电子学报 ›› 2017, Vol. 45 ›› Issue (7): 1686-1694.DOI: 10.3969/j.issn.0372-2112.2017.07.019

• 学术论文 • 上一篇    下一篇

基于参数化模型的FPGA时钟网络设计和优化

余乐1, 陈岩1, 李洋洋1, 吴超1, 王瑶1, 苏童1, 谢元禄2   

  1. 1. 北京工商大学计算机与信息工程学院食品安全大数据技术北京市重点实验室, 北京 100048;
    2. 中国科学院微电子学研究所中国科学院微电子器件与集成技术重点实验室, 北京 100049
  • 收稿日期:2016-06-13 修回日期:2016-12-09 出版日期:2017-07-25
    • 作者简介:
    • 余乐(通信作者),男,1983年6月生于湖北黄冈,现为北京工商大学计算机与信息工程学院讲师.现在的研究兴趣:计算视觉与类脑芯片.E-mail:yule@btbu.edu.cn;陈岩,女,1963年7月生于黑龙江哈尔滨,现为北京工商大学计算机与信息工程学院教授,研究方向:无线传感器网络技术应用、数字通信抗干扰技术.
    • 基金资助:
    • 北京市自然科学基金 (No.4174086)

Design and Optimization of FPGA Clock Network Based on Parameterized Model

YU Le1, CHEN Yan1, LI Yang-yang1, WU Chao1, WANG Yao1, SU Tong1, XIE Yuan-lu2   

  1. 1. Beijing Key Laboratory of Big Data Technology for Food Safety, School of Computer and Information Engineering, Beijing Technology and Business University, Beijing 100048, China;
    2. Key Laboratory of Microelectronic Devices & Integrated Technology, Institute of Microelectronics, Chinese Academy of Sciences, Beijing 100049, China
  • Received:2016-06-13 Revised:2016-12-09 Online:2017-07-25 Published:2017-07-25

摘要:

本文在FPGA时钟网络(Clock Distributed Network,CDN)关键结构尺寸的参数化建模基础上,提出一种针对全定制FPGA CDN的设计和优化方法.本文所建立的参数化模型将结构尺寸分为拓扑结构和电路与互连两类,分别给出了这两类尺寸参数的设计原则.在标准CMOS 0.13μm工艺下,对H树型、鱼骨型以及混合型三种类型时钟网络设计了2组结构参数,分别代表优化前和优化后,对比分析延时、偏斜、功耗和面积等性能参数.实验结果显示:混合型结构在绝对延时和时钟偏斜上减小最多,分别达到20.89%和63.20%;鱼骨型结构的面积减小达到50.14%;H树型结构的绝对延时和功耗则均降低了7.37%和8.33%.以上结果充分证明了本文所提设计优化方法的有效性.

关键词: FPGA, 时钟网络, 参数化, 建模

Abstract:

This paper proposes a methodology for the design and optimization of CDN (Clock Distributed Network) in full custom FPGA systems based on parameterized models of the structural sizes of the key components of CDN systems.The characteristic model proposed herein divides structural sizes into two categories:topology structure,and circuit and interconnect,and provides the design methodology of these two types of sizes.The paper establishes two sets of structure parameters for H-tree,fish-bone and mixed clock networks in standard CMOS 0.13 μm technology,each representing CDN with or without optimizations.And then we compares the performance in terms of transmission delay,clock skew,power consumption,chip area,among a multitude of other parameters.Experiment results indicate that mixed structure results in the greatest reductions of transmission delay and clock skew by 20.89% and 63.20%,respectively;fish-bone structure achieves the best reductions in chip area by 50.14%,while H-tree structure reduces in transmission delay and power consumption by 7.37% and 8.33%,respectively,which strongly confirms the proposed design and optimization methodology.

Key words: FPGA, clock distributed network (CDN), parameterized, modeling

中图分类号: