Synchronization Issues in OFDM Systems

Hao Zhou, Amaresh V. Malipatil and Yih-Fang Huang

Department of Electrical Engineering University of Notre Dame Notre Dame, Indiana 46556 Email: {hzhou,amalipat,huang}@nd.edu
Abstract Orthogonal frequency division multiplexing (OFDM) is a bandwidth efcient modulation technique suitable for broadband wireless communications in frequency selective fading channels, yet, its performance is sensitive to timing and carrier synchronization errors. This paper provides an overview on the causes and the effects of these synchronization errors. We survey various symbol timing and carrier acquisition schemes available in the literature including coarse and ne synchronization techniques, using cyclic prex or training symbols. Blind estimators, which also do not require training symbols, are discussed as viable alternatives when data throughput is very critical. KeywordsOFDM, synchronization, carrier frequency offset, ISI, ICI

I. I NTRODUCTION Future wireless systems are expected to meet the high speed and high mobility requirements. Typical examples include those of high-speed internet, multimedia, and streaming video. The third generation (3G) systems being developed are designed to attain high data-rate service at 2Mb/s for xed users and 384kb/s for mobile users. The fourthgeneration (4G) systems will support even more demanding physical layer requirements, such as streaming video. The high data-rate requirement is a motivation to consider frequency-selective wideband wireless channels. Orthogonal frequency-division multiplexing (OFDM), one of the multicarrier modulation schemes, turns out to be a strong candidate for the future wideband wireless systems due to its high spectral efciency and simplicity in equalization. OFDM has been employed in various commercial applications that include wireless local area networks (WLAN/IEEE 802.11a/g/n and HIPERLAN/2), wireless metropolitan area networks (WMAN/WiMax, IEEE 802.16), terrestrial digital audio broadcasting (DAB) and terrestrial digital video broadcasting (DVB) systems. Its intrinsic usage of both time-division and frequency-division multiple access makes it also suitable for multi-user access. Furthermore, OFDM technique has been combined with code-division multiple access (CDMA) in multicarrier-CDMA systems [1]. However, OFDM also has its drawbacks. The most notable ones are high peak-to-average power ratio (PAPR) and sensitivity to synchronization errors. The PAPR of OFDM is proportional to the number of sub-carriers used in OFDM systems. A large PAPR makes the analog-to-digital converter (ADC) and digital-to-analog converter (DAC) implementation difcult and needs linear RF ampliers which are less power efcient. The synchronization problems of

OFDM include timing and frequency synchronizations. The synchronization errors come from two sources. One is the local oscillator frequency difference between transmitter and receiver, and the other is the Doppler spread due to the relative motion between the transmitter and the receiver. Both timing and frequency synchronizations introduce extra interference to OFDM systems and result in performance degradation. In addition, timing synchronization may affect the performance of channel estimation [2]. Therefore, to improve the performance of an OFDM system, effective synchronization is a key. The primary purpose of this paper is to address the synchronization issues for OFDM systems and offer some insights into the cause, effects and means of mitigating them. II. T IMING S YNCHRONIZATION Timing synchronization for OFDM involves two issues: symbol synchronization and sampling clock synchronization. Symbol synchronization in OFDM deals with the problem of detecting the start of an OFDM symbol. The symbol border gets blurred due to the multipath effect. The cyclic prex (CP) used in the OFDM signal slightly relaxes the timing synchronization requirement. The symbol timing denes the data window for the FFT processing window in the OFDM receiver, i.e., selects the set of samples to calculate the frequency domain symbol for each received OFDM symbol. The sampling clock offset, on the other hand, leads to the OFDM symbol block slowly drifting away in the receiver FFT window. The sampling clock offset is usually measured in parts per million (ppm). The drift of the OFDM symbol block due to sampling clock offset is a sustaining effect, which needs to be taken care of in the tracking mode. A. Effect of Timing Offset As mentioned above, symbol timing is aimed at detecting the start of the OFDM symbol. The detection becomes difcult when the channel experiences multipath fading. The cyclic prex, on the other hand, alleviates some of the burden of symbol timing. If the estimated symbol boundary falls within the cyclic prex, it does not impair the performance. Otherwise, the demodulated OFDM signal will be distorted by inter-symbol interference (ISI) and inter-carrier interference (ICI) [2], [4]. The ISI and ICI are usually modeled as additive Gaussian noise whose power is strongly dependent on the channel prole. They cause system performance to

