Logo image
Layout-Oriented Defect Set Reduction for Fast Circuit Simulation in Cell-Aware Test
Thesis

Layout-Oriented Defect Set Reduction for Fast Circuit Simulation in Cell-Aware Test

Liu, Hsuan Wei
Masters, 國立清華大學, 電機工程學系
2015

Abstract

自動測試形樣產生系統 單元認知測試 缺陷基底測試 缺陷模型的建立和模擬 Automatic test pattern generation (ATPG) Cell-aware test (CAT) Defect-based testing Fault modeling and simulation
Because of the high complexity and density of modern CMOS circuits due to the continuous technology advancement, it is more and more difficult to detect physical defects, which can occur during any step of the fabrication process, on a modern CMOS chip using advanced technology. Therefore, the cell-aware test (CAT) methodology was previously proposed to target the cell-internal defects that cannot be easily detected by the gate-level stuck-at fault (SAF) patterns generated by the conventional ATPG tools. It was shown to reduce the defect level on industrial CMOS-based designs, with the help of detailed defect injected transistor-level circuit simulation and defect-enhanced SAF ATPG. However, the detailed transistor-level circuit simulation has been considered an issue in CAT, as it is very time consuming. The problem mainly lies in that all parasitic capacitors and resistors extracted from cell layout are considered as defect targets, so the defect set is large. To reduce the number of the defect set, and therefore the circuit simulation time, we take layout into consideration when we construct the defect set for each cell, effectively removing the redundant or unnecessary defects and therefore reducing the circuit simulation time dramatically. We propose a generalized approach that can be used to build the fault models based on the cell layout, where the generated faults are closer to the realistic physical defects on the layout, so the number of faults can be significantly reduced. The proposed method is verified by commercial 180nm and 350nm CMOS standard cell library, and the circuit simulation time is reduced to about only 20% or even lower as in comparison with the original CAT methodology.

Metrics

1 Record Views

Details

Logo image