Predistortion Of RF Power Amplifiers Research Articles

Block-Oriented Recurrent Neural Network for Digital Predistortion of RF Power Amplifiers

Block-Oriented Recurrent Neural Network for Digital Predistortion of RF Power Amplifiers

Complexity-Optimized Adaptive Model for Multiband Digital Predistortion of RF Power Amplifiers

Complexity-Optimized Adaptive Model for Multiband Digital Predistortion of RF Power Amplifiers

Iterative Multimetric Model Extraction for Digital Predistortion of RF Power Amplifiers Using Enhanced Quadratic SPSA

Iterative Multimetric Model Extraction for Digital Predistortion of RF Power Amplifiers Using Enhanced Quadratic SPSA

Dual Feature Indexed Quadratic Polynomial-Based Piecewise Behavioral Model for Digital Predistortion of RF Power Amplifiers

Dual Feature Indexed Quadratic Polynomial-Based Piecewise Behavioral Model for Digital Predistortion of RF Power Amplifiers

Gated Dynamic Neural Network Model for Digital Predistortion of RF Power Amplifiers With Varying Transmission Configurations

The future intelligent transmitter will dynamically adjust the transmission configuration on demand, which will bring new challenges to digital predistortion (DPD). In this article, we present a gated dynamic neural network (GDNN) DPD model to linearize the power amplifier (PA) with varying transmission configurations. The proposed GDNN model is composed of a gating network and a backbone network that can be any NN-based DPD model designed for a fixed configuration. The core idea of the GDNN model is that the backbone model can be dynamically adjusted using the configuration-dependent weights generated by the gating network to achieve transmission configuration-adaptive DPD. To further reduce the running complexity of the GDNN DPD, a sparse GDNN (SGDNN) DPD model is also proposed, which selectively activates the neurons of the backbone network according to the transmission configuration. Experiments are performed with a Doherty PA to validate the proposed method, where the varying transmission configuration includes power level, signal bandwidth (BW), and peak-to-average power ratio. The test results demonstrate that the proposed method can effectively linearize the PA with dynamic transmission configuration and has excellent configuration generalization capability. Moreover, the sparse gating technique can reduce the running complexity of the GDNN DPD by more than 50 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\%$</tex-math> </inline-formula> with only a slight performance loss.

Multi-Output Recurrent Neural Network Behavioral Model for Digital Predistortion of RF Power Amplifiers

In this letter, we propose a method for behavioral modeling and digital predistortion (DPD) of RF power amplifiers (PAs) based on multi-output recurrent neural networks (RNNs). RNN has high modeling accuracy, but it also has high running complexity due to the recurrent mechanism. For this reason, we propose a multi-output model architecture, which means that the DPD model produces multiple adjacent outputs simultaneously for a single input sample group. This approach greatly reduces the running complexity of DPD based on RNNs with essentially no deterioration in performance. The proposed multi-output mechanism is applied to both long short-term memory (LSTM) and gate recurrent unit (GRU), and excellent linearization performances are maintained.

Low Complexity Adaptive Model for Digital Predistortion of RF Power Amplifiers in Time-Varying Configurations

A novel behavioral modeling approach called adaptive model tree (AMT) is proposed for digital predistortion (DPD) of RF power amplifiers (PAs) in fixed and time-varying configurations. The AMT model is piecewise based on the decision tree and the reduced-complexity full basis-propagating selection (RC-FBPS) model. A novel two-step joint iterative algorithm is proposed to achieve a good match between the decision tree and the submodels obtained from the RC-FBPS model. The AMT model inherits and enhances the respective advantages of the decision tree and RC-FBPS model to have a powerful adaptive capability potentially. The experimental tests on a Doherty PA confirm that the AMT model can achieve a better trade-off between linearization performance and complexity than the state-of-the-art model in the fixed configuration. Furthermore, to characterize and compensate for the complex dynamic nonlinear distortions of PAs in time-varying configurations, the piecewise modeling technique in time-varying configurations is proposed and applied to the AMT model in this article. The experimental results confirm that the AMT model achieves excellent linearization performance with very low complexity in time-varying configurations and good generalization performance for new configuration combinations that are not used for training.

A Low Running Complexity Model for Digital Predistortion of RF Power Amplifiers

In this letter, two basis function multiplexing-based behavioral modeling methods for digital predistortion (DPD) of RF power amplifiers (PAs) are proposed to reduce the running complexity of DPD. The proposed full basis-propagating selection (FBPS) model and reduced-complexity FBPS (RC-FBPS) model give two reasonable ways to multiplex even-order basis functions, extending the basis-propagating selection (BAPS) model which only uses basis function delay and odd-order basis functions. The experimental results confirm that both the proposed FBPS and RC-FBPS models can achieve a good tradeoff between running complexity and performance.

