On Receiver Algorithms for OFDM Systems

Doktorsavhandling, 2007

This thesis addresses the problem of designing a receiver for an Orthogonal FrequencyDivisionMultiplexing (OFDM)communication system operating over fading channels, taking into account the effects of hardware imperfections. We discuss solutions for efficiently estimating the parameters of a channel model and for compensating oscillator phase noise and analog to digital converter (ADC) quantisation errors at the receiver. The thesis consists of a general introduction to the topics of digital communication (OFDM, channelmodelling, estimation) and of threemain parts addressing the issues of phase noise compensation, channel estimation and ADC quantisation noise cancellation, respectively. In part I, after characterising the oscillator phase noise using aWiener process model, we analyse the effects of phase noise on an OFDM system operating over fading channels and provide an analytical bit error rate (BER) result. We then discuss the problem of estimating and reducing the effects of phase noise. Our solution is based on a two step iterative algorithm that estimates and removes the phase noise-induced common phase error (CPE) and intercarrier interference (ICI). The method is compared to existing algorithms and shows improvement at the cost of additional complexity. In part II, we propose a channel estimation algorithm for a coded multiband-OFDM system operating over the UWB channel. The algorithm is based on the classic Wiener filter-based MMSE channel estimator, which is extended to a decision directed iterative algorithm, taking into account the reliability of the information feedback. Different feedback methods with different orders of complexity are proposed and compared. The performance of the receiver algorithms are presented for the standardMB-OFDM UWB system which uses a punctured convolutional code as well as for a modified system which replaces the channel code for a rate and complexity equivalent turbo code. The results show that introducing the reliability measure in the estimation procedure improves the error rate performance of the system. Moreover, for an equivalent complexity, simulation showed an improvement of the performance using a turbo code over a convolutional code. In part III, we discuss the effect of ADC quantisation of the received signal for an OFDM system using a memoryless uniform quantiser. After providing an analysis of the effect of quantisation of an OFDM signal on the bit error rate and the channel capacity for AWGN and fading channels, we discuss a compensation algorithm. The algorithm is based on a decision feedback method with a feedback loop describing the quantisation error with a power-series type non-linearity. The algorithm is tested on a perfectly known channel as well as with a channel estimated with the algorithm in part II and proved to be able to effectively cancel the effects of quantisation on the received OFDM signal, as well as being robust to channel estimation error.