Duplexer multi-band performance optimization method and system

Abstract

The present invention relates to a method and system for optimizing the multi-band performance of a duplexer. The method comprises the following steps: collecting frequency response data and temperature characteristic data of the duplexer in the transmitting frequency band and the receiving frequency band; performing segmented mapping, establishing a dynamic frequency characteristic diagram, and performing feature extraction to generate frequency band coupling correlation characteristics; calculating and optimizing the metal layer thickness ratio and interface characteristics in the multi-layer electrode structure to generate multi-layer electrode stress dispersion control parameters; dynamically adjusting the compensation voltage to generate a temperature compensation control matrix; performing decoupling processing of the transmitting frequency band and the receiving frequency band on the frequency band coupling correlation characteristics, and performing parallel calculation in combination with the multi-layer electrode stress dispersion control parameters and the temperature compensation control matrix to generate duplexer frequency band performance optimization parameters. The present invention enhances the anti-interference capability of the duplexer through decoupling processing of the transmitting frequency band and the receiving frequency band and parallel optimization calculation.

Description

Technical Field

[0001] The present invention relates to the technical field of duplexers, and in particular to a method and system for optimizing the multi-band performance of a duplexer. Background Art

[0002] In the process of duplexer design optimization, the existing technology mainly faces three difficulties: the frequency band coupling characteristics in the multi-band working mode are complex, and the traditional performance optimization method is difficult to effectively deal with the mutual influence between frequency bands; secondly, the duplexer is affected by temperature changes during operation, resulting in frequency drift, affecting the communication quality; in high-power application scenarios, the stress distribution of the multi-layer electrode structure is uneven, which can easily cause device performance degradation and reliability problems. Summary of the invention

[0003] The main purpose of the present invention is to provide a method and system for optimizing the multi-band performance of a duplexer. The present invention enhances the anti-interference capability of the duplexer by decoupling the transmitting frequency band and the receiving frequency band and performing parallel optimization calculation.

[0004] To achieve the above object, the present invention provides a method for optimizing the multi-band performance of a duplexer, comprising the following steps:

[0005] Collect frequency response data and temperature characteristic data of the duplexer in the transmit frequency band and the receive frequency band;

[0006] Segmentally map the frequency response data according to the transmitting frequency band and the receiving frequency band, establish a dynamic frequency characteristic graph, and extract features from the dynamic frequency characteristic graph through a graph neural network to generate frequency band coupling correlation features;

[0007] Using an interlayer thickness analysis model and an interface stress calculation model, the thickness ratio of metal layers and interface characteristics in the multilayer electrode structure are calculated and optimized according to the frequency band coupling correlation characteristics, and the multilayer electrode stress dispersion control parameters are generated;

[0008] Inputting the temperature characteristic data into a frequency drift compensation model, establishing a temperature-frequency mapping relationship, and dynamically adjusting the compensation voltage through an active bias circuit to generate a temperature compensation control matrix;

[0009] The frequency band coupling correlation characteristics are decoupled into a transmitting frequency band and a receiving frequency band, and parallel calculations are performed in combination with the multi-layer electrode stress dispersion control parameters and the temperature compensation control matrix to generate duplexer frequency band performance optimization parameters.

[0010] The present invention also provides a duplexer multi-band performance optimization system, comprising:

[0011] An acquisition module, used to acquire frequency response data and temperature characteristic data of the duplexer in the transmitting frequency band and the receiving frequency band;

[0012] A mapping module, used to segmentally map the frequency response data according to the transmitting frequency band and the receiving frequency band, establish a dynamic frequency characteristic graph, and extract features from the dynamic frequency characteristic graph through a graph neural network to generate frequency band coupling correlation features;

[0013] A calculation module, for calculating and optimizing the thickness ratio and interface characteristics of the metal layers in the multilayer electrode structure according to the frequency band coupling correlation characteristics by using an interlayer thickness analysis model and an interface stress calculation model, and generating a multilayer electrode stress dispersion control parameter;

[0014] An adjustment module, used for inputting the temperature characteristic data into a frequency drift compensation model, establishing a temperature-frequency mapping relationship, and dynamically adjusting the compensation voltage through an active bias circuit to generate a temperature compensation control matrix;

[0015] A generation module is used to decouple the transmission frequency band and the reception frequency band of the frequency band coupling correlation characteristics, and to perform parallel calculations in combination with the multi-layer electrode stress dispersion control parameters and the temperature compensation control matrix to generate duplexer frequency band performance optimization parameters.

[0016] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0017] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.

[0018] In summary, the technical solution provided by the present invention effectively solves the performance optimization problem under multi-band coupling conditions and improves the transmission efficiency of the duplexer by establishing a dynamic frequency characteristic graph and applying a graph neural network for feature extraction. The multi-layer electrode structure optimization design is adopted, combined with the interlayer thickness analysis model and the interface stress calculation model, to enhance the stability and reliability of the duplexer under high-power working conditions. The temperature compensation technology is innovatively combined with the active bias circuit to solve the frequency drift problem caused by temperature changes and improve the temperature stability of the device. Through the decoupling processing of the transmitting frequency band and the receiving frequency band and parallel optimization calculation, excellent frequency band isolation characteristics are achieved and the anti-interference ability of the duplexer is enhanced. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the steps of a method for optimizing the multi-band performance of a duplexer in one embodiment of the present invention;

[0020] Figure 2It is a structural block diagram of a duplexer multi-band performance optimization system in one embodiment of the present invention;

[0021] Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0022] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0024] Reference Figure 1 This embodiment provides a method for optimizing the multi-band performance of a duplexer, comprising the following steps:

[0025] S1, collecting frequency response data and temperature characteristic data of the duplexer in the transmitting frequency band and the receiving frequency band;

[0026] Among them, the transmission frequency band of the duplexer is frequency scanned, the frequency scan is performed within the selected target range, and the insertion loss data of the transmission port is collected. The insertion loss data can reflect the transmission loss of the duplexer in the transmission frequency band, and the insertion loss characteristic matrix of the transmission frequency band is obtained. Similarly, a similar frequency scan is performed on the receiving frequency band, and the insertion loss data of the receiving port is collected to obtain the insertion loss characteristic matrix of the receiving frequency band. These data constitute the frequency response information of the duplexer in the transmission and reception frequency bands. The insertion loss characteristic matrix of the transmission and reception frequency bands is input into the spectrum analysis model. The spectrum analysis model calculates the isolation characteristics between the two based on the insertion loss characteristics of the transmission frequency band and the reception frequency band. The isolation reflects the degree of signal interference between the transmission and reception ports. Through this process, the frequency band isolation characteristic vector is obtained, which contains the degree of mutual interference between transmission and reception in different frequency bands. The insertion loss characteristic matrix of the transmitting frequency band, the insertion loss characteristic matrix of the receiving frequency band, and the frequency band isolation characteristic vector are fused to form a complete frequency response data set, so as to comprehensively consider the insertion loss of the transmitting frequency band and the receiving frequency band and the isolation between the frequency bands, and reflect the performance of the duplexer in the entire working frequency band. Within the preset temperature range, the duplexer is scanned in the full frequency band, and the gain change data of the duplexer under different working conditions is recorded to obtain the temperature-gain characteristic matrix. The change in gain is closely related to temperature. When the temperature rises, the gain of the duplexer will drift. This characteristic matrix can help the system understand the specific impact of temperature change on the gain of the duplexer. Based on the temperature-gain characteristic matrix, a thermodynamic analysis is performed to calculate the temperature distribution of the duplexer under maximum power conditions. Through numerical simulation or thermodynamic model, the temperature distribution characteristic vector is obtained to describe the temperature distribution of the duplexer under different working conditions. The thermal expansion caused by temperature change will cause the material stress distribution inside the duplexer to change, further affecting its deformation. The stress distribution and deformation of the duplexer are calculated based on the temperature distribution eigenvector, and the stress-deformation eigenvector is obtained, which reflects the mechanical deformation and stress distribution of the duplexer caused by temperature changes during actual operation. The temperature-gain characteristic matrix, temperature distribution eigenvector and stress-deformation eigenvector are fused to form complete temperature characteristic data, which fully describes the performance characteristics of the duplexer under temperature changes, including gain changes, stress distribution caused by temperature, and deformation.

