A Cost-Effective Adaptive Overlapped Cluster-Based MIMO Detector in a Frequency Domain Reconfigurable Modem

Abstract

Not only to reduce the candidates but also to maintain detection performance in multiple-input multiple-output (MIMO) detection, an adaptive overlapped cluster (AOC) scheme to balance detection error and computing cost is built for $4\times 4$ and $8\times 8$ MIMO-OFDM systems with up to 256 quadrature amplitude modulation (QAM). The constellations are partitioned into several clusters. A cluster with size decided by channel status is chosen for signal decoding. Different partition schemes are combined to minimize the numbers of clusters required to cover a candidate symbol as the pre-estimated signal falls at cluster edges, namely overlapped clustering. The simulations of a $4\times 4$ MIMO OFDM with 64 QAM and $8\times 8$ MIMO OFDM with 256 QAM hint that the AOC-based detection requires an additional 0.57 dB and 1.02 dB compared to maximum likelihood (ML). Compared with K-best sphere decoding (SD), it is reduced the computing complexity to 24.50% ~ 56.25% in $4\times 4$ MIMO OFDM and, 35.00% ~ 56.25% in $8\times 8$ MIMO OFDM. In addition, the proposed scheme is ported to a reconfigurable frequency-domain (FD) modem, which is designed and implemented via TSMC 45-nm technology, with multi-rate clocking and processing elements (PEs) upgrading for supporting the proposed MIMIO detection. The results show that the throughput is 1077.8 Mbps with $4\times 4\,\,64$ -QAM modulations.