Near field scanning fast imaging device for radio frequency performance of multi-channel electrically adjustable base station antenna

Abstract

The application provides a near-field scanning fast imaging device for radio frequency performance of a multi-channel electrically adjustable base station antenna, and relates to the technical field of radio frequency antenna testing.The device comprises a scanning component and a data processing module.The scanning component is composed of a radio frequency probe installed on a two-dimensional precision displacement platform and a connected vector network analyzer, is used for point-by-point scanning along a preset planar grid path in the near-field area of the antenna, and collects scattering parameters representing electromagnetic field characteristics at each sampling point to form a near-field scanning data set.The data processing module receives the data set and performs a fast Fourier transform operation to convert the near-field data into far-field radiation characteristics, thereby generating a far-field radiation pattern representing the radio frequency performance of the antenna.A preset static compensation matrix is applied to the near-field data by the data processing module before transformation to suppress the mutual coupling effect between the probe and the antenna and improve the accuracy of the measurement results.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of radio frequency antenna testing, in particular to a near-field scanning fast imaging device for radio frequency performance of a multi-channel electrically adjustable base station antenna. BACKGROUND

[0002] As a core component of a mobile communication system, the radio frequency performance of a base station antenna is directly related to signal coverage, system capacity and interference suppression capability. With the rapid development of 5G and future 6G communication technologies, base station antennas are gradually evolving towards multi-channel and large-scale electrically adjustable arrays. Such antennas usually have wideband, multi-beam and dynamic adjustment characteristics, and higher requirements are put forward for accurate testing of radio frequency performance indicators such as radiation pattern, gain and side lobe level.

[0003] Existing radio frequency performance testing methods mainly include far-field testing and traditional darkroom testing. Far-field testing usually requires a large-size test site to be laid out in an area where the antenna beams are fully expanded, and is subject to site conditions and cost limitations, with low testing efficiency; while darkroom far-field testing can simulate a free space environment, but has strict requirements for equipment arrangement and testing procedures, with a long testing period. For multi-channel electrically adjustable base station antennas, due to the large number of ports and complex beam combinations, traditional far-field testing is difficult to complete full coverage measurement within a limited time.

[0004] Therefore, how to provide a near-field scanning device capable of realizing fast measurement and imaging of radio frequency performance of a multi-channel electrically adjustable base station antenna in a limited space has become a technical problem to be solved in the field. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a near-field scanning fast imaging device for radio frequency performance of a multi-channel electrically adjustable base station antenna, comprising:

[0006] a scanning assembly, the scanning assembly comprising:

[0007] a radio frequency probe mounted on a two-dimensional precision displacement platform;

[0008] and a vector network analyzer connected to the radio frequency probe;

[0009] The two-dimensional precision displacement platform is used to drive the radio frequency probe to scan along a preset planar grid path in the near-field region of a multi-channel electrically adjustable base station antenna under test, and the vector network analyzer measures and collects scattering parameters representing electromagnetic field characteristics at each sampling point on the grid path to form a near-field scanning data set.

[0010] a data processing module connected to the scanning assembly, the data processing module being configured to:

[0011] receive the near-field scanning data set, and perform a fast Fourier transform operation on the near-field scanning data set to complete a near-field to far-field transformation to generate a far-field radiation pattern representing the radio frequency performance of the antenna.

[0012] Optionally, the data processing module is further configured to, before performing the near-field to far-field transformation operation, apply a preset, static compensation matrix to correct the near-field scanning data set to suppress mutual coupling effects between the probe and the antenna to be tested.

[0013] Optionally, comprising:

[0014] The scanning module comprises a scanning component that scans the near-field region of the multi-channel electrically adjustable base station antenna to generate near-field scanning data and transmits the near-field scanning data to the compensation module;

[0015] The antenna parameter module is configured to store and provide target parameters of the multi-channel electrically adjustable base station antenna to the compensation module;

[0016] The compensation module comprises an electromagnetic suppression structure and a large language model, wherein the electromagnetic suppression structure comprises an electromagnetic bandgap structure or a split ring resonator, and is configured to suppress mutual coupling components and cascaded interactions of edge scattering components in the near-field scanning data, and input the suppression results as suppression data to the large language model;

[0017] The large language model generates a block strategy and adaptive parameters based on the target parameters, and calculates a sub-mutual coupling scattering matrix by performing block processing on the near-field scanning data and the suppression data;

[0018] The result merging module is configured to splice the sub-mutual coupling scattering matrix to generate a mutual coupling scattering matrix;

[0019] The processing module is configured to perform matrix solving on the near-field scanning data based on the mutual coupling scattering matrix, and generate a radio frequency performance image by far-field transformation.

[0020] Optionally, the calculation of the sub-mutual coupling scattering matrix comprises:

[0021] The target parameters are converted into serialized prompts and input into the large language model;

[0022] The large language model generates a block strategy based on the serialized prompts, wherein the block strategy comprises sub-region division instructions for the near-field scanning data;

[0023] The large language model generates adaptive parameters based on the serialized prompts, wherein the adaptive parameters comprise prompts for adjusting the calculation;

[0024] generating block data based on the blocking strategy on the near-field scanning data and the suppression data;

[0025] processing the block data based on the adaptive parameters to calculate the sub-intercoupling scattering matrix.

[0026] Optionally, converting the target parameters into serialized prompts and inputting the serialized prompts into the large language model comprises:

[0027] obtaining the target parameters from the antenna parameter module, wherein the target parameters include scanning data size, compensation sub-block capacity, electromagnetic frequency range and dielectric constant gradient of the multi-channel electrically adjustable base station antenna;

[0028] converting the target parameters into serialized prompts, wherein the serialized prompts include numerical sequence of the scanning data size, limit sequence of the compensation sub-block capacity, and physical model description of the electromagnetic frequency range and the dielectric constant gradient;

[0029] inputting the serialized prompts into the large language model, wherein the large language model generates adaptive parameters containing electromagnetic constraints based on the physical model description.

[0030] Optionally, generating block data based on the blocking strategy on the near-field scanning data and the suppression data comprises:

[0031] identifying edge scattering sensitive areas in the near-field scanning data based on the blocking strategy, wherein the blocking strategy includes sub-region division instructions, and the sub-region division instructions integrate boundary intercoupling constraints;

[0032] performing dynamic blocking on the near-field scanning data according to the sub-region division instructions to generate sub-scanning data, wherein the dynamic blocking integrates wavelength-shortened scattering distribution;

[0033] performing corresponding blocking on the suppression data based on the blocking strategy to generate sub-suppression data, wherein the sub-suppression data corresponds to intercoupling component edge values in the sub-scanning data;

[0034] combining the sub-scanning data and the sub-suppression data to generate block data.

[0035] Optionally, splicing the sub-intercoupling scattering matrix to generate an intercoupling scattering matrix comprises:

[0036] receiving the sub-intercoupling scattering matrix from the compensation module and identifying a first boundary phase region based on the sub-intercoupling scattering matrix, wherein the first boundary phase region includes a phase difference value sequence between the sub-intercoupling scattering matrices;

[0037] calculating a first weight sequence according to the first boundary phase region, wherein the first weight sequence is generated based on an edge reflection distribution of the sub-intercoupling scattering matrix;

[0038] fusing and splicing the sub-intercoupling scattering matrix by applying the first weight sequence to generate a first preliminary intercoupling scattering matrix, wherein the fusing and splicing includes a weighted superposition operation of the sub-intercoupling scattering matrix;

[0039] performing array response verification on the first preliminary intercoupling scattering matrix, wherein the array response verification calculates a response consistency index of the first preliminary intercoupling scattering matrix, and adjusts the first preliminary intercoupling scattering matrix based on the response consistency index to generate the intercoupling scattering matrix.

[0040] Optionally, performing matrix solving on the near-field scanning data based on the intercoupling scattering matrix to generate a radio frequency performance image through far-field transformation includes:

[0041] receiving the intercoupling scattering matrix from the result merging module, and receiving the near-field scanning data from the scanning module;

[0042] performing preliminary matrix solving on the near-field scanning data based on the intercoupling scattering matrix to generate a first solving matrix, wherein the preliminary matrix solving includes inverse operation application of the intercoupling scattering matrix;

[0043] calculating a first distortion index according to the first solving matrix, wherein the first distortion index is generated based on the response consistency index, and includes a multiple scattering deviation value of the first solving matrix;

[0044] adjusting the first solving matrix by applying the first distortion index to generate a second solving matrix, wherein the adjustment includes adaptive regularization operation of the first solving matrix;

[0045] performing far-field transformation on the second solving matrix to generate the radio frequency performance image, wherein the far-field transformation includes Fourier integral operation of the second solving matrix.

[0046] Optionally, performing far-field transformation on the second solving matrix to generate the radio frequency performance image includes:

[0047] receiving the second solving matrix from the processing module, and calculating a first far-field integral sequence based on the second solving matrix, wherein the first far-field integral sequence includes a Fourier integral operation result of the second solving matrix;

[0048] identifying a first radiation pattern region according to the first far-field integral sequence, wherein the first radiation pattern region includes a polarization deviation value sequence in the first far-field integral sequence;

[0049] generating a first compensation sequence based on the first radiation pattern region, wherein the first compensation sequence comprises multiple scattering correction values of the polarization deviation value sequence;

[0050] adjusting the first far-field integral sequence by the first compensation sequence to generate a second far-field integral sequence, wherein the adjusting comprises a weighted correction operation of the first far-field integral sequence;

[0051] performing imaging mapping on the second far-field integral sequence to generate the radio frequency performance image, wherein the imaging mapping comprises a two-dimensional projection operation of the second far-field integral sequence.

[0052] Optionally, converting the target parameters into the serialization prompt comprises:

[0053] receiving the target parameters from the antenna parameter module, and extracting numerical data of the scan data size, limit data of the compensation sub-block capacity, range data of the electromagnetic frequency range, and gradient data of the permittivity gradient;

[0054] serializing the numerical data of the scan data size to generate a first numerical sequence, wherein the first numerical sequence comprises a dimension array of the scan data size;

[0055] serializing the limit data of the compensation sub-block capacity to generate a first limit sequence, wherein the first limit sequence comprises a boundary threshold array of the compensation sub-block capacity;

[0056] serializing the range data of the electromagnetic frequency range and the gradient data of the permittivity gradient by a physical model to generate a first physical model sequence, wherein the first physical model sequence comprises a frequency spectrum vector of the electromagnetic frequency range and a gradient vector of the permittivity gradient;

[0057] combining the first numerical sequence, the first limit sequence, and the first physical model sequence to generate the serialization prompt, wherein the serialization prompt comprises an integrated parameter sequence array.

[0058] Compared with the prior art, the application realizes compensation of mutual coupling and edge scattering composite effect through integration of electromagnetic suppression structure and large language model. Specifically, the device utilizes the antenna parameter module to provide relevant parameters, generates a block strategy and adaptive parameters by the large language model, performs block processing on the near-field scanning data, calculates a sub-mutual coupling scattering matrix, and generates a complete mutual coupling scattering matrix through the result merging module, which is used for matrix solving and far-field transformation of the processing module. The scheme improves the efficiency and accuracy of imaging processing, and is suitable for large-scale radio frequency data scenarios; at the same time, through the adaptive block mechanism, the computational overhead is reduced, and the problems of wavelength shortening and scattering non-uniformity in the millimeter wave band are adapted; in addition, the introduction of the electromagnetic suppression structure reduces the parasitic field interference and improves the accuracy of the radiation pattern; overall, the device enhances the real-time performance of batch testing and is suitable for optimization diagnosis of 5G / 6G antenna arrays. BRIEF DESCRIPTION OF DRAWINGS

[0059] Figure 1 A schematic diagram of a near-field scanning fast imaging device for multi-channel electrically adjustable base station antenna radio frequency performance provided by an embodiment of the application;

[0060] Figure 2 A flowchart of a method for calculating a sub-mutual coupling scattering matrix provided by an embodiment of the application;

[0061] Figure 3 A flowchart of a method for inputting target parameters into a large language model provided by an embodiment of the application;

[0062] Figure 4 A flowchart of a method for generating a mutual coupling scattering matrix provided by an embodiment of the application.

[0063] BRIEF DESCRIPTION OF DRAWINGS DETAILED DESCRIPTION

[0064] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application.

[0065] An embodiment of the application provides a near-field scanning fast imaging device for multi-channel electrically adjustable base station antenna radio frequency performance, comprising:

[0066] A scanning assembly, the scanning assembly comprising:

[0067] A radio frequency probe mounted on a two-dimensional precision displacement platform;

[0068] and a vector network analyzer connected to the radio frequency probe;

[0069] The two-dimensional precision displacement platform is configured to drive the radio frequency probe to scan along a preset planar grid path in a near-field region of the multi-channel electrically tunable base station antenna to be measured, and the vector network analyzer is configured to measure and collect scattering parameters representing electromagnetic field characteristics at each sampling point on the grid path to form a near-field scanning data set.

[0070] The data processing module is connected to the scanning assembly and is configured to:

[0071] receive the near-field scanning data set and perform a fast Fourier transform operation on the near-field scanning data set to complete a near-field to far-field conversion;

[0072] to generate a far-field radiation pattern representing the radio frequency performance of the antenna.

[0073] Optionally, the data processing module is further configured to apply a preset static compensation matrix to correct the near-field scanning data set before performing the near-field to far-field conversion operation, to suppress the mutual coupling effect between the probe and the antenna to be measured.

[0074] In a specific implementation, the scanning assembly includes a radio frequency probe installed on a two-dimensional precision displacement platform and a vector network analyzer electrically connected thereto. The two-dimensional precision displacement platform can accurately move the radio frequency probe along a preset planar grid path under the control of a control system, so that the radio frequency probe scans point by point in the near-field region of the multi-channel electrically tunable base station antenna to be measured. The vector network analyzer measures and collects scattering parameters (including amplitude and phase information) representing electromagnetic field characteristics at each scanning point, thereby forming a complete near-field scanning data set.

[0075] The data processing module is configured to receive the near-field scanning data set transmitted by the scanning assembly and perform a fast Fourier transform operation on the data set to realize the near-field to far-field conversion, thereby generating a far-field radiation pattern representing the radio frequency performance of the antenna to be measured. Through the above device, the antenna radio frequency performance imaging can be quickly completed under limited space conditions, and key parameters such as main lobe, side lobe, beam width and side lobe level can be output.

[0076] The data processing module is further configured to preprocess the original near-field scanning data set before performing the near-field to far-field conversion operation. Specifically, a preset static compensation matrix is applied to correct the near-field scanning data to suppress the mutual coupling effect between the probe and the antenna to be measured. The compensation matrix can be obtained through prior modeling or calibration experiments, and is used to correct the additional coupling component introduced by the probe. After the correction, the physical accuracy of the near-field data is improved, and the far-field radiation pattern generated by the subsequent conversion can more truly reflect the intrinsic radio frequency performance of the antenna.

[0077] In this way, the test efficiency can be greatly improved while maintaining the test accuracy, and the method is especially suitable for rapid radio frequency performance measurement and engineering application of multi-channel electrically adjustable base station antennas.

[0078] Referring to Figure 1 As shown in the figure, a schematic diagram of a near-field scanning fast imaging device for radio frequency performance of a multi-channel electrically adjustable base station antenna is provided, and the device comprises:

[0079] The scanning module 10 comprises a scanning component that scans the near-field region of the multi-channel electrically adjustable base station antenna, generates near-field scanning data, and transmits the near-field scanning data to the compensation module 30.

[0080] The antenna parameter module 20 is used to store and provide target parameters of the multi-channel electrically adjustable base station antenna to the compensation module 30.

[0081] The compensation module 30 comprises an electromagnetic suppression structure and a large language model, wherein the electromagnetic suppression structure comprises an electromagnetic bandgap structure or a split ring resonator, and is used to suppress the mutual coupling component and the cascaded interaction of the edge scattering component in the near-field scanning data, and input the suppression result as suppression data to the large language model.

[0082] The large language model generates a block strategy and adaptive parameters based on the target parameters, and calculates a sub-mutual coupling scattering matrix by block processing the near-field scanning data and the suppression data.

[0083] The result merging module 40 is used to splice the sub-mutual coupling scattering matrix to generate a mutual coupling scattering matrix.

[0084] The processing module 50 is used to perform matrix solving on the near-field scanning data based on the mutual coupling scattering matrix, and generate a radio frequency performance image through far-field transformation.

[0085] In a specific implementation, the scanning module 10 is internally provided with a scanning component, such as an open waveguide probe installed on a two-dimensional precision displacement platform driven by a stepping motor. In operation, the scanning component performs point-by-point scanning along a preset planar grid path in the near-field region of the multi-channel electrically adjustable base station antenna to be measured, usually several wavelengths away from the antenna surface. At each sampling point, a vector network analyzer (VNA) connected to the probe measures and records the complex transmission coefficient, such as the S21 parameter, between the probe and the antenna to be measured, and the complex data including amplitude and phase information collectively constitute the near-field scanning data. After the scanning is completed, the scanning module 10 transmits the data set to the compensation module 30.

[0086] The antenna parameter module 20 in the device, which can be a solid state drive or other non-volatile digital storage unit, provides the prior information required for the compensation module 30 to calculate. The module stores the key target parameters of the multi-channel electrically tunable base station antenna to be tested, which are the physical basis for the subsequent adaptive adjustment of the compensation algorithm.

[0087] The compensation module 30 is composed of an electromagnetic suppression structure and a large language model. In this embodiment, the electromagnetic suppression structure is a physical device, which can specifically adopt an electromagnetic band gap (EBG) structure of etching periodic metal patches on the antenna substrate to be tested, or a split ring resonator (SRR) array arranged between the antenna elements. The structure suppresses the surface wave coupling between the antenna elements through its frequency selection characteristics, thereby weakening the mutual coupling components in the near-field scanning data, especially the complex cascade interaction of the scattering components generated by the antenna edge structure. The measurement signals processed by the structure are collected to form suppression data, which are input to the large language model together with the original near-field scanning data.

[0088] The large language model in the compensation module 30 can be a pre-trained model based on the Transformer architecture, configured to perform a planning rather than a direct solving task. It receives the target parameters from the antenna parameter module 20 and generates a set of blocking strategies and adaptive parameters for guiding the subsequent numerical calculation based on these parameters. The blocking strategy specifically represents a set of coordinate boundaries that define the division method of the near-field scanning data matrix, while the adaptive parameters can be a series of numerical calculation recommended values for each sub-region, such as iteration convergence threshold or regularization coefficient. Subsequently, the numerical calculation unit in the compensation module 30, such as a solver based on the method of moments (MoM), divides the input near-field scanning data and suppression data according to the strategy generated by the large language model, and independently calculates each block data to obtain multiple sub-mutual coupling scattering matrices.

[0089] The result merging module 40 receives each sub-mutual coupling scattering matrix calculated in the foregoing steps. The module performs a matrix splicing operation, specifically, according to the spatial position information provided by the blocking strategy, assembles each sub-mutual coupling scattering matrix as a block element into a large sparse matrix with a size corresponding to the complete near-field scanning data, thereby generating a complete mutual coupling scattering matrix.

[0090] The processing module 50 performs the final compensation and imaging. This module receives the complete mutual coupling scattering matrix from the result merging module 40 and the original near-field scanning data from the scanning module 10. The processing module 50 performs a matrix solving operation on the near-field scanning data based on the mutual coupling scattering matrix. Specifically, this operation is to solve a large linear equation set to mathematically eliminate the influence of the mutual coupling effect on the original data. After obtaining the compensated near-field distribution data that is closer to the ideal state, the processing module 50 applies a far-field transformation to it, for example, by performing a two-dimensional fast Fourier transform (2D-FFT), to finally calculate the radio frequency performance image representing the antenna performance, such as a two-dimensional or three-dimensional far-field radiation pattern.

[0091] In this way, through the integration of the electromagnetic suppression structure and the large language model, the compensation of the mutual coupling and edge scattering combined effect is realized. Specifically, the device uses the antenna parameter module 20 to provide related parameters, generates a blocking strategy and adaptive parameters by the large language model, performs blocking processing on the near-field scanning data, calculates a sub-mutual coupling scattering matrix, and generates a complete mutual coupling scattering matrix through the result merging module 40 for matrix solving and far-field transformation by the processing module 50. This scheme improves the efficiency and accuracy of imaging processing and is suitable for large-scale radio frequency data scenarios; at the same time, through the adaptive blocking mechanism, the computational overhead is reduced, which adapts to the problem of wavelength shortening and scattering non-uniformity in the millimeter wave frequency band; in addition, the introduction of the electromagnetic suppression structure reduces the parasitic field interference and improves the accuracy of the radiation pattern; overall, the device enhances the real-time performance of batch testing and is suitable for optimization diagnosis of 5G / 6G antenna arrays.

[0092] As an optional implementation, refer to Figure 2 A flowchart of a method for calculating a sub-mutual coupling scattering matrix provided by the embodiment of the present application, comprising steps S101-S105, wherein:

[0093] S101: convert the target parameters into serialized prompts and input them into the large language model;

[0094] S102: the large language model generates a blocking strategy based on the serialized prompts, wherein the blocking strategy includes sub-region division instructions for the near-field scanning data;

[0095] S103: the large language model generates adaptive parameters based on the serialized prompts, wherein the adaptive parameters include prompts for adjusting the calculation;

[0096] S104: based on the blocking strategy, block the near-field scanning data and the suppression data to generate blocked data;

[0097] S105: based on the adaptive parameters, process the blocked data to calculate the sub-mutual coupling scattering matrix.

[0098] To address the technical problem of how to effectively translate high-level physical parameters into inputs understandable by large language models, and ensure that the computational strategies of their outputs can be accurately executed by downstream numerical computing units, the present application further optimizes the process of calculating the sub-intercoupling scattering matrix in the compensation module 30 described above.

[0099] In specific implementation, the process begins with a structured data conversion step, which converts the target parameters provided by the antenna parameter module 20 into a serialized prompt and provides it as input to the large language model. This step ensures that numerical and conceptual values of the physical world, such as the geometric dimensions of the antenna, material properties, etc., can be accurately translated into text format that can be processed by the large language model.

[0100] For example, a set of target parameters such as {"antenna size": [0.5, 0.5], "frequency": 28e9} can be formatted into a JSON (JavaScript Object Notation) text string. This string is then embedded in a more extensive text template containing task instructions to form the final, complete serialized prompt, which is then input to the large language model.

[0101] Upon receiving the serialized prompt, the large language model generates two types of structured outputs based on its internal representations and reasoning capabilities. The first type of output is the blocking strategy, which contains specific sub-region division instructions. These instructions are not vague suggestions, but explicit, machine-readable commands.

[0102] The second type of output is adaptive parameters, which include specific numerical or configuration suggestions for adjusting the calculation of each sub-region.

[0103] For example, for the above-mentioned sub-region, its corresponding adaptive parameters may be {"regularization coefficient": 0.01, "convergence threshold": 1e-6}, while for another sub-region that may contain more complex edge effects, its parameters may be {"regularization coefficient": 0.05, "convergence threshold": 1e-7}.

[0104] Next, the numerical computing unit within the compensation module 30, such as the aforementioned MoM solver, begins to execute the strategies generated by the large language model. First, the unit parses the sub-region division instructions in the blocking strategy and, based on these precise coordinate indices, performs matrix slicing operations on the complete near-field scanning data matrix and the suppression data matrix, thereby generating sub-blocked data corresponding to each sub-region.

[0105] Finally, for each piece of chunked data, the numerical computing unit processes it based on the adaptive parameters generated by the large language model. For example, when calculating the sub-intercoupling scattering matrix of the first sub-region, the computing unit reads the corresponding adaptive parameters and sets the regularization coefficient inside the solver to 0.01 and the convergence threshold to 1e-6. By performing this adaptive parameter-based, customized processing flow on all chunked data, the final sub-intercoupling scattering matrices are calculated.

