Analysis of mixed-signal manufacturability with statistical TCAD

We have developed a method which combines TCAD simulation with statistical analysis of empirical data. As a result, we can determine worst case models before a significant sample size of fabricated devices can be characterized. As measured data is collected we use models built from simulated data to analyze measured in-line and e-test data. We determine what is the minimum set of e-test or in-line measurements to control the SPICE model parameters. As a result we can provide predictive worst case models, update them based on fabrication data as development proceeds, and finally use them to control the manufacturing line