ASSESSMENT of the EGNOSsignalavailabilityFORLANDmobile user
Pavel Kovář, František Vejražka, Libor Seidl, Pavel Puričer
CzechTechnicalUniversity in Prague, Department of Radio Engineering, Faculty of Electrical Engineering
Technicka 2, 166 27 Praha 6, Czech Republic
phone +420 224 352 246
fax +420 224 355 829
Abstract
Keywords: EGNOS, mobile user, Markov model
The EGNOS system is a satellite based augmentation system (SBAS), which has been developed for improvement of the existing GNSS systems in order to be applicable in safety critical applications. The availability of the EGNOS system for land mobile user is limited by the shadowing of the satellite signals by various obstacles, terrain, buildings, vegetation etc.The EGNOS land mobile channel is analyzed in this project. The shadowing of the EGNOS signal is modelled by two-state Markov model in which a time variable is equal to covered distance. The aim of this contribution is to discuss topics concerning the experimental part of the project. The measurementswere prepared for various typical environments in the CzechRepublic like Prague city, urban and suburban areas, countryside, etc. The experiments were planned to make possible statistical processing and generalisation of their results.The experimental equipment for research of satellite signal shadowing process, a new version of the Experimental GNSS receiver, has been developed. For rapid EGNOS signal detection the fully parallel EGNOS software correlator has been developed and implemented into the receiver.
Introduction
The EGNOS (European Geostationary Navigation Overlay System) [1]is the first step on the way to European satellite navigation system. It belongs under satellite based augmentation systems (SBAS) that offer improvement of present GNSS systems. In the family of SBAS we can find besides EGNOS also US WAAS that is fully compatible with EGNOS. In Asian region there were developed Japanese MSAS and Indian GAGAN that would help to cover entire world by SBAS signal in the future.
Main purpose of the EGNOS is to improve performance and integrity of present systems, US system GPS-NAVSTAR and Russian system GLONASS, in order to enable use of these two systems for safety critical applications. Primary purpose of EGNOS, which has been developed for, is to support position determination and navigation of civilian aircrafts.
EGNOS operation principle is based on distribution of wide area differential corrections of GPS and GLONASS systems. Differential corrections carry information about measurement errors for individual satellites, information about precision of position determination, and information about integrity of the satellite signal. The broadcasted information is provided by the network of ground stations that execute monitoring the navigation signals.
The signal is disseminated by four geostationary satellites (GEO). The structure of EGNOS signal is similar to signal of GPS L1 C/A, i.e. ranging signal that enables distance measurement in addition to differential corrections dissemination.EGNOS satellites then extend the current navigational satellites constellation for several satellites which results in increase of precision of position determination due to PDOP decrease.
The system should be in the status of full operational capability since 2004. The full operational status is preceded by test in the frame of ESTB (EGNOS System Test-Bed). Since the current plans of EGNOS use cover land applications, it is crucial to investigate its availability to land mobile user.
1. Mobile satellite channel modelling
Mobile channel models of satellite systems are based on general channel models. The satellite mobile channels differ from terrestrial mobile channels in the fact that the direct signal path for mobile channels is blocked only in the close surrounding of the receiver antenna. The fluctuation of signal strength caused by the medium-term fading is then not characterized by log-normal distribution known from terrestrial channels but rather by its discrete states (open, closed).
Fig.1shows block scheme of the model of mobile satellite channelwith frequency flat fading[2]. The channel is characterised by its two states: open (unblocking), closed (blocking).The state open means that direct path signal is not blocked. In this state, the flat fading is modelled by the Ricean random process. The parameter expresses ratio of direct path signal power to power of diffracted and reflected signals.
The flat fading in the channel state closed is modelled by the multiplicative random process with log-normal distribution with parametersand. The mechanism of the opening and closing of direct path is usually modelled by Markov chain for the discrete time model and Markov process for the continuous time, respectively.
Fig.1. Model of a frequency-flat fading satellite channel
Mobile satellite channel at the frequency 1.5 GHz can be considered as a frequency flat fading channel for narrow bandwidth only. In case that bandwidth is wider than 30 MHz the delay of the reflected signals has to be taken into account. This leads to significant increase of model complexity[3].
2. Probability of EGNOS message reception
The analytical solution of the EGNOS signal reception issues for the mobile user is presented here. The two-state Markov model of the channel that was described in previous text is used. The solution is represented by messages reception statistics.
Let be a successful reception of i-th message and the beginning of the transmission of i-thmessage is (). Length of broadcasting of the message is. The messages can not be overlapping each other in time.
Assuming the steady state of the random process, i.e. probability of the states 0 and 1 is and, the probability of reception of one message can be then expressed as
Probability of reception of at least one message from N messages can be expressed by probability of union of random sets
.
The term can be computed recursively with use of previous terms. Two sets case can be solved by
The conditional probability can be computed by the terms
The described term contains parts that can be computed recursively.The remaining parts, for example conditional probability of type, can be computed with respect to assumptions above according to
for.
