IEEE COMMUNICATIONS LETTERS, VOL. 4, NO.3, MARCH 2000 95 Frequency-Offset Synchronization and Channel Estimation for OFDM-Based Transmission Hyoung-Kyu Song, Young-Hwan You, Jong-Ho Paik, and Yong-Soo Cho Abstract—This letter is concerned with a frequency-offset esti- mation technique for OFDM-based transmission systems. The fre- quency estimation technique utilizes a repetitive signal structure inside of a OFDM symbol which is used to enlarge the range and increase the accuracy of offset estimation. Also, an averaged deci- sion-directed channel estimation (ADDCE) technique suitable for burst data is proposed. Index Terms—Channel estimation, frequency-offset estimation. I. INTRODUCTION O RTHOGONAL frequency division multiplexing (OFDM) is often used as a modulation scheme for high-speed wireless transmission, due to its simple receiver structure com- posed of frequency-domain one-tap equalizers. Furthermore, the intersymbol interference (ISI) in OFDM can be easily prevented by inserting a guard interval before each transmitted block. However, OFDM-based transmission systems are known to be sensitive to symbol and frequency synchronization errors, and nonlinear distortion [1], [2]. A number of frequency esti- mation methods have been developed for OFDM applications and a reliable synchronization is achieved by transmitting training symbols [3], [4]. In this letter, a multistage technique to acquire frequency offset synchronization for burst data is proposed and shown to have a good accuracy with a large estimation range. Also, an ADDCE technique for burst data is proposed to track time-variation of a wireless channel as well as to reduce noise effects at subchannels. II. FREQUENCY SYNCHRONIZATION One OFDM preamble, lasting over 4 OFDM symbol periods consists of 1 symbol period for silence, 2 symbol periods for symbol and frequency-offset synchronization, and 1 symbol period for equalization (or correction of symbol timing-offset when noncoherent modulation is used). As shown in Fig. 1, the OFDM preamble for the proposed frequency-offset estimator is composed of repetitive slots. The Manuscript received April 5, 1999. The associate editor coordinating the re- view of this letter and approving it for publication was Prof. N. C. Beaulieu. H.-K. Song was with System IC Research Center, Korea Electronics Tech- nology Institute (KETI), KyungGi-Do, 451-860 Korea. He is now with the De- partment of Informtion and Communications Engineering, Se-Jong University, Seoul, 143-150 Korea. Y.-H. You and J.-H. Paik are with System IC Research Center, Korea Elec- tronics Technology Institute (KETI), KyungGi-Do, 451-860 Korea (e-mail: yhyou@nuri.keti.re.kr). Y.-S. Cho is with the School of Electronical Engineering, Chung-Ang Uni- versity, Seoul, 156-756 Korea. Publisher Item Identifier S 1089-7798(00)01250-3. Fig. 1. Preamble structure for the multistage frequency-offset estimator when . OFDM preamble with periodic signals in the time domain can be generated in the frequency domain as follows otherwise (1) where is the training symbol at the th subcarrier for the fre- quency offset estimation, refers to a random symbol other than 0, is the number of samples over 1 OFDM symbol, and is inserted to prevent the signal power of the preamble from being decreased due to zero insertion. A. Multistage Approach In such a symbol structure, a number of base subblocks are bundled to obtain a new primary subblock such that two pri- mary subblocks of an samples come to exist in the 1 OFDM symbol. Further, an number of base subblocks are bundled to obtain a new second subblock such that four secondary sub- blocks come to exist in the 1 OFDM symbol. In this way, of base subblocks are bundled to obtain a number of th sub- blocks. Then, the frequency-offset for the th-stage is estimated by correlating the adjacent th-stage subblocks as follows: (2) The received signal of the OFDM preamble for frequency-offset estimation, , is given by (3) where is the transfer function of the channel at the fre- quency of the th carrier, is the relative frequency offset of the channel, and is the complex envelope of additive white Gaussian noise (AWGN). 1089–7798/00$10.00 © 2000 IEEE