Nonuniform Computational Complexity Allocation for OFDM Detectors

This paper concerns receiver design for multiple-input multiple-output orthogonal frequency-division multiplexing (MIMO-OFDM) communications systems, where frequencyselective fading causes the quality of the channel to vary significantly from one subcarrier to the next. It is common practice for MIMO-OFDM detectors to implement the same detector at each subcarrier, in which case the overall performance is dominated by the weakest subcarrier. In this paper we propose a receiver strategy for MIMO-OFDM channels called nonuniform computational complexity allocation (NCCA), whereby the receiver adapts the computational resources of the MIMO detector at each subcarrier to match a metric of the corresponding channel quality. For concreteness, we investigate this architecture in the special case when each subcarrier uses a list detector called B-Chase, where each list length is dependent on a channel signal-to-noise ratio. The performance of each B-Chase detector is compactly captured by the so-called list detection error probability. We compute an exact expression and upper bound for the list detection error probability in AWGN. We then propose an algorithm for assigning list lengths to subcarriers that aims to minimize the maximum list detection error probability over the different subcarriers. Numerical results for a 4-input 4-output Rayleigh-fading channel with 16-QAM show that the proposed MIMO-OFDM detector outperforms a conventional detector with the same complexity by 4.4 dB. Index Terms — MIMO-OFDM, list detection, list error probability, Chase detection, decision region.