Master of Science Thesis

Transcription

1 TAMPERE UNIVERSITY OF TECHNOLOGY Degree program in Information Technology A.K.M.NAJMUL ISLAM CNR ESTIMATION AND INDOOR CHANNEL MODELING OF GPS SIGNALS Master of Science Thesis Examiners: Docent Elena-Simona Lohan and Prof. Markku Renfors Examiners and topic approved in the council meeting of the Faculty of Computing and Electrical Engineering on the 16th Jan, 2008

2 Abstract TAMPERE UNIVERSITY OF TECHNOLOGY Degree Program in Information Technology, Department of Communications Engineering ISLAM, A.K.M.NAJMUL: CNR estimation and indoor channel modeling of GPS signals Master of Science Thesis, 68 pages, 3 appendix pages March, 2008 Major: Communications Engineering Examiners: Dr. Elena-Simona Lohan, Prof. Markku Renfors Keywords: Binary Offset Carrier, Binary Phase Shift Keying, Carrier to Noise Ratio, Global Positioning System, Galileo, Pseudolites, Satellites In recent studies, accurate positioning of terminals has received much attention in wireless communications research. One reason is that of the requirement for emergency call positioning imposed by the authorities. The positioning algorithms based on the Global Positioning System (GPS) have limitations in the indoor environments because the signal experiences severe attenuation in such situations. Estimation of the Carrier to Noise Ratio (CNR) is one of the most important functionalities of the Global Navigation Satellite Systems (GNSSs) receivers. However, the conventional GPS receivers are not able to estimate the CNR accurately enough in moderate or severe indoor reception. In this thesis, several moment-based CNR estimators are derived and the author shows the results for both Binary Offset Carrier (BOC) and Binary Phase Shift Keying (BPSK) modulated signals. BOC modulation is to be used in modernized GPS signals and for Galileo, the European navigation system, while BPSK is currently employed by basic GPS signals. The results of different estimators are compared in order to find the most robust estimator. On the other hand, the indoor propagation characteristics of the GPS signals are required to be well understood in order to derive good navigation algorithms suitable for indoor environments. In this thesis, the indoor propagation channel using pseudolites and satellites are also analyzed, based on the measurement data collected in different scenarios. i

3 Preface I have written this Master of Science thesis for the Department of Communications Engineering, Tampere University of Technology, Finland. I have done the work for this thesis at the Department of Communications Engineering under the projects, Advanced Techniques for Personal Navigation (ATENA) and Future GNSS Applications and Techniques (FUGAT) funded by the Finnish Funding Agency for Technology and Innovation (Tekes) and some participating companies. I would like to express my gratitude to my thesis supervisors Dr. Elena Simona Lohan and Prof. Markku Renfors for their valuable guidance and assistance during the thesis work. I would also like to thank Dr. Yuan Yang, Danai Skournetou, and Hu Xuan for their friendly support during the work. Finally, I express my gratitude to my parents for their endless love and inspiration. This endeavor is dedicated to my wife, Nasreen Azad. Tampere, Finland. 25th March, 2008 A.K.M. Najmul Islam Insinöörinkatu 60 B TAMPERE najmul.islam(at)tut.fi Tel. int ii

4 Contents Abstract Preface List of Abbreviations List of Symbols i ii vi viii 1 Introduction Motivation for the research topic Objective of the thesis Thesis contributions Thesis outline Fading channel modeling overview Multi-path channel parameters Time dispersion parameters Coherence bandwidth Doppler shift and Doppler spread Coherence time Fading channel models Rician fading channel Rayleigh fading channel Log-normal fading channel Loo fading channel Nakagami fading channel Overview of GPS and Galileo systems Basic GPS overview iii

5 CONTENTS iv 3.2 Modernized GPS system overview Galileo system overview Measurement campaigns Pseudolite based measurement campaign Single pseudolite based measurement Multiple pseudolite based measurement Satellite based measurement campaign Measurement data analysis Acquisition of C/A code Search window Search strategy Correlation Drift estimation Navigation data bit estimation Coherent integration CNR Estimation Signal model Coherent integration outputs Non-coherent integration outputs PDFs and CDFs Moment-based CNR estimators First-order moments (1STO) Second-order moments, method 1 (2NDO-M1) Second-order moments, method 2 (2NDO-M2) Combined second (central) and first (non-central) order moments (2NDO-1STO-M1) Combined second (non-central) and first (central) order moments (2NDO-1STO-M2) Combined fourth and second order moment (4THO-2NDO) Combined fourth and first order moments (4THO-1STO) Simulation results Results for single path channel Results for multi-path channel Envelope vs. squared envelope as nonlinearity

6 CONTENTS v 6.4 CNR mappings CNR estimators results with measurement data CNR estimation results for 4THO-1STO using different navigation bit estimators Computational complexity of the estimators Channel models based on measurement data Proposed combined fading channel models Fading distribution matching Pseudolite results Satellite results Average path number and time dispersion parameters Pseudolite results Satellite results Comparison between pseudolites and satellite results Conclusions and Future Works Conclusions Future Works References 69 Appendix A: Phase variation and Delay error estimation 74

7 List of Abbreviations ATENA AWGN BOC BPSK C/A CDF CIR CNR DS-SS E2-L1-E1 ESA FFT FUGAT GNSS GPS I Advanced Techniques for Personal Navigation Additive White Gaussian Noise Binary Offset Carrier Binary Phase Shift Keying Coarse/Acquisition code of GPS Cumulative Distribution Function Channel Impulse Response Carrier-to-Noise Ratio Direct sequence Spread Spectrum Frequency band centered in MHz European Space Agency Fast Fourier Transform Future GNSS Applications and Techniques Global Navigation Satellite System Global Positioning System In-Phase vi

