Adjacent multiband digital pre-distortion system and method

Abstract

The invention relates to the technical field of digital pre-distortion, in particular to an adjacent multiband digital pre-distortion system and method, and the method comprises the steps: S1, extracting an intermodulation feature vector of an adjacent multiband input signal, and constructing a nonlinear coupling feature matrix based on the intermodulation feature vector; and S2, generating a frequency band cooperative compensation signal by using the nonlinear coupling characteristic matrix, wherein the frequency band cooperative compensation signal is used for compensating an asymmetric cross modulation effect between adjacent frequency bands. And S3, constructing a dynamic pre-distortion parameter optimization model based on the distortion characteristic analysis result of the frequency band cooperative compensation signal and the power amplifier output signal, and generating a final pre-distortion output signal through real-time feedback adjustment. According to the method, the nonlinear primary function set including the high-order intermodulation item and the inter-band cross modulation component is constructed, and the nonlinear coupling feature matrix extracted by singular value decomposition is combined, so that inter-band collaborative modeling and compensation signal generation are realized, and the adaptive capacity under the complex working condition is enhanced.

Description

Technical Field

[0001] This invention relates to the field of digital predistortion technology, and in particular to an adjacent multi-band digital predistortion system and method. Background Technology

[0002] In modern wireless communication systems, spectrum resources tend to be fragmented and highly aggregated, with a large number of adjacent frequency bands being used to carry high-speed data signals. To improve spectrum utilization efficiency, operators generally adopt multi-carrier aggregation and frequency band superposition technologies, which leads to power amplifiers facing composite drive inputs from multiple frequency bands. However, under nonlinear conditions, traditional power amplifiers will produce varying degrees of cross-modulation and intermodulation distortion on these frequency band signals. In particular, asymmetric inter-band modulation interference will seriously affect modulation accuracy, spectrum purity, and adjacent channel performance, becoming a challenge for system linearization control.

[0003] Existing digital predistortion techniques mostly establish nonlinear compensation models for single frequency bands, typically relying on memory polynomials or neural network structures to model amplitude-phase distortion. While these methods perform well in single-frequency scenarios, they struggle to effectively capture the asymmetric coupling behavior between frequency bands when faced with complex cross-modulation paths in adjacent multi-frequency band environments. The compensation strategies lack synergy, leading to decreased modeling accuracy and severe distortion residue. Furthermore, traditional methods often employ static parameter optimization, failing to dynamically adjust the compensation strategy based on system state changes, thus lacking sufficient real-time performance and adaptability. Summary of the Invention

[0004] This invention provides an adjacent multi-band digital predistortion system and method.

[0005] A method for adjacent multi-band digital predistortion includes the following steps: S1: Extract the intermodulation feature vectors of adjacent multi-band input signals, and construct a nonlinear coupling feature matrix based on the intermodulation feature vectors; S2: Generate a band-coordinated compensation signal using the nonlinear coupling feature matrix. The band-coordinated compensation signal is used to compensate for the asymmetric cross-modulation effect between adjacent bands. S3: Based on the distortion characteristic analysis results of the frequency band collaborative compensation signal and the power amplifier output signal, a dynamic predistortion parameter optimization model is constructed, and the final predistortion output signal is generated through real-time feedback adjustment.

[0006] Optionally, S1 includes: S11: Perform nonlinear transformation on adjacent multi-band input signals to construct a basis function set including multiple order intermodulation terms. The basis function set includes self-modulation components and cross-modulation components of signals from different frequency bands. S12: Perform cross-correlation calculations between each basis function signal in the basis function set and the output signal of the power amplifier to obtain a cross-correlation vector representing the coupling strength between each intermodulation term and the output signal; S13: The cross-correlation vectors at multiple time points are concatenated to construct a cross-correlation matrix that includes time-series variation features. Singular value decomposition is then performed on the cross-correlation matrix to extract the singular values ​​corresponding to the dominant components as a measure of the nonlinear coupling strength between frequency bands, thus forming a nonlinear coupling feature matrix.

[0007] Optionally, S13 includes: S131: Stack and concatenate the cross-correlation vectors extracted at multiple different times to construct a cross-correlation matrix that includes time-series change information; S132: Perform singular value decomposition on the cross-correlation matrix, extract the singular value features corresponding to the dominant components, and construct them into a nonlinear coupling feature matrix.

[0008] Optionally, the singular value decomposition process extracts the dominant components of the nonlinear coupling features, and determines the coupling strength between frequency bands based on the energy proportion of each singular value, so as to generate a nonlinear coupling feature matrix that reflects the dominant nonlinear interaction relationship between frequency bands.

