作者: David Bryan Lavo , Tracy Larrabee
DOI:
关键词:
摘要: Determining the source of failure in a defective circuit is an important but difficult task. Important, since finding and fixing root cause defects can lead to increased product quality greater profitability; difficult, because number locations variety mechanisms whereby modern fail are increasing dramatically with each new generation circuits. This thesis presents method for diagnosing faults combinational VLSI While it consists several distinct stages specializations, this designed be consistent three main principles: practicality, probability precision. The proposed approach practical, as uses relatively simple modeling algorithms, limited computation, enable diagnosis even very large It also probabilistic , imposing probability-based framework resist inherent noise uncertainty fault diagnosis, allow combined use multiple models, data sets towards single diagnostic result. Finally, precise, using iterative move from abstract models complex specific behaviors. The system address both initial stage when nothing known about or types present, well end-stage which arbitrarily-specific applied reach desired level deals logic fails quiescent current (IDDQ) test failures. addresses problem size dictionary-based doing so introduces concept low-resolution diagnosis.