8 LIST OF ABBREVIATIONS vii IF LOS MBOC MSE NLOS OS PL PRN PDF Q RHCP RMS RMSE RF SNR TUT Intermediate Frequency Line of Sight Multiplexed Binary Offset Carrier Mean Square Error Non Line of Sight Open Service Pseudolite Pseudo Random Number Probability Distribution Function Quadrature Right Hand Circular Polarization Root-Mean-Square Root-Mean-Square Error Radio Frequency Signal-to-Noise Ratio Tampere University of Technology

9 List of Symbols A α Envelope of the LOS delay Complex channel coefficient B w Bandwidth c f D τ θ E( ) Speed of Light Doppler Error Delay Error Phase difference of two consecutive millisecond outputs Statistic average E b Signal Power Ẽ b Signal Power in a correct bin f c Carrier frequency f D Maximum Doppler shift f ds Doppler spread f Combined M1 () f Combined M2 () f Combined M3 () Combined distribution method-1 Combined distribution method-2 Combined distribution method-3 viii

10 List of Symbols ix f Logn () f Loo () f Naka () f Rayl () f Rice () Lognormal distribution Loo distribution Nakagami distribution Rayleigh distribution Rician distribution K Rice Rician factor I 0 Modified 0th order Bessel function n m Degree of freedom of chi-square distributed variable Nakagami factor µ Mean N c Coherent Integration length N nc Non-coherent Integration length N 0 Noise Power N () Normal Distribution R BOC BOC/BPSK autocorrelation function σ 2 Variance CorridorSAT RoomSAT Satellite data captured in office corridor Satellite data captured in office room

11 List of Symbols x MultiplePL SinglePL Data captured with multiple pseudolites Data captured with single pseudolite θ n Phase of nth millisecond output T coh Coherence Time T ms Integration time in ms v m Maximum Doppler velocity v s Doppler velocity of satellite w Signal level x Q,o Imaginary part of correlation out-of-peak x I,o Real part of correlation out-of-peak x I,p Real part of correlation peak x Q,p Imaginary part of correlation peak y I,p Real part of correlation peak after N c ms y Q,p Imaginary part of correlation peak after N c ms y I,o Real part of correlation out-of-peak after N c ms y Q,o Imaginary part of correlation out-of-peak after N c ms z o Correlation out-of-peak after N nc ms z p Correlation peak after N nc ms χ 2 () Chi-square distribution

12 Chapter 1 Introduction This chapter gives an introduction to the thesis, beginning with the motivation to investigate the particular research topic. This is followed by a discussion of the research objectives and, finally, the thesis outline. 1.1 Motivation for the research topic The satellite-based navigation was started in the early 1970s. After some small-scale system studies, the GPS program was approved in December Since its launch, GPS has emerged as the most dominant technology for providing precise location and navigation capability to the end users. But GPS cannot provide adequate accuracy in some environments, such as indoor and densely populated urban areas. As a result, there is a clear requirement for developing a new navigation system that will overcome the limitations of the GPS and will be compatible with the GPS. The European navigation system, Galileo is planned to meet the overall requirements. Galileo is expected to operate by 2010 [15]. Galileo is an initiative of the European Commission and the European Space Agency (ESA). The new satellites are not yet in the orbit but the signal properties are already standardized in a first phase so that we can start to analyze the characteristics of these signals. Most of the applications of GPS are considered as outdoor applications but nowadays, the indoor personal navigation applications are getting popularity. In such situations, the typical GPS receivers suffer degraded performance or sometimes even complete failure because the signal experiences severe attenuation in the indoor environments. One of the most important personal positioning applications is the emergency call positioning in the cellular network, imposed by the authorities. The accuracy is very critical for such applications. Galileo is planned to increase the accuracy level for such applications. Still 1

13 Introduction 2 the characteristics of the indoor propagation need to be well understood to be able to develop the solution for the indoor positioning problem. In the outdoors, there are combinations of Line-Of-Sight (LOS) and Non Line-Of- Sight (NLOS) signals available, whereas in the indoors there is NLOS propagation only. Most of the time, there is no LOS signals available in the indoors due to the various obstructions. For the purpose of deriving indoor navigation algorithms, the satellite-toindoor propagation and its fading statistics have great importance. There are several studies that attempted to develop channel models for GPS-indoor channel. In [35] the authors used a strong reference outdoor signal to augment indoor processing capabilities and conduct coherent integrations of up to 160 ms. The existence of deep fades and their impact on indoor signals were observed. In [48] the author analyzed high bandwidth raw GPS data with high sensitivity techniques to characterize fading and the multi-path indoor characteristics. In [27], the authors showed that the GPS-indoor channel fading amplitudes of the first arriving peak matches well with the Nakagami-m fading model. In [26, 28], satellite-to-indoor propagation channel characteristics have been analyzed. It was shown that that the indoor signal is expected to be very weak and embedded in noise. Thus, long coherent and non-coherent integrations are required. Pseudolites (PLs), placed on earth surface, especially indoors, are a relatively new technology with great potential for a wide range of positioning and navigation applications. They can be used either as augmentation of space-based positioning systems or as independent systems for indoor positioning and capable of showing better performance [36, 47]. That is why, the PL-to-indoor propagation and its fading statistics have also great importance. The receivers should have the capability to estimate the Carrier to Noise Ratio (CNR) as accurately as possible. In the indoors the signal power remains very low, which affects the delay estimation accuracy, and, thus, the position accuracy. As a result, the conventional GPS receivers are not able to estimate the CNR accurately for the location services in moderate indoor reception. Although the topic of CNR estimation in the GPS receivers is addressed in the literature [22, 33, 38], not much of the published analysis is based on the correlation of the incoming signal. Also few moment-based CNR estimators are found in literature, but these estimators have not been developed based on the correlation function of the BPSK/BOC modulated signals [40]. Furthermore, the author is not aware of extensive comparisons between different CNR estimation methods.

