A Vector Signal Analyzer Test Data Intelligent Processing System
By using a dimensionality reduction module and local density index processing to process the complex baseband signal of the vector signal analyzer, the problem of distinguishing and compensating for interference points in high-order modulation signals is solved, achieving high-precision signal equalization and improved anti-interference capability.
Patent Information
- Application Number
- CN202611368369.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-04
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies struggle to accurately distinguish normal high-energy modulation symbols from abnormal interference points in high-order modulation signal testing environments with occasional transient electromagnetic pulse interference, and also struggle to perform robust and high-precision equalization compensation processing on contaminated complex baseband signals.
A dimensionality reduction module is used to calculate the multi-loop error value. Combined with the local density index and iterative weighted least squares method, the complex baseband signal and background noise power are obtained through a vector signal analyzer to generate a constellation diagram and error vector amplitude, thereby realizing digital linear filtering.
It effectively overcomes the multi-loop energy difference caused by instantaneous amplitude changes in high-order quadrature amplitude modulation signals, accurately locks abnormal interference characteristics, and improves the anti-interference capability of the test system and the reliability of test data.
Smart Images

Figure CN123123082A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data processing technology, and in particular to an intelligent processing system for test data from a vector signal analyzer. Background Technology
[0002] In the field of modern electronic communication and electromagnetic compatibility testing, vector signal analyzers are the core testing equipment for evaluating the quality of radio frequency and microwave signals. In the research and development and production of high-speed communication equipment, radio frequency signals often face the influence of complex electromagnetic environments. For example, occasional transient electromagnetic pulse interference can cause severe amplitude and phase distortion of the received signal. In order to accurately evaluate the true transmission performance of the device under test, it is necessary to rely on vector signal analyzers to perform high-precision data processing and channel compensation on the received baseband signal, thereby extracting the true error vector amplitude and constellation diagram characteristics. This is irreplaceable for ensuring the reliability of the communication system. In response to the electromagnetic interference problem in the test environment, the invention patent with publication number CN121703539A discloses a high-precision EMI diagnosis and pre-test system based on a vector signal analyzer. This scheme determines the superposition risk type and interference signal source combination in the test target by analyzing the phase characteristics and frequency domain characteristics of the electromagnetic interference signal in the vector signal analyzer. It also formulates corresponding environmental interference identification strategies by utilizing the deviation of vector signal characteristics between different test targets, and constructs a diagnostic test method accordingly. However, the aforementioned existing technologies primarily focus on macroscopic-level risk classification and identification strategies for interference signal sources, failing to delve into the microscopic level of baseband modulation symbols for data feature and distortion analysis. When faced with sudden, strong interference such as transient electromagnetic pulses, these existing technologies cannot effectively overcome the inherent multi-ring energy differences in high-order orthogonal amplitude modulation signals due to instantaneous amplitude variations. They struggle to distinguish between normal high-energy modulation symbols in the outermost ring and anomalous interference points superimposed with transient pulses on the complex baseband complex plane. Furthermore, this approach lacks a quantitative evaluation mechanism for the local data distribution density at the single-symbol level, failing to provide a low-level baseband signal equalization and compensation algorithm that maintains numerical solution stability even in the presence of extreme anomalous sample interference. Therefore, when the test signal is contaminated by sporadic, sparse transient electromagnetic pulses, the aforementioned systems cannot effectively compensate for the damaged baseband signal and restore the constellation diagram through high-precision error suppression and digital linear filtering. Summary of the Invention
[0003] The technical problem solved by this invention is that existing technologies have difficulty in accurately distinguishing normal high-energy modulation symbols from abnormal interference points in high-order modulation signal test environments with occasional transient electromagnetic pulse interference, and have difficulty in performing robust and high-precision equalization compensation processing on contaminated complex baseband signals.
[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an intelligent processing system for test data of a vector signal analyzer, comprising: The acquisition module is used to acquire the complex baseband signal of the device under test and the background noise power of the test environment; The dimension reduction module is used to calculate the multi-loop error value of each modulation symbol in the complex baseband signal based on the modulation order of the modulation signal; The local density calculation module is used to calculate the local density index of each modulation symbol based on the background noise power and the multi-loop error value. The solution module is used to solve for the optimal tap coefficient vector based on the complex baseband signal, the local density index, and the reference modulation symbol sequence by using an iterative weighted least squares method. The output module is used to perform digital linear filtering on the complex baseband signal using the optimal tap coefficient vector to obtain the compensated complex baseband signal and generate a constellation diagram and error vector amplitude.
[0005] Preferably, the process of acquiring the complex baseband signal in the acquisition module specifically includes: The modulated radio frequency signal emitted by the device under test is processed to obtain a complex baseband signal; The processing includes: downconversion, analog-to-digital conversion, digital downconversion, matched filtering, timing synchronization, decimation processing, and normalization processing.
[0006] Preferably, the process of acquiring the background noise power in the acquisition module specifically includes: Extract the background sampling sequence before the arrival of the modulated radio frequency signal; The average power of the background sampling sequence is calculated as the normalized background noise power.
[0007] Preferably, the process of obtaining the multi-loop error value in the dimensionality reduction module specifically includes: Based on the modulation order of the modulated radio frequency signal, generate an ideal constellation diagram coordinate set; Based on the ideal constellation coordinate set, calculate the normalized radius values of several reference amplitude rings; For each element in the complex baseband signal, calculate its measured complex amplitude; The geometric distance between the measured complex amplitude and the normalized radius of each reference amplitude loop is calculated as the radial deviation.
[0008] Preferably, the process of obtaining the multi-loop error value in the dimensionality reduction module further includes: Based on the radial deviation, the membership degree of each element to the reference ring of each reference amplitude ring is calculated using the fuzzy membership function; Using the membership degree of the reference ring as a weighting coefficient, the squared values of the radial deviations are weighted and accumulated to calculate the multi-ring error value.
[0009] Preferably, the process of obtaining the local density index in the local density calculation module specifically includes: For each modulation symbol, calculate the squared distance between its multi-loop error value and the multi-loop error values of all other modulation symbols in the sequence; Using the background noise power as the variance control parameter of the kernel function, a Gaussian kernel function transformation is performed on the squared distance to obtain a similarity measure between each modulation symbol; The local density index of the current modulation symbol is obtained by summing and normalizing the similarity measures of all other modulation symbols to the current modulation symbol.
