基于DMX512协议的LED解码芯片验证技术研究

发布时间：2018-04-04 19:15

本文选题：验证技术　切入点：功能覆盖率　出处：《浙江大学》2015年硕士论文

【摘要】：近年来,集成电路产业发生了翻天覆地的变化。随着制造工艺进入纳米阶段,芯片的速度和集成度不断上升,由此引发的芯片功能验证问题日益显著。芯片的验证技术影响着芯片的性能、成本和设计周期,在一定程度上决定了芯片在市场中的生存能力,因此验证技术的提高已成为IC设计者重点关注的热点问题之一。本文从芯片验证方法的发展历程出发,探讨目前功能验证技术的分类及各自优缺点。形式验证主要采用静态形式验证方法,适合小规模组合电路测试；FPGA验证虽能进行软硬结合仿真,但信号可视性较差；仿真性功能验证由于其自身动态验证的特点,已成为IC验证的主流方法之一。其中,基于功能覆盖率驱动的验证方式能提高仿真性功能验证的完备性和可靠性,但提高效率还得取决于随机向量的生成。针对此问题,本文提出一种基于遗传算法的测试激励生成技术,对基于DMX512协议的LED解码芯片进行仿真性功能验证,加速功能覆盖率收敛速度,有效提高验证效率。本文具体分析研究,主要包括以下几方面：第一,对基于DMX512协议的LED解码芯片进行设计分析,特别是对验证对象——DMX512解码模块的内部架构、数据帧、时序等特点进行详细研究,通过分析验证对象特点提取关键信息,制定相应的验证计划。第二,简要介绍遗传算法在功能覆盖率收敛上的应用,详细分析遗传算法中适应度函数和三大遗传算子的选取方法,利用概率分布函数分析各遗传算子在本文应用中的优越性,得到基于比例选择算子、均匀交叉算子和二元变异算子的优秀遗传算法。第三,根据LED解码模块的工作特点,搭建基于VMM形式的验证平台,利用System Verilog(SV)高级验证语言实现遗传算法并将其嵌入于验证平台中。第四,利用VMM形式的验证平台,对DMX512解码模块分别进行全随机向量测试和基于遗传算法的验证向量测试并分析在不同激励下,芯片功能覆盖率的收敛速度。最后说明基于遗传算法的测试向量生成技术能够加速功能覆盖率收敛速度,提高验证效率。
[Abstract]:In recent years, the integrated circuit industry has undergone earth-shaking changes.As the manufacturing process enters the nanometer stage, the speed and integration of the chip are rising, and the problem of chip function verification is becoming more and more obvious.Chip verification technology affects the chip performance, cost and design cycle, to a certain extent, determines the viability of the chip in the market, so the improvement of verification technology has become one of the hot issues that IC designers focus on.Starting from the development of chip verification methods, this paper discusses the classification of current functional verification techniques and their respective advantages and disadvantages.The formal verification mainly adopts the static formal verification method, which is suitable for the FPGA verification of small-scale combinational circuit testing, although it can carry out the soft and hard combination simulation, but the signal visibility is poor, the simulation function verification is due to its own characteristics of dynamic verification.It has become one of the main methods of IC verification.Among them, the verification method based on function coverage driving can improve the completeness and reliability of simulation function verification, but improving efficiency depends on the generation of random vectors.To solve this problem, this paper presents a test excitation generation technique based on genetic algorithm, which simulates the performance of LED decoding chip based on DMX512 protocol, accelerates the convergent speed of function coverage, and effectively improves the efficiency of verification.In this paper, the specific analysis and research, mainly including the following aspects:First, the LED decoding chip based on DMX512 protocol is designed and analyzed, especially the internal structure, data frame, timing and other characteristics of the verification object-DMX512 decoding module are studied in detail, and the key information is extracted by analyzing the characteristics of the verification object.Make corresponding verification plan.Secondly, the application of genetic algorithm in the convergence of function coverage is briefly introduced. The selection methods of fitness function and three genetic operators in genetic algorithm are analyzed in detail, and the superiority of each genetic operator in this paper is analyzed by using probability distribution function.An excellent genetic algorithm based on proportional selection operator, uniform crossover operator and binary mutation operator is obtained.Thirdly, according to the working characteristics of LED decoding module, a verification platform based on VMM is built, and the genetic algorithm is implemented by System Verilog SVS and embedded in the verification platform.Fourthly, using the VMM verification platform, the full random vector test and the verification vector test based on genetic algorithm are carried out for the DMX512 decoding module, and the convergence rate of chip function coverage is analyzed under different excitations.Finally, it is shown that the test vector generation based on genetic algorithm can accelerate the convergence rate of function coverage and improve the efficiency of verification.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN312.8

【参考文献】