988 1-4244-0387-1/06/$20.00 c 2006 IEEE

degrade. Apart from the interference noise effect, timing offset affects channel estimation too [2]. Channel estimation is performed in a time-domain estimation window and the estimated channel impulse response (CIR) is required to be within this estimation window to produce an unbiased channel estimate. The timing offset shifts the location of the CIR along the time axis and degrades the channel estimation once some portions of CIR are shifted outside the channel estimation window. Generally, the closer the window is matched to the CIR, the better the resulting channel estimate would be. The channel estimation performance affects the system performance directly, therefore, accurate timing is critical to good system performance [2]. Sampling frequency synchronization is another timing issue due to sampling clock offset. The sampling clock offset leads to symbol timing drift and has a long-term timing effect in the tracking mode. Furthermore, sampling frequency offset is also considered together with the carrier frequency offset, because they have the combined effect on the subcarrier phase rotation and the ICI [2], [5]. B. Existing Techniques for Timing Synchronization Timing synchronization is usually divided into a coarse timing phase and a ne timing phase. Different systems would have different requirements for symbol timing. For example, WLANs can only use the preambles whereas broadcast systems can employ many symbols to gradually acquire more accurate symbol timing. 1) Coarse Timing: Coarse timing is usually the rst task performed in the OFDM synchronization procedure. During the timing acquisition phase, it has to be assumed (in practice) that the carrier frequency offset is completely unknown. Hence the orthogonality of the subcarriers may not be retained to provide a useful post-FFT signal. Consequently, coarse timing is employed in time domain. Depending on different system requirements, the coarse timing can be carried out by utilizing either the OFDM inherent cyclic prex [3], [4], [8], [9] or the specially designed training sequence (preamble) [10], [16]. By utilizing the cyclic prex, the algorithms can work blindly without the overhead of a known training sequence, whereas employment of a training sequence can usually result in a faster and more robust synchronization. For both the cyclicprex and the training-sequence-based schemes, there are basically three methods for coarse timing, i.e. the maximumlikelihood (ML) [8], minimum mean squared error (MMSE) [4] and the maximum correlation (MC) [3][7] methods. We consider the training-sequence based synchronization cases in the following. The MC method is designed to maximize the correlation metric given by
L1

time index as subscript. The MMSE method can be expressed as maximizing V (d) 1 = |C(d)| P (d) 2
L1

(2)

where C(d) is the same as in (1) and P (d) = (|rd+m |2 + |rd+m+L |2 ) (3)

m=0

The ML method is to maximize Q(d) = |C(d)| P (d) 2 (4)

with the same C(d) and P (d) as dened above and =

E{rd rd+L }

E{|rd |2 }E{|rd+L |2 }

SN R SN R + 1

(5)

Besides, Schmidl and Cox [10] proposed an estimator that maximizes M (d) with
L1

|C(d)|2 (R(d))2

(6)

R(d) =
m=0

|rd+m+L |2

(7)

C(d)

(rd+m rd+m+L )

(1)

m=0

There are many variations made with regard to the above schemes. Lee [9] derived an ML coarse symbol synchronization algorithm by considering the channel statistical information of multipath channels. Lin [15] proposed an ML timing scheme for fast rayleigh-fading channels. Minn [16], considering the false detection under low SNR, proposed the normalized metric for the MMSE scheme, and optimized the training symbol pattern design. 2) Fine Timing: Fine timing is usually the last step in the timing acquisition process. Fine timing can be achieved through correlation either in the time domain [12], [13] or in the frequency domain [4], [6], [14]. Tufvesson [12] proposed a timing scheme based on time-domain PN-sequence correlation which shows better performance than repeated OFDM symbols. Landstrom [13] presented a weighted time correlation scheme exploiting the redundancy in both the cyclic prex and available pilot symbols. Frequency-domain based schemes, [4], [6], [14], rst estimate the channel impulse response after the FFT, then ne-tune the position of the estimated CIR with respect to the FFT-window. Because the frequency-domain based schemes can effectively resolve channel multipaths, they generally render better performance than the time-domain schemes in multipath fading channels. For the frequency-domain based schemes, there are different approaches to tracking the estimated CIR in the FFTwindow. One method is to locate the peak position of the estimated CIR {hd } [6], [14] d = arg max{|hd |}
d

where L is the length of the repeated pattern in the training symbol and r is the received time-domain sequence with a

(8)

APCCAS 2006

989

There are many modied peak-nding versions when the noise effect is considered and when the rst peak is not the largest peak in the CIR [16]. The other method is the maximum energy windowing method proposed in [3], [4], which can be expressed as, d =
d+W

channels. In [22], the exact SNR degradation caused by CFO is determined by deriving the ICI power. B. Estimation of CFO Similar to timing synchronization, CFO offset estimation is usually done in two phases, i.e., initial frequency acquisition or coarse synchronization and then ne frequency tuning. 1) Coarse carrier synchronization: In coarse synchronization, the usual approach is to bring down the CFO estimation error ( ) to within one-half of the subcarrier spacing, i.e., | | < 1/2. As in [3], the acquisition stage can be performed through two stages, i.e., pre-FFT and postFFT. The pre-FFT stage, like coarse timing part, uses the received time-domain signal to acquire the fractional carrier frequency offset. The pre-FFT CFO estimate comes along with the coarse timing estimate with the metric as = (C(d))/(T ) (11)

arg max{
d k=d

|hd | }
2

(9)

where W is the gliding window length. The author proposed a ne timing tracking scheme based on the channel statistical information [18], with the timing metric, d = arg min{h(d)H C1 h(d)} n
d

(10)

where h(d) is the d-shifted estimated CIR vector and Cn is the channel autocorrlation matrix (including the noise effect). The proposed scheme has been shown to reduce the tracking error variance signicantly under time-varying multipath fading channels. 3) Sampling Clock Offset Compensation: Sampling clock offset can be taken care of in two ways. One method is by tracking the symbol timing in the ne timing procedure, the sampling frequency can be adjusted accordingly [3], [6], [16], [18]. The other way is to be considered together with the carrier frequency offset, using the successive OFDM symbol phase rotation to estimate the sampling frequency offset as well as the carrier frequency offset [2], [5]. III. C ARRIER F REQUENCY S YNCHRONIZATION The baseband OFDM signal is upconverted using a radiofrequency (RF) carrier at the transmitter and the receiver downconverts it back to the baseband for demodulation. The carrier frequencies at the transmitter and the receiver are usually not exactly matched because of various reasons like temperature changes, oscillator frequency drifts, tolerance of RF components, etc. This frequency offset destroys the orthogonality of the subcarriers leading to inter-carrier interference (ICI). Moreover, in mobile environments, Doppler shift due to the relative motion between transmitter and receiver, also introduces additional CFO and compounds the problem of ICI. The ICI results in an irreducible noise oor which limits the performance of the OFDM system. Hence it is vital that the CFO be estimated and compensated at the receiver. In this paper, we will study the effect of CFO on the OFDM system and discuss various methods to estimate and compensate the CFO. A. Effect of CFO The effect of CFO on OFDM system performance has been well studied by many authors [20], [21], [22]. The CFO has three major effects on the received OFDM signal, i.e., decrease in desired signal power, phase rotation of the desired signal and introduction of ICI. In [20], the authors derive the BER degradation caused by CFO in an AWGN channel. In [21], Pollet et al. provide approximate expressions for BER in both AWGN and Rayleigh fading

where C(.) is the correlation metric (1), d is the coarse timing estimate and T is the OFDM effective symbol period (excluding cyclic prex). The post-FFT section, on the other hand, is used for fast acquisition of the integer carrier frequency offset. With the aid of some knowledge of frequency domain data symbol (such as pilot), post-FFT signal can be used to locate the correct subcarrier position quickly. 2) Fine carrier synchronization: After initial acquisition, training symbols or pilots subcarriers inserted between data subcarriers can be used to estimate or track the residual CFO. In [24], Moose derived an ML estimate for the CFO using two successive repeated training symbols. The CFO is embedded in the phase difference between samples which are N samples apart (N is the FFT size). The estimator, like most ne acquisition schemes, assumes that the normalized CFO | | < 1/2. The ML estimator is conditionally unbiased for small frequency offsets and high SNRs. Also, the estimation error variance decreases linearly with SNR. The MLE is consistent since the error variance is inversely proportional to the number of subcarriers. In [16], Minn et al. use a carefully designed training symbol that aids in coarse and ne carrier synchronization along with the symbol timing. In [19], the authors have proposed using time-domain channel estimates to reduce the complexity of ML estimation of CFO without performance degradation. Also the scheme offers a novel approach to estimating the CFO in fast time-varying channels with the knowledge of second-order statistics of the fading process. As mentioned before, the sampling clock frequency offset can also be estimated in conjunction with the CFO. In [5], the authors use the frequency domain channel estimate to deduce the weights for a weighted-least-squares sampling clock offset and CFO estimator. Likewise, in [3], the residual carrier and sampling frequency offset are tracked using the temporal correlation in the data-aided post-FFT stage.