[0009] Optionally, S2 includes: S21: Construct a nonlinear basis function vector based on the input signals of each frequency band, wherein the nonlinear basis function vector includes a modulation term composed of multiple frequency band signals and their nonlinear combination forms; S22: Map the nonlinear coupling feature matrix to the nonlinear basis function vector to obtain the initial compensation component for compensating nonlinear distortion in different frequency bands; S23: Apply dynamic adjustment weights to the initial compensation components, and adaptively adjust the compensation intensity of each component according to the current system operating conditions to offset the asymmetric cross-modulation effect between adjacent frequency bands and generate the final frequency band collaborative compensation signal.

[0010] Optionally, S21 includes: S211: Perform nonlinear transformation on the input signals of each frequency band to generate basis function components including different power forms; S212: Combine multiple frequency band signals and their complex conjugate forms to generate a nonlinear basis function vector including cross-modulation terms.

[0011] Optionally, the asymmetric cross-modulation effect includes: Intermodulation interference occurs between adjacent frequency band signals due to the nonlinearity of the power amplifier. The distortion in one frequency band mainly originates from the modulation effect of the other frequency band. The distortion distribution across different frequency bands is not symmetrical, including coupling characteristics of inconsistent modulation path directionality and uneven spectral energy distribution.

[0012] Optionally, S3 includes: S31: Perform multi-dimensional distortion feature analysis on the output signal of the power amplifier and extract relevant indicators that constitute the distortion feature vector. The relevant indicators include spectral features, modulation accuracy indicators, dynamic range information and statistical characteristics. S32: Based on the distortion feature vector and the frequency band collaborative compensation signal, a dynamic predistortion parameter optimization model with a multi-objective optimization structure is constructed. The dynamic predistortion parameter optimization model takes multiple performance indicators as optimization objectives and sets weight factors to reflect the priority relationship of different optimization objectives, so as to realize the dynamic adjustment of the predistortion control parameters. S33: The optimization model is solved in real time using the minimum mean square error algorithm. The predistorter parameters are continuously updated iteratively based on the current feedback results. The updated predistorter parameters are then applied to the original input signal to generate the final predistorted output signal.

[0013] Optionally, S33 includes: S331: The minimum mean square error algorithm is used to solve the dynamic predistortion parameter optimization model in real time. The control parameters in the predistorter are iteratively updated based on the feedback error between the power amplifier output signal and the target response. S332: Apply the updated control parameters to the original input signal and generate the final predistorted output signal through the predistortion function structure.

[0014] An adjacent multi-band digital predistortion system, used to implement the above-mentioned adjacent multi-band digital predistortion method, includes the following modules: Intermodulation feature extraction module: performs nonlinear analysis on adjacent multi-band input signals, extracts intermodulation feature vectors, and constructs a nonlinear coupling feature matrix based on the intermodulation feature vectors; Frequency band cooperative compensation generation module: receives the nonlinear coupling feature matrix, constructs a compensation vector including cross-modulation terms, and generates a frequency band cooperative compensation signal for suppressing asymmetric modulation effects between adjacent frequency bands; Dynamic predistortion optimization module: Based on the distortion characteristic analysis results of the frequency band collaborative compensation signal and the power amplifier output signal, a dynamic predistortion parameter optimization model is constructed, and the final predistortion output signal is generated through a real-time feedback optimization mechanism.

[0015] The beneficial effects of this invention are: This invention constructs a set of nonlinear basis functions including higher-order intermodulation terms and inter-band cross-modulation components, and combines it with the nonlinear coupling feature matrix extracted by singular value decomposition to achieve inter-band collaborative modeling and compensation signal generation. This effectively suppresses the asymmetric cross-modulation effect that traditional single-frequency predistortion methods cannot cover, improves the linearization effect in multi-band scenarios such as carrier aggregation, significantly reduces adjacent channel leakage and modulation error, realizes a refined compensation mechanism for the coupling characteristics of adjacent multi-bands, and significantly improves predistortion accuracy and spectral purity.

