X-ray Inspection Model Validation with Physical Dosimetry

Abstract

A non-functional printed circuit board assembly was developed using ULTIboard and a simple computational transport model was created for use with the Monte Carlo software MCNP. EBT3 film radiography was used to compare with Monte Carlo simulation to validate the board representation. The computational transport model was then revised to include connection pins, solder balls, highly attenuating internal structures, and copper trace distribution. Results from the revised model were compared to the EBT3 films to observe improvements to dose profiles. It was found that this method was useful in verifying the placement of components, as the dose profiles were observed to follow the same trends. The experiment was then repeated using XRQA2 films to achieve the same level of contrast with 1% the dose of EBT3 films. It was found that high contrast may be achieved using these films to identify major issues with the model geometry, at a cost of dose profile accuracy. A second validation method was applied to the model using 37 CaF2 thermoluminescent dosimeters (TLDs). TLD measurements were compared with the simplified and complex transport models to identify the features that have the greatest impact on simulation accuracy. The TLD calibration to CaF2 was found to be accurate within 5.6%, while calibration to dose in Si was found to be accurate within 4.7%. It was observed that the accurate representation of solder balls and proper modeling of highly attenuating internal structures had the greatest impact on simulation accuracy.