[0010] Preferably, the process of solving the optimal tap coefficient vector in the solution module specifically includes: The complex baseband signals are traversed by a sliding window of a preset length, and the center-aligned data slices are extracted as feature vectors. The missing data outside the window boundary is filled with zero values. A reference modulation symbol sequence time-aligned with the complex baseband signal is constructed, the reference modulation symbol sequence being divided into a training region sequence and a tracking region sequence; The training region sequence is filled with a known preceding training sequence; The elements in the tracking region sequence are dynamically updated in each iteration.
[0011] Preferably, the process of solving the optimal tap coefficient vector in the solution module further includes: The feature vector is linearly filtered using the tap coefficient vector of the current iteration period to obtain the estimated modulation symbol, and the residual vector between the estimated modulation symbol and the corresponding elements in the reference modulation symbol sequence is calculated. Based on the magnitude of the residual vector, the basic weight of each modulation symbol is calculated using a piecewise weighting function. Modulation symbols with residuals less than a preset insensitive threshold are assigned zero weight, modulation symbols with residuals in the transition region are assigned gradual weight, and modulation symbols with large residuals are assigned amplitude limiting weight.
[0012] Preferably, the process of solving the optimal tap coefficient vector in the solution module further includes: The local density exponent is converted into a penalty factor using an exponential mapping function; The penalty factor is multiplied by the corresponding basic weight to obtain the final weight of each modulation symbol, and a diagonal weight matrix is constructed accordingly. A weighted least squares objective function is constructed based on the diagonal weight matrix, and a system of normal equations is constructed by combining it with regularization terms; Solve the system of normal equations to obtain the optimal tap coefficient vector.
[0013] Preferably, the output module is used to perform digital linear filtering on the complex baseband signal using the optimal tap coefficient vector to obtain the compensated complex baseband signal; Based on the distribution of each modulation symbol in the compensated complex baseband signal on the I / Q complex plane, draw a constellation diagram; The compensationd modulation symbols are compared with the nearest ideal constellation point in the ideal constellation diagram coordinate set, and the error vector amplitude is calculated.
[0014] The beneficial effects of this invention are as follows: By introducing a dimension reduction module to calculate the multi-ring error value, this invention effectively overcomes the inherent multi-ring energy difference defect of high-order orthogonal amplitude modulation signals due to the continuous change of instantaneous amplitude. By mapping the distribution of complex baseband signals on the two-dimensional complex plane to one-dimensional multi-ring error values, the error value of normal modulation symbols, whether originally located in the inner or outer ring, is significantly reduced to an order of magnitude comparable to the background noise power. Meanwhile, abnormal interference points superimposed with transient pulses retain their abnormally deviated original energy levels, thus increasing them significantly. This invention constructs a Gaussian kernel local density exponent and penalty factor based on the background noise power through a local density calculation module and a nonlinear exponential mapping mechanism. This can quantify the sparseness and isolation of the distribution of modulation symbols in the multi-ring error value space, thereby further widening the weight difference between normal modulation symbols and interference modulation symbols. Attached Figure Description
[0015] Figure 1 This is a basic flowchart of a vector signal analyzer test data intelligent processing system provided in one embodiment of the present invention. Detailed Implementation
[0016] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0017] Example, refer to Figure 1 A vector signal analyzer test data intelligent processing system is provided, comprising: The acquisition module is used to acquire the complex baseband signal of the device under test and the background noise power of the test environment; The dimension reduction module is used to calculate the multi-loop error value of each modulation symbol in the complex baseband signal based on the modulation order of the modulation signal; The local density calculation module is used to calculate the local density index of each modulation symbol based on the background noise power and the multi-loop error value. The solution module is used to solve for the optimal tap coefficient vector based on complex baseband signals, local density exponents, and reference modulation symbol sequences using an iterative weighted least squares method. The output module is used to perform digital linear filtering on the complex baseband signal using the optimal tap coefficient vector to obtain the compensated complex baseband signal and generate a constellation diagram and error vector amplitude.
[0018] This invention introduces a dimension reduction module to calculate the multi-loop error value of the modulation symbol and combines the local density exponent and iterative weighted least squares method for solution. This overcomes the limitations of traditional baseband signal processing in dealing with high-order modulation and transient interference, and realizes the reconstruction and feature extraction of complex baseband signals in the error space. It can accurately lock abnormal interference features in complex noise environments, thus providing a high-precision compensation basis for subsequent digital linear filtering and constellation diagram generation, effectively improving the overall anti-interference capability of the test system and the reliability of test data.
[0019] The process of acquiring complex baseband signals in the acquisition module specifically includes: The modulated radio frequency signal emitted by the device under test is processed to obtain a complex baseband signal; The processing includes: downconversion, analog-to-digital conversion, digital downconversion, matched filtering, timing synchronization, decimation processing, and normalization processing.
[0020] In a microwave anechoic chamber, a vector signal analyzer collects modulated radio frequency signals, such as 64QAM or 256QAM signals, emitted by the device under test through a receiving antenna. The modulated radio frequency signal emitted by the device under test is acquired by a vector signal analyzer and processed into a complex baseband signal. Each element For the first Within one modulation symbol period, the modulation symbols received by the vector signal analyzer For a complex number, its in-phase component is: Orthogonal components are ,Right now subscript The time sequence number of the modulation symbol. ; In this embodiment, the modulated radio frequency signal transmitted by the device under test is a short burst high-order quadrature amplitude modulated radio frequency signal, such as a 256QAM signal; The processing includes: The received modulated RF signal is down-converted and analog-to-digital converted to generate a digital intermediate frequency (IF) signal. This IF signal is then digitally down-converted and matched-filtered to generate a complex baseband sampling stream. The complex baseband sampling stream is then timed, synchronized, and decimated to extract the signal. Each received modulation symbol, for Each received modulation symbol is normalized to eliminate its physical dimensions, and its overall average power is normalized to 1. The normalized received modulation symbols constitute 3D complex column vectors as complex baseband signals In this embodiment, the normalized received modulation symbol All subsequent symbols will be referred to as modulation symbols. In this embodiment, the modulation baud rate is set to 12 Msps. In the configuration of the vector signal analyzer in this embodiment, the automatic gain control (AGC) function of the vector signal analyzer is set to burst lock-in mode or slow response mode to prevent occasional transient electromagnetic pulses from triggering gain readjustment, thereby ensuring that the receiving gain remains constant throughout the short burst acquisition period.