[0016] This invention designs a dynamic predistortion parameter optimization model that integrates the results of frequency band collaborative compensation with the power amplifier output distortion characteristics. Combined with the minimum mean square error algorithm, it realizes the stepwise adaptive update of parameters, which can achieve rapid convergence and stable control in response to changes in power amplifier characteristics, channel fluctuations or modulation mode switching. This ensures the robustness and real-time response capability of the system under various actual operating conditions. A dynamic parameter update mechanism based on multi-objective optimization and real-time feedback adjustment is constructed to enhance the adaptive capability and stability under complex operating conditions. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a method according to an embodiment of the present invention; Figure 2 This is a system block diagram of an embodiment of the present invention. Detailed Implementation

[0019] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. Those skilled in the art may employ other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0020] like Figure 1 As shown, an adjacent multi-band digital predistortion method includes the following steps: S1: Extract the intermodulation feature vectors of adjacent multi-band input signals, and construct a nonlinear coupling feature matrix based on the intermodulation feature vectors; S1 specifically includes: S11: Input signal to adjacent multi-band signals A nonlinear transformation is performed to generate a set of basis function signals including intermodulation components of various orders, represented as: ; in, It is the first The input signal of each frequency band represents the complex envelope of different subcarriers or frequency band channels, and is used to model multi-band composite signal input. This is the highest nonlinear order, ranging from 3 to 5, which can cover most dominant intermodulation components (such as IM3, IM5, and IM7). Higher orders will increase the computational burden. This refers to the number of frequency bands, ranging from 2 to 4. Scenarios with multiple adjacent frequency bands typically involve 2 to 4 subcarriers. It is the first The complex conjugate of the input signal in each frequency band It is the power order, controlling the order of the intermodulation basis functions. Used to construct cross terms between multiple frequency bands, including self-coupling and intermodulation components; In scenarios where adjacent multi-band signals drive power amplifiers, due to the nonlinear characteristics of the power amplifier itself, cross-modulation effects will occur between different frequency bands, which will introduce intermodulation distortion into the output. Traditional digital predistortion (DPD) methods often only consider polynomial modeling of a single frequency band, making it difficult to accurately characterize the interaction between multiple frequency bands.

[0021] To address this issue, this step systematically constructs a set of basis function signals containing intermodulation components of various orders by performing nonlinear power transformations on each frequency band signal and considering its complex conjugate combination with signals from other frequency bands. These basis functions can not only characterize the self-modulation effects of a single frequency band (such as third-order and fifth-order distortion), but also effectively cover the coupling components between frequency bands.

[0022] In radio frequency systems, intermodulation products originate from the nonlinear interaction between signals in multiple frequency bands. Their mathematical expression is naturally manifested as the power product and conjugate product of signals in different frequency bands. Therefore, by using the power combination of multi-band complex envelope signals, the distortion modes that may appear in the power amplifier output can be systematically restored. By limiting the total number of powers to no more than a certain order, it is possible to ensure that the main intermodulation components are included in the constructed basis function set while controlling the computational complexity, thus providing a solid input foundation for subsequent feature extraction and compensation models. The basis function set is not used to directly regress the output, but rather serves as the basis for the next step of cross-correlation calculation. It helps to extract the characteristic response intensity corresponding to each intermodulation term from the output signal, and then deduce the nonlinear coupling structure between frequency bands.

[0023] In summary, the design of this step not only conforms to the theory of multi-band signal intermodulation, but also lays the foundation for the subsequent construction of the coupling matrix and the frequency band collaborative compensation mechanism.

[0024] S12: Convert each basis function signal With power amplifier output signal Perform cross-correlation calculations to obtain a cross-correlation vector. Stack all cross-correlation results to form a new cross-correlation vector, represented as: ; ; in, For the first Each basis function component is used for cross-correlation analysis and compared with the output signal to quantify the degree of influence. This is the output signal of the power amplifier, the actual output signal of the power amplifier, which includes non-ideal characteristics such as intermodulation distortion and memory effect. This is the integration window length (or sampling duration), ranging from 0.01 to 10. Too short a window will compromise statistical stability, while too long a window will increase latency and computational burden. It is the number of basis functions (i.e. (the base number) It is a cross-correlation vector used to quantify the nonlinear coupling strength between each intermodulation term and the output. For the first A number of cross-correlation coefficients, the magnitude of which depends on the degree of intermodulation; After a multi-band signal passes through a nonlinear power amplifier (such as a Doherty amplifier), the output signal includes not only the original components of each band but also intermodulation distortion caused by nonlinear effects. The set of intermodulation basis functions constructed in the previous step covers potential cross-modulation terms, but it cannot determine which terms truly dominate the nonlinear behavior of the current system.

[0025] Therefore, this step compares each basis function with the actual output signal and calculates its response strength in the output. This process is essentially a matched filtering or projection analysis, which can quantify the degree of expression of each intermodulation basis function in the power amplifier output. The higher the cross-correlation value, the greater the influence of the nonlinear mode represented by the basis function on the output signal, reflecting stronger band coupling or nonlinear modulation components.

[0026] By calculating the cross-correlation of all basis functions and arranging them into a cross-correlation vector, we can not only capture all intermodulation features, but also provide a foundation for subsequent dimensionality reduction and modeling operations such as principal component extraction and singular value decomposition.

[0027] S13: Cross-correlation vectors Perform singular value decomposition (SVD) to construct a nonlinear coupled characteristic matrix, represented as: ; in, It uses a sliding window to view multiple moments. The multi-time-dimensional cross-correlation matrix formed by splicing is used to enhance robustness and time-dimensional statistical property analysis. It is a left singular vector matrix, representing the feature dimension, used for principal component decomposition, preserving the main energy directions. It is the singular value matrix, reflecting the energy intensity of each coupled principal component; the larger the value, the more important it is. It is the conjugate transpose of the right singular vector matrix, representing the main coupling direction between frequency bands, and the singular value matrix. As the main indicator reflecting the strength of nonlinear coupling between frequency bands, the nonlinear coupling characteristic matrix is ​​defined as follows: This serves as the core input feature for subsequent frequency band collaborative compensation design.

[0028] In the previous stage, cross-correlation calculations were performed on multiple intermodulation basis functions and the power amplifier output signal to obtain cross-correlation vectors reflecting the contribution strength of each intermodulation term. As time progresses or channel conditions change, these cross-correlation vectors will form a matrix sequence with time-series characteristics, namely the cross-correlation characteristic matrix.

[0029] This step uses the Singular Value Decomposition (SVD) method to process the matrix, aiming to identify the most representative nonlinear coupled principal components from numerous intermodulation features. Singular Value Decomposition is a classic matrix decomposition method that can decompose the original high-dimensional data into a set of orthogonal basis directions (singular vectors) and their corresponding energy intensities (singular values). The left singular vector is used to characterize the spatial distribution among different intermodulation features, while the singular values ​​reflect the energy concentration of each principal component, i.e., its importance in the overall coupling behavior. The right singular vector corresponds to the main coupling direction under different time segments or operating conditions.

[0030] By extracting the singular value matrix (diagonal matrix), a concise and physically meaningful nonlinear coupling characteristic matrix can be obtained, which can be used to represent the dominance of each order of intermodulation components in the current system and the cross-coupling strength between their frequency bands.

[0031] In summary, this step maps high-dimensional raw data into low-dimensional, structured coupling feature matrices with clear physical meaning, which not only improves modeling efficiency but also enhances the adaptability and responsiveness of the entire predistortion system in multi-frequency coupling scenarios.

[0032] S2: A frequency band collaborative compensation signal is generated using a nonlinear coupling feature matrix. The frequency band collaborative compensation signal is used to compensate for the asymmetric cross-modulation effect between adjacent frequency bands. S2 specifically includes: S21: Input signal for each frequency band Construct its nonlinear basis function vector, including self-modulation terms and cross-modulation terms, as follows: ; in, It is the first The input signal of each frequency band represents the baseband complex envelope of each sub-band or carrier, reflecting the input of different communication resource blocks. It is the first The complex conjugate of the input signal in each frequency band is used to construct the cross-modulation term, which, together with complex multiplication, reflects the asymmetric intermodulation behavior. It is a non-negative integer, representing a non-linear exponentiation. It is the first Input signal of each frequency band The power of the power reflects its nonlinear enhancement effect and is used to model higher-order self-modulation components. It is the first The complex conjugate of the input signal in each frequency band The power is a higher-order conjugate term used to simulate cross-modulation coupling behavior. It is the first The frequency band and the first The nonlinear combination of conjugates of signals in each frequency band is the basic unit for constructing intermodulation basis functions, used to extract interband coupling features or generate compensation signals. It is the highest nonlinear order. It is a nonlinear basis function vector, including cross-modulation components across multiple frequency bands, representing self-modulation and cross-modulation components; In practical systems where adjacent multi-band signals drive power amplifiers, asymmetric cross-modulation often occurs between different frequency bands, introducing nonlinear distortion. This distortion cannot be effectively described by single-band modeling methods. Therefore, in order to more accurately characterize the nonlinear coupling features between frequency bands, this step proposes to construct a set of nonlinear basis function vectors for multi-band signals.

[0033] Specifically, this vector is formed by combining multiple frequency band input signals after performing power transformation and complex conjugation operations. It not only includes the nonlinear modulation terms of each frequency band itself, but also explicitly considers the cross-modulation terms generated between two frequency bands through nonlinear channels. These cross-modulation terms are the main source of inter-band interference and play a role in compensation performance. The construction of this basis function vector provides rich and physically meaningful feature inputs for subsequent collaborative compensation, which is an important technical link to improve the predistortion performance of multi-band and suppress cross-modulation distortion.

[0034] S22: The nonlinear coupling characteristic matrix With nonlinear basis function vectors After transformation, the initial frequency band cooperative compensation component is obtained, which is expressed as: ; in, It is the nonlinear coupling characteristic matrix obtained from the aforementioned singular value decomposition, reflecting the intensity of inter-band intermodulation. It is the initial collaborative compensation component vector, and each element corresponds to a compensation candidate signal in a different frequency band; This step aims to fuse the nonlinear coupling features between frequency bands extracted in the previous stage with the constructed nonlinear basis function vector to generate the initial form of the frequency band cooperative compensation signal components. Specifically, the nonlinear coupling characteristic matrix is ​​mapped one-to-one with the nonlinear basis function vector at the current moment. This process can be understood as, given multiple intermodulation coupling directions and intensities, weighting and combining all candidate intermodulation components according to these coupling modes to form a representative set of compensation candidate signals. In the generated initial compensation components, each component corresponds to a specific frequency band compensation effect, and its amplitude and phase characteristics reflect the nonlinear interference intensity that the frequency band may currently face. These initial compensation signals have not yet been optimized and adjusted, but they already possess the basic structure of frequency band synergy, laying the physical foundation and structural prior for the next stage of dynamic weighting and signal injection.

[0035] S23: For the initial collaborative compensation component Apply dynamically adjusted weights The final frequency band cooperative compensation signal is generated, represented as: ; in, It is the first An initial compensation component It is a dynamic weighting factor that can be adaptively adjusted based on the current operating conditions (such as input signal amplitude, output distortion feedback, etc.). Its value ranges from 0 to 1, and it is used to dynamically control the compensation intensity of each channel based on input characteristics, feedback errors, etc. The resulting band-coordinated compensation signal is used to incorporate into the predistortion system to reduce asymmetric cross-modulation distortion.

[0036] This step aims to dynamically optimize and adjust the initial frequency band collaborative compensation components generated by the previous steps, so as to further improve the relevance and adaptability of the compensation signal.

[0037] Specifically, the system assigns an independent dynamic weight factor to each compensation component. This weight is not fixed but is adjusted in real time according to the current operating state. For example, it can be adaptively corrected based on the strength and spectral structure of the input signal, or the distortion features and intermodulation error magnitude extracted from the output signal. By adjusting the weight factor, the system can dynamically enhance the compensation strength of the current main distortion channel while suppressing redundant intervention on insignificant components, thereby achieving the goal of simplifying the compensation structure and focusing on the effective path.

[0038] All weighted compensation components are aggregated in the time domain to form a complete band-coordinated compensation signal. This signal is directly incorporated into the digital predistortion system and superimposed on the original signal before being injected into the power amplifier input to offset the asymmetric cross-modulation distortion generated in the actual power amplifier, thereby improving the linearity and spectral purity of the output signal.

[0039] S3: Based on the distortion characteristic analysis results of the frequency band collaborative compensation signal and the power amplifier output signal, a dynamic predistortion parameter optimization model is constructed, and the final predistortion output signal is generated through real-time feedback adjustment; S3 specifically includes: S31: Power amplifier output signal Frequency domain, time domain, and statistical characteristic analyses are performed to extract multidimensional indices that constitute the distortion feature vector, which are represented as follows: ; in, The adjacent channel leakage ratio (OLL) represents the ratio of the energy leaked from the main channel signal to adjacent channels. It is an important frequency domain indicator for measuring nonlinear spectral spread distortion, and its value ranges from [value range missing]. The more linear the power amplifier output, the smaller the adjacent channel leakage, and the more negative the value. -45 is usually the acceptable baseline for 5G systems. This is the error vector magnitude, representing the vector error between the reconstructed signal at the receiver and the ideal modulated signal. It reflects modulation fidelity and belongs to the time-frequency joint domain index. Its value ranges from 0 to 0.15. The better the system performance, the smaller the error. The peak power factor (PGF) represents the ratio of a signal's maximum envelope power to its average power. It is an important parameter for measuring a signal's dynamic range and is often used to evaluate the compression tendency of a power amplifier. Its value ranges from 6 to 12. Signals with high PGFs are more likely to enter the nonlinear region of the power amplifier. Different modulation methods result in different values; OFDM signals generally have higher PGFs. Kurtosis, or kurtosis, represents the steepness of the amplitude distribution of signal samples. It is a higher-order moment index in a statistical sense, used to help identify non-Gaussian distortion. Its value ranges from 1.8 to 6. Higher kurtosis indicates a more concentrated signal and more prominent outliers. It is typically used in pre-distortion design to detect whether signal spikes cause clipping distortion. It is a distortion feature vector, which serves as a comprehensive representation of system performance degradation. It consists of multiple sub-features representing different dimensions of distortion mechanisms and is used to construct input feature groups for multi-objective optimization functions or control dynamic compensation strategies. It integrates and models multiple distortion indices, which helps to improve the global optimization capability and adaptability of the compensation model. In this step, multi-dimensional distortion features are extracted and analyzed for the output signal of the power amplifier. The purpose is to fully perceive the nonlinear distortion modes present in the output signal and provide an accurate basis for subsequent pre-distortion parameter optimization. The extracted distortion features are not limited to frequency domain information, but also cover time domain statistical characteristics, characterizing the degree to which the signal deviates from the ideal state from multiple perspectives.

[0040] Specifically, this step extracts representative feature parameters from the dimensions of spectral leakage, modulation error, signal dynamic range, and waveform morphology distribution to construct a distortion feature vector. Typical features include adjacent channel leakage ratio, used to measure the spectral spread caused by the power amplifier; error vector amplitude, used to measure modulation accuracy; peak factor, used to assess the risk of nonlinear compression of the signal; and kurtosis, used to reflect whether the signal amplitude distribution is abnormally concentrated or exhibits peak behavior. These features together constitute the distortion feature vector, which serves as the input index for the subsequent predistortion parameter adjustment model, helping the system dynamically perceive the distortion level under the current operating state. It is the core data foundation for realizing closed-loop optimization control. Through this step, a quantitative characterization of nonlinear distortion can be achieved, improving the adaptability and accuracy of the predistortion model.

[0041] S32: Combine with current frequency band cooperative compensation signal With distorted feature vectors Construct a parameter tuning model that includes a multi-objective optimization function, expressed as: ; in, This represents the adjustable parameter vector of the predistorter, with dimensions ranging from 10 to 200 and element values ​​ranging from -1 to 1. This vector controls the transformation characteristics of the predistortion structure. To adapt to different nonlinear characteristics, its values ​​need to be both positive and negative, and are usually normalized to facilitate optimization convergence. It is the total loss function, which includes multiple objective terms and represents the comprehensive error of the entire optimization objective. It must be continuously differentiable; the closer it is to zero, the more accurate the compensation. It is the first Each objective function, known as EVM (Error Vector Magnitude), measures a specific distortion mode and provides direction for model optimization. EVM is a metric for modulated signal distortion, assessing the combined deviation of amplitude and phase errors between the actual transmitted signal and the ideal modulated signal. A smaller EVM indicates better signal quality and lower modulation distortion. In power amplifier output, nonlinear distortion, noise interference, and phase noise all contribute to an increased EVM. It effectively reflects the predistorter's compensation effect on modulation fidelity. Minimizing the EVM objective function directly improves bit error rate performance and spectral compliance. These are the weighting factors for each objective function, used to control the optimization priority of each indicator, with a value range of 0.1-10; In this step, based on the frequency band collaborative compensation signal generated in the previous stage and the multi-dimensional distortion feature vector extracted in real time, a dynamic predistortion parameter optimization model including multiple optimization objectives is constructed. The core objective of this model is to dynamically adjust the control parameters in the predistorter by jointly considering the distortion performance in the frequency domain, time domain and statistical domain, so that its output signal approximates the ideal linear response as much as possible after passing through the power amplifier.

[0042] Specifically, the model uses the parameter vector in the predistorter as the variable to be optimized and sets multiple optimization objective functions. These objectives include multiple directions such as adjacent channel leakage control, modulation error suppression, peak control, and non-Gaussianity mitigation. Each objective function is assigned a weight coefficient according to different system indicators, forming a multi-objective joint optimization framework, which can dynamically balance the compensation strategy according to the actual operating state of the system.

[0043] This optimization model is not only scalable, allowing for the addition or removal of objective function terms according to actual needs, but also supports rapid iterative solutions, making it easy to integrate into the real-time compensation loop of actual communication links. It is an important technical support module for improving the accuracy of nonlinear modeling and adaptive compensation capabilities in this invention.

