Adaptive vector network scanning method, vector network analyzer, medium and product

Abstract

The application discloses a kind of self-adapting vector network scanning method, vector network analyzer, medium and product, it is related to the field of measurement electric variable.Method includes: through chaos mapping generation sparse frequency point sequence to preliminary detection S parameter;Subsequently, the time-domain reflection peak of data and the features such as frequency domain curvature, information entropy, wavelet transform are calculated;Based on these features, the frequency band is intelligently divided into high sensitive area, flat area, and the time-domain key area is mapped into frequency domain ripple area;Optimized frequency step and point number are distributed to each area using quantum genetic algorithm, forming scanning strategy;According to this strategy, segmented scanning is performed, and real-time verification is based on local fitting residual;If residual is over threshold, then dynamically insert supplementary frequency point and perform supplementary scanning;Integrate all data to generate complete and accurate S parameter curve.The application can relieve the contradiction between scanning speed and accuracy in the test of new energy automobile parts.

Description

Technical Field

[0001] This invention relates to the field of electrical variable measurement technology, and in particular to an adaptive vector network scanning method, a vector network analyzer, a medium, and related products. Background Technology

[0002] The Vector Network Analyzer (VNA) is an indispensable core test instrument in the fields of modern radio frequency, microwave and high-speed digital circuits. It accurately characterizes the electrical performance of the device under test (DUT) by scanning and measuring the scattering parameters (S-parameters) of the device under test (DUT) over a wide frequency range.

[0003] As a core instrument for accurately characterizing electrical properties, the VNA (Vehicle Detector) has been applied throughout the entire chain of new energy vehicle R&D and production. From the quality control of 800V high-voltage wiring harnesses, connectors, and power battery packs that are crucial to vehicle safety, to the signal integrity verification of onboard high-speed data links, millimeter-wave radar antennas, and V2X communication modules that ensure the stability of intelligent driving systems, the VNA plays an indispensable role and is a key testing tool to ensure the high performance and high reliability of new energy vehicles.

[0004] However, in large-scale industrial applications, existing VNA testing technology faces a challenging technical contradiction—the conflict between the speed and accuracy of vector network scanning. To accurately capture the key electrical characteristics of components tested in new energy vehicles, such as weak safety hazard signals caused by minor manufacturing defects in high-voltage wiring harnesses, or sharp resonance peaks in millimeter-wave radar antennas, VNAs need to scan with extremely small frequency steps. This results in excessively long test times, far exceeding the pace requirements of mass production in the automotive industry. Conversely, if large scanning steps are used to pursue testing speed, these crucial details are likely to be missed or distorted, leading to inaccurate test results and an inability to reliably assess product performance and safety.

[0005] This contradiction between speed and precision has become a technical bottleneck restricting the improvement of quality and efficiency in the new energy vehicle industry. Summary of the Invention

[0006] To address the aforementioned technical problems and deficiencies, the purpose of this invention is to provide an adaptive vector network scanning method, vector network analyzer, medium, and product that can alleviate the contradiction between vector network scanning speed and accuracy in the testing of new energy vehicle components.

[0007] To achieve the above objectives, in a first aspect, the present invention provides an adaptive vector network scanning method, comprising: generating a non-uniformly distributed sparse frequency point sequence through chaotic mapping within a set test frequency band; detecting S-parameter measurement data of a vector network of a new energy vehicle component based on the sparse frequency point sequence; calculating time-domain features and frequency-domain features based on the S-parameter measurement data, wherein the time-domain features include reflection peak features in the time-domain reflection profile, and the frequency-domain features include curvature features, information entropy features, and wavelet transform features; dividing the frequency domain into a high-sensitivity region and a flat region based on a preset frequency-domain feature mutation threshold and frequency-domain features, and determining a time-domain key region based on the reflection peak features in the time-domain features; and mapping the time-domain key region back to the frequency domain to define the frequency domain. The ripple region, or frequency domain ripple region, represents a frequency band where the S-parameters exhibit periodic, weak fluctuations over a wide bandwidth. Using a quantum genetic algorithm, optimized frequency step sizes and the number of scanning sampling points are assigned to the high-sensitivity region, the flat region, and the ripple region to obtain an optimized scanning strategy. Based on the optimized scanning strategy, segmented scanning is performed on the high-sensitivity region, the flat region, and the ripple region to obtain real-time scanning data. Local fitting residuals for each region are calculated based on the real-time scanning data. If any local fitting residual exceeds a confidence threshold, supplementary frequency points are dynamically inserted into the corresponding frequency domain interval, and supplementary scanning is performed to obtain supplementary scanning data. Based on the real-time scanning data and the supplementary scanning data, the S-parameter curves of the new energy vehicle components in the test frequency band are generated.

[0008] This invention employs the aforementioned method, utilizing an innovative multi-stage, multi-dimensional intelligent scanning strategy to achieve precise on-demand allocation of test resources (i.e., scanning frequency points). This significantly improves testing speed while ensuring accurate capture of key electrical characteristics, effectively alleviating the contradiction between speed and accuracy in vector network scanning. Specifically, this invention first utilizes chaotic mapping for preliminary sparse detection to quickly obtain an overview of the overall response of the device under test. Next, through joint time-domain and frequency-domain feature analysis of the initial data, it accurately identifies "high-sensitivity regions," "ripple regions," and "flat regions" with smooth responses, containing key information (such as sharp resonances, periodic ripples, impedance abrupt changes, etc.). Then, it uses a quantum genetic algorithm to tailor the optimal scanning step size and number of points for different regions, achieving the highest information acquisition efficiency with the fewest test points. Finally, through real-time monitoring and dynamic supplementary scanning mechanisms of local fitting residuals, it performs closed-loop verification and correction of the scanning results, ensuring that the final S-parameter curve has neither the omission of key details nor the drag of redundant data. Compared with the fixed and uniform scanning method in the prior art, the present invention changes the "carpet search" to "precision guidance", focusing the limited test time on the frequency band with the most diagnostic value, thereby effectively alleviating the technical bottleneck that large step scans are prone to losing details and small step scans are too time-consuming, and can meet the urgent needs of the new energy vehicle industry for high-efficiency and high-precision online testing.

[0009] Optionally, in some embodiments, mapping the time-domain key region back to the frequency domain to define the frequency-domain ripple region includes: extracting the characteristic time delay corresponding to each reflection peak from the time-domain key region; determining the theoretical ripple period in the frequency domain based on the characteristic time delay; performing frequency-dependent correction on the theoretical ripple period based on the dispersion characteristics of the signal transmission path to obtain the actual ripple period; and extending a symmetrical frequency band in the frequency domain based on the actual ripple period to define the frequency-domain ripple region.

[0010] The technical solution of the above embodiments, by introducing a correction to the dispersion characteristics of the signal transmission path, transforms the idealized theoretical ripple period into a more physically realistic, frequency-varying actual ripple period. This significantly improves the accuracy of defining the frequency domain ripple region, ensuring that subsequent scanning strategies can use truly effective step sizes to accurately capture weak periodic fluctuations caused by physical defects, avoiding missed detections due to inaccurate models. Optionally, in some embodiments, based on the dispersion characteristics of the signal transmission path, the theoretical ripple period is frequency-dependently corrected to obtain the actual ripple period. Specifically, this includes: obtaining the medium parameters of the vector network transmission path in new energy vehicle components from a preset material property database; constructing a phase velocity frequency dependence characteristic model of electromagnetic wave propagation in the vector network transmission path based on the medium parameters; and performing frequency-variable compensation on the theoretical ripple period according to the phase velocity frequency dependence characteristic model to obtain the actual ripple period.

[0011] The technical solution described above provides a concrete and precise engineering path for dispersion correction. It transforms abstract physical concepts into executable computational steps by calling a pre-defined material database and constructing a phase velocity frequency-dependent model. This makes the correction of the theoretical ripple period no longer an estimation, but a precise compensation based on the actual medium characteristics of the measured component, thereby greatly improving the accuracy and reliability of the actual ripple period calculation.

[0012] Optionally, in some embodiments, determining the theoretical ripple period in the frequency domain based on the characteristic time delay includes: determining the two-way propagation time delay of the electromagnetic wave in the signal transmission path based on the characteristic time delay and the reflection physical mechanism; converting the two-way propagation time delay into the period parameters of the corresponding frequency domain ripple to obtain the theoretical ripple period.

[0013] The technical solution adopted in the above embodiments clarifies the physical basis of time-domain to frequency-domain conversion, ensuring the fundamental accuracy of theoretical ripple period calculation. It first correctly interprets the measured characteristic time delay as the round-trip propagation time delay of the electromagnetic wave, and then converts it into a frequency-domain period parameter. This avoids calculation errors that may result from misusing single-trip time delay, laying a correct theoretical foundation for subsequent refined scanning strategies.

[0014] Optionally, in some embodiments, supplementary frequency points are dynamically inserted and supplementary scanning is performed within the corresponding frequency domain interval. Specifically, this includes: determining the frequency domain interval where the local fitting residual exceeds the confidence threshold as the target sub-interval where the scanning density needs to be increased; determining the insertion density of supplementary frequency points based on the number of scanned frequency points and the rate of change of fitting curvature within the target sub-interval; inserting supplementary frequency points within the target sub-interval according to the insertion density of supplementary frequency points, and performing supplementary scanning based on the currently allocated frequency step size.

[0015] The technical solution described in the above embodiments realizes an intelligent dynamic supplementary scanning mechanism. It abandons a fixed supplementary point strategy and instead adaptively determines the insertion density of supplementary frequency points based on the existing scan density within the target sub-interval and the complexity of the fitted curve. This approach ensures that the resource investment in the correction process is precisely matched to the severity of the problem, minimizing the time required for supplementary scanning while ensuring correction accuracy.

[0016] Optionally, in some embodiments, the insertion density of supplementary frequency points is determined based on the number of scanned frequency points and the rate of change of the fitted curvature within the target sub-interval, including: counting the total number of scanned frequency points within the target sub-interval; calculating the absolute value of the average fitted curvature of the scanned frequency points based on the total number; determining the density level based on the total number and the absolute value of the average fitted curvature; and determining the corresponding insertion density of supplementary frequency points based on the density level.

[0017] The technical solution described above provides a quantitative and hierarchical decision-making model for determining the insertion density of supplementary frequency points. It categorizes the need for supplementary scanning into different levels by comprehensively considering the existing sampling density of the target sub-interval and the inherent complexity of the signal. This method transforms a vague concept of "on-demand supplementation" into a clear and repeatable lookup table operation, enhancing the logic and automation of the dynamic correction strategy.

[0018] Optionally, in some embodiments, generating a non-uniformly distributed sparse frequency point sequence through chaotic mapping includes: setting initial values ​​and nonlinear control parameters for the chaotic mapping; performing multiple iterative calculations based on the nonlinear control parameters to obtain a chaotic variable sequence; and mapping the chaotic variable sequence to a test frequency band to form a non-uniformly distributed sparse frequency point sequence.

[0019] The technical solution described above provides an efficient and highly ergodic frequency point generation method for the initial sparse detection stage. By employing chaotic mapping, the generated sparse frequency point sequence is non-uniformly distributed throughout the entire test frequency band. This avoids the frequency aliasing problem that may occur with traditional periodic sampling, and can quickly cover the entire frequency band with extremely high computational efficiency, laying a higher-quality data foundation for subsequent accurate partitioning and intelligent analysis.