[0027] S2, segmentally map the frequency response data according to the transmitting frequency band and the receiving frequency band, establish a dynamic frequency feature map, and extract features from the dynamic frequency feature map through a graph neural network to generate frequency band coupling correlation features;

[0028] Specifically, the frequency response data is segmented and mapped. According to the different characteristics of the transmitting frequency band and the receiving frequency band, each frequency band is divided into several sub-frequency bands. Each sub-frequency band reflects the local characteristics of the frequency response, more accurately captures the change rules in different frequency intervals, and forms a frequency band segmentation matrix. Corresponding nodes are established for each sub-frequency band in the frequency band segmentation matrix, and these nodes represent specific characteristics in each frequency band interval. According to the frequency interval between frequency bands, the connection weights between nodes are calculated. The interval size between frequency bands determines the connection strength between nodes, and this calculation can reflect the closeness of the relationship between different frequency bands. The initial frequency feature map is constructed by connecting weights. The graph represents the coupling relationship between frequency bands as a graph structure, in which nodes represent frequency bands and the weights of edges represent the degree of mutual influence between frequency bands. Based on the initial frequency feature map, a graph convolution layer is constructed. The graph convolution operation realizes the aggregation of node information by performing spatial domain convolution operations on node features. The graph convolution layer fuses the feature information of a node with the feature information of its neighboring nodes to obtain an aggregated node information matrix. This process can effectively capture the local relationship between nodes and enhance the information transmission between nodes. Through the convolution operation, the features of the nodes are updated, so that each node in the graph can contain more features about its neighbors and global frequency band information. The node information aggregation matrix is ​​subjected to nonlinear transformation and pooling operations. The nonlinear transformation is to enhance the expressive power of the features so that the graph network can handle more complex patterns, while the pooling operation helps to reduce the feature dimension and improve the computational efficiency of the model. Through this process, the graph convolutional network can extract the topological relationship features between nodes. The processed features are encoded into a graph feature encoding matrix to reflect the higher-level relationship structure between frequency bands. In order to enhance the model's sensitivity to the correlation between nodes in different frequency bands, the graph feature encoding matrix is ​​input into a network based on the attention mechanism. The attention mechanism can automatically identify which frequency bands are more closely related by assigning different weights to each frequency band node. The attention mechanism network generates a frequency band attention weight matrix by calculating the degree of correlation between nodes in different frequency bands. Based on the frequency band attention weight matrix, the graph feature encoding matrix is ​​weighted updated to adjust the information of each node so that the network pays more attention to the parts with strong coupling between frequency bands, thereby optimizing the frequency band performance and obtaining a dynamic frequency feature map, which reflects the complex coupling relationship of the duplexer in different frequency bands. Using the graph structure feature learning method, the dynamic frequency feature graph is analyzed to extract the coupling relationship between frequency bands, which helps the model understand the mutual influence of different frequency bands. On this basis, a coupling feature vector is generated to represent the coupling strength and interaction relationship between frequency bands. The coupling feature vector is input into the fully connected neural network for feature fusion to generate comprehensive frequency band coupling correlation features.

[0029] The node feature matrix in the dynamic frequency feature graph is decomposed to convert the complex node features into eigenvalue matrix and eigenvector matrix. The eigenvalue matrix describes the importance of each node feature, while the eigenvector matrix reflects the relative relationship between nodes. Based on the eigenvalue matrix and eigenvector matrix, the graph Fourier transform kernel is constructed. Graph Fourier transform is a technology for frequency domain analysis in graph structure. By performing frequency domain transformation on node features, information of different frequency components is extracted. These frequency components represent the characteristics of nodes in the graph at different scales and can reveal deeper relationships between nodes. The frequency domain feature matrix obtained by frequency domain transformation can help the system understand the performance of nodes in different frequency domains and extract high-order features related to frequency band coupling relationships. The frequency domain feature matrix is ​​input into the graph convolutional neural network. The graph convolutional neural network can perform information propagation and feature extraction in the graph structure. When dealing with complex topological relationships, the graph convolutional neural network can effectively capture high-order topological relationships between nodes. Through graph convolution operations, the feature information of nodes not only comes from direct neighbors, but can also gradually integrate feature information in a wider range through the deep propagation mechanism of the network. Through this process, the deep graph feature matrix obtained contains the complex relationship between nodes in the graph and reveals the coupling characteristics between frequency bands. Edge attention calculation is performed on the deep graph feature matrix to assign different weights to each edge in the graph, highlighting the more critical node relationship in frequency band coupling. The edge weight matrix provides a dynamic weight value for each edge by calculating the dependency and importance between nodes, reflecting the strength of the association between frequency bands. The node features are updated based on the edge weight matrix so that the node features more accurately reflect the degree of coupling between frequency bands, and the coupling feature matrix is ​​obtained. Jump connections are performed on the coupling feature matrix. Jump connections can fuse feature information from different levels and avoid information loss between layers. In this way, coupling features between frequency bands are extracted at different scales to enhance the expression ability of the model. The multi-scale feature vectors are input into the residual network for feature transformation. The residual network avoids the problem of gradient vanishing or information loss in the deep network by introducing short-circuit connections. The purpose of feature transformation is to further nonlinearly map the extracted features, thereby enhancing the expression ability of the features and helping the network capture the complex patterns of frequency band coupling features. Through the processing of the residual network, the obtained coupling feature vector is more expressive and can better adapt to the complex coupling relationship between frequency bands. The coupled feature vector is batch normalized and activated. Batch normalization can make the data distribution more balanced by normalizing the feature vector and reduce the gradient explosion or vanishing problem during training. The activation function enables the network to learn more complex patterns through nonlinear transformation. The processed feature vector is more stable and has higher generalization ability in subsequent tasks. The obtained normalized feature vector is converted into frequency band coupling correlation features through nonlinear mapping.

[0030] S3, using the interlayer thickness analysis model and interface stress calculation model, calculates and optimizes the metal layer thickness ratio and interface characteristics in the multilayer electrode structure according to the frequency band coupling correlation characteristics, and generates the multilayer electrode stress dispersion control parameters;

[0031] It should be noted that the frequency band coupling correlation characteristics are input into the thermal expansion coefficient difference calculation model. The model analyzes the thermal expansion behavior of each metal layer in the multilayer electrode structure based on the difference in thermal expansion coefficients between different metal layers. Through this analysis, the interlayer thermal expansion coefficient matrix is ​​obtained, revealing the difference in expansion degree of different metal layers when the temperature changes. Based on the interlayer thermal expansion coefficient matrix, a stress distribution model is constructed to calculate the stress distribution characteristics at each interface in the multilayer electrode structure. Through the difference in thermal expansion coefficients, the model can simulate the stress generated by different metal layers under temperature changes, and obtain the interface stress distribution vector, which represents the stress intensity generated by the thermal expansion effect at different interfaces of the electrode structure. By understanding the specific situation of stress distribution, the thickness and interface characteristics of each metal layer are adjusted to achieve a more uniform stress distribution. The stress gradient of the interface stress distribution vector is calculated. The stress gradient reflects the rate of change of stress along the interface of the electrode structure. By calculation, it is determined whether the stress is evenly distributed between the layers. If the stress is too concentrated in some areas, it will cause local deformation or material damage. Establish constraints for uniform stress dispersion to ensure that the stress of the multilayer electrode structure is reasonably distributed and avoid structural failures caused by stress concentration. Through the calculated stress dispersion constraint matrix, the thickness ratio of each metal layer in the multilayer electrode structure is iteratively optimized to obtain the interlayer thickness optimization parameters. By continuously adjusting the thickness ratio of each metal layer, the optimization process can find an optimal thickness distribution scheme so that the entire electrode structure reaches the optimal state in terms of thermal expansion and stress distribution. The interlayer thickness optimization parameters are input into the acoustic migration effect analysis model to analyze the stress change of the electrode structure under the action of continuous wave power. The acoustic migration effect analysis model simulates the stress response of the electrode structure under different power conditions. Especially in the case of continuous wave signal transmission, the power change will cause the thermal effect and mechanical response of the structure. By analyzing the power stress eigenvector, the model can accurately describe the stress change of the electrode structure during operation. Based on the power stress characteristics, a temperature field distribution model is established to simulate the temperature change of the electrode structure under different power conditions. The temperature field distribution matrix reveals the temperature change law of the electrode structure under thermal load. Through the analysis of the temperature field distribution matrix, thermal stress analysis is performed to calculate the stress and deformation parameters of the electrode structure under maximum power conditions. These parameters represent the deformation of the electrode under high power working conditions and the stress change caused by temperature change, and the stress deformation parameter matrix is ​​obtained. The interlayer thickness optimization parameters and stress deformation parameter matrix are weightedly fused to obtain the multi-layer electrode stress dispersion control parameters, which effectively manages and adjusts the stress distribution in the electrode structure and improves the performance stability of the duplexer in different working environments.

[0032] S4, inputting the temperature characteristic data into the frequency drift compensation model, establishing a temperature-frequency mapping relationship, and dynamically adjusting the compensation voltage through an active bias circuit to generate a temperature compensation control matrix;

[0033] Specifically, the temperature characteristic data is input into the frequency drift compensation model for frequency response analysis. The model analyzes the frequency drift characteristics according to the relationship between temperature change and frequency response, and generates a frequency drift feature mapping matrix. The matrix reveals the changing trend of frequency response under different temperature conditions, which can help the system accurately capture the frequency drift caused by temperature. The frequency drift feature mapping matrix is ​​input into the temperature correlation analysis model. This model analyzes the change of frequency response in the temperature range of -55°C to 125°C. Through the correlation between temperature and frequency, the model establishes a temperature-frequency correspondence matrix to reveal how the frequency drift of the duplexer under different temperature conditions affects its performance. The drift characteristics in the temperature-frequency correspondence matrix are segmented. The influence of temperature change on frequency is divided into multiple different temperature intervals for more accurate compensation. In this process, the compensation parameter distribution vector is extracted to guide the subsequent compensation voltage design. Based on the compensation parameter distribution vector, the compensation control strategy of the active bias circuit is designed, so as to calculate the corresponding compensation voltage for each temperature interval and obtain the voltage compensation parameter matrix. The voltage compensation parameter matrix is ​​input into the dynamic feedback controller to establish a closed-loop regulation model for the compensation voltage. The closed-loop regulation model ensures that the voltage compensation can be dynamically adjusted according to the temperature change through real-time feedback and adjustment of the compensation voltage to maintain the frequency stability of the duplexer under different temperature conditions. Through the feedback regulation process, the dynamic regulation control vector is obtained. The temperature compensation effect of the dynamic regulation control vector is analyzed, the gain change of each frequency band is calculated, the actual effect of temperature compensation in each frequency band is evaluated, and the influence of temperature compensation on the performance of the duplexer is quantified by the gain change. The obtained compensation effect evaluation matrix records the influence of the compensation voltage on the gain change in different frequency bands, reflecting the effectiveness of the compensation strategy. The compensation effect evaluation matrix is ​​input into the compensation optimization model to iteratively optimize the compensation parameters. The model optimizes the control strategy of the compensation voltage through multiple iterations to improve the effect of temperature compensation. The optimization process continuously adjusts the compensation parameters to achieve the best compensation performance. After the optimization of the compensation parameters is completed, the optimized compensation control vector is weightedly fused with the original compensation parameter distribution vector to obtain the final temperature compensation control matrix.

[0034] S5, decoupling the transmission frequency band and the reception frequency band based on the frequency band coupling correlation characteristics, and performing parallel calculations in combination with the multi-layer electrode stress dispersion control parameters and the temperature compensation control matrix to generate duplexer frequency band performance optimization parameters.

[0035] Among them, the frequency band coupling correlation characteristics are subjected to frequency band decoupling operation. Frequency band decoupling is performed for the transmitting frequency band and receiving frequency band of the duplexer, especially for the transmitting frequency band of 703-748MHz and the receiving frequency band of 758-803MHz. The coupling effect between the two frequency bands is separated so that the characteristics of each frequency band can be described independently. After the frequency band decoupling operation, independent transmitting frequency band feature vectors and receiving frequency band feature vectors are obtained respectively, and the two vectors represent the performance characteristics of the transmitting and receiving frequency bands respectively. Based on the transmitting frequency band feature vector and the receiving frequency band feature vector, the multilayer electrode stress dispersion control parameters are decoupled and calculated. The correlation between the frequency band characteristics and the stress control parameters is analyzed to separate the different stress distribution conditions at the transmitting end and the receiving end. In the calculation process, the transmitting end stress parameters and the receiving end stress parameters are obtained for the transmitting frequency band and the receiving frequency band respectively, which accurately reflect the stress distribution characteristics caused by the electrode structure and working conditions in different frequency bands. According to the transmitting end stress parameters and the receiving end stress parameters, the stress distribution data of the duplexer under 30dBm continuous wave power is calculated. The stress response of the duplexer under actual working conditions is simulated to obtain the power stress response matrix, which shows the stress distribution of the duplexer under different powers and reflects the mechanical performance of the electrode structure under the action of continuous wave signals. The temperature compensation control matrix and the power stress response matrix are input into the multi-band joint compensation model for calculation. The influence of temperature change on the performance of the duplexer is considered, and the compensation parameters of frequency drift are calculated in the temperature range of -55℃ to 125℃. Through calculation, the frequency-temperature compensation parameter matrix is ​​obtained to adjust the working performance of the duplexer under different temperature conditions to ensure that its frequency response remains stable in a wide temperature range. According to the frequency-temperature compensation parameter matrix, the Tx (transmit) passband insertion loss and Rx (receive) passband insertion loss of the duplexer are calculated in parallel. By optimizing the transmit and receive passband insertion loss, the efficiency and performance of the duplexer in actual work are improved. According to the calculation results, the duplexer insertion loss optimization vector is obtained to show the insertion loss of each band after compensation, reflecting the optimization effect of frequency response. Based on the insertion loss optimization vector, the isolation data between the Tx band and the Rx band is calculated. Isolation is an important factor affecting the performance of the duplexer, reflecting the degree of interference between the transmitted and received signals. The signal isolation capability of the duplexer between the transmit and receive frequency bands is optimized by calculating the frequency band isolation characteristic matrix, thereby improving its anti-interference capability and stability. The frequency band isolation characteristic matrix is ​​input into the performance optimization model for optimization calculation. The goal of this model is to make the insertion loss of the duplexer in the transmit frequency band less than 1.6dB and the insertion loss in the receive frequency band less than 1.7dB through iterative calculation. Through multiple iterations, the performance optimization model will continuously adjust various parameters until the above insertion loss requirements are met. In the final optimization process, the performance optimization parameter vector is obtained, which represents the final optimized state of the duplexer in each frequency band, ensuring its optimal performance at work.The performance optimization parameter vector is normalized so that all optimization parameters can be compared and adjusted under a unified standard, and the final parameters for duplexer frequency band performance optimization are obtained.

[0036] In one example, frequency response data and temperature characteristic data of a duplexer in a transmitting frequency band and a receiving frequency band are collected, including:

[0037] Performing a frequency scan on the transmitting frequency band of the duplexer within the target range, collecting insertion loss data of the transmitting port, and obtaining an insertion loss characteristic matrix of the transmitting frequency band; and performing a frequency scan on the receiving frequency band of the duplexer within the target range, collecting insertion loss data of the receiving port, and obtaining an insertion loss characteristic matrix of the receiving frequency band;

[0038] Input the transmission frequency band insertion loss characteristic matrix and the reception frequency band insertion loss characteristic matrix into the spectrum analysis model, calculate the isolation characteristic between the transmission frequency band and the reception frequency band, and obtain the frequency band isolation characteristic vector;

[0039] Perform data fusion on the transmission frequency band insertion loss characteristic matrix, the reception frequency band insertion loss characteristic matrix and the frequency band isolation characteristic vector to obtain frequency response data;

[0040] Perform full-band scanning on the duplexer within a preset temperature range, record gain change data, obtain a temperature-gain characteristic matrix, perform thermodynamic analysis based on the temperature-gain characteristic matrix, calculate the temperature distribution of the duplexer under maximum power conditions, and obtain a temperature distribution characteristic vector;

[0041] The stress distribution and deformation of the duplexer are calculated based on the temperature distribution eigenvector to obtain the stress-deformation eigenvector, and the temperature-gain characteristic matrix, the temperature distribution eigenvector and the stress-deformation eigenvector are fused to obtain the temperature characteristic data.

