A Novel Scheme for Symbol Timing in OFDM WLAN Systems

Transcription

1 86 ECTI TRANSACTIONS ON ELECTRICAL ENG., ELECTRONICS, AND COMMUNICATIONS VOL.3, NO.2 AUGUST 2005 A Novel Scheme for Symbol Timing in OFDM WLAN Systems Yong Wang 1, Ge Jian-hua 1, Bo Ai 2, Li Zong-qiang 3, and Nie Yuan-fei 1,Non-members ABSTRACT In this paper we focus our research on the coarse symbol timing synchronization technique in Wireless Local Area Network (WLAN) COFDM systems. An analysis is made on the difference between the symbol timing synchronization technique in continuous mode transmission systems such as DVB-Tand the burst mode transmission systems such as WLAN. A new method to do coarse symbol timing in time domain is proposed. Simulations show that it has distinct advantage over the conventional Maximum Likelihood (ML) method under multi-path wireless channel conditions. The corresponding Field Programmable Gate Array circuit through test in HDTV prototype in Team of Engineering Expert Group (TEEG) proves that the new method is not only suitable for burst mode but also suitable for continuous mode transmission systems. Keywords: OFDM, WLAN, Symbol Timing 1. INTRODUCTION Coded Orthogonal Frequency Division Multiplexing (COFDM) is a promising technique for future mobile wireless data systems in order to achieve highbit data rate [1]. It has many advantages over conventional single carrier systems, such as robustness against multi-path delay spread [2], the improvement of channel throughput, adaptive modulation of subcarriers according to the channel conditions, extremely high spectral efficiency etc. Many digital transmission systems have adopted OFDM as their key modulation technique such as digital audio broadcasting (DAB), digital video broadcasting terrestrial TV (DVB-T)[3], WLAN systems based on IEEE (a) or Hiperlan2 [4] and asymmetric digital subscriber lines (ADSL). Some related technology such as vector OFDM (V-OFDM), wide-band OFDM (W-OFDM), flash OFDM(F-OFDM) have presented their great advantages in certain application areas. 04PSI06: Manuscript received on December 5, 2004 ; revised on July 5, The authors are with the National key Lab. of ISN in Xidian University, Xi an, China, xdwy@yahoo.com.cn 2 The author is with the National key E&E Lab. on microwave and digital communications in Tsinghua University, Beijing, China 3 The author is with the College of Armed Police Force, Shaanxi, Xi an, China However, there are also some disadvantages in OFDM systems, for example, the large Peak-to Average Power Ratio (PAPR) and its sensitivity to synchronization errors. These errors will not only cause inter-symbol interference (ISI) but also introducing inter-carrier interference (ICI) due to the loss of orthogonality between subcarriers. The analysis of synchronization errors to OFDM systems can be found in [5-7]. Fundamental theory about symbol timing synchronization are simply described in Section 2 and in Section 3, we propose a new method for the symbol timing synchronization with some conclusions drawn in Section SYMBOL TIMING SYNCHRONIZATION The OFDM system symbol timing synchronization is to find the start of OFDM symbol, i.e., the FFT window position. Just as what is shown in fig.1, we call the shadow field the ISI free region. Guard Interval Channel Impulse Response Maximum Delay of Multi-Path Fig.1: nth OFDM Symbol ISI Free Region ISI free region Time (n+1)th OFDM Symbol If the estimated start position of OFDM symbol is located within the ISI free region, data will not be affected by ISI, the phase rotation caused by timing offset can be easily corrected after FFT; if the estimated start position locates within the data interval, the sampled OFDM symbol will contain some samples that belong to other symbols, causing the dispersion of signal constellation which will reduces the system performance [8]. There is a close relationship among the symbol timing synchronization, the frequency synchronization and the sampling clock synchronization. The bit-error-rate (BER) degradation in db with relation to frequency offset is expressed in equation (1) [7]. D(dB) 10 ( π. N.δF ). E s (1) 3ln10 W N 0