[0021] The process of acquiring background noise power in the acquisition module specifically includes: Extract the background sampling sequence before the arrival of the modulated radio frequency signal; Calculate the average power of the background sampling sequence as the normalized background noise power.
[0022] A background sampling sequence with no signal before the arrival of the modulated radio frequency signal is extracted using a vector signal analyzer. The length of the background sampling sequence is [missing information]. ,constitute 3D complex column vector In this embodiment ; The background sampling sequence z is scaled proportionally using the same scaling factor obtained when normalizing the average power of the received modulation symbols; Calculate the normalized background noise power of the current test environment using the background sampling sequence. Background noise power The mathematical expression is: ; in, Indicates the background sampling sequence number 1 sampling points The complex modulus value, since it has been normalized, Dimensionless.
[0023] The process of obtaining multi-loop error values in the dimensionality reduction module specifically includes: Based on the modulation order of the modulated radio frequency signal, generate an ideal constellation diagram coordinate set; Because the instantaneous amplitude of the 256QAM high-order quadrature amplitude modulation signal changes continuously, the sensing results of the signal energy of multiple control loops in the vector signal analyzer are different, resulting in uncoordinated control of each loop. This manifests as multiple concentric rings with significant differences in energy levels on the ideal constellation diagram. This inherent multi-ring energy difference poses a problem for distinguishing the normal high-energy signal point located in the outermost ring from the abnormal interference point superimposed with transient pulses in the subsequent steps of this invention. This step maps the distribution of each element in the complex baseband signal on the two-dimensional complex plane to a one-dimensional multi-loop error value. This significantly reduces the multi-loop error value in the interference-free state to an order of magnitude comparable to the background noise power, regardless of whether the element was originally located in the inner or outer loop. However, since the transient electromagnetic pulse interference point does not belong to any known amplitude loop, its multi-loop error value will increase significantly. Therefore, the multi-loop error value can be used to distinguish between the normal high-energy signal point located in the outermost loop and the abnormal interference point superimposed with the transient pulse. The process of calculating multi-loop error values specifically includes: Based on the quadrature amplitude modulation order of the modulated radio frequency signal acquired by the vector signal analyzer, pre-generate... The radius value of each reference amplitude ring; This embodiment takes 256QAM as an example. Before normalization, the I and Q coordinates of the baseband complex plane are ±1, ±3, ..., ±15. The I axis and Q axis each contain 16 levels. The two are combined to form a total of 256 ideal constellation points of 16×16, which constitute the coordinate set of the ideal constellation diagram. The ideal constellation coordinate set is a complete set of standard constellation coordinates specified by the communication protocol. It is pre-stored in the vector signal analyzer. When the tester sets the modulation format of the signal under test, such as 256QAM, the coordinates of all legal constellation points corresponding to that modulation format are retrieved from the memory. Taking 256QAM as an example, under the condition that the average signal power is normalized to 1, its ideal constellation diagram contains 256 defined complex coordinate values. These coordinate values are defined by the communication standard and constitute the ideal constellation coordinate set. An ideal constellation point corresponds to a geometric coordinate on the I / Q complex plane. It is also a complex number. ; The normalized radius values of several reference amplitude rings are calculated based on the coordinate set of the ideal constellation diagram; Calculate the complex plane magnitude corresponding to each ideal constellation point sequentially, and remove duplicates from all magnitudes. Since the average power of the ideal constellation set has been normalized to 1, its global root mean square amplitude is 1. Divide each original radius after deduplication by this global root mean square amplitude to obtain the constant radius term after average power normalization. ; The standard square 256QAM has 32 unique normalized radii, corresponding to 32 amplitude loops, i.e. The C normalized radii are arranged in a preset order to form a C-dimensional real row vector. , of which elements This represents the normalized radius corresponding to the c-th ideal amplitude loop; For each element in the complex baseband signal, calculate its measured complex amplitude; The mathematical expression for the measured amplitude is: ; in, For the first One modulation symbol, For the first The in-phase component of each modulation symbol, For the first The quadrature components of each modulation symbol; Calculate the geometric distance between the measured complex amplitude and the normalized radius of each reference amplitude loop, and use it as the radial deviation.
[0024] The mathematical expression for radial deviation is: ; in, For the first The modulation symbol is the first Radial deviation of each reference ring, For the first One modulation symbol, For the first The normalized radius of each reference amplitude loop; Here, , The radial deviation is a dimensionless, non-negative real number. The amplitude distance of the first The geometric deviation of the radius of an ideal ring is calculated by taking the absolute value. This operation only extracts the geometric distance in the amplitude dimension, thus eliminating interference from the phase information of the modulation symbol.
[0025] This invention utilizes the standard constellation diagram coordinates of a high-order orthogonal amplitude modulation signal to generate the normalized radius of the reference amplitude loop, and calculates the geometric radial deviation between the measured complex amplitude and each reference amplitude loop. This calculation method eliminates the interference of modulation symbol phase information and extracts only the pure amplitude dimension geometric distance, laying a solid geometric foundation for subsequently mapping the complex and variable amplitude distribution in high-order modulation signals into a one-dimensional error index.
[0026] The process of obtaining multi-loop error values in the dimensionality reduction module also includes: Based on the radial deviation, the fuzzy membership function is used to calculate the reference ring membership degree of each element to each reference amplitude ring. The reference ring membership degree is used to characterize the degree to which an element belongs to the corresponding reference amplitude ring. Reference ring membership degree is used for quantization Topologically belonging to the first The degree of each reference ring; The mathematical expression for the membership degree of the reference ring is: ; in, For the first The modulation symbol pairs the first Membership degree of each reference ring. The total number of reference amplitude loops, The summation index for traversing all cycles, For the first The modulation symbol is the first Radial deviation of each reference ring, For the first The modulation symbol is the first Radial deviation of each reference ring, For the preset smallest positive number, The fuzzy weighted index is set to 2 in this embodiment; To prevent the denominator from being zero, a very small positive constant is introduced. Since the signal amplitude has been normalized, the reference radius... The order of magnitude is 1. Take as ; in, The summation index for traversing all cycles; The fuzzy weighted index is a preset real number greater than 1; The membership degree of the reference loop is used as a weighting coefficient to accumulate the squared values of the radial deviations, and the multi-loop error value is calculated.