[0042] In this example, the frequency of the duplexer's transmit frequency band is scanned within the target range to collect the insertion loss data of the transmit port. The target range refers to the upper and lower limits of the duplexer's operating frequency band. Within this range, the insertion loss of each frequency point in the band is measured by the scanner to obtain the insertion loss characteristic matrix of the transmit frequency band. Insertion loss refers to the loss caused by the insertion of the device during the transmission process of the signal, which is expressed as a complex matrix ,in is the insertion loss of the duplexer transmit port, in dB. The insertion loss characteristic matrix of the transmit frequency band is expressed as:

[0043] ;

[0044] in, It is at the frequency Insertion loss values ​​at the receiving port. This matrix contains the insertion loss data of all frequency points in the transmit frequency band. Perform a similar frequency scan on the receiving frequency band of the duplexer, collect the insertion loss data of the receiving port, and obtain the insertion loss characteristic matrix of the receiving frequency band. Within the range of the receiving frequency band, the insertion loss of each frequency point is measured to obtain the insertion loss characteristic matrix of the receiving frequency band. , which is expressed as:

[0045] ;

[0046] The insertion loss characteristic matrix of the transmit frequency band and the insertion loss characteristic matrix of the receive frequency band are input into the spectrum analysis model, and the isolation characteristic between the transmit frequency band and the receive frequency band is calculated through the model. The isolation degree indicates the degree of mutual interference between the transmit and receive signals, and reflects whether the duplexer can effectively isolate the transmit signal from the receive signal. This characteristic is calculated by the following formula:

[0047] ;

[0048] in, is the power of the received signal, is the power of the transmitted signal. The frequency band isolation eigenvector is expressed as:

[0049] ;

[0050] The isolation characteristic vector calculated by the spectrum analysis model reflects the isolation capability between the transmitting and receiving frequency bands. The data of the transmission frequency band insertion loss characteristic matrix, the receiving frequency band insertion loss characteristic matrix and the frequency band isolation characteristic vector are fused to form an overall frequency response data matrix. Frequency response data matrix It is expressed by the following formula:

[0051] ;

[0052] The matrix combines the insertion loss data of the transmit band, the insertion loss data of the receive band, and the band isolation characteristic data to form a complete frequency response data set, reflecting the performance characteristics of the duplexer in different frequency bands. Within the preset temperature range, the duplexer is scanned across the entire frequency band and the gain change data is recorded. The temperature range is usually between -55°C and 125°C. At each temperature point, the gain of the duplexer will change, and these changes will form a temperature-gain characteristic matrix , expressed as:

[0053] ;

[0054] in, is at temperature The temperature-gain characteristic matrix can provide data on the effect of temperature on gain. Based on the temperature-gain characteristic matrix, a thermodynamic analysis is performed. By calculating the temperature distribution of the duplexer under maximum power conditions, the temperature distribution characteristic vector is obtained. , expressed as:

[0055] ;

[0056] This vector reflects the temperature distribution of the duplexer under different frequency bands and operating power conditions, revealing how temperature changes affect the internal structure and performance of the duplexer. Based on the temperature distribution eigenvector, the stress distribution and deformation of the duplexer are calculated. These calculation results are obtained through the stress-deformation eigenvector express:

[0057] ;

[0058] The stress-deformation eigenvector reveals the deformation caused by internal stress at different frequency bands and temperatures, which helps to evaluate the mechanical stability and long-term reliability of the duplexer. The temperature-gain characteristic matrix, temperature distribution eigenvector and stress-deformation eigenvector are fused to obtain the temperature characteristic data matrix. The matrix contains data on temperature, gain, stress, deformation, etc., which can fully describe the performance changes of the duplexer under different operating temperature conditions. The temperature characteristic data matrix is ​​expressed by the following formula:

[0059] .

[0060] In one example, the frequency response data is segmented and mapped according to the transmitting frequency band and the receiving frequency band, a dynamic frequency feature graph is established, and features of the dynamic frequency feature graph are extracted through a graph neural network to generate frequency band coupling correlation features, including:

[0061] The frequency response data is segmented according to the transmitting frequency band and the receiving frequency band, and each frequency band is divided into several sub-frequency bands to obtain a frequency band segmentation matrix;

[0062] A corresponding node is established for each sub-band in the frequency band segmentation matrix, and the connection weights between the nodes are calculated according to the frequency interval between the frequency bands to obtain an initial frequency feature map;

[0063] A graph convolution layer is constructed based on the initial frequency feature map, and spatial domain convolution operations are performed on node features to obtain a node information aggregation matrix;

[0064] Perform nonlinear transformation and pooling operations on the node information aggregation matrix to extract the topological relationship features between nodes and obtain the graph feature encoding matrix;

[0065] Input the graph feature encoding matrix into the attention mechanism network, calculate the correlation between nodes in different frequency bands, obtain the frequency band attention weight matrix, and perform weighted update on the graph feature encoding matrix based on the frequency band attention weight matrix to generate a dynamic frequency feature map;

[0066] The graph structure features of the dynamic frequency feature graph are learned to extract the coupling relationship between frequency bands to obtain the coupling feature vector, which is then input into a fully connected neural network for feature fusion to obtain the frequency band coupling correlation features.

[0067] In this example, the frequency response data is segmented according to the transmit band and the receive band. Assume that there is a duplexer frequency response data matrix , record the frequency response values ​​in different frequency bands. Divide the entire frequency response data into several sub-bands according to the transmit frequency band and the receive frequency band. Each sub-band contains several frequency points, each of which has a corresponding frequency response value. Therefore, the frequency band segmentation matrix is ​​expressed as:

[0068] ;

[0069] in and Represents the transmit frequency band and the receive frequency band respectively. The frequency response data of each sub-band. The partitioning matrix divides the frequency band into several sub-bands, each of which corresponds to a specific frequency range. Construct the initial frequency feature map. Create a corresponding node for each sub-band. Assume that each sub-band corresponds to a node in the frequency band partitioning matrix, and regard these nodes as vertices in the graph. According to the frequency interval between each sub-band, calculate the connection weights between nodes to form a weighted adjacency matrix . Adjacency Matrix Elements in Indicates The node and The connection strength between nodes (i.e., frequency interval). The smaller the frequency interval, the closer the connection between nodes, and the greater the weight. The connection weight is defined according to the inverse relationship of the frequency interval:

[0070] ;

[0071] in, and The nodes are and nodes The center frequency of the corresponding frequency band, is a parameter that controls the attenuation of connection strength. Based on the initial frequency feature map, a graph convolution layer is constructed. The graph convolution operation propagates information between nodes in the graph and updates the features of the nodes by aggregating the features of the neighboring nodes of each node. The mathematical expression of graph convolution is:

[0072] ;

[0073] in, It is The node feature matrix of the layer, is the normalized adjacency matrix, It is The weight matrix of the layer, is an activation function (such as ReLU). Through the graph convolution operation, the frequency response data is convolved in the spatial domain to obtain the aggregation matrix of node information. , which reflects the characteristic information of each sub-band and integrates the influence of surrounding bands. Perform nonlinear transformation and pooling operations on the node information aggregation matrix to extract the topological relationship features between nodes and reduce the data dimension. Perform nonlinear activation on the node information aggregation matrix, such as using the ReLU function to perform nonlinear transformation on the node information:

[0074] ;

[0075] Perform pooling operations, using maximum pooling or average pooling, to retain important features and downsample the features of each node:

[0076] ;

[0077] The pooling operation helps to extract more representative topological features, thereby reducing the computational complexity and overfitting risk of the model. , get the graph feature encoding matrix , which represents the high-order features of the topological structure between nodes. The graph feature encoding matrix is ​​input into the attention mechanism network. The attention mechanism strengthens important node features by adaptively calculating the degree of association between nodes in different frequency bands. By calculating the attention weight matrix , and the correlation strength between frequency bands is obtained:

[0078] ;

[0079] in, Representation Node The set of neighbor nodes of and The nodes are and nodes Features, is the weight matrix. By calculating the attention weights, we get the weighted graph feature matrix:

[0080] ;

[0081] The weighted graph feature matrix contains the correlation between frequency bands and can adaptively emphasize the frequency bands that are more important for performance optimization. Through graph structure feature learning, the coupling relationship between frequency bands is extracted. Perform deep learning to obtain the coupling feature vector , which contains the coupling information between different frequency bands. The coupling feature vector is input into the fully connected neural network for feature fusion to obtain the final frequency band coupling correlation feature :

[0082] ;

[0083] Through these steps, features related to frequency band coupling are extracted from the frequency response data and used for subsequent performance optimization.

[0084] In one example, graph structure feature learning is performed on the dynamic frequency feature graph to extract the coupling relationship between frequency bands to obtain a coupling feature vector, and the coupling feature vector is input into a fully connected neural network for feature fusion to obtain frequency band coupling correlation features, including:

[0085] Perform graph spectrum decomposition on the node feature matrix in the dynamic frequency feature graph to obtain the eigenvalue matrix and eigenvector matrix;

[0086] Based on the eigenvalue matrix and eigenvector matrix, the graph Fourier transform kernel is constructed, and the node features are transformed into the frequency domain to obtain the frequency domain feature matrix;

[0087] The frequency domain feature matrix is ​​input into the graph convolutional neural network to extract the high-order topological relationship between nodes and obtain the deep graph feature matrix;

[0088] Perform edge attention calculation on the deep graph feature matrix to obtain the edge weight matrix, and update the node features based on the edge weight matrix to obtain the coupling feature matrix;

[0089] Perform skip connection on the coupled feature matrix, fuse feature information at different levels, obtain multi-scale feature vectors, and input the multi-scale feature vectors into the residual network for feature transformation to obtain coupled feature vectors;

[0090] The coupled feature vector is batch normalized and activated to obtain a normalized feature vector, which is then nonlinearly mapped to obtain the frequency band coupling correlation feature.