[0020] In a second aspect, embodiments of the present invention provide a vector network analyzer, comprising: one or more processors and a memory; the memory is coupled to the one or more processors, the memory being used to store computer program code, the computer program code including computer instructions, the one or more processors invoking the computer instructions to cause the vector network analyzer to perform the methods described in the first aspect and any possible implementation thereof.

[0021] Thirdly, the present invention provides a computer-readable storage medium including instructions that, when executed on the vector network analyzer, cause the vector network analyzer to perform the method described in the first aspect and any possible implementation thereof.

[0022] Fourthly, the present invention provides a computer program product containing instructions that, when the computer program product is run on the vector network analyzer, cause the vector network analyzer to perform the method described in the first aspect and any possible implementation thereof.

[0023] Understandably, the vector network analyzer provided in the second aspect, the storage medium provided in the third aspect, and the computer program product provided in the fourth aspect are all used to execute the method provided by this invention. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here. Attached Figure Description

[0024] Figure 1 This is a flowchart of an adaptive vector network scanning method according to an embodiment of the present invention;

[0025] Figure 2 This is an example diagram of an S-parameter curve in an embodiment of the present invention;

[0026] Figure 3 This is a magnified view of a highly sensitive region of an S-parameter curve in an embodiment of the present invention;

[0027] Figure 4 This is a schematic diagram of the hardware architecture of a vector network analyzer according to an embodiment of the present invention. Detailed Implementation

[0028] In some application scenarios of testing new energy vehicle components, such as the final inspection of an 800V high-voltage wiring harness on the production line, when the test engineer uses the existing vector network scanning technology, in order to meet the production cycle (e.g., to complete the test within tens of seconds), he may set a large fixed frequency step, such as sampling one point every 5MHz, and performing a fast scan in the frequency band from 1MHz to 1GHz.

[0029] While this method is fast, if there is a tiny impedance discontinuity inside the wire harness caused by slight pressure on the insulation layer, this defect may only produce a weak reflection signal (a spike in the S11 parameter) within a very narrow frequency band around 253MHz. Because the large 5MHz scan step skips this critical 253MHz point, the test result curve will appear smooth and normal, thus misjudging the potentially safety-hazardous wire harness as a qualified product and allowing it to enter the market. Conversely, if a fine step size of 0.1MHz is used to ensure absolute accuracy, the number of scan points increases dramatically, and a single test will take several minutes, which is completely unsuitable for the fast pace of the production line, leading to low production efficiency and soaring costs.

[0030] However, the technical solution of this invention scans the 800V high-voltage line harness in the same scenario, but the implementation method is completely different.

[0031] First, the vector network analyzer initiates chaotic mapping, instantly generating and testing approximately 100 non-uniformly distributed sparse frequency points in the 1MHz to 1GHz frequency band, obtaining a preliminary profile of the harness response in less than a second.

[0032] Next, the algorithm immediately analyzed this sparse data. On one hand, through frequency domain curvature calculation, it found an abnormally increasing trend in the rate of change of data points in the vicinity of 250MHz, initially marking it as a "highly sensitive region in the frequency domain." On the other hand, by obtaining the time-domain reflection profile through inverse fast Fourier transform (IFFT), a tiny reflection peak was clearly observed. The algorithm identified this as a "critical region in the time domain" and accurately mapped it back to the frequency domain, defining a "frequency domain ripple region" centered at 253MHz. The remaining frequency bands with gentler responses were classified as "frequency domain flat regions."

[0033] Subsequently, the quantum genetic algorithm was activated, assigning a fine step size of 0.1 MHz to the "high-sensitivity region" and the "ripple region," while allocating a sparse step size of 10 MHz to the vast "flat region." The VNA then executed this optimized segmented scanning strategy, rapidly scanning the flat region and performing fine scanning only in the critical area around 253 MHz. During the fine scanning process, real-time residual monitoring detected drastic fluctuations in data points and automatically inserted several supplementary frequency points at the sharpest peaks.

[0034] Ultimately, the entire adaptive scanning process was completed within tens of seconds. Not only did the test speed meet the production line cycle time requirements, but it also accurately and completely captured the S11 parameter spike at 253MHz caused by the minor defect, providing reliable test assurance for product quality and safety.

[0035] In this embodiment, the new energy vehicle components under test may include, but are not limited to: 800V high-voltage wiring harnesses, connectors, power battery packs to ensure driving safety, and a series of high-frequency, high-speed electrical performance sensitive components such as vehicle-mounted high-speed data links, millimeter-wave radar antennas and V2X communication modules to support intelligent driving.

[0036] The technical solution of this invention is applied to a vector network analyzer (VNA). This vector network analyzer firstly possesses the core RF / microwave transceiver link of a traditional VNA, including at least a broadband, low-noise signal source, a high-performance directional coupler or bridge for separating incident and reflected signals, and a multi-channel coherent receiver for accurately measuring amplitude and phase.

[0037] Simultaneously, the VNA is also equipped with a high-performance computing unit, such as a CPU or FPGA (Field-Programmable Gate Array) combined with a DSP (Digital Signal Processor). This computing unit is not only responsible for conventional instrument control and data acquisition, but also specifically designed for high-speed execution of complex algorithms in this embodiment, including chaotic sequence generation, real-time calculation of multi-dimensional features (time-domain reflection, frequency-domain curvature, information entropy, wavelet transform), and iterative optimization of quantum genetic algorithms.

[0038] To match this, the signal source used is an agile frequency synthesizer with extremely fast frequency switching speed and extremely short settling time. This design enables it to instantaneously jump to any specified frequency within nanoseconds to microseconds and complete stable measurement based on non-uniformly spaced, dynamically generated frequency point commands issued by the computing unit. This provides fundamental hardware support for the efficient closed-loop scanning process of "sparse detection - intelligent analysis - partitioned fine scanning - dynamic supplementation".

[0039] Below, in conjunction with Figure 1 The following describes an adaptive vector network scanning method provided by an embodiment of the present invention, comprising the following steps:

[0040] Step 101: Within the set test frequency band, generate a non-uniformly distributed sparse frequency point sequence through chaotic mapping.

[0041] This step aims to quickly obtain an overview of the electromagnetic response of the tested new energy vehicle component across the entire target frequency band using an efficient non-uniform sampling strategy, laying a data foundation for subsequent intelligent analysis and key area identification.

[0042] In practice, the VNA first sets the start frequency f_start and stop frequency f_stop to be tested. The test frequency band can be specifically set based on the design operating frequency range, application requirements, and standard specifications of the new energy vehicle component being tested. For example, for signal integrity testing of onboard high-speed data harnesses, this frequency band can be set to 100kHz to 20GHz. Simultaneously, based on the trade-off between testing efficiency and initial detection accuracy, a preset total number of sparse frequency points N is established. This value is much smaller than the number of points required for traditional full-band fine scanning; for example, N can be between 100 and 500 points.

[0043] Next, VNA calls its built-in chaotic mapping algorithm module. Chaotic mapping is a non-linear, deterministic mathematical iterative process that exhibits stochastic-like behavior. Its core advantage lies in its ability to generate pseudo-random sequences with ergodicity and uniform distribution with minimal computational cost. For example, the logistic mapping can be used, with its iterative formula being x... n+1 =r*x n *(1-x n The initial value x0 and the control parameter r are chosen within an interval that can produce chaotic behavior (e.g., r is close to 4). The VNA generates a chaotic sequence {x1, x2, ..., xn} containing N values ​​in the interval (0,1) by iterating this formula N times. N}

[0044] Finally, the chaotic sequence is mapped to a set test frequency band through a linear transformation to generate the final sparse frequency point sequence {f1,f2,...,f...}. N}, where each frequency point f i =f start +(f stop -f start )*x i The characteristic of this sequence is that the frequency points are not uniformly distributed throughout the frequency band, which avoids the frequency aliasing problem that may occur with periodic sampling. At the same time, its good ergodicity ensures initial coverage of the entire frequency band and can capture preliminary clues of broadband features or significant resonant regions with a very high probability.

[0045] Step 102: Based on the sparse frequency point sequence, detect the S-parameter measurement data of the vector network of the new energy vehicle component.

[0046] The core task of this step is to transform the aforementioned sparse frequency sequence into actual, analyzable physical measurement data. Among these, the S-parameters (scattering parameters) are core complex indicators used to comprehensively characterize the electrical performance of radio frequency, microwave, and high-speed digital networks. They describe how signal energy is reflected and transmitted within the network by quantifying the linear relationship between the incident, reflected, and transmitted waves at each port. In the S-parameters, S11 is the input port reflection coefficient, used to measure how much signal is reflected back due to impedance mismatch when entering the device's input; S21 is the forward transmission coefficient, used to measure how much signal is successfully transmitted from the input to the output.

[0047] First, the VNA receives a sparse frequency sequence {f1,f2,...,f} generated by a chaotic mapping. N Unlike the linear scanning mode of traditional VNAs, the VNA in this embodiment operates in a "spot frequency" or "list scan" mode. In this mode, the agile frequency synthesizer inside the VNA no longer performs continuous and smooth frequency scanning, but instead performs rapid frequency jumps according to list instructions.

[0048] Specifically, the VNA will sequentially assign frequency points f1, f2, ..., f N As the target frequency, the frequency synthesizer is driven to generate the corresponding radio frequency excitation signal precisely and point-by-point. For each frequency point f in the list... i The VNA performs one complete S-parameter measurement cycle: it f i A frequency signal is injected into the new energy vehicle component under test (such as the impedance matching network of the power battery pack) through port 1. At the same time, the directional coupler and coherent receiver inside the component simultaneously measure the signal reflected back from port 1 and the signal transmitted from port 2 (or other ports).

[0049] By performing precise calculations of the amplitude and phase of these measured vector waveforms, the VNA obtains the value at a specific frequency f. i The S-parameters (scattering parameters) under the given conditions, such as S11(f) which characterizes the reflection loss. i ) and S21(f) characterizing transmission properties i This process is repeated for each frequency point in the sparse frequency point sequence until all N frequency points have been measured.

[0050] Finally, the output of this step is a set of discrete, non-uniformly spaced S-parameter measurement datasets, in the form {(f1,S(f1)),(f2,S(f2)),...,(f...}. N ,S(f N This sparse but information-rich data set forms the direct basis for subsequent feature extraction and intelligent analysis.

[0051] Step 103: Calculate time-domain features and frequency-domain features based on the S-parameter measurement data. The time-domain features include reflection peak features in the time-domain reflection profile, and the frequency-domain features include curvature features, information entropy features, and wavelet transform features.

[0052] This step involves in-depth analysis of the aforementioned sparse S-parameter measurement data. Through multi-dimensional mathematical transformations and feature extraction, key patterns of the electrical behavior of the device under test are identified from seemingly chaotic discrete data points, providing a basis for subsequent scanning strategy optimization. This process consists of two parallel analysis paths: time domain and frequency domain.