[0027] Using the membership degree of the reference ring and radial deviation Calculate multi-loop error values The mathematical expression for the multi-loop error value is: ; in, For the first Multi-ring error value of each modulation symbol For the index of the reference amplitude loop, The total number of reference amplitude loops, For the first The modulation symbol pairs the first Membership degree of each reference ring. For the first The modulation symbol to the first Radial deviation of each reference ring, For the first The nearest amplitude ring index to which the modulation symbol belongs; For amplitude ring The normalized radius value; The mathematical expression is: ; in, For the first The nearest amplitude ring index to which the modulation symbol belongs; For the first The modulation symbol to the first Radial deviation of each reference amplitude loop, This represents the total number of reference amplitude loops; To ensure the unbiasedness of the error value in terms of the physical amplitude, this invention uses a first-order reference loop membership degree. As weights, the squared radial deviations of each ring are weighted and summed by expectation. For a normal quadrature amplitude modulation symbol, the modulation symbol must closely adhere to a certain reference ring, and its relationship with that reference ring is... Close to 0 and Close to 1, while with other rings The result is close to 0, therefore the calculated value is... It is significantly suppressed, and its mean and floor noise power are similar. They are on the same order of magnitude; For modulation symbols affected by transient electromagnetic pulse interference, the measured amplitude is far from all reference loops, and each... The values are all relatively large and the confidence levels are dispersed, leading to It fully preserved its abnormally deviated original energy level, which was several orders of magnitude higher than normal. Although multi-ring error value First-order weighting is used, but the membership degree of the reference ring is... The degree of fuzziness is still determined by the fuzziness weighting index. Control, therefore The value of will still affect the multi-loop error values of normal modulation symbols and interference modulation symbols. The optimal distribution of magnitude needs to be determined through experiments. value; In this experiment, the environmental sensitivity constant Intensity of punishment The simulation environment was a MATLAB platform. The number of received modulation symbols was 1200. A random transient pulse with a pulse width of 20 nanoseconds and an envelope amplitude uniformly distributed between 0 and 5 times the signal root mean square value was injected at a random position in each frame. This pulse contaminated only one modulation symbol. The average signal power was normalized to 1. The measured background noise power was... ; fixed , Change the fuzzy weighted index The method is run to obtain the penalty factor for each modulation symbol. Based on the penalty factor, each modulation symbol is marked as an interference point or a normal point. The marking results are compared with the recorded true categories to determine the modulation symbols that are correctly classified and those that are misclassified. In the set of correctly classified normal modulation symbols, the mean of their multi-loop error values is calculated and denoted as the normal point. The mean value is calculated by averaging the multi-loop error values of correctly classified pulse contamination modulation symbols, and is denoted as the contamination point. Mean, calculate the discrimination index, discrimination index is the number of contamination points Mean and normal point The ratios of the means are shown in Table 1. Table 1
[0028] As can be seen from Table 1, when At that time, the discrimination reaches its maximum value. This is normal. The mean remains at The low level, with background noise power They are at the same level, but the pollution point The mean value is as high as 23.5, indicating the best separation between the two; when At this time, minute physical noise is easily amplified exponentially, causing the scalar base of the normal point error to be abnormally raised, resulting in a decrease in discrimination; when At this time, fuzzy transition occurs, and the membership degree, due to the averaging effect, causes the weighting of the normal modulation symbol to each ring to become diffuse, resulting in multi-ring error values of the normal modulation symbol. The index rebounded, but the distinguishability actually declined. Therefore, this embodiment will use a fuzzy weighted index. The preferred version is 2.0.
[0029] This invention introduces a fuzzy membership function to quantify the topological membership of signal points, and uses this as a weighting coefficient to accumulate the squared radial deviation, ultimately calculating the multi-ring error value. The dimension reduction mapping mechanism of this invention greatly reduces the error value of normal high-energy signal points in the outermost ring to the level of the noise floor due to their high membership, while unconstrained transient pulse interference points retain extremely high error levels because they are outside all reference rings. This solves the technical problem of the difficulty in distinguishing between the inherent peaks of high-order modulation signals and transient abnormal pulses.
[0030] The process of obtaining the local density index in the local density calculation module specifically includes: For each modulation symbol, calculate the squared distance between its multi-loop error value and the multi-loop error values of all other modulation symbols in the sequence; Using the background noise power as the variance control parameter of the kernel function, a Gaussian kernel function transformation is performed on the squared distance to obtain a similarity measure between each modulation symbol; The similarity metrics of all other modulation symbols to the current modulation symbol are accumulated and normalized to obtain the local density index of the current modulation symbol. The local density index is used to characterize the sparsity of the distribution of modulation symbols in the error space.
[0031] After obtaining the background noise power and the multi-loop error value calculated above, for each modulation symbol, calculate the squared distance between its multi-loop error value and the multi-loop error values of all other modulation symbols in the sequence; Based on background noise power and multi-loop error value sequence The local density index is calculated, which characterizes the degree of isolation of a modulation symbol. The mathematical expression for the local density index is: ; in, For the first The local density index of each modulation symbol, The total number of modulation symbols, For traversal Sequence excluding the current modulation symbol Indexes of all other modulation symbols, For the first Multi-ring error value of each modulation symbol For the first Multi-ring error value of each modulation symbol The preset environmental sensitivity constant is a dimensionless constant; in this embodiment, it is set to 3.0. Background noise power; Normalization factor This is to make the density index The numerical magnitude is freed from the influence of the sample size N, limiting it to... Within the interval, this ensures that the input domain of the subsequent penalty factor mapping equation remains in a stable interval, preventing numerical overflow or passivation of the exponential mapping; The mathematical expression for the local density index is represented by the current modulation symbol. Centered on the symbol, calculate its relationship with all other modulation symbols. The square distance, and using As a variance control quantity for the Gaussian kernel; For normal modulation symbol points, its Smaller, and compared to most modulation symbols Values are similar, the squared distance is within the noise margin, the exponent term is close to 1, sum and divide by back, It will remain at a relatively high value close to 1; When the modulation symbol is contaminated by a transient electromagnetic pulse, its The anomaly is large, and the squared distance from the majority of points in the main group far exceeds the noise scale; the exponential term rapidly decays to 0. Approaching 0; The local density indices of all modulation symbols form a density index sequence. .
