Study of RF power amplifier behavior models based on BP improved algorithm

How to model the power amplifier behavior accurately is the key to system-level simulation. BP neural network can be used to simulate random nonlinear system, but it easily falls into the local minimum points and has no enough precision. So, this article proposes two improved model based on BP neural network model, one is cascading model BP-RBF, and the other is PSO_BP neural network. Design amplifier circuit in ADS2009 utilizing the freescale semiconductor chip MRF6S21140, and then extract voltage data as the simulation data. Carry on the MATLAB fitting simulation by BP, BP-RBF as well as PSO_BP, compared with voltage RMS error (RMSE), epochs and convergence time. Eventually, the results show that the improved algorithm BP-RBF, PSO_BP models have better fitting function than BP model, and fit the characteristics of power amplifier accurately, then have the important application value to construct system simulation.