MINISTRY OF EDUCATION AND TRAINING HANOI UNIVERSITY OF SCIENCE AND TECHNOLOGY NGUYEN THI THANH TU RESEARCH AND DEVELOPMENT OF ADVANCED INTERFERENCE MITIGATION TECHNIQUES IN GNSS RECEIVERS MASTER OF SCIENCE THESIS COMPUTER AND COMMUNICATION ENGINEERING ACADEMIC SUPERVISORS: Dr Tạ Hải Tùng Dr Beatrice Motella Hanoi – 2015 BỘ GIÁO DỤC VÀ ĐÀO TẠO TRƯỜNG ĐẠI HỌC BÁCH KHOA HÀ NỘI NGUYỄN THỊ THANH TÚ NGHIÊN CỨU VÀ PHÁT TRIỂN KỸ THUẬT CHỐNG NHIỄU GIAO THOA TRONG BỘ THU GNSS LUẬN VĂN THẠC SĨ KHOA HỌC KỸ THUẬT MÁY TÍNH VÀ TRUYỀN THÔNG NGƯỜI HƯỚNG DẪN KHOA HỌC : TS Tạ Hải Tùng TS Beatrice Motella Hà Nội – 2015 ACKNOWLEDGMENTS In the first words of this thesis, I am extremely grateful to those who in various ways contributed to all the results presented in this thesis Foremost, I would like to express my sincere gratitude to my supervisors Dr Ta Hai Tung and Dr Beatrice Motella for their patience, encouragement, and broad research vision Their guidance helped me in all the time of this study and writing of this thesis In addition, I deeply acknowledge Prof Gustavo Belforte for his enthusiastic supports I would like to thank the supports from all the NAVIS members I am thankful to Tran Trung Hieu, Nguyen Dinh Thuan, Truong Minh Duc, who gave me a lot of helpful advice I also appreciate Ms Nguyen Thi Ha Phuong, who has always encouraged me since the very beginning of my study The list could not be complete without the Growing NAVIS project, funded by the European Commission under the FP7 Call Galileo.2011.4.3-1 – International Activities (Grant Agreement No 287203), for supporting my internship at ISMB from May to July, 2013 Last, and most importantly, I would like to thank my parents, my sister and her family for their love that definitely is my limitless support COMMITMENT I commit myself to be the person who was responsible for conducting this study All reference figures were extracted with clear derivation The presented results are truthful and have not published in any other person‟s work NGUYEN Thi Thanh Tu TÓM TẮT LUẬN VĂN Ngày nay, hệ thống định vị toàn cầu sử dụng vệ tinh (GNSS – Global Navigation Satelite System) đóng vai trò quan trong nhiều lĩnh vực khác Một số ứng dụng GNSS dân quan trọng kể đến như: giám sát phương tiện giao thông, đồng truyền thông đồ Do đó, yêu cầu tính xác liên tục thu GNSS ngày trở nên cấp thiết Việc sử dụng kỹ thuật trải phổ trực tiếp, mang lại nhiều ưu điểm chống nhiễu cho hệ thống GNSS GPS, Galileo Tuy nhiên, tín hiệu GNSS nhận antenna thường có lượng thấp, dẫn đến việc chúng dễ bị tác động nguồn nhiễu khác Tất tín hiệu truyền phạm vi tần số gần với băng tần tín hiệu GNSS trở thành nguồn gây nhiễu cho thu GNSS Có nhiều loại nhiễu khác tác động lên thu GNSS, đó, nhiễu băng hẹp (narrow-band interference – NBI) nhiễu có ảnh hưởng lớn tới hiệu thu GNSS Các nghiên cứu gần sử dụng Notch filter lọc nhiễu băng hẹp cho thu GNSS cho thấy phương pháp khả thi hiệu Các lọc notch filter đáp ứng xung vô hạn có độ phức tạp không cao dễ dàng triển khai cách hiệu Trong khuôn khổ luận văn, tác giả đề xuất thiết kế cho khối chống nhiễu băng hẹp thu GNSS Khối có khả phân loại nhiễu băng hẹp tự động điều chỉnh cấu hình notch filter, cho chất lượng tín hiệu đầu đạt tốt Luận văn xây dựng gồm chương sau:  Chương Fundamentals Mô tả sơ lược hệ thống GNSS nói chung, nguyên tắc hoạt động hệ thống Trình bày khái niệm liên quan đến nhiễu (interference), ảnh hưởng nhiễu lên hiệu thu tổng quan phương pháp phòng chống  Chương Notch filter for GNSS narrowband interference mitigation Giới thiệu Notch filter, lọc có khả loại bỏ nhiễu băng hẹp thu GNSS, tham số quan trọng Notch filter, cải tiến Notch filter giúp lọc có khả tự phát tần số nhiễu băng hẹp  Chương New design of notch