Home         Authors   Papers   Year of conference   Themes   Organizations        To MES conference

Direct Learning Digital Predistorters for Power Amplifiers
Authors
	Djigan V.I.
Date of publication
	2020
DOI
	10.31114/2078-7707-2020-3-151-157
Abstract
	The power amplifier linearization problem arises in equipment of communication systems, where an amplifier of a transmitter has to operate in a saturation region, providing a high efficiency, but suffering from nonlinear distortions. The problem is often solved by means of the power amplifier in-put signal predistortion in a way, when the output signal of the predistorter and power amplifier in cascade becomes distortionless. As the predistortion is generally carried out in the digital domain, the technology is called a digital predistortion and the predistortion device is called a Digital Predistort-er (DPD). A DPD produces a nonlinear transfer function, which is an inverse to that of a power amplifier. This action linearizes the Amplitude-to-Amplitude Modulation (AM-AM) and Amplitude-to-Phase modulation (AM-PM) functions of the DPD and power amplifier in cascade. DPDs are generally built on the non-linear adaptive filters base, which are represented by multichannel linear filters. Presently there are two main kinds of the DPDs: with indirect learning and with direct learning. The given paper presents a new direct learning DPD and demonstrates its efficiency comparing with the traditional direct learning DPD. The new DPD is presented in two mathematically equivalent forms. One of the forms proposes an efficient implementation, if DPD output signal delay in samples is much less then total number of weights of an adaptive filter, based on Least Mean Square (LMS) algorithm. The DPD simulation results are presented. The results in form of transient response an power spectral density graphs demonstrate the proposed DPD efficiency when the LMS and Recursive Least Squares (RLS) adaptive filtering algorithms are used comparing with traditional direct learning DPD, which can use only gradient search adaptive filter-ing algorithms like LMS, Normalized LMS (NLMS) or Affine Projection (AP).
Keywords
	Digital Predistorter (DPD), direct learning, non-linear adaptive filter, Least Mean Square (LMS) algorithm, Recursive Least Squares (RLS) algorithm.
Library reference
	Djigan V.I. Direct Learning Digital Predistorters for Power Amplifiers // Problems of Perspective Micro- and Nanoelectronic Systems Development - 2020. Issue 3. P. 151-157. doi:10.31114/2078-7707-2020-3-151-157
URL of paper
	http://www.mes-conference.ru/data/year2020/pdf/D058.pdf