• DocumentCode
    2596422
  • Title

    Black-box modelling of conducted electromagnetic emissions by adjustable complexity support vector regression machines

  • Author

    Ceperic, Vladimir ; Gielen, Georges ; Baric, Adrijan

  • Author_Institution
    FER, Univ. of Zagreb, Zagreb, Croatia
  • fYear
    2012
  • fDate
    21-24 May 2012
  • Firstpage
    17
  • Lastpage
    20
  • Abstract
    A black-box method for modelling of conducted electromagnetic emissions (EME) at an integrated circuit (IC) power supply pin and ground level by adjustable complexity support vector regression machines (ACSVR) is presented. The ACSVR provides a basis for representing the nonlinear dynamic conducted EME. The ACSVR enables accurate modelling of conducted EME according to the IEC 61967-4 1Ω method and allows the adjustment of accuracy versus model simulation speed. As a test case, the EME model of conducted emissions of a 242-transistor voltage reference with offset compensation is presented. The resulting models (implemented in Verilog A) are accurate and fast to execute.
  • Keywords
    IEC standards; electromagnetic compatibility; electronic engineering computing; integrated circuit modelling; nonlinear dynamical systems; regression analysis; support vector machines; ACSVR; IC power supply pin; IEC 61967-4 1Ω method; Verilog A; adjustable-complexity support vector regression machines; black-box modelling; conducted electromagnetic emission; integrated circuit power supply pin; model simulation speed; nonlinear dynamic conducted EME model; offset compensation; transistor voltage reference; Complexity theory; Data models; Integrated circuit modeling; Optimization; Predictive models; Support vector machines;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Electromagnetic Compatibility (APEMC), 2012 Asia-Pacific Symposium on
  • Conference_Location
    Singapore
  • Print_ISBN
    978-1-4577-1557-0
  • Electronic_ISBN
    978-1-4577-1558-7
  • Type

    conf

  • DOI
    10.1109/APEMC.2012.6238017
  • Filename
    6238017