一种高性能极化码SC译码器设计

doi:10.16180/j.cnki.issn1007-7820.2022.08.003

摘要/Abstract

摘要：

针对极化码SC译码器存在的高延时、低吞吐率、低资源效率等问题,文中提出了一种高性能SC译码器硬件架构。通过剪枝冻结比特结点的方式化简SC译码二叉树,设计跨周期的PE单元存储模块,并在译码最后一个阶段利用2b-SC算法,保证译码器具有较低的延时和较高的吞吐率。采用资源复用的方法,提高译码器资源效率。测试结果表明,文中所提出的译码器周期为330,吞吐率为388.85 Mbit·s^-1,资源效率为2.204 Mbit·s^-1·kGE^-1。与其他SC译码器的对比试验表明,该高性能SC译码器的延时、吞吐率、资源效率均得到了有效改善。此外,该译码器的功耗较低,应用前景良好。

关键词: 极化码, 串行抵消, 延时, 功耗, 吞吐率, 资源效率, 资源复用, 专用集成电路

Abstract:

In view of high latency, low throughput and low area efficiency of polar code SC decoder, a high-performance hardware architecture of SC decoder is proposed. The decoder becomes low-latency and high-throughput by pruning frozen bit nodes to simplify the SC decoding binary tree, designing cross-cycle storage for PE, and using 2b-SC algorithm in the last stage. The resource-reused method is adopted to increase the decoder area efficiency. The testing results show that the cycle of the proposed decoder is 330, the throughput is 388.85 Mbit·s^-1, and the area efficiency is 2.204 Mbit·s^-1·kGE^-1. Compared with other SC decoders, latency, throughput and area efficiency of the high-performance SC decoder proposed in this study are significantly improved. Additionally, the decoder has lower power consumption and broad application prospect.

Key words: polar code, successive cancellation, latency, power consumption, throughput, area efficiency, resource reuse, ASIC

中图分类号:

TN47

王晓蕾,戴吴骏,杜高明,李桢旻,张多利. 一种高性能极化码SC译码器设计[J]. 电子科技, 2022, 35(8): 14-20.

WANG Xiaolei,DAI Wujun,DU Gaoming,LI Zhenmin,ZHANG Duoli. Design of a High-Performance SC Decoder for Polar Codes[J]. Electronic Science and Technology, 2022, 35(8): 14-20.