[0044] S33: An adaptive optimization algorithm (minimum mean square error algorithm) is used to solve the above optimization model in real time to obtain the optimal parameters. Based on this, the final predistorted output signal is generated, expressed as: ; in, It is the original input signal, representing the system input before processing, and is the starting point for all predistortion calculations. This is a predistortion function structure employing a memory polynomial model to represent the nonlinear mapping relationship from the original signal to the compensated signal. It needs to possess certain expressive power and differentiability for easy optimization. The memory polynomial model combines nonlinear and delay terms, simultaneously characterizing the amplitude nonlinear distortion and memory effect of the power amplifier. It boasts advantages such as simple structure, high computational efficiency, and ease of online updating, making it suitable for the real-time requirements of this invention for multi-band cross-modulation distortion modeling and compensation. The memory polynomial structure combines multiple time-delayed versions of the input signal with nonlinear terms of different orders, assigning independent parameter weights to each combination to form the predistorted output signal. This structure allows for flexible adjustment of the order and memory depth to balance modeling accuracy and computational complexity. It is the output predistortion signal, used to cancel the dynamic nonlinear distortion terms generated in the power amplifier. The adaptive optimization algorithm is the minimum mean square error algorithm. In the process of dynamic predistortion parameter optimization, in order to achieve real-time response and compensation adjustment to changes in nonlinear distortion, this invention selects the minimum mean square error algorithm as an adaptive optimization mechanism. This algorithm iteratively adjusts the predistortion parameters by comparing the error signal between the power amplifier output signal and the desired output, so that the overall mean square error value gradually decreases, thereby realizing online self-learning and self-correction of the parameter vector.

[0045] Specifically, at each frame of signal or at each sampling point level, the error between the output of the current predistorted signal after power amplifier transformation and the reference target is calculated. This error is then multiplied by the corresponding input basis function as a gradient direction estimate, and the parameters in the predistorter are updated in small steps. This algorithm has the advantages of simple structure, low computational cost, and ease of implementation, and is suitable for scenarios where the predistortion function structure is a memory polynomial type. By introducing the minimum mean square error algorithm, the predistortion system possesses rapid adaptability and stable convergence characteristics under scenarios of frequency band cross-modulation and dynamic changes in nonlinear distortion, effectively improving compensation accuracy and system robustness. This is a crucial technical step in achieving closed-loop adaptive optimization.

[0046] like Figure 2 As shown, an adjacent multi-band digital predistortion system is used to implement the above-mentioned adjacent multi-band digital predistortion method, and includes the following modules: Intermodulation feature extraction module: Performs nonlinear analysis on adjacent multi-band input signals, extracts intermodulation feature vectors, and constructs a nonlinear coupling feature matrix based on the intermodulation feature vectors; Band Co-compensation Generation Module: Receives the nonlinear coupling feature matrix, constructs a compensation vector including cross-modulation terms, and generates a band co-compensation signal for suppressing asymmetric modulation effects between adjacent bands; Dynamic predistortion optimization module: Based on the distortion characteristic analysis results of the frequency band collaborative compensation signal and the power amplifier output signal, a dynamic predistortion parameter optimization model is constructed, and the final predistortion output signal is generated through a real-time feedback optimization mechanism.

[0047] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0048] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for adjacent multi-band digital predistortion, characterized in that, Includes the following steps: S1: Extract the intermodulation feature vectors of adjacent multi-band input signals, and construct a nonlinear coupling feature matrix based on the intermodulation feature vectors; S2: Generate a band-coordinated compensation signal using the nonlinear coupling feature matrix. The band-coordinated compensation signal is used to compensate for the asymmetric cross-modulation effect between adjacent bands. S3: Based on the distortion characteristic analysis results of the frequency band collaborative compensation signal and the power amplifier output signal, a dynamic predistortion parameter optimization model is constructed, and the final predistortion output signal is generated through real-time feedback adjustment.

2. The adjacent multi-band digital predistortion method according to claim 1, characterized in that, S1 includes: S11: Perform nonlinear transformation on adjacent multi-band input signals to construct a basis function set including multiple order intermodulation terms. The basis function set includes self-modulation components and cross-modulation components of signals from different frequency bands. S12: Perform cross-correlation calculations between each basis function signal in the basis function set and the output signal of the power amplifier to obtain a cross-correlation vector representing the coupling strength between each intermodulation term and the output signal; S13: The cross-correlation vectors at multiple time points are concatenated to construct a cross-correlation matrix that includes time-series variation features. Singular value decomposition is then performed on the cross-correlation matrix to extract the singular values ​​corresponding to the dominant components as a measure of the nonlinear coupling strength between frequency bands, thus forming a nonlinear coupling feature matrix.