[0053] In the time-domain analysis path, the VNA primarily focuses on the measurement data of the reflection parameter S11. This non-uniformly sampled S11 data is interpolated (e.g., by spline interpolation) to obtain a uniformly spaced frequency domain sequence, which is then subjected to an inverse fast Fourier transform (IFFT). The physical meaning of this operation is to transform the frequency-domain reflection response to the time domain, generating a time-domain reflection profile (TDR). The TDR visually displays the impedance discontinuities encountered by the signal as it propagates along the measured component (e.g., a high-voltage power line). Reflection peaks are significant peaks appearing in this time-domain profile; each peak corresponds to a physical location of an impedance abrupt change (e.g., a connector, solder joint, or manufacturing defect), and the height of the peak reflects the severity of the discontinuity.

[0054] In the frequency domain analysis path, VNA directly computes various features on the original sparse frequency domain data.

[0055] Curvature characteristics quantify the curvature of a curve by calculating the approximate second derivative of the S-parameter curve near each measurement point. High curvature values ​​correspond to sharp resonance peaks or steep filter sidebands.

[0056] Information entropy is a concept based on information theory. It divides the frequency band into several sub-windows and calculates the Shannon entropy of the S-parameter amplitude distribution within each window. Regions with high entropy values ​​mean that the signal behavior is more complex and unpredictable, and may hide important details.

[0057] Wavelet transform features utilize wavelet transform as a mathematical microscope to perform multi-scale analysis of S-parameter data. It is particularly good at detecting transient or local singularities in signals and can accurately locate those frequency features that are extremely narrow but have a critical impact, such as narrowband notches in millimeter-wave radar antennas.

[0058] By integrating these time-domain and frequency-domain characteristics, the system can comprehensively perceive the electrical characteristics of the DUT (Device Under Test, referring to new energy vehicle components in this embodiment).

[0059] Step 104: Based on the preset frequency domain feature mutation threshold and the frequency domain features, divide the frequency domain into a high-sensitivity region and a flat region, and determine the time domain key region based on the reflection peak features in the time domain features.

[0060] The purpose of this step is to intelligently divide the entire test frequency band into regions based on the features extracted in the previous stage, thereby directing scanning resources to the frequency band with the most informational value.

[0061] In practice, the VNA first retrieves the frequency domain feature mutation threshold preset in the system configuration. This frequency domain feature mutation threshold is not a single value, but a set of quantitative standards for different frequency domain features (curvature, information entropy, wavelet transform coefficients). It defines what degree of signal change is considered significant. This threshold can be preset according to the type of component under test (such as filter, antenna, cable) or adaptively adjusted through machine learning.

[0062] Subsequently, the VNA iterates through each sparse frequency point calculated in step 103 and its corresponding frequency domain feature value. For any frequency point, if its curvature value exceeds the curvature threshold (meaning that the S-parameter curve bends sharply at this point), or the information entropy of its window exceeds the entropy threshold (meaning that the signal has high local complexity), or the modulus of its wavelet transform coefficients exceeds the wavelet threshold (meaning that there is a transient or singular feature), the frequency point will be marked by the system.

[0063] The VNA then merges all the marked frequency points and their neighboring regions to form one or more consecutive frequency bands, which are formally defined as high-sensitivity regions in the frequency domain. All unmarked, broad frequency bands with flat eigenvalues ​​are classified as flat regions in the frequency domain. Simultaneously, the VNA's processor independently analyzes the time-domain reflection profile, identifying all significant reflection peaks using peak detection algorithms (e.g., finding local maxima exceeding a specific signal-to-noise ratio).

[0064] The time delay position of each reflection peak is precisely recorded, and a time window is defined around that position. This time window is determined as the time-domain critical region, which directly corresponds to a specific physical discontinuity in the tested new energy vehicle component. In this embodiment, the time-domain critical region can specifically refer to the local reflection peak region on the time-domain reflection profile of the tested component, whose amplitude significantly exceeds the noise floor, indicating the existence of an impedance discontinuity in the physical link.

[0065] Step 105: Map the time-domain key region back to the frequency domain to define the frequency domain ripple region, which represents the frequency band where the S-parameters exhibit periodic and weak fluctuations over a wide bandwidth.

[0066] This step is a crucial link between time-domain analysis and frequency-domain scanning strategies. Its purpose is to accurately convert the physical structural defects of the test device (which manifest as reflections in the time domain) into specific scanning requirements in the frequency domain.

[0067] The VNA receives the time-domain critical region information determined in step 104, which is essentially a series of time delay values ​​{τ1, τ2, ...}, where each τ value represents the location of an impedance discontinuity. According to the basic principles of the Fourier transform, an isolated event in the time domain (such as a reflected pulse at τ) will produce a full-frequency, periodic response in the frequency domain.

[0068] Specifically, a reflection with a delay of τ will interfere with the incident signal, thereby superimposing a periodic weak fluctuation, i.e., a ripple, on the amplitude curve of the S-parameters (especially S11). The frequency period Δf of this ripple is exactly the reciprocal of the time delay τ, i.e., Δf = 1 / τ.

[0069] Therefore, mapping the key time-domain region back to the frequency domain is not about locating a narrow frequency point, but rather about identifying the full-band periodic characteristics caused by the time-domain event. The VNA calculates the ripple period Δf corresponding to each τ value. Subsequently, the VNA defines the entire or most of the test frequency band affected by this reflection (which could be the entire scan band) as a special region, namely the "frequency domain ripple region." This definition tells the subsequent scanning algorithm that within this broad frequency band, there exists a small but important periodic signal with a period of Δf. To accurately capture and quantify this ripple caused by the potential defect, the scan step size must be significantly smaller than Δf.

[0070] Step 106: Using a quantum genetic algorithm, assign corresponding optimized frequency step size and number of scanning sampling points to the high-sensitivity region, the flat region, and the ripple region in the frequency domain, respectively, to obtain an optimized scanning strategy.

[0071] This step is the core decision-making process of the entire adaptive scanning method. VNA uses an advanced global optimization algorithm—Quantum Genetic Algorithm (QGA)—to tailor the optimal combination of scanning parameters for different types of frequency domain regions divided in the previous steps.

[0072] The goal of optimizing the frequency step size and the number of scan sampling points is to minimize the total number of scan points (i.e., the shortest test time) while meeting the minimum fidelity requirements of each region.

[0073] Quantum genetic algorithms (QGAs) are intelligent optimization techniques that simulate quantum computing and biological evolution. Their key feature is the use of "qubits (Q-bits)" to represent individuals in the solution space. Unlike the binary bits used in traditional genetic algorithms, Q-bits can exist in a superposition of 0 and 1 states. This results in a more diverse population, allowing QGAs to explore more possibilities simultaneously, thus possessing stronger global search capabilities and faster convergence speeds, effectively avoiding getting trapped in local optima.

[0074] In its implementation, the VNA encodes the scanning strategy (i.e., the frequency step size set for the three types of regions: high-sensitivity region, flat region, and ripple region) into a Q-bit chromosome. Next, the VNA defines a "fitness function" that comprehensively evaluates the merits of a scanning strategy. Its evaluation criteria can include: predicted measurement accuracy (e.g., ensuring the step size in the high-sensitivity region is sufficient to resolve the sharpest resonance peak, and the step size in the ripple region is less than the ripple period) and total testing time (proportional to the total number of scan points). The QGA iteratively updates the Q-bit chromosome population by simulating the evolution of quantum states. Each generation undergoes a quantum rotation gate operation, causing the population to evolve towards higher fitness.

[0075] After several iterations, the quantum genetic algorithm converges to an optimal solution. The VNA decodes this optimal Q-bit chromosome to obtain the final optimized scanning strategy, which is to clearly allocate specific frequency step sizes and corresponding number of scan points to the three types of regions.

[0076] In this embodiment, the fitness function is a mathematical expression used by the quantum genetic algorithm to evaluate the merits of the "scanning strategy." The scanning strategy includes the frequency step size and the number of scan sampling points allocated to different frequency domain regions (high-sensitivity region, flat region, ripple region). The fitness function can be constructed as follows:

[0077] Fitness function (F) = Accuracy score - Total test point penalty; where:

[0078] Accuracy score: If the allocation step size of both the high-sensitivity region and the ripple region meets or is less than the preset minimum step size (e.g., high-sensitivity region step size ≤ 0.1MHz, ripple region step size ≤ 0.2MHz), then the accuracy score is a large positive value (e.g., 100 points).

[0079] If the step size in any region exceeds the preset minimum step size requirement, the accuracy score is set to a very small negative value or zero (e.g., -1000 points) to strongly penalize strategies that do not meet the accuracy requirements. This ensures that the algorithm will not choose strategies that sacrifice accuracy for speed.

[0080] Total test point penalty: Calculates the total number of test points required to scan all frequency domain regions under this scanning strategy. The more test points, the greater the penalty.

[0081] The penalty value can simply be equal to the total points multiplied by a small negative weighting coefficient (e.g., -0.01), or the penalty value = the total points.

[0082] For example: Total number of points = (length of high-sensitivity area / step size of high-sensitivity area) + (length of flat area / step size of flat area) + (length of ripple area / step size of ripple area).

[0083] Ultimately, the quantum genetic algorithm attempts to maximize the value of this fitness function. This means that the algorithm will look for those that: a) meet or exceed the measurement accuracy requirements of critical regions (highly sensitive regions and ripple regions) (by ensuring that the step size is small enough) to obtain a high "accuracy score".

[0084] b) At the same time, while meeting the accuracy requirements, try to reduce the total number of scan points to reduce the "total test point penalty".

[0085] In this way, the fitness function guides the quantum genetic algorithm to optimize the scanning speed while ensuring measurement accuracy, thus alleviating the contradiction between speed and accuracy.

[0086] Step 107: Perform segmented scanning on the frequency domain high-sensitivity region, the frequency domain flat region, and the frequency domain ripple region according to the optimized scanning strategy to obtain real-time scanning data.

[0087] This step transforms the abstract scanning blueprint developed in the previous steps into concrete physical measurement actions.

[0088] First, the VNA analyzes and optimizes the scanning strategy. This optimized scanning strategy is essentially a detailed "segmented scanning task list", which clearly specifies the starting frequency, ending frequency, and optimal frequency step size or number of sampling points for each segment (such as the high-sensitivity region, the flat region, or the ripple region in the frequency domain).

[0089] Subsequently, the VNA performs measurements sequentially according to this segmented scan task list. When the scan enters the flat frequency region, the VNA performs a rapid frequency-hopping scan with a large frequency step size (e.g., 10 MHz), quickly covering this broad, flat frequency band with minimal sampling points. When the scan enters the marked high-sensitivity frequency region, the scan mode seamlessly switches, and the VNA performs a fine linear scan with an extremely small step size (e.g., 0.1 MHz), accurately depicting the details of resonance peaks or steep sidebands with extremely high resolution. Similarly, in the ripple region, the VNA scans with a precisely calculated step size (e.g., 0.5 MHz) ensured to be smaller than the ripple period, guaranteeing that enough points are sampled to reconstruct the periodic, weak fluctuations caused by time-domain reflections.

[0090] Throughout the segmented scanning process, the VNA receiver and data acquisition unit work synchronously, storing the S-parameter amplitude and phase data measured at each frequency point into memory in real time, forming a raw measurement dataset containing different regions and different sampling densities, namely "real-time scan data".

[0091] Step 108: Calculate the local fitting residuals for each region based on the real-time scanning data.

[0092] This step is a key quality control step in the entire adaptive scanning method. Its purpose is to quickly and quantitatively verify the validity of the collected data after the initial optimized scan is completed.

