Algorithms and Architectures of Energy-Efficient Error-Resilient MIMO Detectors for Memory-Dominated Wireless Communication Systems

摘要

In a broadband MIMO-OFDM wireless communication system, embedded buffering memories occupy a large portion of the chip area and a significant amount of power consumption. Due to process variations of advanced CMOS technologies, it becomes both challenging and costly to maintain perfectly functioning memories under all anticipated operating conditions. Thus, Voltage over Scaling (VoS) has emerged as a means to achieve energy efficient systems resulting in a tradeoff between energy efficiency and reliability. In this paper we present the algorithm and VLSI architecture of a novel error-resilient K-Best MIMO detector based on the combined distribution of channel noise and induced errors due to VoS. The simulation results show that, compared with a conventional MIMO detector design, the proposed algorithm provides up-to 4.5 dB gain to achieve the near-optimal Packet Error Rate (PER) performance in the 4 $times$ 4 64-QAM system. Furthermore, based on experimental results, when jointly considering the detector and memory power consumption, the proposed resilient scheme with VoS memory can achieve up to 32.64% savings compared to the conventional K-Best detector with perfect memory.