Coefficient extraction for MPM using LSE, ORLS and SLS applied to RF-PA modeling

Three methods for extracting the behavioral modeling coefficients of the memory polynomial model are compared herein. The first one is the ordinary least square regression, which is widely used for adjusting model parameters; the second is the order recursive least squares, which is suitable for exploring the optimal nonlinearity order and memory depth by comparing subsequent errors while increasing the complexity of the model; and the third is called sequential least square, which is very attractive to be implemented and it only requires identifying the behavior of a power amplifier, and calculating the most accurate model coefficients for each measurement. The equations of the three methods were simulated in Matlab for the NXP 10W power amplifier with complex baseband data, and their implementation was evaluated with normalized mean square error. Also a comparison of their computational complexity based on Halstead metrics is given herein.