| | GSRC Student Profile:
Research Overview: A Hierarchical Technique for Efficient Simulation of Large-Scale RF/Mixed-Signal Systems
As sizes of integrated circuits keep growing, traditional Newton-Raphson algorithm for solving nonlinear equations is greatly challenged by vulnerability of convergence and huge memory requirements. To cope with these issues, multilevel Newton algorithm (MLNA) has been reported. In this work, we propose a generic hierarchical equation framework for differential-algebraic equations based on MLNA, which enables mixed-domain analyses and could be extended to various types of simulation. Hierarchical DC, transient and harmonic balance analyses have been developed thereafter. To demonstrate accuracy and efficiency of this method, different simulations of large scale circuits are performed. Experimental results show that while high accuracy is reached, our method contributes to 10x speedup and 100x memory saving for systems with large size, comparing to traditional flattened methods.
| |
|