3. EGNOS fast detection algorithm
First step to successful reception of the EGNOS message is correct and fast detection of the EGNOS signal. The advantage of the EGNOS is its use of geostationary satellites for signal transmission. Thus the signal is not affected by the Doppler offset of the carrier frequency caused by the movement of the satellite. The frequency offset of the received signal is then caused by user receiver movement and inaccuracy of the receiver clock source. The sources of errors and the related frequency offsets are summarised in the following table.
Sources of Errors / Max. Frequency Offset[Hz]Satellite movement / 0
User movement (max. radialvelocity 72 km/h) / 100
Inaccuracy of the receiver clocks 10-7 / 150
Total / 250
Table 1.Contributions to the frequency error
The results in the table show that maximal frequency offset in the discussed case will be 250 Hz, which is one half of the frequency step during searching state. The detection of the presence of the EGNOS signal for designed receiver can be done by the search of the maximum of the correlation function for one frequency only, which is equal to L1.The task is thus simplified to a search in signal delay.
The other method that will speed up the algorithm of the search of the correlation function maximum is use of parallel computation of correlation function. The receiver in that case has to be able to computeduring time interval value of correlation function for all of investigated delays.
The signal processing algorithm of EGNOS signal detection implemented in the experimental GNSS receiver can be according the block scheme at Fig.2separated into four consecutive parts:
- Transformation of the intermediate frequency signal to the baseband and signal decimation
- Parallel computation of cross correlation function of received signal and its replica
- Noncoherent integration
- Search of the correlation function maximum and signal presence detection
Fig.2. Block scheme of signal processing algorithm
The whole signal processing algorithm was developed and tested in Matlab Simulink environment with use of Xilinx DSP Blockset add-on and Xilinx System Generator [4]. The described combination was used because of planned implementation into the experimental GNSS receiver based on FPGA device Xilinx Virtex-II, described bellow. The implementation is shown on the Fig.3. BlockGPS_EGNOS_parallel_correlatortransforms signal to the baseband andcomputes cross correlation of the received signal and replica. Block Noncoherent_integratorexecutes noncoherent integration. The last block Peak_detectorexecutes search of the maximum of correlation function andevaluates signal presence. The implementation scheme is for testing purposes also equipped by blockof logical analyzerChipScope, whichis connected to the output of parallel correlator and noncoherent integrator.
Fig.3. Matlab Simulink model of signal processing algorithm
4. Experimental measurements
This chapter describes realization and results of experimental measurements of shadowing process of EGNOS satellites for land mobile user. The experimental GNSS receiver based on Software Defined Radio concept was used for experimental measurements. The equipment was installed in the car with carriage. The measurements were realized during common traffic operations in typical environments:
•city of Prague
•industrial town
•village
•highway
•main and secondary roads in the country
The measured data were processed by mathematical software and compared to theoretical results.
4.1. Experimental GNSS receiver
High performance analogue to digital and digital to analogue converters together with high computational power of programmable devices and digital signal processors (DSP) enable to convert analogue RF signal to its digital form and process this signal as a sets of binary data. The realization of basic blocks and even whole signal processing stages in present systems, for example radio receivers and radio transmitters for communications or navigation, is then transferred to digital domain. It includes signal filtering, modulation/demodulation, synchronization, etc. The border of converting analogue signal to digital moves to the higher frequencies (e.g. intermediated frequencies 70 and 140 MHz). At this level, the signal with bandwidth of tens of MHz can be now processed with sufficient dynamic range thanks to high performance of digital devices.
As all operations with signal are done by digital programmable circuits or devices, the signal processing is determined by a implemented software algorithms or devices software configurations. The concept of radio receiver, whose main parts of signal processing chain are realized digitally by programmable digital circuits, is usually called Software Defined Radio (SDR) or simply Software Radio.
The SDR concept of embedded processor in FPGA device was chosen for design and development of experimental GNSS receiver at the CzechTechnicalUniversity in Prague. The main aim of the design was to provide flexible and versatile platform for implementation, testing, and verification of GNSS signal processing algorithms. The receiver should serve also as a highly configurable device for GNSS signal measurements and tests. To achieve these requirements the modular concept of the receiver was chosen.
Generally, the receiver consists of two main parts: RF unit and DSP unit. In the first version, the RF unit consisted of two channels, providing down converted analogue signal to DSP unit connected to High Power Computer (HPC) (Fig.4). The DSP part was based on Xilinx FPGA Virtex-II device platform realized as a PCI interface card installed in HPC represented by PC workstation running under MS Windows 2000 operating system [5].
Fig.4. Experimental GNSS software receiver architecture – first version
The initial test and experiments with first version receiver have shown that chosen concept is appropriate for planned purposes, however with several restrictions. The main issues related to the first version of GNSS receiver were represented by following problems:
- The Windows 2000 proved not to be reliable operation system for such application, where frequent software handling is required. The interrupt dropouts were sometime observed.