[0091] In this example, the node feature matrix is ​​decomposed to extract the core feature patterns from the frequency response data of the graph for subsequent frequency domain transformation and high-order graph structure learning. The goal of graph spectral decomposition is to obtain the eigenvalue matrix and eigenvector matrix of the graph through eigenvalue decomposition. Given a graph adjacency matrix and the node feature matrix ,in is a dimensional matrix, representing The characteristics of each node are Features. For the adjacency matrix Perform spectral decomposition, that is, perform eigenvalue decomposition on it:

[0092] ;

[0093] in, is the eigenvector matrix, is the eigenvalue matrix, is the inverse of the eigenvector matrix. Through spectral decomposition, the adjacency matrix of the graph is Convert to frequency domain representation to process the relationship between nodes. Based on the eigenvalue matrix and the eigenvector matrix , construct the graph Fourier transform kernel. Graph Fourier transform is a method to convert the signal of a graph into a frequency domain representation, and the frequency domain transformation of the node features is performed by multiplying the eigenvalue and the eigenvector. The frequency domain transformation of the node features is performed by the following steps:

[0094] ;

[0095] in, It is the frequency domain representation of node features. Through the graph Fourier transform, we get the frequency domain feature matrix , reflecting the distribution of node features in the frequency domain. The spectral information of the graph is used to effectively capture the global structure of the graph and extract important patterns in the frequency domain. The frequency domain feature matrix Input into the graph convolutional neural network. The graph convolutional neural network is a deep learning model that transfers and aggregates information through adjacency matrices and node features. The graph convolution layer performs convolution operations on frequency domain features to extract high-order topological relationships between nodes. Suppose there is a graph convolution layer, whose operation is represented as:

[0096] ;

[0097] in, It is The node feature matrix of the layer, is the normalized adjacency matrix, It is The weight matrix of the layer, is the activation function (such as ReLU). Through graph convolution operation, node features As the number of network layers increases, more information from neighboring nodes will be gradually aggregated to obtain a deep graph feature matrix , which reflects the high-order topological relationship between nodes in the graph. Edge attention is calculated on the deep graph feature matrix. By calculating the edge weights between nodes, the information of important edges in the graph is strengthened. The edge attention mechanism adjusts the update process of node features by evaluating the relative importance of each edge. The edge weight matrix is ​​calculated using the following formula :

[0098] ;

[0099] in, and Is a node and nodes The eigenvector of Representation Node The set of neighbor nodes of is a shared weight matrix. The purpose of edge weight calculation is to assign different importance to different edges through the attention mechanism, so that the model can adaptively focus on the most critical graph structure information. By normalizing the edge weight matrix, we can get the weighted update of each node under the influence of its neighbor nodes. , update the node features and obtain the coupling feature matrix , which represents the coupling relationship between frequency bands. The update of node features is expressed as:

[0100] ;

[0101] The coupling information between frequency bands is extracted through the updated coupling feature matrix. These coupling features are the result of mutual influence and adjustment between frequency bands, and can reflect the complex coupling characteristics of the duplexer under multi-band operation. Processed by skip connection. Skip connection is a neural network structure that can fuse feature information at different levels to enhance information flow. Skip connection directly concatenates or adds the features of the previous layer with the features of the current layer, thereby improving the expressiveness of the model. Skip connection helps fuse frequency band coupling features at different levels to obtain more comprehensive multi-scale feature information. The process of skip connection is expressed as:

[0102] ;

[0103] After skip connection, we get a multi-scale feature vector , contains information from different levels. After further feature transformation of the residual network, the final coupled feature vector is obtained The residual network directly jumps to connect features at different levels to reduce the problem of gradient disappearance and make feature transformation more efficient. The operation process of the residual network is expressed as:

[0104] ;

[0105] Coupled eigenvector Perform batch normalization and activation function processing to obtain a normalized feature vector Batch normalization helps speed up the training process and stabilize the learning process of the model. Activation functions (such as ReLU) introduce nonlinear transformations, allowing the model to learn more complex features. The normalized feature vector Represents the final frequency band coupling correlation characteristics.

[0106] In one example, the interlayer thickness analysis model and the interface stress calculation model are used to calculate and optimize the metal layer thickness ratio and interface characteristics in the multilayer electrode structure according to the frequency band coupling correlation characteristics, and generate the multilayer electrode stress dispersion control parameters, including:

[0107] The frequency band coupling correlation characteristics are input into the thermal expansion coefficient difference calculation model to analyze the thermal expansion coefficient of each metal layer in the multi-layer structure and obtain the inter-layer thermal expansion coefficient matrix;

[0108] A stress distribution model is constructed based on the interlayer thermal expansion coefficient matrix, and the stress distribution characteristics of each interface in the multilayer electrode structure are calculated to obtain the interface stress distribution vector;

[0109] The stress gradient of the interface stress distribution vector is calculated, and the uniform stress dispersion constraint condition is established to obtain the stress dispersion constraint matrix. The thickness ratio of each metal layer is iteratively optimized according to the stress dispersion constraint matrix to obtain the interlayer thickness optimization parameter.

[0110] The optimized parameters of interlayer thickness are input into the acoustic migration effect analysis model, the stress change of the electrode structure under the action of continuous wave power is calculated, and the power stress characteristic vector is obtained. Based on the power stress characteristic vector, a temperature field distribution model is established to calculate the temperature distribution characteristics of the electrode structure and obtain the temperature field distribution matrix.

[0111] Thermal stress analysis is performed on the temperature field distribution matrix, and the stress and deformation parameters of the device under maximum power conditions are calculated to obtain the stress deformation parameter matrix. The stress deformation parameter matrix and the interlayer thickness optimization parameters are weightedly fused to obtain the multilayer electrode stress dispersion control parameters.

[0112] In this example, the frequency band coupling correlation feature is input into the thermal expansion coefficient difference calculation model to analyze the thermal expansion coefficient of each metal layer in a multilayer structure. The thermal expansion coefficient is an important parameter that describes the rate of volume change of a material due to temperature changes. It is represented by the symbol For each layer of metal in a multilayer structure, its thermal expansion coefficient It is calculated by the following relationship:

[0113] ;

[0114] in, Display material The length of the layer varies, is the initial length, is the thermal expansion coefficient of the material, is the temperature change. Through the input of the frequency band coupling correlation feature, the temperature change response of each layer of material is obtained, and the thermal expansion coefficient of each metal layer is calculated. These thermal expansion coefficients form a matrix , each element of which Indicates Layer metal and The difference in thermal expansion coefficient between the layers of metal. Based on this interlayer thermal expansion coefficient matrix , construct a stress distribution model. Since multi-layer metal structures usually generate stress under the action of heat and force, especially when the thermal expansion coefficients or material properties of different layers are different, the stress difference between the interfaces will have a significant impact on the overall structure. The stress distribution model needs to take into account the stress caused by the thermal expansion difference between different material layers on the interface. Assume that there is an interface stress between each metal layer , then the interface stress distribution vector Calculated by the following model:

[0115] ;

[0116] in, It is The elastic modulus of the layer material, is the strain of the layer, is the cross-sectional area of ​​the layer. In this way, according to the stress distribution and thermal expansion coefficient differences of different layers, the stress distribution characteristics of each interface are obtained and expressed as the interface stress distribution vector ,in is the number of interfaces. The stress distribution vector on the interface Perform stress gradient calculation. Stress gradient indicates the rate at which stress distribution changes with position. By calculating the gradient of stress distribution, the constraint condition for uniform stress dispersion is obtained, that is, the distribution of stress between different layers should be as uniform as possible. Stress gradient It is expressed by the following formula:

[0117] ;

[0118] in, is the distance between two adjacent layers, and They are Layer and The stress value between layers. The location with large stress gradient is usually the stress concentration point, which needs to be optimized. Based on the stress gradient, the stress uniform dispersion constraint condition is established to obtain the stress dispersion constraint matrix . This matrix describes the stress distribution and gradient constraints of each metal layer, and can guide the subsequent interlayer thickness optimization. Through iterative optimization, the thickness ratio of each metal layer is adjusted according to the stress dispersion constraint matrix to make the stress distribution more uniform. The interlayer thickness optimization is solved by an optimization algorithm (such as the gradient descent method) to obtain the interlayer thickness optimization parameters. . Optimize the interlayer thickness parameter Input into the acoustic migration effect analysis model. The acoustic migration effect refers to the propagation of sound waves in multilayer structures, which is used to analyze the thermal and mechanical effects caused by electromagnetic waves. The stress change of the electrode structure under continuous wave power is affected by temperature and mechanics. This change is expressed by the power stress eigenvector Modeling is performed. The power stress eigenvector is expressed as:

[0119] ;

[0120] in, is the time-varying stress, is the duration. Through this model, the stress response of the electrode structure under the action of continuous wave power is calculated. Based on the power stress eigenvector , build a temperature field distribution model. Since the effect of power will cause the temperature of the electrode structure to change, the change of temperature field is simulated by the heat conduction equation:

[0121] ;

[0122] in, is the temperature, is the thermal diffusivity, is the material density, is the specific heat capacity, is the power stress eigenvector. By solving this equation, we can obtain the temperature field distribution matrix of the electrode structure: , describing the temperature distribution characteristics at different locations. The temperature field distribution matrix Perform thermal stress analysis. Thermal stress analysis takes into account the stress changes caused by temperature changes and uses the thermal stress formula for calculation:

[0123] ;

[0124] in, is heat stress, is the elastic modulus, is the coefficient of thermal expansion, is the temperature change. Through analysis, the stress and deformation parameters of the electrode structure under maximum power conditions are calculated, and the stress deformation parameter matrix is ​​obtained. . The stress-deformation parameter matrix and interlayer thickness optimization parameters Perform weighted fusion to obtain the final multilayer electrode stress dispersion control parameters This parameter can achieve stress optimization of the multilayer electrode structure, thereby improving the performance and stability of the duplexer.

[0125] In one example, the temperature characteristic data is input into the frequency drift compensation model to establish a temperature-frequency mapping relationship, and the compensation voltage is dynamically adjusted through an active bias circuit to generate a temperature compensation control matrix, including:

[0126] The temperature characteristic data is input into the frequency drift compensation model to perform frequency response analysis, and a frequency drift characteristic mapping matrix is ​​obtained;

[0127] The frequency drift characteristic mapping matrix is ​​input into the temperature correlation analysis model, and the frequency response changes in the temperature range of -55°C to 125°C are analyzed to obtain the temperature-frequency correspondence matrix;

[0128] The drift characteristics in the temperature-frequency correspondence matrix are fitted piecewise to obtain the compensation parameter distribution vector, and the compensation control strategy of the active bias circuit is designed based on the compensation parameter distribution vector. The compensation voltage in each temperature interval is calculated to obtain the voltage compensation parameter matrix.

[0129] Input the voltage compensation parameter matrix into the dynamic feedback controller, establish a closed-loop adjustment model of the compensation voltage, obtain a dynamic adjustment control vector, perform temperature compensation effect analysis on the dynamic adjustment control vector, calculate the gain change of each frequency band, and obtain a compensation effect evaluation matrix;

[0130] The compensation effect evaluation matrix is ​​input into the compensation optimization model, the compensation parameters are iteratively optimized to obtain the optimized compensation control vector, and the optimized compensation control vector and the compensation parameter distribution vector are weightedly fused to obtain the temperature compensation control matrix.

[0131] In this example, the temperature characteristic data is input into the frequency drift compensation model for frequency response analysis. The effect of temperature on frequency response is manifested as frequency drift. Temperature changes will cause the parameters of components in the circuit to change, which in turn affects their operating frequency. In order to compensate for this drift, the frequency response is modeled. Assume that the frequency response of the device Changes with temperature, temperature characteristics data As input, frequency drift analysis is performed. Frequency drift feature map matrix Calculated by:

[0132] ;

[0133] in, It is the temperature The frequency response of is the frequency response at standard temperature, is the frequency drift characteristic matrix, which indicates the change of frequency response under different temperatures. In this way, the frequency drift characteristic matrix under different temperature conditions is obtained, which reflects the trend of device frequency response changing with temperature. The frequency drift characteristic mapping matrix is ​​input into the temperature correlation analysis model, and the frequency response change in the temperature range of -55°C to 125°C is analyzed to obtain the temperature-frequency correspondence matrix The matrix describes the change of frequency response at different temperatures and is obtained by regression analysis or interpolation method. In this process, the corresponding relationship between temperature and frequency is established through the following relationship:

[0134] ;

[0135] in, are the various temperature points within the temperature range, is at temperature By analyzing these data, the trend of frequency response changing with temperature is obtained. In order to accurately compensate for frequency drift, the drift characteristics in the temperature-frequency correspondence matrix are segmented and fitted to obtain the compensation parameter distribution vector This fitting can be done using polynomial fitting, piecewise linear fitting or spline fitting. The specific method is determined by the temperature characteristics of the system. Assume that in each temperature range, the frequency drift characteristic Expressed as a linear function:

[0136] ;

[0137] in, and are the fitting parameters, is the temperature. By performing segmented fitting on the frequency drift characteristics in different intervals, the compensation parameter distribution vector is obtained: , which represents the compensation parameters required in different temperature ranges. On this basis, based on the compensation parameter distribution vector , design the compensation control strategy of the active bias circuit. The design of the circuit mainly compensates the frequency drift caused by temperature by adjusting the voltage. According to the compensation parameter distribution vector, calculate the compensation voltage for each temperature range , the voltage and frequency drift characteristics have the following relationship:

[0138] ;

[0139] in, is the voltage compensation coefficient, is the frequency drift, is the voltage applied to compensate for temperature drift. In this way, the required compensation voltage is calculated for each temperature range, and the voltage compensation parameter matrix is ​​obtained. , which reflects the voltage to be applied at different temperatures. The input is sent to the dynamic feedback controller to establish a closed-loop regulation model for the compensation voltage. The dynamic feedback controller is used to adjust the voltage in real time to ensure the accuracy and stability of frequency drift compensation. Assuming the output frequency response of the system With compensation voltage The following relationship exists:

[0140] ;

[0141] in, is the frequency response after compensation, is the feedback gain coefficient, which indicates the effect of the compensation voltage on the frequency response. Through the feedback controller, the compensation voltage is dynamically adjusted to ensure that the frequency response reaches the expected value. Analyze the temperature compensation effect, calculate the gain change of each frequency band, and obtain the compensation effect evaluation matrix The gain change is expressed by the following formula:

[0142] ;

[0143] in, is the gain after compensation, is the standard gain, is the gain change in the compensation effect evaluation matrix. By analyzing the gain change, the compensation effect is evaluated and a basis is provided for subsequent optimization. Input the compensation optimization model to iteratively optimize the compensation parameters. The optimization algorithm uses gradient descent method, genetic algorithm and other methods to iteratively update the control vector of the compensation voltage with the goal of minimizing the gain change or maximizing the stability of the frequency response. , and obtain the optimized compensation control vector . Compensation control vector is optimized by weighted fusion and compensation parameter distribution vector , and finally get the temperature compensation control matrix , the matrix provides an accurate temperature compensation control strategy to maintain the frequency stability and performance of the device at different temperatures.

[0144] In one example, the frequency band coupling correlation characteristics are decoupled into the transmitting frequency band and the receiving frequency band, and the multi-layer electrode stress dispersion control parameters and the temperature compensation control matrix are combined for parallel calculation to generate the duplexer frequency band performance optimization parameters, including:

[0145] Perform frequency band decoupling operation on the frequency band coupling correlation characteristics in the 703-748MHz transmission frequency band and the 758-803MHz reception frequency band to obtain independent transmission frequency band feature vectors and reception frequency band feature vectors;

[0146] Based on the eigenvector of the transmitting frequency band and the eigenvector of the receiving frequency band, the stress dispersion control parameters of the multilayer electrode are decoupled and calculated to obtain the stress parameters of the transmitting end and the stress parameters of the receiving end;

[0147] The stress distribution data of the duplexer under 30dBm continuous wave power is calculated according to the stress parameters of the transmitting end and the receiving end, and the power stress response matrix is ​​obtained;

[0148] The temperature compensation control matrix and the power stress response matrix are input into the multi-band joint compensation model, the frequency compensation parameters in the temperature range of -55°C to 125°C are calculated, and the frequency-temperature compensation parameter matrix is ​​obtained;

[0149] The Tx passband insertion loss and the Rx passband insertion loss of the duplexer are calculated in parallel according to the frequency-temperature compensation parameter matrix to obtain the duplexer insertion loss optimization vector, and the isolation data between the Tx frequency band and the Rx frequency band is calculated based on the duplexer insertion loss optimization vector to obtain the frequency band isolation characteristic matrix;

[0150] The frequency band isolation characteristic matrix is ​​input into the performance optimization model. Through iterative calculation, the Tx passband insertion loss is less than 1.6dB and the Rx passband insertion loss is less than 1.7dB. The performance optimization parameter vector is obtained, and the performance optimization parameter vector is normalized to obtain the duplexer frequency band performance optimization parameters.