14 Introduction Objective of the thesis As a part of the Advanced Techniques for Personal Navigation (ATENA) and Future GNSS Applications and Techniques (FUGAT) projects, the objective of this thesis has been to analyze the different measurement data captured from different satellites and pseudolites in different indoor scenarios. The purpose has been to estimate the CNR accurately and derive a suitable channel model. The ATENA and FUGAT projects are research projects carried out at Tampere University of Technology (TUT) in cooperation with some industrial partners during the years The overall objectives of the ATENA and FUGAT projects are very wide scale and thus this thesis only covers a very small part of that. 1.3 Thesis contributions The novel contributions of the thesis are given below: A study on different navigation data bit estimation methods for GPS signals. Derivation of three moment-based CNR estimators based on the correlation of the incoming signal. A comparison of these estimators has been performed including four other estimators derived in similar way. A procedure has also been proposed for choosing the required noise samples for estimating the CNR accurately. Combined fading channel models have been proposed for matching with the measurement data. 1.4 Thesis outline This thesis consists eight chapters. The structure of the thesis is given below. Chapter 1 has introduced the motivation, related previous studies and the overall objective of the research. Chapter 2 discusses the currently available fading channel modeling techniques which are used in the communication systems. Chapter 3 introduces the GPS and Galileo systems to the readers from the point of view of signal characteristics. Chapter 4 discusses the measurement setups for the different indoor measurement campaigns for GPS based pseudolites and satellites.

15 Introduction 4 Chapter 5 is dedicated to the measurement data analysis. In the navigation data estimation part of the thesis, different methods are studied. Chapter 6 describes a signal model and the derived moment-based CNR estimators based on the correlation of the incoming signal. The performance of each estimator is tested for simulation based BOC modulated signal and measurement based BPSK modulated signal. The results for different estimators are compared in order to find the most robust estimator. A comparative analysis of the navigation data estimation methods described in Chapter 5 is presented in this chapter too. Chapter 7 presents proposed theoretical fading channel models along with the channel model based on the raw data to the readers. Chapter 8 finally presents the conclusions of the overall research.

16 Chapter 2 Fading channel modeling overview Fading is the term used to describe the fluctuations in the envelope of a transmitted radio signal. Fading is a common phenomenon in wireless communication channels caused by the superposition of two or more versions of the transmitted signals which arrive at the receiver at slightly different times. The resultant received signal can vary widely in amplitude and phase, depending on various factors such as the relative propagation time of the waves and bandwidth of the transmitted signal [3, 16]. This chapter starts by discussing the multi-path channel parameters. Then it discusses the currently available fading channel modeling techniques commonly used in the communication systems. The most commonly known statistical representations of fading are: Rayleigh [48], Rice [24, 48], Nakagami-m [27, 49], Log Normal [48], and Loo [31] distributions. Finally, it presents the combined fading channel models by combining two or more models. 2.1 Multi-path channel parameters Time dispersion parameters In order to compare different multi-path channels, time dispersion parameters such as the Mean excess delay, τ and Root Mean Square (RMS) delay spread, σ τ are used. The mean excess delay is the first moment of the power delay profile and it can be given by [39]: τ = P(τ k )τ k k P(τ k ) where P(τ k ) is the relative amplitude of the multi-path component at kth delay (τ). k (2.1) 5

17 Fading channel modeling overview 6 The RMS delay spread is the square root of the second central moment of the power delay profile and is given by [39]: σ τ = τ 2 (τ) 2 (2.2) where τ 2 = P(τ k )τ 2 k k P(τ k ) k (2.3) Coherence bandwidth Coherence bandwidth, B coh is the maximum transmission bandwidth over which the channel can be assumed to be approximately constant in frequency. That is, a signal having frequencies within a bandwidth B coh will be affected approximately similarly by the channel. The RMS delay spread and coherence bandwidth are inversely proportional to each other. If the coherence bandwidth is defined as the bandwidth over which the frequency correlation function is above 0.9, then the coherence bandwidth is given by [29]: B coh 1 50σ τ (2.4) If the frequency correlation function is above 0.5, then the coherence bandwidth is given by [39]: B coh 1 5σ τ (2.5)

18 Fading channel modeling overview Doppler shift and Doppler spread The movement of the satellites introduces frequency shifts on the carrier and the code of the received signal. This phenomenon is known as the Doppler effect. The maximum Doppler shift, f D can be given by [45]: f D = v s f c (2.6) c where v s is the speed of the satellite, f c is the carrier frequency and c is the speed of light. The Doppler spread, f ds is defined as the range of frequencies over which the received Doppler spectrum is essentially non-zero. An example Doppler spectrum is given in Figure 2.1 based on Jake s model. The maximum Doppler shift used in this figure is 10 Hz. 1 Doppler power spectrum 0.9 Normalized Doppler power spectral density Frequency shift from the carrier [Hz] Figure 2.1: Example of Doppler spectrum Coherence time Coherence time, T coh is the maximum difference in time such that two states of the channel, measured less than T coh seconds apart, are still correlated at some extent. The Doppler spread and coherence time are inversely proportional to each other. A popular rule of thumb for digital communications is to define the coherence time by [39]: T coh = f D (2.7) where f D is the maximum Doppler shift.

