V. K. Poulkov is with the Departmenyt of Telecommunications, Technical University of Sofia, Sofia 1000, Bulgaria (e-mail: vkp@tu-sofia.bg). G. L. Iliev is with the Departmenyt of Telecommunications, Technical University of Sofia, Sofia 1000, Bulgaria (e-mail: gli@tu-sofia.bg). Abstract—This paper proposes a specific approach to channel equalization for Orthogonal Frequency Division Multiplex (OFDM) systems. Inserting an equalizer realized as an adaptive system before the FFT processing, the influence of variable delay and multipath could be mitigated in order to remove or reduce considerably the guard interval and to gain some spectral efficiency. The adaptive algorithm is based on adaptive filtering with averaging (AFA) for parameter update. Based on the development of a model of the OFDM system, through extensive computer simulations, we investigate the performance of the channel equalized system. The results show much higher convergence and adaptation rate compared to one of the most frequently used algorithms - Least Mean Squares (LMS). Index Terms—Adaptive algorithms and filters, channel equalization, orthogonal frequency division multiplexing. I. BACKGROUND FDM is widely accepted as an attractive transmission technique for different types of broadband transmission systems. It is also a very promising technique for delivering high data rate multimedia services over the mobile radio channel. The performance of such systems is limited by the severe effects of multiple delay spread. In order to overcome this problem a guard interval (cyclic prefix/suffix) is added to each OFDM symbol, to efficiently combat the multi-path effect, at the price of some loss in spectral efficiency, depending on the ratio of unextended and extended symbol periods. Other techniques are also used to combat impairments in time and frequency varying radio channels to obtain high spectral efficiencies in cellular systems, such as channel coding and interleaving, adaptive modulation, equalization, spectrum spreading, dynamic channel allocation etc. For channel equalization in OFDM there are two general approaches. First one is to apply an equalizer after the FFT block [1]. Second option is to make equalization before the FFT processing [2]. Our design uses the second approach. II. METHOD Channel estimation in coherent OFDM can be performed by inserting pilots into the two-dimensional time-frequency lattice, since the mobile channel can be viewed as a two-dimensional stochastic signal sampled at scattered pilot positions, where a noisy sample is obtained. To be able to interpolate channel estimates both in time and frequency, the pilot spacing has to fulfill the Nyquist sampling theorem. In the design of channel estimators for OFDM, pilot information must be distributed optimally in the time-frequency grid. For the 2D pilot patterns a trade-off has to be found; pilots have to be placed close enough to guarantee reliable estimation of the channel frequency response and at the same time, pilot density must be kept as low as possible to avoid reduction of the data rate. For our case the minimum pilot spacing in time and frequency is determined, having in mind some expected bandwidth of the channel variation in time and frequency. We use a cyclic pilot pattern to ensure reliable estimation of the channel in time and frequency. To shorten the delay before the first channel estimates can be calculated, which is undesirable in packet transmission, we use a preamble of symbols for initial training. Then the inserted pilots within the data symbols, besides channel estimation, are used for tracking the remaining offset after the initial training [3,4]. The main processing blocks are presented in Fig. 1. Concerning channel equalization, we use the approach of equalization before the FFT processing. Using this arrangement it is possible to realize the equalizer as an adaptive system with relatively simple structure. Concerning the adaptive algorithm, we employ a recently developed in [5] method based on adaptive filtering with averaging for parameter update. Comparison with one of the most frequently used algorithms – Least Mean Squares (LMS) [6,7] shows that the present design has three very attractive features: high adaptation rate, relatively low computational complexity and robustness in fixed-point implementations. The equalizer works in two modes. First, in training mode for channel estimation. Estimation error is defined as the difference between the estimate and the original signal available through the pilot signals. Second, in decision direction fashion, when the estimation error is determined as the difference between the estimate and the detected data symbols at the decision device output. Channel Equalization for OFDM Vladimir Poulkov and Georgi Iliev O