A methodology for process capability analysis to optimize specifications of laser trimmed embedded resistors for advanced electronics manufacturing applications

The continued need for size, weight and power reduction is driving electronics manufacturing towards embedding of components attached to the substrate surface of Printed Circuit Boards (PCBs). Embedded components include thin film resistors placed in the build-up layers of PCBs. One challenge of embedding resistors is that applications require tighter specification range achieved by standard processes. A laser tool integrated with an electrical test tool is used to ablate small amounts of resistor material while measuring its ability to produce a resistor with tight tolerances. For this technique to succeed, resistors must be manufactured with a resistance value below a target and then trimmed with the laser. As resistors are trimmed, resistance values increase.;Process capability analysis is used to assess how parts produced conform to the required specifications (in other words, this analysis can be used to determine yields). Some authors emphasize that the classical Process Capability Indices (PCIs) such as Cp, Cpk, Cpm, and Cpmk should be applied to in-control and normally distributed processes. Several studies have been conducted to demonstrate how poorly the capability studies based on the normal distribution perform as a predictor of yields when the process is not normally distributed.;In industrial practice, the normality assumption is frequently utilized to calculate PCIs. However, the importance of defining a good distribution fit for the data (rather than using the normal if the data are not normal) to apply the correct PCIs and the investigation of the statistical status of the process (either in-control or out-of-control) are commonly omitted when performing capability studies.;In this thesis, a new methodology for process capability analysis is proposed to assist industry to either confirm the assumption of normality and statistical control or, if not, to properly account for lack of normality and/or state of control. This methodology helps define the appropriate PCIs and yields to provide reliable results, and investigate whether the process is capable of delivering more stringent specifications. While the techniques included in the proposed methodology are applied in the electronics arena, they have wide applicability to any organization's domain.;The methodology includes three different cases. Case 1 is applied when the data are in statistical control and follow the normal distribution. Case 2 is suitable when the data are in statistical control and do not follow the normal distribution. Case 3 is used when the data are not in statistical control and the traditional PCIs cannot be applied. Every case is applied to 15,000 resistance values to determine whether three different targets of embedded resistors, in a production line of an electronics manufacturing company, are capable of delivering more stringent specification ranges. A software application is proposed to real-time monitor the basic statistics, the point estimates of the PCIs, their confidence intervals, the observed yields and the natural tolerance limits of the process of the resistance values. According to the analysis and results, two resistor targets are capable of delivering tighter specification ranges than the actual specification range; one resistor target seems to be not capable of delivering a tighter specification range than the actual ones.