Vector Decomposed Long Short-Term Memory Model for Behavioral Modeling and Digital Predistortion for Wideband RF Power Amplifiers

Abstract

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. INDEX TERMS Nonlinear power amplifier, behavioral modeling, digital predistortion, neural network, long short-term memory, vector decomposed.

Digital Predistortion of RF Power Amplifiers With Phase-Gated Recurrent Neural Networks

“…During the experiment, the proposed model, PG-JANET, the vector decomposed LSTM (VDLSTM) as a recurrent network-based model [15], the augmented vector decomposed time-delay neural network (AVDTDNN) [14] as a feedforward network-based model, and GMP as a conventional method are tested and compared. The measurements have been conducted with different test signals that are 60-, 100-, and 200-MHz orthogonal frequency-division multiplexing (OFDM) signals with a 7.7-dB peak-to-average power ratio (PAPR).…”

Digital Predistortion of RF Power Amplifiers With Decomposed Vector Rotation-Based Recurrent Neural Networks

“…Throughout the experiment, the proposed model DVR-JANET was compared with our prior work PG-JANET, DVR model itself, and GMP as the conventional model. It is worth to mention that we have compared PG-JANET model with other neural network-based models, particularly the state-of-the-art time-delayed neural network and LSTM-based models, such as the augmented vector-decomposed time-delayed network (AVDTDNN) model [15] and the vector decomposed long short-term memory (VDLSTM) model [17], in [22]. To avoid replication, we did not conduct tests with other models, because they are all worse than the PG-JANET.…”

“…Relying on the impactful success of deep learning and its enormous potential for improvement, neural network learning emerges an appealing alternative in PA modeling. Recently, many neural network-based DPD models have been proposed [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20]. When deploying a deep learning model, selecting a proper network structure to accurately model nonlinear dynamics of the PA becomes significantly challenging.…”

“…Recurrent models can involve an infinite number of past samples into the learning and, therefore, potentially can lead to higher accuracy [21], but the conventional recurrent networks suffer from stability issues. Some recent works have proposed recurrent DPD models [16], [17], [18], [19], but none of the revised networks concentrate on the search for the most suitable network structure. In [22], a useful link between the lightweight Just Another NETwork (JANET) [23] and the physical behavior of the PA is established.…”

Segmented Spline Curve Neural Network for Low Latency Digital Predistortion of RF Power Amplifiers

“…While the performance of these techniques is promising, additional validation would be required to show their viability for DPD. More recently, approaches, such as the vector decomposed time-delayed neural network (VDTDNN) [9] and the vector decomposed long short-term memory (VDLSTM) [10], have been proposed and shown to have good performance. These approaches, while well performing, require additional high-latency operations on top of what is required for RVTDNN.…”