Search

Due to the importance of metal layers in the product yield, serpentine test structures are usually fabricated on test chips to extract parameters for yield prediction. In this paper, the confidence level and estimation precision of the average defect density on metal layers are investigated to minimize the randomness of experimental results and make the measured parameters more convincing. On the basis of the Poisson yield model, the method to determine the total area of all serpentine test structures is obtained using the law of large numbers and the Lindeberg-Levy theorem. Furthermore, the method to determine an adequate area of each serpentine test structure is proposed under a specific requirement of confidence level and estimation precision. The results of Monte Carlo simulation show that the proposed method is consistent with theoretical analyses. It is also revealed by wafer experimental results that the method of designing serpentine test structure proposed in this paper has better performance.

A novel method for fault Diagnosis of High-current converter, which is constructed on the basis of Wavelet Neural Network and Improved Adaptive Genetic Algorithms (IAGA), is presented here. In the proposed method, IAGA is employed to optimize the structure and the parameters of WNN and enhance the complexity, convergence and generalization ability of the network. By training and testing under MATLAB/SIMULINK, it is clearly shown that WNN based on IAGA performs better than WNN based on BP（Back-Propagation Neural Network）as well as linear adaptive GA (LAGA).