A new approach for generating a system model from its input-output data is presented. The model is approximated as a linear combination of simple basis functions. The number of basis functions is kept as small as possible to prevent over-fitting and to make the model efficiently computable. Based on these conditions, genetic programming is employed for the generation and selection of the appropriate basis. Since the obtained model can be expressed in simple mathematical expressions, it is suitable for using the model as a macro or behavior model in system level simulation. Experimental results are shown.

Mixed Quasi Newton simulation algorithm that is capable of calculating analog circuits containing mixed level of element models is presented. Conventional circuit simulators usually apply Newton method to solve nonlinear system equations resulted from circuit equations. At each Newton iteration step, it is necessary to reevaluate the Jacobian stamp of circuit elements. However, obtaining the Jacobian stamp of elements described by complex behavior models is a computationally expensive process. To reduce the number of Jacobian evaluations, we combine Newton method and Quasi-Newton method as a new updating scheme. The simulation results show that our algorithm can reduce the number of Jacobian evaluations and improve the simulation time, particularly when simulating circuits containing many behavior model elements.

A new numerical procedure called asymptotic periodic waveform evaluation (APWE) for finding the steady state solution of nonlinear circuits driven by one tone periodic input signal is presented. APWE starts by constructing a virtual system which gives the same periodic steady state waveform as the original system's but with a shorter transient duration. Thus the periodic steady state (PSS) response can be obtained by simply performing transient analysis of the newly derived system for a few periods. An efficient method for solving the nonlinear equations occurred during the transient analysis is presented. To improve the convergence rate of PSS waveform, APWE is combined with extrapolation method. Some simulation results are shown.