CHANNEL ESTIMATION & ICI SUPPRESSION IN OFDM SYSTEMS

Abstract

The use of Orthogonal Frequency Division Multiplexing (OFDM) for high rate data transmission over fading dispersive channels has been of wide interest in recent years. OFDM, which already forms a part of several wireless broadcasting standards, is being viewed as a potential candidate for the design of upcoming 4G systems. When the intervening channel is doubly selective, the transmitted signal undergoes impairments due to multipath fading and Doppler spread. In order to facilitate the use of coherent modulation techniques, the receiver has to employ efficient channel estimation and equalization schemes. Relying on a block fading channel assumption, we propose a channel estimation scheme that makes use of cyclic prefix (CP) part of the OFDM symbols as a training sequence. CP serves the added task of channel estimation in the proposed system; although it is appended to ease the equalizer design in OFDM systems, and is normally discarded at the receiver. Channel estimation is carried out using Kalman filter algorithm and is compared against a class of adaptive filtering algorithms. The proposed method is shown to outperform variants of LMS and LS algorithms, which are commonly proposed for tracking timevarying systems. Comparison with conventional frequency domain pilot based method also illustrates the gain of the proposed method at low training overhead and computational complexity. The use of orthogonal frequency division multiplexing (OFDM) in frequency selective fading environments has been explored. The use of block fading channel assumption facilitates the use of OFDM in slowly varying environments. However, when the channel undergoes rapid time variations, the orthogonality between OFDM subcarriers is perturbed and inter-carrier interference (ICI) is introduced. In presence of time selective fading, OFDM is found to be more prone to errors as compared to single carrier systems. Besides this, obtaining reliable channel estimates in rapidly fading environments for receiver equalization is a non-trivial task. Channel estimation and equalization for OFDM are found to be the two major bottlenecks in applying in OFDM to systems with large Doppler spread. We have addressed the problem of channel estimationand ICI suppression by viewingthe system as a state space model. Kalman filter is employed to estimate the channel; this is followed by a time domain ICI mitigation filter which maximizes the SINR (signal to interference plus noise ratio) at the receiver. The method is seen to provide good tracking performance for moderate to high Doppler spreads at low training overhead. Apart from this, the performance is validated with suitable bounds on the performance of the system; the BER performance over a TIV Rayleigh fading channel serves as the lower bound, whereas the BER performance of the system over a double selective channel, with ICI as the dominant impairment forms the upper bound. Apart from this, the method provides significant SINR gain compared with a method that uses pilot based channel estimation followed by interpolation. As the order of channel mobility increases, the channel estimation techniques fail to keep up with the channel variations unless the channel is approximated with a simple model, (linearly varying or AR-1 model for instance), or a large amount of training is provided. The use of ICI mitigation schemes for effective equalization is limited by the need for reliable channel estimates at the receiver. The problem of channel estimation at high Doppler spread is addressed by relying on generalized complex exponential basis expansion model (GCE-BEM), which allows the modeling of each time varying channel tap in a deterministic manner, via a set of time-invariant (TIV) parameters and a set of complex exponential basis functions. Of the various kinds of BEM available in literature, the efficacy of GCE-BEM in approximating a wireless fading channel is demonstrated. Beginning with the development of a CPbased state-space model, the receiver carries out channel estimation with Kalman filtering, and equalizes the received OFDM symbols with a time-domain filter that maximizes the SINR at each received subcarrier. The method attempts to estimate the BEM parameters from the CP part of received symbols and is highly effective in tracking channels with a normalized Doppler spread up to 0.2. Simulations show

good MSEE and BER performance with low training overhead, when compared with a method based on conventional training. ICI forms the dominant impairment for the performance ofOFDM systems at high Doppler spreads. Even with perfect channel estimates available at the receiver, demodulation using FFT, which is simply matched filtering operation, introduces errors since the subcarriers are not orthogonal. Thus receiver equalization in doubly selective fading channels, in presence of ICI, is a non-trivial task. The conventional block equalizer designs in literature, which provide optimal performance in some sense (MMSE or ZF) incur a prohibitively large computational complexity. The number of floating-point operations FLOPs is usually a function of the square or cube of the number of data subcarriers in the OFDM symbol. The time-varying finite impulse response (TV-FIR) equalizer designs too involve a large computational overhead. This is a major concern for recent systems which study OFDM at large carrier frequencies with thousands of subcarriers. We address the problem of equalization by formulating a state space model. Kalman filter, which is known to yield the optimum solution to a linear filtering problem, is employed as an equalizer to estimate the unknown state of the system, which comprises the transmitted symbols. Besides this, the use of convolutional coding is explored. The proposed transceiver structure provides good performance at high Doppler spreads, which falls close to the theoretical performance over a TIV Rayleigh fading channel. Compared with other equalizers, the method provides a better performance, and incurs a considerably low computational complexity, which is a function of the number of discrete channel taps, and is only linear in the number ofOFDM subcarriers. The concept of Multiple Input Multiple Output (MIMO) systems allows improved communications by exploiting the multipath structure available between a set of transmitting and receiving antennas. The systems are not taxing in terms of the two basic communications resources, namely the transmit power and bandwidth; and this advantage comes at the cost of increased computational complexity and equipment. MIMO-OFDM systems have been studied recently as a means of providing high-rate data transmission over wireless channels. The use of space-time coding (STC) is known to achieve diversity and/or coding gain in MIMO configurations. However, the decoding and demodulation of STC-MIMO-OFDM requires channel knowledge at the receiver, unless differential modulation technique is employed. Semi-blind techniques like EM algorithm perform well but involve large computational complexity, when applied directly for MIMO channel estimation. We propose a receiver design for STC-MIMO-OFDM systems based on semi-blind channel estimation and maximum likelihood (ML) decoding of STC. Simplification of the conventional EM algorithm and its coupling with the ML decoder yields an algorithm, which converges within a few iterations and provides a performance comparable to the case of ideal channel knowledge. Simulation results show that the system tracks doubly selective channels with significant accuracy and exhibits good MSEE and BER performance at low to moderate Doppler spreads. Unlike space-time coded systems that encode data across transmit antennas over multiple symbols to exploit coding gain, spatial multiplexing MIMO systems transmit independent data streams from different antennas, and attempt to achieve higher data rates through diversity gain. MIMO-OFDM has been recently explored for providing high speed data communication over fading dispersive channels. Bit interleaved coded modulation (BICM) architecture with iterative detection and decoding techniques (IDD) at the receiver has been found to attain good performance (close to Turbo MIMO) with moderate computational complexity. The scheme however relies on the availability of perfect channel knowledge at the receiver, for VBLAST decoding algorithm as well as MMSE interference cancellation. We propose a semiblind channel estimation scheme for MIMO-OFDM systems based on VBLAST decoding and modified EM algorithm. The proposed VBLAST-EM channel estimator is coupled with MMSE based interference cancellation and demapping-decoding block for improved performance. The technique is seen to converge within a few iterations and provides a performance comparable to that of the ideal CSI case for 2x2 as well as 4x4 MIMO-OFDM systems.