[0106] In this way, by serializing physical parameters into structured prompts, a clear and unambiguous communication protocol is established, ensuring that the large language model can accurately understand the physical context of the computing task. Secondly, the strategy generated by the large language model, which includes specific instructions and parameters, allows its high-level planning to be accurately and repeatedly executed by the downstream numerical computing unit, avoiding the uncertainty caused by ambiguous instructions. This complete process from parameters to prompts to specific strategies enables the entire device to dynamically and intelligently adjust its core computing process according to the specific circumstances of different antennas, improving computing efficiency and automation levels while ensuring accuracy.

[0107] For example, the large language model can adopt a Transformer-based encoder-decoder architecture. Specifically, the model can contain 12 encoder layers and 12 decoder layers, each layer with 8 self-attention heads, with a total parameter size of approximately 110 million. The model is pre-trained on a dataset containing general scientific literature, programming code, and professional electromagnetic papers to enable basic logical reasoning and code generation capabilities.

[0108] Further, to enable the large language model to perform this task, it can be fine-tuned in a specific field. The training dataset used for fine-tuning consists of samples, each of which contains a set of antenna target parameters and the corresponding optimal chunking strategy and adaptive parameters designed by electromagnetic simulation experts in advance. Among them, simulation data can be generated by commercial electromagnetic simulation software (such as CST Studio Suite or ANSYS HFSS), covering near-field distribution data of multi-channel electrically adjustable base station antenna arrays of different sizes, frequencies, and substrate materials. The training process uses a supervised learning method to minimize the difference between the strategy generated by the large language model and the expert strategy, for example, a cross-entropy loss function can be used. The training uses the Adam optimizer with an initial learning rate of 0.001 and iterates for a total of 100 cycles.

[0109] For example, during runtime, the serialized prompts input to the large language model have been specially engineered. For example, a complete prompt can include the following text structure:

[0110] Task: Generate a computational strategy for antenna near-field compensation.

[0111] Input parameters: {"antenna size_m": [0.1, 0.1], "frequency_GHz": 28, "dielectric constant": 3.5, "number of scan points":}. Please output the JSON-formatted partitioning strategy and adaptive parameters. The large language model understands the task description and parameters in this prompt through its attention mechanism and generates structured JSON output, such as: {"partitioning strategy": {"type": "coordinate indexing", "boundaries": [,,...]}, "adaptive parameters": [{"region index": 0, "regularization coefficient": 0.01}, {"region index": 1, "regularization coefficient": 0.05},...]}. This mechanism ensures the accuracy and machine readability of input and output.

[0112] It can be understood that the principle of the present application for generating a computational strategy using a large language model is that the mutual coupling and edge scattering effects in the antenna near-field are a highly complex, nonlinear physical process. Traditional fixed algorithms are difficult to adapt to subtle changes in different antenna configurations. However, a large language model fine-tuned on domain data can learn the deep, nonlinear mapping relationship between antenna physical parameters such as frequency, dielectric constant, and optimal computational strategy from a large number of samples. The optimal computational strategy may, for example, be where to perform finer partitioning to capture edge effects. This is similar to the model learning certain implicit rules of electromagnetic wave propagation and scattering, thereby being able to generate a more optimal, customized numerical solution for specific problems than general fixed algorithms, which is essentially a data-driven, heuristic algorithm selection and parameter optimization process.

[0113] As an optional implementation, the large language model can choose a base model with strong general reasoning and code generation capabilities. These models are usually based on the Transformer architecture and have obtained the ability to understand complex instructions and generate structured output by pre-training on massive text and code data.

[0114] For example, in different embodiments of the present application, the following models can be used as base models, including but not limited to: GPT-4 or its subsequent versions, Gemini 1.5 Pro, Llama series models.

[0115] As an optional implementation, see Figure 3 A flowchart of a method for inputting target parameters into a large language model according to an embodiment of the present application is provided, including steps S201-S203, wherein:

[0116] S201: Obtain the target parameters from the antenna parameter module 20, wherein the target parameters include the scan data size of the multi-channel electrically adjustable base station antenna, the compensation sub-block capacity, the electromagnetic frequency range, and the dielectric constant gradient;

[0117] S202: Convert the target parameters into a serialized prompt, wherein the serialized prompt includes a numerical sequence of the scan data size, a limit sequence of the compensation sub-block capacity, and a physical model description of the electromagnetic frequency range and the dielectric constant gradient.

[0118] S203: Input the serialized prompt into the large language model, wherein the large language model generates adaptive parameters containing electromagnetic constraints based on the physical model description.

[0119] The technical problem solved by the embodiment is how to convert discrete and purely numerical engineering parameters into an input format that can be understood by a large language model (LLM) and its underlying physical meaning and inherent constraints. Simply inputting a string of numbers into the LLM may not understand whether the numbers represent antenna size or frequency, and may not understand the physical relationship between them, resulting in a lack of physical authenticity in the generated strategy.

[0120] In specific implementation, a set of explicit target parameters are obtained from the antenna parameter module 20. In this embodiment, the set of target parameters is defined to include at least: the scan data size of the multi-channel electrically adjustable base station antenna, such as a 128x128 matrix, representing the number and layout of near-field scan sampling points; the compensation sub-block capacity, such as 1024, representing the maximum number of matrix units that can be processed by the downstream numerical calculation unit at a time, which is a hardware or algorithm performance limitation; the electromagnetic frequency range, such as 28GHz to 30GHz; and the dielectric constant gradient, which describes the spatial variation of the dielectric constant of the antenna substrate material.

[0121] Next, the step of converting the target parameters into a serialized prompt is performed. This step is not simply a list of numbers, but maps different types of parameters into different types of serialized expressions:

[0122] For the scan data size, it is converted into a numerical sequence.

[0123] For the compensation sub-block capacity, it is converted into a limit sequence, explicitly informing the LLM that the generated blocking strategy must comply with this computational resource limitation.

[0124] For the electromagnetic frequency range and the permittivity gradient, they are converted into a richly contextualized physical model description. This is a textual description that not only gives numerical values, but also explains their physical meaning. For example, {"frequency_GHz": 28, "dielectric_description": "Substrate permittivity varies linearly from 2.2 at the center to 2.5 at the edges."}. This way of describing puts pure numbers (28, 2.2, 2.5) into a physical framework.

[0125] Finally, this constructed serialized hint containing numerical sequences, restriction sequences, and physical model descriptions is input into a large language model. Thanks to the learning of similar physical descriptions during the fine-tuning process, the large language model can analyze the hint. In particular, it can understand the physical model description "the permittivity varies linearly from the center to the edge" and infer that this means the electromagnetic wave wavelength will be shortened in the edge area of the antenna, and the field change will be more drastic. Therefore, when generating adaptive parameters, the model will naturally introduce instructions containing electromagnetic constraints, such as assigning smaller sizes or higher computational accuracy to sub-blocks in the edge area to accurately capture this uneven scattering effect.

[0126] In this way, by introducing physical model descriptions, abstract physical concepts are conveyed to the large language model, improving the physical accuracy of the model's generated strategies. By converting hardware or algorithmic restrictions into explicit restriction sequences, it ensures that the strategies generated by the LLM are executable in reality, avoiding resource overflow. It enables the LLM to generate adaptive parameters containing electromagnetic constraints based on its understanding of physical laws, making the entire compensation process more intelligent and refined.

[0127] As an optional implementation, based on the blocking strategy, the near-field scanning data and the suppression data are blocked to generate blocked data, including:

[0128] Based on the blocking strategy, an edge scattering sensitive area in the near-field scanning data is identified, wherein the blocking strategy includes sub-region division instructions that integrate boundary mutual coupling constraints;

[0129] According to the sub-region division instructions, the near-field scanning data is dynamically blocked to generate sub-scanning data, wherein the dynamic blocking integrates the wavelength-shortened scattering distribution;

[0130] Based on the blocking strategy, the suppression data is correspondingly blocked to generate sub-suppression data, wherein the sub-suppression data corresponds to the mutual coupling component edge value in the sub-scanning data;

[0131] combining the sub-scan data and the sub-suppression data to generate patch data.

