The Compensation Problem and Solution Using Design of Experiments for Dense Multilayer Printed Circuit Boards

Armed with the model building techniques of design of experiments (DOE), compensation predictions can be greatly improved. Using results from a real production job, this paper will present the most common compensation models used in the industry today and demonstrate their limited capability. The limited capability of these models is their faulty assumptions. One of these erroneous assumptions is that the prepreg in the multilayer board build plays a minor roll to the overall movement of the core layers. This paper will demonstrate that prepreg is one of the most important variables that affects the movement of thin cores and is an important variable in the experimental designs. Each DOE focused on an individual core type. The DOEs used only two control variables and two noise variables with 16 experimental runs. Configurations of the cores and the prepreg glass styles were the two control variables. The time period during lamination and the lamination press used in production were the two noise variables. From these small experiments coefficients were derived for each core type. These coefficients were used to generate a general compensation model. A model validation process is demonstrated on panels built in a different production plant. Validation of the model is an ongoing process and indicates where the model is performing well, where the model is weak, and where model improvements are needed to keep pace with rising technology expectations.