990

APCCAS 2006

IV. B LIND TIMING AND CARRIER SYNCHRONIZATION Blind estimation algorithms exploit known statistical properties and inherent structure of the signal to estimate the parameters. In [26], Tureli et al. use the inherent orthogonality among OFDM subcarriers and propose two estimators, i.e., MUSIC-like approach and ESPIRIT-like approach. The MUSIC-like estimator estimates the CFO by forming a cost function based on the orthogonal subsapce obtained from subspace decomposition of the observations. The MUSIClike approach has better accuracy and requires about 4 data symbols to attain its best accuracy but it is computationally intensive. In [28], Blcskei exploits the cyclostationarity of o OFDM signals to estimate both symbol timing and CFO in pulse shaped OFDM. The estimators are asymptotically unbiased and need only the second order statistics and do not require CP or pilots. In [29], Negi et al. employ the second order statistics and rank minimization to derive a blind symbol timing estimate. V. C ONCLUSION In this paper, we have discussed the effect of timing and frequency synchronization errors on the performance of OFDM system. Symbol timing, CFO and sampling frequency offset are critical issues, which, if uncompensated for, will result in ISI, ICI, reduced signal power, SNR degradation and ultimately signicantly deteriorated system performance. We surveyed various existing techniques for coarse and ne synchronization. Cyclic prex based synchronization schemes can be applied without training sequence, and training-sequence-based schemes can achieve faster and more robust performance. In bursty communications like WLAN, coarse synchronization may be sufcient. On the other hand, in continuous transmission like DVB, ne synchronization and tracking is necessary to track and correct the residual errors. Blind estimation algorithms that avoid the overhead of training symbols were also briey discussed. R EFERENCES
[1] L. Hanzo, M. Munster, B. J. Choi and T. Keller, OFDM and MC-CDMA for Broadband Multi-User Communications, WLANs and Broadcasting, IEEE Press, 2003. [2] M. Speth, S. Fechtel, G. Fock, and H. Meyr, Optimum Receiver Design for Wireless Broad-Band Systems Using OFDM - Part I, in IEEE Trans. on Comm., vol. 47, no. 11, pp. 1668 - 1677, Nov. 1999. [3] M. Speth, S. Fechtel, G. Fock, and H. Meyr, Optimum Receiver Design for OFDM-Based Broadband Transmission - Part II: A Case Study, in IEEE Trans. on Comm., vol. 49, no. 4, pp. 571 - 578, April. 2001. [4] M. Speth, F. Classen, and H. Meyr, Frame synchronization of OFDM systems in frequency selective fading channels, in Proc. VTC, pp. 1807 - 1811, 1997. [5] P. Y. Tsai, H. Y. Kang and T. D. Chiueh, Joint Weighted Least-Squares Estimation of Carrier-Frequency Offset and Timing Offset for OFDM Systems Over Multipath Fading Channels, in IEEE Trans. on Veh. Technol., vol. 54, no. 1, pp. 211-223, Jan 2005. [6] B. Yang, K. B. Letaief, R. S. Cheng, and Z. Cao, Timing Recovery for OFDM Transmission, in IEEE Journal on Selected Areas in Communications, vol. 18, no. 11, pp. 2278 - 2291, Nov. 2000. [7] T. Keller, L. Piazzo, P. Mandarini, and L. Hanzo, Orthogonal Frequency Division Multiplex Synchronization Techniques for FrequencySelective Fading Channels, in IEEE Journal on Selected Areas in Communications, vol. 19, no. 6, pp. 999 - 1008, June. 2001.

