Automated system infrastructure to facilitate design of experiments (DOE) data analysis

Abstract

The complex semiconductor fabrication process needs to be optimized quickly to increase process yield and efficiency in fab operations. The Design of Experiments (DOE) technique is frequently applied in semiconductor domain by separating a wafer lot into wafer splits, and studying the effects of varying process parameters in relation to a baseline split across diffusion, photo, etch and other fabrication tools. An automated solution implemented at Kilby Center (KFAB) of Texas Instruments to capture and store the lot split DOE data for analysis, which satisfies the requirements of data volume, accuracy, and accessibility, is presented. Guidelines on how lot splits are defined consistently (the "rule set"), design and implementation of the system infrastructure components, and system's applications are included.