[0132] The embodiment aims to solve the problem of waste of computing resources and distortion of physical effects caused by traditional uniform patching when processing large near-field scanning data. The traditional uniform patching method either cannot accurately capture the complex electromagnetic effects of the antenna edge region due to too coarse grid, or leads to a large amount of unnecessary redundant calculation in the region where the field changes gently due to too fine grid.

[0133] In a specific implementation, based on the patching strategy, the edge scattering sensitive region in the near-field scanning data is identified. In the patching strategy generated by the LLM, the core sub-region division instruction is not only a set of segmentation coordinates, but also an intelligent instruction set integrating the boundary mutual coupling constraint. For example, according to the antenna geometric size provided by the antenna parameter module 20, the instruction set will explicitly mark the regions around the entire scanning data matrix, such as the outermost 10% of rows and columns, as edge scattering sensitive regions. At the same time, the boundary mutual coupling constraint in the instruction will provide specific numerical suggestions for the boundary processing of these regions, to ensure the accuracy of subsequent sub-matrix calculation.

[0134] Secondly, the numerical calculation unit performs dynamic patching on the near-field scanning data according to the sub-region division instruction to generate sub-scan data. The dynamic patching here is a non-uniform and physically aware segmentation process, which integrates the physical phenomenon of wavelength shortening of scattering distribution. Specifically, for the aforementioned identified edge scattering sensitive regions, due to the discontinuity of the antenna dielectric substrate at the edge, which causes the change of effective permittivity, and in turn causes the wavelength shortening of surface waves and more severe field gradient, the sub-region division instruction will instruct to use finer division at this place, for example, divide these regions into multiple 8x8 pixel sub-blocks. For the non-sensitive regions in the center of the antenna, the instruction will use coarser division, such as 32x32 pixel sub-blocks. In this way, the entire near-field scanning data matrix is cut non-uniformly to obtain a set of sub-scan data with different sizes.

[0135] Further, based on the same patching strategy, the suppression data is correspondingly patched to generate sub-suppression data. The numerical calculation unit uses the same coordinate index as that used for dividing the near-field scanning data to cut the suppression data matrix obtained after processing by electromagnetic suppression structures such as EBG or SRR. Each piece of sub-suppression data generated has its content corresponding to the edge value of the mutual coupling component in the corresponding sub-scan data in physical sense, providing more pure boundary condition information preprocessed by physics for subsequent calculation of the mutual coupling scattering matrix of the sub-region.

[0136] The sub-scan data and the sub-suppression data are combined to generate final block data. For each divided sub-region, the corresponding sub-scan data and sub-suppression data are packaged into a unified data structure.

[0137] For example, for the i-th sub-region, the generated block data i is a data pair including two matrices, i.e., the sub-scan data i and the sub-suppression data i.

[0138] In this way, by identifying the edge scattering sensitive region and performing dynamic blocking, intelligent on-demand allocation of computing resources is realized, the computing efficiency is improved, and the accurate capture of physical effects on key regions such as the antenna edge is ensured. The method integrates physical constraints such as wavelength shortening, so that the blocking strategy is no longer a blind geometric segmentation, but an adaptive process based on electromagnetic principles, thereby improving the accuracy of the final imaging result. By co-blocking the original scan data and the suppression data and combining them, more complete and reliable input information is provided for subsequent sub-matrix calculation.

[0139] As an optional implementation, refer to Figure 4 A flowchart of a method for generating a mutual coupling scattering matrix provided by an embodiment of the present application includes steps S301-S304, wherein:

[0140] S301: receiving the sub-mutual coupling scattering matrix from the compensation module 30, and identifying a first boundary phase region based on the sub-mutual coupling scattering matrix, wherein the first boundary phase region includes a sequence of phase difference values between the sub-mutual coupling scattering matrices;

[0141] S302: calculating a first weight sequence according to the first boundary phase region, wherein the first weight sequence is generated based on the edge reflection distribution of the sub-mutual coupling scattering matrix;

[0142] S303: applying the first weight sequence to fuse and splice the sub-mutual coupling scattering matrix to generate a first preliminary mutual coupling scattering matrix, wherein the fusion and splicing includes a weighted superposition operation of the sub-mutual coupling scattering matrix;

[0143] S304: performing array response verification on the first preliminary mutual coupling scattering matrix, wherein the array response verification calculates a response consistency index of the first preliminary mutual coupling scattering matrix, and adjusts the first preliminary mutual coupling scattering matrix based on the response consistency index to generate the mutual coupling scattering matrix.

[0144] The embodiment aims to solve the technical problem of how to accurately fuse multiple sub-intercoupling scattering matrices into a physically continuous and self-consistent complete matrix after obtaining them. If these independently calculated sub-matrices are simply directly block spliced, non-physical phase jumps and amplitude discontinuities are extremely easy to be introduced at the boundaries of the sub-regions. These mutations at the splicing joints will be amplified as error sources in subsequent matrix solving, and ultimately lead to the generation of serious artifacts in the radio frequency performance image, distorting the real radiation characteristics. The embodiment solves this problem through a refined splicing method including weighted fusion and response verification.

[0145] For the above S301:

[0146] In specific implementation, the result merging module 40 first receives all the calculated sub-intercoupling scattering matrices from the compensation module 30. Then, the module analyzes any two spatially adjacent sub-matrices, such as sub-matrix A and sub-matrix B, on their common boundary. Specifically, the module extracts the complex values of a column (or a row) of elements at the boundary of the sub-matrix A, and the complex values of the corresponding column (or row) of elements at the adjacent boundary of the sub-matrix B, and calculates the element-by-element phase difference between them. All these calculated phase difference values on the boundary together constitute a first boundary phase region, whose data form is a sequence of phase difference values. The sequence intuitively quantifies the degree of phase discontinuity on the splicing boundary.

[0147] For the above S302:

[0148] In specific implementation, according to the identified boundary phase region, a first weight sequence for subsequent fusion operation is calculated. The calculation of the weight sequence is based on the edge reflection distribution of the sub-intercoupling scattering matrix.

[0149] Specifically, the amplitude of the non-diagonal elements in the edge region of each sub-matrix is analyzed. If the edge reflection distribution of a certain sub-matrix is strong, that is, the amplitude of the non-diagonal elements near the edge is large, it indicates that the coupling effect between this sub-region and other regions is truncated sharply at the boundary, and its boundary data should be given a lower weight in fusion to reduce its damage to the overall continuity. Conversely, the boundary data of a sub-matrix with weak edge reflection is given a higher weight. Through this method, a first weight sequence corresponding to all boundary points is generated.

[0150] For the above S303:

[0151] In specific implementation, the calculated first weight sequence is applied to fuse and splice all the sub-intercoupling scattering matrices. This process is not a simple block splicing, but a weighted superposition operation.

[0152] For example, at the boundary region of two sub-matrices, the module creates a smooth transition zone, in which the new matrix element values are determined by the weighted average of the corresponding boundary elements of the two sub-matrices, and the weights are the first weight sequence calculated in the previous step.

[0153] By performing this operation on all boundaries, the phase jump identified by the first boundary phase region can be effectively smoothed out, thereby generating a first preliminary mutual coupling scattering matrix that smoothly transitions at the boundaries.

[0154] For the above S104:

[0155] In order to ensure the global physical consistency of the entire matrix, the generated first preliminary mutual coupling scattering matrix needs to be subjected to an array response verification. This verification process is completed by calculating a response consistency index.

[0156] For example, a known, ideal excitation vector, such as a plane wave incident from the front of the antenna, can be multiplied by the preliminary mutual coupling scattering matrix to obtain a theoretical array response. Then, it is analyzed whether there are high-frequency spatial oscillations or abnormal values in the response vector that do not conform to physical laws. The response consistency index is a quantitative measure of these abnormalities. If the index exceeds a pre-set threshold, the module will adjust the first preliminary mutual coupling scattering matrix, for example, by applying a low-pass spatial filter or an iterative smoothing algorithm, until the response consistency index meets the requirements. The matrix obtained after this verification and adjustment is the final, high-quality mutual coupling scattering matrix.

