Block-Wise Parallel Detection for OFDM with Fast Fading

The bit-error rate of the standard decoding algorithms for Orthogonal Frequency Division Multiplexing (OFDM) degrades when the channel is not stationary. Existing algorithms that decode OFDM signals in fast fading have significant computational cost and are inherently sequential which creates a large decoding latency even on systems with large processing capabilities. To reduce the decoding latency, and maintain high data rates, we introduce a new algorithm where blocks of the received OFDM symbol are detected in parallel. A number of iterations are performed to estimate and remove the interference between the blocks of the OFDM symbol. This algorithm gives error performance nearly as good as the best detection algorithms under fast fading, but it is also parallel in nature. We also show that the algorithm can be implemented in fixed-point arithmetic. This means that the algorithm we propose can be implemented on fixed-point parallel embedded processors, and thus a significant reduction of the decoding latency is indeed possible at a low implementation cost.