edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 26 NO. 3 20 for the DFI parameters and observing the critical features under x-ray inspection. The latter requires setting up a simulation environment[19] to observe how x-ray radiation interacts with a 3D model of the sample (Fig. 8). These two approaches together can help to set thresholds for DFI parameters of the target design collectively referred to as PDFI. The x-ray Monte Carlo (XRMC) simulation method utilizes a standard experimental setup comprising three main components: an x-ray source, a sample, and a detector. The x-ray source can be either point-like or have a threedimensional Gaussian distribution. The emitted radiation from the source may be unpolarized, partially polarized, or fully polarized. The energy spectrum of the x-rays can contain continuous components and discrete energy lines, which can be monochromatic or follow Gaussian distributions with specified standard deviations. The sample is a CAD model which in case of commercial adoption needs to be provided by the package designer. Such a model will consist of multiple materials or phases, each characterized by its mass density and elemental composition (atomic numbers and weight fractions). Each phase is assumed to be homogeneous, and the sample’s geometry is defined using three-dimensional objects bounded by quadric surfaces. The detector in the setup can be a pixel array or a single-element detector. Each pixel or element may have a rectangular or elliptical shape. The detector can record various properties, such as the number of x-ray photons collected by each pixel over a specific exposure time, the total deposited energy of these photons, or the full spectrum of the detected radiation at each pixel. Importantly, all components (source, sample, and detector) can be positioned at arbitrary locations and orientations within the simulation. DFI Heuristics: In the development of the CMx-ray framework, a key focus has been on establishing heuristics to determine the appropriate thresholds for DFI parameters. These heuristics are crucial in ensuring that the DFI parameters are set in a way that maximizes the efficacy of x-ray inspection in IC packaging. The approach to defining these heuristics encompasses several critical aspects. • Empirical Testing and Simulation Models: The initial heuristic in- volves a combination of empirical testing of sample packages and the use of simulation models. This approach is aimed at identifying features within the IC package that are more susceptible to defects and noise perturbations. By analyzing real-world data from sample tests and leveraging the predictive power of simulation models, insights into which features pose higher risks can be obtained. This information is instrumental in setting more accurate and effective DFI parameter thresholds. • Complexity and Size of Package Features: The second heuristic considers the complexity and size of features within a package. For instance, capturing the intricacies of numerous tightly stacked RDL layers in an x-ray image can be challenging. Therefore, it is logical to set larger thresholds for such complex features. This consideration ensures that the thresholds are realistic and align with the practical limitations of x-ray imaging capabilities. • Impact of Noise Scattering on Imaging: Another critical heuristic is the assessment of how noise scattering affects the imaging of specific features, particularly those involving high atomic number (high Z) materials. These materials, often found in various types of bump structures, are prone to causing significant imaging issues due to their propensity to scatter x-rays more intensely. Figure 9 shows the difference the presence of noise scattering makes in image reconstruction. Understanding the impact of these materials on image quality is essential for setting appropriate DFI thresholds that account for potential noise-related challenges. • Iterative Optimization of DFI Thresholds: The final heuristic emphasizes the importance of continuously refining the DFI thresholds. This optimization Fig. 9 Effect of high-Z noise scattering in x-ray image of tight pitch solder bumps. Fig. 8 Setting up a simulation environment for x-ray interaction.
RkJQdWJsaXNoZXIy MTYyMzk3NQ==