39. A PARALLEL ANALYSIS FLOW FOR POWER DISTRIBUTION NETWORKS USING DISCRETE FOURIER TRANSFORM

Department: Computer Science & Engineering
Faculty Advisor(s): Chung-Kuan Cheng

Primary Student
Name: Xiang Hu
Email: x2hu@ucsd.edu
Phone: 858-534-8174
Grad Year: 2012

Abstract
A frequency domain simulation flow for large power distribution network (PDN) analysis is proposed in this work. Based on given PDN linear circuit models and time-domain load current input profiles, the proposed flow first converts the time-domain input currents to their frquency domain representations using discrete Fourier transform (DFT). Then frequency-domain voltage response of the PDN is computed using efficient linear matrix solvers. Finally, time-domain voltage response of the PDN is obtained by inverse discrete Fourier transform (IDFT). The proposed flow not only allows designers to gain an insight to the frequency domain characteristics of the PDN but also obtains accurate time domain voltage responses according to different load switching profiles. The contributions of this work are as follows. (1) The simulation results with the proposed flow have an error of less than 4% compared to HSPICE. (2) The closed-form mean square error bounds are derived to validate the accuracy of the proposed flow. (3) Parallel processing is incorporated to the flow which results in a significant reduction in total simulation time. Experimental results show that nearly 1800x speedup is achieved compared to HSPICE transient simulation. (4) It is demonstrated that for certain input profiles the proposed flow obtains moare accurate results than HSPICE.

« Back to Posters or Search Results