19 Fading channel modeling overview Fading channel models The general term, fading is used to describe the fluctuations in the envelope of a received radio signal. However, when speaking of such fluctuations, it is of interest to consider whether the observation has been made over short distances or long distances. For a wireless channel, the former case will show rapid fluctuations in the signal s envelope, while the latter will give a more slowly varying, averaged view. For this reason, the first scenario is formally called small-scale or multipath fading, while the second scenario is referred to as large-scale fading [5]. Small-scale fading is explained by the fact that the instantaneous received signal strength is a sum of many contributions coming from different directions due to many reflections of the transmitted signal reaching the receiver [3]. Large-scale fading is due to shadowing. Rayleigh and Rician models are the common small-scale fading models. The Nakagami distribution also falls in this class. Log-normal can be used for large-scale fading. Loo model combines both small-scale fading and large-scale fading. A brief overview of the fading channel models is presented in the following sub-sections Rician fading channel The Rician distribution models the channel in the situation when there is strong LOS signal with the presence of some weaker, randomly-distributed multipath components. The envelope of a signal undergoing Rician fading can be expressed by [24, 37]: f Rice (w) = w σ 2 Rice ( exp (w2 + µ 2 Rice ) ) ( ) wµrice 2σ 2 I 0 Rice σ 2 Rice where f Rice (w) is the probability of the signal amplitude level w, σ 2 Rice is the variance (can be given by σ 2 Rice = (var(i) + var(q))/2, where I and Q are the in-phase and quadrature components of LOS coefficient), µ Rice = mean(i) 2 + mean(q) 2, and I 0 (x) is the modified 0th order Bessel function and can be given by [9]: µ 2 Rice 2σ 2 Rice where i is the imaginary unit. I 0 (x) = m=0 ( 1) m m!(m + 1) (2.8) ( ) ix 2m (2.9) 2 The Rician factor K Rice (i.e., the ratio of LOS to multi-path power) is given by K Rice = [37]. Example of Rician distributions is given in Figure 2.2.

20 Fading channel modeling overview Rician PDF with σ Rice = 1 µ Rice =0 µ Rice =1 µ Rice =2 µ =4 Rice Figure 2.2: Example of Rician distributions for different µ Rice Rayleigh fading channel Rayleigh represents the worst case fading case and it can be considered as special case of Rician distribution when no LOS component is present. The envelope of a signal undergoing Rayleigh fading can be expressed by [37, 48]: f Rayl (w) = w ( ) σ 2 exp w2 Rayl 2σ 2 (2.10) Rayl where σ 2 Rayl is given by σ 2 Rayl = 2 π mean( I 2 + Q 2 ) [37]. An example of Rayleigh distributions is given in Figure Rayleigh PDF σ =0.5 Rayl σ =1.0 Rayl σ =1.5 Rayl σ Rayl = Figure 2.3: Example of Rayleigh distributions for different σ Rayl

21 Fading channel modeling overview Log-normal fading channel Signals that propagate through some attenuating medium have a log-normal power distribution [13]. If z is the signal amplitude, the lognormal distribution can be expressed by [48]: ( 1 f Logn (w) = w exp (log 10 (w) µ Logn) 2 ) (2.11) 2πσ Logn 2σ 2 Logn where σ Logn represents the standard deviation of log 10 (A) and µ Logn represents the mean of log 10 (A), where A = I 2 + Q 2 is the envelope corresponding to LOS delay. The parameters σ Logn and µ Logn depend on the medium of propagation and possibly on the motion of transmitter and receiver. An example of Log-normal distributions with various σ logn is given in Figure Log normal PDF with µ logn = 1 σ logn =0.5 σ logn =1 σ logn =1.5 σ =2.0 logn Figure 2.4: Example of Log-normal distributions for different σ logn Loo fading channel Loo [31] has developed a statistical distribution assuming that the fading of the shadowed LOS signal is log-normally distributed and the multi-path signals fading is Rayleigh distributed. Loo s distribution model can be expressed by [24, 48]: ( ( 1 f Loo (w) = 0 x exp (log 10 (x) µ Logn) 2 2σ 2 Logn ) ( ) ( ) (x2 +w 2 ) I xz w 2σ 2 0 dx) Loo σ 2 Loo σ 2 Loo 2πσLogn (2.12)

