Prototype Implementation of Real-Time ML Detectors for Spatial Multiplexing Transmission

SUMMARY We developed two types of practical maximum-likelihooddetectors (MLD) for multiple-input multiple-output (MIMO) systems, us-ing a field programmable gate array (FPGA) device. For implementations,we introduced two simplified metrics called a Manhattan metric and a cor-relation metric. Using the Manhattan metric, the detector needs no multi-plication operations, at the cost of a slight performance degradation within1dB. Using the correlation metric, the MIMO-MLD can significantly re-duce the complexity in both multiplications and additions without any per-formance degradation. This paper demonstrates the bit-error-rate perfor-mance of these MLD prototypes at a 1Gbps-order real-time processingspeed, through the use of an all-digital baseband 4 ×4 MIMO testbed inte-grated on the same FPGA chip. key words: multiple-input multiple-output (MIMO), spatial multiplex-ing, maximum-likelihood detection (MLD), field programmable gate array(FPGA), real-time MIMO detector 1. Introduction Recently, wireless communications systems that exploitmultiple transmitting and receiving antennas have receiveda significant amount of attention. These multiple-inputmultiple-output (MIMO) systems are expected to achievea dramatic increase in spectral efficiency because of theantenna diversity in scattering-rich wireless environments.Therefore, the MIMO system has been one of currentlypromising techniques to realizeGbps-class high-speed wire-less transmission for future communications systems [1].Many researchers have extensively investigated severaltechniques for multiple-antenna systems, such as the BellLaboratories layered space-time architecture (BLAST) [2]and space-time coding (STC) schemes [3].In MIMO spatial multiplexing systems, maximum-likelihood detection (MLD) schemes offer excellent perfor-mance [4]. Since the receiver estimates the most-likely sig-nals out of all the possible transmitting signals, the compu-tational complexity of distance metric calculations becomesextremely high in general. Consequently, many researchershave focused on developing more computationally efficientschemes. As a kind of low-complexity detectors, we canuse spatial filters based on the minimum mean-square error