Independent Digital Predistortion Parameters Estimation Using Adaptive Principal Component Analysis

Abstract

This paper presents an estimation/adaptation method based on the adaptive principal component analysis (APCA) technique to guarantee the identification of the minimum necessary parameters of a digital predistorter. The proposed estimation/adaptation technique is suitable for online field-programmable gate array or system on chip implementation. By exploiting the orthogonality of the resulting transformed matrix obtained with the APCA technique, it is possible to reduce the number of coefficients to be estimated which, at the same time, has a beneficial regularization effect by preventing illconditioning or overfitting problems. Therefore, this identification/adaptation method enhances the robustness of the parameter estimation and simplifies the adaptation by reducing the number of estimated coefficients. Due to the orthogonality of the new basis, these parameters can be estimated independently, thus allowing for scalability. Experimental results will show that it is possible to determine the minimum number of parameters to be estimated in order to meet the targeted linearity levels while ensuring a robust well-conditioned identification. Moreover, the results will show how thanks to the orthogonality property of the new basis functions, the coefficients of the digital predistorter can be estimated independently. This allows to tradeoff the digital predistorter adaptation time versus performance and hardware complexity.

Dynamic Selection and Estimation of the Digital Predistorter Parameters for Power Amplifier Linearization

“…The dynamic DPD parameter update scheme using the PCA-DPLS method has been experimentally evaluated considering the complex waveform scenario and the test setup detailed in [20]. Four 20-MHz bandwidth and 64-QAM modulated fastconvolution filter bank multi-carrier (FC-FBCM) signals, each with different sub-carrier group deactivation configurations, have been carrier aggregated.…”

“…Some of the aforementioned feature extraction techniques have been proposed by the authors in [20]- [22] as an alternative to one of the most commonly used solutions based on QR factorization combined with recursive least squares (QR-RLS) [23]. The objective of using these techniques is not only to ensure a proper well-conditioned estimation, but also a reduction in the number of parameters required in the identification or adaptation process.…”

“…For that reason, in [22] the authors proposed the combination of an off-line OMP search for reducing the DPD basis in the forward path with PLS extraction in the adaptation (or observation) path. In comparison to PCA [20], where the new basis of components with maximal variance are obtained taking into account only the input data, with PLS the new basis of components show maximal covariance in relation to the signal to be estimated and thus, the reduction capabilities using PLS are significantly better than using PCA, as discussed in [22]. In [24], the authors extended the reduction capabilities by presenting a dynamic adaptation approach based on PLS where the basis of new components used in the DPD estimation is dynamically adjusted and thus, at every iteration of the adaptation process, only the minimum number of required components are used to minimize the linearization error.…”

“…In this paper, the authors go a step further by proposing a new approach to dynamically estimate and update the coefficients of the DPD that not only shows better dimensionality reduction capabilities than previous published solutions in [20], [22], [24], but also introduces less computational complexity than QR-LS (i.e., any method based on a QR decomposition with a fixed number of coefficients used to solve the LS problem, such as Matlab's mldivide or backslash) thanks to the inherent dimensionality reduction introduced by the proposed algorithm.…”

Sparse Identification of Volterra Models for Power Amplifiers Without Pseudoinverse Computation

“…In this framework, a special interest has risen in lowcomplexity DPD implementations [14]- [20] due to the desired use of limited resources and memory in field-programmable gate arrays (FPGAs) and the numerical stability benefits observed from the signal processing standpoint.…”