A Sparse-Bayesian Approach for the Design of Robust Digital Predistorters Under Power-Varying Operation

Abstract

In this article, a sparse-Bayesian treatment is proposed to solve the crucial questions posed by power amplifier (PA) and digital predistorter (DPD) modeling. To learn a model, the advanced Bayesian framework includes a group of specific processes that maximize the likelihood of the measured data: regressor pursuit and identification, coefficient estimation, stopping criterion, and regressor deselection. The relevance vector machine (RVM) method is reformulated theoretically to be implemented in complex-valued linear regression. In essence, given an initial set of candidate regressors, the result of this sparse-Bayesian learning approach is the most likely model. Experimental results are provided for the linearization of class AB and class J PAs driven by a 30-MHz fifth-generation new radio signal for a fixed average power, where the evolution of the figures of merit versus the number of active coefficients is examined for the proposed sparse-Bayesian pursuit (SBP) algorithm in comparison to other greedy algorithms. The SBP presents a good performance in terms of linearization capabilities and computational cost. Furthermore, the proposed Bayesian framework enabled the design of a DPD model structure, deselect regressors, and readjust coefficients in a direct learning architecture, demonstrating the robustness to changes in the power level over a 10-dB range.

A Sparse Neural Network-Based Power Adaptive DPD Design and Its Hardware Implementation

“…11 show the relationship between the utilization of DSP48 and EVM for FPGA-implemented MP-DPD and ETDNN-DPD with pruning. To check the trade-off between hardware utilization and EVM, we implemented 4 types of DPD on FPGA for both MP-DPDs (i -iv) and ETDNN-DPDs with pruning (a -d) whose parameters are optimized by using the equations ( 12), ( 13) and ( 12), (14), respectively. we can see that ETDNN-DPD required the fewer DSP48s than MP-DPD did for the same EVM.…”

“…However, changing the DPD structure by pruning sometimes leads to the degradation of compensation performance since the pruned DPD structure is calculated by using the training data under the limited condition. To validate the stability of pruned DPDs, pruning based on Bayesian framework for the GMP-DPD [14] demonstrated the robustness to changes in the power level. However, this pruning approach cannot apply to the NN-based DPDs that have multistage connections of neural networks and as far as we know, there is no report about the robustness of pruned NN-based DPDs to changes in the power level at the present time.…”

A Comparison of Surrogate Behavioral Models for Power Amplifier Linearization under High Sparse Data

“…Sparse systems are a recurrent phenomenon that occurs while acquiring PA measurement curves for systems that operate with MIMO technologies and applications related to the new 5G standard [ 17 ]. In recent research [ 18 ], an extension of a sparse-Bayesian treatment is developed to remove non-active regressors and re-estimate the model coefficient in a system with varying power levels. Additionally, in [ 19 ], a dynamic deflection reduction algorithm was developed due to the high performance of beneficial terms and reduced schemes compared to the substantial full Volterra model.…”