filters on GNSS narrowband interference mitigation Đề xuất thiết kế khối chống nhiễu băng hẹp thu GNSS  Chương Performance analysis Trình bày kết đánh giá đề xuất nêu chương  Chương Conclusion Tóm tắt nội dung luận văn, kết đạt TABLE OF CONTENTS Acknowledgments Commitment Tóm tắt Luận văn List of Figures List of Tables 11 List of Abrreviations 12 Introduction 13 Chapter 1.1 Fundamentals 16 GNSS Overview 16 1.1.1 Fundamentals of satellite navigation 17 1.1.2 GNSS receiver structure 19 1.2 Interference Threat 20 1.3 Interference Mitigation Techniques 23 Chapter Notch Filter for GNSS Narrowband Interference Mitigation 26 2.1 Notch Filter 26 2.2 Adaptive Frequency Notch Filter 28 2.3 Using Notch Filter on GNSS Narrowband Interference Mitigation 31 Chapter New Design of Notch Filter on GNSS Narrowband Interference Mitigation 33 3.1 Impact of Notch Filter on GNSS Signal 33 3.2 New Design of Narrowband Interference Mitigation Module for GNSS Receiver 36 3.2.1 Interference Characterization Block 37 3.2.2 Notch Filter Configuration 53 Chapter Performance Analysis 55 4.1 Bandwidth Estimation Algorithm 55 4.2 Quality of Filtered Signal 57 Chapter Conclusion 61 References 62 LIST OF FIGURES Figure 1.1 Fundamental of a satellite navigation system 17 Figure 1.2 Clock misalignments in a GNSS system 18 Figure 1.3 High-level architecture of a GNSS receiver 19 Figure 2.1 Frequency response of notch filter for two difference pole factors 26 Figure 2.2 Spectrum at the front-end output at different time instants (from [4]) 28 Figure 2.3 Power of filter output when notch frequency varies 29 Figure 2.4 Spectrum of an interfered signal before and after filtering 30 Figure 2.5 Adapted notch frequency of an ANF 31 Figure 2.6 High-level block diagram of the GNSS receiver chain 32 Figure 3.1 Proposed Interference Detection/ Estimation Unit 37 Figure 3.2 Notch frequency varies 38 Figure 3.3 Spectrum of the input signal 40 Figure 3.4 The signal power at the output of a NF and its derivative Bandwidth of NF is fixed at 30 kHz and whose frequency sweeps over the spectrum range 42 Figure 3.5 Bandwidth estimation algorithm 43 Figure 3.6 Interference bandwidth estimation for different NFs bandwidth 45 Figure 3.7 Proposed method for characterizing the incoming NBI 46 Figure 3.8 The accuracy of the interference bandwidth estimation algorithm versus 47 Figure 3.9 Notch frequency varies in the presence of a CWI 48 Figure 3.10 versus 52 Figure 3.11 Zoom of Figure 3.10 52 Figure 4.1 Simulation results: estimation of the interference bandwidth ( ) 55 Figure 4.2 Simulation results: estimation of the interference bandwidth ( ) 56 Figure 4.3 Estimated interference bandwidth over time, in case of variable band 57 Figure 4.4 Theoretical and practical C/N0 when bandwidth of interference (ACARS Harmonics) and 58 Figure 4.5 Theoretical and simulated C/N0 in case of CWI ( ) 59 Noted that: the used two-pole IIR NF is a combination of two one-pole NFs Thus, without loss of generality, it is possible to just consider the one-pole NF From [1]: ( | ( )) | (3.20) ( ) And: ( (| ) | ) ( ( ( | | | ) ) (3.21) | ) The extremes of the filtered power are also the extremes of (| | ) , where They are also solutions of the problem: (| ( ( Since and | | | | ) (3.22) | ) ( | | ) | ) | ( (| | ) | 49 | ) Thus: ( )* ( ( ( ( Since for )) ) ], ( ) √( ( ) [ ( ) √( ( )+ ) ) ) , this solution can be ignored Hence (3.22) has two solutions: √ ( ) For this reason, in the presence of CWI, if ANF is utilized to estimate the interference band, the estimated value ̂ is also the distance between solutions of (3.22): ̂ √ ( From [1], the bandwidth of NF is ̂ From (3.13) and (3.24), ( ) (3.23) Hence: √ ) (3.24) can be investigated in both cases of band-limited