2 A Novel Scheme for Symbol Timing in OFDM WLAN Systems 87 where N denotes the number of subcarriers in one OFDM symbol, f represents the frequency offset, the system bandwidth and SNR is denoted by W and E s /N 0 respectively. If SNR and the W is fixed, a larger frequency offset estimation range will need an increased subcarrier spacing, thus shortening the OFDM symbol length. This means a higher requirement for symbol timing synchronization. The sampling clock offset also has effects on the symbol timing synchronization. For example, assume that 1ppm sampling clock offset in 2Kmode with a guard interval of 512 samples in DVB-T [2], then the FFT window will move one sample about every 400 symbols. The higher the sampling clock offset, the more the influence has on the symbol timing synchronization. Therefore, all factors related to symbol timing should be taken into consideration. Many papers have been concerned with this technique, and they can generally be classified into 2 categories: one is based on the pilots embedded in OFDM symbols or the training symbols [9-13]; another one utilizes the inherent structure of OFDM symbols, e.g. cyclic prefix (CP) of OFDM symbol [14-18], a typical method exploits both the CP and the pilots is presented in [17]. There will be a larger error floor to do the coarse symbol synchronization with CP correlation method due to the destroyed data by ISI. In [18] only the last few samples free from ISI in guard interval are utilized, but just as most CP methods are, there also exists a prerequisite assuming that guard interval should be greater than channel impulse response. This may lead to failure in synchronization under some serious channel conditions. A blind algorithm is proposed in [19], but it requires enough signal statistical information and a high SNR with higher system complexity. After all, the above algorithms have their advantages and disadvantages. Some has high complexity and others have poor efficiency. Some is suitable for continuous transmission mode [11,20] and the other is only suitable for burst mode [21]. The OFDM symbol timing estimation with windowing function or pulse shaping is presented in [22], such schemes will reduce the channel throughput. In this paper, our analysis and simulations on the basic OFDM structure such as defined in [1-2]. 3. PROPOSED COARSE SYMBOL TIMING SYNCHRONIZATION There is difference in the symbol timing synchronization technique between burst mode and continuous mode transmission systems. Usually, the former requires a fast acquisition time for synchronization compared with the latter one, in which the acquisition time is not so important and the feedback (FB) system is often adopted to deal with the more accurate symbol timing synchronization in frequency domain. Therefore, the time domain coarse symbol timing synchronization is much more important and should be more accurate in burst mode systems than that in continuous mode systems. The OFDM training structure in IEEE802.11(a) standard [4] for WLAN system is show in figure2. Where t 1 to t 10 are ten identical short OFDM symbols, T 1 and T 2 denote two identical long OFDM symbols, followed by the SIGNAL field and DATA. Four subcarriers are inserted as pilots into positions -21,-7,7 and 21 with 52 total subcarriers. The total training length is 16µs and the dashed boundaries in the figure denote repetitions due to the periodicity of the inverse Fourier transform. t 1 to t 7 are often used for signal detection, automatic gain control (AGC) and diversity selection, while t 8 to t 10 are utilized to do coarse frequency offset estimation and timing synchronization. Two long OFDM symbols are adopted to do channel and fine frequency offset estimation. Usually, pilot symbols embedded in such OFDM structures as [3] have good properties of autocorrelation, which can be used to do coarse symbol timing synchronization in time domain [23]. However, due to few pilot (only four pilots) and the lack of power lifting property in IEEE802.11(a) standard, it is not suitable to do symbol timing utilizing pilots. As a result, we must make full use of the preamble. 