GMSK iterative carrier recovery for AIS burst-mode

Transcription

1 Available online at June 017, 4(3): The Journal of China Universities of Posts and Telecommunications GMSK iterative carrier recovery for AIS burst-mode Lu Xiwen 1, Luo Yijun 1 (), Li Jie 1. School of Electronic Information, Wuhan University, Wuhan , China. Xi an Institute of Space Radio Technology, Xi an , China Abstract In this paper, an iterative carrier recovery algorithm of Gaussian filtered minimum shift keying (GMSK) in burst-mode was designed. The data utilization rate and precision of the traditional demodulation method for recovering carrier signal in burst data packet with limited-length is poor. In order to solve this problem, this paper proposed an iterative carrier recovery algorithm. This algorithm can improve the estimated precision of carrier recovery and data utilization rate of burst data packet in a large extent by performing multiple forward and backward iterations. And the algorithm can be implemented in Simulink environment. As the communication mode of automatic identification system (AIS) is abrupt, the algorithm is especially suitable for satellite-based AIS. Keywords GMSK, iteration, carrier recovery, AIS, burst signal 1 Introduction With the rapid development of modern technology, high efficiency communication systems are increasingly essential to our daily life. What s more, people s demands of communication system are also boosted. The maritime traffic has become increasingly intensive and busy, due to the development of shipping business, the obvious tendency of ships large-scale and high-speed, as well as the increase of ship number and load. As early as Before Christ (BC), people had already applied vessel traffic management system (VTMS) [1] which used to ensure the safety and efficiency of vessel within a certain area. The development of VTMS was divided into three generations. The first generation system before the 1950s used wireless telegraph and telephone to communicate. Due to this system merely utilize simple light, sound, electricity and mechanical signal, it can only manage ports, canals and narrow river channels. During the period of 1950s to 190s, we applied radar and very high frequency (VHF) radio telephone to manage vessel traffic which belong to the Received date: Corresponding author: Luo Yijun, Lyj@whu.edu.cn DOI: /S1005-5(17)6005-X second generation system. Additionally, the artificial recognition method acquires information through VHF radio telephone, which has some drawbacks, such as difficult to identify, low identifying speed, high error rate and so on []. Although the next improved method used a very high frequency direction finder (VHF-DF) recognition to improve recognition speed, it is still liable to cause traffic problem which is caused by the low positional accuracy, little useful information, and crowded VHF communication. Since the 190s, the VTMS applied computer technology as a core tool was the third generation system [3]. The Rader data processing with the computer technology is able to avoid crashing, but it cannot identify certain information including vessel speed, steering rate, and vessel specific information. Moreover, it is also apt to be influenced by weather and sea state. Therefore, it is difficult to work normally and to monitor other near vessels as soon as possible. In order to solve the above problems adequately and to strengthen the protection of the marine environment, the satellite-based AIS [4] was raised, which is a new kind of digital aid-to-navigation system including network technology, modern communication technology, computer technology and electronic information display technology.

2 The Journal of China Universities of Posts and Telecommunications 017 AIS was used to establish effective communication between stations and ships as well as ships and vessels to improve the safety and efficiency of navigation, and also to protect maritime trade environment at the same time. All the ships with an AIS equipment only send information on a regular time, including identification, location, direction, and speed, to realize the information interaction. Since the vessel identification problems in VTMS have been drawn abundant attention and the opinion of AIS has been proposed, the identification of vessel and research of AIS, such as discussion, research, argumentation, limit test, calibration, and promote have never been interrupted in the international maritime organization. AIS is a type of burst communication method and the signal is non-continuous and non-periodic. In order to track AIS signals, we use GMSK modulation mode to modulate [5]. Due to the envelope of GMSK signal has neither sharping edge nor turning point, which has great anti-interference characteristics as well as a perfect frequency and power density spectrum [6]. All of these properties make it highly suitable for wireless and satellite communication. In practice, different users were divided into different areas with time division multiple access (TDMA) technology. Different users can communicate with each other, but it is inevitable to receive overlapping burst signals because the receiving range covers multiple areas. Thus, in the design of ship communication system receiving algorithm, the key part of the method is how to accurately extract the corresponding coherent carrier and timing signal from the contaminated and burst data, which have overlap interference in the frequency and time domain. Traditional demodulation way can be split into coherent demodulation and non-coherent demodulation [7]. Non-coherent demodulation has the simple structure, but its anti-interference capability is very poor []; coherent demodulation has the great anti-interference capability, but its structure is extremely complex. Under the burst-mode, as a result of the limited data packet length, no matter the non-coherent demodulation and the coherent demodulation do not demodulate all data completely in the burst data packet and the demodulate precision much lower than continuous mode GMSK [9]. In recent years, demodulating AIS signal has not made considerable progress, people just view to consider improving AIS signal receiver and detector to avoid the effect accused by demodulation performance [10]. Analogy to the channel decoding iteration algorithm [11], we found that the iterative algorithm can build up the reliability and optimization performance of decoding, so this article put forward a kind of GMSK carrier recovery algorithm based on iterative. This algorithm can improve the burst data carrier frequency estimation precision significantly, furthermore, accurately and completely demodulate it. Theory analysis and calculate.1 Theory analysis of feed-forward square loop carrier recovery algorithm The GMSK modulation principle is shown in Fig. 1. Fig. 1 GMSK modulation principle block diagram GMSK signal can be described as: Eb πak st ( ) = cos ft t φ c k T π + + T b b (1) where E b is the symbol energy. T b is the symbol period. a k is data of the first k code whose value is either 1 or 1. φ k is phase constant of the first k code. When a k =±1, signal frequency must be ƒ + =πƒ c +π/(t b ) and ƒ - =πƒ c π/(t b ), so the carrier s frequency is ƒ c =(ƒ - +ƒ + )/. Firstly, we carry out the synchronous estimation by fast Fourier transform (FFT) to detect preamble of received signal from which we can get the estimation of the frequency offset [1], the time offset, and the initial phase. After the synchronous estimation, carrier can be recovered by feed-forward carrier recovery algorithm and Fig. shows the achieve structure. Fig. GMSK feed-forward carrier and timing recovery algorithm

3 Issue 3 Lu Xiwen, et al. / GMSK iterative carrier recovery for AIS burst-mode 3 In Fig., BPF is band-pass filter and n(t) is noise. Although the received GMSK digital intermediate frequency (IF) signal does not contain carrier component, double carrier frequency component can be generated after square transformation and extract it with a narrow band filter or an equivalent phase-locked loop, by which we can get coherent carrier by an operation of frequency division. Setting the input signal of carrier recovery circuit is st ( ) = mt ( ) P cos( ωt + θ ); ω = πf () s c 1 c c where P s is received signal power. m(t) represents original data signal. θ 1 is unknown phase. ω c is the carrier angular frequency. After squaring, we can obtain: s ( t) = P[ mt ( )] + P[ mt ( )] cos(ω t + θ ) (3) s s c If using α to represent direct component in [m(t)], then = + = + ( ) (4) [ mt ( )] α {[ mt ( )] α} α N t Substituting Eq. (4) into Eq. (3) s ( t) = αp + PN ( t) + αp cos(ω t + θ ) + s s m s c 1 N ( t) P cos(ω t + θ ) (5) m s c 1 where the first item of the right side in Eq. (5) is the direct component, and the second in Eq. (5) is the low frequency component. As for the third, it is a discrete spectrum component whose frequency is ω c, which is exactly what we need. And the fourth is sideband components which symmetrically distribute on both sides of ω c. We can filter out the third part that we need through narrow band filter whose center frequency is ω c, then conduct frequency division can get the required frequencies, f + and f, that is to obtain s I (t)=cos(πƒ + t) and s Q (t)=cos(π f t). Adding and subtracting these two parts respectively can obtain carrier φ I (t) and φ Q (t). The timing signal can be extracted by multiplying these parts and making them pass through a LPF. Fig. 3 shows the structure diagram of square loop carrier recovery. Fig. 3 Extracting process of f + and f m 1. Theory analysis of carrier recovery based on iterative As mentioned in Sect. 1, in order to adapt to the burst data scenario in marine communication and improve the system ability of anti-interference signal. We shall propose an iterative algorithm that execute many times forward and backward operations for burst data packet to extract the real carrier frequency successively, which based on feed-forward square loop carrier recovery algorithm, as shown in Fig. 4. (a) Forward calculation (b) Backward calculation Fig. 4 Square loop feedback method carrier frequency Now, we provide a theory analysis of this algorithm. Fig. 4(a) shows the process of forward calculation. Assuming IF signal is cos(ω c t+θ e ), then the useful part after squaring is cos(ω c t+θ e ). In the phase detector, we multiply the square of IF signal and the local carrier from NCO output that is predicted value obtained by synchronization estimation, which can be expressed as: 1 cos(ωct + θe)[ sin( ωct)] = sin( θe) 1 sin(4 ωct) cos( θ 1 e) cos(4 ωct) sin( θe) (6) The high frequency part of Eq. (6) and the noise can be filtered out by LPF, and then get phase error information, 0.5sin(θ e ). Moreover, we can get the first forward iterative carrier frequency and phase by loop filter and NCO. After a forward calculation, the new output carrier frequency and phase are taken as the initial value of the second calculation (see Fig. 4(b)). Comparing with the calculation process of forward iterative, the backward iterative calculation has three differences: 1) Reverse the square data flow of the IF signal.

