A Novel Algorithm for Automated Model Generation of Analog Circuits Using Chebyshev-Newton Interpolation

In this paper a novel automated model generation approach is proposed for macromodeling of analog circuits utilizing a combination of Chebyshev and Newton interpolating polynomials. The technique generates single interpolating polynomial macromodel for the entire training trajectory as compared to existing piecewise linearization approaches that generate multiple linear models that cover full nonlinear trajectories. The macromodels generated using combinations of Chebyshev-Newton (CN) polynomials are efficient in terms of both speed and accuracy than existing macromodeling solutions that utilize Taylor polynomials for linearization of nonlinear systems. We demonstrate our approach through an illustrative example and simulation results confirm that CN macromodels are faster as well as more accurate than Taylor macromodels; the proposed approach is also applicable for the inputs far from training inputs.