[0093] The VNA processes the "real-time scan data" obtained in step 107 for each region. For each region (such as a high-sensitivity region or a flat region in the frequency domain), the VNA selects a mathematical model suitable for the characteristics of that region to locally fit the discrete data points within that region. For example, for a "flat region in the frequency domain" with a expected smooth response, a low-order polynomial (such as a linear or quadratic function) may be used for fitting; while for a "high-sensitivity region in the frequency domain" with drastic changes, a higher-order polynomial or spline interpolation function may be used.

[0094] The local fit residual is defined as the absolute or squared value of the difference between the actual measured value and the predicted value of the fitted model at each measured frequency point (data point). Simply put, the local fit residual quantifies the degree to which the actual data deviates from a smooth trend.

[0095] The VNA calculates the residuals of all data points within the region and identifies the maximum value or calculates its root mean square value, which serves as an evaluation metric for the overall fit quality of the region. A small residual value means that the current sampling point can well describe the signal behavior of the frequency band, while a large residual value is a warning sign, indicating that there may be undetected, drastic signal changes between two sampling points, and the existing scan density may be insufficient.

[0096] Step 109: If any of the local fitting residuals exceeds the confidence threshold, then supplementary frequency points are dynamically inserted in the corresponding frequency domain interval and supplementary scanning is performed to obtain supplementary scanning data.

[0097] This step constitutes the closed-loop feedback and correction mechanism of the method of the present invention, ensuring the reliability and completeness of the final measurement results and avoiding the omission of key features due to minor deviations in the initial strategy.

[0098] First, the VNA compares the local fitting residuals of each region calculated in step 108 with a preset confidence threshold. This confidence threshold is an engineering-defined upper tolerance, representing the maximum acceptable fitting error. The confidence threshold is set to balance accuracy requirements with test time control; for components with high requirements, this threshold is set lower. If the residuals of all regions are below this threshold, the preliminary scan results are considered reliable, and the process will proceed directly to the final step.

[0099] However, once the VNA detects that the local fitting residual in any frequency domain interval exceeds the confidence threshold, it determines that the scan density in that interval is insufficient. At this point, the VNA immediately initiates a dynamic supplementation procedure. It precisely locates the frequency neighborhood of the data points or data points with the largest residual and "dynamically inserts supplementary frequencies" within this neighborhood.

[0100] In this embodiment, the insertion strategy can be to perform bisection interpolation between the two adjacent measurement points with the largest residuals, that is, to generate a new frequency point to be measured at their midpoint. Subsequently, the VNA will only perform an efficient supplementary scan, i.e., point frequency measurement, on these newly generated supplementary frequency points, without rescanning the entire region. The obtained measurement data is the supplementary scan data, which will be used to fill in the previous data gaps and improve the detail resolution of local areas.

[0101] In some embodiments, drastic changes in the S-parameter curve (leading to excessive fitting residuals) are not random noise, but are caused by specific physical characteristics of the tested new energy vehicle components (such as resonance, impedance abrupt changes, etc.). Therefore, the precise frequency location of this "hidden" feature point can be predicted by quickly fitting a physical model to existing data, and then supplementary scanning can be performed preferentially at this predicted point to achieve "precision-guided" supplementation, rather than "carpet-bombing" encryption. The specific implementation method is as follows:

[0102] 1. Local Feature Recognition and Model Selection: When a local fitting residual exceeding the limit is detected between frequency points f_A and f_B, interpolation is not performed immediately. Instead, the local variation trend of the S-parameters (including amplitude and phase) of f_A and f_B and several neighboring measured data points (e.g., f_{A-1}, f_A, f_B, f_{B+1}) is analyzed first.

[0103] If a local curve exhibits a "U"-shaped or inverted "U"-shaped trend, it is determined that a resonance peak or notch may exist at that location. In this case, a simplified second-order resonator model (such as the transfer function model of an RLC parallel / series circuit) will be automatically selected as the fitting target.

[0104] If the phase of a local curve flips rapidly, it is determined that there may be a zero or pole, and a simple rational fraction (pole-zero) model is selected.

[0105] 2. Fast parameter identification and feature frequency prediction: Using these few (e.g., 4) local data points, a lightweight optimization algorithm (such as a simplified version of particle swarm optimization or genetic algorithm) is used to quickly identify the parameters of the physical model selected in the previous step.

[0106] For the resonator model, the goal of the algorithm is to quickly estimate the three key parameters of the model: the resonant center frequency (f_peak), the quality factor (Q value), and the peak amplitude.

[0107] The estimated "resonance center frequency (f_peak)" is the precise frequency of the hidden feature point predicted by VNA that is most likely to cause a huge change in the curve.

[0108] 3. Priority-based dynamic interpolation strategy: Based on the identified model parameters, a priority list of supplementary frequency points is generated: First priority (P1): The predicted resonant center frequency (f_peak) is used as the first scanning point that must be supplemented. This is because peak or valley points contain the most information and have the greatest impact on the curve shape.

[0109] Second priority (P2): Based on the estimated Q value, calculate the half-power point bandwidth of the resonance (3dB bandwidth, approximately equal to f_peak / Q). Then, the two frequency points f_peak ± (bandwidth / 2) are used as supplementary points for the second priority. These two points are crucial for accurately characterizing the steepness and width of the resonance peak.

[0110] Third priority (P3): If more points are needed according to the set supplementary density level, then uniform or curvature-weighted interpolation will be performed between the above feature points.

[0111] Execution and Iteration: The VNA first performs a supplementary scan on frequency point P1. If the fitting residual for this interval has fallen below the confidence threshold after one supplementary scan, the supplementary scan process ends, greatly saving time. If the residual is still high, the scan continues on frequency point P2, and so on.

[0112] In new energy vehicles, high-frequency components, such as poor contact in connectors and manufacturing tolerances in antenna units, are highly susceptible to introducing unexpected parasitic resonances. This embodiment employs the method described above, which can effectively and sensitively capture these narrow-band, sharp characteristic signals caused by physical defects, demonstrating strong practical relevance.

[0113] Step 110: Based on the real-time scanning data and supplementary scanning data, generate the S-parameter curve of the new energy vehicle component in the test frequency band.

[0114] This step is the final result summary and presentation stage of the entire adaptive scanning process. Its aim is to integrate the discrete data points collected in all previous stages into a complete, accurate, and user-friendly S-parameter curve. Specifically, the S-parameter curve is a two-dimensional graph plotted with the amplitude (usually in dB) or phase (in degrees) of the scattering parameters (S-parameters) measured by the vector network analyzer as the vertical axis and frequency as the horizontal axis. It visually demonstrates how the electrical performance (such as reflection, transmission, and isolation characteristics) of the device under test changes with frequency.

[0115] The VNA first creates a unified dataset and then integrates all valid measurement data from different stages. This includes: the initial sparse frequency point data obtained in step 102, the main "real-time scan data" obtained from segmented scanning in step 107, and the "supplementary scan data" that may be generated in step 109 for correction.

[0116] The VNA sorts all these data points in ascending order of frequency, forming a single, non-uniformly spaced final frequency-S-parameter data list. This list is the most original and accurate result produced by the method in this embodiment. It records the electrical response of the tested new energy vehicle component in the entire test frequency band in the most efficient way (i.e., dense where needed and sparse where not needed).

[0117] Finally, to facilitate engineers' observation and analysis, the VNA display system graphically processes this final discrete dataset. By invoking interpolation algorithms (such as linear interpolation, spline interpolation, or sinc interpolation), it smoothly connects adjacent data points, thereby generating a visually continuous "S-parameter curve" on the screen. This S-parameter curve realistically reflects all key details, including sharp peaks and valleys in highly sensitive areas as well as subtle fluctuations in ripple areas, while avoiding wasting display resources in flat areas. This provides a comprehensive and reliable basis for evaluating the performance and quality of new energy vehicle components.

[0118] In some embodiments, the S-parameter curve can be referenced. Figure 2 , Figure 2 The S11 reflection coefficient curves covering the entire test frequency band from 1MHz to 1GHz are shown after using the adaptive vector network scanning method of this embodiment. Figure 2 The algorithm's intelligent frequency band segmentation results are visually indicated by different colored background areas: the red background area represents the "high-sensitivity frequency domain region" containing sharp resonance peaks, the orange background area represents the "frequency domain ripple region" caused by physical reflections, and the broad green background area corresponds to the "frequency domain flat region" with a smooth response. The entire curve not only accurately captures the key defect peak at 253MHz and the connector resonance at 680MHz, but also fully analyzes the periodic ripples caused by the wiring harness structure. At the same time, the different smoothness of the curves in different regions also indicates that the system dynamically matches the optimal measurement parameters (such as scan step size, intermediate frequency bandwidth, etc.) for each region, thereby achieving efficient scanning of the entire frequency band while ensuring the accuracy of key details.

[0119] Furthermore, such as Figure 3 As shown, a magnified view of the region where the 253MHz defect peak is located is displayed, revealing in detail the execution process of the core closed-loop feedback mechanism of "real-time verification and dynamic supplementary scanning" in this embodiment. Figure 3The blue dots represent sparse scan points generated by the initial chaotic mapping. While these points initially identify features, their insufficient density prevents accurate reconstruction of the peak shape, leading to excessive local fitting residuals. In response, the algorithm automatically initiates a correction process, dynamically inserting supplementary scan points (shown as orange squares) at the most critical frequency points near the peak. The point where the residual exceeds the threshold triggers the supplementary point is where the fitting residual between the initial sparse scan points is largest, revealing the causal logic of the entire dynamic correction process: it is precisely because a large fitting error (cause) is detected that the algorithm is triggered and performs a precise supplementary scan (effect) at that point, thus ensuring the reliability of the final result. Finally, by integrating the initial and supplementary data, the system generates a smooth and accurate red solid line curve, ensuring that even critical defect signals in extremely narrow bands can be reliably and faithfully captured by this method, fully demonstrating the robustness and high precision of this scheme.

[0120] This invention, through a series of innovative adaptive steps, alleviates the technical contradiction between speed and accuracy in traditional vector network scanning. First, through chaotic mapping sparse detection and multi-dimensional feature extraction in steps 101-103, this method can rapidly construct the electrical feature profile of the device under test in a very short time, achieving efficient "reconnaissance." Next, in steps 104-105, based on joint time-frequency analysis, the entire frequency band is accurately divided into high-sensitivity, flat, and ripple regions, achieving intelligent "zoning planning" of test resources. The core advantage is reflected in step 106, where a quantum genetic algorithm is used to tailor the optimal scanning strategy for different regions, precisely allocating limited testing time to the most critical frequency bands, achieving "optimal configuration" of scanning efficiency. Finally, through the closed-loop feedback mechanism of segmented scanning, real-time residual monitoring, and dynamic supplementary scanning in steps 107-110, it is ensured that no potentially critical details are missed during high-speed scanning, achieving "high confidence and completeness" of the measurement results. In summary, this method transforms the traditional, time-consuming, and inefficient "carpet-bombing" scanning process into a highly intelligent testing workflow of "reconnaissance-analysis-precision strike-results evaluation and remediation." This reduces testing time by several orders of magnitude while ensuring the accurate characterization of key performance indicators (such as weak defects and sharp resonances) of new energy vehicle components, perfectly meeting the dual demands of high efficiency and high precision testing for large-scale automotive production.