interference and CWI It should be noted that for L1 band, interferences with are considered as CWI, and interferences with are considered as NBI (which is also the band-limited interference) [16] Thus, investigated with CWI and the interferences which satisfy is The results are shown in Figure 3.10 and Figure 3.11 It can be seen that in the range of ), (because of the condition: of CWI is always smaller than those of band-limited interference, and there is a threshold separating them (Figure 3.11) The simulations based on the achieved equations show that the threshold The first value of should be about 0.585 should be large enough to ensure the speed of convergence of the interference bandwidth estimation algorithm for even a wide narrow-band interference (for example, ) However, if is too wide, the difference of between CWI and a narrow band-limited interference is very small Therefore, it might lead to a false alarm in determining if interference is CWI or band-limited interference To overcome this limitation, this threshold can be adjusted depending on when , In particular, should be set at 0.5, it means that for this case, a CWI is considered as a band-limited interference Then will be decreased, and the process of estimating the interference bandwidth is continued If , will be reconfigured as 0.6, which gives a higher accuracy of the interference classification 51 CWI Bi = 0.7kHz 3.5 Bi = 4kHz Bi = 50kHz Bi = 100kHz B r 2.5 1.5 0.5 1.5 2.5 BN (Hz) Figure 3.10 3.5 4.5 x 10 versus 0.65 CWI Bi = 0.7kHz 0.64 Bi = 4kHz 0.63 B r 0.62 0.61 0.6 T 0.59 0.58 0.57 0.5 1.5 2.5 BN (Hz) 3.5 Figure 3.11 Zoom of Figure 3.10 4.5 x 10 3.2.2 Notch Filter Configuration As mentioned in section 2.1, notch frequency and notch bandwidth are two important parameters of a NF Notch frequency can be adapted by LMS algorithm, while notch bandwidth, which is decided by pole factor authors recommended that best , is often set as a constant In [14], should be set at 0.9, whereas [7] recommended that the is from 0.99 to 0.998 However, it should be noted that: with the same notch bandwidth can be different if the sampling frequency relation between , the is different, because the and the notch bandwidth can be estimated as follow [16]: (3.25) For this reason, these recommended values might not be reliable for different receiver configurations (or different In this section, ) is estimated by optimizing C/N0 of the filtered signal in the expression (3.12), which presents the quality of the filtered signal through the filter magnitude response and the characteristics of all components of received signal These characteristics are: (i) the power of useful signal interference bandwidth Among them, since , (ii) the noise density , (iii) the and frequency , and (iv) the power of interference does not change the shape of the C/N0 line, it is not necessary to know this value in advance; can be calculated from the temperature of the environment As restricted at the beginning, in this thesis, two considered specific cases of NBI are CWI and band-limited interference In the case of band-limited interference, automatically adapted by an ANF, characterization block, and is can be provided by the Interference can be estimated by AGC or spectrum analysis [19] Therefore, it can be assumed that all these factors are known Consequently, the 53 expression (3.12) has only one variable However, it is difficult to find the value of ̂ directly from (3.26) ̂ { } (3.26) However, equation (3.12) can be discretized and calculated fast and easily with varying Hence, the problem (3.26) can be solved approximately by a numerical method Then, the configuration of NF can be completed For CWI, it has to note that the zeros of NF