An Integrated Scheme of FIR and Augmented Real-Valued Time-Delay Neural Network of Harmonic Cancellation Digital Predistortion Model for High-Frequency Power Amplifier

Intermodulation distortion (IMD) and higher-order harmonics caused by the nonlinearity of power amplifiers (PAs) are the main problems of multioctave PAs used in high-frequency (HF) transmitters. This letter proposes a new digital predistortion (DPD) model that integrates FIR low-pass filter characteristics and augmented real-valued time-delay neural network (ARVTDNN) based on the characteristics of shortwave nonlinear distortion signals to improve baseband IMD and suppress higher-order harmonics of multioctave HF PA. Based on the ARVTDNN structured DPD model, the FIR low-pass filtering architecture in the rate adaptation submodule of the baseband processing in the HF transmitter is further integrated and reused. The experimental results on actual HF PA show that compared with the traditional memory polynomial model (MPM), general memory polynomial (GMP), and ANN-based models, the proposed method reduces baseband IMD without increasing algorithm complexity, while the second and third harmonics are below - 67 dBc.