Power-ground plane modeling using artificial neural network
Loading...
Date
2019-06
Authors
Low, Chen Eng
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
As technology advances, the increased complexity of electrical devices has
gradually increased the number of variables that affect the output. This raises a
problem whereby conventional simulation-based circuit design tools take time to run
simulation, and circuit simulation must be done every time a design parameter is
changed. Increasing design parameter in circuit design results in increasing time to run
simulation, hence it takes longer time to solve circuit design problem. As a result, it
has become prominent to explore new method to solve integrity analyses, without
sacrificing accuracy and time. For past few decades, artificial neural network (ANN)
neural network has emerged as a popular tool for solving electrical circuit problem.
Neural network’s learning ability has become an alternative for conventional
simulation method. In this work, power-ground plane is modelled based on their
physical design parameter using both conventional method and neural network
method. Sonnet Lite is used as conventional method to model the power-ground plane,
whereas MATLAB is implemented to develop the neural network. Z-Parameter of the
power-ground plane obtained from each method are compared, to justify the findings.
The results indicate that the ANN achieved an accuracy of above 0.9 (with a reference
of 1.0) and successfully modelled power-ground plane.