- Communication with DSP via PCI bus, that is block oriented, caused some problems. The direct access to the correlator registers would be more convenient.
- The first version of the experimental GNSS receiver is incompact; that complicates for example mobile experiments realization.
These problems lead to design of advanced version of the receiver. The main change was done in DSP unit design that is now represented by FPGA device Virtex-II Pro with two embedded IBM PowerPC PPC–405 cores (Fig.5). That platform enables single chip integration of all digital processing parts, i.e. correlators and computer for tracking and navigation task resolving. For achieving higher reliability of the whole system, the true real time multitasking operating system is chosen instead of Windows 2000.
Fig.5. Experimental GNSS software receiver architecture – advanced version
Moreover, the number of RF channels was increased to four due to modernization of the GNSS systems, where the new GNSS signals on the new frequencies will be available. The receiver should process these signals simultaneously. The RF channels are unified to keep the simplicity and compactness of the receiver. Each channel is equipped with SAW intermediated frequency (IF) filter of unified bandwidth 24 MHz. The intermediate frequency is increased on 140 MHz. It insures higher suppression of undesirable signals on the mirror frequency. The resolution of the analogue to digital converters will be 8 bits. The sample frequency is designed 80 MHz. Such frequency can be easily derived from 10 MHz normal by frequency multiplication.
4.2.Installation of measuring equipment
The measuring equipment has to measure and record presence of EGNOS signal and travelled distance.Because of complicated direct installation to the car the sensor of travelled distance was mounted to the carriage towed behind the car.The construction of the sensor is based on combination of permanent magnets and magnetic contacts similar to sensors used for bicycle computers. The wheel of carriage was equipped by three permanent magnets (Fig.6). The magnetic contact of sensor is based on miniature dry-reed relay in DIL14 package, mounted on the base of carriage.
Fig.6. Sensor of the travelled distance – wheel with magnets and relay mounting
The measured signals are pre-processed in 8051 series microcontroller, which processes pulses from distance sensor and information about presence of EGNOS signal from experimental GNSS receiver. Processed data are then stored in notebook. The experimental receiver and notebook were installed in the cabin of the car (Fig.7)
EGNOS signal was received by the common consumer GPS L1 antenna with magnetic mounting on the roof of the car.
Fig.7. Experimental GNSS receiver and notebook for data storage
4.3. EGNOS satellite selection
The mobile experiments were carried out in January to March of 2005. EGNOS was operating during time of experiments in its testing phase ESTB. EGNOS signal was regularly transmitted by Inmarsat IOR (PRN 131) satellite. Another signal from Inmarsat AOR-E (PRN 120) satellite was available in the unscheduled intervals besides IOR satellite signal.
The Inmarsat IOR satellite is located above the Indian Ocean in the position 64º East and the elevation of the satellite in the Czech Republic is around 15º [6]. In the full operational phase will the system transmit EGNOS signals from the satellites listed in the Table 2.The elevation angle of new satellites will be more than 30º, which will result in better availability of EGNOS signal for land mobile user. The regular broadcast from these new satellites was not available in time of experiment, that's why the Inmarsat IOR (PRN 131) satellite was used for experimental measurements. Several measurements were done for AOR-E satellite too.
Satellite / position / ELEVAtioninCR / PRNInmarsat III (AOR-E) / 15,5W / 26º / 120
ESA Artemis / 21,5E / 32º / 124
Inmarsat III F5 / 25E / 32º / 126
Inmarsat IOR / 64,5ºE / 15º / 131
Table 2. Constellation of EGNOS Satellites
4.4. Measurements results
Figures 8, 9, and10show measured processes of EGNOS satellite shadowing in typical environments (village, main and secondary roads –Fig.8a, industrial town –Fig.9a, city of Prague–Fig.10a). The computed estimates of distribution of time of blocking and unblockingstates of satellite signal path based on measured shadowing process are shown at Fig.8b, Fig.9b, and Fig.10b. The pictures show comparison of the described distribution with exponential distributiondensity function, which relies to theoretical distribution of persistence of Markov process in each of the states. The graphs show that estimate of distribution of open or closed state is in correspondence with theoretical distribution except for part closeto zerodistance.
(a) / (b)Fig.8. The EGNOS satellite shadowing (a),Distribution of blocking and unblocking states (b)
–village, main and secondary roads
Fig.9. The EGNOS satellite shadowing (a),Distribution of blocking and unblocking states (b)
–industrial town
Fig.10. The EGNOS satellite shadowing (a),Distribution of blocking and unblocking states (b)
–city of Prague
The next step of measured data analysis is an evaluation of message reception statistics. The probabilities of message reception were obtained from the simulation based on measured shadowing processes. The statistics obtained from real measurements were compared with probability statistics computed from mathematical model.
(a) /
(b)
Fig.11. The probability of reception of at least one message from N for real data (a) and mathematical model (b)
– village, main and secondary roads
(a) /
(b)
Fig.12. The probability of reception of at least one message from N for real data (a) and mathematical model (b)
– industrial town