Memory Correction of a Doherty Power Amplifier with a WCDMA Input using Digital Predistortion

A digital predistortion algorithm is presented that compensates for severe nonlinearities and memory effects introduced by a high efficiency RF power amplifier (PA) with a Doherty structure. The input signal, multicarrier WCDMA, has a wide bandwidth and large crest factor that stimulates the nonlinear modes of the PA and generates significant out-of-band distortion. The predistortion uses filtered nonlinear modes, derived from the input signal, to reduce the out-of-band distortion. Computationally efficient two-stage filters are implemented using static filters chosen to match the statistical properties of the WCDMA signal followed by a smaller set of adaptive FIR coefficients. The estimation favors out-of-band distortion avoiding biases in the predistortion coefficients due to quadrature imbalances. The algorithm performance, demonstrated on a test bench including a Doherty PA, exceeds the ACLR specifications and provides as much as 29 dB of distortion reduction