[0121] The following is combined Figure 2 The adaptive vector network scanning method of this invention further illustrates the following steps:

[0122] Step 201: Set the initial values ​​and nonlinear control parameters for the chaotic mapping.

[0123] This step is the initialization and configuration stage of the entire chaotic sequence generation process. Its core is to set a precise mathematical starting point and behavioral pattern for subsequent iterative operations.

[0124] Based on the Logistic mapping mentioned in the foregoing embodiments, its iterative formula is x n+1 =r*x n *(1-x n In this step, two key parameters need to be set.

[0125] The first is the initial value x0, which is the seed or starting point for iterative calculation. It can be selected as an irrational number in the interval (0,1) or a value that avoids special points, so as to ensure that the chaotic characteristics of the system can be fully stimulated.

[0126] The second is the nonlinear control parameter *r*, which directly determines the dynamic behavior of the mapping function. To generate a sequence with good ergodicity and pseudo-randomness, the value of *r* is set within the chaotic region of the system. For the logistic mapping in this embodiment, the value of *r* is between 3.57 and 4.0. When the value of *r* is less than 3.57, the system will exhibit periodicity or converge to a fixed point, failing to generate a non-uniform sequence with good ergodicity; while values ​​exceeding 4.0 will cause the sequence to diverge. Therefore, in this embodiment, *r* is preferably set to a constant value close to 4.0, such as 3.99, to maximize the chaotic characteristics of the sequence and its coverage in the (0,1) interval, thereby providing the most effective frequency distribution for subsequent sparse scanning.

[0127] By precisely setting these two parameters, VNA constructs a deterministic mathematical model that is ready to generate a unique, aperiodic numerical sequence that is extremely sensitive to initial conditions, laying the foundation for subsequent non-uniform frequency point generation.

[0128] Step 202: Perform multiple iterative calculations based on nonlinear control parameters to obtain a sequence of chaotic variables.

[0129] This step is the core execution stage for actually generating the chaotic sequence. In the VNA's calculation program, the initial value x0 and the nonlinear control parameter r set in the previous step are used as inputs to start the iterative calculation. Specifically, x0 is substituted into the chaotic mapping formula (such as the logistic mapping formula) to calculate the first value x1; then, x1 is used as the new input and substituted into the formula again to calculate x2; this process is repeated, using the output of the previous iteration as the input of the next iteration, and the process is executed continuously for a preset number of iterations (N times). This value N corresponds to the total number of sparse frequency points that are ultimately desired to be generated.

[0130] After N iterations, a sequence {x1,x2,...,xN} containing N values ​​is obtained, which is the chaotic variable sequence. This sequence has the typical characteristics of a chaotic system: all its values ​​are restricted to the normalized interval (0,1); the sequence itself does not exhibit any simple periodicity, but behaves like a random number; at the same time, it has ergodicity, that is, with a sufficient number of iterations, the values ​​in the sequence will be almost uniformly distributed throughout the interval (0,1).

[0131] This feature is crucial for ensuring that the initial sparse scan can cover the entire test band unbiasedly, avoiding the risk of spectral aliasing and information loss that may be caused by periodic sampling.

[0132] Step 203: Map the chaotic variable sequence to the test frequency band to form a non-uniformly distributed sparse frequency point sequence.

[0133] This step aims to transform the abstract, normalized sequence of chaotic variables generated in the previous step into a list of specific frequency values ​​that can be directly measured on the VNA. This process is a simple linear scaling transformation. First, the system acquires the complete "test band" to be measured, which is explicitly defined by a start frequency f_start and an end frequency f_stop, for example, 1MHz to 1GHz for automotive Ethernet harness testing. Next, the system processes the "chaotic variable sequence" {x1,x2,...,x...} N Each chaotic variable x in} i Perform a linear mapping operation. The specific mapping formula is: f i =f start +(f stop -f start )*x i The function of this formula is to distribute each x, which was originally distributed in the interval (0,1), into a single integer. i The values ​​are proportionally stretched and shifted to the frequency range [f_start, f_stop]. After this transformation, the final output {f1, f2, ..., f_stop} is obtained. N This refers to a "non-uniformly distributed sparse frequency point sequence." "Sparse" means the total number of points N is much smaller than the number of points in a traditional scan, ensuring high detection efficiency. "Non-uniform distribution" inherits the inherent characteristics of chaotic sequences, making the intervals between these detection frequency points vary and appear random. This allows for a rapid and comprehensive preliminary "snapshot" of the full-band response of the device under test in an unconventional way, providing an efficient and information-rich data foundation for subsequent intelligent analysis and key area identification.

[0134] Step 204: Based on the sparse frequency point sequence, detect the S-parameter measurement data of the vector network of new energy vehicle components.

[0135] This step can be referred to the description in the foregoing embodiments, and will not be repeated here.

[0136] Step 205: Calculate the time-domain and frequency-domain characteristics based on the S-parameter measurement data.

[0137] This step can be referred to the description in the foregoing embodiments, and will not be repeated here.

[0138] Step 206: Based on the preset frequency domain feature mutation threshold and frequency domain features, divide the frequency domain into a high-sensitivity region and a flat region, and determine the time domain key region based on the reflection peak features in the time domain features.

[0139] This step can be referred to the description in the foregoing embodiments, and will not be repeated here.

[0140] Step 207: Extract the characteristic time delay corresponding to each reflection peak from the key time domain region.

[0141] The core task of this step is to accurately quantify the location information of each physical discontinuity point in the time-domain reflection profile. The VNA will initiate a sophisticated peak detection algorithm within the critical region of the time domain. This algorithm will scan the time-domain reflection profile curve point by point, looking for local maximum points whose amplitudes significantly exceed the preset noise floor; these points are called "reflection peaks". Once a reflection peak is identified, the algorithm will accurately record the position of the peak point on the horizontal axis; this time value is the "characteristic delay" (τ). Physically speaking, this characteristic delay represents the total time it takes for the test signal emitted by the VNA to propagate from the test port to a certain impedance discontinuity point inside the new energy vehicle component (such as a poor connector crimp or a damaged insulation layer), then reflect back, and be captured by the receiver again; that is, the two-way propagation delay.

[0142] Therefore, each extracted feature delay uniquely and quantitatively corresponds to the spatial location of a specific physical feature or potential defect within the test component. Ultimately, the output of this step is a list containing one or more feature delays {τ1, τ2, ...}, providing a precise time reference for subsequent frequency domain feature analysis.

[0143] Step 208: Determine the theoretical ripple period in the frequency domain based on the characteristic time delay.

[0144] This step is a crucial theoretical transition linking time-domain analysis and frequency-domain scanning strategies, based on the duality principle of the Fourier transform in signal processing. An isolated event in the time domain (such as a reflected pulse with a delay of τ) will produce a periodic response throughout the entire frequency spectrum.

[0145] VNA uses this basic principle to convert the time delay τ of each feature extracted in the previous step into its corresponding feature in the frequency domain.

[0146] Specifically, when a reflected signal with a delay of τ interferes with the original incident signal, a sinusoidal or cosine-shaped fluctuation is superimposed on the amplitude response curve of the S-parameters (especially S11). This fluctuation is called the ripple. The ripple exhibits periodicity on the frequency axis, and the frequency interval between two adjacent peaks or troughs is constant; this interval is called the "theoretical ripple period" (Δf_theoretical). Its calculation formula is very straightforward: Δf_theoretical = 1 / τ. Therefore, the VNA processor simply takes the reciprocal of each characteristic delay τ to calculate the corresponding theoretical ripple period. The term "theoretical" is used because this calculation is based on an ideal assumption that the signal propagation speed in the transmission path does not change with frequency. The output of this step is a set of theoretical ripple period values, which preliminarily reveal the minimum frequency resolution required to resolve these frequency domain details caused by reflection.

[0147] In some embodiments, step 208 may specifically include steps S2081-S2082:

[0148] S2081, based on characteristic time delay and reflection physical mechanism, determines the two-way propagation time delay of electromagnetic waves in the signal transmission path.

[0149] The core of this step is to give the time value extracted in the previous step, which is purely mathematical in nature, a clear physical meaning, so as to provide a solid foundation for subsequent calculations.

[0150] In this embodiment, the characteristic time delay refers to the time coordinate corresponding to the reflection peak directly read from the time-domain reflection profile. In the physical mechanism of reflection, this time is not the time it takes for the signal to propagate unidirectionally from the starting point to the endpoint. In fact, it is the excitation signal emitted by the VNA that propagates along the signal transmission path of the new energy vehicle component (such as a high-voltage wire harness) to a certain impedance discontinuity point, where the signal energy is partially reflected. The reflected wave then returns along the original path to the test port of the VNA and is received. Therefore, the physical essence of this measured characteristic time delay is the total time taken for the electromagnetic wave to complete the entire round-trip path of "going out" and "returning".

[0151] The implementation of this step is the logical formal confirmation of this physical model by the VNA algorithm, which directly equates each extracted feature delay (τ) to the "two-way propagation delay". This confirmation step is crucial because it ensures that subsequent calculations based on this time value are grounded in a correct understanding of the actual physical process, avoiding a doubling of the error in subsequent ripple period calculations due to misuse of the one-way delay.

[0152] S2082 converts the two-way propagation delay into the periodic parameters of the corresponding frequency domain ripple, thus obtaining the theoretical ripple period.

[0153] This step is a crucial mathematical transformation linking time-domain analysis and frequency-domain scanning strategies, based on the principle of duality of the Fourier transform in signal processing. The VNA will convert the two-way propagation delay (τ) determined in the previous step into its corresponding characteristic in the frequency domain.

[0154] In practice, the VNA uses a fundamental mathematical conversion formula: Δf = 1 / τ. The physical meaning of this formula is that the interference between the reflected wave with a delay of τ and the main signal will superimpose a periodic ripple on the amplitude curve of the S-parameters. The frequency interval between two adjacent peaks or troughs of this ripple (the ripple period) is exactly the reciprocal of the two-way propagation delay. The result calculated in this step is called the "theoretical ripple period." The term "theoretical" is used because the calculation is based on an ideal assumption: the propagation speed of electromagnetic waves in the transmission medium does not change with frequency (i.e., no dispersion effect).

[0155] Therefore, the output of this step is a constant frequency value, which provides an initial benchmark for subsequent more refined dispersion correction and initially reveals the order of magnitude of frequency resolution required to capture this particular defect signal.

[0156] Step 209: Based on the dispersion characteristics of the signal transmission path, the theoretical ripple period is corrected for frequency dependence to obtain the actual ripple period.

[0157] This step is a precise correction to the idealized calculation in the previous step, aiming to match the theoretical model with the actual physical characteristics of new energy vehicle components, thereby significantly improving the accuracy of subsequent scanning strategies.

[0158] Dispersion characteristics refer to the phenomenon where the propagation speed (phase velocity) of electromagnetic waves in actual transmission media (such as the insulation material of high-voltage wire harnesses and the polymer medium of high-speed data cables) changes with frequency. This is a real physical phenomenon. This frequency dependence of the speed directly results in the frequency domain ripple period caused by reflection not being a fixed constant, but rather changing slowly with frequency throughout the entire test frequency band.