4 4 The Journal of China Universities of Posts and Telecommunications 017 ) The loop filter has the narrower bandwidth to improve the precision of the loop filter in backward iteration. 3) In the phase accumulation part, previous phase quantization value is obtained by a later phase quantization value minus the phase error. Similarly, the backward iterative frequency and phase are the initial value of the next forward iterative operation. The algorithm gradually aims at the frequency and phase to achieve carrier synchronization in the mode of burst data packet by alternating the forward and backward operation. 3 Concrete simulation procedure 3.1 Base band signal source Base band signal source is burst GMSK signal which was generated as the format of AIS signal [13], and the length of each burst data packet is limit (56 bit, baud). The AIS frame format is shown as Fig. 5. Fig. 5 AIS every 56 bit detailed data format The generating of interference signal and main signal are basically the same, which just staggered for a little period of configurable time. In the Simulink program block diagram, the input source has a main signal and two interference signals, whose strength are 1/ and /15 of main signal, respectively. Then these signals will pass the additive white Gaussian noise (AWGN) channel whose E b /N 0 = db, where E b is the power of the signal and N 0 is the power spectral density of the noise. In the simulation, carrier frequency is MHz, main signal relative frequency offset is , and interference signals relative frequency offset are 10 5 and The initial phases are different and the sample rate is khz. 3. Signal modulation and carrier recovery We use the Matlab, and utilize sim function in M procedure to call Simulink, so that we only execute M procedure to avoid complicated steps that operating forward calculation and backward calculation repeatedly. It is also convenient to set the previous operate results as the operating parameter of the next step. Furthermore, we can control Simulink circulation times by for function. Following, we modulate the signal source and use the iteration algorithm to recover carrier. In the loop filter for carrier recovery, the parameter C 1 and C are combined to determine the capture bandwidth, capture time and tracking accuracy. When C 1 /C is smaller, the loop has broader bandwidth but lower accuracy. On the contrary, when C 1 /C is larger, the loop has narrower bandwidth but higher accuracy. The calculation formulas of the two parameters are 1 ξωnt C1 = (7) K K ξωt + ( ω T) C 0 d n n 1 4( ωnt) = K 0Kd ξωnt + ( ωnt) where ω n is natural angular frequency. K 0 is the gain of NCO. K d is the gain of phase discriminator gain and ξ is damping factor. T is symbol period. After many times tests, we conclude that setting the parameters C 1 and C as the Table 1 can recover carrier more accurate. ()

5 Issue 3 Lu Xiwen, et al. / GMSK iterative carrier recovery for AIS burst-mode 5 Table 1 The value of C 1 and C at different times The number of disposal C 1 C The first disposal The second disposal 3 The third disposal 4 The forth disposal 5 When we execute the first disposal, we set the C 1 /C to a small value due to we need the broad bandwidth. In the reverse disposal, that is second disposal, as the frequency offset curve is already stable and it does not need broad bandwidth, we adopt a relative large value of C 1 /C to achieve high accuracy. At the following disposal, we can increase the value of C 1 /C constantly to improve the accuracy. After the preliminary estimate, f + and f are taken as corresponding initial fixed value whose values are 150/ 19 and 0/ 19, respectively. After repeated dispose, Figs. 6(a)~6(c) show the results of f + every time disposal. After the carrier recovery, then we organize together the signal from LPF with delaying time to synchronize symbol timing [14]. (a) Frequency offset estimation of the 1st disposal (c) Frequency offset estimation of the 6th disposal Fig. 6 Frequency offset estimation results of f + Through Fig. 6 we can get that the loop convergence accuracy is becoming much more accurate and the frequency offset estimation signal to noise ratio (SNR) is also enlarged gradually with the increasing of iterative times. Comparing with the simulation result, the practical relative error of first forward estimation is After being processed for six times, the relative error is about and the surplus relative error is less than , which is much smaller than the demodulation algorithm referred by Ref. [15]. So after processing several times, this iterative algorithm increased the precision of the estimated frequency offset of the burst data packet by about 7 30 times, reaching Hz = 0.64 Hz, which gain excellent precision comparing to the existing algorithms. Similarly, after operating so many times the precision of phase error has greatly improved, and the total phase error can be within 3 ~ 4, which is also much smaller than the phase-recovery referred by Ref. [16]. The estimated precision can be further improved if we increase disposal time. 3.3 Signal demodulation (b) Frequency offset estimation of the 4th disposal This study simulates the whole GMSK demodulation system which to verify the recovery precision of carrier and synchronous timing signal based on recovered carrier. Due to the fact that the valid data is only a frame, and every frame data is only 56. In order to compare the influence of iteration times to the algorithm anti-noise performance, a simulation for the demodulation system of 50 frames burst data packet under different iterative times was executed when E b /N 0 = db, the results shown in Fig. 7.

