基于深度学习算法的数模混合芯片测试方法研究  

Research on the testing method of digital analog hybrid chips based on deep learning algorithms

作　　者：张永华 ZHANG Yonghua(Beijing Zhaoxin Electronic Technology Co.,Ltd.,Beijing 100094,China)

机构地区：[1]北京兆芯电子科技有限公司,北京100094

出　　处：《电子设计工程》2025年第4期77-81,共5页Electronic Design Engineering

摘　　要：数模混合芯片内部结构的复杂性和信号交互的多样性,使得测试难以有效覆盖芯片的所有工作模式和边界条件,进而导致测试覆盖率偏低。为此,提出了基于深度学习算法的数模混合芯片测试方法研究。根据数模混合芯片的基本组成和运行原理以及测试需求,采用卷积神经网络算法构建测试信号生成模型,结合芯片状态融合信号特征输出测试信号,联合测试项目的覆盖范围编排测试用例,并利用混合信号仿真测试策略实现对芯片的性能测试。实验结果表明,该文方法能够全面考虑芯片的功能测试项目,增加了测试的广度和深度,芯片测试覆盖率始终保持在75%以上,测试结果的可靠性较高。Due to the complexity of the internal structure of digital analog hybrid chips and the diversity of signal interactions,it is difficult to effectively cover all operating modes and boundary conditions of the chip,resulting in low test coverage.Therefore,a research on the testing method of mixed signal chips based on deep learning algorithms has been proposed.Based on the basic composition and operating principles of mixed signal chips,as well as testing requirements,a convolutional neural network algorithm is used to construct a test signal generation model.Combined with the chip state fusion signal characteristics,the test signal is output,and the coverage range of joint testing projects is combined.Test cases are arranged,and mixed signal simulation testing strategies are used to achieve performance testing of the chip.The experimental results show that this method can fully consider the functional test items of the chip,and increase the breadth and depth of the test.The coverage rate of the chip test is always above 75%,and the reliability of the test results is high.

关 键 词：深度学习算法 数模混合芯片 性能测试 测试用例 测试信号 

分 类 号：TN407[电子电信—微电子学与固体电子学]