[0159] In practice, the VNA first extracts the phase information of the S21 parameter obtained from the initial sparse scan, {∠S21(f i Next, the group delay of phase S21 is calculated. The group delay is defined as the negative value of the derivative of the phase response with respect to the angular frequency (τ). g=-dφ / dω), which physically represents the time required for the signal envelope to pass through the device under test. Since the phase data is discrete, the VNA obtains a set of discrete group delay estimates {τ} by performing finite difference calculations on the phase difference and frequency difference between adjacent sparse frequency points. g (f i This set of data directly reflects the trend of the actual propagation delay of the signal in the device under test as a function of frequency, that is, its measured dispersion characteristics.

[0160] Subsequently, a function fitting (e.g., polynomial or spline fitting) is performed on this set of sparse group delay data points to generate a continuous, empirical group delay model τ_g(f).

[0161] Finally, the VNA uses a group delay model to correct the theoretical ripple period. Since the ripple period is inversely proportional to the two-way propagation delay that produces it, the actual ripple period Δf_actual(f) is inversely proportional to the group delay τ_g(f) at that frequency.

[0162] By using a proportional relationship, the theoretical ripple period is dynamically adjusted using the group delay model, thereby obtaining a more accurate actual ripple period that varies with frequency.

[0163] In some embodiments, step 209 may specifically include steps S2091-S2093:

[0164] S2091, obtain the medium parameters of the vector network transmission path in the new energy vehicle components from the preset material property database.

[0165] This step is the starting point for precise correction. Its core is establishing a precise digital electromagnetic material profile for the component under test within the internal system of the vector network analyzer (VNA). In practice, the VNA first selects the specific model of the new energy vehicle component being tested from a list, such as "XX brand 800V high-voltage power cable - Model A" or "YY brand automotive Ethernet connector - Model B". Once selected, the VNA's processor uses this model as an index to query its internally stored or accessible "pre-configured material property database". This database, pre-configured by the user before the instrument leaves the factory or according to the specific application, stores various "dielectric parameters" of engineering materials commonly used in automotive electronics. These dielectric parameters mainly refer to the relative permittivity (εr) and loss tangent (tanδ), which together determine the speed and attenuation of electromagnetic waves propagating in the material. Crucially, the database does not store single numerical values, but rather functional models or data tables of these parameters as a function of frequency (i.e., εr(f) and tanδ(f)), because the electrical properties of almost all insulating materials exhibit frequency dependence, i.e., dispersion characteristics. Ultimately, the output of this step is that the VNA successfully retrieves and loads a set of frequency-related medium parameter data that perfectly matches the device under test, providing the most basic physical input for the subsequent construction of an accurate propagation model.

[0166] The material property database is constructed as follows:

[0167] First, standardized material testing methods, such as the resonant cavity method, transmission line method, or free space method, are used to accurately measure the original S-parameters of the target material (such as XLPE cross-linked polyethylene commonly used in new energy vehicle wiring harnesses, PBT plastic in connectors, etc.) over a wide frequency range using a vector network analyzer.

[0168] Next, using specialized materials analysis software, the measured S-parameter data are used to derive the curve or data table showing the change of the material's complex permittivity (including the relative permittivity εr and the loss tangent tanδ) with frequency.

[0169] Finally, these verified structured datasets, which include material type, manufacturer, test conditions, and frequency-related dielectric parameters, are integrated, stored, and indexed into an electronic database that can be quickly queried and accessed by the VNA system software. This provides reliable and traceable physical parameter support for subsequent dispersion correction.

[0170] S2092, based on medium parameters, constructs a phase velocity frequency dependence characteristic model of electromagnetic wave propagation in the vector network transmission path.

[0171] The core task of this step is to transform the relatively abstract material-level medium parameters obtained in the previous step into a more physically meaningful mathematical model that directly describes the signal propagation behavior.

[0172] VNA utilizes fundamental formulas from electromagnetic field theory to build its model. Phase velocity (vp) refers to the speed at which a single frequency component of an electromagnetic wave, at its point of equal phase (e.g., a wave crest), propagates through a medium. Phase velocity is related to the effective relative permittivity (ε) of the medium. eff Closely related, the basic relationship is phase velocity v p = c / sqrt(ε eff ), where c is the speed of light in a vacuum, sqrt(ε eff The effective refractive index is ε. The key is the effective relative permittivity ε. eff It is itself a function of frequency, and it is determined by the medium parameter ε obtained in the previous step. r (f) is determined by the geometry of the transmission line.

[0173] The VNA will call the corresponding electromagnetic model based on the type of the device under test (such as coaxial cable, twisted pair, etc.) and convert ε r (f) Substitute the values ​​to calculate the phase velocity v. p As a function of frequency, i.e., v p (f). This function v p (f) is the phase velocity frequency dependence characteristic model. The establishment of this phase velocity frequency dependence characteristic model means that the VNA no longer assumes that the signal propagates at a constant speed, but accurately describes that signal components of different frequencies will travel at different speeds on the same path, which is the essence of the dispersion phenomenon.

[0174] In some embodiments, the phase velocity frequency-dependent characteristic model, i.e., the function v p The mathematical expression for (f) may include:

[0175]

[0176] Among them, v p (f) represents the final phase velocity frequency dependence characteristic model.

[0177] f represents a specific frequency point within the test band.

[0178] c represents the speed of light in a vacuum.

[0179] ε eff (f) represents the dielectric constant value that electromagnetic waves experience when propagating in a vacuum, under the combined effect of the geometry of the transmission line and the dielectric material at a frequency of f.

[0180] ε r(f) represents the relative permittivity of the material at frequency f. It is an input obtained directly from the VNA material property database and represents the material dispersion.

[0181] ε eff,dc DC Effective Permittivity is a dimensionless constant that describes the equivalent dielectric constant of a transmission line structure (e.g., twisted pair, microstrip line) as the frequency approaches zero; it is determined by the ε-coefficient of the material. r (0) and the geometric layout of the conductor together determine, where ε r (0) refers to the relative permittivity of the material at a frequency of f = 0 Hz, which is the relative permittivity of the material under a DC electric field, and is usually also called the static permittivity.

[0182] f p The frequency of the structural poles, measured in Hertz (Hz), represents the characteristic frequency at which structural dispersion begins to become significant. This frequency is closely related to the cross-sectional dimensions of the transmission line (such as conductor spacing and insulation thickness). For example, in a microstrip line, fp... p Typically, the substrate thickness h, conductor width W, and dielectric constant ε of the microstrip line are related. r Related, can be expressed as f p =F(W,h,ε) r )express.

[0183] In this embodiment, f p This can be obtained through simulation or experimental fitting. For example, using professional electromagnetic simulation software (such as ANSYS HFSS, CST Studio Suite), different values ​​(W, h, ε) can be scanned. r Parameter combinations. Frequency domain simulation is performed for each combination to extract the effective relative permittivity ε. eff The curve is obtained, and then the simulation results are matched with the phase velocity frequency dependence characteristic model through a curve fitting algorithm, thereby deriving the optimal f. p Finally, the values ​​obtained from a large number of simulations (W,h,ε) are... r ) as input and corresponding output f p By combining these data into a dataset and utilizing multivariate fitting or machine learning techniques, a high-precision f-value can be constructed. p Mapping function, i.e., f p =F(W,h,ε) r ), or form f p With (W,h,ε) r Look-up Table.

[0184] This embodiment's phase velocity frequency-dependent characteristic model clearly separates the two physical effects:

[0185] ε r (f) directly represents the physical properties of the material itself (i.e., material dispersion).

[0186] This represents the effect of frequency-dependent field distribution caused by the transmission line geometry (structural dispersion).

[0187] For a specific type of test object (such as coaxial cable or a specific type of twisted pair), ε eff,dc and f p These two structural parameters are fixed. They can be obtained in advance through electromagnetic simulation calculations or calibration measurements and are stored in the VNA's database along with the material data. The VNA can automatically retrieve these parameters when the user selects the type of part under test.

[0188] v in this embodiment p (f) provides a high-precision model that dynamically reflects real physical properties. Through a carefully designed mathematical structure, it integrates two core dispersion effects—namely, material dispersion caused by the insulating medium itself (manifested in ε)—within the model. r (f) and structural dispersion caused by conductor geometry (reflected in ε) eff,dc and f p The parameters have been effectively fused. Therefore, the model no longer simply assumes a fixed propagation speed, but can accurately depict the actual propagation speed of signal components of different frequencies in complex transmission paths such as wiring harnesses in new energy vehicles.

[0189] Especially considering the wide signal bandwidth (from low-frequency CAN bus to GHz-level automotive Ethernet) in applications such as high-voltage wiring harnesses and high-speed data links in new energy vehicles, dispersion effects are very significant within such a wide frequency band. Ignoring dispersion and using a fixed phase velocity to calculate the ripple period would lead to huge calculation errors at high frequencies, making it impossible to accurately define the "frequency domain ripple zone," potentially causing the adaptive scanning strategy to fail. This embodiment designs a phase velocity frequency-dependent characteristic model that improves calculation accuracy and ensures the accuracy of subsequent ripple period correction, thus providing a solid physical model foundation for the reliability and accuracy of the entire adaptive scanning method.

[0190] S2093, based on the phase velocity frequency dependence characteristic model, performs frequency-varying compensation on the theoretical ripple period to obtain the actual ripple period.

[0191] This step is the final step in the correction process. It applies the dynamic velocity model established in the previous step to the actual ripple period calculation, achieving a precise transition from theory to practice. VNA first reviews the fundamentals of the theoretical ripple period calculation: Δf_theoretical = 1 / τ, where τ is the two-way propagation delay. τ is also equal to 2L / vp, where L is the physical distance to the reflection point and vp is the phase velocity.

[0192] In theoretical calculations, vp is assumed to be a constant. Now, VNAs have a more accurate phase velocity model, vp(f). Frequency compensation involves substituting this frequency-varying vp(f) into the formula for calculating the ripple period.

[0193] The VNA first estimates the physical distance L based on the initial τ and vp at a certain reference frequency. Then, it recalculates the ripple period using this fixed L and the dynamic vp(f). Therefore, the corrected "actual ripple period" Δf_actual is no longer a constant, but a function that varies with frequency: Δf_actual(f) = vp(f) / (2L). This means that the ripple period may be wider at low frequencies, while at high frequencies, the ripple period may become narrower due to changes in phase velocity caused by dispersion effects.

[0194] Through this compensation, the VNA can predict that the frequency step size required to resolve ripple will be different when scanning at different frequency bands, thus providing the most accurate and dynamic basis for subsequent adaptive scan step size allocation.

[0195] Step 210: Based on the actual ripple period, extend a symmetrical frequency band in the frequency domain to define the frequency domain ripple region.

[0196] The purpose of this step is to formally define a "frequency domain ripple region" that requires special handling in the VNA scan plan and to provide clear quantitative guidance for its subsequent fine scanning.

[0197] After calculating the actual ripple period, the VNA needs to determine the frequency range affected by this ripple characteristic. Since the interference ripple generated by a single discrete reflection point can theoretically extend across the entire frequency domain, the frequency domain ripple region will cover the entire or most of the test frequency band.

[0198] Using the actual ripple period as a reference, a symmetrical frequency band is extended in the frequency domain. The core operation is not to define a narrow new frequency band, but rather to label a broad frequency range (e.g., from f_start to f_stop) as the "frequency domain ripple region" and assign it a special property: the scan step size within this region must be strictly controlled by the "actual ripple period" Δf_actual(f). According to the sampling theorem, to reconstruct a signal with a period of Δf without distortion, the sampling frequency should be at least 2 / Δf, meaning the sampling interval (step size) should be less than Δf / 2. Therefore, the VNA uses a value less than Δf_actual(f) / k (where k is a coefficient greater than 2, such as 4 or 5, to ensure sufficient resolution) as the upper limit of the scan step size for this region.

