis (row of the 2-D search space corresponding to the estimated Doppler frequency) and Doppler frequency axis (column of the 2-D search space corresponding to the estimated code delay) 2) Repeat the generation of the signal sequence, with the insertion of the front-end filter. Repeat the signal acquisition and compare the obtained search spaces and estimated acquisition parameters. 3) Repeat the generation of the signal sequence at different levels of C/N 0 (e.g. 33 dbhz, 38 dbhz, 44 dbhz, 50 dbhz). Repeat the signal acquisition and compare the obtained search spaces and estimated acquisition parameters. 4) Repeat point 3 by using the other two acquisition schemes: Parallel in Time and Parallel in Frequency. hibit the performance required to elaborate the signal from an antenna in real-time, but its architecture directly derives from a real-time fully software receiver able to do it [4]. Consequently, the insertion of the new acquisition strategy in SOPRANO allows the students to understand the architecture of a software GNSS receiver able to elaborate real data. Moreover, they can experience which are the most critical parts of the receiver and why they demand an optimized implementation, in order to cope with performance requirements for the elaboration of a real signal. IV. CONCLUSION The use of N-FUELS and SOPRANO in hands-on sessions of courses in the Specializing Master on Navigation have improved students understanding of the characteristics of GNSS signals and related processing techniques. On the one hand, providing students with well-structured environments where they can play with parameters, such as N-FUELS, allows them to acquire the needed sensibility to deeply understand the structure of GNSS signals and receivers. On the other hand, the SOPRANO tool permits to smoothly move to an open

5 TABLE IV. TEXT OF THE SECOND EXERCISE ABOUT ACQUISITION Exercise 2b Objectives: Implementation of one acquisition scheme in a fully software receiver. Description: Replace the Parallel in Time acquisition employed in the SOPRANO fully software receiver with a C implementation of the Parallel in Frequency strategy. In particular, open the file /acquisition/cafevaluation.c and modify the PerformCaf function, whose prototype is the following: TCafEvaluationStatus PerformCaf (TSopranoStruct * r, int channelindex, char *data1, int size1, char *data2, int size2); where: TSopranoStruct * r is the pointer to the main structure, containing all data structures needed by the receiver; int channelindex is the index of the channel assigned to a satellite. This parameter must not be changed; char *data1, char *data2, int size1 and int size2 are related to the circular buffer which contains the ADC output samples. The first two parameters are pointers to two different chunks of data, while size1 and size2 are the number of samples available at the location data1 and data2 respectively. All these parameters must not be changed, since they involve the receiver framework. TCafEvaluationStatus is the type of the return parameter, indicating the status of the CAF evaluation. It assumes one of the following values: {CAF_EVALUATION_IN_EXECUTION, CAF_EVALUATION_SUCCESSFUL, CAF_EVALUATION_UNSUCCESSFUL}. The PerformCaf function has access to the data structure TCafEvaluation *caf, needed by the acquisition and contained in r, as clearly appears looking at this assignment: caf =&((r->channelarray[channelindex]).caf) where channelarray is the vector of channels, each one assigned to a different satellite. All the variables that you need to employ to implement the Parallel in Time strategy are contained in that structure, while the acquisition parameters are read from the file configurationfile/acquisitionparameters.txt and stored in TAcquisitionParameters *acqpars structure. Questions: 1) Repeat simulations performed in Exercise 2a, and compare the obtained results. 2) Discuss the strategies employed in the software receiver in order to reduce the complexity of the employed algorithm. 3) Perform an execution time profiling of the code and list the most critical points that must be further optimized in order to enable a straight execution with samples from an antenna. In order to be compliant with SOPRANO, input files described in Exercise 2a must be generated using 8 bits per sample. Fig. 3. Graphic representing the result of the acquisition procedure implemented in Exercise 2a. The position of the peak reflects actual values of Doppler shift and code delay. environment, where students can start experiencing receivers implementations which are more similar to those available in real-time software receivers. In the on-going edition of the Specializing Master on Navigation we have used for the first time both N-FUELS and SOPRANO in our classes. With respect to past years, when only N-FUELS was used together with a collection of MATLAB R routines, we have appreciated an easiest conduction of hands-on sessions, and a better understanding and acquired knowledge by students. As a matter of fact, being the geographic origin of our students quite variegated (from Africa to Australia passing by South East Asia and, of course, Europe), their background knowledge is not always homogeneous. The use of the two tools allowed us to better overcome lacks of background in programming which we sometimes verify in some of our students. In addition, the exercises discussed in previous subsections proved to be effective for increasing the motivation and the satisfaction of our students, encouraging them in improving their programming skills and providing a practical feeling about GNSS signals and receiver algorithms. At the end of the courses, we received positive and encouraging feedbacks from our students about their experience with N-FUELS and SOPRANO: in detail, we received positive comments from anonymous questionnaires and multiple expressions of interest for possible internships in our research group. In consideration of the positive results achieved, we are planning to extend the use of these tools to other courses on GNSS in the M.Sc. in Telecommunication at Politecnico di Torino. REFERENCES [1] A. Davies, Education, research, and innovation and technology transfer in GNSS (ERIG). Final report. GSA Virtual Library. Aug [2] F. Dovis, O. Julien, B. Deisting, G. Povero, M. Bousquet, and R. Blasi, G-TRAIN supporting education and training in the field of satellite

6 navigation in europe, in Global Engineering Education Conference (EDUCON 2012), (Marrakeshi, Morocco), Apr [3] F. Dovis, A. Lee, L. Lo Presti, and G. Povero, Master program on navigation and related applications: a partnership between university, private sector, and international bodies, in IEEE International Conference on Frontiers in Education (FIE 2006), (San Diego, California), Oct [4] M. Fantino, A. Molino, and M. Nicola, N-GENE GNSS receiver: Benefits of software radio in navigation, in Proceedings of the European Navigation Conference (ENC 2009), (Napoli, Italy), May [5] M. Troglia Gamba, E. Falletti, D. Rovelli, and A. Tuozzi, FPGA implementation issues of a two-pole adaptive notch filter for GPS/Galileo receivers, in The Institute Of Navigation GNSS Conference (ION GNSS 2012), (Nashville, Tennessee), Sept [6] E. Falletti and B. Motella, Combination of squared correlators for multipath mitigation in SoL GNSS receivers, IET Radar, Sonar and Navigation, vol. 6, issue 7, pp , [7] NavSAS Group, Contact! first acquisition and tracking of IOV Galileo signals, Inside GNSS, pp , Jan [8] C. Fernández-Prades, L. Lo Presti, and E. Falletti, Satellite radiolocalization from GPS to GNSS and beyond GNSS: Novel technologies and applications for civil mass-market, Proceedings of the IEEE, vol. 99, no. 11, pp , Nov [9] E. Falletti, D. Margaria, and B. Motella, Educational library of GNSS signals for navigation, Coordinates, vol. V, issue 8, pp , Aug [10] E. Kaplan and C. Hegarty, Understanding GPS: Principles And Applications. Norwood, MA: Artech House, [11] J. B.-Y. Tsui, Fundamentals of Global Positioning System Receiver. A Software Approach. Ney York: Wiley, 2nd ed., [12] J. Betz and B. Titus, Intersystem and intrasystem interference with signal imperfections, in Position Location and Navigation Symposium, PLANS 2004, pp , [13] M. Irsigler, G. W. Hein, and B. Eissfeller, Multipath performance analysis for future GNSS signals, in ION NTM meeting, (San Diego, CA), The Institute Of Navigation, January [14] SIGE semiconductor, SE4120L GNSS Receiver IC datasheet, DST , Rev 3.5, 26 May [15] K. Borre, D. Akos, N. Bertelsen, P. Rinder, and S. H. Jensen, A Software-Defined GPS And Galileo Receiver: A Single-Frequency Approach. Boston, MA: Birkhauser, [16] K. Borre, The GPS Easy Suite-Matlab code for the GPS newcomer, GPS Solutions, vol. 7, no. 1, pp , [17] K. Borre, GPS EASY SUITE II: A Matlab Companion, Inside GNSS, pp , May/June [18] C. Fernandez-Prades, C. Aviles, L. Estove, J. Arribas, and P. Closas, Design patterns for gnss software receivers, in Satellite Navigation Technologies and European Workshop on GNSS Signals and Signal Processing (NAVITEC), th ESA Workshop on, pp. 1 8, [19] A. Molino, M. Nicola, M. Pini, and M. Fantino, N-GENE GNSS Software Receiver for Acquisition and Tracking Algorithms Validation, in European Signal Processing Conference, EUSIPCO 2009, (Glasgow, Scotland), August 2009.