[0151] In this example, the frequency band coupling correlation features are decoupled, and the frequency band features of the 703-748MHz transmit frequency band and the 758-803MHz receive frequency band are extracted respectively, and independent transmit frequency band feature vectors and receive frequency band feature vectors are obtained. The coupling relationship between the transmit frequency band and the receive frequency band cannot be ignored. The frequency band decoupling operation is used to remove these coupling effects and ensure the independence of the features of each frequency band. Assume that the frequency band coupling feature matrix is , the decoupled transmission frequency band feature vector and the receiving frequency band feature vector Extraction is done as follows:

[0152]

[0153] ;

[0154] The decoupling operation is based on the correlation matrix between the frequency bands, and uses methods such as singular value decomposition or other matrix decomposition methods to extract independent feature vectors of the transmit frequency band and the receive frequency band. and the receiving frequency band feature vector , decouple the multi-layer electrode stress dispersion control parameters to obtain the transmitter stress parameters and the receiver stress parameters. The purpose of stress dispersion control is to reduce the stress concentration between different metal layers caused by temperature or power fluctuations to ensure the reliability of the structure. By establishing a relationship model between the characteristics of the transmitting and receiving frequency bands and the electrode stress, the transmitter stress parameters are obtained. and receiving end stress parameters :

[0155] ;

[0156] ;

[0157] in, and Represent the stress distribution matrix of the transmitter and the receiver respectively. Based on the finite element analysis model, the stress characteristics are decoupled from the frequency band characteristics through decoupling technology to obtain the stress parameters of two different frequency bands. After obtaining the stress parameters of the transmitter and the receiver, the stress distribution data of the duplexer under 30dBm continuous wave power is calculated to obtain the power stress response matrix. Power stress response matrix Describe the stress distribution of the duplexer under given power conditions:

[0158] ;

[0159] in, is the continuous wave power, Reflects the stress response of each position inside the duplexer under a specific power. This matrix is ​​obtained by numerical calculation or simulation. and the power stress response matrix Input the multi-band joint compensation model, calculate the frequency compensation parameters in the temperature range of -55°C to 125°C, and obtain the frequency-temperature compensation parameter matrix The matrix represents the compensation parameters of the frequency response at different temperatures, calculated by the following model:

[0160] ;

[0161] in, Indicates the temperature range, is the temperature compensation control matrix, is the power stress response matrix. Through this model, the compensation parameters of the frequency response in the entire temperature range are obtained. According to the frequency-temperature compensation parameter matrix , the Tx passband insertion loss and Rx passband insertion loss of the duplexer are calculated in parallel. Assume that the insertion loss of the duplexer is expressed as and , respectively represent the insertion loss of the transmit passband and the receive passband, and the calculation formula is:

[0162] ;

[0163] ;

[0164] By calculation, the insertion loss optimization vector of the duplexer in the transmitting and receiving frequency bands is obtained. Insertion loss optimization vector , calculate the isolation data between the Tx band and the Rx band , and obtain the frequency band isolation characteristic matrix . Frequency band isolation It is expressed by the following formula:

[0165] ;

[0166] This formula shows the relationship between the isolation between frequency bands and the insertion loss. The greater the isolation, the smaller the interference between the transmit and receive frequency bands of the duplexer. By calculating the isolation characteristic matrix , to evaluate the inter-band isolation performance of the duplexer. The frequency band isolation characteristic matrix Input the performance optimization model and optimize the duplexer performance through iterative calculation to ensure that the Tx passband insertion loss is less than 1.6dB and the Rx passband insertion loss is less than 1.7dB. Use an optimization algorithm, such as gradient descent or genetic algorithm, to optimize the performance optimization parameter vector Solve it to satisfy the performance constraints:

[0167] ;

[0168] Get the performance optimization parameter vector , and normalize the vector to obtain the duplexer frequency band performance optimization parameters Through these optimized parameters, the performance of the duplexer in the transmit and receive frequency bands is improved to ensure that it meets the design requirements.

[0169] Reference Figure 2 This embodiment provides a duplexer multi-band performance optimization system, including:

[0170] Acquisition module 1, used to collect frequency response data and temperature characteristic data of the duplexer in the transmitting frequency band and the receiving frequency band;

[0171] Mapping module 2, used to segmentally map the frequency response data according to the transmitting frequency band and the receiving frequency band, establish a dynamic frequency characteristic graph, and extract features from the dynamic frequency characteristic graph through a graph neural network to generate frequency band coupling correlation features;

[0172] Calculation module 3, used to calculate and optimize the metal layer thickness ratio and interface characteristics in the multilayer electrode structure according to the frequency band coupling correlation characteristics by using the interlayer thickness analysis model and the interface stress calculation model, and generate the multilayer electrode stress dispersion control parameters;

[0173] The adjustment module 4 is used to input the temperature characteristic data into the frequency drift compensation model, establish a temperature-frequency mapping relationship, and dynamically adjust the compensation voltage through an active bias circuit to generate a temperature compensation control matrix;

[0174] The generation module 5 is used to decouple the transmission frequency band and the reception frequency band according to the frequency band coupling correlation characteristics, and to perform parallel calculations in combination with the multi-layer electrode stress dispersion control parameters and the temperature compensation control matrix to generate duplexer frequency band performance optimization parameters.

[0175] In this embodiment, for the specific implementation of each unit in the above system embodiment, please refer to the above method embodiment, which will not be repeated here.

[0176] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0177] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0178] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0179] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0180] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0181] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for optimizing the multi-band performance of a duplexer, characterized in that: The following steps are involved: Collect frequency response data and temperature characteristic data of the duplexer in the transmit frequency band and the receive frequency band; Segmentally map the frequency response data according to the transmitting frequency band and the receiving frequency band, establish a dynamic frequency characteristic graph, and extract features from the dynamic frequency characteristic graph through a graph neural network to generate frequency band coupling correlation features; Using an interlayer thickness analysis model and an interface stress calculation model, the thickness ratio of metal layers and interface characteristics in the multilayer electrode structure are calculated and optimized according to the frequency band coupling correlation characteristics, and the multilayer electrode stress dispersion control parameters are generated; Inputting the temperature characteristic data into a frequency drift compensation model, establishing a temperature-frequency mapping relationship, and dynamically adjusting the compensation voltage through an active bias circuit to generate a temperature compensation control matrix; The frequency band coupling correlation characteristics are decoupled into a transmitting frequency band and a receiving frequency band, and parallel calculations are performed in combination with the multi-layer electrode stress dispersion control parameters and the temperature compensation control matrix to generate duplexer frequency band performance optimization parameters.

2. The method for optimizing the multi-band performance of a duplexer according to claim 1, characterized in that: The collecting of frequency response data and temperature characteristic data of the duplexer in the transmitting frequency band and the receiving frequency band includes: Performing a frequency scan on the transmitting frequency band of the duplexer within the target range, collecting insertion loss data of the transmitting port, and obtaining an insertion loss characteristic matrix of the transmitting frequency band, and performing a frequency scan on the receiving frequency band of the duplexer within the target range, collecting insertion loss data of the receiving port, and obtaining an insertion loss characteristic matrix of the receiving frequency band; Inputting the transmission frequency band insertion loss characteristic matrix and the reception frequency band insertion loss characteristic matrix into a spectrum analysis model, calculating the isolation characteristic between the transmission frequency band and the reception frequency band, and obtaining a frequency band isolation characteristic vector; Performing data fusion on the transmission frequency band insertion loss characteristic matrix, the reception frequency band insertion loss characteristic matrix and the frequency band isolation characteristic vector to obtain frequency response data; Performing a full-band scan on the duplexer within a preset temperature range, recording gain change data, obtaining a temperature-gain characteristic matrix, and performing a thermodynamic analysis based on the temperature-gain characteristic matrix to calculate the temperature distribution of the duplexer under maximum power conditions, and obtaining a temperature distribution characteristic vector; The stress distribution and deformation of the duplexer are calculated based on the temperature distribution characteristic vector to obtain the stress-deformation characteristic vector, and the temperature-gain characteristic matrix, the temperature distribution characteristic vector and the stress-deformation characteristic vector are fused to obtain temperature characteristic data.