22 Fading channel modeling overview 11 where σ Logn = std(log 10 (A)) and µ Logn = mean(log 10 (A)) represent the standard deviation and mean of the logarithm of the measured envelope A, respectively, and σ Loo = std(a), µ Loo = mean(a) are the standard deviation and mean of the measured envelope, respectively. An example of Loo distributions with various σ Loo is given in Figure Loo PDF σ Loo =0.5 σ Loo =1 σ Loo =1.5 σ =2.0 Loo Figure 2.5: Example of Loo distributions for different σ Loo Nakagami fading channel Three types of Nakagami distributions are found in the literature namely Nakagami-q, Nakagami-n and Nakagami-m distributions. The Nakagami-q distribution is also known as Hoyt distribution. It can span from one-sided Gaussian fading (q=0) to Rayleigh fading (q=1). The Nakagami-n distribution is also known as the Rician distribution. The Rician factor of the Rician distribution and the parameter, n of the Nakagami-n distribution are related by K Rice = n 2. The Nakagami-n distribution spans the range from Rayleigh fading (n = 0) to no fading (n = ) [42]. Nakagami-m distribution is a generic model of fade statistics that is used in the study of mobile radio communications [7, 32]. A wide class of fading channel conditions can be modeled with Nakagami-m distribution [32]. This fading distribution has gained a lot of attention lately, since the Nakagami-m distribution often gives the best fit to land-mobile [2, 41, 43] and indoor mobile multi-path propagation as well as scintillating ionospheric radio links [42]. The PDF of a Nakagami-m fading amplitude can be expressed by [25, 49]: f Naka (w) = 2 ( ) m m ) w exp( 2m 1 mw2, (2.13) Γ(w) µ Naka µ Naka

23 Fading channel modeling overview 12 where µ Naka = mean( α 2 ) = mean(a 2 ) is the mean of the envelope power (α is the complex channel coefficient), m is the Nakagami-m factor and Γ(.) is the Gamma function. The following estimate of m factor can be used (i.e., m is equal to the inverse of the normalized variance of the squared envelope) [25, 37]: m = µ 2 Naka mean(a 2 mean(a 2 )) 2 (2.14) For m = 1, Nakagami-m is equivalent to Rayleigh distribution [32]. Nakagami-m distribution can closely approximate the Nakagami-q distribution by a one-to-one mapping between m parameter and the q parameter. The mapping is given by [42]: m = (1 + q2 ) 2 2(1 + 2q 4 ) 2 (2.15) Another one-to-one mapping can be found between the m parameter and the n parameter allowing the Nakagami-m distribution to closely approximate the Nakagami-n distribution. The mapping is given by [42]: m = (1 + n2 ) n 2 (2.16) An example of Nakagami-m distribution with various m and µ Naka values is given in Figure Nakagami PDF m=1,µ Naka =1 m=1.0,µ =2 Naka m=2,µ Naka =3 m=3,µ Naka = Figure 2.6: Example Nakagami distribution for different m and µ Naka values

24 Chapter 3 Overview of GPS and Galileo systems This chapter presents the concepts of GPS and Galileo systems. It first starts with the basic position measurement concepts that have been used in the GPS. Then, it describes the modernized GPS system, and finally, the new European navigation system, Galileo that is planned to be launched within few years. 3.1 Basic GPS overview The GPS system is based on the time-of-arrival measurements. The distance between a satellite and a receiver is calculated from the knowledge of how much time it takes for the signal from the satellite to get to the receiver [8]. The signal from the satellite to the receiver travels at the speed of light. If we know the travel time of the signal from the satellite to the receiver, we can easily calculate the distance. Figure 3.1 demonstrates the distance based positioning in two-dimensional case. In order to determine the receiver position, three distances from three satellites are required. Two satellites give two possible solutions because two circles intersect at two points. Hence, a third satellite is needed to determine the receiver position uniquely. For similar reason, to calculate the position in the three-dimensional plane, four satellites and four distances are required. In the above discussion, it is assumed that the distance measured from the satellite to the receiver is very accurate and does not contain any bias error. But actually the measured distance has an unknown bias error because the receiver clock and the GPS clock are not fully synchronized. In order to resolve this bias error, one more satellite is required and thus in order to find the accurate position five satellites are needed [45]. However, the general statement is that four satellites can be used to find the receiver position, even though the measured distance has a bias error [45]. Although, it is enough to know four distances in the process of getting an accurate 13

25 Overview of GPS and Galileo systems 14 Satellite 1 Satellite 2 Receiver Satellite 3 Figure 3.1: Two-dimensional user position. positioning result, it is not enough to have four random satellites flying in the sky. The satellites must also know their positions. A global network of ground stations is needed to give the position data to the satellites. In general, the GPS system can be considered as comprising three segments: the space segment, the control segment and the user segment. The space segment contains all the satellites. The basic GPS system has a total of 24 satellites divided into six orbits. Each orbit has four satellites and has an inclination angle of 55-degree. The orbits are separated by 60 degrees and each orbit has a radius of 26,560 km. The control segment consists of five control stations. The purpose of this segment is to monitor the performance of the GPS satellites, generate and upload the navigation data to the satellites. The user segment consists of the GPS receivers and the user community. There are two main signals: the coarse/acquisition (C/A) and the precision (P) codes. The P code is modified by Y code, which is refereed as P(Y) code. P(Y) is used for military purpose and it is not available for civilian users. The basic GPS signal contains two carrier frequencies: L1 and L2. The center frequency of L1 is at MHz and L2 is at MHz. L1 frequency contains C/A and P(Y) signals and L2 frequency contains only the P(Y) signal. This thesis focuses mostly on the civilian C/A signal. The C/A code is BPSK modulated with a chip rate of MHz while the chip rate of P(Y) signal is MHz. The navigation message, which contains information about the satellites,