[0199] Through this step, a feature originating from a time-domain physical defect is thoroughly and precisely transformed into a specific, executable frequency-domain scanning constraint.

[0200] Step 211: Using a quantum genetic algorithm, the corresponding optimized frequency step size and number of scanning sampling points are assigned to the high-sensitivity region, the flat region, and the ripple region in the frequency domain, respectively, to obtain the optimized scanning strategy.

[0201] In some embodiments, in order to achieve comprehensive and in-depth optimization of test speed, accuracy and signal-to-noise ratio, this embodiment further includes in this step: allocating corresponding optimized intermediate frequency bandwidth and optimized average number of times to the frequency domain high-sensitivity region, the frequency domain flat region and the frequency domain ripple region respectively.

[0202] In this embodiment, the optimization objective is not limited to frequency step size and number of scan points. The optimization dimension of the quantum genetic algorithm is expanded to simultaneously include the optimization of two key hardware measurement parameters of the vector network analyzer: intermediate frequency bandwidth (IFBW) and averaging factor. The intermediate frequency bandwidth directly affects the speed and noise floor of single-point measurements; a wide IFBW results in high speed but high noise, while a narrow IFBW has the opposite effect. The averaging factor effectively suppresses random noise by averaging multiple measurements, but it significantly increases the test time.

[0203] Therefore, the chromosomes of the quantum genetic algorithm in this embodiment are re-encoded to carry a complete, multi-dimensional scanning configuration scheme. This scanning configuration scheme assigns an optimal combination of three parameters to each frequency domain region (frequency domain high-sensitivity region, frequency domain flat region, frequency domain ripple region): optimized frequency step size, optimized intermediate frequency bandwidth, and optimized averaging count.

[0204] To adapt to the needs of multi-dimensional collaborative optimization, this embodiment also upgrades the fitness function of the quantum genetic algorithm to more comprehensively evaluate the overall performance of a complete scan configuration scheme that includes frequency step size, intermediate frequency bandwidth, and average number of scans. The upgraded fitness function consists of a mandatory "quality score" and a "time cost penalty". Its core design idea is to first ensure that the measurement quality is absolutely up to standard, and then select the scheme with the shortest total test time among all schemes that meet the quality requirements.

[0205] Specifically, the upgraded fitness function guides the optimization process through a two-part reward and penalty mechanism:

[0206] The first part is a hard constraint check on measurement quality. It checks whether the scheme decoded by the chromosome has configured sufficiently high measurement accuracy for the two key regions of high sensitivity and ripple. That is, whether the intermediate frequency bandwidth is lower than the preset high-precision bandwidth threshold and whether the average number of times is higher than the high signal-to-noise ratio number of times threshold. Only the scheme that fully meets these conditions can get a large basic reward score. Otherwise, it will be subject to a very large penalty and will be immediately eliminated by the algorithm during the evolution process.

[0207] The second part is the time cost quantification based on the instrument's physical model. It accurately calculates the estimated total time required to complete the entire scan based on the frequency step size, intermediate frequency bandwidth, and average number of scans allocated to each region in the scheme, and deducts this time as a penalty from the basic reward score.

[0208] By maximizing this upgraded fitness function, the quantum genetic algorithm can automatically find the optimal balance among the three mutually constraining dimensions of speed, accuracy, and signal-to-noise ratio, and finally select a globally optimal scanning strategy that can both ensure that the details of key signals are not distorted and shorten the testing cycle to the greatest extent.

[0209] In this embodiment, the chromosome of the quantum genetic algorithm is a sequence of qubits encoding a complete multidimensional scanning strategy. It structurally includes three core measurement parameters—frequency step size, intermediate frequency bandwidth, and average number of scans—set for three types of regions: high-sensitivity region, flat region, and ripple region. Each measurement parameter is represented by a set of qubits, allowing the entire chromosome to simultaneously explore all possible combinations of scanning parameters in a quantum superposition state. This efficiently finds the optimal scanning scheme—the one with the shortest global testing time—while ensuring measurement accuracy in key regions; i.e., optimizing the scanning strategy.

[0210] Step 212: Perform segmented scanning on the high-sensitivity region, flat region, and ripple region in the frequency domain according to the optimized scanning strategy to obtain real-time scanning data.

[0211] In some embodiments, this step may further include: when switching between the high-sensitivity region, the flat region, and the ripple region in the frequency domain, dynamically adjusting the frequency step size, intermediate frequency bandwidth, and average number of scans according to the assigned optimized scanning strategy.

[0212] Specifically, when performing a scan according to an optimized scanning strategy, the vector network analyzer (VNA) not only switches the frequency step size based on the region but also dynamically adjusts the instrument's hardware settings in real time. For example, when scanning into a frequency band marked as a flat region in the frequency domain, the VNA uses a larger frequency step size, a wider intermediate frequency bandwidth (e.g., 50kHz), and a lower averaging count (e.g., 1 count) to achieve extremely fast scanning. Conversely, when the scan pointer is about to enter a highly sensitive or rippled region in the frequency domain, the VNA switches to a very small frequency step size, a very narrow intermediate frequency bandwidth (e.g., 100Hz), and a higher averaging count (e.g., 16 counts), sacrificing local speed in exchange for the highest fidelity capture of sharp resonance peaks or weak ripple signals.

[0213] In this way, this embodiment improves adaptive scanning from one-dimensional frequency point planning to three-dimensional measurement parameter collaborative optimization, making the allocation of test resources (time, signal-to-noise ratio) more intelligent and refined, thereby more effectively solving the contradiction between speed and accuracy in complex test scenarios.

[0214] Step 213: Calculate the local fitting residuals for each region based on real-time scanning data.

[0215] This step can be referred to the description in the foregoing embodiments, and will not be repeated here.

[0216] Step 214: If any local fitting residual exceeds the confidence threshold, then the frequency domain interval where the local fitting residual exceeds the confidence threshold is determined as the target sub-interval where the scanning density needs to be increased.

[0217] This step is the diagnosis and localization stage of the adaptive correction mechanism. Its core VNA automatically identifies the precise frequency bands that are under-scanned based on data quality self-assessment.

[0218] In practice, the VNA systematically traverses all segmented scanning regions, comparing the calculated "local fitting residual" within each region with a system-preset "confidence threshold." This confidence threshold represents the upper limit of acceptable fitting error. Once the residual value in a certain frequency domain interval exceeds this threshold, it serves as a clear warning signal, indicating that the actual S-parameter curve behavior within that interval is far more complex than what the current sparse sampling points can depict; that is, there may be undetected sharp resonances or rapid fluctuations between two adjacent measurement points.

[0219] At this point, the VNA will immediately mark and define the frequency domain range defined by two adjacent measured frequency points with high residual values ​​as the "target sub-interval that requires increased scan density". The determination of this "target sub-interval" is equivalent to accurately locking the problem area, providing a clear target for the subsequent "precise point supplementation" operation, avoiding blindly rescanning the entire large area, thus ensuring the efficiency of the correction process.

[0220] Step 215: Determine the insertion density of supplementary frequency points based on the number of scanned frequency points and the rate of change of fitted curvature within the target sub-interval.

[0221] This step is the core of the adaptive correction strategy. Its purpose is to intelligently determine the intensity of the "remedy" based on the characteristics of the problem area, i.e., how many new measurement points to add. This process abandons the simple fixed-point addition strategy and instead adopts a more refined dynamic adjustment mechanism.

[0222] VNA comprehensively analyzes two key metrics. The first is the "number of scanned frequency points within the target sub-interval," reflecting the current baseline sampling density. The second is the "rate of change of fitted curvature," a more profound metric that measures not only the curvature of the curve but also the rate of change of that curvature. A high rate of change indicates extremely complex and irregular signal behavior, such as a narrow-band spike superimposed on the edge of another broadband resonance. VNA's algorithm inputs these two metrics into a decision model. For example, if an extremely high rate of change of fitted curvature is detected within an already sparse sub-interval, it determines the presence of a very fine feature and thus assigns a very high "insertion density of supplementary frequency points." Conversely, if only a moderate rate of change is detected within a relatively dense sub-interval, a lower insertion density is allocated. The final output "insertion density" is a quantitative instruction that precisely specifies the number of additional scan points needed within the target sub-interval, ensuring that the resources invested in correction are proportional to the severity of the problem.

[0223] In some embodiments, step 215 may further include the following steps:

[0224] S2151, count the total number of scanned frequency points within the target sub-interval.

[0225] This step is a basic data statistics operation performed within the VNA, the purpose of which is to provide crucial contextual information for subsequent decision-making.

[0226] In practice, once a frequency range is identified as a "target sub-range" due to excessive fitting residuals, the VNA immediately accesses its memory-stored list of real-time scan data. The VNA uses the start and end frequencies of the target sub-range as boundaries, searching and counting all frequency points within these boundaries (including boundary points). This count is the "total number of scanned frequency points." For example, a simple target sub-range might be defined by only two adjacent measured frequency points, in which case the total number is 2. This count itself does not directly assess the complexity of the signal, but it is crucial because it quantifies the current sampling density in the problem area. A lower count indicates that the initial probes in the area are very sparse, providing a key context for subsequently assessing the severity of high curvature: drastic changes between sparse sampling points are often more alarming than the same changes between dense sampling points, requiring more aggressive supplementary scans.

[0227] S2152, calculate the absolute value of the average fitted curvature of the scanned frequency points based on the total number.

[0228] This step aims to quantitatively assess the geometric severity of the S-parameter curve within the target sub-interval. The fitting curvature, calculated during the preceding residual calculation, is the second derivative of the locally fitted curve at each measured frequency point; it directly reflects the degree of curvature of the S-parameter curve at that point. A sharp resonance peak corresponds to a maximal curvature value.

[0229] In this step, the VNA first extracts the fitted curvature value corresponding to each scanned frequency point within the target sub-interval from its internal computation cache. Next, the processor takes the "absolute value" of each curvature value, because the decision-making process is concerned with the degree of curvature, not its direction (convex upwards or concave downwards).

[0230] Then, the absolute values ​​of all these curvatures are summed up, and the sum is divided by the "total number of scanned frequency points" counted in the first step. The final calculated result is the "average absolute value of the fitted curvature". This value becomes a centralized, standardized indicator that concisely summarizes whether the signal behavior of the target sub-interval is smooth or highly volatile at the current sampling density.

[0231] S2153, density level is determined based on the total quantity and the absolute value of the average fitted curvature.

[0232] This step is a crucial intelligent decision-making process. Its core involves comprehensively analyzing the inherent complexity of the signal and the current level of detection, thereby prioritizing the urgency and intensity of supplementary scanning. In a VNA, this process is accomplished through a pre-defined decision matrix or rule set, as detailed below:

[0233] The VNA uses the "total number of scanned frequency points" (representing the current sampling density) and the "absolute value of the average fitted curvature" (representing signal complexity) as two input dimensions. These two input values ​​are then compared to preset thresholds, and a discrete density level is output based on the combination interval they fall into. For example, the rule can be set as follows: if the total number of frequency points is "low" and the average curvature is "high," it is determined to be the highest density level (e.g., level 4); if the total number of frequency points is "high" but the average curvature is still "high," it may be determined to be the next lower level (e.g., level 3); if the total number of frequency points is "high" and the average curvature is "low," it may be determined to be the lowest density level (e.g., level 1).

