ARVTDNN-Based Digital Predistortion for High-Power Amplifier Nonlinearity Compensation in Broadband Radiated Immunity Test

This paper proposes the use of test signals with an applied augmented real-valued time-delay neural network (ARVTDNN)-based digital predistortion (DPD) to mitigate test signal distortion caused by the nonlinearity of high-power amplifiers (HPAs) in radiated immunity (RI) tests. Orthogonal frequency division multiplexing (OFDM)-based broadband signals were generated to obtain the input and output IQ data of the HPA, and these data were used to build two neural network (NN) models: one for the HPA and the other for the predistorter. Based on these models, DPD simulations were conducted, and the results demonstrated that the signal distortion was effectively compensated, as verified through amplitude modulation (AM) and phase modulation (PM) characteristics and the power spectral density (PSD). Specifically, the simulations showed that the DPD restores the crest factor (CF), one of the key test signal specifications defined in the IEC 61000-4-41, thereby recovering the original input signal characteristics. In conclusion, this paper proposes a method for establishing a robust test setup for broadband RI test and suggests the potential for future research to analyze whether the CF is a risk factor for the EUT.