Semiconductor Process Variation Analysis using Implicit Differentiation Chain Rule  
Author Tongdan Jin


Co-Author(s) Qiyu Huang; Xiulan Cheng; Yisha Xiang


Abstract Modeling and propagating the variance of performance measures becomes a formidable task as today’s integrated circuits consist of multiple layers with dozens of process variables. We propose a top-down variance decomposition method to estimate the variation of system performance considering the correlation of dependent variables across different layers. Using implicit differentiation chain rule, the idea is to sequentially decompose the variance of system performance down to the process level without explicitly computing the covariance of dependent variables at the intermediate levels. Compared with simulations and other analytical methods, the new method reduces the computational time while keeping good estimation accuracy.


Keywords process variation, chain rule, statistical static timing analysis, variance propagation
    Article #:  RQD25-233
Proceedings of 25th ISSAT International Conference on Reliability & Quality in Design
August 1-3, 2019 - Las Vegas, NV, U.S.A.