A 2.69 Mbps/mW 1.09 Mbps/kGE Conjugate Gradient-based MMSE Detector for 64-QAM 128×8 Massive MIMO Systems

Guiqiang Peng,Leibo Liu,Qiushi Wei,Yao Wang,Shouyi Yin,Shaojun Wei
DOI: https://doi.org/10.1109/asscc.2018.8579260
2018-01-01
Abstract:This paper proposes a very-large-scale-integratíon (VLSI) architecture for an MMSE detector in an uplink 128×8 64-QAM massive MEMO system to achieve high energy and area efficiencies. A soft-output recursion conjugate gradient (RCG)-based minimum mean square error (MMSE) detector is fabricated onto a 3.5 mm 2 silicon with TSMC 65 nm CMOS technology for a 64-QAM 128×8 massive MEMO system. A processing element (PE) array and a user-level pipeline are introduced for this architecture to achieve high energy and area efficiencies. The chip achieves a 1.5 Gbps throughput under a 500 MHz working frequency while dissipating 557 mW at 1.2 V. The energy efficiency (throughput/power) and area efficiency (throughput/area) are 2.69 Mbps/mW and 1.09 Mbps/kGE, which are 2.39-to-2.47× and 1.15-to-8.81× those of the normalized state-of-the-art designs, respectively.
What problem does this paper attempt to address?