26 Overview of GPS and Galileo systems 15 GPS time, clock behavior and system status is modulated on both the L1 and L2 carriers at a chip rate of 50 bits per second (bps) with a bit duration of 20 ms. 3.2 Modernized GPS system overview The only navigation system that can be used worldwide is the GPS system. GPS is military operated, but it is used for many commercial and civilian services nowadays. It shows very poor performance for the indoor location based services too. As a result GPS does not offer enough accuracy or warranty of service and it cannot be used in many vital positioning applications. The GPS system has been upgraded to meet the requirements, but still its functionality cannot be trusted in many scenarios. The modernized GPS frequency plan is shown in Figure 3.2. The modernized GPS includes a new frequency band L5 ( MHz) that provides a wide-band signal. In addition, the new L2C signal will provide the civilian community a more robust signal that is capable of improving resistance to interference and allowing longer integration times to the receivers. A new military M-code will also be added to L1 and L2 bands, but will be spectrally separated from the civil codes. It has been decided that the new modulation type for the new M signal will be Binary Offset Carrier (BOC) modulation. Figure 3.2: The modernized GPS frequency plan [14].

27 Overview of GPS and Galileo systems Galileo system overview The European navigation system, Galileo is planned to achieve European sovereignty and service guarantees through dedicated system under civil control [15]. The overall Galileo system consists of 30 satellites (27 operational+3 active spares), positioned in three circular Medium Earth Orbit (MEO) planes at 23,222 km altitude above the Earth, and at an inclination angle of the orbital planes of 56 degrees [5]. The services that will be provided by the Galileo are: Open Service (OS), Safety of Life Service (SoL), Commercial Service (CS), Public Regulated Service (PRS) and Search and Rescue Service (SAR). The reliability of the Galileo services is higher than that of the GPS [44]. Galileo is meant to provide better navigation accuracy due to its signal properties. The BOC modulation is planned to be used in the Galileo signals [5]. The frequency plan for the Galileo system is shown in the Figure 3.3 which consists of four frequency bands: E5a, E5b, E6 and E2-L1-E1. The E2-L1-E1 band with the center frequency MHz is the most interesting band as the current GPS signal (C/A) is in it. This thesis mostly focuses on this frequency band. The readers who are interested in other frequency bands are referred to [17]. Both the GPS C/A code and Galileo OS signals are transmitted in the same frequency band. But still the signals do not interfere significantly with each other since different modulation is used. The most important characteristics of the Galileo signals, in comparison with the GPS signals, are the different modulation types and code lengths. SinBOC(1,1) (briefly denoted as SinBOC) has been the candidate modulation type for the Galileo OS signal in the E2-L1-E1 band for many years. The code length for the OS signal is 4092 chips, which is four times higher than the GPS C/A code length. Recently the GPS-Galileo working group on interpretability and compatibility has recommended an optimized Multiplexed Binary Offset Carrier (MBOC) spreading modulation that would be used by Galileo for its OS service and also by GPS for its L1C signal [18]. However, this thesis presents the simulation results for the SinBOC(1,1) only. For the technical details of the BOC modulation, the readers are referred to [18],[4] and [5].

28 Overview of GPS and Galileo systems 17 Figure 3.3: Galileo frequency plan [46]. The use of GPS and Galileo at the same time is very interesting. The accuracy can be increased a lot by using the two systems together. The indoor reception might be improved in this way to provide the location based services to the users.

29 Chapter 4 Measurement campaigns This chapter presents the measurement campaign descriptions to the readers. It first starts with the pseudolite (PL) based measurement campaign where single PL-based and multiple PL-based measurement campaign descriptions are discussed. Then it discusses the satellite-based measurement campaign. These measurements were captured with the help of Space Systems Finland (SSF) and u-nav Microelectronics. 4.1 Pseudolite based measurement campaign Using PLs, two types of measurement campaigns were undertaken: single PL-based and multiple PL-based. These measurement setup descriptions are given in the following subsections. In these measurements, PRN indexes higher or equal to 32 were used, which are mostly reserved for non-satellite use. The sampling frequency of the GPS receiver was MHz. The L1 carrier in the data was down-converted to an intermediate frequency (IF) of Hz Single pseudolite based measurement The measurements were first carried out in the Tamppi arena building, then in the Festia building of TUT, Finland during June, Two synchronized GPS receivers were used. One receiver was used as reference receiver and was connected to the PL via cable. The other receiver was connected to an indoor antenna measuring the signal coming from the air. The transmit antenna was placed in a fixed position at the first floor, with an elevation of around 7 meters with respect to the receiver. The radiation patterns of the helix antenna (transmit antenna) used in the measurements was Right Hand Circular Polarization (RHCP), where the main beam was within ±30/35 from the antenna pointing direction. 18

30 Measurement campaigns 19 Attenuation in PL software was 20 db. The measurement process was carried out in 5 sets: SET 1: It was captured in the Tamppi arena sports hall. The receiver was moved from inside the main beam to the outside. The photo of the environment is shown in Figure 4.1 and the schematic representation of the measurement set is shown in Figure 4.2. Figure 4.1: Photo taken in the Tamppi arena sports Hall from the transmitter position. SET 2: It was also captured in the Tamppi Arena sports hall. But this time the receiver was moved inside the main beam only. The schematic representation of the measurement set is shown in Figure 4.2. SET 3: It was the last set that was captured in the Tamppi Arena sports hall. The receiver was moved outside of the main beam. The antenna pointing direction was parallel to the direction of the movement. The schematic representation of the measurement set is shown in Figure 4.3. SET 4: It was captured in the Festia main hall. The receiver movement was within the main beam with few times out of LOS.

