一种低功耗高压缩率测试图形生成方法的实现与应用研究

It needs a great amount of data to test future's nano-meter circuits. Moreover, excessive test power affects test quality greatly. Therefore, high test compression ratio and low-power testing have been the difficulties and research focuses of nano-meter circuit testing. Utilizing multi-dimension similarity test patterns, this research explores a solution to solve these problems. Firstly, it proposes a method to generate multi-dimension similarity test patterns, and studies the implemention method utilizing two seeds. Then, it proposes a new test pattern theory, and a seed optimization method based on this theory is developed. It can solve the problems of low test pattern generation efficiency and high test hardware overhead. Thirdly, experiments on the circuit under test (CUT) will be carried out, and the relationship between the delay fault coverage, test power reduction and the configuration of two seeds of the CUT will be drawn. Based on this relationship the optimization method about the structures of two seeds is developed. Further, a method to re-compress seeds will be studied, and the method to implement the decompression circuit will be conducted. Finally, experimental results on International Test Benchmarks and related industrial circuits will be provided to demonstrate the test efficency and quality of the proposed method. This study utilizes the multi-dimension similarity test patterns to solve the problems of high test power and low test data compression ratios. There are few related researchs abroad or at home. Therefore, it is an advanced research topic about nano-meter circuit test. This research explores a new way to enhance test compression ratio, and satifies the needs of very high test compression ratios in the furture.

由于纳米数字集成电路测试数据量巨大，且其过高的测试功耗极大地影响了测试质量，因此测试数据压缩和低功耗测试是难点与研究热点。本项目探索一种采用多维相似性测试图形的解决方案。研究多维相似性测试图形的构造方法，提出基于两个种子向量的实现方法；建立一种新型的测试图形表达和分析方法，研究种子向量赋值方法，以解决测试生成效率低和硬件开销大等问题；实验与分析种子向量构成方法对故障覆盖率和测试功耗减少率等的影响，提出种子向量优化方法；研究种子向量的再次压缩方法，根据多维线性和再次压缩关系建立硬件实现方法；最后用国际实验电路和相关工业电路的实验结果验证所提方法的测试效率和质量。本研究提出的低功耗高压缩率的多维相似性测试图形实现与应用方法，国内外尚处于探索阶段，是前沿性课题，其意义在于探索出解决测试压缩的一个新方向，满足未来测试压缩率大幅度提高之迫切需求。

由于集成电路测试面对着半导体器件特征尺寸小、集成电路集成度和复杂度高导致的芯片测试功耗高、面积开销和测试数据量大等问题，因此测试数据压缩和低功耗测试是难点与研究热点。本项目探索了几种低功耗低成本测试图形的生成方法的解决方案。研究多维相似性测试图形的构造方法，提出基于两个种子向量的实现方法；建立一种新型的测试图形表达和分析方法，以解决测试生成效率低和硬件开销大等问题；实验与分析种子向量构成方法对故障覆盖率和测试功耗减少率等的影响；最后用国际实验电路的实验结果验证所提方法的测试效率和质量。除以上之外，本项目还探索了将机器学习算法应用到集成电路测试流程中，提出几种可能的解决方案，为满足未来海量测试数据生成做了有益的尝试。