Efficient High-frequency Nonlinear Systems Modeling and Performance Optimization Using Symbolic Computation and Neural Network Techniques

摘要

In this paper we present an original and efficient method for optimum prediction of large signal high-frequency circuit performance using symbolic computation and neural network techniques. The proposed approach consists in two levels. The device level allows optimum input/output powers of all nonlinear devices to be predicted by symbolic computation and then optimum performance to be deduced. Once the optimum harmonic powers are obtained, a neural network model is generated in the circuit level and used to train the circuit performance. The proposed method has been applied to practical circuits design and optimization and their computed performance have been compared successfully to published results.