Using neural networks and 3D polynomial interpolation for the study of probe yield vs. E-test correlation. Application to sub-micronics mixed-signal technology

In the present paper we propose the use of neural networks for statistical modelling of data, as well as the use of 3D surface in order to visualise results in a very intuitive way. The scope of the paper is to present a method for extracting qualitative information from the confrontation of yield and E-test data in order to easily identify best process conditions and potential process marginality issues. The neural network approach is a new way to face determining the huge amount of raw data that yield analysis involves in the microelectronics industry.