[0157] In this way, by identifying and weighting the fusion of boundary regions, the non-physical splicing gaps introduced by independent calculation of sub-matrices are effectively eliminated, ensuring the continuity and accuracy of the final matrix. The determination of the weight based on the edge reflection distribution enables the fusion process to intelligently handle the boundary characteristics of different sub-regions, improving the physical reasonableness of the splicing. The final array response verification and adjustment steps provide a global quality guarantee for the entire splicing process, enhancing the reliability and robustness of the final generated mutual coupling scattering matrix, and laying a foundation for subsequent high-precision radio frequency performance imaging.

[0158] As an optional implementation, generating a radio frequency performance image by matrix solving of the near-field scanning data based on the mutual coupling scattering matrix includes:

[0159] Receiving the mutual coupling scattering matrix from the result merging module 40 and receiving the near-field scanning data from the scanning module 10;

[0160] Preliminary matrix solving of the near-field scanning data based on the mutual coupling scattering matrix, generating a first solving matrix, wherein the preliminary matrix solving includes an inverse operation application of the mutual coupling scattering matrix;

[0161] calculating a first distortion index according to the first solution matrix, wherein the first distortion index is generated based on the response consistency index and includes a multiple scattering bias value of the first solution matrix;

[0162] adjusting the first solution matrix by applying the first distortion index to generate a second solution matrix, wherein the adjustment includes an adaptive regularization operation of the first solution matrix;

[0163] performing a far-field transform on the second solution matrix to generate the radio frequency performance image, wherein the far-field transform includes a Fourier integral operation of the second solution matrix.

[0164] The embodiment aims to solve the technical problem of how to overcome the inherent instability of numerical calculation and suppress residual errors when compensation solution is performed by using the finally generated mutual coupling scattering matrix. In theory, the mutual coupling effect can be eliminated by a simple matrix inverse operation. However, in actual application, due to the existence of measurement noise and the possible introduction of small residual errors in the aforementioned splicing and fusion process, direct matrix inversion is often an ill-posed problem. This ill-posed problem is extremely sensitive to noise and errors, and often contains severely amplified, non-physical artifacts in the solution result, so that a clear and accurate radio frequency performance image cannot be obtained.

[0165] In a specific implementation, the processing module 50 first calls the complete mutual coupling scattering matrix that has been fused and verified from the result merging module 40, and calls the original, uncompensated near-field scanning data from the scanning module 10.

[0166] The processing module 50 first performs a preliminary matrix solution. This step is a standard inverse operation application, such as solving the linear equation group by conjugate gradient method or direct inversion, where C is the mutual coupling scattering matrix and v is the near-field scanning data vector. Among them, “ ” represents the multiplication operation between the matrix and the vector, that is, the matrix C is multiplied by the vector x to obtain the vector v, and the solved x is the preliminary estimate of the ideal uncoupled near-field distribution, denoted as the first solution matrix.

[0167] After obtaining the preliminary solution, the module does not directly use it for imaging, but rather performs a quality assessment. The module calculates a first distortion indicator based on the first solution matrix. The calculation of this indicator is composite: on the one hand, it is based on the aforementioned response consistency indicator, i.e. it calculates the response that the first solution matrix should produce on the antenna array with the first solution matrix as the new excitation source, and assesses the physical consistency of this response; on the other hand, it also specifically calculates a multiple scattering bias value for the first solution matrix. This bias value is obtained by performing a two-dimensional spectral analysis of the first solution matrix, with the aim of identifying and quantifying those high-frequency spatial ripples or non-physical oscillations that are specific to the high-order multiple scattering effects between the probe and the antenna under test, which have not been fully compensated.

[0168] Further, the processing module 50 applies the calculated first distortion indicator to fine-tune the first solution matrix. This tuning is an adaptive regularization operation, such as Tikhonov regularization. Unlike traditional methods that use a fixed regularization parameter, the adaptation is embodied in the fact that the strength of the regularization is spatially modulated by the first distortion indicator.

[0169] Specifically, in regions where the first distortion indicator shows a large distortion, e.g. regions where a high multiple scattering bias value is identified, the module applies a stronger regularization effect to suppress artifacts; whereas in regions where the distortion is small, a weaker regularization is applied to preserve the true details to the maximum extent. After this adaptive tuning, a second solution matrix is generated, which is effectively free of noise and artifacts, and in which the physical details are more realistic.

[0170] Finally, the processing module 50 inputs this optimally tuned second solution matrix, which represents the final, high-fidelity compensated near-field distribution, into a far-field transformation. This transformation is mathematically a Fourier integral operation, which is usually implemented in practice through an efficient two-dimensional fast Fourier transform (2D-FFT) algorithm. The result of the transformation is the final, clear and accurate radio frequency performance image, such as the far-field radiation pattern.

[0171] As an optional embodiment, it is also necessary to address the problem of how to eliminate the distortion of the far-field image converted from the residual errors of the near field in the final far-field transformation and imaging stage. Although the aforementioned steps have performed fine compensation and adjustment on the near-field data, some weak, highly spatially correlated residual errors, e.g. errors caused by probe polarization leakage or high-order scattering effects that have not been fully modeled, can still exist. When the far-field transformation is performed through the Fourier integral operation, these small errors of the near field are propagated and can be amplified in specific regions of the far field, manifesting as non-physical artifacts on the radiation pattern, e.g. unrealistic cross-polar lobes or side lobe elevations, thereby affecting the accurate assessment of the true performance of the antenna.

[0172] In this embodiment, it specifically comprises: receiving the second solution matrix from the processing module 50, and calculating a first far-field integral sequence based on the second solution matrix.

[0173] In a specific implementation, the processing module 50 first receives the second solution matrix with high fidelity after adaptive regularization adjustment from the previous step. Then, the module performs a standard fast Fourier integral operation on the matrix, for example, by two-dimensional fast Fourier transform, and the operation result is the preliminary calculation of the antenna far-field radiation characteristics, forming the first far-field integral sequence. The sequence is a complex matrix, representing the far-field amplitude and phase at different azimuth and elevation angles.

[0174] Further, a first radiation pattern region is identified according to the first far-field integral sequence.

[0175] In a specific implementation, the preliminary far-field result is analyzed to identify the region where distortion may exist. Specifically, the module analyzes the polarization purity of each point in the first far-field integral sequence according to the physical characteristics of the antenna design itself, such as its expected main polarization direction. For an antenna designed for vertical polarization, the horizontal polarization component in each far-field direction is calculated. Those regions with abnormally high horizontal polarization components are identified as the first radiation pattern region. The abnormal cross-polarization values of all points in this region together form a polarization deviation value sequence.

[0176] Further, a first compensation sequence is generated based on the first radiation pattern region.

[0177] After identifying the region with polarization deviation, a first compensation sequence for correction is generated. The generation of this sequence is based on a physical assumption: the significant polarization deviation in the far field is often caused by the multiple scattering components in the near field that are not fully compensated. Using a pre-set physical model or lookup table, the polarization deviation values in the far field are inversely mapped to a set of multiple scattering correction values. These correction values aim to offset the impact caused by residual multiple scattering propagating to the far field.

[0178] Further, the first compensation sequence is applied to adjust the first far-field integral sequence to generate a second far-field integral sequence.

[0179] In a specific implementation, the generated first compensation sequence is applied to adjust the preliminary far-field result. The adjustment is a weighted correction operation.

[0180] Specifically, the correction values are modulated by a spatial weighting function that takes its maximum value at the center of the first radiation pattern region and smoothly decays outward. This ensures that the correction operation is concentrated in the region where the problem is most severe, while avoiding introducing new discontinuities at the region boundaries. After this weighted correction, the polarization bias is suppressed, resulting in a second far-field integral sequence that is purer and more physically realistic.

[0181] Further, the second far-field integral sequence is mapped to generate the radio frequency performance image.

[0182] In a specific implementation, the second far-field integral sequence after the final correction is mapped to generate the radio frequency performance image. This step includes: first, calculating the modulus of each complex value in the sequence and converting it to decibel (dB) units; second, normalizing the result; and finally, performing a two-dimensional projection operation to project the three-dimensional spherical radiation data onto a two-dimensional plane to generate a radio frequency performance image that can be directly interpreted by the user, such as a two-dimensional cross-section of the E-plane and H-plane, or a pseudo-color image of the entire visible hemisphere.