are constrained to be on the unit circle, thus, (3.12) becomes: |∫ | (3.27) ∫ It is clear that C/N0 is no longer depending on the power of interference In this case, the smaller notch bandwidth is, the little loss of useful signal is However, notch bandwidth must be large enough to ensure the stability of the adaptive filter For this reason, should be chosen as a predefined constant if the incoming interference is a single tone From above analysis and results, the NF can be configured as following:  If the incoming interference is CWI,  If the incoming interference is NBI, is a predefined constant is selected by a numerical method so that the C/N0 of signal is the highest The performance analysis of this design will be presented in the next chapter CHAPTER 4.1 PERFORMANCE ANALYSIS Bandwidth Estimation Algorithm The simulation result obtained by implementing the bandwidth estimation algorithm is presented in Figure 4.1 In this experiment, the characteristics of input signal are introduced in Figure 3.3 (Conf 1) The notch bandwidth of two ANFs is fixed at 30 kHz The estimated frequencies at the output of them are shown on Figure 4.1 The two black lines correspond to the actual limits of the interference bandwidth, while the red and blue lines are their estimation 140 120 100 Frequency (kHz) 80 60 40 20 -20 -40 -60 0.5 1.5 2.5 time (samples) 3.5 4.5 x 10 Figure 4.1 Simulation results: estimation of the interference bandwidth ( ) Figure 4.2 shows achieved result with a typical signal available to a commercial GPS receiver with interference bandwidth (ACARS Harmonics) and 55 (data demodulation threshold for the band-limited interference) [15] The interference frequency matches the GPS signal frequency (Conf 2) Similar to Figure 4.1, ANF and ANF work well on estimating the interference bandwidth Estimated interference band over time 20 15 Frequency (kHz) 10 -5 -10 -15 -20 0.5 1.5 2.5 time (samples) 3.5 4.5 Figure 4.2 Simulation results: estimation of the interference bandwidth ( x 10 ) Furthermore, the performance of the algorithm has also been tested in case of variable interference bandwidth Figure 4.3 shows the results when the algorithm is applied to a data set of s containing an interference source, which characteristics vary over time (i.e., the interference bandwidth is 10 kHz over the first 500 ms, and 18 kHz over the last 500 ms of simulation) 20 Interference bandwidth estimation (kHz) 18 16 14 12 10 0 100 200 300 400 500 600 time(ms) 700 800 900 1000 Figure 4.3 Estimated interference bandwidth over time, in case of variable band 4.2 Quality of Filtered Signal To validate the performance of Notch Filter configuration block, some simulations were implemented Since a too large filter bandwidth leads to the distortion of useful signal, the notch filter is investigated with higher than 0.9 [12, 7] Figure 4.4 shows the theoretical line of C/N0 with the simulated signal which has the characteristics as introduced the signal in Figure 4.2 It can be seen that the best is about 0.97, which corresponds to the notch bandwidth of 155 kHz This value does not match the interference bandwidth, because C/N0 depends on not only the interference bandwidth, but also the power of interference (This is the reason that it is not possible to use the estimated interference bandwidth to calculate directly by (3.26)) Furthermore, compare with other recommended pole factors in [7] (0.9, 0.99 to 0.998), there is an improvement in C/N0 of filtered signal about more than to dB 57 NBI = 25kHz; CNo = 45dB-Hz; Pi = -126.5dB 44 42 C/No eff 40 38 36 34 32 30 0.9 Simulation Theory 0.91 0.92 0.93 0.94 0.95 0.96 pole factor 0.97 0.98 0.99 Figure 4.4 Theoretical and practical C/N0 when bandwidth of interference (ACARS Harmonics) and In case of CWI, the theoretical and simulated C/N0 are shown on the Figure 4.5 Since is the worst case of PRN1 GPS L1 signal [5], if this interference is not filted, C/N0 will downgrade about 25 dB (see Figure 4.5) These simulations also suggest that the best for CWI is in the range [0.99; 0.998], which corresponds to the notch bandwidth from 10 to 50 kHz For the stable of the ANF, the value of 0.9 is chosen The performance of the proposed architecture with simulated signals is shown on Table 1.1 and Table 4.2 Table 1.1 shows the performance of the proposed architecture for CWI The interference frequency is from the center L1 band Because of the configuration as shown on Table 4.1, PRN and PRN are in the worst case ( [5]) It can be seen that their channels lost track of signal if the ANF is not applied The other channels are affected by the interference; thus, their C/N0 is much smaller than the configured value of simulated signal (45dB) The proposed method uses than , this configuration brings a higher performance The higher is not used because it might lead to the unstable of the ANF CWI: CNo = 45dB-Hz; Pi = -130dB 45 40 C/No eff 35 30 25 20 Simulation Theory 15 0.9 0.91 0.92 0.93 0.94 0.95 0.96 pole factor 0.97 0.98 0.99 Figure 4.5 Theoretical and simulated C/N0 in case of CWI ( ) For a band-limited interference, C/N0 of the filtered signal is shown on Table 4.2 The input signal characterizations are as follows:  PRN and Doppler shift are chosen randomly  Bandwidth of interference  Interference frequency matches the intermediate frequency: 59 It can be seen that the when the power of the interference is high (-126.5 dB), a too narrow notch does not remove enough the interference power; then it leads to the signal loss of track Whereas, when the interference power is lower (-135 dB), both a too narrow and wide notch not give the best performance Table 4.1 Performance of the proposed architecture in case of CWI PRN Proposed Method Doppler No filtering shift ̂ 40.11 Loss of track 43.58 1500 40.00 34.46 43.58 5000 40.09 Loss of track 43.67 12 4500 40.09 35.54 43.52 Table 4.2 Performance of the proposed architecture in case of band-limited interference PRN Doppler Proposed Method shift ̂ 2500 38.90 34.79 38.85 1600 39.26 35.23 39.41 5000 38.68 Loss of track 38.59 12 4200 38.93 35.23 38.95 2500 40.16 40.82 41.66 1600 39.85 41.29 41.91 5000 39.85 40.97 41.61 12 4200 40.19 41.25 42.07 ̂ kHz ̂ kHz CHAPTER CONCLUSION The focus of this thesis is the use of notch filter on narrowband interference mitigation for GNSS signal Through assessing the impact of notch filter on GNSS signal, a new design of narrowband interference mitigation module is proposed The proposal includes:  The interference bandwidth estimation algorithm: As its name, the output of this algorithm is an estimated interference band The algorithm is based on ANFs, which are designed originally to track the interference frequency However, the cost function is modified so that the notch frequencies converge to the boundaries of the interference band instead of the interference frequency Simulations show that the algorithm works well even with time-varying bandwidth interference  Interference classification algorithm: This algorithm allows classifying the incoming interference into CWI or band-limited interference Then, if the interference is band-limited, the input parameters of interference bandwidth estimation algorithm are configured appropriately  Notch filter configuration block: Through assessing C/N0 of the signal after notch filtering, a mechanism of configuring the notch filter is proposed It uses a numerical method to give a set of notch filter parameters, which optimize the