Multi-variable statistical process controller for discrete manufacturing

A method of controlling a plurality of process control variables in processing discrete products, said method comprising:

• a. utilizing process models relating a plurality of product quality parameters to said plurality of process control variables;
• b. measuring said plurality of product quality parameters on a first plurality of products;
• c. exercising statistical quality control tests on said plurality of product quality parameters of said first plurality of products;
• d. continuing processing if statistical quality control tests are successful;
• otherwise
• e. measuring said plurality of product quality parameters on a second plurality of products;
• f. tuning said process models to create tuned process models using said plurality of product quality parameters from said first and said second plurality of products, wherein said tuning estimates a changed state of said processing;
• g. estimating new values for said plurality of process control variables from said tuned models;
• h. processing a third plurality of products at said new values of said plurality of process control variables;
• i. measuring said plurality of product quality parameters on said third plurality of products;
• j. repeating said tuning of said process models and said estimating new values for said plurality of process control variables if said quality control parameters from said third plurality of products are not acceptable;
• otherwise
• k. continuing processing products with said new values of said plurality of process control variables if said quality control parameters from said third plurality of products are acceptable.

A method of controlling a plurality of process control variables in semiconductor wafer processing, said method comprising:

• a. utilizing process models relating a plurality of product quality parameters to said plurality of process control variables;
• b. measuring said plurality of product quality parameters on a first plurality of wafers;
• c. exercising statistical quality control tests on said plurality of product quality parameters of said first plurality of wafers;
• d. continuing processing if statistical quality control tests are successful;
• otherwise
• e. measuring said plurality of product quality parameters on a second plurality of wafers;
• f. tuning said process models to create tuned process models using said plurality of product quality parameters from said first and said second plurality of wafers, wherein said tuning estimates a changed state of said semiconductor wafer processing;
• g. estimating new values for said plurality of process control variables from said tuned models to optimize said plurality of product quality parameters;
• h. processing a third plurality of wafers at said new values of said plurality of process control variables;
• i. measuring said plurality of product quality parameters on said third plurality of wafers;
• j. repeating said tuning of said process models and said estimating new values for said plurality of process control variables if said quality control parameters from said third plurality of wafers are not acceptable;
• otherwise
• k. continuing processing wafers with said new values of said plurality of process control variables if said quality control parameters from said third plurality of wafers are acceptable.