[0183] As an optional implementation, converting the target parameters into a serialized prompt includes:

[0184] Receiving the target parameters from the antenna parameter module 20 and extracting the numerical data of the scan data size, the limit data of the compensation sub-block capacity, the range data of the electromagnetic frequency range, and the gradient data of the permittivity gradient;

[0185] Serializing the numerical data of the scan data size to generate a first numerical sequence, wherein the first numerical sequence includes a dimension array of the scan data size;

[0186] Serializing the limit data of the compensation sub-block capacity to generate a first limit sequence, wherein the first limit sequence includes a boundary threshold array of the compensation sub-block capacity;

[0187] Serializing the range data of the electromagnetic frequency range and the gradient data of the permittivity gradient to generate a first physical model sequence, wherein the first physical model sequence includes a frequency spectrum vector of the electromagnetic frequency range and a gradient vector of the permittivity gradient;

[0188] Combining the first numerical sequence, the first limit sequence, and the first physical model sequence to generate the serialized prompt, wherein the serialized prompt includes an integrated parameter sequence array.

[0189] This embodiment aims to solve the technical problem of how to ensure the standardization, repeatability and machine readability of the conversion process in the process of converting target parameters into serialized prompts. The aforementioned embodiment proposes the concept of converting parameters into numerical sequences, limit sequences and physical model descriptions, but if there is a lack of a systematic conversion and organization process, the generated prompt format may not be consistent, leading to unstable responses of LLM.

[0190] In specific implementation, the device receives target parameters from the antenna parameter module 20 and extracts and classifies these parameters. Specifically, the module extracts from the target parameters: numerical data representing the scan data size of the scan point array size; limit data representing the compensation sub-block capacity of the computing unit processing capability; range data representing the electromagnetic frequency range of the operating frequency point; and gradient data representing the permittivity gradient of the substrate material characteristics.

[0191] Secondly, the numerical data of the scan data size is serialized and converted to generate a first numerical sequence. This conversion converts the human-readable size descriptor into a machine-readable dimension array.

[0192] Thirdly, the limit data of the compensation sub-block capacity is serialized and converted to generate a first limit sequence. This conversion converts a single limit value (1024) into an array of boundary threshold values containing the value.

[0193] Subsequently, the range data of the electromagnetic frequency range and the gradient data of the permittivity gradient are physically modeled and serialized to generate a first physical model sequence. This step converts physical parameters into vectorized expressions:

[0194] The electromagnetic frequency range (28 GHz) is converted into a frequency spectrum vector, for example [28e9].

[0195] The permittivity gradient (linear change from 2.2 to 2.5) is converted into a gradient vector, for example ["linear", 2.2, 2.5], where the first element defines the gradient type and the subsequent elements are key parameters.

[0196] These two vectors together constitute the first physical model sequence.

[0197] Finally, the first numerical sequence, the first limit sequence and the first physical model sequence generated above are combined to generate the final serialized prompt. This combination process integrates all serialized arrays and vectors into a unified, structured data object to form an integrated parameter sequence array.

[0198] For example, the final generated serialized prompt can be a JSON object with the following format:

[0199] {"task": "generate_strategy", "params": {"scan_dims": , "block_capacity": , "physics": {"freq_vector": [28e9], "diel_vector": ["linear",2.2, 2.5]}}}.

[0200] This lowers the barrier for LLM to understand complex physical and engineering constraints, thereby ensuring that it can generate high-quality computational strategies more stably and accurately.

[0201] As an optional implementation, the step of dynamically dividing the near-field scanning data into blocks according to the sub-region division instruction to generate sub-scan data includes:

[0202] The sub-region partitioning instruction is received from the block partitioning strategy, and a first scattering vector is calculated based on the sub-region partitioning instruction, wherein the first scattering vector includes the wavelength distribution sequence of the near-field scanning data;

[0203] A first block boundary sequence is generated based on the first scattering vector, wherein the first block boundary sequence includes an array of mutual coupling constraint values ​​of the sub-region partitioning instruction;

[0204] The first block boundary sequence is applied to the near-field scan data to perform block operations, generating a first sub-scan block sequence, wherein the block operations include the region segmentation operation of the near-field scan data;

[0205] The first sub-scan block sequence is integrated to generate the sub-scan data, wherein the integration includes the sequential arrangement operation of the first sub-scan block sequence.

[0206] This embodiment aims to address the algorithm implementation problem: how the numerical computation unit can accurately translate the abstract partitioning instructions, which include physical constraints, provided by the Large Language Model (LLM), into specific segmentation operations on the near-field scan data matrix. Without a clear computational path and only a human-defined concept, the accuracy and repeatability of dynamic segmentation are difficult to guarantee. This embodiment transforms the high-level strategies of LLM into precise and executable segmentation boundaries through a step-by-step, data-driven computational process.

[0207] In a specific implementation, the numerical computing unit receives sub-region division instructions from the patching strategy generated by the LLM. Then, it calculates the first scattering vector based on these instructions. This step is to convert the physical insight of the LLM into quantitative indicators. For example, the sub-region division instructions may contain a hint such as "pay attention to the edge wavelength shortening effect caused by the dielectric constant gradient". Based on this, the numerical computing unit performs a two-dimensional phase gradient analysis on the entire near-field scanning data, calculates the local spatial frequency of each sampling point position, and calculates the effective wavelength distribution sequence of the point from this. This sequence, i.e. the first scattering vector, quantifies the actual propagation characteristics of electromagnetic waves at various locations on the antenna surface.

[0208] Secondly, according to the first scattering vector, a first patch boundary sequence is generated. This step is to use the quantitative physical indicators of the previous step to determine the best segmentation line. The numerical computing unit extracts an intercoupling constraint value array from the sub-region division instructions, for example, the array may specify that "the intercoupling truncation error at any sub-block boundary should be less than -40dB". Then, the unit takes the first scattering vector (wavelength distribution sequence) as input, and uses optimization algorithms such as dynamic programming to find a set of segmentation boundaries that can meet the intercoupling constraints and preferentially be placed in areas with gentle wavelength changes to avoid cutting areas with steep field gradients. This set of optimized, non-uniformly distributed boundary coordinates, i.e. the first patch boundary sequence.

[0209] Thirdly, the first patch boundary sequence is applied to the near-field scanning data for patching operation. This operation is an accurate region segmentation operation. The numerical computing unit takes the first patch boundary sequence as the cutting instruction, and performs matrix slicing on the complete near-field scanning data matrix to generate a set of matrix blocks with different sizes and clear physical meanings, i.e. the first sub-scanning block sequence.

[0210] Finally, the first sub-scanning block sequence is integrated to generate the final sub-scanning data. The integration here is a sequential arrangement operation, for example, all cut sub-scanning blocks are stored in a list or structured array according to their spatial position in the original matrix. This integrated data structure is the final sub-scanning data that can be used for downstream parallel processing.

[0211] In this way, the implementability and reproducibility of the entire technical solution are enhanced, and the patching decision is directly driven by the physical characteristics of the measurement data itself, which is more accurate and adaptive than simply relying on geometric position division, ensuring that the segmented sub-problems are in good physical condition, laying a foundation for subsequent high-precision calculation of the sub-intercoupling matrix.

