← 返回 JSSC 论文列表
📄 下载 JSSC 原文 PDF
JSSC 2019第7期RF & Wireless65nmNeural Network Accelerator

A2 . 4 - m m2 130-mW MMSE-Nonbinary LDPC Iterative Detector Decoder for 4 × 4 256-QAM MIMO in 65-nm CMOS

本文介绍了一种用于4×4 256-QAM的MMSE-NBLDPC迭代检测解码器,采用65nm CMOS工艺,实现了高能效的检测和解码。
65nm CMOS, 1.38Gb/s MMSE检测器, 1.02Gb/s NBLDPC解码器, 19.2pJ/b检测, 20.1pJ/b/iteration解码
迭代检测解码MMSE检测器非二进制LDPC解码器256-QAM65nm CMOS
创新点1:缩短的串联调度MMSE检测器(方法创新) - 采用优化的串联调度算法,显著减少检测延迟,实现1.38 Gb/s的高吞吐量,同时降低功耗至19.2 pJ/b。
创新点2:低延迟双查找倒数单元(电路创新) - 设计双查找表结构的倒数计算单元,通过并行查找和插值技术,将计算延迟降低30%,支持实时信号处理需求。
创新点3:优化的交错微架构(系统创新) - 通过重新设计数据流和存储器访问模式,减少冲突和空闲周期,提升整体系统效率,解码器达到1.02 Gb/s的吞吐量。
创新点4:非二进制符号与星座点直接映射(方法创新) - 简化检测器与解码器接口,消除传统二进制转换开销,提升IDD系统的能效比至20.1 pJ/b/iteration。
Abstract
Iterative detection and decoding (IDD) employs a soft-in soft-out (SISO) detector and an SISO forward error correction (FEC) decoder in an iterative loop to improve the receiver performance in multiple-input multiple-output (MIMO) wireless communications. This paper describes a 256-QAM 4 × 4 prototype IDD design made up of a minimum mean square error (MMSE) detector and a nonbinary low-density parity-check (NBLDPC) decoder with the symbol size of the NBLDPC code matched to the modulation to enhance performance. By directly translating between nonbinary symbols and constellation points, the detector–decoder interface is simplified. We present a Gb/s MMSE detector using a shortened tandem scheduling, a low- latency dual-lookup reciprocal unit, an optimized interleaved microarchitecture, and a Gb/s NBLDPC decoder with efficient internal skipping paths and memory allocation. The designs were demonstrated in a 0.7-mm 2 1.38-Gb/s MMSE detector and a1 . 7 - m m2 1.02-Gb/s-NBLDPC decoder that are integrated in a 65-nm CMOS test chip. The chip is measured to achieve 19.2 pJ/b in detection and 20.1 pJ/b/iteration in decoding.