[0032] In the sporadic, sparse transient electromagnetic pulse interference scenario targeted by this invention, the calculation of this local density can accurately identify isolated abnormal modulation symbols. Regarding environmental sensitivity constant The value of was determined experimentally, in relation to the fuzzy weighted index. Under the same simulation conditions as the experiment, the fuzzy weighting index is fixed. Intensity of punishment Change the environmental sensitivity constant The values were taken, and the proportion of normal points that were misjudged as interference, the proportion of real pulses that were missed, and the final measured error vector amplitude value were statistically analyzed. The results are shown in Table 2. The error vector amplitude is related to the degree of vector deviation between the actual received modulation symbol and the ideal reference constellation point. The normal point false positive rate represents the proportion of all known undisturbed normal modulation symbols that are incorrectly marked as interference points. The lower the proportion, the better the protection of normal signals. The pulse miss rate represents the proportion of all known modulation symbols contaminated by transient electromagnetic pulses that are incorrectly marked as normal points. The lower the proportion, the more sensitive the interference is to be detected. Table 2
[0033] As can be seen from Table 2, At this point, the false positive rate at normal points drops to 0.1%, while the pulse false negative rate is 0%, and the measured error vector amplitude value is as low as 0.51%. If the value is too low, a large amount of normal noise will be mistakenly suppressed. If the density is too large, the nuclear density assessment will be sluggish, leading to missed pulse detections. Therefore, this embodiment will... Set to 3.0.
[0034] This invention utilizes background noise power as the variance control parameter of the Gaussian kernel function to calculate the local density exponent of the modulation symbol in the multi-loop error value space. This can capture the local clustering state and isolation degree of error features. Since normal signals are highly dense in the error space, while signals contaminated by electromagnetic pulses are extremely free and isolated, this local density exponent directly quantifies this topological difference into a smooth exponent that approximates zero or one, thereby amplifying and identifying the characteristics of occasional and sparse transient pulse interference.
[0035] The process of finding the optimal tap coefficient vector in the solution module specifically includes: The complex baseband signal is traversed by a sliding window of preset length, and the center-aligned data slices are extracted as feature vectors. The missing data outside the window boundary is filled with zero values. Obtaining complex baseband signals and density index sequence Subsequently, this invention is based on complex baseband signals. and density index sequence In the presence of transient pulse interference, the optimal tap coefficient vector of the baseband channel compensation equalizer is solved. In this embodiment, the physical structure of the equalizer is composed of It consists of a tap. For odd numbers, this embodiment sets Each tap stores a specific complex coefficient. The vector formed by the sequential arrangement of the complex coefficients is the tap coefficient vector to be solved. ; For the equalizer, the center alignment offset is In this embodiment ; Regarding the first modulation symbols Extraction length is The sliding data window constitutes the feature vector. For index values that are out of bounds, i.e. Modulation symbols with subscripts less than or equal to 0 or greater than N are padded with complex zero values, and the feature vectors are... The center tap is aligned with the complex baseband signal. The first in element Since the sliding data window has achieved center tap alignment, the filter output... In physical time, it directly corresponds to the first... The modulation symbol at each moment; A reference modulation symbol sequence time-aligned with the complex baseband signal is constructed. The reference modulation symbol sequence is divided into a training region sequence and a tracking region sequence. The training sequence is filled with a known preceding training sequence; Construction and complex baseband signals Reference modulation symbol sequences of the same length Refer to each element in the modulation symbol sequence For the first The ideal modulation symbol that the device under test should transmit at a given time. , physical time Alignment; In this implementation, the elements in the reference modulation symbol sequence The values are segmented and filled. Specifically, in this embodiment, the physical frame structure transmitted by the device under test is assumed to conform to the standard protocol, and its header contains a known preamble training sequence. The length of the preamble training sequence is... Each modulation symbol in the preceding training sequence is known; In the reference modulation symbol sequence, arrive As a training region sequence, arrive Directly fill in the values from the known preceding training sequence; The elements in the tracking sequence are dynamically updated in each iteration.
[0036] In the reference modulation symbol sequence, the remaining Each element in the tracking region sequence is dynamically updated in each iteration. The dynamic update process specifically includes: Before the iterative update begins, this embodiment initializes the tap coefficient vector to... , The One element is 1, and the rest are 0; pass For eigenvectors Perform filtering operations to obtain , For each element of the tracking region sequence Find the distance from the ideal constellation coordinate set The nearest point, take the coordinates of this point as The initial guess value is used to fill each element of the tracking area sequence based on the ideal constellation map coordinate set; pass For eigenvectors After filtering, the result is It is a complex number that corresponds to a point in the I / Q complex plane, and iterates through the set of coordinates of the ideal constellation diagram. Calculate each of the 256 ideal constellation points in the data. Calculate the Euclidean distances to all ideal constellation points, and select the complex value of the ideal constellation point with the shortest distance. As .
[0037] This invention constructs a reference modulation symbol sequence mechanism with training area sequences and tracking area sequences, and introduces zero-value padding in the filtering operation to ensure the center alignment of the taps. This allows the use of known preamble training sequences to provide an absolutely reliable initial reference. At the same time, during the tracking phase, the reference guess value is dynamically updated using the optimal constellation point distance mapping. This not only ensures the correctness of the direction of convergence of the initial filter coefficients, but also endows the system with a powerful dynamic correction capability to continuously track and adaptively correct the channel state in the blind demodulation interval.
