基于深度神经网络的高速信道自适应均衡器

计算机工程与科学 ›› 2022, Vol. 44 ›› Issue (04): 605-610.

基于深度神经网络的高速信道自适应均衡器

翦杰，罗章，赖明澈，肖立权，徐炜遐

（国防科技大学计算机学院,湖南长沙 410073）

收稿日期:2021-10-12 修回日期:2021-12-19 接受日期:2022-04-25 出版日期:2022-04-25 发布日期:2022-04-20
基金资助:
国家重点研发计划（2018YFB2202303）;并行与分布处理重点实验室基金（6142110200301）

An adaptive high-speed channel equalizer based on deep neural network

JIAN Jie,LUO Zhang,LAI Ming-che，XIAO Li-quan,XU Wei-xia

（College of Computer Science and Technology,National University of Defense Technology,Changsha 410073,China）

Received:2021-10-12 Revised:2021-12-19 Accepted:2022-04-25 Online:2022-04-25 Published:2022-04-20

摘要/Abstract

摘要： 高速串行接口是提高高性能互连网络带宽的关键技术，而信道均衡器则是提高信号完整性的核心部件。利用现代数字信号处理(DSP)结构，提出了基于深度神经网络(DNN)的高速信道均衡研究方法，此方法在面向未来50 GB以上的高速信道时，克服了传统判决反馈均衡器（DFE）的判决速度受限于反馈回路的固有缺陷问题。仿真结果表明，在采用PAM4编码方式，高速信道波特率为28 GB，信道损耗为15 dB，或者波特率为56 GB，信道损耗为30 dB时，与传统的15阶FFE 组合2阶DFE的均衡器结构相比，本文所提出的3层DNN结构，具有更好的均衡效果，以及更快的均衡收敛速度。

关键词: 深度神经网络, 快速串行链路, 数字信号处理器, 均衡器

Abstract: High-speed serial interface is a key technology to improve the bandwidth of high-performance interconnection networks, and the channel equalizer is the core component to improve signal integrity. This paper uses the modern digital signal processing (DSP) structure to propose a deep neural network (DNN)-based high-speed channel equalization research method. This method overcomes the inherent shortcoming that the decision speed of the traditional decision feedback equalizer (DFE) is limited by the feedback loop when facing high-speed channels above 50GB in the future. The simulation results show that, compared with the traditional 2-tap DFE architecture with 15-tap FFE, the proposed 3-layer DNN structure has better equalization effect and faster equalization convergence speed, when the PAM4 encoding method is used, the high-speed channel baud rate is 28GB, and the channel loss is 15dB, or the baud rate is 56GB and the channel loss is 30dB.

Key words: deep neural network, high-speed serial link, digital signal processor, equalizer

翦杰, 罗章, 赖明澈, 肖立权, 徐炜遐. 基于深度神经网络的高速信道自适应均衡器[J]. 计算机工程与科学, 2022, 44(04): 605-610.

JIAN Jie, LUO Zhang, LAI Ming-che, XIAO Li-quan, XU Wei-xia. An adaptive high-speed channel equalizer based on deep neural network[J]. Computer Engineering & Science, 2022, 44(04): 605-610.

参考文献［4］

［1］	Kao S-Y, Liu S-I.A 7.5-Gb/s one-tap-FFE transmitter with adaptive far-end crosstalk cancellation using duty cycle detection［J］.IEEE Journal of Solid-State Circuits,2013,48(2):391-404.
［2］	Momtaz A,Green M M.An 80 mW 40 Gb/s 7-tap T/2-spaced FFE in 65 nm CMOS［C］∥Proc of 2009 IEEE International Solid-State Circuits Conference,2009:364-365.
［3］	Sant S R,Waikar S S,Dave M,et al.A 16-Gbps 9 mW transmitter with FFE in 90 nm CMOS technology for offchip communication［C］∥Proc of 2011 IEEE Computer Society Annual Symposium on VLSI,2011:78-83.
［4］	Su Peng-zhou, Huang Lu, Fang Yi,et al.Design of a new 6.25 Gb/s CTLE equalizer［J］.Microelectronics,2016,46(2):215-218.(in Chinese)
［5］	Jung J W, Razavi B. A 25 Gb/s 5.8 mW CMOS equalizer［J］.IEEE Journal of Solid-State Circuits,2015,50(2):515-526.
［6］	Hsieh C,Liu S.Decision feedback equalizers using the back-gate feedback technique［J］.IEEE Transactions on Circuits and Systems II:Express Briefs,2011,58(12):897-901.
［7］	Dickson T O,Bulzacchelli J F,Friedman D J.A 12-Gb/s 11-mW half-rate sampled 5-tap decision feedback equalizer with current-integrating summers in 45-nm SOI CMOS technology［C］∥Proc of 2008 IEEE Symposium on VLSI Circuits,2008:58-59.
［8］	Nazari M H,Emami-Neyestanak A.A 15 Gb/s 0.5 mW/Gb/s 2-tap DFE receiver with far-end crosstalk cancellation［C］∥Proc of 2011 IEEE International Solid-state Circuits Conference,2011:446-448.
［9］	Bjerke B A. Pulse amplitude modulation in handbook of computer networks:Key concepts,data transmission,and digital and optical networks［M］.New Jersey:John Wiley & Sons,Inc.,2016.
［10］	Kin H. Neural networks:A comphrehensive foundation［M］.2nd ed.NJ:Prentice-Hall,Upper Saddle River,2001.
［11］	Hinton R,Willians R J.Learning representations by backpropagating errors［J］.Nature,1986,323(9):533-536.
	附中文参考文献：
［4］	苏鹏洲，黄鲁，方毅，等.一种新型6.25 Gb/s CTLE均衡器的设计［J］.微电子学，2016,46(2):215-218.

[1]	马铭苑, 李虎, 王梓斌, 况晓辉. 深度神经网络模型后门植入与检测技术研究综述[J]. 计算机工程与科学, 2022, 44(11): 1959-1968.
[2]	马鹏勇陈书明孙锁林. 双簇结构DSP的数据Cache优化[J]. J4, 2008, 30(9): 119-121.
[3]	柳笛黎福海闫旭. 基于DSP的嵌入式图像监控系统的设计与实现[J]. J4, 2008, 30(2): 39-41.
[4]	胡贯荣舒行正涂刚. 嵌入式μCLinux引导装载程序的设计与实现[J]. J4, 2007, 29(11): 31-33.