3. The method for optimizing the multi-band performance of a duplexer according to claim 1, characterized in that: The step of segmentally mapping the frequency response data according to the transmitting frequency band and the receiving frequency band, establishing a dynamic frequency feature graph, and extracting features from the dynamic frequency feature graph through a graph neural network to generate frequency band coupling correlation features includes: The frequency response data is segmented according to the transmitting frequency band and the receiving frequency band, and each frequency band is divided into a plurality of sub-frequency bands to obtain a frequency band segmentation matrix; Establishing a corresponding node for each sub-band in the frequency band segmentation matrix, and calculating the connection weights between the nodes according to the frequency intervals between the frequency bands to obtain an initial frequency characteristic graph; Constructing a graph convolution layer based on the initial frequency feature graph, performing spatial domain convolution operations on node features, and obtaining a node information aggregation matrix; Performing nonlinear transformation and pooling operations on the node information aggregation matrix to extract topological relationship features between nodes and obtain a graph feature encoding matrix; Input the graph feature encoding matrix into the attention mechanism network, calculate the correlation degree between nodes in different frequency bands, obtain the frequency band attention weight matrix, and perform weighted update on the graph feature encoding matrix based on the frequency band attention weight matrix to generate a dynamic frequency feature map; The dynamic frequency feature graph is subjected to graph structure feature learning, the coupling relationship between frequency bands is extracted, a coupling feature vector is obtained, and the coupling feature vector is input into a fully connected neural network for feature fusion to obtain frequency band coupling association features.

4. The method for optimizing the multi-band performance of a duplexer according to claim 3, characterized in that: The step of performing graph structure feature learning on the dynamic frequency feature graph, extracting the coupling relationship between frequency bands, obtaining a coupling feature vector, and inputting the coupling feature vector into a fully connected neural network for feature fusion to obtain a frequency band coupling correlation feature includes: Performing graph spectrum decomposition on the node feature matrix in the dynamic frequency feature graph to obtain an eigenvalue matrix and an eigenvector matrix; Constructing a graph Fourier transform kernel based on the eigenvalue matrix and the eigenvector matrix, performing frequency domain transformation on node features, and obtaining a frequency domain feature matrix; Inputting the frequency domain feature matrix into a graph convolutional neural network to extract high-order topological relationships between nodes and obtain a deep graph feature matrix; Performing edge attention calculation on the deep graph feature matrix to obtain an edge weight matrix, and updating node features based on the edge weight matrix to obtain a coupling feature matrix; Performing skip connection on the coupling feature matrix, fusing feature information at different levels to obtain a multi-scale feature vector, and inputting the multi-scale feature vector into a residual network for feature transformation to obtain a coupling feature vector; The coupling feature vector is batch normalized and activated to obtain a normalized feature vector, and the normalized feature vector is nonlinearly mapped to obtain a frequency band coupling correlation feature.

5. The method for optimizing the multi-band performance of a duplexer according to claim 1, characterized in that: The interlayer thickness analysis model and the interface stress calculation model are used to calculate and optimize the metal layer thickness ratio and interface characteristics in the multilayer electrode structure according to the frequency band coupling correlation characteristics, and generate the multilayer electrode stress dispersion control parameters, including: The frequency band coupling correlation characteristics are input into the thermal expansion coefficient difference calculation model, the thermal expansion coefficient of each metal layer in the multi-layer structure is analyzed, and the inter-layer thermal expansion coefficient matrix is ​​obtained; Constructing a stress distribution model based on the interlayer thermal expansion coefficient matrix, calculating stress distribution characteristics of each interface in the multilayer electrode structure, and obtaining an interface stress distribution vector; Performing stress gradient calculation on the interface stress distribution vector, establishing stress uniform dispersion constraint conditions, obtaining a stress dispersion constraint matrix, and iteratively optimizing the thickness ratio of each metal layer according to the stress dispersion constraint matrix to obtain an interlayer thickness optimization parameter; Inputting the interlayer thickness optimization parameter into the acoustic migration effect analysis model, calculating the stress change of the electrode structure under the action of continuous wave power, obtaining the power stress characteristic vector, and establishing a temperature field distribution model based on the power stress characteristic vector, calculating the temperature distribution characteristics of the electrode structure, and obtaining the temperature field distribution matrix; A thermal stress analysis is performed on the temperature field distribution matrix, and the stress and deformation parameters of the device under maximum power conditions are calculated to obtain a stress deformation parameter matrix. The stress deformation parameter matrix and the interlayer thickness optimization parameters are weightedly fused to obtain multilayer electrode stress dispersion control parameters.

6. The method for optimizing the multi-band performance of a duplexer according to claim 1, characterized in that: The temperature characteristic data is input into a frequency drift compensation model to establish a temperature-frequency mapping relationship, and a compensation voltage is dynamically adjusted through an active bias circuit to generate a temperature compensation control matrix, including: Inputting the temperature characteristic data into a frequency drift compensation model to perform frequency response analysis to obtain a frequency drift characteristic mapping matrix; The frequency drift characteristic mapping matrix is ​​input into the temperature correlation analysis model, and the frequency response change in the temperature range of -55°C to 125°C is analyzed to obtain a temperature-frequency correspondence matrix; Performing piecewise fitting on the drift characteristics in the temperature-frequency correspondence matrix to obtain a compensation parameter distribution vector, and designing a compensation control strategy for the active bias circuit based on the compensation parameter distribution vector, calculating the compensation voltage in each temperature interval to obtain a voltage compensation parameter matrix; Input the voltage compensation parameter matrix into a dynamic feedback controller, establish a closed-loop adjustment model of the compensation voltage, obtain a dynamic adjustment control vector, perform temperature compensation effect analysis on the dynamic adjustment control vector, calculate the gain change of each frequency band, and obtain a compensation effect evaluation matrix; The compensation effect evaluation matrix is ​​input into the compensation optimization model, the compensation parameters are iteratively optimized to obtain an optimized compensation control vector, and the optimized compensation control vector and the compensation parameter distribution vector are weightedly fused to obtain a temperature compensation control matrix.

7. The method for optimizing the multi-band performance of a duplexer according to claim 1, characterized in that: The decoupling of the frequency band coupling correlation characteristics between the transmitting frequency band and the receiving frequency band is performed, and the multi-layer electrode stress dispersion control parameters and the temperature compensation control matrix are combined for parallel calculation to generate duplexer frequency band performance optimization parameters, including: Performing frequency band decoupling operation on the frequency band coupling correlation characteristics in a 703-748 MHz transmission frequency band and a 758-803 MHz reception frequency band to obtain independent transmission frequency band feature vectors and reception frequency band feature vectors; Perform parameter decoupling calculation on the multi-layer electrode stress dispersion control parameter based on the transmitting frequency band characteristic vector and the receiving frequency band characteristic vector to obtain a transmitting end stress parameter and a receiving end stress parameter; Calculate stress distribution data of the duplexer at 30 dBm continuous wave power according to the transmitting end stress parameter and the receiving end stress parameter to obtain a power stress response matrix; Input the temperature compensation control matrix and the power stress response matrix into a multi-band joint compensation model, calculate the frequency compensation parameters in the temperature range of -55°C to 125°C, and obtain a frequency-temperature compensation parameter matrix; According to the frequency-temperature compensation parameter matrix, the Tx passband insertion loss and the Rx passband insertion loss of the duplexer are calculated in parallel to obtain a duplexer insertion loss optimization vector, and based on the duplexer insertion loss optimization vector, the isolation data between the Tx frequency band and the Rx frequency band is calculated to obtain a frequency band isolation characteristic matrix; The frequency band isolation characteristic matrix is ​​input into the performance optimization model, and the Tx passband insertion loss is made less than 1.6dB and the Rx passband insertion loss is made less than 1.7dB through iterative calculation, so as to obtain a performance optimization parameter vector, and the performance optimization parameter vector is normalized to obtain the duplexer frequency band performance optimization parameters.

8. A duplexer multi-band performance optimization system, characterized in that: For implementing the steps of the method according to any one of claims 1 to 7, the system comprises: An acquisition module, used to acquire frequency response data and temperature characteristic data of the duplexer in the transmitting frequency band and the receiving frequency band; A mapping module, used to segmentally map the frequency response data according to the transmitting frequency band and the receiving frequency band, establish a dynamic frequency characteristic graph, and extract features from the dynamic frequency characteristic graph through a graph neural network to generate frequency band coupling correlation features; A calculation module, for calculating and optimizing the thickness ratio and interface characteristics of the metal layers in the multilayer electrode structure according to the frequency band coupling correlation characteristics by using an interlayer thickness analysis model and an interface stress calculation model, and generating a multilayer electrode stress dispersion control parameter; An adjustment module, used for inputting the temperature characteristic data into a frequency drift compensation model, establishing a temperature-frequency mapping relationship, and dynamically adjusting the compensation voltage through an active bias circuit to generate a temperature compensation control matrix; A generation module is used to decouple the transmission frequency band and the reception frequency band of the frequency band coupling correlation characteristics, and to perform parallel calculations in combination with the multi-layer electrode stress dispersion control parameters and the temperature compensation control matrix to generate duplexer frequency band performance optimization parameters.

9. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images