Automatic PCB Material Characterization Using Design of Experiments and Mixed-Integer Sequential Quadratic Programming

The growing demand for higher data rate SerDes channel calls for better PCB channel modelling. Conductor surface roughness can no longer be ignored, and it must be accurately modelled. PCB material characterization is a reliable methodology to extract roughness parameters from the lab measurement. The characterization process calibrates a transmission line model against lab measurement in terms of insertion loss and phase. However, manual calibration is not an easy task to get excellent broadband fitting for both insertion loss and phase. This paper introduces an automatic calibration approach based on Design of Experiments (DoE) and Mixed-Integer Sequential Quadratic Programming (MISQP).