[0038] The process of finding the optimal tap coefficient vector in the solution module also includes: The feature vector is linearly filtered using the tap coefficient vector of the current iteration period to obtain the estimated modulation symbol, and the residual vector between the estimated modulation symbol and the corresponding elements in the reference modulation symbol sequence is calculated. After obtaining the complete reference modulation symbol sequence, the tap coefficient vector iterative optimization process begins; pass For eigenvectors Filtering is performed to obtain the estimated modulation symbol. ; Calculate the residual vector. The mathematical expression for the elements of the residual vector is: ; in, For the first The residual value of the modulation symbol, i.e., the value of the th modulated symbol in the residual vector. One element, For the first One reference modulation symbol, For the first The estimated modulation symbol corresponding to each eigenvector; Based on the magnitude of the residual vector, the basic weight of each modulation symbol is calculated using a piecewise weighting function. Modulation symbols with residuals less than a preset insensitive threshold are assigned zero weight, modulation symbols with residuals in the transition region are assigned gradual weight, and modulation symbols with large residuals are assigned amplitude limiting weight.
[0039] Assign basic weights to each modulation symbol, and assign basic weights to modulation symbols located in the training region sequence. For modulation symbols located in the tracking region sequence, the basic weights are calculated piecewise using the Huber weight function based on the magnitude of the residuals. ; Set an insensitive threshold With Huber weight function threshold Due to background noise power The normalized noise power is its arithmetic square root. This is the root mean square standard deviation of the physical noise floor. To match the magnitude dimensions of the residual vector, a Gaussian noise distribution is used. The criterion is to dynamically and adaptively set the Huber weight function threshold to... To ensure that residuals caused by normal noise are not excessively penalized; at the same time, set As an insensitive threshold, The settings are intended to filter out minute residual fluctuations caused by thermal noise; The base weights are calculated piecewise using the Huber weight function. The process specifically includes; when hour, That is, the weight contribution of the modulation symbol to this iteration is zero, and it does not participate in the update of the tap coefficient vector; when hour, ,when hour, ; All basic weights It constitutes 3D real vector ,all All are non-negative real numbers, representing the contribution of the corresponding modulation symbol in solving the tap coefficient vector in the current iteration step.
[0040] This invention introduces an adaptive Huber piecewise weighting function to calculate the basic weights of modulation symbols, directly linking both the insensitivity threshold and the weighting function threshold to the background noise power. This dynamic threshold setting based on the Gaussian noise distribution criterion effectively filters out small residual fluctuations caused purely by thermal noise. At the same time, it assigns gradual weights to signals with residuals in the transition region and amplitude-limiting weights to signals with extremely large residuals, ensuring that the weight allocation can both accommodate normal thermal noise and initially weaken the influence of distorted signals on the optimization direction.
[0041] The process of finding the optimal tap coefficient vector in the solution module also includes: The local density exponent is converted into a penalty factor using an exponential mapping function. Local density index Characterized the first The degree of isolation of a modulation symbol in the multi-loop error value space: For normal modulation symbols, their multi-loop error values are small and similar to the error values of other normal modulation symbols. Therefore, their local density index remains at a relatively high value close to 1. For interference modulation symbols contaminated by transient electromagnetic pulses, their multi-loop error values are abnormally large and extremely far away from other modulation symbols. Therefore, their local density index approaches 0. If the local density index is directly used as the final weight of the modulation symbol when solving the tap coefficient vector, the weight of the interference modulation symbol is very low, but it still retains a small non-zero value. For transient pulse interference with extremely high amplitude and energy, even if its weight is low, the huge residual of the interference point will still produce a non-negligible traction force in the subsequent weighted least squares normal equation, causing the solved tap coefficient vector to deviate from the optimal direction. Therefore, this embodiment introduces a penalty factor. Nonlinear exponential mapping is performed on the local density index to further widen the weight difference between normal modulation symbols and interference modulation symbols, and force the weight of interference points to be compressed to near zero. The mathematical expression for the penalty factor is: ; in, For the first The penalty factor for each modulation symbol, To determine the severity of the punishment, For the first The local density index of each modulation symbol; The calculation of the penalty factor locally maps the density exponent to The factor of the interval, when the modulation symbol is a normal point, Approaching 1, Approaching 0, Approaching 1; when the modulation symbol is a transient pulse interference point, Approaching 0, Approaching 1, Approaching 0; Punishment factor The strength of the penalty polarization is determined. This embodiment experimentally determines the optimal value of the penalty factor. In this experiment, the optimal value is determined in conjunction with the fuzzy weighting index. Under the same simulation conditions as the experiment, the fuzzy weighting index is fixed. Environmental sensitivity constant This experiment examines the penalty factors for contaminated and normal points. The mean and the number of iterations for iterative reweighted least squares convergence are shown in Table 3. Table 3
[0042] Analysis of Table 3 shows that when At that time, the pollution point The average is only 0.06, which is normal. The mean is as high as 0.99, showing a clear polarization between the two, and convergence is achieved in just 4 iterations. Insufficient penalty at certain times leads to excessive weakening of the weights of normal points, resulting in slow convergence; when At that time, due to the characteristics of the exponential function, the normal point It was quickly saturated to near 1.00, while the interference point Also because It increased and then rebounded to 0.16, which is too high. This causes the weights of all normal points to converge to 1, weakening the algorithm's ability to distinguish potential outliers in normal modulation symbols, thus affecting the weight matrix. The reduced dynamic range is equivalent to degenerating into an ordinary least squares problem, leading to a deterioration in the condition number of the normal equations and a surge in the number of iterations to convergence to 21. Therefore, this embodiment will... Curing setting is 5.0; The penalty factor is multiplied by the corresponding basic weight to obtain the final weight of each modulation symbol, and a diagonal weight matrix is constructed based on this to achieve algebraic isolation of interference points.
[0043] Construct a weight matrix, the weight matrix is as follows: diagonal matrix , It is a diagonal matrix, and the first element on its diagonal is... Each element is composed of and The product of these elements constitutes the weight matrix, and the mathematical expression for the diagonal elements is as follows: ; in, The weight matrix is the first One diagonal element, For the first The penalty factor for each modulation symbol, For the first The basic weights of each modulation symbol; A weighted least squares objective function is constructed based on a diagonal weight matrix, and a system of normal equations is constructed by combining it with a regularization term; Solve the system of equations to obtain the optimal tap coefficient vector.
