Automatic analog fault simulation

This paper presents an approach for using analog simulation together with artificial neural networks as a tool in the development of a test strategy. Relationships between circuit components and node states provided by simulation are used to create relationship matrices for a neural network. This approach provides a guide to diagnosing circuits by developing inferences about circuit component problems. The approach presented in this paper is a component of an automated test program generation (ATPG) concept. Problems associated with this approach are identified and defined. An industry accepted simulation engine generates data for operational circuit behavior. The simulation engine uses a netlist consisting of a to-from list, model definitions and an input description. Failures are simulated by one of two methods: (1) modifying circuit netlist, (2) modifying components. The circuit netlist is modified by adding opens and shorts to circuit nodes. The component models are modified by changing the component parameters within the component library. Simulation faults are chosen by the test program developer. Measures of component and node deviations from normal are used in the development of neural network matrices. An example analog circuit demonstrates the approach and typical results. This work may lead to development of a test strategy engine for automated analog test program development