Nonlinear Relaxation Algorithms for Circuit Simulation

Resve A. Saleh

EECS Department
University of California, Berkeley
Technical Report No. UCB/ERL M87/21
April 1987

http://www2.eecs.berkeley.edu/Pubs/TechRpts/1987/ERL-87-21.pdf

Circuit simulation is an important Computer-Aided Design (CAD) tool in the design of Integrated Circuits (IC). However, the standard techniques used in programs such as SPICE result in very long computer run times when applied to large problems. In order to reduce the overall run time, a number of new approaches to circuit simulation have been developed and are described in this dissertation. These methods are based on nonlinear relaxation techniques and exploit the relative inactivity of large circuits. Simple waveform processing techniques are described to determine the maximum possible speed improvement which can be obtained by exploiting this property of large circuits. Three simulation algorithms are described, two of which are based on the Iterated Timing Analysis (ITA) method and a third based on the Waveform-Relaxation Newton (WRN) method. New programs which incorporate these techniques hake been developed and used to simulate a variety of industrial circuits. The results from these simulations are also provided. The techniques are shown to be much faster than the standard approach. In addition, a number of parallel aspects of these algorithms are described and a general space-time model of parallel task scheduling is developed.

Advisor: A. Richard Newton


BibTeX citation:

@phdthesis{Saleh:M87/21,
    Author = {Saleh, Resve A.},
    Title = {Nonlinear Relaxation Algorithms for Circuit Simulation},
    School = {EECS Department, University of California, Berkeley},
    Year = {1987},
    Month = {Apr},
    URL = {http://www2.eecs.berkeley.edu/Pubs/TechRpts/1987/858.html},
    Number = {UCB/ERL M87/21},
    Abstract = {Circuit simulation is an important Computer-Aided Design (CAD)
tool in the design of Integrated Circuits (IC). However, the
standard techniques used in programs such as SPICE result in
very long computer run times when applied to large problems. In
order to reduce the overall run time, a number of new approaches  
to circuit simulation have been developed and are described
in this dissertation. These methods are based on nonlinear
relaxation techniques and exploit the relative inactivity of large
circuits. Simple waveform processing techniques are described
to determine the maximum possible speed improvement which can
be obtained by exploiting this property of large circuits. Three
simulation algorithms are described, two of which are based on
the Iterated Timing Analysis (ITA) method and a third based on
the Waveform-Relaxation Newton (WRN) method. New programs which
incorporate these techniques hake been developed and used to simulate
a variety of industrial circuits. The results from these simulations
are also provided. The techniques are shown to be much faster than
the standard approach. In addition, a number of parallel aspects
of these algorithms are described and a general space-time model
of parallel task scheduling is developed.}
}

EndNote citation:

%0 Thesis
%A Saleh, Resve A.
%T Nonlinear Relaxation Algorithms for Circuit Simulation
%I EECS Department, University of California, Berkeley
%D 1987
%@ UCB/ERL M87/21
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/1987/858.html
%F Saleh:M87/21