[0044] The reconstructed weight matrix Substituting into the weighted least squares iterative objective equation, specifically, in the... In this iteration, a weighted least squares objective function is constructed. The mathematical expression of the weighted least squares objective function is: ; in, The weighted least squares objective function value, The total number of modulation symbols, For the first One reference modulation symbol, For the solution to be found dimensional tap coefficient vector It is the conjugate transpose of h. For the first 1 eigenvector; To reconstruct the diagonal matrix The One diagonal element, The preset regularization constant is taken in this embodiment. , Represents the vector of tap coefficients of Norm square regularization; Transforming the above weighted least squares objective function into matrix form, the mathematical expression in matrix form is: ; in, The weighted least squares objective function value, It is composed of the corresponding reference modulation symbol The complex conjugates constitute 3D column vector, It consists of all eigenvectors conjugate transpose Composition matrix, For the solution to be found dimensional tap coefficient vector This indicates the conjugate transpose. This is the weight matrix. This is a preset regularization constant; Convert the matrix form of the objective function to Taking the partial derivative of the conjugate vector and setting the derivative to zero, we obtain the normal equation, whose mathematical expression is: ; in, It consists of all eigenvectors conjugate transpose Composition Matrix, i.e., the first Behavior , This is the weight matrix. for The conjugate transpose of . The preset regularization constant is used. for identity matrix It is composed of the corresponding reference modulation symbol The complex conjugates constitute 3D column vector, For the solution to be found dimensional tap coefficient vector; Solving this system of linear equations yields the tap coefficient vector for the current iteration step; For the The tap coefficient vector is obtained by the next iteration. After that, regarding the first The next iteration, utilizing Re-evaluate the eigenvectors Perform filtering operations to obtain The mathematical expression for filtering is: ; in, for The corresponding estimated modulation symbol, Tap coefficient vector The conjugate transpose of . For feature vectors; For each element of the tracking region sequence Find the distance again in the ideal constellation coordinate set. The nearest point, take the coordinates of this point as The initial guess value is dynamically updated for each element of the tracking area sequence based on the ideal constellation map coordinate set; This embodiment employs a phased update strategy during the iteration process: When the number of iterations In this embodiment, the following is taken: The pilot phase is the stage where only the known preceding training sequence is used to update the tap coefficient vector. The weights of the tracking region sequence are forced to zero. In other words, only the training region modulation participates in the iterative update during the pilot phase, so that the tap coefficient vector can quickly converge from the initial state to a basically accurate solution. When the number of iterations At this point, the tracking phase begins, during which the tap coefficient vector has initially converged. The weights of all modulation symbols are restored, allowing the tracking region sequence to participate in subsequent iterations. Based on the updated reference modulation symbol sequence, calculate the residual, base weights, and penalty factor to construct the weight matrix, solve the objective function, and repeat the above iterative process until the tap coefficient vectors obtained in two consecutive iterations are obtained. and The Euclidean distance between them is less than 10 -4 The iteration stops when the maximum number of iterations reaches 50, and the vector signal analyzer uses the finally converged tap coefficient vector as the optimal tap coefficient vector. For complex baseband signals Filtering is performed to obtain the compensated complex baseband signal, and then a constellation diagram is drawn based on the compensated modulation symbol and the error vector amplitude is calculated.
[0045] This invention utilizes an exponential mapping function to convert the local density exponent into a polarized penalty factor, and multiplies it with the basic weights to construct a diagonal weight matrix. This directly forces the weights of high-energy pulse interference to near zero, blocking the erroneous pull of the huge residuals when solving the weighted least squares equations. Combined with a phased iterative update strategy in the guidance and tracking phases, the tap coefficient vector can ignore severe electromagnetic pulse interference and quickly approach the optimal solution with very few iterations, improving the robustness and convergence efficiency of the algorithm.
[0046] The output module is used to perform digital linear filtering on the complex baseband signal using the optimal tap coefficient vector to obtain the compensated complex baseband signal. Based on the distribution of each modulation symbol in the compensated complex baseband signal on the I / Q complex plane, draw a constellation diagram; Each compensated modulation symbol is compared with the nearest ideal constellation point in the ideal constellation diagram coordinate set, and the error vector amplitude is calculated. The error vector amplitude is used to measure the quality of the transmitted signal of the device under test.
[0047] This invention utilizes the solved optimal tap coefficient vector with strong anti-interference capability to perform digital linear filtering compensation on the original complex baseband signal. This process removes complex channel distortion and transient electromagnetic pulse pollution, and truly restores the original baseband signal appearance of the transmitter. This makes the final generated high-order constellation diagram clearer and more convergent. The calculated error vector amplitude can accurately and objectively measure the actual transmission quality of the device under test, and eliminates performance misjudgment caused by environmental interference.
[0048] This invention effectively overcomes the inherent multi-ring energy difference defect of high-order orthogonal amplitude modulation signals due to the continuous change of instantaneous amplitude by introducing a dimension reduction module to calculate the multi-ring error value. By mapping the distribution of complex baseband signals on the two-dimensional complex plane to one-dimensional multi-ring error values, the error value is significantly reduced to an order of magnitude comparable to the background noise power, regardless of whether the normal modulation symbol is originally located in the inner or outer ring. Meanwhile, abnormal interference points superimposed with transient pulses retain the abnormally deviated original energy level and thus increase significantly. This processing method can distinguish between normal high-energy signal points located in the outermost ring and abnormal interference points superimposed with transient pulses, significantly improving the accuracy of extracting abnormal features of the underlying baseband signal. Based on this, the present invention constructs a Gaussian kernel local density exponent and penalty factor based on background noise power through a local density calculation module and a nonlinear exponential mapping mechanism. This mechanism can quantify the sparsity and isolation of the distribution of modulation symbols in the multi-loop error value space, thereby further widening the weight difference between normal modulation symbols and interference modulation symbols. For transient pulse interference points with extremely high amplitude and energy, the present invention forces its penalty factor to be close to zero, thereby achieving algebraic isolation of interference points when constructing the diagonal weight matrix. This avoids the erroneous pull of huge residuals on the weighted least squares equations, ensuring the optimality of the tap coefficient vector solution direction and the stability of the numerical solution.