6 6 The Journal of China Universities of Posts and Telecommunications 017 Due to the frame header data of burst data packet cannot lock without iteration, the bit error rate (BER) is much higher than under the iterative algorithm. From Fig. 7, we can see that as soon as proceeding iterative algorithm, the BER gets smaller at a large extent and the anti-interference performance also becomes stronger. Further analysis can get the conclusion that the BER fitting with the curve of the ideal [17] just after once iteration. Fig. 7 The demodulate BER under different iterative times 4 Conclusions In conclusion, as the traditional demodulation algorithms have the defects of poor data utilization rate and low carrier recovery precision in burst mode. Thus, we proposed an iterative algorithm to recover carrier based on traditional feed-forward square loop coherent demodulation structure. In addition, we make a simulation for this algorithm on Simulink circumstance, which achieve a success for burst-mode GMSK demodulation. This algorithm can maximize the data utilization rate and demodulate the burst packet data completely. Moreover, it greatly improves the accuracy of the frequency offset estimation in the presence of interferences. We used two methods to verify the frequency offset estimation precision for this algorithm. The first one, observing simulated waveform of frequency offset estimation after every iteration calculation, and then comparing result from the waveform with the estimated value by FFT. From this way, we can put forward a conclusion that the frequency offset becoming more and more accurate. Second, we used the recovered carrier and timing signal to demodulate corresponding signal. We find that the BER of demodulated signal is very close to the ideal value and the simulated waveform of demodulated signal achieves a higher level of match with the signal source, so we can also get the conclusion that precision of carrier and timing signal through this algorithm is accurate. Moreover, the algorithm fully indicates the excellent performance of sacrifice time for precision brought by iteration. References 1. Van Westrenen F. Modelling arrival control in a vessel traffic management system. Cognition, Technology and Work, 014, 16(4): Wang S Y, Xu K Y. The present, prospect and counter measure of AIS. Marine Technology, 001, (5): (in Chinese) 3. Qi Q, Yu T. The development of vessel traffic management system (VTS). China Radio, 013, (4): 36 3 (in Chinese) 4. Holsten S. Global maritime surveillance with satellite-based AIS. Proceedings of the 009 OCEANS Conference Europe, Mar 11 14, 009, Bremen, Germany. Piscataway, NJ, USA: IEEE, 009: 4p 5. Zheng K, Hu Q, Zhang J B. Positioning error analysis of ranging-mode using AIS signals in China. Journal of Sensors, 016: ID Murota K. Spectrum efficiency of GMSK land mobile radio. IEEE Transactions on Vehicular Technology, 195, 34(): Huang Y L, Fan K D, Huang C C. A fully digital noncoherent and coherent GMSK receiver architecture with joint symbol timing error and frequency offset estimation. IEEE Transactions on Vehicular Technology, 000, 49(3): Elnoubi S M. Analysis of GMSK with discriminator detection in mobile radio channels. IEEE Transactions on Vehicular Technology, 196, 35(): D Andrea A, Mengali U, Reggiannini R. Carrier phase and clock recovery for continuous phase modulated signals. IEEE Transactions on Communications, 197, 35(10): Xia P Y, Zhang T S, Ding L H, et al. Practical concern analysis on the detection probability for satellite-based AIS. Proceedings of the nd Asia-Pacific Conference on Communications (APCC 16), Aut 5 7, 016, Yogyakarta, Indonesia. Piscataway, NJ, USA: IEEE, 016: Xu Z, Wang K, Li Z. Decoding structure of Turbo code based on parallel prediction control. Journal of Beijing University of Posts and Telecommunications, 009, 3(6): (in Chinese ) 1. Peng H, Li J, Ge L D. Non-data-aided carrier frequency offset estimation of GMSK signals in burst mode transmission. Proceedings of the th IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 03): Vol 4, Apr 6 10, Hong Kong, China. Piscataway, NJ, USA: IEEE, 003: Cheng H, Wang X H, Cao Z F. An improved non-coherent demodulation algorithm for satellite-based AIS signals. Telecommunication Engineering, 014, 54(): (in Chinese ) 14. Costas J P. Synchronous communications. Proceedings of the IRE, 1956, 44(1): Cai Y. The research and implementation of AIS signal receiving demodulation algorithm. Master Thesis. Nanjing, China: Nanjing University of Science and Technology, 016 (in Chinese ) 16. Prévost R, Coulon M, Bonacci D, et al. Joint phase-recovery and demodulation-decoding of AIS signals received by satellite. Proceedings of the 3th IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 13), May 6 31, 013, Vancouver, Canada. Piscataway, NJ, USA: IEEE, 013: Guo T Y, Liu Z J, Wang X M, et al. Data transmission. nd ed. Beijing, China: The People s Posts and Telecommunications Press, 199: 373 (in Chinese ) (Editor: Wang Xuying)