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

At present, we are the dawn of a new era for wireless communication systems. The new dynamic approach to the operation of cellular networks requires an extension of the essential input–output hardware mechanisms, to include intelligence. Traditional neural network structures can be used to embed artificial intelligence into cellular network base stations; however, standard network structures are not an optimal solution for these use cases. In this article, we present a neural network structure, which is specifically designed to provide a more accurate and computationally efficient solution compared with the previous neural network solutions for predistortion of RF power amplifiers (PAs). The proposed network structure is directly compared with alternative neural network solutions, which have been successfully employed for digital predistortion (DPD). The operation of this network is validated for DPD with experimental measurements with wideband signals using the latest generation of commercially available RF hardware. The novel network structure proposed in this work is demonstrated, in practice, to have better performance for a normalized mean square error (NMSE), an adjacent channel leakage ratio (ACLR), and an error vector magnitude (EVM) compared with the most popular previously published neural network.

Residual RNN Models With Pruning for Digital Predistortion of RF Power Amplifiers

The paper presents several novel residual recurrent neural network (RNN) models for digital predistortion of radio-frequency power amplifiers. Then, unstructured and structured neural network (NN) pruning algorithms are proposed to reduce the per-sample computational complexity and the required memory, without compromising their modeling performance. The mathematical equations of the complexity and the memory are derived to analyze the hardware cost and design low cost models. An experimental platform is developed to measure the modeling performance in terms of adjacent channel leakage ratio (ACLR), normalized mean square error, error vector magnitude (EVM) and bit error rate. The platform uses an envelope tracking PA excited by 80 MHz IEEE 802.11ac signals. Compared with the state-of-the-art RNN model in (Sun <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">et al.</i> , 2019), the proposed models reduce the per-sample complexity and required memory by <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$94\hbox{--}99$</tex-math></inline-formula> %, they also improve the ACLR and EVM by at least 3 dB and 1 dB at 0.45 dB and 3.15 dB gain compression, respectively. Furthermore, an observation path with time alignment is used to remove the transmit latency in the state-of-the-art approach. At 3.15 dB gain compression, the proposed models outperform the state-of-the-art feedforward and convolutional NN models by at least 1.3 dB in EVM, outperform the general memory polynomial model by at least 3.3 dB in EVM. The above evidences indicate that the proposed RNN models can be implemented with relatively low hardware cost and achieve the best EVM in high power efficiency scenario.

Behavioral modeling and digital predistortion of RF power amplifiers based on time-delay kernel ridge regression

This paper, investigates and presents the behavioral modeling and digital pre-distortion (DPD) of radio frequency power amplifiers (RFPAs) using a time-delay kernel ridge regression (KRR) algorithm. The KRR is an advanced machine learning algorithm that can be effectively used for modeling the baseband characteristics of the RFPA considering both the effects of memory and transistor non-linearity. Compared to the traditional artificial neural network (ANN) based approach which is computationally intensive, the proposed approach using KRR with radial basis kernel function and min-max normalization method, extracts the PA behavioral and DPD model in a short time and yield consistently better results. The performance of the proposed approach in extracting PA and DPD model is demonstrated experimentally on a GaN based class AB power amplifier. The experimental results illustrates that, when compared to ANN based model, the proposed approach yields more accurate PA and DPD models, with an improvement in modeling performance by 2 dB in terms of normalized mean square error (NMSE). In addition, compared to the ANN based approach, the DPD model developed using the proposed approach exhibit improved linearization performance in suppressing spectral regrowth due to PA non-linearity.

Self-Sensing Digital Predistortion of RF Power Amplifiers for 6G Intelligent Radio

The future intelligent communication systems will dynamically adjust the transmitted signal according to the radio environment and human behavior, which will lead to the rapid change of the characteristics of power amplifier (PA) and bring new challenges for digital predistortion (DPD). In this letter, a novel self-sensing DPD (SS-DPD) technique is proposed to linearize PA driven by fast time-varied signals. By automatically sensing the features of input signal and integrating them into the neural network, the proposed model is capable of linearizing the PA operated in such time-varied scenarios without updating DPD coefficients. Furthermore, the polynomial basis functions are embedded into neural network to reduce the complexity. Experimental results on a Doherty PA driven by the fast time-varied signal show that the proposed method can achieve good performance constantly with low complexity.

Block-Oriented Time-Delay Neural Network Behavioral Model for Digital Predistortion of RF Power Amplifiers

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.

