CN-121981055-A - Simulation method for describing nonlinear distortion of digital predistortion system

Abstract

The invention discloses a simulation method for describing nonlinear distortion of a digital predistortion system, and belongs to the technical field of power amplifier digital predistortion. The method comprises the steps of constructing an MP model, adding partial lead items into the MP model to realize GFMP model, and compensating nonlinear distortion generated by signal amplification by using GFMP model. Compared with the traditional MP digital predistortion model, the GFMP model of the invention has more excellent linearization capability under the condition of little increase of algorithm complexity in the digital predistortion system.

Inventors

• Wei hengzhou
• MA QIANLI
• CHEN WENJIE
• GAO HAONAN
• SUN XIAOMAN
• GAO XIAN
• MA MINGYI
• WEN PEIBO
• ZHAO XIAOSHUAI
• MA ZIANG

Assignees

• 中国电子科技集团公司第五十四研究所

Dates

Publication Date

20260505

Application Date

20260113

Claims (3)

1. 1. A simulation method for describing nonlinear distortion of a digital predistortion system is characterized by comprising the following steps: s1, constructing an MP model; Step S2, adding partial lead items in the MP model to realize GFMP model; and S3, compensating nonlinear distortion generated by signal amplification by using GFMP model.
2. 2. The simulation method for describing nonlinear distortion of a digital predistortion system according to claim 1, wherein the specific manner of step S1 is: (101) Establishing a Volterra series model: In the formula, Representing the memory depth of the model, The non-linear order is represented by a number of non-linear orders, Representing nonlinear order as Volterra series model coefficients at the order, Representing the input signal of the power amplifier corresponding to the memory depth instant, An output signal representative of the power amplifier; (102) Expanding the Volterra series model to obtain: On the basis of the Volterra series model, all cross terms are omitted, only diagonal terms are reserved, and the obtained low-complexity model, namely the MP model, has the following expression: In the formula, 、 Respectively representing the nonlinear order and the memory depth of the model, 、 Representing the power amplifier input and output signals respectively, Is the coefficients of the model.
3. 3. The simulation method for describing nonlinear distortion of a digital predistortion system according to claim 2, wherein the specific manner of step S1 is: (201) Adding partial lead terms into the expression of the MP model to obtain a generalized complete memory polynomial model, namely GFMP model, wherein the expression is as follows: In the formula, 、 Respectively representing the nonlinear order and the memory depth of the model, 、 Respectively representing the input and output signals of the power amplifier, As the coefficients of the model, Representing the advanced memory depth of the model.

Description

Simulation method for describing nonlinear distortion of digital predistortion system Technical Field The invention belongs to the technical field of power amplifier digital predistortion, and particularly relates to a simulation method for describing nonlinear distortion of a digital predistortion system. Background As an essential component in a communication system, a Power Amplifier (PA) has been a hot spot problem in the communication field in research on nonlinear modeling thereof. The nonlinear modeling work of early PA is mainly directed to narrowband signals, so memory effects of PA are often ignored and nonlinear distortion thereof near the saturation region is mainly focused. In this period, a static nonlinear model is often used to model the PA, where representative nonlinear models include a Rapp model, a Saleh model, a taylor series model, etc., which can accurately fit the nonlinear characteristics of the PA under excitation of a narrowband signal. As the demands of communication systems increase and change, the bandwidth of the communication signals increases, and the memory effect of the PA becomes more and more pronounced. The memory effect of PA is embodied in that its output signal is not only related to the input signal at the present moment, but also affected by the input signal at the past moment, and the larger the signal bandwidth, the more pronounced the memory effect. In this period, dynamic nonlinear models are mainly adopted for broadband signal modeling, and the dynamic nonlinear models can be divided into two types, wherein one type is nonlinear modeling comprising linear memory, and the most common models are two-box (two-box) models, including two model structures of Wiener models and Hammerstein models. Both models are obtained by decomposing a nonlinear model with linear memory into a cascade of a linear time-invariant system and a nonlinear system, wherein the structure of the Wiener model is a cascade structure of the linear time-invariant system and the nonlinear system, and the Hammerstein model structure is opposite to the Wiener model and is expressed as a cascade sequence of the nonlinear system and the linear time-invariant system. The other class refers to a nonlinear model containing nonlinear memory, and the model outputs nonlinear characteristics which are closer to those of an actual power amplifier under the excitation of a broadband signal, wherein a representative model is a volt-draw (Volterra) series model and a simplified model thereof. The Volterra series model can accurately describe the memory effect and the nonlinear distortion characteristic of the PA output signal, however, as the memory depth and the nonlinear order increase, model parameters are increased in geometric multiples, so that the complexity of model realization is greatly improved, and the complete Volterra series model is often only applied to a theoretical analysis stage. In order to reduce the complexity of the model and facilitate the implementation of hardware, a series of simplified models of Volterra series models are proposed. The memory polynomial MP model only keeps diagonal terms of the Volterra series model, but omits other cross terms, thereby obviously reducing model coefficients, greatly reducing model realization complexity, and simultaneously ensuring certain model fitting precision, so that the MP model is widely applied to nonlinear modeling research of PA at present. The model obtains better fitting precision at the cost of improving part complexity and is suitable for being applied to power amplifier fitting research with stronger memory effect and nonlinear distortion characteristic. In addition, a learner aims at a weak nonlinear system and provides an envelope memory polynomial (Envelope Memory Polynomial, EMP) model which only delays the signal envelope, so that the number of parameters is further reduced, and the implementation complexity is reduced. Considering the different influences of even-order terms and odd-order terms on nonlinear distortion characteristics, on the basis of an MP model, the predistortion linearization performance of the model is improved by discarding even-order terms with smaller nonlinear influences and replacing the even-order terms with fractional-order terms, and simulation shows that the fractional-order model has higher fitting precision and predistortion linearization performance on the premise of basically the same complexity compared with the MP model. Based on the existing model, even order terms with weak nonlinear influence on PA are omitted, and the characteristics of an MP model, an EMP model and a static envelope (Static Envolope, SE) model are combined, so that a HEMP model is provided by adding a large number of cross terms, and the model is lower in complexity than a GMP model and ensures high accuracy of model fitting. The digital predistortion system aims to generate an inverse model opposite to the nonlinear chara