31 Measurement campaigns 20 Antenna Pointing Direction Antenna Pointing Direction m 7m Receiver Receiver Receiver 9m 6m 9m 6m Figure 4.2: Transmitter and receiver positions for the PL-based indoor propagation measurement in the Tamppi Arena. Left: during SET-1, SinglePL. Right: during SET-2, SinglePL. Sports Hall Receiver Antenna Pointing Direction Figure 4.3: Transmitter and receiver positions for the PL-based indoor propagation measurement in the Tamppi Arena during SET-3, SinglePL.

32 Measurement campaigns 21 Figure 4.4: Transmitter and receivers position for the PL-based indoor propagation measurement in the Fiesta building during SET-4, SinglePL and SET-5, SinglePL. SET 5. This set was also captured in Festia main hall. The receiver movement was within NLOS condition (almost always behind obstructions). In Figure 4.4, two pictures of the environment taken both from the transmitter antenna position and receiver antenna position are shown for SET-4, SinglePL and SET-5, SinglePL. The schematic representation of these measurement sets is shown in Figure Multiple pseudolite based measurement Multiple PLs based measurements were also carried out in TUT during November, The measurements were first carried out in Tietotalo main corridor, and then in the Institute of Communications Engineering (ICE) offices in 5 sets: SET 1: It was captured in Tietotalo main corridor with 2 active PLs: PL1 (PRN 33, used as reference) and PL3 (PRN 34) were placed according to Figure 4.6. The PLs were placed in the second floor at a height of 5 meters, and the receivers were in the ground level. The receiver movement was started from 10.5 m away from PL1, first towards PL3, then towards PL1, and so on. The attenuation in PL1 and PL3 were about db.

33 Measurement campaigns 22 Obstructions Obstructions 1st Festia Main Hall Level Receiver Antenna Pointing Direction Ground Level Figure 4.5: Transmitter and receivers position for the PL-based indoor propagation measurement in the Fiesta Main Hall during SET-4, SinglePL and SET-5, SinglePL. SET 2: This set was captured in the same environment setup as SET multiple 1. But one more active PL was used: PL2 (PRN 32, attenuation 50 db). The receiver was first moved towards PL1, then towards PL3 and at last below the bridge where PL2 was placed. SET 3: This set was also captured with three active PLs (PL1: PRN 33, attenuation 55 db; PL2: PRN 32, attenuation 40 db, PL3: PRN 34, attenuation 55 db). PL1 and PL3 were placed at the same location as before, but PL2 was placed one level up, at the third floor, above the window ceiling of Tietotalo. SET 4: It was captured in the office corridor of ICE with three active PLs. They were placed in a triangle. PL1 also used as reference (PRN 33, attenuation 60 db). PL2 and PL3 attenuation was 60 db and 55 db respectively. The receiver movement was along the corridor. SET 5: It was captured with the same configuration as for SET 4, MultiplePL (same attenuation and same PRNs), but PL2 was used as reference. Figure 4.7 shows a schematic representation of the measurement setup for SET 4, MultiplePL and SET 5, MultiplePL.

34 Measurement campaigns m 4 m 15.8 m PL1 H=5 m PL3 H=5m Receiver PL2 H=5m Figure 4.6: Schematic representation of measurement SET 1, SET 2 and SET 3, MultiplePL (the PLs are shown in red and the receiver in black). PL 3 H=1.4m 6.1 m PL 2 H=0.97m Offices Offices 15.4m PL 1 H=1.45m Receiver Offices Figure 4.7: Schematic representation of measurement SET 4, MultiplePL and SET 5, MultiplePL.

35 Measurement campaigns Satellite based measurement campaign Two satellite based measurement campaigns were undertaken by TUT and u-nav microelectronics, Finland during March, In both campaigns, the transmitters were the different GPS satellites available in view during the measurement date and the receivers were the integrated GPS receivers with sampling rate of MHz. Two GPS receivers synchronized to a common clock, operating in parallel were used. The first one was used to acquire the signal from an outdoor antenna placed on the roof of the building. This signal was quite strong, and it was used as the reference signal for code-phase and Doppler frequency acquisition, as well as for frequency drift estimation and correction. The second receiver was moved in the indoor environment to capture the indoor signal. The down-converted intermediate frequency (IF) was same as the PLs. Among the two measurements, one was carried out in typical office-room scenario and the other was carried out in typical office-corridor scenario. The first scenario, denoted by RoomSAT, shown in Figure 4.8 (left), corresponds to a small room without any window (about 5m 2 ), where in the front there was small corridor with large windows. Here the LOS signal is more likely to be absent. The second scenario, denoted by CorridorSAT, shown in Figure 4.8 (right), corresponds to a long corridor with open windows and doors. The receiver movement inside the environments was random and it was at the walking speed. All the measurements were taken for a duration of 1-2 mins for reliable statistics. Figure 4.8: Photo of RoomSAT (left) and CorridorSAT (right) scenario.

