Parallel detection of MC-CDMA in fast fading

Prior work has developed a variation of MCCDMA, block-spread OFDM (BSOFDM), providing good error performance over stationary channels with low detection cost and allowing parallel computation to reduce the receiver latency. The previous detection algorithms developed for BSOFDM are not robust to fast fading. In this paper, it is demonstrated through the use of performance bounds that the performance degradation under fast fading is caused by limitations of the previous decoding algorithms, and is not an inherent limitation of BSOFDM. A novel iterative detection algorithm is introduced with intrinsic data-level parallelism for detecting BSOFDM in the presence of fast fading. Detection is first performed independently on subblocks of the received block vector. Information is exchanged between these parallel detectors in an iterative manner by estimating the interference between the blocks and removing it from the signal vector. It is shown that the computational complexity of this algorithm is not significantly higher than the prior detection algorithms for BSOFDM, and achieves excellent BER for fast fading with fixed-point arithmetic, making it suitable for use on embedded systems. In addition the inherent parallelism of this algorithm means multiple computational units can be exploited, if available, to reduce receiver latency.