3.1 Conventional Method According to the maximum likelihood theory, the following equation is hold: ˆP N = max S ( 32, +32) ( 16 (10 N) i=1 x i+s x 16 N+i+S ) (2) Where ˆP N is the estimated coarse symbol timing position, N denotes the number of short OFDM symbols for correlation. x i+s is the received short OFDM symbol at the ith subcarrier, S is the sliding window, S ( 32, 31) is the estimated range for the symbol timing offset. Two long identical OFDM symbols or averaging over several OFDM symbols can be used to improve the estimation accuracy. Lots of simulations show that such method can get good performance under AWGN channel with high SNR. However, it has poor performance under multi-path fading channels due to its inherent defect. Simulation plot 3 presents the distinct peak value of the proposed method, while there exists the severe peak ambiguity in conventional correlation method. As is analyzed above, the ML method fails under multi-path fading channels due to the destroyed data in guard interval, and thus, limit its application area. Considering that the IEEE802.11(a) is set for the indoor fixed communication environment, the correlation method based on ML is enough, however, if we want to deal with outdoor environment and under severe multi-path fading channel such as those described in DVB-T[3] or IEEE802.16[24], the conven-

3 88 ECTI TRANSACTIONS ON ELECTRICAL ENG., ELECTRONICS, AND COMMUNICATIONS VOL.3, NO.2 AUGUST s s s s s s t 1 t t 2 3 t t 4 5 t 6 t t t t 10 GI2 T1 T2 GI SIGNAL GI DATA 1 GI DATA 2 Fig.2: IEEE802.11(a) OFDM training structure Fig.3: Comparison of the ML and the proposed method. N=16 samples in one short OFDM symbol, 64 samples in one long OFDM symbol, symbol timing offset:16 samples; frequency offset ( normalized to subcarrier spacing ) is 0.2; AWGN channel, SNR=15dB. Top figure is the conventional correlation method; bottom figure is the proposed method. tional method has poor performance and is not practical under these conditions. As we have mentioned before, N denotes the number of short OFDM symbols for correlation or convolution, which follows the IEEE standard. It is obvious that the longer duration of length N will provide more accurate estimation results. In fact, we can also design some new appended prefix or training symbols to improve the estimation accuracy, however, there is a tradeoff between size of prefix and the transmission efficiency, which should be taken into accout when in actual usage. 3.2 The Proposed Method To overcome problems the conventional method faced, we proposed a new method which has not only good property to deal with multi-path fading channel but also has fast symbol timing acquisition time. The algorithm is expressed in equation (3): ˆP = max (1, ) (Re(y n) Re(I n1,n2 )) 16 9 (3) Fig.4: Comparison of the ML and the proposed method. N=16 samples in one short OFDM symbol, 64 samples in one long OFDM symbol, symbol timing offset:16 samples; frequency offset ( normalized to subcarrier spacing ) is 0.2; Rayleigh fading channel with 20 paths[3]; SNR=5dB. Top figure is the conventional correlation method; bottom figure is the proposed method. y n is the received sequence in preamble, I n1,n2 is the transmitted preamble data belong to region (n1,n2) before passing through the wireless channel. In equation (3), it is assumed that there is 32 samples symbol timing offset, i.e. the estimation range is 32 2, as a result, our research range is from 1 to The term 16 9 is substracted because the convolution peak is in the 10 th short OFDM symbol, the first 9 short symbol length should be eliminated. Figure 5 and 6 are the performance simulation with variation of SNR under AWGN and Rayleigh fading channel environment respectively. It is not difficult to see from figure 6 that the ten short symbols or ten short symbols plus two long symbols can all make the final estimated deviation converging to zero. While the latter one has 1dB performance superiority over that only with ten short OFDM symbols. Compared with conventional correlation method, the estimated deviation come to convergence at zero at about over 10dB, while with the proposed method it can come to zero at -10dB which means an about 20dB supe-