[0234] This density level is a standardized, qualitative classification result that transforms a complex continuous numerical problem into an easily processed hierarchical instruction, clearly indicating the system's judgment on the degree of insufficient scanning of the current sub-interval.

[0235] S2154, determine the corresponding supplementary frequency insertion density according to the density level.

[0236] This step is the final output of the decision-making process. It transforms the relatively abstract "density level" determined in the previous step into a specific, executable quantitative scanning instruction.

[0237] The VNA system configuration includes a pre-stored supplementary frequency insertion density level mapping table. This table maps each possible density level to a specific supplementary frequency insertion density. The supplementary frequency insertion density is a specific numerical value that specifies how many measurement points need to be added within the target sub-interval. For example, the mapping relationship can be set as follows: density level 1 corresponds to inserting 1 supplementary frequency point (i.e., bisection); level 2 corresponds to inserting 3 supplementary frequency points (i.e., quartering); and the highest density level 4 may correspond to inserting 15 supplementary frequency points, or directly switch to rescanning within the sub-interval with a very small fixed step size.

[0238] The VNA will look up this mapping table based on the density level obtained in the previous step and obtain the final insertion density value. This value will directly serve as the basis for the next stage of supplementary scanning hardware execution, ensuring that the strength of the corrective measures is precisely matched with the previously assessed severity of the problem, and achieving efficient use of resources.

[0239] In this embodiment, the construction of the supplementary frequency point insertion density level mapping table can be based on a combination of expert experience, simulation analysis, and experimental verification. First, based on the type of device under test (e.g., antenna, filter) and its typical failure modes, RF engineers estimate the approximate sampling density required to recover the true curve under different signal complexities, and draft an initial correspondence between density levels and the number of supplementary points. Second, a large number of S-parameter curves containing resonant peaks of different widths and Q values ​​are generated as "virtual samples" using electromagnetic simulation software. The scanning and residual detection process of this invention is simulated on these samples, and different numbers of supplementary points are repeatedly tested. By comparing the degree of agreement between the supplemented curves and the original simulation curves, the optimal number of supplementary points that should match different density levels is systematically optimized and quantified. Finally, actual measurements are performed on a real vector network analyzer using standard parts or known defective parts, and the mapping table parameters are fine-tuned until they can achieve the fastest convergence speed and the highest testing efficiency while ensuring measurement accuracy.

[0240] Step 216: Insert supplementary frequency points in the target sub-interval according to the insertion density of the supplementary frequency points, and perform supplementary scanning based on the currently allocated frequency step size.

[0241] This step translates the preceding decisions into the execution of specific physical measurements. First, based on the "insertion density of supplementary frequencies" determined in the previous step, the VNA calculates the precise frequency values ​​of a series of new frequencies to be inserted within the locked "target sub-interval." A feasible strategy is to perform uniform interpolation within this sub-interval. For example, if it is determined that three supplementary frequencies need to be inserted, the frequency range of the sub-interval will be divided into four equal parts, and the frequencies to be measured will be generated at the three new division points.

[0242] Subsequently, these newly generated frequency points are compiled into a temporary, short scan list. Then, the VNA performs an efficient "supplementary scan." During this process, the VNA's agile frequency synthesizer no longer performs continuous scanning, but instead, according to the instructions of this temporary list, rapidly jumps to each supplementary frequency point and completes a full S-parameter measurement. Here, "performing a supplementary scan based on the currently assigned frequency step size" means that the measurement mechanism of the supplementary scan is consistent with the main scan, but its frequency points are dynamically determined by the insertion density, rather than following the original fixed step size.

[0243] The results of this supplementary scan are a new set of high-value measurement data, which are used to precisely fill in the gaps in the previous data, thereby ensuring that the final S-parameter curves can faithfully reproduce the true electrical characteristics of the device under test in this critical region.

[0244] Step 217: Based on real-time scan data and supplementary scan data, generate S-parameter curves of new energy vehicle components in the test frequency band.

[0245] This step can be referred to the description in the foregoing embodiments, and will not be repeated here.

[0246] The methods provided in the above embodiments can be executed by a vector network analyzer. The vector network analyzer in this embodiment of the invention is described below from a hardware processing perspective; please refer to [link to relevant documentation]. Figure 4 This is a schematic diagram of a physical device structure of a vector network analyzer in an embodiment of the present invention.

[0247] It should be noted that, Figure 4 The structure of the vector network analyzer shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0248] like Figure 4 As shown, the vector network analyzer includes a Central Processing Unit (CPU) 401, which can perform various appropriate actions and processes based on programs stored in Read-Only Memory (ROM) 402 or loaded from storage section 408 into Random Access Memory (RAM) 403, such as performing the methods described in the above embodiments. The RAM 403 also stores various programs and data required for system operation. The CPU 401, ROM 402, and RAM 403 are interconnected via bus 404. An Input / Output (I / O) interface 405 is also connected to bus 404.

[0249] The following components are connected to I / O interface 405: input section 406 including audio input devices, push-button switches, etc.; output section 407 including a liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 408 including a hard disk, etc.; and communication section 409 including a network interface card such as a LAN (Local Area Network) card, modem, etc. Communication section 409 performs communication processing via a network such as the Internet. Drive 410 is also connected to I / O interface 405 as needed. Removable media 411, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 410 as needed so that computer programs read from them can be installed into storage section 408 as needed.

[0250] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 409, and / or installed from removable medium 411. When the computer program is executed by central processing unit (CPU) 401, it performs the various functions defined in the present invention.

[0251] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0252] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, program segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0253] Specifically, the vector network analyzer of this embodiment includes a processor and a memory. The memory is coupled to one or more processors and is used to store computer program code, which includes computer instructions. One or more processors call the computer instructions to cause the vector network analyzer to execute the method provided in the above embodiment.

[0254] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the vector network analyzer described in the above embodiments; or it may exist independently and not assembled into the vector network analyzer. The storage medium carries one or more computer programs that, when executed by a processor of the vector network analyzer, cause the vector network analyzer to implement the methods provided in the above embodiments.

[0255] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

[0256] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. An adaptive vector network scanning method, characterized in that, include: Within the set test frequency band, a non-uniformly distributed sparse frequency point sequence is generated through chaotic mapping; Based on the sparse frequency point sequence, the S-parameter measurement data of the vector network of new energy vehicle components are detected; The time-domain features and frequency-domain features are calculated based on the S-parameter measurement data. The time-domain features include reflection peak features in the time-domain reflection profile, and the frequency-domain features include curvature features, information entropy features, and wavelet transform features. Based on the preset frequency domain feature mutation threshold and the frequency domain features, the frequency domain high-sensitivity region and the frequency domain flat region are divided in the frequency domain, and the time domain key region is determined in the time domain based on the reflection peak features in the time domain features. The time-domain key region is mapped back to the frequency domain to define the frequency-domain ripple region, which represents the frequency band where the S-parameters exhibit periodic, weak fluctuations over a wide bandwidth. This step specifically includes: extracting the characteristic time delay corresponding to each reflection peak from the time-domain key region; determining the theoretical ripple period in the frequency domain based on the characteristic time delay; performing frequency-dependent correction on the theoretical ripple period based on the dispersion characteristics of the signal transmission path to obtain the actual ripple period; and extending a symmetrical frequency band in the frequency domain based on the actual ripple period to define the frequency-domain ripple region. Using a quantum genetic algorithm, corresponding optimized frequency step size and number of scanning sampling points are assigned to the high-sensitivity region, the flat region, and the ripple region in the frequency domain, respectively, to obtain an optimized scanning strategy; According to the optimized scanning strategy, segmented scanning is performed on the high-sensitivity region, the flat region, and the ripple region in the frequency domain to obtain real-time scanning data; Calculate the local fitting residuals for each region based on the real-time scanning data; If any of the local fitting residuals exceeds the confidence threshold, supplementary frequency points are dynamically inserted in the corresponding frequency domain interval and supplementary scanning is performed to obtain supplementary scanning data; Based on the real-time scanning data and supplementary scanning data, the S-parameter curves of the new energy vehicle components in the test frequency band are generated.

2. The method according to claim 1, characterized in that, The frequency-dependent correction of the theoretical ripple period based on the dispersion characteristics of the signal transmission path is used to obtain the actual ripple period, specifically including: Obtain the medium parameters of the vector network transmission path in the new energy vehicle component from a preset material property database; Based on the aforementioned medium parameters, a phase velocity frequency dependence characteristic model of electromagnetic wave propagation in the vector network transmission path is constructed. The theoretical ripple period is frequency-variable compensated according to the phase velocity frequency dependence characteristic model to obtain the actual ripple period.

3. The method according to claim 1, characterized in that, Determining the theoretical ripple period in the frequency domain based on the characteristic time delay includes: Based on the aforementioned characteristic time delay and the physical mechanism of reflection, the two-way propagation time delay of electromagnetic waves in the signal transmission path is determined. The two-way propagation delay is converted into the periodic parameters of the corresponding frequency domain ripple to obtain the theoretical ripple period.

4. The method according to any one of claims 1-3, characterized in that, The step of dynamically inserting supplementary frequency points and performing supplementary scanning within the corresponding frequency domain interval specifically includes: The frequency domain intervals where the local fitting residuals exceed the confidence threshold are identified as target sub-intervals where the scanning density needs to be increased. Based on the number of scanned frequency points and the rate of change of fitted curvature within the target sub-interval, the insertion density of supplementary frequency points is determined; Supplementary frequency points are inserted in the target sub-interval according to the insertion density of the supplementary frequency points, and supplementary scanning is performed based on the currently allocated frequency step size.

5. The method according to claim 4, characterized in that, The step of determining the insertion density of supplementary frequency points based on the number of scanned frequency points and the rate of change of fitted curvature within the target sub-interval includes: Count the total number of scanned frequency points within the target sub-interval; Calculate the absolute value of the average fitted curvature of the scanned frequency points based on the total number; The density level is determined based on the total quantity and the absolute value of the average fitted curvature. The corresponding supplementary frequency insertion density is determined based on the density level.

6. The method according to claim 1, characterized in that, The generation of a non-uniformly distributed sparse frequency point sequence through chaotic mapping includes: Set the initial values ​​and nonlinear control parameters for the chaotic mapping; Based on the aforementioned nonlinear control parameters, multiple iterative calculations are performed to obtain a chaotic variable sequence; The chaotic variable sequence is mapped to the test frequency band to form a non-uniformly distributed sparse frequency point sequence.

7. A vector network analyzer, characterized in that, Includes one or more processors and memory; The memory is coupled to the one or more processors, the memory being used to store computer program code, the computer program code including computer instructions, the one or more processors invoking the computer instructions to cause the vector network analyzer to perform the method as described in any one of claims 1-6.

8. A computer-readable storage medium storing computer instructions, characterized in that, When the computer instructions are run on the vector network analyzer, the vector network analyzer performs the method as described in any one of claims 1-6.

9. A computer program product, characterized in that, When the computer program product is run on a vector network analyzer, the vector network analyzer performs the method as described in any one of claims 1-6.

Images