3. The adjacent multi-band digital predistortion method according to claim 2, characterized in that, S13 includes: S131: Stack and concatenate the cross-correlation vectors extracted at multiple different times to construct a cross-correlation matrix that includes time-series change information; S132: Perform singular value decomposition on the cross-correlation matrix, extract the singular value features corresponding to the dominant components, and construct them into a nonlinear coupling feature matrix.

4. The adjacent multi-band digital predistortion method according to claim 3, characterized in that, The singular value decomposition process extracts the dominant components of the nonlinear coupling features, and the coupling strength between frequency bands is determined based on the energy proportion of each singular value, so as to generate a nonlinear coupling feature matrix that reflects the dominant nonlinear interaction relationship between frequency bands.

5. The adjacent multi-band digital predistortion method according to claim 3, characterized in that, S2 includes: S21: Construct a nonlinear basis function vector based on the input signals of each frequency band, wherein the nonlinear basis function vector includes a modulation term composed of multiple frequency band signals and their nonlinear combination forms; S22: Map the nonlinear coupling feature matrix to the nonlinear basis function vector to obtain the initial compensation component for compensating nonlinear distortion in different frequency bands; S23: Apply dynamic adjustment weights to the initial compensation components, and adaptively adjust the compensation intensity of each component according to the current system operating conditions to offset the asymmetric cross-modulation effect between adjacent frequency bands and generate the final frequency band collaborative compensation signal.

6. The adjacent multi-band digital predistortion method according to claim 5, characterized in that, S21 includes: S211: Perform nonlinear transformation on the input signals of each frequency band to generate basis function components including different power forms; S212: Combine multiple frequency band signals and their complex conjugate forms to generate a nonlinear basis function vector including cross-modulation terms.

7. The adjacent multi-band digital predistortion method according to claim 1, characterized in that, The asymmetric cross-modulation effect includes: Intermodulation interference occurs between adjacent frequency band signals due to the nonlinearity of the power amplifier. The distortion in one frequency band mainly originates from the modulation effect of the other frequency band. The distortion distribution across different frequency bands is not symmetrical, including coupling characteristics of inconsistent modulation path directionality and uneven spectral energy distribution.

8. The adjacent multi-band digital predistortion method according to claim 1, characterized in that, S3 includes: S31: Perform multi-dimensional distortion feature analysis on the output signal of the power amplifier and extract relevant indicators that constitute the distortion feature vector. The relevant indicators include spectral features, modulation accuracy indicators, dynamic range information and statistical characteristics. S32: Based on the distortion feature vector and the frequency band collaborative compensation signal, a dynamic predistortion parameter optimization model with a multi-objective optimization structure is constructed. The dynamic predistortion parameter optimization model takes multiple performance indicators as optimization objectives and sets weight factors to reflect the priority relationship of different optimization objectives, so as to realize the dynamic adjustment of the predistortion control parameters. S33: The optimization model is solved in real time using the minimum mean square error algorithm. The predistorter parameters are continuously updated iteratively based on the current feedback results. The updated predistorter parameters are then applied to the original input signal to generate the final predistorted output signal.

9. The adjacent multi-band digital predistortion method according to claim 8, characterized in that, S33 includes: S331: The minimum mean square error algorithm is used to solve the dynamic predistortion parameter optimization model in real time. The control parameters in the predistorter are iteratively updated based on the feedback error between the power amplifier output signal and the target response. S332: Apply the updated control parameters to the original input signal and generate the final predistorted output signal through the predistortion function structure.

10. An adjacent multi-band digital predistortion system, used to implement the adjacent multi-band digital predistortion method as described in any one of claims 1-9, characterized in that, Includes the following modules: Intermodulation feature extraction module: performs nonlinear analysis on adjacent multi-band input signals, extracts intermodulation feature vectors, and constructs a nonlinear coupling feature matrix based on the intermodulation feature vectors; Frequency band cooperative compensation generation module: receives the nonlinear coupling feature matrix, constructs a compensation vector including cross-modulation terms, and generates a frequency band cooperative compensation signal for suppressing asymmetric modulation effects between adjacent frequency bands; Dynamic predistortion optimization module: Based on the distortion characteristic analysis results of the frequency band collaborative compensation signal and the power amplifier output signal, a dynamic predistortion parameter optimization model is constructed, and the final predistortion output signal is generated through a real-time feedback optimization mechanism.

Images