4 A Novel Scheme for Symbol Timing in OFDM WLAN Systems 89 Fig.5: Estimated deviation for conventional method. N=16 samples in one short OFDM symbol, 64 samples in one long OFDM symbol, symbol timing offset:16 samples; frequency offset ( normalized to subcarrier spacing ) is 0.2;AWGN channel. riority. Also, the proposed method is simulated under multi-path fading channel, while the conventional method is simulated under AWGN channel. 4. CONCLUSION In this paper, we propose a new method to do coarse symbol timing synchronization with high accuracy and a fast acquisition time. Lots simulations show that it has excellent performance compared with the conventional ML method especially under multi-path fading channels and at low SNR. It has been verified that the method is also suitable for continuous mode transmission systems such as DVB-T. It is achieved with APEX TM 20 series FPGA chips produced by Altera company. The corresponding FPGA circuit with the proposed method through test in Broad-band digital broadcasting terrestrial TV (BDB-T) prototype in TEEG (Team of Engineering Expert Group ) in China proves good performance of the scheme. ACKNOWLEDGMENT The authors would like to express their thanks to the support of National Natural Science Funds in China ( and ). References [1] W.Y.Zou, and Y.Y.Wu, COFDM: an overview, IEEE Trans. On Broadcasting, 41(1):1-8, Mar [2] H.Sari, G.Karam, and I.Jeanclaude, Techniques for Digital Terrestrial TV Broadcasting, IEEE Fig.6: Estimated deviation for the proposed method. N=16 samples in one short OFDM symbol, 64 samples in one long OFDM symbol, symbol timing offset:16 samples; frequency offset ( normalized to subcarrier spacing ) is 0.2; Rayleigh fading channel with 20 paths[3]. Communications Magazine, 1995, 33(2), pp [3] European Telecommunication Standard ETS , Digital broadcasting systems for television, sound and data services; Framing structure, channel coding and modulation for digital terrestrial television, May [4] IEEE P802.11a/D7.0, DRAFT Suplement to STANDARD for Information Technology Telecommunications and information exchange between systems - Local and metropolitan area networks - Specific Requirements - Part II Wireless LAN Medium Access Control ( MAC ) and Physical Layer ( PHY )specifications: High Speed Physical Layer in the 5GHz Band, July [5] J.A.C. Bingham, Multi-carrier Modulation for Data Transmission: An Idea Whose Time Has Come, IEEE Communications Magazine, Mar 1990, vol. 28, pp [6] L.Wei, and C.Schlegel, Synchronization requirements for multi-user OFDM on satellite mobile and two-path Rayleigh-fading channels, IEEE Trans. on Commun., Feb/Mar/Apr. 1995, 43(2/3/4), pp [7] T.Pollet, M.van Bladel, and M.Moeneclaey, BER sensitivity of OFDM systems to carrier frequency offset and Wiener phase noise, IEEE Trans. on Commun., Feb/Mar/Apr. 1995, 43(2/3/4), pp [8] M.H.Hsieh, and C.H.Wei, A Low-Complexity Frame Synchronization and Frequency offset Compensation Scheme for OFDM Systems over Fading Channels, IEEE Trans. on Vehicular

5 90 ECTI TRANSACTIONS ON ELECTRICAL ENG., ELECTRONICS, AND COMMUNICATIONS VOL.3, NO.2 AUGUST 2005 Technology, Sept. 1999, 48(5), pp [9] P.H.Moose, A technique for orthogonal frequency-division multiplexing frequency offset correction, IEEE Trans. on Communications, Oct. 1994, 42(10), pp [10] F.Classen, and H.Meyr, Frequency synchronization algorithms for OFDM systems suitable for communication over frequency selective fading channels, In Proc. IEEE Vehic. Tech. Conf., vol.3, June 1994, Stockholm, Sweden, pp [11] T.M.Schmidl, and D.C.Cox, Robust frequency and timing synchronization for OFDM, IEEE Trans. on COM., Dec. 1997, 45(12), pp [12] W.D.Warner, and C.Leung, OFDM/FM frame synchronization for mobile radio data communication, IEEE Trans. on Vehic. Tech., Aug. 1997, 42(3), pp [13] F.Tufvesson, M.Faulkner, and