Modeling and optimization of multilayer LTCC inductors for RF/wireless applications using neural network and genetic algorithms

The design and modeling of multilayer inductors offers considerable challenges to circuit designers because of their complex 3D topology. In this paper, we present a neural network based modeling scheme for multilayer ceramic system-on-package (SOP) inductor library development. A genetic algorithm based optimizer is coupled with the obtained neural network model, for subsequent design and optimization of inductor circuit model parameters. This methodology is validated by characterization data collected from multilayer inductors fabricated in a 12 metal layer low-temperature co-fired ceramic (LTCC) fabrication process. The embedded inductors considered are of great interest for W-CDMA and C-Band applications. The layout parameters predicted by the genetic optimizer match the measured results over the 1-5 GHz range to within 5%. The proposed neuro-genetic algorithm based design promises to minimize the time and cost for multilayer passive design while providing high accuracy.