An equivalent processing method for integrated circuit electrical parameter data using BP neural networks

Abstract

In response to the current problem of significant differences between old and new equipment in testing electrical parameter data, as well as the lack of a unified parameter measurement standard in the IC testing industry, we propose an equivalent processing method for electrical parameter data using BP neural networks. Firstly, the method involves mapping the electrical parameter data measured by both the old and new equipment and subsequently trimming the test data. Then, the BP neural network is applied to the equivalent processing of the dataset, aiming to unify the test limits and electrical parameter data. This approach facilitates the extrapolation of new data from old equipment data, enabling a seamless data transfer and the harmonization of industry standards. This method effectively reduces the testing difficulty and misjudgment rate, thereby improving the testing efficiency for semiconductor manufacturers. The experimental results demonstrate that the proposed algorithm can accurately infer the new equipment's test data from the old equipment's data, ensuring data smooth transition. Furthermore, the modification of electrical parameter data does not affect the yield.