I.E. Kashchenko1, A.P. Pavlov2, A.V. Bakhmutskaya3

1-3 Omsk Scientific Center SB RAS (Institute of Radiophysics and Physical Electronics) (Omsk, Russia)

At design the modern digital predistortion input systems, it is necessary to take into account the adjustment of their parameters over time. This is because the characteristics of the amplifying path can change under the influence of external and internal factors. The main external factor is the ambient temperature, the main internal factor is the degradation of the parameters of the active and passive components of the amplifying path. When using digital pre-distortion input systems based on an analytical description of the nonlinear properties of a power amplifier, the most optimal, from the point of view of suppressing nonlinear distortions, is the use of algorithms with indirect learning, in which the parameters are adapted. From the point of view of the linear least squares problem, there are several well-established methods applicable for adapting the parameters of the predistortion system, each of which differs in speed and quality of non-linear distortion suppression along with the amount of computed resources. The method of least squares has the least efficiency, however, it requires a minimum amount of computing resources, which is an important indicator in the design of modern digital pre-distortion input systems as part of small-sized and low-power FPGA and VLSI. So, the study of methods and ways to reduce the convergence time of the adaptation algorithm based on the least squares method is a relevant task.

The aim of the article is to study methods and ways to reduce the convergence time of the adaptation algorithm based on the least squares method.

The article presents the implementation of the normalized least squares method for adapting the parameters of the digital pre-distortion input system. The use of the normalized least squares method makes it possible to reduce the convergence time of the adaptation algorithm in comparison with algorithms based on the ordinary least squares method. In addition, the proposed algorithm uses the adaptation of the step constant in time, which depends on the result of the error signal at the output of the system. The arithmetic operations used in the proposed algorithm are adapted for implementation as part of FPGA and VLSI.

Kashchenko I.E., Pavlov A.P., Bakhmutskaya A.V. The algorithm of adaptation of the input system of predistortion based on the
normalized least squares method. Radiotekhnika. 2022. V. 86. № 8. P. 21−28. DOI: https://doi.org/10.18127/j00338486-202208-02 (In Russian)

1. Eun C., Powers E.J. A new Volterra predistorter based on indirect learning architecture. IEEE Trans. on Signal Processing. 1997. V. 45. № 1.
2. Gan L., Abd-Elrady E. Digital predistortion of memory polynomial systems using direct and indirect learning architectures. Proc. of the IASTED Conference on Signal and Image Processing. 2009.
3. Ding L., Ma Z., Morgan D.R., Zierdt M., Pastalan J. A least square/Newton method for digital predistortion of wideband signals. IEEE Trans. on Communications. May 2006. V. 54. № 5. Р. 833–840.
4. Gan L. Adaptive digital predistortion of nonlinear systems. Ph.D. Thesis, Faculty of Electrical and Information Engineering. Graz University of Technology. Graz, Austria. 2009.
5. Yang B., B¨ohme J.F. Rotation-based RLS algorithms: Unified derivations, numerical properties, and parallel implementations. IEEE Trans. on Signal Processing. May 1992. V. 40. Р. 1151-1166.
6. Shoaib M., Werner S., Apolin´ario J.A. (Jr.), Laakso T.I. Solution to the weight extraction problem in fast QR-decomposition RLS algorithms. Proc. IEEE Intern. Conf. on Acoust. Speech, Signal Processing. Toulouse, France. 2006. Р. III-572-III-575.
7. Dzhigan V.I. Istorija, teorija i praktika adaptivnoj obrabotki signalov. Sb. trudov «Problemy razrabotki perspek-tivnyh mikro- i nanojelektronnyh sistem - 2012». Pod obshh. red. akad. RAN A.L. Stempkovskogo. M.: IPPM RAN. 2012. S. 30-37 (In Russian).
8. Conte E., de Maio A., Ricci G. Covariance matrix estimation for adaptive CFAR detection in compound-Gaussian clutter. IEEE Trans. Aerosp. Electron. Syst. 2002.V. 38. № 2. Р. 415-426.
9. Montoro G., Gilabert P.L., Bertran E., Cesari A., Garca J.A. An LMS-Based Adaptive Predistorter for Cancelling Nonlinear Memory Effects in RF Power Amplifiers. Proc. Asia Pacific Microwave Conference (APMC'07). Bangkok, Thailand. Dec. 2007.
10. Parhami B. Computer Arithmetic: Algorithms and Hardware Designs. N.Y.: Oxford University Press. 2000.
11. Morgan D.R., Ma Z., Kim, Zierdt, Pastalan J. A generalized memory polynomical model for digital predistortion of RF power amplifiers. IEEE Trans. Sig. Proc. 2006. V. 54. Р. 3852-3860.