interference suppression Simulations show that C/N0 of the filtered signal is improved more than 1dB 61 REFERENCES [1] Borio D (2008), A Statistical Theory for GNSS Signal Acquisition PhD dissertation, Politecnico di Torino [2] Borio, D.; Camoriano, L.; Lo Presti, L (2007), Two-Pole and Multi-Pole Notch Filters: a Computationally Effective Solution for Interference Detection and Mitigation, 7th International Conference on ITS [3] Falletti, E., Margaria, D., and Motella, B (2009), "A Complete Educational Library of GNSS Signals and Analysis Functions for Navigation Studies", Coordinates, 2009, V, (8), pp 30-34 [4] Giordanengo G (2009), Impact of Notch Filtering on Tracking Loops for GNSS Applications, Master‟s thesis, Politecnico di Torino [5] Kaplan E and Hegarty C (2006), Understanding GPS: Principles and Applications, Artech House, second ed [6] Kim, N (2006), Interference Effects on GPS Receivers in Weak Signal Environments, University of Calgary, Department of Geomatics Engineering Masters Thesis [7] Landry, R., Calmettes V., Bousquet M (1998), Impact of interference on a generic GPS receiver and assessment of mitigation techniques Spread Spectrum Techniques and Applications, 1998 IEEE 5th International Symposium on Vol IEEE, 1998 [8] Landry, R , Renard, A (1997), Analysis of potential interference sources and assessment of present solutions for GPS/GNSS receivers, 4th Saint-Petersburg on INS, May [9] Motella B (2008), Interference Monitoring for Global Navigation Satellite Systems, PhD dissertation, Politecnico di Torino [10] Motella B., Pini M., Dovis F (2008) “Investigation of the Effect of Strong Out- of-Band Signals on Global Navigation Satellite Systems receivers,” GPS Solutions, Volume 12, Issue 2, pp 77-86 [11] Parseval's theorem, http://en.wikipedia.org/wiki/Parseval%27s_theorem [12] Pini, M., Falletti, E., and Fantino, M (2008), Performance evaluation of C/N0 estimators using a real time GNSS software receiver, In Spread Spectrum Techniques and Applications, 2008 ISSSTA'08 IEEE 10th International Symposium on (pp 28-31) IEEE [13] Plancherel theorem, http://en.wikipedia.org/wiki/Plancherel_theorem [14] Raasakka, J., and Orejas, M.: Analysis of notch filtering methods for narrowband interference mitigation In Position, Location and Navigation Symposium-PLANS 2014, 2014 IEEE/ION (pp 1282-1292) IEEE [15] Satellite navigation, http://en.wikipedia.org/wiki/Satellite_navigation [16] Tabatabaei Balaei, A (2007), Detection, characterization and mitigation of interference in receivers for global navigation satellite systems, Doctor of Philosophy, The University of New South Wales [17] Trinkle, M., and Gray, D (2001, July), GPS interference mitigation; overview and experimental results, Proceedings of the 5th International Symposium on Satellite Navigation Technology & Applications (pp 1-14) [18] Ta Hai Tung (2009), Acquisition Architecture for Modern GNSS Signals, PhD dissertation, Politecnico di Torino [19] Yang, J H., Kang, C H., Kim, S Y., Park, C G (2012) Intentional GNSS interference detection and characterization algorithm using AGC and adaptive IIR notch filter Int J Aeronaut Space Sci, 13(4), 491-498.B8 63 ...BỘ GIÁO DỤC VÀ ĐÀO TẠO TRƯỜNG ĐẠI HỌC BÁCH KHOA HÀ NỘI NGUYỄN THỊ THANH TÚ NGHIÊN CỨU VÀ PHÁT TRIỂN KỸ THU T CHỐNG NHIỄU GIAO THOA TRONG BỘ THU GNSS LUẬN VĂN... gives an overview of Global Navigation Satellite Systems, interferences and their effects of interference on performance of GNSS receivers And a summary of interference mitigation techniques is presented... 28 2.3 Using Notch Filter on GNSS Narrowband Interference Mitigation 31 Chapter New Design of Notch Filter on GNSS Narrowband Interference Mitigation 33 3.1 Impact of Notch Filter on GNSS Signal