图/表 15

图1

图2

图3

图4

图5

图6

图7

图8

图9

图10

图11

表1

表2

表3

表4

参考文献 19

[1]	Arikan E. Channel polarization:A method for constructing capacity-achieving codes[C]. Toronto:Proceedings of IEEE International Symposium on Information Theory, 2008.
[2]	Arikan E. Channel polarization:A method for constructing capacity-achieving codes for symmetric binary-input memoryless channels[J]. IEEE Transactions on Information Theory, 2009, 55(7):3051-3073. doi: 10.1109/TIT.2009.2021379
[3]	妙琳, 彭亚雄, 陆安江. RA码与网络编码的联合设计[J]. 电子科技, 2020, 33(10):40-44.
	Miao Lin, Peng Yaxiong, Lu Anjiang. Joint design scheme of RA code and network coding[J]. Electronic Science and Technology, 2020, 33(10):40-44.
[4]	Gao J, Liu R. Neural network aided SC decoder for polar codes[C]. Chengdu:Proceedings of IEEE the Fourth International Conference on Computer and Communications, 2018.
[5]	Zhang R, Liu F, Zeng Z, et al. Neural network based successive cancellation decoding algorithm for polar codes in URLLC[C]. Oulu:Proceedings of the Sixteenth International Symposium on Wireless Communication Systems, 2019.
[6]	窦征立, 王亚刚. 基于Elman神经网络的联合循环机组燃烧室温度模型建模[J]. 电子科技, 2021, 34(3):60-64.
	Dou Zhengli, Wang Yagang. Modeling of combustion chamber temperature model of combined cycle unit based on Elman neural network[J]. Electronic Science and Technology, 2021, 34(3):60-64.
[7]	Feng B, Liu R. Efficient-memory and low-latency BP decoding algorithm for polar codes[J]. IEEE Communications Letters, 2020, 24(6):1236-1239. doi: 10.1109/LCOMM.2020.2982643
[8]	Khoshnevis H, Marsland I, Yanikomeroglu H. Throughput-based design for polar-coded modulation[J]. IEEE Transactions on Communications, 2019, 67(3):1770-1782. doi: 10.1109/TCOMM.2018.2880204
[9]	Shih X Y, Huang P C, Chen Y C. High-speed low-area-cost VLSI design of polar codes encoder architecture using radix-k processing engines[C]. Kyoto:IEEE the Fifth Global Conference on Consumer Electronics, 2016.
[10]	林青. 极化码编译码算法研究及VLSI设计[D]. 合肥: 合肥工业大学, 2020.
	Lin Qing. Research on encoding and decoding algorithm of polar codes and VLSI design[D]. Hefei: Hefei University of Technology, 2020.
[11]	刘建航, 何怡静, 李世宝, 等. 基于预译码的极化码最大似然简化连续消除译码算法[J]. 电子与信息学报, 2019, 41(4):959-966.
	Liu Jianhang, He Yijing, Li Shibao, el al. Pre-decoding based maximum-likelihood simplified successive-cancellation decoding of polar codes[J]. Journal of Electronics & Information Technology, 2019, 41(4):959-966.
[12]	Shrestha R, Bansal P, Srinivasan S. High-throughput andhigh-speed polar-decoder VLSI-architecture for 5Gnew radio[C]. Delhi:The Thirty-second International Conference on VLSI Design and the Eighteenth International Conference on Embedded Systems, 2019.
[13]	Li Q, Niu K, Dong C. A high throughput pipeline HARQ scheme of polar codes[C]. Shenzhen:Proceedings of the Nineteenth International Symposium on Wireless Personal Multimedia Communications, 2016.
[14]	Tan W, Wang A, Xu Y, et al. Area-efficient pipelined VLSI architecture for polar decoder[C]. Limassol:Proceedings of IEEE Computer Society Annual Symposium on VLSI, 2020.
[15]	Mishra A, Raymond A J, Amaru L G, et al. A successive cancellation decoder ASIC for a 1 024 bit polar code in 180 nm CMOS[C]. Kobe:Proceedings of IEEE Asian Solid State Circuits Conference, 2012.
[16]	Shrestha R, Sahoo A. High-speed and hardware-efficient successive cancellation polar-decoder[J]. IEEE Transactions on Circuits and Systems II: Express Briefs, 2019, 66(7):1144-1148. doi: 10.1109/TCSII.2018.2877140
[17]	Yoon H Y, Kim T H. Efficient successive-cancellation polar decoder based on redundant LLR representation[J]. IEEE Transactions on Circuits and Systems II: Express Briefs, 2018, 65(12):1944-1948. doi: 10.1109/TCSII.2018.2811378
[18]	Kalluru H S, Abbas Z. Optimal power-area polar decoder design based on iterative decomposition technique[C]. Rajkot:Proceedings of IEEE the Sixteenth India Council International Conference, 2019.
[19]	李山山, 乔树山, 李智. 一种读写分离结构的SRAM译码器设计[J]. 电子设计工程, 2019, 27(14):6-9.
	Li Shanshan, Qiao Shushan, Li Zhi. A read-write separated decoder design for SRAM[J]. Electronic Design Engineering, 2019, 27(14):6-9.

冻结信号		使能信号		输入		输出
fr₁	fr₂	en₁	en₂	LLR(a)	LLR(b)	u_odd	u_even
0	0	1	0	MSB(a)MSB(b)=1		0	MSB(a)
0	0	1	0	MSB(a)MSB(b)≠0		1	MSB(b)
1	0	0	1	LLR(a)+LLR(b)<0		0	1
1	0	0	1	LLR(a)+LLR(b)≥0		0	0
x	1	0	0	x	x	0	0

数据类型	结果
工艺/nm	40
频率/MHz	250.6
效率/ Mbit·s^-1·kGE^-1	2.204
电压/V	1.21
面积/mm²	0.18
电路门数/kGE	176.46
吞吐率/ Mbit·s^-1	388.85

设计	文献[15]	文献[16]	文献[17]	本文
算法	SC	SC	SC	SC
工艺/nm	180	180	180	180
架构	parallel	Tree	Tree	Line
频率/MHz	150.0	446.7	337.0	222.2
周期/clk	1 560	1 534	1 534	330
吞吐率/ Mbit·s^-1	49	149	126	343.60
电路门数/kGE	183.64	295.44	256.34	176.46
面积/mm²	1.71	5.35	-	2.19
效率/ Mbit·s^-1·kGE^-1	0.267	0.505	0.492	1.949

数据类型	文献[15]	文献[16]	本文
工艺/nm	180	180	180
电压/V	1.50	1.80	1.61
功耗/mW	67.0	1 072.9	200.0