36 Chapter 5 Measurement data analysis This chapter discusses the data analysis steps that are followed to compute the Channel Impulse Response (CIR) estimates. The setup block diagram is detailed in Figure 5.1. An initial Doppler drift estimate (incorporating the drift due to low IF sampling) and an initial delay estimate are obtained based on the reference signal by scanning the whole delay- Doppler space. Also, the code drift, frequency drift and navigation data are estimated based on this reference signal and then removed from the wireless signal. Then, the wireless signal is correlated with the replica code, by taking into account the delay-doppler estimates and their drifts. In the indoor environment the coherent integration must be longer than 20 ms in order to compensate for the increase in the noise level. For this reason, the removal of the navigation data must be done before the coherent integration, similar with [26], [27] and [28]. The most important parts of Figure 5.1 are discussed in the following sections. replica GPS code Indoor signal Ref signal Correlators bank Doppler, delay and drift estimates Correlation on 1 ms Data removal Integration on Nc ms CIR estimates Navigation data estimates Figure 5.1: Block diagram of the measurement data processing. 25

37 Measurement data analysis Acquisition of C/A code For any DS-SS (Direct sequence Spread Spectrum) system, it is necessary to estimate the timing and the frequency shift of the received signal in order to be able to de-spread the received signal and to obtain the original data. For that, it is required to define the position where there is an alignment between the received signal and the spreading code [23]. This is done to estimate the Doppler shift and initial delay. The process is done using cross-correlation, which measures the similarity of the code and the delayed replica of the same code. The search process is usually two-dimensional by which both the time shift (i.e., delay) and Doppler shift can be determined [22]. The value of the Doppler shift changes over time according to the place and speed of the satellite. It is naturally much more easier to look for the correct frequency, if the probable Doppler shift is known in advance. According to [45] the maximum Doppler velocity, v s of the satellite is 929 m/s. The Doppler frequency caused by the land vehicle is often very small. For a stationary observer the maximum Doppler shift on the carrier is f D = v s f c c = = 5kHz where c is the speed of light and f c is the L1 carrier frequency. In the measurement campaigns, the receiver was moved in a low speed which was around 1 2 km/hr. So the maximum Doppler shift was around ±5 khz. However, if a receiver is moved by using a high speed aircraft, it is reasonable to assume the maximum Doppler shift is ±10 khz [45]. On the other hand the Doppler shift on the C/A code is quite small because of the low frequency of the C/A code. The C/A code has a frequency of MHz which is 1, 540 ( /1.023) times lower than the carrier frequency. For a stationary observer the maximum Doppler shift on the C/A code is f D = v s f c c = = 3.2Hz where f c is the C/A code frequency. If the receiver is moved at high speed, the Doppler shift can be assumed as ±6.4 Hz.

38 Measurement data analysis 27 In the search process, all possible code delays and frequencies are searched through with some predefined search steps. The search space is typically equal to the length of the spreading code in the code-delay domain. In the Doppler frequency domain, the search interval can be several khz or even tens of khz [11]. If some a priori information (i.e., assistance data) about the Doppler frequency is available, the frequency interval may be just few tens of Hz [11]. In order to accomplish the search in a short time, the bandwidth of the searching program cannot be very narrow. Using a narrow bandwidth for searching means taking many steps to cover the desired frequency range and it is time consuming. On the other hand, searching with wide bandwidth provides low sensitivity. So the bandwidth should be selected properly. For measurement data, samples are searched in the initial stage with a frequency step of 400 Hz. In the next stage a window of 200 correlators is used with smaller frequency step of 20 Hz to get better phase estimate and check the correctness of the frequency estimated from the previous stage. The correct delay of the C/A code for SET-1, SinglePL is shown as an example in Figure 5.2 after the initial stage Normalized correlation power Delay [chips] Figure 5.2: The correct delay of C/A code of SET-1, SinglePL Search window Each tentative code-phase is called a code bin (or a time bin) and each tentative frequency shift is denoted as a Doppler bin (or a frequency bin). One code bin together with one Doppler bin compose a search bin (or a test cell). The whole code-frequency uncertainty region can be divided into several search windows and each window can be divided into several time-frequency bins. The time-frequency search window defines the decision re-

39 Measurement data analysis 28 gion [22] Search strategy In the search stage, the search windows are examined to see whether the time-frequency estimate is correct or not. The search process is started from one search window, with a certain tentative Doppler frequency and a certain tentative delay. All delays and frequencies, which correspond to the size of the search window at issue, are searched through with the predefined search steps. If the window is decided to be dismissed, the search process moves on to the next search window, and the same procedure is continued, until the correct window and the correct delay-frequency combination is found [22, 23]. A hybrid search is used in this thesis Correlation The tentative time-frequency bins are tested and possible signals are detected via crosscorrelation. This means that the received signal is correlated with the reference code with different tentative delays and frequencies, and the resulting values are then combined to achieve a two-dimensional correlation output for the whole search window. From the correlation output, it can be further determined whether the search window is correct or not via a correlation peak which appears for correct delay-frequency combination. The correlation process is described in detail in [22, 34]. The correlation properties of the spreading codes are very important. If the auto and cross-correlation properties are perfect, the correlation function would appear as a pure impulse at the correct delay and will have zero values elsewhere. But actually there is always some interference and noise present, which affects the correlation output of the received signal and reference code. An example correlation function is shown in Figure 5.3 for SET-1, SinglePL data. The initially estimated carrier and delay for each set of PL signals are shown in Table 5.1. Also the carrier and delay for each set of satellite signal for different PRNs are shown in Table Drift estimation As discussed in the previous section, the satellite orbital motion can cause a Doppler shift up to 5 khz for stationary receiver. In addition, satellite clock drift affects the actual frequency emitted from the GPS satellites, causing a further Doppler effect. The carrier phase estimation is highly affected by the frequency and phase drifts. According to [48],