Work

A novel enhancing method for terahertz imaging of integrated circuits flaw detection

Year: 2024

Type: article

Abstract: This research paper addresses the current gap in understanding the intricate damage types within packaged integrated circuits (ICs). We introduce a novel and multiscale latent low-rank representation (M-LatLRR) methodology, meticulously tailored to enhance the quality of terahertz (THz) IC images and reveal latent features. By integrating THz imaging with our proposed M-LatLRR method, we aim to facilitate the precise identification of damage types within inside the packaged ICs. Firstly, the multiscale Gaussian functions are used to remove the blur and the components are obtained via averaging. Secondly, the image matrices are utilized to extract base matrices and detail matrices. With LatLRR, the multiscale detail matrices are extracted at several representation levels. The final enhanced image is reconstructed by average strategy for dealing with the detail and pre-enhancement parts. The M-LatLRR framework is universal and can be effectively applied to extract multi-level features of packaged IC images. In a comparative analysis, our method demonstrates superior capabilities in determining the failure types of fractured wire bonds, cracks, and delamination of the dielectric layer, outperforming alternative methodologies.

Source: Physica Scripta

DOI: 10.1088/1402-4896/ad8400