P.Hoeher, etal, OFDM Time and Frequency synchronization by Spread Spectrum Pilot Technique, In Proc. of 8th IEEE Communication Theory Mini Conference in conjunction to ICC 99, June , Vancouver, Canada. [14] F.Daffara, and A.Chouly, Maximum-likelihood frequency detectors for orthogonal multi-carrier systems, In Proc. Int. Conf., May 1993, Geneva, Switzerland, pp [15] J.J.van de Beek, M.Sandell, and P.O.Börjesson, ML estimation of time and frequency offset in OFDM systems, IEEE Trans. on Signal Processing, July 1997, 45(7), pp [16] J.J.van de Beek, M.Sandell, and M.Isaksson etal, Low-complex frame synchronization in OFDM systems, In Proc. Int. Conf. Universal Personal Communications, Nov. 1995, Tokyo, Japan, pp [17] D.Landström, S.K.Wilson, and J.J.van de Beek etal, Symbol time offset estimation in coherent OFDM systems, In Proc. Int. Conf. On Communications, June 1999, Vancouver, BC, Canada, pp [18] M.Speth, F.Classen, and H.Meyr, Frame synchronization of OFDM systems in frequency selective fading channels, In Proc. Vehicular Tech. Conf., May 1997, Phoenix, AZ, pp [19] R.Negi, and J.M.Cioffi, Blind OFDM symbol synchronization in ISI channels, IEEE Trans. on Communication, Sept. 2002, 50(9), pp [20] B.G.Yang, K.B.Letaief, and S.Roger, Timing recovery for OFDM Transmission, IEEE Journal on Selected Areas in Communications, Nov. 2000, 18(22), pp [21] E.G.Larsson, G.Q.Liu, and J.Li etal, Joint Symbol Timing and Channel Estimation for OFDM Based WLANs, IEEE Communications Letters, Aug. 2001, 5(8), pp [22] J.M.Arenas, D.Landström, and J.J.van de Beek etal, Synchronization in OFDM systems- sensitivity to the choice of pulse shape, InProc.ofthe 16 th GRETSI Symposium on Signal and Image Processing, Sept , Grenable, France, pp [23] Bo AI, Jian-hua GE and Yong WANG, Symbol timing synchronization technique in COFDM systems, IEEE Trans. on Broadcasting. March 2004, 50(1): [24] IEEE Std TM. IEEE Standard for Local and metropolitan area networks. Part 16:Air Interface for Fixed Broadband Wireless Access Systems, IEEE Computer Society and the IEEE Microwave Theory and Techniques Society, 8th April, web security. Yong Wang (M 2003) was born in Shannxi Province in China in He received a B Sc., Master and Ph.D degree from Xidian University in China in 1997,2002 and 2005 respectively, and now working as an associate professor on communications in the Key Lab. of ISN in Xidian University. He has once participated in the key research project on HDTV in TEEG in China and his interests are wireless communications. Ge Jian-hua was born in JiangSu Province in China in He received the B Sc., Master and Ph.D degeree from Xidian University in 1982, 1985 and 1989 respectively. He is now the professor in both Xidian University and Shanghai JiaoTong University. He is the senior member of Chinese Electronics Institute. He has won lots of scientific and technical prizes in China. His interests are wireless communications and Bo Ai (M 2001) was born in Shannxi Province in China in He received a BSc. Degree from Engineering Institute of Armed Police Force in 1997, a Master and dr. degree from Xidian University in 2002 and 2004 in China respectively, and now working as a post dr. in dept. of E&E, state of key lab. on microwave and digital communications in Tsinghua University in China. His current interests are the research and applications of MIMO-OFDM technique with emphasis on synchronization and HPA linearization techniques.

6 A Novel Scheme for Symbol Timing in OFDM WLAN Systems 91 Li Zong-qiang was born in Shannxi Province in China in He received a B Sc. degree from Zhejiang University in China in 1995 and now working as a lecturer on communications in the College of Armed Police Force. His interests are broadband multimedia communications. Nie Yuan-fei received both the B.A. and M.S.E degree from Xidian University, xi an, China. He is currently working toward the Ph.D. degree in Communication and Information system in it. His research interests include OFDM, space-time signal processing and iterative decoding in serial concatenated systems.