31 March 2020 in IEEE Access
This paper proposes two novel vector decomposed neural network models for behavioral modeling and digital predistortion (DPD) of radio-frequency (RF) power amplifiers (PAs): vector decomposed long short-term memory (VDLSTM) model and simplified vector decomposed long short-term memory (SVDLSTM) model. The proposed VDLSTM model is a variant of the classic long short-term memory (LSTM) model that can model long-term memory effects. To comply with the physical mechanism of RF PAs, VDLSTM model only conducts nonlinear operations on the magnitudes of the input signals, while the phase information is recovered by linear weighting operations on the output of the LSTM cell. Furthermore, this study modifies the LSTM cell by adding phase recovery operations inside the cell and replacing the original hidden state with the output magnitudes that are recovered with phase information. With the modified LSTM cell, a low-complexity SVDLSTM model is proposed. The experiment results show that the proposed VDLSTM model can achieve better linearization performance compared with the state-of-the-art models when linearizing PAs with wideband inputs. Besides, in wideband scenarios, SVDLSTM model with much fewer parameters can present comparable linearzation performance compared to VDLSTM model.
9 November 2021 in IEEE Transactions on Microwave Theory and Techniques
A novel block-oriented time-delay neural network (BOTDNN) model for dynamic nonlinear modeling and digital predistortion (DPD) of RF power amplifiers (PAs) is proposed. The proposed model consists of a dynamic linear network and a static nonlinear network to characterize dynamic nonlinear systems. The dynamic linear network simulates multiple linear filters using a fully connected layer with the linear activation function. The static nonlinear network is constructed based on vector decomposition and phase recovery mechanism. To validate the proposed model, experiments have been carried out with two different PAs operating at 2.4 and 39 GHz, respectively. The test results demonstrate that the proposed model has better PA modeling and nonlinear compensation capabilities than state-of-the-art PA behavioral models, while with significantly lower model complexity. Furthermore, to reduce the system cost, we investigate the problems that arise when the neural network-based behavioral models are applied to low feedback sampling rate DPD and propose an improved method. The experiments confirm that the proposed low feedback sampling rate DPD method can effectively alleviate the deterioration of linearization performance caused by undersampling.
9 April 2020 in IEEE Access
This article presents two novel neural network models based on recurrent neural network (RNN) for radio frequency power amplifiers (RF PAs): instant gated recurrent neural network (IGRNN) model and instant gated implict recurrent neural network (IGIRNN) model. In IGRNN model, two state control units are introduced to ensure the linear transmission of hidden state and solve the problem of vanishing gradients of RNN model. In contrast with conventional RNN model, IGRNN can better describe the long-term memory effect of power amplifier, more in line with the physical distortion characteristics of power amplifier. Furthermore the instantaneous gates are used to express the input information implicitly to reduce the redundancy of the input information, and a simpler IGIRNN model is proposed. The complexity analysis indicates that the proposed models have significantly lower complexity than other RNN-based variant structures. A wideband Doherty RF PA excited by 100MHz and 120MHz OFDM signals was employed to evaluate the performance. Extensive experimental results reveal that the proposed IGRNN and IGIRNN models can achieve better linearization performance compared with RNN model and traditional GMP model, and have comparable performance with lower computational complexity compared with the state-of-the-art RNN-based variant models, such as gated recurrent unit (GRU) model.
15 November 2021 in International Journal of RF and Microwave Computer-Aided Engineering
29 November 2021 in IEEE Access
Power amplifiers (PAs) are widely used in RF broadcasting applications. However, they exhibit nonlinear behavior and deteriorate the quality of transmitted signals. Digital predistortion (DPD) is developed to linearize the distortion generated by PAs. Due to the contradiction between modeling precision and complexity, to reduce the complexity and ensure comparable performance, it is necessary to determine an appropriate or optimized structure before applying DPD. Heuristic method is a good method to solve such multivariate problems with a considerable large search domain. In this paper, a novel approach to determine the structure of DPD based on heuristic method is proposed, including the enhanced hill-climbing (EHC) with stronger global search ability and enhanced genetic algorithm (EGA). Ridge regression is introduced to ensure the correctness of the search direction. Besides, orthogonal matching pursuit is used to further reduce the basis number while keeping good performance. The validation of the proposed techniques on a Doherty PA using generalized memory polynomial (GMP) model demonstrates the capability to efficiently find the dimensions of an appropriate GMP model. As shown by the convergence curve and the final model performance, the model searched by the proposed method has satisfactory results, and achieves a good balance between complexity and performance.