Extended Hammerstein model for RF power amplifier behavior modeling

This paper presents a novel extended Hammerstein model for predicting the dynamic nonlinearity of wideband radio frequency power amplifier (RFPA). This new model tackles the inaccuracy of conventional Hammerstein scheme in accounting for the short term memory effects of the RFPA by adding an extra branch. For that, an error signal between the output and the input signal of the memoryless submodel is first computed and processed by an additional filter. The filtered error signal is then post-injected at the output of the model. This extra modeling mechanism leads to remarkable accuracy when modeling two different RFPAs operating at different biasing points and driven with 4 carrier (4C) WCDMA signals. Despite its simple structure and identification algorithm, the extended Hammerstein model demonstrated excellent capacity in mimicking the dynamic characteristics (AM/AM & AM/AM) and the output signal spectrum of the RFPA under test.