• DocumentCode
    3093955
  • Title

    Statistical decoupling capacitance allocation by efficient numerical quadrature method

  • Author

    Eguia, Thom Jefferson A ; Mi, Ning ; Tan, Sheldon X D

  • Author_Institution
    Dept. of Electr. Eng., Univ. of California, Riverside, CA
  • fYear
    2009
  • fDate
    16-18 March 2009
  • Firstpage
    309
  • Lastpage
    316
  • Abstract
    In this paper, we propose a novel statistical decap allocation method to reduce the voltage drop noise in the presence of variational leakage current sources. The new method can derive the closed form of decoupling capacitance (decaps) in terms of variational parameters from the variational leakage currents. It treats the deterministic decap method as a black box and can work with any existing simulation-based methods. The new method employs the orthogonal polynomials to represent the variational gate leakages in a closed form first. Decap distributions are then computed by a fast multi-dimentional Gaussian quadrature method with sparse grid technique. Experimental results show that the proposed method can deliver order of magnitudes speedup over the Monte Carlo method with almost the same accurancy.
  • Keywords
    leakage currents; microprocessor chips; polynomials; statistical analysis; deterministic decap method; fast multi-dimentional Gaussian quadrature method; numerical quadrature method; sparse grid technique; statistical decap allocation method; statistical decoupling capacitance allocation; variational leakage current sources; Capacitance; Distributed computing; Gate leakage; Grid computing; Leakage current; Noise reduction; Polynomials; Random variables; Threshold voltage; Very large scale integration;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Quality of Electronic Design, 2009. ISQED 2009. Quality Electronic Design
  • Conference_Location
    San Jose, CA
  • Print_ISBN
    978-1-4244-2952-3
  • Electronic_ISBN
    978-1-4244-2953-0
  • Type

    conf

  • DOI
    10.1109/ISQED.2009.4810313
  • Filename
    4810313