Increasing Yields with Entropy-based Analysis of Test Data

Manufacturing test data is often analyzed with the direct goal of increasing production yields and reduction of variance within the product. Six Sigma and statistical process control methodologies are the standard tool sets for analysis of test data. The normal course of events results in the analyst seeking to reduce failures by taking steps to correct the tests that fail most often. Unfortunately, the complex nature of modern electronics is such that the systems often exhibit chaotic dynamics with self-organized criticality. In these particular cases standard analysis, such as Six Sigma, and correcting the tests which fail most often, are insufficient due to the chaotic nature of the data. This paper presents an entropy-based analysis of these particular types of systems. The algorithms for detection of self-organized criticality and chaotic dynamics within the test data are given. Corrective action solutions are also offered to assist in dealing with the chaotic dynamics in these systems. Finally, a case study is presented in which the given analysis techniques were utilized to effect a 52% yield increase on a product at Rockwell Collins.