[0212] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit the present application; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A near-field scanning fast imaging device for radio frequency performance of a multi-channel electrically-tunable base station antenna, characterized in that, The application relates to a method for measuring the radio frequency performance of a multi-channel electrically tunable base station antenna. The application comprises: a scanning assembly, which comprises: a radio frequency probe mounted on a two-dimensional precision displacement platform; and a vector network analyzer connected to the radio frequency probe; wherein the two-dimensional precision displacement platform is used to drive the radio frequency probe to scan along a preset planar grid path in the near-field region of a multi-channel electrically tunable base station antenna, and the vector network analyzer measures and collects scattering parameters representing the electromagnetic field characteristics of each sampling point on the grid path to form a near-field scanning data set; a data processing module connected to the scanning assembly, which is configured to: receive the near-field scanning data set and perform a fast Fourier transform operation on the near-field scanning data set to complete the near-field to far-field conversion and generate a far-field radiation pattern representing the radio frequency performance of the antenna; The application further comprises: a scanning module comprising a scanning assembly, which scans the near-field region of a multi-channel electrically tunable base station antenna to generate near-field scanning data and transmits the near-field scanning data to a compensation module; an antenna parameter module for storing and providing target parameters of the multi-channel electrically tunable base station antenna to the compensation module; a compensation module comprising an electromagnetic suppression structure and a large language model, wherein the electromagnetic suppression structure comprises an electromagnetic bandgap structure or a split ring resonator, which is used to suppress the mutual coupling component in the near-field scanning data and the cascaded interaction of the mutual coupling component with the edge scattering component, and input the suppression result as suppression data into the large language model; The large language model generates a block strategy and adaptive parameters based on the target parameters, and calculates a sub-mutual coupling scattering matrix by performing block processing on the near-field scanning data and the suppression data; a result merging module for splicing the sub-mutual coupling scattering matrix to generate a mutual coupling scattering matrix; 2. The near-field scanning quick imaging device for radio frequency performance of multi-channel electrically adjustable base station antenna according to claim 1, characterized in that, a processing module for performing matrix solving on the near-field scanning data based on the mutual coupling scattering matrix to generate a radio frequency performance image through far-field conversion.

3. The near-field scanning quick imaging device for radio frequency performance of multi-channel electrically adjustable base station antenna according to claim 1, characterized in that, The data processing module is further configured to apply a preset static compensation matrix to correct the near-field scanning data set before performing the near-field to far-field conversion operation, so as to suppress the mutual coupling effect between the probe and the antenna under test. The calculation of the sub-mutual coupling scattering matrix comprises: converting the target parameters into serialized prompts and inputting the serialized prompts into the large language model; the large language model generates a block strategy based on the serialized prompts, wherein the block strategy comprises sub-region division instructions for the near-field scanning data; the large language model generates adaptive parameters based on the serialized prompts, wherein the adaptive parameters comprise prompts for adjusting the calculation; the near-field scanning data and the suppression data are blocked based on the block strategy to generate block data; 4. The near-field scanning quick imaging device for radio frequency performance of a multi-channel electrically-steerable base station antenna according to claim 3, wherein, the block data is processed based on the adaptive parameters to calculate the sub-mutual coupling scattering matrix. Converting the target parameters into serialized prompts and inputting the serialized prompts into the large language model comprises: obtaining the target parameters from the antenna parameter module, wherein the target parameters comprise the scanning data size, compensation sub-block capacity, electromagnetic frequency range and dielectric constant gradient of the multi-channel electrically tunable base station antenna; convert the target parameter into a serialization hint, wherein the serialization hint comprises a numerical sequence of the scan data size, a limit sequence of the compensation sub-block capacity, and a physical model description of the electromagnetic frequency range and dielectric constant gradient; input the serialization hint into the large language model, wherein the large language model generates adaptive parameters containing electromagnetic constraints based on the physical model description.

5. The near-field scanning quick imaging device for radio frequency performance of a multi-channel electrically-steerable base station antenna according to claim 4, wherein, based on the blocking strategy, the near-field scan data and the suppression data are blocked to generate blocked data, which comprises: based on the blocking strategy, an edge scattering sensitive area in the near-field scan data is identified, wherein the blocking strategy comprises sub-region division instructions that integrate boundary mutual coupling constraints; according to the sub-region division instructions, the near-field scan data is dynamically blocked to generate sub-scan data, wherein the dynamic blocking integrates the scattering distribution of wavelength shortening; based on the blocking strategy, the suppression data is correspondingly blocked to generate sub-suppression data, wherein the sub-suppression data corresponds to the mutual coupling component edge value in the sub-scan data; the sub-scan data and the sub-suppression data are combined to generate blocked data.

6. The near-field scanning quick imaging device for radio frequency performance of a multi-channel electrically-steerable base station antenna according to claim 1, wherein, splicing the sub-mutual coupling scattering matrix to generate a mutual coupling scattering matrix, which comprises: receiving the sub-mutual coupling scattering matrix from the compensation module and identifying a first boundary phase region based on the sub-mutual coupling scattering matrix, wherein the first boundary phase region comprises a phase difference value sequence between the sub-mutual coupling scattering matrices; calculating a first weight sequence according to the first boundary phase region, wherein the first weight sequence is generated based on the edge reflection distribution of the sub-mutual coupling scattering matrix; applying the first weight sequence to fuse and splice the sub-mutual coupling scattering matrix to generate a first preliminary mutual coupling scattering matrix, wherein the fusion and splicing includes a weighted superposition operation of the sub-mutual coupling scattering matrix; verifying the array response of the first preliminary mutual coupling scattering matrix, wherein the array response verification calculates a response consistency index of the first preliminary mutual coupling scattering matrix, and adjusts the first preliminary mutual coupling scattering matrix based on the response consistency index to generate the mutual coupling scattering matrix.

7. The near-field scanning quick imaging device for radio frequency performance of a multi-channel electrically-steerable base station antenna according to claim 6, wherein, based on the mutual coupling scattering matrix, the near-field scan data is matrix solved to generate a radio frequency performance image through far-field transformation, which comprises: receiving the mutual coupling scattering matrix from the result merging module and receiving the near-field scan data from the scanning module; based on the mutual coupling scattering matrix, the near-field scan data is preliminarily matrix solved to generate a first solving matrix, wherein the preliminary matrix solving includes the application of inverse operation of the mutual coupling scattering matrix; calculating a first distortion index according to the first solving matrix, wherein the first distortion index is generated based on the response consistency index and includes multiple scattering deviation values of the first solving matrix; applying the first distortion index to adjust the first solving matrix to generate a second solving matrix, wherein the adjustment includes adaptive regularization operation of the first solving matrix; performing a far-field transform on the second solution matrix to generate the radio frequency performance image, wherein the far-field transform comprises a Fourier integral operation of the second solution matrix.

8. The near-field scanning quick imaging device for radio frequency performance of a multi-channel electrically-steerable base station antenna according to claim 7, wherein, performing a far-field transform on the second solution matrix to generate the radio frequency performance image comprises: receiving the second solution matrix from the processing module and calculating a first far-field integral sequence based on the second solution matrix, wherein the first far-field integral sequence comprises a Fourier integral operation result of the second solution matrix; identifying a first radiation pattern region according to the first far-field integral sequence, wherein the first radiation pattern region comprises a polarization deviation value sequence in the first far-field integral sequence; generating a first compensation sequence based on the first radiation pattern region, wherein the first compensation sequence comprises a multiple scattering correction value of the polarization deviation value sequence; adjusting the first far-field integral sequence by applying the first compensation sequence to generate a second far-field integral sequence, wherein the adjusting comprises a weighted correction operation of the first far-field integral sequence; performing an imaging mapping on the second far-field integral sequence to generate the radio frequency performance image, wherein the imaging mapping comprises a two-dimensional projection operation of the second far-field integral sequence.

9. The near-field scanning quick imaging device for radio frequency performance of multi-channel electrically adjustable base station antenna according to claim 4, characterized in that, converting the target parameters into a serialized prompt comprises: receiving the target parameters from the antenna parameter module and extracting numerical data of the scan data size, limit data of the compensation sub-block capacity, range data of the electromagnetic frequency range, and gradient data of the permittivity gradient; serializing the numerical data of the scan data size to generate a first numerical sequence, wherein the first numerical sequence comprises a dimension array of the scan data size; serializing the limit data of the compensation sub-block capacity to generate a first limit sequence, wherein the first limit sequence comprises a boundary threshold array of the compensation sub-block capacity; serializing the range data of the electromagnetic frequency range and the gradient data of the permittivity gradient according to a physical model to generate a first physical model sequence, wherein the first physical model sequence comprises a frequency spectrum vector of the electromagnetic frequency range and a gradient vector of the permittivity gradient; combining the first numerical sequence, the first limit sequence, and the first physical model sequence to generate the serialized prompt, wherein the serialized prompt comprises an integrated parameter sequence array.

Images