Statistical Compact Model Extraction for Skewed Gaussian Variations

  • V. Janakiraman
  • Shrinivas J. Pandharpure
  • Josef Watts
Conference paper
Part of the Environmental Science and Engineering book series (ESE)

Abstract

A technique for extracting Statistical Compact Model (SCM) parameters for skewed Gaussian parameters is proposed. Existing techniques handle non-Gaussian variations through non-linearity in model equations. However, hardware data on certain technologies suggest that non-Gaussian variations are observed even on linear parameters like Idlin/Idsat. We propose to model such variations through skewed Gaussian random variables. Analytical expressions for the statistics of the skewed Gaussian process and performance parameters are derived. SCM parameters are extracted by setting up a skewed back propagation of variance (SBPV) algorithm.

Keywords

Statistical  Skew Gaussian Model 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    C. C. McAndrew. Efficient statistical modeling for circuit simulation, pages 97–122. Kluwer Academic Publishers, Norwell, MA, USA, 2004.Google Scholar
  2. 2.
    J. Viraraghavan, S.J. Pandharpure, and J. Watts. Statistical compact model extraction: A neural network approach. Computer-Aided Design of Integrated Circuits and Systems, IEEE Transactions on, 31(12):1920–1924, 2012.CrossRefGoogle Scholar
  3. 3.
    I. Stevanovic and C.C. McAndrew. Quadratic backward propagation of variance for nonlinear statistical circuit modeling. Computer-Aided Design of Integrated Circuits and Systems, IEEE Transactions on, 28(9):1428 –1432, sept. 2009.Google Scholar

Copyright information

© Springer International Publishing Switzerland 2014

Authors and Affiliations

  • V. Janakiraman
    • 1
  • Shrinivas J. Pandharpure
    • 1
  • Josef Watts
    • 2
  1. 1.Semiconductor Research and Development CentreIBM India Pvt. Ltd.BangaloreIndia
  2. 2.Semiconductor Research and Development CentreIBM BurlingtonBurlingtonUSA

Personalised recommendations