[8] J. J. van de Beek, M. Sandell, and P. O. Borjesson, ML estimation of timing and frequency offset in OFDM systems, in IEEE Trans. Signal Processing, vol. 45, no. 6, pp. 1800 - 1805, July. 1997. [9] D. Lee and K. Cheun, Coarse Symbol Synchronization Algorithms for OFDM Systems in Multipath Channels, IEEE Communication Letters, vol. 6, no. 10, pp. 446 - 448, Oct. 2002. [10] T. M. Schmidl and D. C. Cox, Robust Frequency and Timing Synchronization for OFDM, IEEE Transactions on Communications, vol. 45, no. 12, pp. 1613 - 1621, Dec. 1997. [11] H. Minn, M. Zeng and V. K. Bhargava, On Timing Offset Estimation for OFDM Systems, in IEEE Communications Letters, vol. 4, no. 7, pp. 242 - 244, July. 2000. [12] F. Tufvesson, O. Edfors and M. Faulkner, Time and frequency synchronization for OFDM using PN-sequence preambles, in VTC 1999, vol. 4, pp. 2203 - 2207, Sept. 1999. [13] D. Landstrom, S. K. Wilson, J-J. Beek, P. Odling, and P. O. Borjesson, Symbol Time Offset Estimation in Coherent OFDM Systems, in IEEE Trans. on Communications, vol. 50, no. 4, pp. 545 - 549, April. 2002. [14] N. Chen, M. Tanaka and R. Heaton, OFDM Timing Synchronisation Under Multi-path Channels, in IEEE VTC, vol. 1, pp. 378 - 382, April. 2003. [15] J. Lin, Maximum-Likelihood Frame Timing Instant and Frequency Offset Estimation for OFDM Communication Over a Fast RayleighFading Channel, in IEEE Trans. on Vehicular Technology., vol. 52, no. 4, pp. 1049 - 1062, July. 2003. [16] H. Minn, V. K. Bhargava, and K. B. Letaief, A Robust Timing and Frequency Synchronization for OFDM Systems, in IEEE Trans. on Wireless Comm., vol. 2, no. 4, pp. 822 - 839, July. 2003. [17] A. J. Coulson, Maximum likelihood synchronization for OFDM using a pilot symbol: algorithms, in IEEE Journal on Selected Areas in Communications., vol. 19, no. 12, pp. 2486 - 2494, Dec. 2001. [18] H. Zhou and Y. F. Huang, A maximum likelihood ne timing estimation for wireless OFDM systems, IEEE 6th workshop on SPAWC, pp. 705 - 709, June. 2005. [19] H. Zhou, A. V. Malipatil and Y. F. Huang, Maximum-Likelihood Carrier Frequency Offset Esimation for OFDM Systems in Fading Channels, IEEE WCNC, Apr 2006. [20] Pollet, T.; Van Bladel, M.; Moeneclaey, M., BER Sensitivity of OFDM Systems to Carrier Frequency Offset and Wiener Phase Noise, IEEE Transactions on Communications., vol. 43, no. 2/3/4, pp. 191 193, Feb-Mar-April 1995. [21] Ma, X.; Kobayashi, H.; Schwartz, S.C, Effect of frequency offset on BER of OFDM and single carrier systems, in IEEE International Conference on Communications, vol. 3, pp. 22-27, June 1996. [22] Jungwon Lee, Hui-Ling Lou, D. Toumpakaris, J. M. Ciof, Effect of carrier frequency offset on OFDM systems for multipath fading channels, in Proc. IEEE GLOBECOMM04, vol. 6, pp. 3721-3725, Nov 2004. [23] Schmidl, T.M.; Cox, D.C., Low-Overhead, Low-Complexity [Burst] Synchronization for OFDM, in IEEE International Conference on Communications, vol. 3, pp. 22-27, June 1996. [24] Paul H. Moose, A Technique for Orthogonal Frequency Division Multiplexing Frequency Offset Correction,, IEEE Trans. Comm.,, vol. 42, no. 10, pp. 2908-2914, Oct. 1994. [25] P. Y. Tsai, H. Y. Kang and T. D. Chiueh, Joint Weighted LeastSquares Estimation of Carrier-Frequency Offset and Timing Offset for OFDM Systems Over Multipath Fading Channels, in IEEE Trans. on Veh. Technol., vol. 54, no. 1, pp. 211-223, Jan 2005. [26] Liu, H.; Tureli, U., A high-efciency carrier estimator for OFDM communications, in IEEE Communication Letters, vol. 2, no. 4, pp. 104-106, April 1998. [27] Zhang Z.; Long K.; Zhao M.; and Liu Y., Joint Frame Synchronization and Frequency Offset Estimationin OFDM Systems,, IEEE Trans. on Broadcasting,, vol. 51, no. 3, pp. 389-394, Sept. 2005. [28] H. Bolcskei, Blind estimation of symbol timing and carrier frequency offset in wireless OFDM system,, IEEE Trans. on Commun.,, vol. 49, pp. 988-999, June 2001. [29] R. Negi and J. M. Ciof, Blind OFDM Symbol Synchronization in ISI Channels,, IEEE Trans. on Commun.,, vol. 50, no. 9, pp. 15251534, Sept 2002.

APCCAS 2006

991