A decomposition approach for testing large analog networks

This article presents an efficient method for testing large scale analog and mixed mode networks. Test equations are derived lbr a partitioned network from Krichhoff current law equations at the partition points. Voltages at the partit ion points are used to identify network pararneters. Thc method has applications to circuit modeling, fault diagnosis, testing and calibration. The conventional testing methods for dynamic, nonlinear networks are based on the sensitivity approach, which uses incremental changes in voltages to estimate changes in network parameters. However, this conventional approach cannot handle large scale circuits because the sensitivity matrix is dense. This results in enormous requirements for memory space and computing time when the circuit size becomes large. The new method overcomes these deficiencies of the sensitivity approach. In this article, we introduce the decomposition method, describe its basic features and its algorithm. and compare this method with a conventional. sensitivity technique using testing network examples.