Reducing Power Consumption of Digital Predistortion for RF Power Amplifiers Using Real-Time Model Switching

In this article, we propose a new behavioral modeling method to reduce the running complexity and power consumption of digital predistortion (DPD) models for radio frequency power amplifiers. By employing the proposed method, different cross terms in a DPD model can be switched dynamically in real time. Each cross-term branch can further choose the model coefficients from multiple coefficient sets, which improves the DPD performance with little extra complexity. The switch of both cross-term branches and model coefficients is realized using a decision tree-based switch controller, leading to very low run-time complexity. By optimizing the selection process, different input samples can choose the most suitable model configuration that contributes most to the linearization performance. As only one branch is activated at a time, the power consumption can be greatly reduced. Based on the experimental results, the proposed method can achieve excellent linearization performance with significantly reduced power consumption.

From MHz to THz: Systems and Applications

This Special Issue is formed by collecting 46 regular papers on the topic of systems and applications, which reflects the latest progress in microwave systems and applications from MHz to THz. All those 46 papers are divided into six subtopics, namely, digital predistortion of RF power amplifiers (five articles), radar systems and applications (14 articles), wireless systems (nine articles), detectors and imaging (ten articles), microwave applications (five articles), and microwave power transfer technologies (three articles).

Boosted Model Tree-Based Behavioral Modeling for Digital Predistortion of RF Power Amplifiers

In this article, we propose a new behavioral modeling approach, called boosted model tree, to characterize and compensate for the complex nonlinear distortions induced by wideband high-efficiency radio frequency power amplifiers. With the proposed model, the input data are classified into different zones by decision trees and each zone is assigned separate submodels. We also employ a model boosting technique to build multiple parallel tree structures that jointly model the desired nonlinear behavior. By designing dedicated optimization procedures, both tree structures and submodel coefficients can be efficiently identified. It is demonstrated that the combination of piecewise and parallel structures provides a powerful and hardware-efficient way to model nonlinear memory effect and cross terms. Based on the experimental results, the proposed method can achieve improved linearization performance with low hardware complexity under challenging wideband predistortion scenarios.

Instant Gated Recurrent Neural Network Behavioral Model for Digital Predistortion of RF Power Amplifiers

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.

Low complexity output generalized memory polynomial model for digital predistortion of RF power amplifiers

A novel output generalized memory polynomial (OGMP) behavioral model was proposed in this article, which is based on the previous output signal for digital predistortion (DPD) of power amplifiers (PA). Traditional MP or GMP model use polynomials of the previous input signal to characterize memory effect. Although the OGMP model use polynomials of the previous output signal to characterize memory effect. Using the previous output signal to characterize polynomials of the previous input signal, the number of coefficients will decrease. Measurement results show that the proposed OGMP model can achieve the similar effect with less coefficients. In detail, the complexity of OGMP model reduced by about 50% comparing with MP model. Compared with GMP model, the complexity of OGMP model reduced by about 60% with the similar effect.

Direct Error-Searching SPSA-Based Model Extraction for Digital Predistortion of RF Power Amplifiers

This paper presents a low-complexity architecture to extract model coefficients for digital predistortion of radio frequency power amplifiers. The proposed approach directly updates the model coefficients online using a stochastic optimization algorithm that utilizes random perturbation of the model coefficients to determine the coefficient updating direction and converge toward the optimum solution. This technique avoids resource-intensive matrix operations and the requirement for an offline error model in the conventional model extraction techniques and thus drastically reduces the implementation complexity. The complete model extraction solution has been implemented on a field-programmable gate array, and it is shown that the hardware resource usage is remarkably low. Experimental measurements were conducted on a gallium nitride Doherty amplifier excited by Long Term Evolution signals and the results showed that the proposed technique can achieve linearization performance comparable to that obtained by using the conventional and significantly more complex solutions.

A Novel Generalized Parallel Two-Box Structure for Behavior Modeling and Digital Predistortion of RF Power Amplifiers at LTE Applications

This paper presents a generalized parallel two-box structure that is proposed for modeling and digital predistortion of power amplifiers and wireless transmitters exhibiting memory effects. The proposed predistortion scheme consists of two separable boxes; the first is utilized to model the static behavior of the power amplifier, while the second is proposed to consider the memory effect and nonlinear distortion of the power amplifier. The coefficients of the proposed model are identified by applying an indirect learning structure and a least square method. The validation of the proposed model is carried out using the simulation of the power amplifier and the digital predistortion excited by a 64QAM signal in the advanced design system software. According to the simulation results, the criterion of adjacent channel power ratio reduced by about 16 dB. The simulation results reveal an adjacent channel power ratio of almost − 48 dB. Indeed, the proposed model leads to a better performance in terms of spectral regrowth in comparison with the memory polynomial model, and it also reduces the number of coefficients by approximately 22%. This proposed model enables a more accurate modeling of nonlinear distortion and memory effects compared to previous linearization methods.