[0049] In the iterative weighted least squares solution process, this invention adaptively combines the Huber piecewise weighting function based on background noise power and dynamically sets the insensitivity threshold and weight function threshold. This not only effectively filters out the small residual fluctuations caused by thermal noise and ensures that the residuals caused by normal noise are not excessively penalized, but also performs reasonable gradient weighting control on the residuals in different intervals. This enables the baseband channel compensation equalizer to converge quickly with very few iterations even in harsh test scenarios with occasional and sparse transient electromagnetic pulse interference.
[0050] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media containing computer-usable program code. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory, electrically erasable programmable read-only memory, erasable programmable read-only memory, programmable read-only memory, read-only memory, magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions can also be stored in a computer-readable storage medium capable of directing a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means implemented in a process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0051] It should be noted that the above 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 preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the protection scope of the present invention.
Claims
1. An intelligent data processing system for a vector signal analyzer test data, characterized in that, include: The acquisition module is used to acquire the complex baseband signal of the device under test and the background noise power of the test environment; The dimension reduction module is used to calculate the multi-loop error value of each modulation symbol in the complex baseband signal based on the modulation order of the modulation signal; The local density calculation module is used to calculate the local density index of each modulation symbol based on the background noise power and the multi-loop error value. The solution module is used to solve for the optimal tap coefficient vector based on the complex baseband signal, the local density index, and the reference modulation symbol sequence by using an iterative weighted least squares method. The output module is used to perform digital linear filtering on the complex baseband signal using the optimal tap coefficient vector to obtain the compensated complex baseband signal and generate a constellation diagram and error vector amplitude.
2. The intelligent data processing system for a vector signal analyzer as described in claim 1, characterized in that, The process of acquiring the complex baseband signal in the acquisition module specifically includes: The modulated radio frequency signal emitted by the device under test is processed to obtain a complex baseband signal; The processing includes: downconversion, analog-to-digital conversion, digital downconversion, matched filtering, timing synchronization, decimation processing, and normalization processing.
3. The intelligent data processing system for a vector signal analyzer as described in claim 2, characterized in that, The process of acquiring the background noise power in the acquisition module specifically includes: Extract the background sampling sequence before the arrival of the modulated radio frequency signal; The average power of the background sampling sequence is calculated as the normalized background noise power.
4. The intelligent data processing system for a vector signal analyzer as described in claim 3, characterized in that, The process of obtaining the multi-loop error value in the dimensionality reduction module specifically includes: Based on the modulation order of the modulated radio frequency signal, generate an ideal constellation diagram coordinate set; Based on the ideal constellation coordinate set, calculate the normalized radius values of several reference amplitude rings; For each element in the complex baseband signal, calculate its measured complex amplitude; The geometric distance between the measured complex amplitude and the normalized radius of each reference amplitude loop is calculated as the radial deviation.
5. The intelligent data processing system for a vector signal analyzer as described in claim 4, characterized in that, The process of obtaining the multi-loop error value in the dimensionality reduction module further includes: Based on the radial deviation, the membership degree of each element to the reference ring of each reference amplitude ring is calculated using the fuzzy membership function; Using the membership degree of the reference ring as a weighting coefficient, the squared values of the radial deviations are weighted and accumulated to calculate the multi-ring error value.
6. The intelligent data processing system for a vector signal analyzer as described in claim 5, characterized in that, The process of obtaining the local density index in the local density calculation module specifically includes: For each modulation symbol, calculate the squared distance between its multi-loop error value and the multi-loop error values of all other modulation symbols in the sequence; Using the background noise power as the variance control parameter of the kernel function, a Gaussian kernel function transformation is performed on the squared distance to obtain a similarity measure between each modulation symbol; The local density index of the current modulation symbol is obtained by summing and normalizing the similarity measures of all other modulation symbols to the current modulation symbol.
7. The intelligent data processing system for a vector signal analyzer as described in claim 6, characterized in that, The process of solving the optimal tap coefficient vector in the solution module specifically includes: The complex baseband signals are traversed by a sliding window of a preset length, and the center-aligned data slices are extracted as feature vectors. The missing data outside the window boundary is filled with zero values. A reference modulation symbol sequence time-aligned with the complex baseband signal is constructed, the reference modulation symbol sequence being divided into a training region sequence and a tracking region sequence; The training region sequence is filled with a known preceding training sequence; The elements in the tracking region sequence are dynamically updated in each iteration.
8. The intelligent data processing system for a vector signal analyzer as described in claim 7, characterized in that, The process of solving the optimal tap coefficient vector in the solution module further includes: The feature vector is linearly filtered using the tap coefficient vector of the current iteration period to obtain the estimated modulation symbol, and the residual vector between the estimated modulation symbol and the corresponding elements in the reference modulation symbol sequence is calculated. Based on the magnitude of the residual vector, the basic weight of each modulation symbol is calculated using a piecewise weighting function. Modulation symbols with residuals less than a preset insensitive threshold are assigned zero weight, modulation symbols with residuals in the transition region are assigned gradual weight, and modulation symbols with large residuals are assigned amplitude limiting weight.
9. The intelligent data processing system for a vector signal analyzer as described in claim 8, characterized in that, The process of solving the optimal tap coefficient vector in the solution module further includes: The local density exponent is converted into a penalty factor using an exponential mapping function; The penalty factor is multiplied by the corresponding basic weight to obtain the final weight of each modulation symbol, and a diagonal weight matrix is constructed accordingly. A weighted least squares objective function is constructed based on the diagonal weight matrix, and a system of normal equations is constructed by combining it with regularization terms; Solve the system of normal equations to obtain the optimal tap coefficient vector.
10. The intelligent data processing system for a vector signal analyzer as described in claim 9, characterized in that, The output module is used to perform digital linear filtering on the complex baseband signal using the optimal tap coefficient vector to obtain the compensated complex baseband signal. Based on the distribution of each modulation symbol in the compensated complex baseband signal on the I / Q complex plane, draw a constellation diagram; The compensationd modulation symbols are compared with the nearest ideal constellation point in the ideal constellation diagram coordinate set, and the error vector amplitude is calculated.
Citation Information
Patent Citations
High-precision EMI diagnosis and pretest system based on vector signal analyzer
CN121703539A
